G protein-coupled receptor-like receptors and modulators thereof
Granted 14 Oct 2003 · 2 office actions
Current assignee: The Upjohn Company · originally Pfizer
Law firm: Law firm · Log in to unlock
Attorney: Attorney · Log in to unlock
Inventors: Martha J. Larsen, David E. Lowery, Teresa M. Kubiak, Timothy G. Geary · Examiner: Marjorie Moran · AU 1631 · TC 1600
Life of the application
9 dated eventsAbstract
The invention provides neuropeptide ligands, G protein-coupled receptors and methods of screening for modulators of receptor activity. Identified modulators, including neuropeptide ligand mimetics, are useful as biostatic and biocidal agents of varying scope, ranging from lethal activity restricted to particular invertebrate parasites to broad spectrum invertebrate parasiticides active on a wide range of invertebrates, including helminths and insects.
Description
133 parts›This application claims benefit of U.S. Provisional Application…
This application claims benefit of U.S. Provisional Application No. 60/167,523, filed Nov. 24, 1999.
›FIELD OF THE INVENTION
The present invention relates generally to G-protein, seven-transmembrane-receptor-like polynucleotides and polypeptides, as well as modulators of the interaction between the polypeptides and their ligand(s) that are useful for inhibiting the neurological activity of organisms, e.g., invertebrates.
›BACKGROUND
It has been estimated that as many as one in three humans is infected by one or more species of parasitic helminth (flatworms and roundworms). Helminths also represent a chronic and continuing threat to the health of livestock and companion animals; their abundance and resistance to anthelmintic drugs effectively prevents animal agriculture in certain environments. Despite the prevalence of helminth-caused disease, little is known of helminth physiology. This ignorance, in turn, hampers efforts to identify agents capable of controlling these pathogens.
Flatworms (platyhelminths) are evolutionarily quite distinct from the roundworms (nematodes) and differ markedly in neuromuscular anatomy and physiology. Only a subset of one class of drugs (some of the benzimidazoles) shows activity against both flatworms and roundworms. The diversity among the animal species that threaten the health of man, livestock, crops, and sensitive environmental niches, presents a challenge to efforts to mount broad-based attacks on the pest organisms.
G protein-coupled receptors (i.e., GPCRs) form a vast superfamily of cell surface receptors which are present in virtually all animal cells and are characterized by an amino-terminal extracellular domain, a carboxy-terminal intracellular domain, and a serpentine structure that passes through the cell membrane seven times. Hence, such receptors are sometimes also referred to as seven transmembrane (7TM) receptors. These seven transmembrane domains define three extracellular loops and three intracellular loops, in addition to the amino- and carboxy-terminal domains. The extracellular portions of the receptor have a role in recognizing and binding one or more extracellular binding partners (e.g., ligands), whereas the intracellular portions have a role in recognizing and communicating with downstream effector molecules.
The G protein-coupled receptors bind a variety of ligands including calcium ions, hormones, chemokines, neuropeptides, neurotransmitters, nucleotides, lipids, odorants, and even photons. Not surprisingly, the GPCRs are important in the normal (and sometimes the aberrant) function of many cell types. [See generally Strosberg, Eur. J. Biochem., 196: 1-10 (1991) and Bohm et al., Biochem J., 322: 1-18 (1997).] When a specific ligand binds to its corresponding receptor, the ligand typically stimulates the receptor to activate a specific heterotrimeric guanine nucleotide-binding regulatory protein (G protein) that is coupled to the intracellular portion or region of the receptor. The G protein, in turn, transmits a signal to an effector molecule within the cell by either stimulating or inhibiting the activity of that effector molecule. These effector molecules include adenylate cyclase, phospholipases and ion channels. Adenylate cyclase and phospholipases are enzymes that are involved in the production of the second messenger molecules cAMP, inositol triphosphate and diacyglycerol. It is through this sequence of events that an extracellular ligand stimulus exerts intracellular changes through a G protein-coupled receptor. Each such receptor has its own characteristic primary structure, expression pattern, ligand binding profile, and intracellular effector system.
Because of the vital role of G protein-coupled receptors in the communication between cells and their environment, such receptors are attractive targets for therapeutic intervention, and drugs that activate or antagonize the activation of such receptors are known. For receptors having a known ligand, the identification of agonists or antagonists may be sought specifically for mimicking, enhancing or inhibiting the action of the ligand. Thus, GPCRs show promise as potential targets of methods for treating infestations and/or infections caused by a variety of invertebrate pests, including both ecto- and endo-parasites. However, such methods must be able to discriminate the GPCRs of the invertebrate pest organisms from the GPCRs found in those species of plants and vertebrate animals upon whom the pests prey.
A large family of peptides (typically 4-15 amino acids in length) that is largely, if not exclusively, found in invertebrate animals such as helminths is a class of neuropeptides known as FMRFamide related peptides (i.e., FaRPs). The prototypical FMRFamide peptides are so named because of the “FMRF” amino acid sequence, including the consensus “RF” sequence, at their C-termini. As neuropeptides, these molecules are involved in vital biological processes requiring controlled neuromuscular activity. Although some neurotransmitters and neuromodulators (including neuropeptides) have been shown to function as ligands for receptors, to date there has been no identification of a FaRP neuropeptide as a ligand of a GPCR.
Because of the toxic potential of broad-spectrum chemical parasiticides, there exists a need in the art for targeted biologicals capable of selectively interfering with the life cycle of harmful invertebrates such as helminths and insects without harming host plant and animal species, as well as the environment.
›SUMMARY OF THE INVENTION · 1 of 4
The present invention generally relates to materials and methods for the targeted interference with vital biological processes of pest invertebrates. By providing materials and methods for modulating the activity of invertebrate GPCRs involved in neuromuscular activity, the invention provides a biological approach to invertebrate pest control that can minimize the deleterious consequences to non-pest species of animals, including man, as well as plants and the environment in general.
One aspect of the invention is a screening method for identifying candidate anti-invertebrate modulators that affect one or more activities of an invertebrate GPCR-like receptor involved in neuromuscular functioning including, e.g., binding of a GPCR-like receptor to a ligand, typically a peptide ligand, and signal transduction. The method comprises the steps of: (a) contacting a test compound with a composition, wherein the composition contains a GPCR-like receptor encoded by a polynucleotide having a sequence selected from the group consisting of SEQ ID NOS: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178, or a polynucleotide hybridizing to the GPCR-like receptor under stringent conditions of hybridizing at 42° C. in a solution comprising 50% formamide, 1% SDS, 1 M NaCl, 10% dextran sulfate, and washing twice for 30 minutes at 60° C. in a wash solution comprising 0.1×SSC and 1% SDS; and (b) measuring the activity of the GPCR-like receptor in the presence and absence of the test compound. In preferred embodiments, the screening method includes a receptor, or fragment, variant or derivative thereof, having a sequence set forth in SEQ ID NOS: 7, 21, 25, 35, 43, 45, 105, 107, 109, 111, 113, 115, 117, 177 and 179.
As one of ordinary skill in the art would recognize, the above-described method can be practiced with a variety of GPCR-like receptors. For example, the GPCR-like receptor used in the screening method may be encoded by a polynucleotide having a sequence set forth in any one of SEQ ID NOS:43, 21, 45, 35, 7, 106, and 104. As noted above, such GPCR-like receptors may be used in screening assays designed to measure a GPCR-like receptor activity, including binding activity. Expressly contemplated are embodiments of the screening method comprising a GPCR-like receptor encoded by a polynucleotide comprising a sequence set forth in SEQ ID NO:43 and a peptide comprising a sequence selected from the group consisting of SEQ ID NOS:85, 86, 88, 89, and 118, wherein the peptide binds to the GPCR-like receptor. encoded by a polynucleotide having the sequence set forth in SEQ ID NO:43 is provided. In an alternative embodiment, the GPCR-like receptor comprises a sequence set forth in SEQ ID NO:21 and the peptide comprises a sequence selected from the group consisting of SEQ ID NOS:78, 79, 80, 84, 87, 92, 98, 100, 120, 171, 143, 122, 123, 97, 85, 83, 101, 102, 93, 88, 91, 94, 93, 90, 152, 153, 154, 155, 156, 157, 80, 158, 119, 159, 160, 161, 162, 163 and 164. Another embodiment involves a GPCR-like receptor comprising a sequence set forth in SEQ ID NO:45 and a peptide comprising a sequence selected from the group consisting of SEQ ID NOS:86, 118, 125, 88, 126, 127, 128, 129, 102, 131, 100, 133, 92, 135, 136, 137, 87, 139, 91, 141 and 83. In yet another embodiment, the GPCR-like receptor comprises a sequence set forth in SEQ ID NO:35 and a peptide comprising a sequence selected from the group consisting of SEQ ID NOS:99, 97, 96, 77, 82, 81, 87, 100, 92, 80, 98, 120, 121, 79 and 84. In still another embodiment, the GPCR-like receptor comprises a sequence set forth in SEQ ID NO:7 and a peptide comprising a sequence selected from the group consisting of SEQ ID NOS:94, 103, 95, 101, 85, 79, 84, 87, 86, 80, 92, 100, and 180. Yet another embodiment involves a GPCR-like receptor comprising a sequence selected from the group consisting of SEQ ID NO:106 and a peptide comprising a sequence selected from the group consisting of SEQ ID NOS:80, 92, 98, 100, 120, 121, 79, 84, 136, 87 and 86. Still another one of the many embodiments of this aspect of the invention involves a GPCR-like receptor comprising a sequence set forth in SEQ ID NO:104 and a peptide comprising a sequence selected from the group consisting of SEQ ID NOS:80, 92, 98, 100, 120, 121, 79, 84, 136, 87, 86, 150, 151, 133, 165, 91, 166, 131 and 167.
Another aspect of the invention is a method of identifying an anti-invertebrate modulator of an activity of an invertebrate GPCR-like receptor comprising the following steps: (a) contacting a test compound and a composition, wherein the composition contains a GPCR-like receptor selected from the group consisting of polypeptides encoded by a polynucleotide having a sequence selected from the group consisting of SEQ ID NOS: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178, or a polynucleotide hybridizing to the GPCR-like receptor under stringent conditions of hybridizing at 42° C. in a solution comprising 50% formamide, 1% SDS, 1 M NaCl, 10% Dextran sulfate, and washing twice for 30 minutes at 60° C. in a wash solution comprising 0.1×SSC and 1% SDS, optionally in the presence of a peptide or other ligand of the receptor; and (b) measuring the activity of the GPCR-like receptor in the presence and absence of the test compound. Modulators are identified as test compounds that alter (i.e., increase or decrease) a GPCR-like receptor function, such as a binding property of a receptor or an activity such as G protein-mediated signal transduction or membrane localization. The composition may contain an isolated GPCR-like receptor; alternatively, the composition may contain a GPCR-like receptor in association with, e.g., an intact cell or cell portion, such as a membrane. Presently preferred embodiments of the method use a GPCR-like receptor having an amino acid sequence set forth in SEQ ID NOS: 8, 22, 26, 36, 44, 105, 107, 109, 111, 113, 115, 117, 177 or 179. Preferred peptides are neuropeptides derived from invertebrates and include the FaRP family of neuropeptides. Particularly preferred are invertebrate neuropeptides having an amino acid sequence selected from the group consisting of SEQ ID NOS: 77-103 and 118-151. The methods of the invention embrace neuropeptides that are attached to a label, such as a radiolabel (e.g., 125 I, 35 S, 32 P, 33 P, 3 H), a fluorescence label, a chemiluminescence label, an enzymic label and an immunogenic label.
›SUMMARY OF THE INVENTION · 2 of 4
In various embodiments of the method, the assay may take the form of an ion flux assay, a yeast growth assay, a non-hydrolyzable GTP assay such as a [ 35 S]-GTPγS assay, a cAMP assay, an inositol triphosphate assay, a diacylglycerol assay, an Aequorin assay, a Luciferase assay, a FLIPR assay for intracellular Ca 2+ concentration, a mitogenesis assay, a MAP Kinase activity assay, an arachidonic acid release assay (e.g., using [ 3 H]-arachidonic acid), and an assay for extracellular acidification rates, as well as other binding or function-based assays of GPCR activity that are generally known in the art. In several of these embodiments, the invention comprehends the inclusion of any of the G proteins known in the art, such as G α16 , G α15 , or chimeric G qi5 , G qs5 , G qo5 , or G qz5 .
In another aspect of the invention, a method of identifying a candidate anti-invertebrate modulator is provided. The method comprises the steps of: (a) contacting a test compound and a composition, wherein the composition contains a GPCR-like receptor encoded by a polynucleotide selected from the group consisting of receptor polyucleotides having a sequence set forth in SEQ ID NOS: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178, and polynucleotides hybridizing to the receptor polynucleotides under stringent conditions of hybridizing at 42° C. in a solution comprising 50% formamide, 1% SDS, 1 M NaCl, 10% dextran sulfate, and washing twice for 30 minutes at 60° C. in a wash solution comprising 0.1×SSC and 1% SDS; and (b) identifying a test compound that binds to or interacts with the composition as a candidate anti-invertebrate modulator. In a preferred embodiment, the polynucleotide encoding the GPCR-like receptor comprises a sequence selected from the group consisting of SEQ ID NOS: 7, 21, 25, 35, 43, 45, 105, 107, 109, 111, 113, 115, 117, 177 and 179.
Another aspect of the invention is a method of identifying an anti-invertebrate agent comprising the following steps: (a) identifying a modulator using the method described above; (b) contacting the modulator and an invertebrate tissue; and (c) measuring the response of the invertebrate tissue, e.g. neural signaling or neuromuscular activity, thereby identifying the modulator as an anti-invertebrate agent. Although any invertebrate tissue may be used in the method, presently preferred tissue sources are helminths and insects. An exemplary type of suitable invertebrate tissue is neuromuscular tissue, e.g. in isolated form or remaining in association with part, or all, of an invertebrate organism. The anti-invertebrate agents, and compositions comprising one of those agents, identified by this method constitute yet another aspect of the invention.
In a related aspect, the invention provides a method for treating an invertebrate comprising the step of contacting the invertebrate with a biologically effective amount of a modulator identified by the methods described herein. A biologically effective amount of a modulator is an amount that is sufficient to induce a desired response in the treated invertebrate. Thus, a biologically effective amount of a modulator may be the amount that interferes with physiological activity of the treated invertebrate in a non-lethal manner (i.e., a biostatic effect) or in a lethal manner (i.e., a biocidal effect). A preferred modulator for use in the treatment methods is an inhibitor of GPCR-like receptor activity. The invention is not limited to particular means for delivering the modulator to an invertebrate, nor is the invention limited as to the compositions comprising the modulator which may be delivered.
Another aspect of the invention is drawn to methods of producing an invertebrate GPCR-like receptor comprising the following steps:(a) incubating a source cell at a temperature of at least about 35° C.; (b) lowering the temperature to at most about 26° C.; and (c) detecting the GPCR-like receptor. In various embodiments, the temperatures may be varied to optimize production using no more than routine experimentation. For example, the cells may be incubated at temperatures higher than about 35° C., e.g., a temperature of at least about 37° C.; the temperature also may be lowered beyond 30° C., for example to at most 29° C., at most 28° C., at most 27° C., at most 26° C., at most 25° C. or at most 24° C. In some embodiments, both the incubation temperature may be raised above 35° C. and the temperature lowering may extend beyond 30° C. The method of producing an invertebrate GPCR-like receptor may further comprise recovering the GPCR-like receptor, which may be native or recombinant in origin. The receptor may be recovered in intact cells, cell portions (e.g., membranes) obtained as a result of cell lysis, or in isolated form. Any of a wide variety of cells may be used as source cells, such as cells derived from mammals, amphibians, arthropods (e.g., insects), mollusks, helminths, and others. It is anticipated that this method of producing a GPCR-like receptor is particularly suited to the recombinant production of GPCR-like receptors using non-invertebrate cells, such as mammalian cells.
The invention also comprehends compositions of matter, such as a modulator of an activity of the GPCR-like receptors identified by the methods described herein. The modulators of the invention exhibit a variety of chemical structures, which can be generally grouped into non-peptide mimetics of natural GPCR-like receptor ligands, peptide and non-peptide allosteric effectors of GPCR-like receptors, and peptides and non-peptide compounds that function as activators or inhibitors (competitive, uncompetitive and non-competitive) (e.g., antibody products) of an ultimate GPCR-like receptor activity. For example, such modulators may be compounds or compositions that are agonists or antagonists of ligand binding to a GPCR-like receptor, allosteric effectors thereof, or compounds (or compositions) that affect the ultimate activity of GPCR-like receptors through direct action on the receptor or an effect introduced downstream of the receptor in, e.g., a signal cascade. In addition, modulators according to the invention may be compounds or compositions that interfere with the expression of a GPCR-like receptor, either through inhibiting transcription of the DNA or translation of the corresponding mRNA. Expression of the GPCR-like receptor can be monitored by any methods known in the art, including Western blot analysis using polyclonal or monoclonal antibodies to the GPCR-like receptor or Northern blot analysis or quantitative polymerase chain reaction (PCR) using suitable probes or primers based on the sequence of the GPCR-like receptor gene. In particular, modulators that interfere with the expression of gene products include anti-sense polynucleotides and ribozymes that are complementary to the gene sequences. The invention further embraces modulators that affect the transcription of gene products of the invention through the formation of oligonucleotide-directed triplet helix formation. The invention does not restrict the sources for suitable modulators, which may be obtained from natural sources such as plant, animal or mineral extracts, or non-natural sources such as small molecule libraries, including the products of combinatorial chemical approaches to library construction, and peptide libraries. Preferred peptide receptors have an amino acid sequence selected from the group consisting of SEQ ID NOS: 8, 22, 26, 36, 44, 105, 107, 109, 111, 113, 115, 117, 177 or 179.
›SUMMARY OF THE INVENTION · 3 of 4
Another aspect of the invention is drawn to an isolated GPCR-like receptor comprising a sequence selected from the group consisting of SEQ ID NOS: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 105, 107, 109, 111, 113, 115, 117, 177, and 179, and variants and fragments thereof. Preferred receptors have sequences set forth in SEQ ID NOS: 8, 22, 26, 36, 44, 105, 107, 109, 111, 113, 115, 117, 177 and 179. Variants and fragments of GPCR-like receptors retain at least one biological or immunological property of the cognate GPCR-like receptor, and are at least 10, at least 15, at least 20, at least 25, at least 30, at least 40, at least 50, at least 60, at least 70, at least 225, 80, at least 90, at least 100, at least 125, at least 150, at least 175, at least 200, at least 225, at least 250, at least 275, at least 300, at least 325, at least 350, at least 375, at least 400, at least 425, or at least 450 amino acids in length. Fragments specifically include GPCR domains of the receptors. Biological activities of GPCR-like receptors according to the invention include, but are not limited to, the binding of a natural or an unnatural ligand, as well as any one of the functional activities of GPCRs known in the art. Other non-limiting examples of GPCR activities include transmembrane signaling of various forms, which may involve G protein association and/or the exertion of an influence over G protein binding of various guanidylate nucleotides; transmembrane localization; or binding of accessory proteins or polypeptides unrelated to known G proteins.
The invention also provides an isolated polynucleotide encoding a GPCR-like receptor. Such polynucleotides may be selected from the group consisting of: (a) a polynucleotide comprising a nucleotide sequence encoding any one of the amino acid sequences set forth in SEQ ID NOS: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 105, 107, 109, 111, 113, 115 117, 177, and 179 (including the nucleotide sequences set forth in SEQ ID NOS: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176, and 178); and (b) a polynucleotide which hybridizes under conditions of high stringency to the complement of the polynucleotide of (a). Exemplary conditions of high stringency are provided below. Such polynucleotides also include polynucleotides that exhibit at least 90%, at least 95%, at least 98%, at least 99% or at least 99.9% sequence identity to either a polynucleotide sequence disclosed in the sequence listing (i.e., SEQ ID NOS: 1, 3, 5, 7, 9, 11, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178) or to a polynucleotide encoding a GPCR-like receptor comprising one of the sequences disclosed in the sequence listing (i.e., SEQ ID NOS: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 105, 107, 109, 111, 113, 115, 117, 177, and 179). Any one of the publicly available algorithms (e.g., the BLAST1 program of GCG) for comparing sequences may be used in determining the degree of sequence similarity (including appropriate penalties for gap introductions). A preferred algorithm is the BLAST algorithm implemented at the GenBank website under the auspices of the National Center for Biotechnology Information A on the World Wide Web at ncbi.nlm.nih.gov/BLAST/ using default parameters. A polynucleotide of the invention may be partially or wholly chemically synthesized and embraces an anti-sense polynucleotide which specifically hybridizes to the complement of one or more of the above-described polynucleotides. In related aspects, the invention comprehends vectors comprising these polynucleotides preferably operably linked to expression control sequences, including expression vectors, as well as non-native host cells transformed or transfected with a polynucleotide in accordance with the invention or a host cell transformed or transfected with a vector of the invention. All suitable native and non-native host cells are embraced by the invention, including mammalian cells (e.g., COS cells, CHO cells, HEK293 cells), insect cells (e.g., Drosophila melanogaster S2 cells, Spodoptera frugiperda Sf9 cells, High-5 cells), yeast cells, bacterial cells (e.g., E. coli ) and helminthic cells. The suitability of a particular cell for use as a host cell in accordance with the invention will depend on the ability to introduce a polynucleotide of the invention into the cell by any known means of transformation or transfection. Preferred host cells will also be capable of stably maintaining the introduced polynucleotide and will present a minimum of obstacles to propagation.
Another aspect of the invention is directed to a genetically modified invertebrate comprising a polynucleotide encoding a heterologous GPCR-like receptor (e.g., transgene) as described above or comprising a modification in a native gene encoding a GPCR-like receptor (e.g., an insertional disruption or deletion of the gene). The GPCR-like receptor encoding gene may be expressed at normal levels for that gene or may be overexpressed or underexpressed. A preferred invertebrate for generation of genetically modified organisms is a member of the helminths.
Yet another aspect of the invention is drawn to diagnostic methods for determining neurological abnormalities associated with aberrant GPCR-like receptor activity. Such methods specifically measure the presence, and optionally quantity, of a GPCR-like receptor polynucleotide according to the invention or the presence, and optionally quantity or activity, of a GPCR-like receptor. Any method known in the art may be used to measure the specific polynucleotides or polypeptides of the invention, and measurements may be performed on intact organisms, isolated tissues, or cell cultures. In a related aspect, the invention contemplates methods for diagnosing invertebrate infestation of an organism (e.g., mammals such as humans, mammalian livestock, other vertebrates such as fish, and non-pest invertebrates such as molluscs) or an environment using a specific measurement of a polynucleotide or polypeptide according to the invention.
›SUMMARY OF THE INVENTION · 4 of 4
Another aspect of the invention is directed to treatment methods. The invention contemplates methods of killing or inhibiting the viability of invertebrates using the modulators described above or identified as described above, including antibodies and antisense polynucleotides. Such methods include methods for treating infestations and/or infections caused by a variety of invertebrate pests, including both ecto- and endo-parasites. A variety of human and other animal ailments, particularly those ailments relating to aberrant neurological functioning, are treated by administering a biologically effective amount of a modulator, GPCR-like receptor polynucleotide or GPCR-like receptor to a cell, tissue, organism, or environment using techniques known in the art.
Use of such modulators in the preparation of a medicament for treating parasitic infection is also contemplated. A variety of administration regimens known in the art are available to deliver a therapeutically effective (i.e., a level of activity capable of deleteriously affecting at least one biological process of a parasite, or more generally, of a pest organism) activity level of a modulator, directly or indirectly, to a parasite or pest, or to an environment associated with the parasite or pest.
Numerous additional aspects and advantages of the invention will become apparent to those skilled in the art upon consideration of the following detailed description of the preferred embodiments of the invention.
›DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS · 1 of 5
The present invention has several aspects, one of which is modulating the activity of invertebrate G protein-coupled receptors, which is useful in controlling a wide variety of pest organisms that affect the health of humans and other animals such as domestic pets and livestock, as well as plants. The G protein-coupled receptors typically transduce signals involved in the neurophysiological functioning of the target invertebrates and interference with that functioning frequently proves fatal to the target organism. Moreover, these receptors are found only in invertebrates, so that specific modulators are safe for use around humans, as well as their pets, livestock, and crops.
A “GPCR-like receptor” is a polypeptide receptor exhibiting the structural characteristics of G protein-coupled receptors (GPCRs), i.e., an N-terminal extracellular region, comprising several loop-like domains (typically three), a transmembrane region comprising seven transmembrane domains arranged in a typical serpentine disposition, and an intracellular region comprising several loop-like domains (typically three).
Invertebrate neuropeptides show profound neuromuscular effects and FaRPs represent the largest family of such neuropeptides known to date. A “FaRP” is a “FMRFamide-related peptide.” A “FMRFamide,” in turn, is a relatively small peptide, typically having 4-15 amino acids, that matches at least three of the four listed amino acids (FMRF) at its C-terminus. Generally, FaRPs exhibit neurophysiological effects and are therefore properly grouped in a class of neuropeptides that frequently function as ligands of GPCR-like receptors. A “modulator” is an effector of a GPCR-like receptor function, which include the binding of one or more ligands, localization to a membrane, and signal transduction. Signal transduction may involve a change in the relative affinities of a GPCR-associated G protein for various guanidylate nucleotides. With respect to polynucleotides and polypeptides of the invention, “synthesized,” as used herein and understood in the art, refers to polynucleotides or polypeptides produced by purely chemical, as opposed to enzymatic, methods. “Wholly” synthesized sequences are therefore produced entirely by chemical means, and “partially” synthesized sequences embrace those wherein only portions of the resulting polynucleotide or polypeptide were produced by chemical means.
The present invention provides purified and isolated polynucleotides (e.g., DNA sequences and RNA transcripts, both sense and complementary antisense strands, including splice variants thereof) encoding invertebrate G protein-coupled receptors. DNA polynucleotides of the invention include genomic DNA, cDNA, and DNA that has been chemically synthesized in whole or in part.
Genomic DNA of the invention comprises the protein coding region for a polypeptide of the invention and includes allelic variants thereof. It is widely understood that, for many genes, genomic DNA is transcribed into RNA transcripts that undergo one or more splicing events wherein intron (i.e., non-coding regions) of the transcripts are removed, or “spliced out.” RNA transcripts that can be spliced by alternative mechanisms, and therefore be subject to removal of different RNA sequences but still encode the same polypeptide, are referred to in the art as splice variants. Splice variants therefore are encoded by the same original genomic DNA sequences but arise from distinct mRNA transcripts found in at least one cell. By way of non-limiting example, several embodiments of the invention are characterized by one of the particular splice variants identified in Table 1, below.
Allelic variants are modified forms of a wild-type gene sequence, the modification resulting from recombination during chromosomal segregation or exposure to conditions which give rise to genetic mutation. Allelic variants, like wild type genes, are naturally occurring sequences (as opposed to non-naturally occurring variants which arise from in vitro manipulation), and are also comprehended by the invention.
The invention also comprehends cDNA that is obtained through reverse transcription of an RNA polynucleotide encoding invertebrate GPCR-like receptors (conventionally followed by second strand synthesis of a complementary strand to provide a double-stranded DNA). In addition to cDNA forms of the polynucleotides identified in Table 1 as splice variants, preferred cDNAs according to the invention include those polynucleotides having sequences selected from the group consisting of SEQ ID NOS: 13, 15 and 21. These sequences differ from sequences in the Wormpep database that exhibit some similarity thereto, as shown in Table 2.
Preferred DNA sequences encoding invertebrate GPCR-like receptor polypeptides are set out in SEQ ID NOs: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178. A preferred DNA of the invention comprises a double stranded molecule (for example, the molecule having one of the sequences set forth in the above-referenced SEQ ID NOs, along with the complementary molecule (the “non-coding strand” or “complement”) having a sequence unambiguously deducible from one of those sequences, according to Watson-Crick base-pairing rules for DNA). Also preferred are other polynucleotides encoding one of the invertebrate GPCR-like receptor polypeptides of SEQ ID NOs: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 105, 107, 109, 111, 113, 115, 117, 177, and 179 which may differ in sequence from the corresponding polynucleotides of SEQ ID NOs: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178 by virtue of the well-known degeneracy of the universal genetic code.
The invention further embraces invertebrate species (preferably helminth and insect) homologs of the disclosed GPCR-like DNAs. Species homologs, sometimes referred to as “orthologs,” in general share at least 35%, at least 40%, at least 45%, at least 50%, at least 60%, at least 65%, at least 70%, at least 75%, at least 80%, at least 85%, at least 90%, at least 95%, at least 98%, or at least 99% similarity with the DNA sequences disclosed herein. Percent sequence “similarity” with respect to polynucleotides of the invention is defined herein as the percentage of nucleotide bases in the candidate sequence that are identical to nucleotides in the relevant sequences set forth in the sequence listing below, after aligning the sequences and introducing gaps, if necessary, to achieve the maximum percent sequence identity. Preferred for comparative sequence analyses is the BLAST program available from GCG, implemented with default parameters. Also preferred is the Blastall program available from NCBI.
›DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS · 2 of 5
The polynucleotide sequence information provided by the invention makes possible large scale expression of an encoded polypeptide by techniques well known and routinely practiced in the art. Polynucleotides of the invention also permit identification and isolation of polynucleotides encoding related GPCR-like polypeptides, such as the aforementioned allelic variants and species homologs, by well known techniques including Southern and/or Northern hybridization, and by polymerase chain reaction (PCR). Examples of related polynucleotides include polynucleotides encoding polypeptides homologous to the invertebrate GPCRs and structurally related polypeptides sharing one or more biological, immunological, and/or physical properties of those GPCRs. Invertebrate genes encoding proteins homologous to the disclosed GPCRs can also be identified by Southern and/or PCR analysis and are useful in methods of the invention described below. Knowledge of the sequence of the disclosed GPCR-encoding DNAs also makes possible, through use of Southern hybridization, polymerase chain reaction (PCR), and other known techniques, the identification of genomic DNA sequences encoding GPCR expression control regulatory sequences such as promoters, operators, enhancers, repressors, and other regulatory sequences known in the art. Polynucleotides of the invention are also useful in hybridization assays to detect the capacity of cells to express an invertebrate GPCR.
The disclosure herein of a full-length polynucleotide encoding invertebrate GPCR-like polypeptides makes readily available to the worker of ordinary skill in the art every possible fragment of the full-length polynucleotides. The invention therefore provides fragments of invertebrate GPCR-encoding polynucleotides comprising at least 14-15, and preferably at least 18, 20, 25, 50, or 75 consecutive nucleotides of a polynucleotide encoding an invertebrate GPCR.
Fragment polynucleotides contemplated by the invention encode immunologically active peptides capable of interacting with specific anti-GPCR antibodies, as well as domains of GPCRs. The GPCR domains (i.e., the N-terminal extracellular domain, one or more of the three extracellular loop domains, one or more of the seven transmembrane domains, one or more of the three intracellular loop domains, or the C-terminal intracellular domain) of invertebrate GPCRs disclosed herein are characterized in Example 2 and Table 6, below. Full length or fragment polynulcleotides can be linked to polynucleotides encoding heterologous polypeptides, e.g., for producing variants that are fusion proteins.
Preferably, fragment polynucleotides of the invention comprise sequences unique to one of the disclosed GPCR-encoding sequences, and therefore hybridize under highly stringent or moderately stringent conditions only (i.e., “specifically”) to polynucleotides encoding a disclosed invertebrate GPCR (or fragments thereof) and not to polynucleotides encoding other GPCRs. Polynucleotide fragments of genomic sequences of the invention comprise not only sequences unique to the coding region, but also fragments of the full-length sequence derived from introns, regulatory regions, and/or other non-translated sequences. Sequences unique to polynucleotides of the invention are recognizable through sequence comparison to other known polynucleotides, and can be identified through use of alignment programs routinely utilized in the art, e.g., those made available in public sequence databases. Such sequences also are recognizable from Southern hybridization analyses to determine the number of fragments of genomic DNA to which a polynucleotide will hybridize. Polynucleotides of the invention can be labeled in a manner that permits their detection, including radioactive, fluorescent, and enzymatic labeling.
Fragment polynucleotides are particularly useful as probes for detection of full-length invertebrate GPCR-like polynucleotides, or other polynucleotide fragments thereof. One or more fragment polynucleotides can be included in kits that are used to detect the presence of a polynucleotide encoding an invertebrate GPCR-like receptor polypeptide, or used to detect variations in a polynucleotide sequence encoding such a polypeptide.
The invention also embraces DNAs encoding invertebrate GPCR-like polypeptides that hybridize under conditions of moderate or high stringency to the non-coding strand, or complement, of a polynucleotide comprising one of the sequences set forth in SEQ ID NOs: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178.
Exemplary conditions of high stringency are as follows: hybridization at 42° C. in a solution (i.e., a hybridization solution) comprising 50% formamide, 1% SDS, 1 M NaCl, 10% Dextran sulfate, and washing twice for 30 minutes at 60° C. in a wash solution comprising 0.1×SSC and 1% SDS. It is understood in the art that conditions of equivalent stringency can be achieved through variation of temperature and buffer, or salt concentration, as described in Ausubel, et al. (Eds.), Protocols in Molecular Biology , John Wiley & Sons (1994), pp.6.0.3 to 6.4.10. Modifications in hybridization conditions can be empirically determined or precisely calculated based on the length and the percentage of guanosine/cytosine (GC) base pairing of the probe. The hybridization conditions can be calculated as described in Sambrook, et al., (Eds.), Molecular Cloning: A Laboratory Manual , Cold Spring Harbor Laboratory Press: Cold Spring Harbor, N.Y. (1989), pp. 9.47 to 9.51.
The invention also provides a purified and isolated invertebrate GPCR-like polypeptide encoded by a polynucleotide of the invention. Presently preferred is a polypeptide comprising one of the amino acid sequences set out in SEQ ID NOs: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 105, 107, 109, 111, 113, 115, 117, 177, and 179, and species homologs thereof. The invention also embraces a GPCR-like polypeptide encoded by a DNA selected from the group consisting of: (a) the DNA sequence set out in any one of SEQ ID NOs: 1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31, 33, 35, 37, 39, 41, 43, 45, 104, 106, 108, 110, 112, 114, 116, 176 and 178 and species homologs thereof, and (b) a DNA molecule encoding a GPCR-like gene product that hybridizes under conditions of moderate or high stringency to the DNA of (a). The invention further embraces polypeptides that have at least about 99%, at least about 95%, at least about 90%, at least about 85%, at least about 80%, at least about 75%, at least about 70%, at least about 65%, at least about 60%, at least about 55%, and at least about 50% identity and/or homology to the preferred polypeptides of the invention. Percent amino acid sequence “identity” with respect to the preferred polypeptides of the invention is defined herein as the percentage of amino acid residues in the candidate sequence that are identical with the residues in the GPCR-like gene product sequence after aligning both sequences and introducing gaps, if necessary, to achieve the maximum percent sequence identity, and not considering any conservative substitutions as part of the sequence identity. Percent sequence “homology” with respect to the preferred polypeptides of the invention is defined herein as the percentage of amino acid residues in the candidate sequence that are identical with the residues in one of the GPCR-like receptor polypeptide sequences after aligning the sequences and introducing gaps, if necessary, to achieve the maximum percent sequence identity, and also considering any conservative substitutions as part of the sequence identity. Conservative substitutions are defined as set out in Tables 3 and 4.
›DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS · 3 of 5
GPCR polypeptides of the invention may be isolated from natural invertebrate cell sources or may be chemically synthesized, but are preferably produced by recombinant procedures involving host cells of the invention. GPCR-like gene products of the invention may be full-length polypeptides, biologically active fragments, or variants thereof which retain specific biological or immunological activity. Variants may comprise GPCR-like polypeptide analogs wherein one or more of the specified (i.e., naturally encoded) amino acids is deleted or replaced or wherein one or more non-specified amino acids are added: (1) without loss of one or more of the biological activities or immunological characteristics specific for the GPCR-like receptor; or (2) with specific disablement of a particular biological activity of the GPCR-like polypeptide. Contemplated deletion variants also include fragments lacking portions of a GPCR-like polypeptide not essential for biological activity, and insertion variants include fusion polypeptides in which the wild-type GPCR-like polypeptide or fragment thereof has been fused to another polypeptide.
Variant GPCR-like polypeptides include those wherein conservative substitutions have been introduced by modification of polynucleotides encoding polypeptides of the invention. Conservative substitutions are recognized in the art to classify amino acids according to their related physical properties and can be defined as set out in Table 3. (See, WO 97/09433, page 10, published Mar. 13, 1997 (PCT/GB96/02197, filed Sep. 6, 1996).) Alternatively, conservative amino acids can be grouped as defined in Lehninger, [ Biochemistry , Second Edition; Worth Publishers, Inc. NY:NY (1975), pp.71-77] as set out in Table 4.
Variant GPCR-like polypeptides of the invention include mature GPCR-like gene products, i.e., wherein leader or signal sequences are removed, having additional amino terminal residues. GPCR-like gene products having an additional methionine residue at position-1 are contemplated, as are GPCR-like receptors having additional methionine and lysine residues at positions-2 and -1. Variants of these types are particularly useful for recombinant protein production in bacterial cell types. Variants of the invention also include gene products wherein amino terminal sequences derived from other proteins have been introduced, as well as variants comprising amino terminal sequences that are not found in naturally occurring proteins.
Another component useful in some methods for identifying modulators of GPCR-like receptor activity, described below, is a class of neuropeptides that includes FaRPs. The FaRPs comprise a large family of peptides that typically function as neurotransmitters in the invertebrate nervous system. The FaRPs that are associated with these neurophysiological functions generally are about four to about nine amino acid residues in length and include a C-terminal amino acid sequence motif of (aromatic)-X-R-Famide, where X is typically M, L, I or V. Day et al., Peptides 20:999-1019 (1999), which is incorporated by reference in its entirety. However, peptides that are exceptions to this rule are also contemplated by the invention, particularly those that are co-encoded on FaRP precursor genes or those that share the characteristics of conforming peptides. Also contemplated by the invention are neuropeptide variants, including FaRP variants, peptides exhibiting an RYamide motif, and other peptides such as retro-inverso neuropeptides (e.g., FaRPs comprising the D stereoisomers of amino acids in a sequence that is reversed from a reference FaRP). For example a retro-inverso variant of A L -F L -M L -R L -F L would be F D -R D -M D -F D -A D .
Neuropeptides such as FaRPs have been found in a wide variety of invertebrates, including arthropods such as insects (e.g., locusts and flies such as Drosophila) and lobsters, mollusks such as the snail Lymnaea stagnalis , and helminths such as nematodes (e.g., C. elegans, A. suum, Haemonchus contortus ), trematodes (e.g., Schistosoma mansoni ) and cestodes, among others, including Manduca. Information relating to the structure, function, and structure-function relationships ofthese neuropeptides is known in the art [Day et al. (1999)]. Table 5 below provides a classification of several known C. elegans FaRPs, which are examples of the class of neuropeptides useful in practicing the invention.
Further characterization of the structure-function relationships of neuropeptides, such as FaRPs, that are embraced by the invention may be readily accomplished by one of ordinary skill in the art. For example, amino acid-scan modifications (e.g., alanine-scan), in which each residue is sequentially replaced with another amino acid such as alanine are available. Additionally, given the knowledge in the art (see generally, Geary et al., 1999), such modifications to known neuropeptides (e.g., FaRPs) as the internal or terminal deletion of one or more amino acids, as well as the internal or terminal addition of residues, involves no more than routine experimentation. Such “variants” of FaRPs and related neuropeptides are among the neuropeptides contemplated for use in the methods of the invention.
The GPCR-like receptor polypeptides and neuropeptides produced by the methods described above are useful in assays for modulators of GPCR-like receptor activities, including binding partner (e.g., ligand) binding. Assays contemplated by the invention include both binding assays and activity assays; these assays may be performed in conventional or high throughput formats. GPCR-like receptor activity is defined as including the binding of any binding partner, such as a ligand, as well as the propagation of any transmembrane signal (e.g., stimulation of a G protein or influence on the flux of an ion across a membrane). Modulator screens are designed to identify stimulatory and/or inhibitory agents. The sources for potential agents to be screened include natural sources, such as a cell extract (e.g., invertebrate cells including, but not limited to, bacterial, fungal, algal, and plant cells) and synthetic sources, such as chemical compound libraries. For proteins with known activity, function assays are established based on the activity, and a large number of potential agents are screened for the ability to either stimulate or inhibit the activity. Binding assays are used to detect GPCR-like receptor binding activity to neuropeptide or non-peptide ligands. Both functional and binding assays of GPCR-like receptor activity are readily adapted to screens for modulators such as inhibitory compounds.
›DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS · 4 of 5
The invention contemplates a multitude of assays to screen and identify inhibitors of ligand binding to GPCR-like receptors. In one example, the GPCR-like receptor is immobilized and interaction with a binding partner is assessed in the presence and absence of a candidate modulator such as an inhibitor compound. In another example, interaction between the GPCR-like receptor and its binding partner is assessed in a solution assay, both in the presence and absence of a candidate inhibitor compound. In either assay, an inhibitor is identified as a compound that decreases binding between the GPCR-like receptor and its binding partner. Another contemplated assay involves a variation of the di-hybrid assay wherein an inhibitor of protein/protein interactions is identified by detection of a positive signal in a transformed or transfected host cell.
Candidate modulators contemplated by the invention include any chemical compounds, including libraries of chemical compounds. There are a number of different libraries used for the identification of small molecule modulators, including: (1) chemical libraries, (2) natural product libraries, and (3) combinatorial libraries comprised of random peptides, oligonucleotides or organic molecules. Chemical libraries consist of random chemical structures, or analogs of known compounds, or analogs of compounds that have been identified as “hits” or “leads” in prior drug discovery screens, some of which may be derived from natural products or from non-directed synthetic organic chemistry. Natural product libraries are collections of microorganisms, animals, plants, or marine organisms which are used to create mixtures for screening by: (1) fermentation and extraction of broths from soil, plant or marine microorganisms or (2) extraction of plants or marine organisms. Natural product libraries include polyketides, non-ribosomal peptides, and variants (non-naturally occurring) thereof. For a review, see Science 282:63-68 (1998). Combinatorial libraries are composed of large numbers of peptides, oligonucleotides, or organic compounds as a mixture. These libraries are relatively easy to prepare by traditional automated synthesis methods, PCR, cloning, or synthetic methods. Of particular interest are non-peptide combinatorial libraries. Still other libraries of interest include peptide, protein, peptidomimetic, multiparallel synthetic collection, recombinatorial, and polypeptide libraries. For a review of combinatorial chemistry and libraries created therefrom, see Myers, Curr. Opin. Biotechnol . 8:701-707 (1997). Identification of modulators through use of the various libraries described herein permits modification of the candidate “hit” (or “lead”) to optimize the capacity of the “hit” to modulate activity.
Candidate modulators contemplated by the invention can be designed and include soluble forms of binding partners, as well as chimeric, or fusion, proteins thereof. A “binding partner” as used herein broadly encompasses non-peptide modulators, peptide modulators (e.g., neuropeptide variants), antibodies (including monoclonal and polyclonal antibodies, single chain antibodies, chimeric antibodies, biftinctional/bispecific antibodies, humlianized antibodies, human antibodies, and complementary determining region (CDR)-grafted antibodies, including compounds which include CDR and/or antigen-binding sequences, which specifically recognize a polypeptide of the invention), antibody fragments, and modified compounds comprising antibody domains that are immunospecific for the expression product of the identified GPCR-like gene.
A number of assays are known in the art that can identify chemical compounds that bind to or interact with a GPCR-like receptor. Such assays are useful, for example, in methods of identifying candidate modulators described herein, or in methods for identifying specific neuropeptide ligands of a GPCR-like receptor. Assays that measure binding or interaction of compounds with target proteins include assays that identify compounds that inhibit unfolding or denaturation of a target protein, assays that separate compounds that bind to target proteins through affinity ultrafiltration followed by ion spray mass spectroscopy/HPLC methods or other physical and analytical methods, capillary electrophoresis assays and two-hybrid assays.
One such screening method to identify direct binding of test ligands to a target protein is described in U.S. Pat. No. 5,585,277, incorporated herein by reference. This method relies on the principle that proteins generally exist as a mixture of folded and unfolded states, and continually alternate between the two states. When a test ligand binds to the folded form of a target protein (i.e., when the test ligand is a ligand of the target protein), the target protein molecule bound by the ligand remains in its folded state. Thus, the folded target protein is present to a greater extent in the presence of a test ligand which binds the target protein, than in the absence of a ligand. Binding of the ligand to the target protein can be determined by any method which distinguishes between the folded and unfolded states of the target protein. The function of the target protein need not be known in order for this assay to be performed. Virtually any agent can be assessed by this method as a test ligand, including, but not limited to, metals, polypeptides, proteins, lipids, polysaccharides, polynucleotides and small organic molecules.
Another method for identifying ligands of a target protein is described in Wieboldt et al., Anal. Chem., 69:1683-1691 (1997), incorporated herein by reference. This technique screens combinatorial libraries of 20-30 agents at a time in solution phase for binding, to the target protein. Agents that bind to the target protein are separated from other library components by simple membrane washing. The specifically selected molecules that are retained on the filter are subsequently liberated from the target protein and analyzed by HPLC and pneumatically assisted electrospray (ion spray) ionization mass spectroscopy. This procedure selects library components with the greatest affinity for the target protein, and is particularly useful for small molecule libraries.
›DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS · 5 of 5
Alternatively, such binding interactions are evaluated indirectly using the yeast two-hybrid system described in Fields et al., Nature, 340:245-246 (1989), and Fields et al., Trends in Genetics, 10:286-292 (1994), both of which are incorporated herein by reference. The two-hybrid system is a genetic assay for detecting interactions between two proteins or polypeptides. It can be used to identify proteins that bind to a known protein of interest, or to delineate domains or residues critical for an interaction. Variations on this methodology have been developed to clone genes that encode DNA binding proteins, to identify peptides that bind to a protein, and to screen for drugs. The two-hybrid system exploits the ability of a pair of interacting proteins to bring a transcription activation domain into close proximity with a DNA binding domain that binds to an upstream activation sequence (UAS) of a reporter gene, and is generally performed in yeast. The assay requires the construction of two hybrid genes encoding (1) a DNA-binding domain that is fused to a first protein and (2) an activation domain fused to a second protein. The DNA-binding domain targets the first hybrid protein to the UAS of the reporter gene; however, because most proteins lack an activation domain, this DNA-binding hybrid protein does not activate transcription of the reporter gene. The second hybrid protein, which contains the activation domain, cannot by itself activate expression of the reporter gene because it does not bind the UAS. However, when both hybrid proteins are present, the noncovalent interaction of the first and second proteins tethers the activation domain to the UAS, activating transcription of the reporter gene.
When the function of the GPCR-like receptor is unknown and no ligands are known to bind the gene product, the yeast two-hybrid assay can be used to identify proteins that bind to the receptor. In an assay to identify proteins that bind to a GPCR-like receptor, or fragment thereof, a fusion polynucleotide encoding both a GPCR-like receptor or fragment (i.e., a first protein) and a UAS binding domain may be used. In addition, a large number of hybrid genes each encoding a different second protein fused to an activation domain are produced and screened in the assay. Typically, the second protein is encoded by one or more members of a total cDNA or genomic DNA fusion library, with each second protein coding region being fused to the activation domain. This system is applicable to a wide variety of proteins, and it is not even necessary to know the identity or function of the second binding protein. The system is highly sensitive and can detect interactions not revealed by other methods; even transient interactions may trigger transcription to produce a stable mRNA that can be repeatedly translated to yield the reporter protein.
In addition, when the GPCR-like receptor or fragment thereof is known to interact with another protein or nucleic acid, the two-hybrid assay can be used to detect agents that interfere with the binding interaction. Expression of the reporter gene is monitored as different test agents are added to the system. The presence of an inhibitory agent, for example, results in lack of or reduction in a reporter signal.
The candidate modulators identified by the initial screens are evaluated for their effect on neuromuscular function using in vivo or ex vivo systems. A preferred evaluation method exposes the candidate modulators to A. suum neuromuscular strips and records neuromuscular activity relative to control strips exposed to compounds other than the candidate modulator under investigation. Other methods of screening for modulators are described in the following examples, and still other screening methods are well known in the art. Other aspects and advantages of the present invention will be understood upon consideration of the following illustrative examples. Example 1 describes the cloning of GPCR-like receptors; Example 2 provides an analysis of the polynucleotide sequences encoding those receptors in the context of their structural features such as regions and domains; Example 3 discloses GPCR-like receptor expression studies; Example 4 describes antibodies to GPCR-like receptors; Example 5 provides a variety of assays to identify modulators of GPCR-like receptor activity; and Example 6 describes approaches to the identification of candidate GPCR-like receptors, as well as the cloning and characterization of such receptors.
›Examples17
›EXAMPLE 1
Cloning of GPCR-Like Receptors
A. Database Search
The Wormpep database, containing all of the predicted protein sequences encoded by the C. elegans genome, used in these studies, was obtained through the Sanger Centre Web site at sanger.ac.ukJProjects/C_elegans on the world wide web. Wormpep versions 13 (Feb. 13, 1998) through 23 (released Sep. 4, 2000). This database contains 19,430 protein sequences, including 388 splice variants. C. elegans genomic DNA sequences were accessed through ACEDB (Release WS3 4-25). The databases were searched and manipulated using programs from the Wisconsin Package GCG programs.
B. Subcloning of the Coding Region via PCR
Standard molecular biology techniques were performed as described in Ausubel, et al. (Eds.), Protocols in Molecular Biology , John Wiley & Sons (1994). Based on the database analyses described above, 22 candidate GPCR-like receptor genes were chosen for PCR amplification. The sequences of primers used for PCR amplification are provided in SEQ ID NOS: 47-76. PCR primers were designed to incorporate an optimized translation initiation sequence (Kozak et al., Nucl. Acids Res. 15(20):8125-8148 (1987)) around the initiator methionine codon and to incorporate a unique restriction site at the 3′ end for subsequent linearization of the PCR-amplified fragments. The Expand™ High Fidelity PCR System (Boehringer-Mannheim Corp., Indianapolis, Ind.), including a mixture of Taq and Pwo DNA polymerases, was used for the PCR experiments. Reaction volumes of 100 μl contained 2.6 Units of enzyme mix (“Units” defined by the supplier), 200 μM of each standard dNTP, 1×Expand™ HF buffer with 15 mM MgCl 2 , 300 nM of each primer, and 2 μl (approximately 10 ng) of a DNA template preparation, described below. DNA template preparations were the products of a first-strand cDNA synthesis reaction (heat-inactivated after completion) catalyzed by SuperScript reverse transcriptase (GIBCO-BRL, Gaithersburg, Md.) with random hexamers to prime the synthesis using a preparation of total C. elegans RNA as template. PCR amplifications were performed in a Perkin-Elmer 9600 thermocycler using the following conditions: 2 minutes at 94° C.; 30 cycles of 15 seconds at 94° C., 30 seconds at 50° C., 2 minutes at 70° C.; and a final incubation at 70° C. for 7 minutes. PCR products were electrophoresed through 0.8% agarose/TAE gels and visualized using ethidium bromide.
Amplified PCR products were cloned using the pCR2.1 TA-tailed vectors (Stratagene, La Jolla, Calif.). PCR products were excised from the agafose gels and purified using QIAquick columns (Qiagen, Valencia, Calif.). Ligation reactions were performed in accordance with the recommendations of the supplier (Stratagene); Ligation reactions were transformed into E. coli DH5α competent cells and plated on LB plates containing 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal). White colonies were picked and examined for the presence of the appropriate size insert by a commercial mini-prep DNA isolation procedure (Qiagen). To facilitate subsequent expression studies, all cloned receptors were subcloned into the mammalian expression vector, pCR3.1 (Stratagene). Two microgram aliquots of vector and each GPCR-like receptor clone were digested using either BstXI (clones CEGPCR1a, CEGPCR1f; each showing some sequence similarity to Wormpep AC7.1) or EcoRI (clones CEGPCR3 (Wormpep C10C6.2), CEGPCR7 (Wormpep C39E6.6), CEGPCR8 (Wormpep C50F7.1), CEGPCR9 (Wormpep C56G3.1), CEGPCR12h and CEGPCR12u (each having some sequence similarity to Wormpep F41E7.3), CEGPCR13 (having some sequence similarity to Wormpep T05A1.1), and CEGPCR15 (Wormpep C49A9.7)). Restriction endonucleases were purchased from New England Biolabs (Beverly, Mass.) and were used according to the manufacturer's instructions. Following digestion, the linear vector and the GPCR-like receptor sequences were isolated from 1% TAE agarose gels and purified using a QIAquick gel extraction kit (Qiagen). Ligation reactions were performed at 14° C. for 16 hours and contained 100 ng vector DNA, 100 ng GPCR-like insert DNA and 1 Unit T4 DNA ligase (Boehringer Mannheim). Plasmids were propagated in E. coli strain DH5α, then isolated and purified using, a Qiagen column. Subclones were identified and insert orientation determined by restriction endonuclease analysis as follows: CEGPCR1a and CEGPCR15 (SEQ ID NOS: 1 and 33), HindIII; CEGPCR3 and CEGPCR8, (SEQ ID NOS: 43 and 27) AvaI; CEGPCR7, (SEQ ID NO:25) BamHI; CEGPCR13, (SEQ ID NO:13) HincII/HindIII; CEGPCR1f, (SEQ ID NO:3) BamHI/BglII; and CEGPCR9, CEGPCR12h and CEGPCR12u, (SEQ ID NOS: 29, 7 and 9) XbaI.
›EXAMPLE 2
Analysis of the GPCR-Like Receptor Sequences
DNA from pCR2.1 recombinant plasmids containing the expected insert size were sequenced using BigDye dye terminator chemistry (PE Applied Biosystems, Foster City, Calif.) on an ABI 377 automated DNA sequencer. Complete sequences were assembled using Sequencher software (GeneCodes, Ann Arbor, Mich.) and compared against the expected sequence as predicted from Wormpep. Each of the 22 putative GPCR-like receptor clones appeared to contain a full-length coding region, although some of the encoded GPCR-like receptor sequences could not be confirmed as having a sequence reported in the Wormpep database. In particular, clones CEGPCR1a, CEGPCR1f, CEGPCR4, CEGPCR12c, CEGPCR12h, CEGPCR12u, CEGPCR12v, CEGPCR13, CEGPCR14, and CEGPCR18a are known to differ from sequences found in the Wormpep database (see Tables 1 and 2, above), while clones CEGPCR5, CEGPCR6, CEGPCR10, CEGPCR11, CEGPCR19 and CEGPCR20 are not known to have sequences found in the Wormpep database.
All 22 of the cloned GPCR-like receptor sequences exhibited the structural characteristics of GPCRs. Each coding region encoded seven hydrophobic stretches of amino acids consistent with transmembrane domains, the characteristic N-terminal extracellular domain, C-terminal cytoplasmic domain and both extracellular and intracellular loops. Table 6 lists the endpoints of all seven transmembrane domains in each of the 22 GPCR-like receptors. In identifying the transmembrane domains, moreover, the N-terminal extracellular domains are effectively identified (the start of the amino acid sequence to the start of the first transmembrane domain), the three extracellular loop domains (the end of 2TM to the start of 3TM, the end of 4TM to the start of 5TM, and the end of 6TM to the start of 7TM, respectively, in Table 6), the three intracellular loop domains (the end of 1TM to the start of 2TM, the end of 3TM to the start of 4TM, and the end of 5TM to the start of 6TM, respectively, in Table 6), and the C-terminal intracellular domain (the end of 7TM to the end of the amino acid sequence).
›EXAMPLE 3 · 1 of 2
Recombinant Expression of GPCR-Like Receptors
Recombinant pCR3.1 clones encoding invertebrate GPCR-like receptors were transformed into wild-type CHO-K1 cells and functional expression was achieved using a temperature-shift incubation protocol. The CHO-K1 cells were cultured at 37° C. in a humidified atmosphere of 5% CO 2 in air in DMEM medium supplemented with 10% heat-inactivated fetal bovine serum (i.e., FBS), 10 μg/ml gentamicin, and 0.1 mM nonessential amino acids to give complete DMEM medium. Additionally, for incubations at temperatures lower than 37° C., the medium was fortified with 10 mM HEPES (pH 7.4). Cells were transfected with recombinant pCR3.1 clones containing GPCR-like receptor DNAs, using LipofectAMINE PLUS™, essentially according to the manufacturer's instructions. Briefly, CHO-K1 cells were plated on 10 cm sterile tissue culture dishes (Coming Glass Works, Corning, N.Y.) and were incubated until the plates showed about 50-60% confluent growth. In separate plastic tubes, PLUS (20 μl/plate; Life Technologies) was added to 0.75 ml/plate of recombinant plasmid DNA in OptiMEM (5 μg/plate; Gibco/BRL—Life Technologies), mixed and incubated at room temperature for 15 minutes. Separately, LipofectAMINE (30 μl/plate) was mixed with 0.75 ml OptiMEM and added to the pre-complexed DNA/PLUS mixture and incubated at room temperature for 15 minutes. The culture medium was then replaced with serum-free transfection medium (DMEM, 5 ml/plate). The DNA-PLUS-LipofectAMINE complex was then added to the plates (1.5 ml/plate) and mixed gently into the medium, followed by a 3 hour incubation at 37° C. in 5% CO 2 . At this time, the medium was supplemented with complete DMEM medium containing 20% FBS (6.5 ml/plate) and incubation was continued at 37° C. in 5% CO 2 . After a total of 24 hours at 37° C. in 5% CO 2 , the medium was replaced with complete DMEM medium fortified with 10 mM HEPES (10 ml/plate) and the cells were moved to an incubator maintained at 28-30° C. in a humidified atmosphere of 5% CO 2 in air. The incubation continued for an additional 24 hours, typically at the lower temperature indicated above. A plasmid encoding Green Fluorescent Protein (i.e., GFP, 4 μg/plate) was used for transient GFP expression in CHO-K1 cells to estimate the transfection yields under the same conditions as used for GPCR-like receptor polynucleotides.
To assess expression, membranes were prepared from transfected CHO-K1 cells. Transfected cells were washed once with ice-cold Dulbecco's phosphate-buffered saline (i.e., PBS), 5 ml per 10 cm plate, and the cells were scraped into 5 ml of the same buffer. Cell suspensions from multiple plates were combined and centrifuged at 500×g for 10 minutes at 4° C. The cell pellet was reconstituted in ice-cold TEE (25 mM Tris, pH 7.4, 5 mM EGTA, 5 mM EDTA). Convenient aliquots were snap-frozen in liquid nitrogen and stored at −70° C. After thawing, the cells were homogenized and centrifuged at 500×g for 5 minutes at 4° C. to pellet nuclei and unbroken cells. Tile supernatant was centrifuged at 47,000×g for 30 minutes at 4° C. The membrane pellet was washed once with TEE, resuspended in 20 mM HEPES, pH 7.4, 100 mM NaCl, 10 mM MgCl 2 , 1 mM EDTA (assay buffer), distributed into aliquots of 0.5-1.0 ml, and frozen in liquid nitrogen. Membrane aliquots were stored at −70° C. Membrane protein concentration was determined using the BCA Protein Assay Reagent from Pierce (Rockford, Ill.), with bovine serum albumin (i.e., BSA) as a standard.
Expression of the recombinant GPCR-like receptors was assessed using a function-based [ 35 S]GTPγS assay, which is described in greater detail in Example 5F, below. In general, the post-transfection reduction in incubation temperature led to significant increases in detectable GPCR-like receptor activity. More specifically, a controlled study of three cloned GPCR-like receptors, CEGPCR3 (Wormpep C10C6.2), CEGPCR12c (exhibiting some sequence similarity to Wormpep F41E7.3), and CEGPCR7 (Wormpep C39E6.6), demonstrated that the temperature reduction resulted in significant increases in the expression levels of all three receptors. The GTPγS assay described in Example 5F was also used to screen for neuropeptide (including FaRP) ligands binding to invertebrate GPCR-like receptors. The results are presented in Tables 8-13, below, which identify GPCR-like receptors binding to one or more neuropeptide (e.g., FaRP) ligands, GPCR-like receptors binding to neuropeptide ligands having specified structures, the sources of those particular ligands, and EC 50 (nM) values. As shown in the tables, such ligands were identified for CEGPCR3, CEGPCR4, CEGPCR5, CEGPCR7, CEGPCR12c, CEGPCR12h, CEGPCR12u, CEGPCR12v, CEGPCR16, CEGPCR19.1, and CEGPCR19.2, and on that basis these receptors have been identified as neuropeptide (FaRP) receptors.
CEGPCR3, CEGPCR4, CEGPCR5, CEGPCR12c and CEGPCR16 were not detectably activated by their matched ligands, GNSFLRFamide (SEQ ID NO:88), PDVDHVFLRFamide (SEQ ID NO:94), GLGPRPLRFamide (SEQ ID NO:86), PDVDHVFLRFamide (SEQ ID NO: 94), and ASEDALFGTMRFamide (SEQ ID NO: 77), respectively, when cell membranes were prepared from cells transfected and maintained at 37° C. In contrast, a temperature shift to 28° C. 24 hours after transfection resulted in significant stimulation of ligand-induced [ 35 S]GTPγS binding.
In the case of the CEGPCR7-AF9 system, cell transfections and constant temperature (37° C.) incubations were performed three times. No stimulation of the GLGPRPLRFamide (AF9)-induced [ 35 S]GTPγS binding was detected in two of these experiments while in the third one, a 1.4 fold elevated [ 35 S]GTPγS binding was recorded. The extent of this stimulation was, however, considerably smaller than the stimulation in activity achieved by using the cooling (28° C.) protocol (2.6 fold elevation in activity). Thus, the temperature downshift was important for the functional expression of C. elegans GPCRs expressed in mammalian cells. In contrast, the responses of the human D2 receptor control to its ligand, dopamine, were the same regardless of the transfection protocol used. Moreover, no temperature-dependent receptor activity was detected when the hD2 receptor control was exposed to a number of neuropeptides, including FaRPs, indicating that the temperature-dependent GPCR-like receptor activity was specific to the cloned receptors.
›EXAMPLE 3 · 2 of 2
All invertebrate GPCR-like receptors, including those receptors identified in Table 6, are screened against the invertebrate neuropeptides, including many FaRPs. It is expected that each one of the invertebrate GPCR-like receptors will be activated by one or more neuropeptides of the class including FaRPs, using an assay disclosed in Example 5, such as the GTPγS assay.
›EXAMPLE 4
Antibodies to GPCR-Like Receptors
Standard techniques are employed to generate polyclonal or monoclonal antibodies to the GPCR-like receptors, and to generate useful antigen-binding fragments thereof or variants thereof. Such protocols can be found, for example, in Sambrook et al., Molecular Cloning: a Laboratory Manual . Second Edition, Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory (1989); Harlow et al. (Eds), Antibodies A Laboratory Manual ; Cold Spring Harbor Laboratory; Cold Spring Harbor, N.Y. (1988). In one embodiment, recombinant GPCR-like polypeptides (or cells or cell membranes containing such polypeptides) are used as antigens to generate the antibodies. In another embodiment, one or more peptides having amino acid sequences corresponding to an immunogenic portion of a GPCR-like receptor (e.g., 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more amino acids) are used as antigen. Peptides corresponding to extracellular portions of GPCR-like receptors, especially hydrophilic extracellular portions, are preferred. The antigen may be mixed with an adjuvant or linked to a hapten to increase antibody production. Polyclonal and monoclonal antibodies, chimeric (e.g., humanized) antibodies, fragments of antibodies, and all other forms of antibody molecules disclosed herein arc referred to collectively as antibody products.
A. Polyclonal or Monoclonal Antibodies
As one exemplary protocol, a recombinant GPCR-like polypeptide or a synthetic fragment thereof is used to immunize a mouse for generation of monoclonal antibodies (or larger mammal, such as a rabbit, for polyclonal antibodies). To increase antigenicity, peptides are conjugated to Keyhole Lympet Hemocyanin (Pierce), according to the manufacturer's recommendations. For an initial injection, the antigen is emulsified with Freund's Complete Adjuvant and injected subcutaneously. At intervals of two to three weeks, additional aliquots of GPCR-like receptor antigen are emulsified with Freund's Incomplete Adjuvant and injected subcutaneously. Prior to the final booster injection, a serum sample is taken from the immunized mice and assayed by Western blot to confirm the presence of antibodies that immunoreact with a GPCR-like polypeptide. Serum from the immunized animals may be used as a polyclonal antisera or used to isolate polyclonal antibodies that recognize a GPCR-like receptor. Alternatively, the mice are sacrificed and their spleens are removed for generation of monoclonal antibodies.
To generate monoclonal antibodies, the spleens are placed in 10 ml serum-free RPMI 1640, and single-cell suspensions are formed by grinding the spleens in serum-free RPMI 1640, supplemented with 2 mM L-glutamine, 1 mM sodium pyruvate, 100 Units/ml penicillin, and 100 μg/ml streptomycin (RPMI) (Gibco, Canada). The cell suspensions are filtered and washed by centrifugation and resuspended in serum-free RPMI. Thymocytes taken from three naive Balb/c mice are prepared in a similar manner and used as a Feeder Layer. NS-1 myeloma cells, kept in log phase in RPMI with 10% (FBS (Hyclone Laboratories, Inc., Logan, Utah) for three days prior to fusion, are centrifuged and washed as well.
To produce hybridoma fusions, spleen cells from the immunized mice are combined with NS-1 cells and centrifuged, and the supernatant is aspirated. The cell pellet is dislodged by tapping the tube, and 2 ml of 37° C. PEG 1500 (50% in 75 mM HEPES, pH 8.0) (Boehringer-Mannheim) is stirred into the pellet, followed by the addition of serum-free RPMI. Thereafter, the cells are centrifuged and resuspended in RPMI containing 15% FBS, 100 μM sodium hypoxanthine, 0.4 μM aminopterin, 16 μM thynidine (HAT) (Gibco), 25 Units/ml IL-6 (Boehringer-Mannheim) and 1.5×10 6 thymocytes/ml and plated into 10 Coming flat-bottom 96-well tissue culture plates (Coming, Coming N.Y.).
On days 2, 4, and 6 after the fusion, 100 μl of medium is removed from the wells of the fusion plates and replaced with fresh medium. On day 8, the fusions are screened by ELISA, testing for the presence of mouse IgG that binds to a GPCR-like receptor polypeptide. Selected fusions are further cloned by dilution until monoclonal cultures producing anti-GPCR-like receptor antibodies are obtained.
B. GPCR-like Receptor-Neutralizing Antibodies from Phage Display
GPCR-like receptor-neutralizing antibodies are generated by phage display techniques such as those described in Aujame et al., Human Antibodies , 8(4):155-168 (1997); Hoogenboom, TIBTECH , 15:62-70 (1997); and Rader et al., Curr. Opin. Biotechnol ., 8:503-508 (1997), all of which are incorporated by reference. For example, antibody variable regions in the form of Fab fragments or linked single chain Fv fragments are fused to the amino terminus of filaments phage minor coat protein pIII. Expression of the fusion protein and incorporation thereof into the mature phage coat results in phage particles that present an antibody on their surface and contain the genetic material encoding the antibody. A phage library comprising such constructs is expressed in bacteria, and the library is screened for GPCR-like receptor-specific phage-antibodies using labeled or immobilized GPCR-like receptor as antigen-probe.
C. GPCR-like Receptor-Neutralizing Antibodies from Transgenic Animals
GPCR-like receptor-neutralizing antibodies are generated in transgenic animals, such as mice, essentially as described in Bruggemann et al., Immunol. Today 17(8):391-97 (1996) and Bruggemann et al., Curr. Opin. Biotechnol . 8:455-58 (1997). Transgenic mice carrying V-genie segments in germline configuration, and expressing the transgenes in their lymphoid tissue, are immunized with a GPCR-like polypeptide composition using conventional immunization protocols. Hybridomas are generated from B cells of the immunized mice using conventional protocols and screened to identify hybridomas secreting anti-GPCR-like receptor antibodies (e.g., as described above).
›EXAMPLE 5 · 1 of 7
Assays to Identify Modulators of GPCR-Like Receptor Activity
Two general approaches exist for the discovery of novel compounds that are binding partners for receptors, especially GPCRs. Each of the general approaches is compatible with high throughput screening (i.e., HTS) formats, which are preferred formats for identifying ligands and other binding partners of GPCR-like receptor polypeptides, such as modulators of receptor activity. The first approach involves measuring the binding of a known ligand, preferably labeled with a radiolabel, to a preparation that contains the receptor, either found in native tissue or based on expression of the gene encoding the receptor, typically in a heterologous system. The recombinant system involves the expression of a recombinant GPCR, and is the presently preferred system for binding assays. The second approach to the identification of GPCR-like receptor binding partners involves the measurement of the activity of the receptor, which may be influenced by either the binding of ligands that are agonists, and elevate receptor activity, or by the binding of antagonists, which interfere with agonist binding, thereby reducing the level of receptor activity. As for binding assays, recombinant systems are the presently preferred forms for function-based assays.
Receptor Binding Assays (RBA) as HTS Systems for Drug Discovery
The literature is replete with examples of the use of radiolabeled ligands in FITS binding assays for drug discovery (see Williams, Med. Res. Rev. 11:147-184 (1991); Sweetnam et al., J. Nat. Prod. 56:441-455 (1993) for review). It is also possible to screen for novel anthielmintic compounds with radiolabeled ligands in HTS binding screens (Geary et al., 1999). However, native nematode tissue, particularly tissue from parasitic nematodes, is difficult to acquire and native nematode tissue binding assays are not well-suited for HTS because the density of receptor sites in these preparations is typically much lower than found in mammalian tissues (Thompson, et al., Parasitology 113:S217-S238 (1996)). Other reasons that recombinant receptors are preferred for HTS binding assays include better specificity (higher relative purity) and ability to generate large amounts of receptor material (see Hodgson, Bio/Technology 10:973-980 (1992)).
A variety of heterologous systems are available for expression of recombinant receptors and are well known to those skilled in the art. Such systems include bacteria (Strosberg et al., Trends in Pharm. Sci. 13:95-98 (1992)), yeast (Pausch, Trends in Biotech. 15:487-494 (1997)), several kinds of insect cells (Vanden Broeck, Intl. Rev. Cytol. 164:189-268 (1996)), amphibian cells (Jayawickreme et al., Curr. Opin. Biotechnol. 8:629-634 (1997)) and several mammalian cell lines (CHO, HEK293, COS, etc.; see Gerhardt et al., Eur. J. Pharmacol. 334:1-23 (1997); Wilson et al., Brit. J. Pharmacol. 125:1387-1392 (1998)). These examples do not preclude the use of other possible cell expression systems, including cell lines obtained from nematodes (WO 98/37177).
A C. elegans orphan GPCR expressed in one of the described recombinant systems can be used for HTS binding assays in conjunction with its defined ligand, in this case the corresponding neuropeptide (e.g., FaRP) that activates it. The identified peptide is labeled with a suitable radioisotope, including, but not limited to, 125 I (preferred; see Geary et al., 1999), 3 H, 35 S or 32 P, by methods that are well known to those skilled in the art. Alternatively, the peptides may be labeled by well-known methods with a suitable fluorescent derivative (Baindur et al., Drug Dev. Res. 33:373-398 (1994); Rogers, Drug Disc. Today 2:156-160 (1997)). Radioactive ligand specifically bound to the receptor in membrane preparations made from the cell line expressing the recombinant protein can be detected in HTS assays in one of several standard ways, including filtration of the receptor-ligand complex to separate bound ligand from unbound ligand (Williams, 1991; Sweetnam et al., 1993). Alternative methods include a scintillation proximity assay (SPA) or a FlashPlate format in which such separation is unnecessary (Nakayama et al., Drug Disc. & Dev. 1:85-91 (1998); Boss et al., J. Biomol. Screening 3:285-292 (1998)). Binding of fluorescent ligands can be detected in various ways, including fluorescence energy transfer (FRET), direct spectrophotofluorometric analysis of bound ligand, or fluorescence polarization (see Rogers, 1997; Hill, Curr. Opin. in Drug Disc. & Dev. 1:92-97 (1998)).
A binding assay that has proven useful is a Flp18 binding assay or screen.
A peptide ligand screening assay that measures ligand binding to a suitable GPCR-like receptor according to the invention is exemplified using a modified C. elegans Flp18 peptide ligand. The modified C. elegans Flp18 peptide, YDVPGVLRFamide (SEQ ID NO:147), is extended by an N-terminal tyrosine residue. Iodination of the peptide is accomplished using a standard chloramine T procedure. Added to a 2 ml glass vial are 10 μl of a 1 mM water solution of the Flp18 peptide, 10 μl of 0.1M (pH 7.99) sodium phosphate buffer, 1.0 mCi [ 125 I] sodium iodide and 5 μl of a 2 mg/ml chloramine T solution (in the phosphate buffer). The mixture is vortexed for 60 seconds and the reaction is stopped by the addition of 25 μl of a 5 mg/ml solution of sodium metabisulfite in phosphate buffer. The mixture is then subjected to HPLC gradient fractionation by injecting it onto a Vydac C18 (0.45×15 cm) column. The gradient used is 70% A and 30% B at time zero to 20% A and 80% B at time 25 minutes (A=0.1M NH 4 acetate in water; B=0.1M NH4 acetate in water 40%: CH 3 CN 60%, v/v). The flow rate is 1.0 ml/minute. Samples are collected into 0.25 ml capture buffer (0.1M sodium phosphate buffer with 0.5% bovine serum albumin, 0.1% Triton X100 and 0.05% Tween 20) at 30 second intervals from t=8 to t=20 minutes. Monoiodo YDVPGVLRFamide (SEQ ID NO:119) typically elutes at t=11 minutes and the yield is approximately 100 μCi in 0.75 ml.
›EXAMPLE 5 · 2 of 7
Membranes for use in the assay are derived from cells stably transfected with a GPCR-like receptor expression construct. The transfected cells are grown in 10 cm dishes to subconfluence, rinsed once with 5 ml of ice-cold Ca 2+ /Mg 2+ —free PBS, and scraped into 5 ml of the same buffer. Cells are pelleted by centrifugation (500×g, 5 minutes), resuspended in 25 mM Tris, pH 7.5, 5 mM EDTA, 5 mM EDTA, pH 7.5 (TEE), and frozen in liquid nitrogen. After thawing, the cells are homogenized using a Dounce homogenizer (one ml TEE per plate of cells), and centrifuged at 1,000×g for 5 minutes to remove nuclei and unbroken cells.
The homogenate supernatant is centrifuged at 20,000×g for 20 minutes to isolate the membrane fraction, and the membrane pellet is washed once with TEE and resuspended in binding buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 1 mM EDTA). The resuspended membranes are optionally frozen in liquid nitrogen and stored at −70° C. until use.
Aliquots of cell membranes prepared as described above and stored at −70° C. are thawed, homogenized, and diluted into assay buffer as described below.
The peptide ligand screening assay is facilitated by use of 96-well plates Millipore Multiscreen® filtration plates (FB opaque 1.0 μM glass fiber type B, catalog number MAFBNOB50). Protein reaction products are collected using a Millipore Multiscreen® solvent resistant manifold (catalog number MAVMO960R). Each replicate of the assay is contained in a single well and has a volume of 100 ul containing 5 ug protein (preparation described above). Each test group contains two replicates. For each test compound, one group is run with [ 125 I]YDVPGVLRFamide (SEQ ID NO:119) only to determine total binding, and one group is run with 1 μM (or as designated) concentration of the test compound and [ 125 I]YDVPGVLRFamide (SEQ ID NO:119; for non-specific binding). The order of adding reagents for each replicate is as follows: assay buffer (20 mM HEPES, 10 mM MgCl 2 , 1% bovine serum albumin, pH 7.4) test compound (made up in assay buffer), [ 125 I]YDVPGVLRFamide (SEQ ID NO:119; in assay buffer) and membrane suspension (in assay buffer). The addition of the membrane suspension initiates the binding reaction which is run for 30 minutes at room temperature (22° C.). Following the 30 minute incubation each plate is placed on the filtration manifold and vacuum is applied, pulling the liquid through the filter (the liquid is discarded) and catching the protein on the filters in each well. For washing, the vacuum is released and 200 μl assay buffer is added to each well followed by reapplication of the vacuum. This washing is repeated twice more (total of 3 washes for each replicate). Following washing, the plastic covering on the underside of each plate is removed and the plate is placed in a bottom-sealed Microbeta® scintillation counting cassette (catalog number 1450-105). Twenty five μl of scintillant is added to each well and the plate is placed on a rotary shaker at 80 rpm for one hour and then allowed to sit overnight. The following day, the plate is counted in a Microbeta® scintillation counter. The mean non-specific binding is subtracted from the mean total binding to yield specific binding for both the standard (YDVPGVLRFamide, SEQ ID NO:147) and the unknowns.
Response-based GPCR HTS Systems
It is well known that activation of heterologous receptors expressed in recombinant systems results in a variety of biological responses, which are typically mediated by G proteins expressed in the host cells. Agonist binding to a GPCR results in exchange of bound GDP for GTP at a binding site on the G subunit; one can use a radioactive, non-hydrolyzable derivative of GTP, such as [ 35 S]GTPγS, to measure binding of an agonist to the receptor. (Seifert et al., Eur. J. Biochem. 255:369-382 (1998).) One can also use this binding to measure the ability of antagonists to bind to the receptor by decreasing binding of GTP [ 35 S] in the presence of a known agonist.
The G proteins required for functional expression of heterologous GPCRs can be native constituents of the host cell or can be introduced through well-known recombinant technology. The G proteins can be intact or chimeric. Often, a nearly universally competent G protein (e.g., G α16 ) is used to couple any given receptor to a detectable response pathway. G protein activation results in the stimulation or inhibition of other native proteins, events that can be linked to a measurable response.
Examples of such biological responses include, but are not limited to, the following responses: the ability to survive in the absence of a limiting nutrient in specifically engineered yeast cells (Pausch, 1997); changes in intracellular Ca 2+ concentration as measured by fluorescent dyes (Murphy et al., Curr. Opin. in Drug Disc. & Dev. 1:192-199 (1998)). Fluorescence changes can also be used to monitor ligand-induced changes in membrane potential or intracellular pH; an automated system suitable for HTS has been described for these purposes (Schroeder et al., J. Biomol. Screening 1:75-80 (1996)). Melanophores prepared from Xenopus laevis show a ligand-dependent change in pigment organization in response to heterologotis GPCR activation; this response is adaptable to HTS formats (Jayawickreme et al., 1997). Assays are also available for the measurement of common second messengers, including cAMP, phosphoinositides and arachidonic acid. Set forth in the following subsections are exemplary function-based assays for identifying modulators (agonists and antagonists) of GPCR-like receptor activity. Among the modulators that can be identified by these assays are natural ligand compounds of the receptor; synthetic analogs and derivatives of natural ligands; antibodies, antibody fragments, and/or antibody-like compounds derived from natural antibodies or from antibody-like combinatorial libraries; and/or synthetic compounds identified by high throughput screening of libraries; and other libraries known in the art. All modulators that bind GPCR-like receptors are useful for identifying GPCR-like polypeptides in tissue samples (e.g., for diagnostic purposes, pathological purposes, and other purposes known in the art). Agonist and antagonist modulators are useful for up-regulating and down-regulating GPCR-like receptor activity, respectively, to treat invertebrate infestations and/or infections of man, other animals, plants or the environment. GPCR-like receptor binding partners also may be used to deliver a therapeutic compound or a label to cells that express a GPCR-like receptor (e.g., by attaching the compound or label to the binding partner). The assays may be performed using single putative modulators; they may also be performed using a known agonist in combination with candidate antagonists (or visa versa).
›EXAMPLE 5 · 3 of 7
A. cAMP Assays
In one type of assay, levels of cyclic adenosine monophosphate (cAMP) are measured in GPCR-like receptor-transfected cells that have been exposed to candidate modulator compounds. Protocols for cAMP assays have been described in the literature. [See, e.g., Sutherland et al., Circulation, 37:279 (1968); Frandsen, et al., Life Sciences, 18:529-541 (1976); Dooley et al., J. Pharm. & Exp. Therap., 283(2):735-41 (1997); and George et al., J. Biomol. Screening, 2(4):235-40 (1997).] An exemplary protocol for such an assay, using an Adenylyl Cyclase Activation FlashPlate® Assay from NEN™ Life Science Products, is set forth below.
Briefly, the GPCR-like receptor coding sequence (e.g., a cDNA or intronless genomic DNA) is subcloned into a commercial expression vector, such as pzeoSV2 (Invitrogen, San Diego, Calif.), and transiently transfected into Chinese Hamster Ovary (CHO) cells using known methods, such as the transfection reagent FuGENE 6 (Boehringer-Mannheim) and the transfection protocol provided in the product insert. The transfected CHO cells are seeded into the 96-well microplates of the FlashPlate® assay kit, which are coated with solid scintillant to which antisera to cAMP has been bound. For a control, some wells are seeded with wild type (untransfected) CHO cells. Other wells on the plate receive various amounts of cAMP standard solution for use in creating a standard curve.
One or more test compounds are added to the cells in each well, with water and/or compound-free medium/diluent serving as a control. After treatment, cAMP is allowed to accumulate in the cells for exactly 15 minutes at room temperature. The assay is terminated by the addition of lysis buffer containing [ 125 I]-labeled cAMP, and the plate is counted using a Packard Topcount™ 96-well microplate scintillation counter. Unlabeled cAMP from the lysed cells (or from standards) competes with the fixed amounts of [ 125 I]-cAMP for antibody bound to the plate. A standard curve is constructed, and cAMP values for the unknowns are obtained by interpolation. Changes in intracellular cAMP level of the cells in response to exposure to a test compound are indicative of GPCR-like receptor modulating activity. Modulators that act as agonists at receptors which couple to the G s subtype of G proteins will stimulate production of cAMP, leading to a measurable 3-10 fold increase. Agonists OF receptors which couple to the G i/o subtype of G proteins will inhibit forskolin-stimulated cAMP production, leading to a measurable decrease of 50-100%. Modulators that act as inverse agonists will reverse these effects at receptors that are either constitutively active or activated by known agonists.
B. Aequorin Assays
In another function based assay, cells (e.g., CHO cells) are transiently co-transfected with both a GPCR-like receptor expression construct and a construct that encodes the photoprotein apoaquorin. In the presence of the cofactor coelenterazine, apoacquorin will emit a measurable luminescence that is proportional to the amount of intracellular (cytoplasmic) free calcium. [See generally Cobbold P. H. and Lee, J. A. C. “Aequorin measurements of cytoplasmic free calcium. In: McCormack J. G. and Cobbold P. H., eds., Cellular Calcium: A Practical Approach . Oxford:IRL Press (1991); Stables et al., Analytical Biochemistry , 252: 115-26 (1997); and Haugland, R. P. Handbook of Fluorescetit Probes and Research Chemicals . Sixth edition. Eugene Oreg.: Molecular Probes (1996).]
In one exemplary assay, a GPCR-like receptor is subcloned into the commercial expression vector pzeoSV2 (Invitrogen, San Diego, Calif.) and transiently co-transfected along with a construct that encodes the photoprotein apoaequorin (Molecular Probes, Eugene, Oreg.) into CHO cells using the transfection reagent FuGENE 6 (Boehringer-Mannheim) and the transfection protocol provided in the product insert. Alternatively, the GPCR-like receptor may be subcloned into pCR3.1 and transfected into CHO cells using LipofectAMINE PLUS™, as described above.
The cells are cultured for 24 hours at 37° C. in DMEM (Gibco/BRL, Gaithersburg, Md.) supplemented with 10% fetal bovine serum, 2 mM glutamine, 10 Units/ml penicillin and 10 μg/ml streptomycin, at which time the medium is changed to serum-free DMEM containing 5 μM coelenterazine (Molecular Probes, Eugene, Oreg.), and the cells are cultured for two additional hours at 37° C. Cells are then detached from the plate using VERSEN (Gibco/BRL), washed, and resuspended at 200,000 cells/ml in serum-free DMEM.
Dilutions of candidate GPCR-like receptor modulator drugs are prepared in serum-free DMEM and dispensed into wells of an opaque 96-well assay plate, 50 μl/well. Plates are loaded onto an MLX microtiter plate luminometer (Dynex Technologies, Inc., Chantilly, Va.). The instrument is programmed to dispense a 50 μl cell suspension into each well, one well at a time, and immediately read luminescence for 15 seconds. Dose-response curves for the modulator candidates are constructed using the area under the curve for each light signal peak. Data are analyzed with SlideWrite, using the equation for 1-site ligand, and EC 50 values are obtained. Changes in luminescence caused by the drugs are considered indicative of modulatory activity. Modulators that act as agonists at receptors which couple to the G q subtype of G proteins give an increase in luminescence of up to 100 fold. Modulators that act as inverse agonists will reverse this effect at receptors that are either constitutively active or activated by known agonists.
C. Luciferase Reporter Gene Assay
The photoprotein luciferase provides another useful tool for assaying for modulators of GPCR-like receptor activity. Cells (e.g., CHO cells or COS 7 cells) are transiently co-transfected with both a GPCR-like receptor expression construct (e.g., a GPCR-like receptor in pzeoSV2 (Invitrogen, San Diego, Calif.)) and a reporter construct which includes a luciferase coding region downstream from a transcription factor, either the cAMP-response element (CRE), AP-1, or NF-κB. Agonist binding to receptors coupled to the G s subtype of G proteins leads to increases in cAMP, activating the CRE transcription factor and resulting in expression of the luciferase gene. Agonist binding to receptors coupled to the G q subtype of G proteins leads to production of diacylglycerol that activates protein kinase C, which activates the AP-1 or NF-κB transcription factors resulting in expression of the luciferase gene. Expression levels of luciferase reflect the activation status of the signaling events. [See generally George et al., Journal of Biomolecular Screening , 2(4): 235-40 (1997); and Stratowa et al., Current Opinion in Biotechnology , 6: 574-81 (1995).] Luciferase activity may be quantitatively measured using, e.g., luciferase assay reagents that are commercially available from Promega (Madison, Wis.).
›EXAMPLE 5 · 4 of 7
In one exemplary assay, CHO cells are plated in 24-well culture dishes at a density of 100,000 cells/well one day prior to transfection and cultured at 37° C. in DMEM (Gibco/BRL, Gaithersburg, Md.) supplemented with 10% fetal bovine serum, 2 mM glutamine, 10 Units/ml penicillin and 10 μg/ml streptomycin. Cells are transiently co-transfected with both a GPCR-like receptor expression construct and a reporter construct containing the luciferase gene. The reporter plasmids CRE-luciferase, AP-1-luciferase and NF-κB-luciferase may be purchased from Stratagene (LaJolla, Calif.). Transfections are performed using FuGENE 6 transfection reagent (Boehringer-Mannheim), following the protocol provided in the product insert. Cells transfected with the reporter construct alone are used as a control. Twenty-four hours after transfection, cells are washed once with phosphate buffered saline (PBS) pre-warmed to 37° C. Serum-free DMEM is then added to the cells either alone (control) or with one or more candidate modulators and the cells are incubated at 37° C. for five hours. Thereafter, cells are washed once with ice-cold PBS and lysed by the addition of 100 μl of lysis buffer/well (from the luciferase assay kit, Promega, Madison, Wis.). After incubation for 15 minutes at room temperature, 15 μl of the lysate is mixed with 50 μl substrate Solution (Promega) in an opaque white 96-well plate, and the luminescence is read immediately on a Wallace model 1450 MicroBeta scintillation and luminescence counter (Wallace Instruments, Gaithersburg, Md.).
Differences in luminescence in the presence versus the absence of a candidate modulator compound are indicative of modulating activity. Receptors that are either constitutively active or activated by agonists typically give a 3-20 fold stimulation of luminescence compared to cells transfected with the reporter gene alone. Modulators that act as inverse agonists will reverse this effect.
D. Intracellular Calcium Measurement Using FLIPR
Changes in intracellular calcium levels are another recognized indicator of C protein-coupled receptor activity, and such assays can be employed to assay for modulators of GPCR-like receptor activity. For example, CHO cells stably transfected with a GPCR-like receptor expression construct are plated at a density of 4×10 4 cells/well in Packard black-walled 96-well plates specially designed to isolate fluorescent signals to individual wells. The cells are incubated for 60 minutes at 37° C. in modified Dulbecco's PBS (D-PBS), containing 36 mg/L pyruvate and 1 g/L glucose with the addition of 2.5 mM probenecid (Sigma Chemical Co.) and one of four calcium indicator dyes (Fluo3™ AM, Fluo4™ AM, Calcium Green™-1 AM, or Oregon Green™ 488 BAPTA-1 AM) at a concentration of 4 μM. Plates are washed once with modified D-PBS without 1% fetal bovine serum and incubated for 10 minutes at 37° C. to remove residual dye from the cellular membrane. In addition, a series of washes with modified D-PBS is performed immediately prior to activation of the calcium response.
Calcium response is initiated by the addition of one or more candidate receptor agonist compounds, or a positive control such as a calcium ionophore A23187 (10 μM), or ATP (4 μM). Fluorescence is measured by Molecular Device's FLIPR with an argon laser, excitation at 488 nm. [See, e.g., Kuntzweiler et al., Drug Development Research , 44(1): 14-20 (1998).] The F-stop for the detector camera was set at 2.5 and the length of exposure was 0.4 msec. Basal fluorescence of cells was measured for 20 seconds prior to addition of agonist, ATP, or A23187, and was subtracted from the response signal. The calcium signal is measured for approximately 200 seconds, taking readings every two seconds. The calcium ionophore and ATP increase the calcium signal 200% above baseline levels. In general, activated GPCRs increase the calcium signal approximately 10-15% above baseline signal.
E. Mitogenesis Assay
In mitogenesis assays, the ability of candidate modulators to induce or inhibit GPCR-like receptor-mediated cell growth is determined. [See, e.g., Lajiness et al., Journal of Pharmacology and Experimental Therapeutics , 267(3): 1573-81 (1993).] For example, CHO cells stably expressing a GPCR-like receptor are seeded into 96-well plates at a density of 5000 cells/well and grown at 37° C. in DMEM with 10% fetal calf serum for 48 hours, at which time the cells are rinsed twice with serum-free DMEM. After rinsing, 80 μl of fresh DMEM, or DMEM containing a known mitogen, is added along with 20 μl DMEM containing varying concentrations of one or more test compounds diluted in serum-free medium. As controls, some wells on each plate receive serum-free medium alone, and some receive medium containing 10% fetal bovine serum. Untransfected cells or cells transfected with the vector alone also may serve as controls.
After culture for 16-18 hours, 1 μCi/well of [ 3 H]-thymidine is added to the wells and cells are incubated for an additional 2 hours at 37° C. The cells are trypsinized and harvested onto filter mats with a cell harvester (Tomtec) and the filters are counted in a Betaplate counter. The incorporation of [ 3 H]-thymidine in serum-free test wells is compared to the results achieved in cells stimulated with serum. Use of multiple concentrations of test compounds permits creation and analysis of dose-response curves using the non-linear, least squares fit equation: A=B×[C/(D+C)]+G where A is the percent of serum stimulation; B is the maximal effect minus baseline; C is the EC 50 ; D is the concentration of the compound; and G is the maximal effect. Parameters B, C and G are determined by Simplex optimization.
Agonists that bind to the receptor are expected to increase [ 3 H]-thymidine incorporation into cells, showing up to 80% of the response to serum. Antagonists that bind to the receptor will inhibit the stimulation seen with a known agonist by up to 100%.
F. [ 35 S]GTPγS Binding Assay
Because G protein-coupled receptors signal through intracellular G proteins whose activity involves GTP/GDP binding and hydrolysis, measurement of binding of the non-hydrolyzable GTP analog [ 35 S]GTPγS in the presence and absence of putative modulators provides another indicator of modulator activity. [See, e.g., Kowal, et al., Neuropharmacology , 37: 179-87 (1998).]
›EXAMPLE 5 · 5 of 7
In one exemplary assay, cells stably transfected with a GPCR-like receptor expression construct are grown in 10 cm dishes to subconfluence, rinsed once with 5 ml of ice-cold Ca 2+ /Mg 2+ —free PBS, and scraped into 5 ml of the same buffer. Cells are pelleted by centrifugation (500×g, 5 minutes), resuspended in 25 mM Tris, pH 7.5, 5 mM EDTA, 5 mM EDTA, pH 7.5 (TEE), and frozen in liquid nitrogen. After thawing, the cells are homogenized using a Dounce homogenizer (one ml TEE per plate of cells), and centrifuged at 1,000×g for 5 minutes to remove nuclei and unbroken cells.
The homogenate supernatant is centrifuged at 20,000×g for 20 minutes to isolate the membrane fraction, and the membrane pellet is washed once with TEE and resuspended in binding buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 1 mM EDTA). The resuspended membranes can be frozen in liquid nitrogen and stored at −70° C. until use.
Aliquots of cell membranes prepared as described above and stored at −70° C. were thawed, homogenized, and diluted into buffer containing 20 mM HEPES, pH 7.4, 10 mM MgCl 2 , 1 mM EDTA, and 100 mM NaCl at a concentration of 10-50 μg/ml. Reaction mixtures were prepared in 96-well polypropylene plate, Nunc. Twenty microliters of a 10× putative modulator compound (including but not limited to potential peptide ligands) solution or a water control (20 μl), 18.2 M GDP (0.11 ml, 10 μM final concentration), and a membrane preparation were mixed and placed on ice. The modulator/ligand-GDP-membrane mixtures were incubated for 20 minutes at room temperature on a shaking platform and then placed on ice. To each sample, 20 μl guanosine 5′-O-(3[ 35 S]thio) triphosphate (NEN, 600-1200 Ci/mmol; [ 35 S]-GTPγS), was added to approximately 40,000 cpm/0.2 ml, or a final concentration of 100 pM. Plates with the incubation mixtures (0.2 ml/well total) were incubated at room temperature for 45 minutes. Reaction mixture aliquots, 0.175 ml each, were then transferred into pretreated (100 μl/well wash buffer) 96-well FB MultiScreen filter plates (Millipore). Membranes were subsequently washed three times with 0.25 ml ice-cold wash buffer (10 mM HEPES, pH 7.4, 10 mM MgCl 2 ) per well each time and vacuum-filtered. After the last wash, Supermix Optiphase scintillation fluid (25 μl/well, Wallac) was added and the plates were sealed and counted in a Trilux 1450 Microbeta counter (Wallac) for one minute per well. As positive controls, membranes from CHO cells stably expressing a rat dopamine type 2 (rD 2 ) receptor, were treated with 1 mM dopamine in 0.025% ascorbic acid (100 μM dopamine final concentration). Non-specific binding was measured in the presence of 100 μM cold GTPγS and was subtracted from the total. Each treatment was typically carried out in triplicate.
Ligand-induced stimulation of [ 35 S]GTPγS binding was expressed as a multiple of the basal activity with no ligand added. Each treatment was run either in triplicate, or, on occasion in duplicate and bindings (cpm) were calculated as means +/− standard deviations. Dose-response curves for the receptor/ligand systems were analyzed using a non-linear least squares SAS model, with y=B max X/(K d +X). Other dose-response curves were analyzed using Prism (GraphPad Software, Inc. San Diego, Calif.) and the following equation y=Bottom+(Top−Bottom)/(1+10 LogEc 50 −X ).
Originally, the GTPγS assay was chosen as a functional assay because agonist-driven stimulation revealed by the GTPγS assay reflects early events in the GPCR activation cascade, regardless of further activation pathways of various down-stream signaling events. This appears especially useful for the assessment of possible activation of orphan GPCRs (i.e., GPCRs having unknown functions and unknown signaling pathways). The GTPγS assay was carried out with membranes prepared from CHO cells transiently transfected with DNA encoding C. elegans GPCRs using a 96-well MultiScreen G/FB filter plates and a MultiScreen vacuum manifold (Millipore) for filtration. Because the GTPγS assay is known to poorly recognize GPCRs coupled to the Gq class of G-proteins, a Ca +2 mobilization assay, based on a FLIPR readout, was used as well to evaluate putative Gq-couplcd orphan GPCRs transiently transfected into CHO cells. It was essential to use a 37° C. to 28° C. temnperature downshift and to incubate the cells for an additional 24-48 hours for functional receptor expression. The temperature shift was implemented 24 hours after transfection. Without the temperature downshift, expression of functional receptor could not be detected.
“The CEGPCR3 receptor (SEQ ID NO:43) was found to be activated by several peptide ligands (Table 7), as determined in the GTPγS assay. Chinese hamster ovary cells were incubated for 24 hours at 37° C. after transfection, followed by an additional 24 hours incubation at 28° C. before cell harvesting for membrane preparation. CEGPCR3 was matched with two C. elegans peptides encoded by flp15, GGPQGPLRF-NH 2 (SEQ ID NO:85; EC 50 152 nM) and GPSGPLRF-NH 2 (SEQ ID NO:89; EC 50 422 nM). A Manduca peptide, GNSFLRF-NH 2 (SEQ ID NO:88; EC 50 7900 nM), also activated the CEGPCR3 receptor, albeit with a potency about 19-52-fold lower than that determined for the two C. elegans flp15 peptides. Based on these data, we identified CEGPCR3 as the receptor for flp15 peptides.
In addition, peptide ligands were identified for the CEGPCR7 receptor (SEQ ID NO:25), as revealed in Table 7. One peptide bears the sequence GLGPRPLRF-NH 2 (SEQ ID NO:86; AF9) (EC 50 207 nM). It is worth noting that [I]Y 0 GLGPRPLRF-NH 2 (SEQ ID NO:118), representing an AF9 analog N-terminally extended with a 3-iodo-Tyr residue, was also active (EC 50 237 nM). The functional activity of [I]Y 0 AF9 indicates that this analog is an agonist and, in labeled (e.g., radioiodinated) form is useful as a probe for binding assays, including high-throughput screening (HTS) assays.”
The CEGPCR4 receptor was activated by several peptide ligands, as determined using the GTPγS assay (Table 8). To perform the assays. Cl-10 cells were incubated as described above in the context of assaying CEGPCR3. The C. elegans peptides encoded on flp18 were the most potent (EC 50 's in the low nM range). Also very potent were several A. suum peptides, with high sequence homology to the C. elegans FLP18 peptides (all having the same sequence, PGVLRF-NH 2 , at their C-termini). In addition to the FLP18 FaRPs, a battery of peptides from different invertebrate species ( Manduca, lobster, locust, Drosophila ), also activated this receptor. The latter were, however, significantly less potent (EC 50 s in the high nanomolar to micromolar range). Based on the EC 50 values, we identified CEGPCR4 as the receptor for C. elegans flp18 peptides. As shown in Table 8, [IodoY]DVPGVLRF-NH 2 , which is a N-terminally iodo-Tyr-extended flp18 analog (SEQ ID NO:119, EC 50 =8), was essentially as potent as its parent peptide, DVPGVLRF-NH 2 . This finding, as well as high affinity binding of [ 125 I-Y]DVPGVLRF-NH 2 (SEQ ID NO:80) to the CEGPCR4/CHO membranes (Kd in the subnanomolar range), showed that [ 125 I-Y]DVPGVLRF-NH 2 (SEQ ID NO:119) is an excellent probe, preferably radioiodinated, for a high throughput binding assay as a screen for modulators of the ligand-receptor interaction.
›EXAMPLE 5 · 6 of 7
Shown in Table 8 are the synthetic peptides, various analogs of DVPGVLRF-NH 2 that acted as agonists of the CEGPCR4 receptor in CHO cells. Most of the synthetic peptides were very active (EC 50 s in the nanomolar range in the GTPγS assay, and IC 50 s in the low nanomolar range in the competition binding assay). The C-terminal tetrapeptide VLRF-NH2 was considerably less potent with an EC 50 of about 4 micromolar in the GTPγS assay, and an IC50 of about 400 nM in the competition binding assay. In general, iodo-Tyr-modified peptide analogs, particularly those disclosed herein (e.g., Table 8), represent good potential probes for binding assays and may be conveniently labeled with 125 I.
Evaluation of CEGPCR5 expressed in CHO cells in the [ 35 S]GTPγS assay, as described above, yielded several active peptide ligands, as shown in Table 9. A peptide ligand designated AF9 (which also binds CEGPCR7, see above) bearing the sequence GLGPRPLRF-NH 2 (SEQ ID NO:86) was the most potent activator (EC 50 in the low nanomolar range) of CEGPCR5. An analog extending AF9 by an iodo-Tyr residue at its N-terminus, [iodoY]-GLGPRPLRF-NH 2 (SEQ ID NO:118), was about 200-fold less potent (EC 50 range of 176-273 nM), but was active enough to serve as a ligand in binding assays designed to identify modulators. Based on the GTPγS assay results, the C. elegans CEGPCR5 was identified as an AF9 or flp21 peptide receptor.
The CEGPCR16 receptor, expressed in CHO cells as described for CEGPRC3, was also used in a [ 35 S]GTPγS assay to screen for peptide ligands, which resulted in the identification of peptides encoded by the C. elegans flp3 gene. Of the peptides investigated, the Flp3 peptides were the most potent activators (Table 10), with EC 50 values of 100-330 nM. One noteworthy ligand is the Flp l8 peptide, SVPGVLRF-NH 2 (SEQ ID NO:100, EC 50 =64 nM), which is comparable to, or more potent than, the Flp3 peptides in activating CEGPCR16; most of the other Flp18 peptides were either inactive or had an EC 50 in the high nanomolar to low micromolar range.
The CEGPCR12 h receptor, and the related splice variant isoforms CEGPCR12c, CEGPCR12u, and CEGPCR12v were expressed in CHO cells as described for CEGPRC3 and used in [ 35 S]GTPγS assays to screen for peptide ligands. The results presented in Table 11, specifically presented for CEGPCR12 h, apply to all CEGPCR12 isoforms. As shown therein, the most active peptide ligand was schistoFLRFamide and its N-terminally iodinated derivative, exhibiting EC 50 values of 330 nM and 478 nM, respectively. Candidate peptide ligands were tested at 5 μM.
The CEGPCR19.1 and CEGPCR19.2 receptors, expressed in CHO cells, were subjected to screens for peptide ligands (present at 10 μM each) using a calcium mobilization assay (FLIPR). This calcium flux assay provides an alternative to the use of other screening assays disclosed herein. The assay, known in the art, was performed on transiently transfected CHO cells that had been incubated for 24 hours at 37° C. immediately following transfection, followed by a 48-hour period at 28° C. Cells were then loaded with 4 M Fluo-3 and 2.5 mM probenecid for calcium signaling (FLIPR) analysis. The results shown in Table 12 indicate that CEGPCR19.1 was activated by Flp18 peptides and some structurally similar A. suum peptides (identical C-terminal sequences of PGVLRF-NF 2 ). EC 50 values for CEGPCR19.1 binding of Flp18 peptides, and AF3 and AF4 peptides, ranged from 20-100 nM (similar EC 50 values were observed for CEGPCR19.2, described below). The results support the conclusion that CEGPCR19.1 is either an isoform of, or closely related to, the CEGPCR4 receptor because both CEGPCR19.1 and CEGPCR4 were activated by the same Flp18 peptides as well as by A. suum peptides sharing a high degree of similarity with those Flp18 peptides.
CEGPCR19.2, which is a splice variant of CEGPCR19.1, was activated (calcium assay) by the same peptides that activated CEGPCR19.1, as well as by several additional A. suum peptides not identified as activators of CEGPCR19.2 (SEQ ID NOS:91, 99, 131, 133, 150 and 151), as shown in Table 13.
G. MAP Kinase Activity Assay
Evaluation of MAP Kinase activity in cells expressing a GPCR-like receptor provide another assay to identify modulators of GPCR-like polypeptide activity. [See, e.g., Lajiness et al., Journal of Pharmacology and Experimental Therapeutics , 267(3): 1573-81 (1993); and Boulton et al., Cell , 65: 663-75 (1991).]
In one embodiment, CHO cells stably transfected with a GPCR-like receptor are seeded into 6-well plates at a density of 70,000 cells/well 48 hours prior to the assay. During this time, the cells are cultured at 37° C. in DMEM medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 10 Units/ml penicillin and 10 μg/ml streptomycin. The cells are starved for serum for 1-2 hours prior to the addition of stimulants.
For the assay, the cells are treated with medium alone or medium containing a putative agonist, or the positive control phorbol myristoyl acetate (i.e., PMA), and the cells are incubated at 37° C. for varying times. To stop the reaction, the plates are placed on ice, the medium is aspirated, and the cells are rinsed with 1 ml of ice-cold PBS containing 1 mM EDTA. Thereafter, 200 μl cell lysis buffer (12.5 mM MOPS, pH 7.3, 12.5 mM β-glycerophosphate, 7.5 mM MgCl 2 , 0.5 mM EGTA, 0.5 mM sodium vanadate, 1 mM benzamidine, 1 mM dithiothreitol, 10 μg/ml leupeptin, 10 μg/ml aprotinin, 2 μg/ml pepstatin A, and 1 μM okadaic acid) is added to the cells. The cells are scraped from the plates and homogenized by 10 passages through a 23¾ needle. The cytosol fraction is then prepared by centrifugation at 53,000 rpm for 15 minutes.
Aliquots (5-10 μl containing 1-5 μg protein) of cytosols are mixed with 1 mM MAPK Substrate Peptide (APRTPGGRR, Upstate Biotechnology, Inc. N.Y.) and 50 μM [γ- 32 P]ATP, (NEN, 3000 Ci/mmol) diluted to a final specific activity of ˜2000 cpm/pmol in a total volume of 25 μl. The samples arc incubated for 5 minutes at 30° C., and reactions are stopped by spotting 20 μl on 2 cm 2 of Whatman P81 phosphocellulose paper. The filter squares are washed in four changes of 1% H 3 PO 4 , and the squares are counted by liquid scintillation spectroscopy. Equivalent cytosolic extracts are incubated without MAPK substrate peptide, and the cpm from these samples are subtracted from the matched samples with the substrate peptide. The cytosolic extract from each well is used as a separate point. Protein concentrations are determined by a dye binding protein assay (Bio-Rad). Agonist activation of the receptor is expected to result in up to a five-fold increase in MAPK enzyme activity. This increase is blocked by antagonists.
›EXAMPLE 5 · 7 of 7
H. [ 3 H]Arachidonic Acid Release
The activation of GPCRs has been observed to potentiate arachidonic acid release in cells, providing yet another useful assay for modulators of GPCR-like receptor activity. [See, e.g., Kanterman et al., Molecular Pharmacology , 39: 364-9 (1991).] For example, CHO cells that are stably transfected with a GPCR-like receptor expression construct are plated in 24-well plates at a density of 15,000 cells/well and grown in DMEM medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 10 Units/ml penicillin and 10 μg/ml streptomycin for 48 hours at 37° C. before use. Cells of each well are labeled by incubation with [ 3 H]-arachidonic acid (Amersham Corp., 210 Ci/mmol) at 0.5 μCi/ml in 1 ml DMEM supplemented with 10 mM HEPES, pH 7.5, and 0.5% fatty-acid-free bovine serum albumin for 2 hours at 37° C. The cells are then washed twice with 1 ml of the same buffer.
Candidate modulator compounds are added in 1 ml of the same buffer, either alone or containing 10 μM ATP as a positive control and the cells are incubated at 37° C. for 30 minutes. Buffer alone and mock-transfected cells are used as controls. Samples (0.5 ml) from each well are counted by liquid scintillation spectroscopy. Agonists which activate the receptor will lead to potentiation of the ATP-stimiulated release of [ 3 H-]-arachidonic acid. This potentiation is blocked by antagonists.
I. Extracellular Acidification Rate
In yet another assay, the effects of putative modulators of GPCR-like receptor activity are assayed by monitoring extracellular changes in pH induced by the putative modulators. [See, e.g., Dunlop et al., Journal of Pharmacological and Toxicological Methods , 40(I): 47-55 (1998).]
CHO cells transfected with a GPCR-like receptor expression construct are seeded into 12 mm capsule cups (Molecular Devices Corp.) at 4×10 5 cells/cup in DMEM supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 10 Units/ml penicillin, and 10 μg/ml streptomycin. The cells are incubated in this medium at 37° C. in 5% CO 2 for 24 hours.
Extracellular acidification rates are measured using a Cytosensor microphysiometer (Molecular Devices Corp.). The capsule cups are loaded into the sensor chambers of the microphysiometer and the chambers are perfused with running buffer (bicarbonate free DMEM supplemented with 4 mM L-glutamine, 10 Units/ml penicillin, 10 μg/ml streptomycin, 26 mM NaCl) at a flow rate of 100 μl/minute. Agonists or other agents are diluted into the running buffer and perfused through a second fluid path. During each 60-second pump cycle, the pump is run for 38 seconds and is off for the remaining 22 seconds. The pH of the running buffer in the sensor chamber is recorded during the cycle from 43-58 seconds, and the pump is re-started at 60 seconds to start the next cycle. The rate of acidification of the running buffer during the recording time is calculated by the Cytosoft program. Changes in the rates of acidification are calculated by subtracting the baseline value (the average of four rate measurements immediately before addition of modulator candidates) from the highest rate measurement obtained after addition of a modulator candidate. The selected instrument detects 61 mV/pH unit. Modulators that act as agonists at the receptor result in an increase in the rate of extracellular acidification as compared to the rate in the absence of agonist. This response is blocked by modulators which act as antagonists at the receptor.
›EXAMPLE 6 · 1 of 5
To avoid impediments imposed by potential artifacts of PCR-based cloning, alternative approaches to the cloning of neuropeptide (e.g., FaRP) receptors was developed. By increasing the versatility of the cloning approach, FaRP receptors having desired relationships (e.g., particular levels of primary, or structural, sequence similarity or identity, common activities, or related sources) could be targeted. For example, targeted neuropeptide (e.g., FaRP) receptors might exhibit 70%, 75%, 80%, 85%, 90%, 95%, 99%, 99.9%, or 99.99% similarity; also, the targeted receptors might be selected from invertebrates, from insects, from flat- and round-worms, from particular round-worm genera, or from particular round-worm species, sub-species or variants.
The identification of several authentic neuropeptide (FaRP) receptors also makes possible additional bioinformatics strategies. Comparison of confirmed receptors with the C. elegans database should aid in the identification of additional neuropeptide receptors, including FaRP receptors. Likewise, the identification of the C. elegans neuropeptide receptors make possible a similar approach using other databases, including Drosophila and human databases.
The sequences of a total of 20 cloned receptors are listed, including all of the receptors exhibiting high levels of similarity to the confirmed neuroepeptide (FaRP) receptors. Finally, the results of bioinformatic analyses using the current databases and biochemical information are presented.
Bioinformatic Selection of Receptors
Later releases of the Wormpep database were searched in a manner similar to that described above for the initial identification of candidate receptors. Wormpep releases through Wormpep 23 were downloaded from the Sanger Centre web site. The database was searched using BLAST 2.0, Altschul et al., Nucl. Acids Res. 25(17):3389-402 (1997), with known C. elegans neuropeptide (e.g., FaRP) receptors (CEGPCR3, CEGPCR4, CEGPCR5, CEGPCR7, CEGPCR12c, CEGPCR12h, CEGPCR12u, CEGPCR12v, and CEGPCR16). A composite BLAST score was assembled from these searches by calculating the product of the “e” values from all of the searches for a given receptor. Release 118 of GenPept was used for general searches against public sequences and as a source of sequences used for the phylogenetic analyses. Alignments and phylogenetic relationships were prepared using the AlignX (ClustalW, Thompson et al., Nucl. Acids Res. 22(22):4673-80 (1994)) alignment program (Vector NTI suite, Informax, N. Bethesda, Md.).
Cloning of Receptors
Receptors were cloned using PCR techniques. Initially, the predicted sequence was used to design PCR primers. The 5′ PCR primer incorporated the C. elegans initiator ATG within an optimized 5′ untranslated region (Kozak, M., Nucl. Acids Res. 15(20):8125-48 (1987). The 3′ PCR primer usually contained a restriction enzyme site not contained within the receptor sequence to simplify later production of cRNA for injection studies.
If the initial approach using primers designed to the predicted sequence did not readily produce a PCR product, or if the predicted sequence was obviously in error (as judged by alignment with other GPCRs), then either an anchored PCR approach (used successfully for CEGPCR4 and CEGPCR22), or a rational primer design approach (see below) was followed to obtain an intact clone.
Typically, a PCR product was obtained using either XL rTth polymerase (Perkin-Elmer Applied Biosystems, Foster City, Calif.) or the Expand HF mix (Boehringer Mannheim, Indianapolis, Ind.). PCR primer sequences are presented in the sequence listing, with “f” and “r” suffixes added to a given GPCR clone designation to unambiguously refer to forward and reverse primer pairs. The DNA template for the reaction was either a first strand cDNA synthesis reaction (using random hexamers to prime synthesis from a preparation of total C. elegans RNA) or a C. elegans cDNA library in pBluescript, constructed by performing an in vivo excision reaction on a cDNA commercial library constructed in a Uni-ZAP XR lambda vector (Stratagene, La Jolla, Calif., catalog number 937006). PCR products were typically gel-purified using QIAquick columns (Qiagen, Valencia, Calif.) and either cloned directly into pCR3.1 (Invitrogen, Carlsbad, Calif.), or indirectly through an initial cloning step into pCR2.1. Plasmids containing the appropriate inserts were prepared by a mini-prep DNA isolation procedure (Qiagen) and were sequenced using BigDye dye terminator chemistry (PE Applied Biosystems, Foster City, Calif.) on an ABI 377 automated DNA sequencer, as described above. Complete sequences were assembled using Sequencher software (Version 3.0, GeneCodes, Ann Arbor, Mich.) and compared against the expected sequence predicted from the Wormpep database.
Cloning CEGPCR2
CEGPCR2 (SEQ ID NO:176) was cloned using PCR. The template for the reaction was a plasmid library of C. elegans cDNA isolated from an in vivo excision of a commercially available library (constructed in a Uni-ZAP XR vector; Stratagene, La Jolla, Calif.) according to the manufacturer's instructions. The PCR reaction was performed using the Expand HF kit (Boehringer Mannheim, Indianapolis, Ind.) and contained: 0.5 μM each primer, 150 nM dNTPs, 0.5 mM MgCl 2 , and 1.4 units Expand High Fidelity enzyme mix. The PCR reaction was carried out in a Stratagene Robocycler, using the following conditions: 30 cycles at 95° C. for 30 seconds, 50° C. for 30 seconds, 72° C. for 2 minutes, followed by 72° C. for 5 minutes. A TA cloning kit (Invitrogen, Carlsbad, Calif.) was used to clone the gel-purified PCR product into pCR2.1. The ligation mixture was transformed into E. coli DH5α, and transformed cells were identified using a colony PCR method to screen transformants for the presence of the appropriate insert. Purified plasmid DNA was obtained using a mini-prep procedure (Qiagen Valencia, Calif.). Plasmids were sequenced using dye terminator chemistry (PE Applied Biosystems, Foster City, Calif.) and an automated sequencer (ABI 377). A clone with an intact open reading frame was obtained and subcloned into pcDNA3.1(+) using EcoRI. XhoI was used to determine the orientation of insert fragments.
›EXAMPLE 6 · 2 of 5
Cloning CEGPCR21
CEGPCR21 was cloned as described above in the context of CEGPCR2, using the same template, PCR conditions, and cloning strategy. HindIII was used to determine the orientation of inserts.
Cloning CEGPCR23
The C. elegans receptor CEGPCR23 was cloned using the C. elegans cDNA plasmid library as template. The PCR reaction was set up using 0.5 μM each primer, 200 μM dNTPs, 1.15 mM Mg(OAc) 2 , and 4 units rTth DNA polymerase, XL. The PCR reactions were performed in a Perkin-Elmer 9600 Thermal Cycler using the following program: 25 cycles at 94° C. for 30 seconds, 50° C. for 10 minutes, and 72° C. for 10 minutes. The resulting PCR product was gel-purified, cloned into pCR2.1, and propagated in E. coli DH5α. Plasmid DNA was isolated and verified by sequencing. An intact clone was selected and subcloned into pcDNA3.1(+) (Invitrogen) using BstXI. Subclones with inserts in the correct orientation were identified using BamHI/BglII.
Cloning CEGPCR22
An anchored PCR approach was used to obtain the correct ends for CEGPCR22. Internal primers were designed to the exon containing 7TM. PCR experiments using these internal primers with the initial flanking primers confirmed that the 3′ PCR primer was nonfunctional. Using the internal primer (SCN162—5′-ACG TTT AAG AGC TCT CAA ATC CCA T-3′; SEQ ID NO:171) and T7/M13 reverse vector primers, along with the C. elegans cDNA plasmid library, a PCR product was successfully obtained that corresponded to the correct 3′ end. Using that information, a new 3′ PCR primer was designed and used in conjunction with the initial 5′ PCR primer. CEGPCR22 was then cloned in a similar manner to CEGPCR23. Template and PCR conditions were identical to CEGPCR23, except that the PCR program included 30 cycles at 94° C. for 15 seconds, 55° C. for 5 minutes, and 72° C. for 10 minutes. The resulting PCR product was directly cloned into pCR3.1.
Cloning CEGPCR4
An anchored PCR approach was used to obtain the correct ends for CEGPCR4. Internal primers were synthesized. PCR experiments using these internal primers with the initial flanking primers confirmed that the 3′ end of the cDNA was incorrectly predicted. Using an internal primer (SCN160—5′-CCAGAGCTCATCAAAACTCAAGAAT-3′; SEQ ID NO:172) and T7/M13 reverse vector primers, along with the C. elegans cDNA plasmid library, a PCR product was successfully obtained that corresponded to the correct 3′ end, extending beyond the stop codon. Using that information, a new 3′ PCR primer (SCN189, 5′-TTACAATTTAAAACTAGGTGCTTCT-3′; SEQ ID NO:174) was designed and used in conjunction with the initial 5′ PCR primer. CEGPCR4 was then cloned using identical conditions to CEGPCR23, except that the template was derived from a first strand cDNA synthesis of C. elegans mRNA and the PCR program included 25 cycles at 94° C. for 15 seconds, 60° C. for 5 minutes, and 72° C. for 10 minutes. The PCR product was cloned initially in pCR2.1, and subcloned into pcDNA3.1(+) using EcoRI, and clones with the correct orientation were found using digestion with SacI.
Cloning CEGPCR5
CEGPCR5 was cloned using the same PCR reaction components as CEGPCR22 and first strand cDNA as template. PCR conditions were: 35 cycles at 94° C. for 30 seconds, 60° C. for 5 minutes, and 72° C. for 10 minutes. The reactions were carried out in a Perkin-Elmer 9600 thermal cycler. The resulting PCR product was cloned directly into pCR3.1 and transformed into E. coli DH5α. Clones were identified by colony PCR. Plasmids were isolated, purified, and sequenced as described above.
Cloning CEGPCR19.1
The short splice variant of clone CEGPCR19 was cloned using conditions identical to those described for CEGPCR5, with the exception of a 50° C. hybridization step for the PCR reaction.
Cloning CEGPCR19.2
The long splice variant of clone CEGPCR19 was cloned using conditions identical to those described for CEGPCR5, with the exception of a 55° C. hybridization step for the PCR reaction.
Cloning CEGPCR1a and CEGPCR1f
The cloning of the two splice isoforms of CEGPCR1 were described in the examples above. Those initial clones were based on the Wormpep 13-predicted sequence. Although the receptors were successfully amplified and cloned, alignment of the sequences suggested that other variant clones might also exist that contained an initiator ATG in a more conventional location than within the first transmembrane region. Subsequent releases of the database included a longer AC7.1 sequence corresponding to CEGPCR1 (but still not the splice variants). Using that new sequence from the Wormpep database, a new primer (SCN-199, 5′-GCCGCCATGAACTTTTCGGCCACCGATTCGA-3′; SEQ ID NO:175) was designed and additional clones were isolated (using the identical procedures and 3′ PCR primer described in the preceding examples for CEGPCR1).
Cloning CEGPCR12c, CEGPCR12h, and CEGPCR12u
Similar to CEGPCR1, alignment of the initially amplified clones for CEGPCR12 suggested that additional variants might exist. In the case of CEGPCR12, it was anticipated that variants encoding polypeptides having additional N-terminal sequence would be identified. Using the rational primer design approach, PCR primers were designed for several potential exons. Primer DEL-1850 (5′-GCCGCCATGTCGAATGATCTCGTGCCTTCAG-3′, SEQ ID NO:173) and the original 3′ PCR primer (SEQ ID NO:70; using PCR conditions as described for the initial amplification of CEGPCR12) were then used to amplify the longer version. The PCR product was cloned directly into PCR3.1 as described above. Intact versions for CEGPCR12c, CEGPCR12h, and CEGPCR12u were obtained.
The C. elegans Wormpep 23 database was searched using BLAST 2.0 (Altschul, et al., Nucl. Acids Res. 25(17):3389-402 (1997)), with the six currently known C. elegans neuropeptide (e.g., FaRP) receptors (CEGPCR3, CEGPCR4, CEGPCR5, CEGPCR7, CEGPCR12c, CEGPCR12h, CEGPCR12u, CEGPCR12v, and CEGPCR16). Using the e-values from these searches, a composite BLAST score was calculated. Using this approach, a group of 11 receptors was discernable from the rest of the C. elegans GPCRs. These receptors include the above-six confirmed neuropeptide (FapP) receptors, and CEGPCR13, CEGPCR17, CEGPCR14, CEGPCR19 (both CEGPCR19.1 and CEGPCR19.2), and CEGPCR11. An additional group of receptors fell into a second tier. This group included CEGPCR20, CEGPCR1 (Both CEGPCR1a and CEGPCR1f), CEGPCR15, CEGPCR18a, CEGPCR25 (SEQ ID NO:178), and CEGPCR3. BLAST comparisons against the GenBank database showed that these receptors were also likely peptide receptors, although they have not yet been matched to ligands.
›EXAMPLE 6 · 3 of 5
An alignment of the receptors using CLUSTALW shows that the receptors exhibit the classic characteristics of the seven transmembrane family of receptors, including higher degrees of similarity through the transmembrane regions and very few absolutely conserved residues (for a review see Probst et al., Cell Biol 11(1): 1-20 (1992)). The only unusual feature for the receptors is a few atypical amino acids in the conserved ‘DRY’ motif immediately following 3TM, e.g., ‘HEF’ in the CEGPCR1a and If sequences and ‘DKF’ for the CEGPCR18a and 18b sequences.
The C. elegans receptors can also be classified using a phylogenetic approach. FIG. 1 shows an alignment of the C. elegans GPCRs described above resulting from a typical phylogenetic analysis. This analysis showed that ten of the eleven members of the above-mentioned family of receptors, including the neuropeptide receptors identified to date, fall into two distinct groups. The only other receptors that fall into these groups are the ‘GRL105’ receptor from Lymnaea stagnalis (an invertebrate NPY receptor) and the DmGPCR1 and DmGPCR2 receptors from Drosophila. We have recently confirmed that the DmGPCR1 receptor has a neuropeptide (i.e., FaRP) ligand.
This analysis also showed that the second tier of C. elegans receptors from the BLAST analysis tends to fall into classes with other known neuropeptide receptors. CEGPCR1 and CEGPCR24 fall into an interesting group with the L. stagnalis lymnokinin receptor (Cox et al., J. Neurosci. 17(4):1197-205 (1997)), the L. stagnalis cardioexcitatory peptide (an RFamide ligand) receptor (Tensen et al., J. Neurosci. 18(23):9812-21 (1998)), the Stomoxys calcitrans (stable fly) ‘tachykinin-like’ receptor (Guerrero, F. D., Annals N.Y. Acad. Sci. 814:310-1 (1997)), a receptor for a leucokinin-like peptide from cattle tick (Holmes, et al., Insect Mol Biol (2000; in press)) and the DmGPCR6 receptor [Li, et al., J. Biol. Chem. 267(1):9-12 (1992). We have recently identified this last receptor as having a neuropeptide (i.e., FaRP) ligand. This group is separate, and contains only invertebrate receptors, but is most closely related to the vertebrate family of neurokinin (NK-1,2,3) receptors. The C. elegans receptors CEGPCR6 and CEGPCR15 are in a group distinct from, but related to, the vertebrate neurokinin receptors. A cluster containing a variety of cholecystokinin receptors contains the C. elegans receptor Y39A3B.5. The receptor CEGPCR20 falls into a group containing the Drosophila receptor for an allatostatin-related peptide (Birgul et al., EMBO Journal 18(21):5892-900 (1999)) and DmGPCR4, both of which recognize allatostatin peptide ligands. CEGPCR18 variants are included in a group containing several vertebrate orphan receptors, most closely related to the clusters containing bombesin receptors, galanin receptors, gastrin releasing peptide receptors and the above mentioned allatostatin receptors. CEGPCR8 (and CEGPCR21) cannot obviously be linked to any particular group of receptors, being most closely related to a broad cluster containing a variety of receptors, including some falling outside the domain of those receptors recognizing peptide ligands (e.g., serotonin receptors, opsins). The receptors CEGPCR9, CEGPCR23 and CEGPCR22 are all listed in a small group apart from a group containing a variety of receptors with peptide ligands, including somatostatin, galanin, and bombesin receptors. BLAST results are slightly in contrast with these results for CEGPCR9, listing the highest scoring hits for this receptor as NPY-type receptors. BLAST results indicate the five highest scoring matches to CEGPCR22 and CEGPCR23 are all other C. elegans receptors (Y116A8B.5, C43C3.2, T02D1.6), but immediately following those are a series of vertebrate somatostatin and kappa-opioid receptors.
The receptors were cloned using several strategies, all based on the use of PCR. The first strategy involved the simple design of PCR primers using the predicted cDNA sequence of the GPCR. In cases where a PCR product was not obtained using primers designed on the basis of the predicted sequence, an alternative approach was used to design new PCR primers. The failure in the initial PCR experiments was typically due to the inability of the gene prediction software to accurately predict the ends of the GPCR sequences. Alignment of the predicted amino acid sequences for the GPCRs often revealed the errors, e.g., one of the ends of the encoded protein sequences would be much longer that any of the other aligned sequences. In some cases the alignment showed that a product that did amplify by PCR had been inaccurately predicted, e.g., showing an initiator ATG that was within the predicted 1TM region. The incorrectly predicted end could usually be pinpointed by the use of internal PCR primers, located within a conserved and unspliced region, in conjunction with the original flanking PCR primers. Using this information, new PCR primers were designed. Taking into account the possible open reading frames, potential mRNA processing sites, and proximity to the presumed correct sequence, a series of alternative PCR primers were synthesized and tested. This rational primer design approach, although labor-intensive, was extremely successful, resulting in a high rate of successful PCR amplification.
Based on comparisons of cloned cDNAs to predicted sequences, 11 of 22 GPCR-like receptor sequences selected from Wormpep were incorrectly predicted. As would be expected, most of these errors have occurred in the prediction of the N- and C-termini of the encoded proteins. Although an anchored PCR approach has been successful in obtaining the ends of an incorrectly predicted GPCR-like receptor in a couple of cases, the rational primer design approach, although laborious, has had a much higher success rate.
Phylogenetic analyses can be used to speculate on possible functions for the receptors. It should be noted that there are numerous methods and variables that can be used to build phylogenetic relationships. Although the results described herein are based on one method Thompson et al., Nucl. Acids Res. 22(22):4673-80 (1994)), several methods and permutations of phylogenetic analyses were investigated, with all methods giving very similar results, especially with regards to the overall relationships observed for the confirmed neuropeptide (e.g., FaRP) receptors.
›EXAMPLE 6 · 4 of 5
At this point in time, a total of six C. elegans receptors and two Drosophila receptors have been matched to neuropeptide (FaRP) ligands. Phylogenetic analyses of the C. elegans and Drosophila receptors, along with all of the peptidergic receptors in the public databases, show that all of the C. elegans FaRP receptors may be grouped together. This group contains an additional five C. elegans receptors (CEGPCR13, CEGPCR14, CEGPCR17, and CEGPCR19.1 and CEGPCR19.2) that are expected to have FaRP ligands. The only other receptors within this group are two Drosophila receptors, one of which has been confirmed to have a FaRP ligand (DmGPCR1), and a single receptor from L. stagnalis (GRL105) (Tensen et al., J. Neurosci. 18(23):9812-21 (1998)). The GRL105 receptor has been linked to a L. stagnalis neuropeptide Y homologue (an RFamide). The phylogenetic analysis would predict that it may also have a FaRP ligand. The grouping of these invertebrate receptors in a category that contains no known vertebrate receptors is consistent with FaRP receptors providing a relatively invertebrate-specific target.
The other C. elegans receptor that received a high score in the BLAST analyses was CEGPCR11. This receptor falls into a group that is highly related to the group containing confirmed FaRP receptors, described above, and includes the dog and human orexin receptors. Slight alterations in the conditions used in the analysis would result in CEGPCR11 being grouped with the above receptors.
Another group of receptors that is of interest is derived from C. elegans neuropeptide receptors. This group includes the C. elegans receptors CEGPCR24a, CEGPCR24b, and CEGPCR1. The recent identification of a FaRP ligand for the Drosophila receptor DmGPCR6, also located within this group of receptors, increases the interest in these C. elegans receptors.
Two clones (CEGPCR6 and CEGPCR20) included in the table have yet to be amplified. These clones do not PCR-amplify using primers designed from the predicted sequences in Wormpep. It is very likely that one, or both, of the termini of these clones is incorrectly predicted. At the time of the receptor cloning, these receptors were not classified in the group of receptors that contained the first verified FapP receptors. CEGPCR20 groups with a family of receptors that include the vertebrate galanin receptors, the Drosophila allatostatin-like peptide receptor (Birgul et al., EMBO Journal 18(21):5892-900 (1999)), the Drosophila ‘galanin’ receptor Lenz et al., BBRC 269(1):91-96 (2000) (ligand assigned by homology, the Drosophila receptor DmGPCR4, which has recently been confirmed to be an additional Drosophila allatostatin receptor, and a series of four orphan L. stagnalis receptors. Based on this information, is expected that CEGPCR20 recognizes an allatostatin-like peptide ligand. CEGPCR6 falls into a grouping with CEGPCR1a, CEGPCR1f, CEGPCR15, and two Drosophila receptors. With the recent discovery that one of the Drosophila receptors recognizes a FaRP ligand, there is a renewed interest in this group of receptors.
Based on analyses to date, the invertebrate GPCR-like receptors do not appear to have highly similar sequences in other organisms, such as vertebrates and plants. However, receptors bearing lower levels of similarity, e.g., 55%, 60%, 65%, 70%, 75%, 80%, 85%, and preferably 90%, 95%, 98%, 99% and more preferably 99.5% similarity to a GPCR-like receptor amino acid sequence disclosed herein are also contemplated by the invention. Analogously, the invention comprehends receptor-encoding polynucleotides exhibiting 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, and preferably 95%, 98%, 99and more preferably 99.5% similarity to a polynucleotide disclosed herein. Similarities can be determined using any of a variety of algorithms known in the art, with the BLAST algorithm implemented at the GenBank website (http://www.ncbi.nlm.nih.gov/BLAST/), using default parameters, being preferred. These receptor sequences are anticipated to be useful in a variety of contexts. For example, it is expected that vertebrate, and more particularly mammalian, receptor sequences showing some similarity to the invertebrate GPCR-like receptors will be useful in diagnosing, and treating, a variety of neurological ailments or conditions.
The following, clones (deposit numbers indicated parenthetically) were deposited with the ARS Patent Culture Collection, 1815 North University Street, Peoria, Ill. 61604 on Nov. 3, 2000: CEGPCR1a (UC20129 NRRL B-30361), CEGPCR1f (UC20130 NRRL B-30362), CEGPCR3 (UC20132 NRRL B-30364), CEGPCR4 (UC20133 NRRL B-30365), CEGPCR5 (UC20134 NRRL B-30366), CEGPCR7 (UC20135 NRRL B-30367), CEGPCR8 (UC20136 NRRL B-30368), CEGPCR9 (UC20137 NRRL B-30369), CEGPCR12c (UC20138 NRRL B-30370), CEGPCR12h (UC20139 NRRLB-30371), CEGPCR12u (UC20140 NRRL B-30372). On Nov. 10, 2000, the following clones were deposited with the ARS Patent Culture Collection: CEGPCR11 (UC20141 B-30381), CEGPCR13 (UC20142 B-30382), CEGPCR14 (UC20143 B-30383), CEGPCR15 (UC20144 B-30384), CEGPCR16 (UC20145 B-30385), CEGPCR17 (UC20146 B-30386), CEGPCR18a (UC20147 B-30387), CEGPCR19.2 (UC20148 B-30388), CEGPCR19.1 (UC20149 B-30389), CEGPCR21 (UC20150 B-30390), CEGPCR22 (UC20151 B-30391), and CEGPCR23 (UC20152 B-30392).
Numerous modifications and variations in the invention as set forth in the above illustrative examples are expected to occur to those skilled in the art. Consequently only such limitations as appear in the appended claims should be placed on the invention.
1
atg aac ttt tcg gcc acc gat tcg ata ttg gca tca acg ata aca acg 48
Met Asn Phe Ser Ala Thr Asp Ser Ile Leu Ala Ser Thr Ile Thr Thr
1 5 10 15
gtg att ggt gga gct gga gtt ttg gca gaa gca ggc gaa gct gaa cta 96
Val Ile Gly Gly Ala Gly Val Leu Ala Glu Ala Gly Glu Ala Glu Leu
20 25 30
tct ggt gat gat gat ttt tat gag ctg act cct gta gaa ttg ata ata 144
Ser Gly Asp Asp Asp Phe Tyr Glu Leu Thr Pro Val Glu Leu Ile Ile
35 40 45
tgg tgc atg ctg tat gca att ata gcc ttc atg gca gtt gtt gga aat 192
›EXAMPLE 6 · 5 of 5
Trp Cys Met Leu Tyr Ala Ile Ile Ala Phe Met Ala Val Val Gly Asn
50 55 60
ctt ctg gtt ctc tac ata aca ctg ttc aga tta aga gtc cgt tcc atc 240
Leu Leu Val Leu Tyr Ile Thr Leu Phe Arg Leu Arg Val Arg Ser Ile
65 70 75 80
aca acc tac ttc att ctg aac ctc gga ttt gct gac ctc ttc act ggt 288
Thr Thr Tyr Phe Ile Leu Asn Leu Gly Phe Ala Asp Leu Phe Thr Gly
85 90 95
att ttt gcg att ccc ttc aag ttt cag gct gct ctt ttt caa gaa tgg 336
Ile Phe Ala Ile Pro Phe Lys Phe Gln Ala Ala Leu Phe Gln Glu Trp
100 105 110
ttc ctg ccg cga tca ctc tgc cgg ata gtt cca tac gtg gaa aca gtt 384
Phe Leu Pro Arg Ser Leu Cys Arg Ile Val Pro Tyr Val Glu Thr Val
115 120 125
gct ctg aca gtt tca gtc ttc aca ctt gtg acg tca gca gtt cat gaa 432
Ala Leu Thr Val Ser Val Phe Thr Leu Val Thr Ser Ala Val His Glu
130 135 140
ttc cgt aca atg ttc ttc tcg aaa tgc tca caa atg agc cca aga tct 480
Phe Arg Thr Met Phe Phe Ser Lys Cys Ser Gln Met Ser Pro Arg Ser
145 150 155 160
gca aaa cga tgt gta ctt ttg ata tgg ata atg gcg gtt ctt gtg tct 528
Ala Lys Arg Cys Val Leu Leu Ile Trp Ile Met Ala Val Leu Val Ser
165 170 175
cta cca cat gga ttg ttc cat aat aca tac gaa ttt cca gat gac aat 576
Leu Pro His Gly Leu Phe His Asn Thr Tyr Glu Phe Pro Asp Asp Asn
180 185 190
aat act tca att gta cag tgt ctc cca gta tat cct gat gct ggt tgg 624
Asn Thr Ser Ile Val Gln Cys Leu Pro Val Tyr Pro Asp Ala Gly Trp
195 200 205
tgg aaa aca tac aat gtc tac ctt gtc ata atc caa tat ttt gtt cca 672
Trp Lys Thr Tyr Asn Val Tyr Leu Val Ile Ile Gln Tyr Phe Val Pro
210 215 220
atg att att ctt gac act gcg tac aca atg att gct gtt aaa ata tgg 720
Met Ile Ile Leu Asp Thr Ala Tyr Thr Met Ile Ala Val Lys Ile Trp
225 230 235 240
tca ttg agt cag tca aga gtt gaa ctt gat gaa aca aaa atg gca acc 768
Ser Leu Ser Gln Ser Arg Val Glu Leu Asp Glu Thr Lys Met Ala Thr
245 250 255
cag aag ctt atg cgt act ctc atc att gtc gtt gcc tgt ttc tca ttg 816
Gln Lys Leu Met Arg Thr Leu Ile Ile Val Val Ala Cys Phe Ser Leu
260 265 270
tgt tgg ttt cca ttg gag acg tat cta ctt ttg aat gaa ttg aaa ccg 864
Cys Trp Phe Pro Leu Glu Thr Tyr Leu Leu Leu Asn Glu Leu Lys Pro
275 280 285
gaa att aat gga tgg aaa tac atc aat ttg gtg ttc ttc ttt tca cat 912
Glu Ile Asn Gly Trp Lys Tyr Ile Asn Leu Val Phe Phe Phe Ser His
290 295 300
tgg ctg gcg atg agc aat tct tgt ctt aat cca att att tat gga ctt 960
Trp Leu Ala Met Ser Asn Ser Cys Leu Asn Pro Ile Ile Tyr Gly Leu
305 310 315 320
tac aat aca aaa tac aac gag gaa tat cgt cgt ttg ttt cgc caa att 1008
Tyr Asn Thr Lys Tyr Asn Glu Glu Tyr Arg Arg Leu Phe Arg Gln Ile
325 330 335
gga tgc att tgg caa cgg cag aaa agt ttg gac gat tcg atg aaa ccg 1056
Gly Cys Ile Trp Gln Arg Gln Lys Ser Leu Asp Asp Ser Met Lys Pro
340 345 350
gag cgt cgt tgg aat tct tca aat gat tgt caa gat caa cag gaa att 1104
Glu Arg Arg Trp Asn Ser Ser Asn Asp Cys Gln Asp Gln Gln Glu Ile
355 360 365
gat caa att gtt gat att cca cca gtt att tct aca aat aat ctt tct 1152
Asp Gln Ile Val Asp Ile Pro Pro Val Ile Ser Thr Asn Asn Leu Ser
370 375 380
ccc tg 1157
Pro
385
2
385
›PRT
Homo sapiens
2
Met Asn Phe Ser Ala Thr Asp Ser Ile Leu Ala Ser Thr Ile Thr Thr
1 5 10 15
Val Ile Gly Gly Ala Gly Val Leu Ala Glu Ala Gly Glu Ala Glu Leu
20 25 30
Ser Gly Asp Asp Asp Phe Tyr Glu Leu Thr Pro Val Glu Leu Ile Ile
35 40 45
Trp Cys Met Leu Tyr Ala Ile Ile Ala Phe Met Ala Val Val Gly Asn
50 55 60
Leu Leu Val Leu Tyr Ile Thr Leu Phe Arg Leu Arg Val Arg Ser Ile
65 70 75 80
Thr Thr Tyr Phe Ile Leu Asn Leu Gly Phe Ala Asp Leu Phe Thr Gly
85 90 95
Ile Phe Ala Ile Pro Phe Lys Phe Gln Ala Ala Leu Phe Gln Glu Trp
100 105 110
Phe Leu Pro Arg Ser Leu Cys Arg Ile Val Pro Tyr Val Glu Thr Val
115 120 125
Ala Leu Thr Val Ser Val Phe Thr Leu Val Thr Ser Ala Val His Glu
130 135 140
Phe Arg Thr Met Phe Phe Ser Lys Cys Ser Gln Met Ser Pro Arg Ser
145 150 155 160
Ala Lys Arg Cys Val Leu Leu Ile Trp Ile Met Ala Val Leu Val Ser
165 170 175
Leu Pro His Gly Leu Phe His Asn Thr Tyr Glu Phe Pro Asp Asp Asn
180 185 190
Asn Thr Ser Ile Val Gln Cys Leu Pro Val Tyr Pro Asp Ala Gly Trp
195 200 205
Trp Lys Thr Tyr Asn Val Tyr Leu Val Ile Ile Gln Tyr Phe Val Pro
210 215 220
Met Ile Ile Leu Asp Thr Ala Tyr Thr Met Ile Ala Val Lys Ile Trp
225 230 235 240
Ser Leu Ser Gln Ser Arg Val Glu Leu Asp Glu Thr Lys Met Ala Thr
245 250 255
Gln Lys Leu Met Arg Thr Leu Ile Ile Val Val Ala Cys Phe Ser Leu
260 265 270
Cys Trp Phe Pro Leu Glu Thr Tyr Leu Leu Leu Asn Glu Leu Lys Pro
275 280 285
Glu Ile Asn Gly Trp Lys Tyr Ile Asn Leu Val Phe Phe Phe Ser His
290 295 300
Trp Leu Ala Met Ser Asn Ser Cys Leu Asn Pro Ile Ile Tyr Gly Leu
305 310 315 320
Tyr Asn Thr Lys Tyr Asn Glu Glu Tyr Arg Arg Leu Phe Arg Gln Ile
325 330 335
Gly Cys Ile Trp Gln Arg Gln Lys Ser Leu Asp Asp Ser Met Lys Pro
340 345 350
Glu Arg Arg Trp Asn Ser Ser Asn Asp Cys Gln Asp Gln Gln Glu Ile
355 360 365
Asp Gln Ile Val Asp Ile Pro Pro Val Ile Ser Thr Asn Asn Leu Ser
370 375 380
Pro
385
3
atg aac ttt tcg gcc acc gat tcg ata ttg gca tca acg ata aca acg 48
Met Asn Phe Ser Ala Thr Asp Ser Ile Leu Ala Ser Thr Ile Thr Thr
1 5 10 15
gtg att ggt gga gct gga gtt ttg gca gaa gca ggc gaa gct gaa cta 96
Val Ile Gly Gly Ala Gly Val Leu Ala Glu Ala Gly Glu Ala Glu Leu
20 25 30
tct ggt gat gat gat ttt tat gag ctg act cct gta gaa ttg ata ata 144
Ser Gly Asp Asp Asp Phe Tyr Glu Leu Thr Pro Val Glu Leu Ile Ile
35 40 45
tgg tgc atg ctg tat gca att ata gcc ttc atg gca gtt gtt gga aat 192
Trp Cys Met Leu Tyr Ala Ile Ile Ala Phe Met Ala Val Val Gly Asn
50 55 60
ctt ctg gtt ctc tac ata aca ctg ttc aga tta aga gtc cgt tcc atc 240
Leu Leu Val Leu Tyr Ile Thr Leu Phe Arg Leu Arg Val Arg Ser Ile
65 70 75 80
aca acc tac ttc att ctg aac ctc gga ttt gct gac ctc ttc act ggt 288
Thr Thr Tyr Phe Ile Leu Asn Leu Gly Phe Ala Asp Leu Phe Thr Gly
85 90 95
att ttt gcg att ccc ttc aag ttt cag gct gct ctt ttt caa gaa tgg 336
Ile Phe Ala Ile Pro Phe Lys Phe Gln Ala Ala Leu Phe Gln Glu Trp
100 105 110
ttc ctg ccg cga tca ctc tgc cgg ata gtt cca tac gtg gaa aca gtt 384
Phe Leu Pro Arg Ser Leu Cys Arg Ile Val Pro Tyr Val Glu Thr Val
115 120 125
gct ctg aca gtt tca gtc ttc aca ctt gtg acg tca gca gtt cat gaa 432
Ala Leu Thr Val Ser Val Phe Thr Leu Val Thr Ser Ala Val His Glu
130 135 140
ttc cgt aca atg ttc ttc tcg aaa tgc tca caa atg agc cca aga tct 480
Phe Arg Thr Met Phe Phe Ser Lys Cys Ser Gln Met Ser Pro Arg Ser
145 150 155 160
gca aaa cga tgt gta ctt ttg ata tgg ata atg gcg gtt ctt gtg tct 528
Ala Lys Arg Cys Val Leu Leu Ile Trp Ile Met Ala Val Leu Val Ser
165 170 175
cta cca cat gga ttg ttc cat aat aca tac gaa ttt cca gat gac aat 576
Leu Pro His Gly Leu Phe His Asn Thr Tyr Glu Phe Pro Asp Asp Asn
180 185 190
aat act tca att gta cag tgt ctc cca gta tat cct gat gct ggt tgg 624
Asn Thr Ser Ile Val Gln Cys Leu Pro Val Tyr Pro Asp Ala Gly Trp
195 200 205
tgg aaa aca tac aat gtc tac ctt gtc ata atc caa tat ttt gtt cca 672
Trp Lys Thr Tyr Asn Val Tyr Leu Val Ile Ile Gln Tyr Phe Val Pro
210 215 220
atg att att ctt gac act gcg tac aca atg att gct gtt aaa ata tgg 720
Met Ile Ile Leu Asp Thr Ala Tyr Thr Met Ile Ala Val Lys Ile Trp
225 230 235 240
tca ttg agt cag tca aga gtt gaa ctt gat gaa aca aaa atg gca acc 768
Ser Leu Ser Gln Ser Arg Val Glu Leu Asp Glu Thr Lys Met Ala Thr
245 250 255
cag aag ata tca gtg gta tca atg gtt tca cca aac act caa tta tcg 816
Gln Lys Ile Ser Val Val Ser Met Val Ser Pro Asn Thr Gln Leu Ser
260 265 270
cag ctt atg cgt act ctc atc att gtc gtt gcc tgt ttc tca ttg tgt 864
Gln Leu Met Arg Thr Leu Ile Ile Val Val Ala Cys Phe Ser Leu Cys
275 280 285
tgg ttt cca ttg gag acg tat cta ctt ttg aat gaa ttg aaa ccg gaa 912
Trp Phe Pro Leu Glu Thr Tyr Leu Leu Leu Asn Glu Leu Lys Pro Glu
290 295 300
att aat gga tgg aaa tac atc aat ttg gtg ttc ttc ttt tca cat tgg 960
Ile Asn Gly Trp Lys Tyr Ile Asn Leu Val Phe Phe Phe Ser His Trp
305 310 315 320
ctg gcg atg agc aat tct tgt ctt aat cca att att tat gga ctt tac 1008
Leu Ala Met Ser Asn Ser Cys Leu Asn Pro Ile Ile Tyr Gly Leu Tyr
325 330 335
aat aca aaa tac aac gag gaa tat cgt cgt ttg ttt cgc caa att gga 1056
Asn Thr Lys Tyr Asn Glu Glu Tyr Arg Arg Leu Phe Arg Gln Ile Gly
340 345 350
tgc att tgg caa cgg cag aaa agt ttg gac gat tcg atg aaa ccg gag 1104
Cys Ile Trp Gln Arg Gln Lys Ser Leu Asp Asp Ser Met Lys Pro Glu
355 360 365
cgt cgt tgg aat tct tca aat gat tgt caa gat caa cag gaa att gat 1152
Arg Arg Trp Asn Ser Ser Asn Asp Cys Gln Asp Gln Gln Glu Ile Asp
370 375 380
caa att gtt gat att cca cca gtt att tct aca aat aat ctt tct ccc 1200
Gln Ile Val Asp Ile Pro Pro Val Ile Ser Thr Asn Asn Leu Ser Pro
385 390 395 400
tg 1202
4
400
›PRT
Homo sapiens
4
Met Asn Phe Ser Ala Thr Asp Ser Ile Leu Ala Ser Thr Ile Thr Thr
1 5 10 15
Val Ile Gly Gly Ala Gly Val Leu Ala Glu Ala Gly Glu Ala Glu Leu
20 25 30
Ser Gly Asp Asp Asp Phe Tyr Glu Leu Thr Pro Val Glu Leu Ile Ile
35 40 45
Trp Cys Met Leu Tyr Ala Ile Ile Ala Phe Met Ala Val Val Gly Asn
50 55 60
Leu Leu Val Leu Tyr Ile Thr Leu Phe Arg Leu Arg Val Arg Ser Ile
65 70 75 80
Thr Thr Tyr Phe Ile Leu Asn Leu Gly Phe Ala Asp Leu Phe Thr Gly
85 90 95
Ile Phe Ala Ile Pro Phe Lys Phe Gln Ala Ala Leu Phe Gln Glu Trp
100 105 110
Phe Leu Pro Arg Ser Leu Cys Arg Ile Val Pro Tyr Val Glu Thr Val
115 120 125
Ala Leu Thr Val Ser Val Phe Thr Leu Val Thr Ser Ala Val His Glu
130 135 140
Phe Arg Thr Met Phe Phe Ser Lys Cys Ser Gln Met Ser Pro Arg Ser
145 150 155 160
Ala Lys Arg Cys Val Leu Leu Ile Trp Ile Met Ala Val Leu Val Ser
165 170 175
Leu Pro His Gly Leu Phe His Asn Thr Tyr Glu Phe Pro Asp Asp Asn
180 185 190
Asn Thr Ser Ile Val Gln Cys Leu Pro Val Tyr Pro Asp Ala Gly Trp
195 200 205
Trp Lys Thr Tyr Asn Val Tyr Leu Val Ile Ile Gln Tyr Phe Val Pro
210 215 220
Met Ile Ile Leu Asp Thr Ala Tyr Thr Met Ile Ala Val Lys Ile Trp
225 230 235 240
Ser Leu Ser Gln Ser Arg Val Glu Leu Asp Glu Thr Lys Met Ala Thr
245 250 255
Gln Lys Ile Ser Val Val Ser Met Val Ser Pro Asn Thr Gln Leu Ser
260 265 270
Gln Leu Met Arg Thr Leu Ile Ile Val Val Ala Cys Phe Ser Leu Cys
275 280 285
Trp Phe Pro Leu Glu Thr Tyr Leu Leu Leu Asn Glu Leu Lys Pro Glu
290 295 300
Ile Asn Gly Trp Lys Tyr Ile Asn Leu Val Phe Phe Phe Ser His Trp
305 310 315 320
Leu Ala Met Ser Asn Ser Cys Leu Asn Pro Ile Ile Tyr Gly Leu Tyr
325 330 335
Asn Thr Lys Tyr Asn Glu Glu Tyr Arg Arg Leu Phe Arg Gln Ile Gly
340 345 350
Cys Ile Trp Gln Arg Gln Lys Ser Leu Asp Asp Ser Met Lys Pro Glu
355 360 365
Arg Arg Trp Asn Ser Ser Asn Asp Cys Gln Asp Gln Gln Glu Ile Asp
370 375 380
Gln Ile Val Asp Ile Pro Pro Val Ile Ser Thr Asn Asn Leu Ser Pro
385 390 395 400
5
atg tcg aat gat ctc gtg cct tca gtg tct tct ata cta aat gaa aca 48
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
aca cct tca tat caa agt aca tgt aaa atc aaa aac aac ccc atg gaa 96
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
atg gaa tat ttc cgc cca ttc ttt att tct atg tat tgt gct gtt ttt 144
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
ttg gtt gct agt tcc ggt aat ttt ttg gtg gtg tac gtt gtg atg acg 192
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
aac aag cga atg cag acg atc acc aac att ttt att aca aat ctc gca 240
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
gtt tct gat ata atg gtt aac ttt aca tcg ttg tgg ctt aca cca aca 288
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
tac acc tca ata gga cat tgg ata ttc gga ggt gga ttg tgt cat ggt 336
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
tta ccc ctg ttc caa ggt aca agt atc ttc atc agt acg tgg aca ctt 384
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
acg gct ata gcc ata gat cga tac ata gtg atc gtg cac aac tca tca 432
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
aat atc aat ata aat gat aga atg tcc atg aga tct tgc ctt tcg ttc 480
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
att gtc ctc att tgg cta tgc tca ttg ctt ctg gtc act ccg tat gcc 528
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
atc aac atg aag ctc aac tac att cat gaa cca tgt gat ttt ctg ata 576
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
tgt agt gag gac tgg agc aat gcc gaa ttt cga tct att ttt gga att 624
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
gtg gtg atg att ctt caa ttc att ttg cca ttt gta ctc att gcc att 672
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
agt tac ata aaa ata tgg ttg ttc cta aat agc cgt caa agt atg acc 720
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
gag aga aaa tcc gat atc aag cgc aaa aag cgt tta cta cgg atg ttg 768
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
att gtc atg gtg gtc atc ttc gca att tgc tgg ttc cca ttc aac ctc 816
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
tta aac tgt ctc cga gat ctc aag ttg gat aat ttc atg cgt ggc tac 864
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
ttc agt ttt gtt ttc ctt tcc gtg cat ttg atg agt atg aca gct acc 912
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
gcc tgg aat cca atc ctc tac gca ttc atg aat gag acc ttc cgt gag 960
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
gag ttc gca aaa gtt gtt ccc tgc ttg ttt gcg cgt cgt cct gga act 1008
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
ggt cca att cgc gtc atc act gaa cgc acc gct atg ata act aac ccg 1056
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
ttt cga cgt gca aac cga aaa aaa aaa gtg gag gaa cag cct gtt acg 1104
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
gtg att tct gat tca gag agt cca ctt caa act gca gtg gag ccg cag 1152
Val Ile Ser Asp Ser Glu Ser Pro Leu Gln Thr Ala Val Glu Pro Gln
370 375 380
cgc agt atc gtg tat ctc gat gag cct gag aac gga tca agt tgt cag 1200
Arg Ser Ile Val Tyr Leu Asp Glu Pro Glu Asn Gly Ser Ser Cys Gln
385 390 395 400
acg ctt ctg ctt ta 1214
Thr Leu Leu Leu
6
404
›PRT
Homo sapiens
6
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
Val Ile Ser Asp Ser Glu Ser Pro Leu Gln Thr Ala Val Glu Pro Gln
370 375 380
Arg Ser Ile Val Tyr Leu Asp Glu Pro Glu Asn Gly Ser Ser Cys Gln
385 390 395 400
Thr Leu Leu Leu
7
atg tcg aat gat ctc gtg cct tca gtg tct tct ata cta aat gaa aca 48
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
aca cct tca tat caa agt aca tgt aaa atc aaa aac aac ccc atg gaa 96
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
atg gaa tat ttc cgc cca ttc ttt att tct atg tat tgt gct gtt ttt 144
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
ttg gtt gct agt tcc ggt aat ttt ttg gtg gtg tac gtt gtg atg acg 192
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
aac aag cga atg cag acg atc acc aac att ttt att aca aat ctc gca 240
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
gtt tct gat ata atg gtt aac ttt aca tcg ttg tgg ctt aca cca aca 288
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
tac acc tca ata gga cat tgg ata ttc gga ggt gga ttg tgt cat ggt 336
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
tta ccc ctg ttc caa ggt aca agt atc ttc atc agt acg tgg aca ctt 384
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
acg gct ata gcc ata gat cga tac ata gtg atc gtg cac aac tca tca 432
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
aat atc aat ata aat gat aga atg tcc atg aga tct tgc ctt tcg ttc 480
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
att gtc ctc att tgg cta tgc tca ttg ctt ctg gtc act ccg tat gcc 528
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
atc aac atg aag ctc aac tac att cat gaa cca tgt gat ttt ctg ata 576
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
tgt agt gag gac tgg agc aat gcc gaa ttt cga tct att ttt gga att 624
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
gtg gtg atg att ctt caa ttc att ttg cca ttt gta ctc att gcc att 672
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
agt tac ata aaa ata tgg ttg ttc cta aat agc cgt caa agt atg acc 720
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
gag aga aaa tcc gat atc aag cgc aaa aag cgt tta cta cgg atg ttg 768
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
att gtc atg gtg gtc atc ttc gca att tgc tgg ttc cca ttc aac ctc 816
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
tta aac tgt ctc cga gat ctc aag ttg gat aat ttc atg cgt ggc tac 864
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
ttc agt ttt gtt ttc ctt tcc gtg cat ttg atg agt atg aca gct acc 912
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
gcc tgg aat cca atc ctc tac gca ttc atg aat gag acc ttc cgt gag 960
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
gag ttc gca aaa gtt gtt ccc tgc ttg ttt gcg cgt cgt cct gga act 1008
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
ggt cca att cgc gtc atc act gaa cgc acc gct atg ata act aac ccg 1056
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
ttt cga cgt gca aac cga aaa aaa aaa gtg gag gaa cag cct gtt acg 1104
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
gtg att tct gag agt cca ctt caa act gca gtg gag ccg cag cgc agt 1152
Val Ile Ser Glu Ser Pro Leu Gln Thr Ala Val Glu Pro Gln Arg Ser
370 375 380
atc gtg tat ctc gat gag cct gag aac gga tca agt tgt cag acg ctt 1200
Ile Val Tyr Leu Asp Glu Pro Glu Asn Gly Ser Ser Cys Gln Thr Leu
385 390 395 400
ctg ctt ta 1208
Leu Leu
8
402
›PRT
Homo sapiens
8
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
Val Ile Ser Glu Ser Pro Leu Gln Thr Ala Val Glu Pro Gln Arg Ser
370 375 380
Ile Val Tyr Leu Asp Glu Pro Glu Asn Gly Ser Ser Cys Gln Thr Leu
385 390 395 400
Leu Leu
9
atg tcg aat gat ctc gtg cct tca gtg tct tct ata cta aat gaa aca 48
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
aca cct tca tat caa agt aca tgt aaa atc aaa aac aac ccc atg gaa 96
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
atg gaa tat ttc cgc cca ttc ttt att tct atg tat tgt gct gtt ttt 144
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
ttg gtt gct agt tcc ggt aat ttt ttg gtg gtg tac gtt gtg atg acg 192
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
aac aag cga atg cag acg atc acc aac att ttt att aca aat ctc gca 240
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
gtt tct gat ata atg gtt aac ttt aca tcg ttg tgg ctt aca cca aca 288
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
tac acc tca ata gga cat tgg ata ttc gga ggt gga ttg tgt cat ggt 336
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
tta ccc ctg ttc caa ggt aca agt atc ttc atc agt acg tgg aca ctt 384
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
acg gct ata gcc ata gat cga tac ata gtg atc gtg cac aac tca tca 432
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
aat atc aat ata aat gat aga atg tcc atg aga tct tgc ctt tcg ttc 480
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
att gtc ctc att tgg cta tgc tca ttg ctt ctg gtc act ccg tat gcc 528
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
atc aac atg aag ctc aac tac att cat gaa cca tgt gat ttt ctg ata 576
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
tgt agt gag gac tgg agc aat gcc gaa ttt cga tct att ttt gga att 624
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
gtg gtg atg att ctt caa ttc att ttg cca ttt gta ctc att gcc att 672
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
agt tac ata aaa ata tgg ttg ttc cta aat agc cgt caa agt atg acc 720
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
gag aga aaa tcc gat atc aag cgc aaa aag cgt tta cta cgg atg ttg 768
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
att gtc atg gtg gtc atc ttc gca att tgc tgg ttc cca ttc aac ctc 816
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
tta aac tgt ctc cga gat ctc aag ttg gat aat ttc atg cgt ggc tac 864
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
ttc agt ttt gtt ttc ctt tcc gtg cat ttg atg agt atg aca gct acc 912
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
gcc tgg aat cca atc ctc tac gca ttc atg aat gag acc ttc cgt gag 960
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
gag ttc gca aaa gtt gtt ccc tgc ttg ttt gcg cgt cgt cct gga act 1008
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
ggt cca att cgc gtc atc act gaa cgc acc gct atg ata act aac ccg 1056
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
ttt cga cgt gca aac cga aaa aaa aaa gtg gag gaa cag cct gtt acg 1104
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
gtg att tct gaa aca gat tca gag agt cca ctt caa act gca gtg gag 1152
Val Ile Ser Glu Thr Asp Ser Glu Ser Pro Leu Gln Thr Ala Val Glu
370 375 380
ccg cag cgc agt atc gtg tat ctc gat gag cct gag aac gga tca agt 1200
Pro Gln Arg Ser Ile Val Tyr Leu Asp Glu Pro Glu Asn Gly Ser Ser
385 390 395 400
tgt cag acg ctt ctg ctt ta 1220
Cys Gln Thr Leu Leu Leu
405
10
406
›PRT
Homo sapiens
10
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
Val Ile Ser Glu Thr Asp Ser Glu Ser Pro Leu Gln Thr Ala Val Glu
370 375 380
Pro Gln Arg Ser Ile Val Tyr Leu Asp Glu Pro Glu Asn Gly Ser Ser
385 390 395 400
Cys Gln Thr Leu Leu Leu
11
atg tcg aat gat ctc gtg cct tca gtg tct tct ata cta aat gaa aca 48
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
aca cct tca tat caa agt aca tgt aaa atc aaa aac aac ccc atg gaa 96
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
atg gaa tat ttc cgc cca ttc ttt att tct atg tat tgt gct gtt ttt 144
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
ttg gtt gct agt tcc ggt aat ttt ttg gtg gtg tac gtt gtg atg acg 192
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
aac aag cga atg cag acg atc acc aac att ttt att aca aat ctc gca 240
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
gtt tct gat ata atg gtt aac ttt aca tcg ttg tgg ctt aca cca aca 288
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
tac acc tca ata gga cat tgg ata ttc gga ggt gga ttg tgt cat ggt 336
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
tta ccc ctg ttc caa ggt aca agt atc ttc atc agt acg tgg aca ctt 384
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
acg gct ata gcc ata gat cga tac ata gtg atc gtg cac aac tca tca 432
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
aat atc aat ata aat gat aga atg tcc atg aga tct tgc ctt tcg ttc 480
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
att gtc ctc att tgg cta tgc tca ttg ctt ctg gtc act ccg tat gcc 528
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
atc aac atg aag ctc aac tac att cat gaa cca tgt gat ttt ctg ata 576
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
tgt agt gag gac tgg agc aat gcc gaa ttt cga tct att ttt gga att 624
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
gtg gtg atg att ctt caa ttc att ttg cca ttt gta ctc att gcc att 672
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
agt tac ata aaa ata tgg ttg ttc cta aat agc cgt caa agt atg acc 720
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
gag aga aaa tcc gat atc aag cgc aaa aag cgt tta cta cgg atg ttg 768
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
att gtc atg gtg gtc atc ttc gca att tgc tgg ttc cca ttc aac ctc 816
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
tta aac tgt ctc cga gat ctc aag ttg gat aat ttc atg cgt ggc tac 864
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
ttc agt ttt gtt ttc ctt tcc gtg cat ttg atg agt atg aca gct acc 912
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
gcc tgg aat cca atc ctc tac gca ttc atg aat gag acc ttc cgt gag 960
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
gag ttc gca aaa gtt gtt ccc tgc ttg ttt gcg cgt cgt cct gga act 1008
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
ggt cca att cgc gtc atc act gaa cgc acc gct atg ata act aac ccg 1056
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
ttt cga cgt gca aac cga aaa aaa aaa gtg gag gaa cag cct gtt acg 1104
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
gtg att tct gaa ctt tta cac ccg aca tcc aac gaa gaa taggaagcca 1153
Val Ile Ser Glu Leu Leu His Pro Thr Ser Asn Glu Glu
370 375 380
ttcattccct gcatgcactc gttgtcagtg aacttttcaa caaaccagca atctttagaa 1213
acagattcag agagtccact tcaaactgca gtggagccgc agcgcagtat cgtgtatctc 1273
gatgagcctg agaacggatc aagttgtcag acgcttctgc tttaa 1318
12
381
›PRT
Homo sapiens
12
Met Ser Asn Asp Leu Val Pro Ser Val Ser Ser Ile Leu Asn Glu Thr
1 5 10 15
Thr Pro Ser Tyr Gln Ser Thr Cys Lys Ile Lys Asn Asn Pro Met Glu
20 25 30
Met Glu Tyr Phe Arg Pro Phe Phe Ile Ser Met Tyr Cys Ala Val Phe
35 40 45
Leu Val Ala Ser Ser Gly Asn Phe Leu Val Val Tyr Val Val Met Thr
50 55 60
Asn Lys Arg Met Gln Thr Ile Thr Asn Ile Phe Ile Thr Asn Leu Ala
65 70 75 80
Val Ser Asp Ile Met Val Asn Phe Thr Ser Leu Trp Leu Thr Pro Thr
85 90 95
Tyr Thr Ser Ile Gly His Trp Ile Phe Gly Gly Gly Leu Cys His Gly
100 105 110
Leu Pro Leu Phe Gln Gly Thr Ser Ile Phe Ile Ser Thr Trp Thr Leu
115 120 125
Thr Ala Ile Ala Ile Asp Arg Tyr Ile Val Ile Val His Asn Ser Ser
130 135 140
Asn Ile Asn Ile Asn Asp Arg Met Ser Met Arg Ser Cys Leu Ser Phe
145 150 155 160
Ile Val Leu Ile Trp Leu Cys Ser Leu Leu Leu Val Thr Pro Tyr Ala
165 170 175
Ile Asn Met Lys Leu Asn Tyr Ile His Glu Pro Cys Asp Phe Leu Ile
180 185 190
Cys Ser Glu Asp Trp Ser Asn Ala Glu Phe Arg Ser Ile Phe Gly Ile
195 200 205
Val Val Met Ile Leu Gln Phe Ile Leu Pro Phe Val Leu Ile Ala Ile
210 215 220
Ser Tyr Ile Lys Ile Trp Leu Phe Leu Asn Ser Arg Gln Ser Met Thr
225 230 235 240
Glu Arg Lys Ser Asp Ile Lys Arg Lys Lys Arg Leu Leu Arg Met Leu
245 250 255
Ile Val Met Val Val Ile Phe Ala Ile Cys Trp Phe Pro Phe Asn Leu
260 265 270
Leu Asn Cys Leu Arg Asp Leu Lys Leu Asp Asn Phe Met Arg Gly Tyr
275 280 285
Phe Ser Phe Val Phe Leu Ser Val His Leu Met Ser Met Thr Ala Thr
290 295 300
Ala Trp Asn Pro Ile Leu Tyr Ala Phe Met Asn Glu Thr Phe Arg Glu
305 310 315 320
Glu Phe Ala Lys Val Val Pro Cys Leu Phe Ala Arg Arg Pro Gly Thr
325 330 335
Gly Pro Ile Arg Val Ile Thr Glu Arg Thr Ala Met Ile Thr Asn Pro
340 345 350
Phe Arg Arg Ala Asn Arg Lys Lys Lys Val Glu Glu Gln Pro Val Thr
355 360 365
Val Ile Ser Glu Leu Leu His Pro Thr Ser Asn Glu Glu
370 375 380
13
atg atc aac gaa aca gaa gag aca tgt gat cga tat ata gac aag cat 48
Met Ile Asn Glu Thr Glu Glu Thr Cys Asp Arg Tyr Ile Asp Lys His
1 5 10 15
cca gat atg aca aat gaa cca aca gtc ctt gtc aca ttc tcc ctg ctt 96
Pro Asp Met Thr Asn Glu Pro Thr Val Leu Val Thr Phe Ser Leu Leu
20 25 30
tat ctt cat att ttt ctt ctt gga atc cta gga aat tcg gct gtt cta 144
Tyr Leu His Ile Phe Leu Leu Gly Ile Leu Gly Asn Ser Ala Val Leu
35 40 45
tat ctt aca atg aaa cat cgt caa tta caa act gtt caa aat ata ttt 192
Tyr Leu Thr Met Lys His Arg Gln Leu Gln Thr Val Gln Asn Ile Phe
50 55 60
att ttg aac tta tgc gca tcg aat gtt tta atg tgc ttg acg agt ctt 240
Ile Leu Asn Leu Cys Ala Ser Asn Val Leu Met Cys Leu Thr Ser Leu
65 70 75 80
cca atc aca ttt atc aca aat gtc tac aaa caa tgg ttc ttc tca tcg 288
Pro Ile Thr Phe Ile Thr Asn Val Tyr Lys Gln Trp Phe Phe Ser Ser
85 90 95
cct gtt tgt aaa ctt att cca ttg gtt cag gga gct tca atc ttt gtc 336
Pro Val Cys Lys Leu Ile Pro Leu Val Gln Gly Ala Ser Ile Phe Val
100 105 110
tcc act ttc tcc ctt tcc gcc att gca ctt gac cga tac aac ctc gtg 384
Ser Thr Phe Ser Leu Ser Ala Ile Ala Leu Asp Arg Tyr Asn Leu Val
115 120 125
gtt cgt cct cat aaa caa aaa ctc agt tca cga agt gca atg atg att 432
Val Arg Pro His Lys Gln Lys Leu Ser Ser Arg Ser Ala Met Met Ile
130 135 140
gct ctt ctc att tgg gtt att agt gta gtt gtt tgt atg cca tat gga 480
Ala Leu Leu Ile Trp Val Ile Ser Val Val Val Cys Met Pro Tyr Gly
145 150 155 160
tgg tac atg gat gtg gag aag ctc aat gga tta tgt gga gag tac tgt 528
Trp Tyr Met Asp Val Glu Lys Leu Asn Gly Leu Cys Gly Glu Tyr Cys
165 170 175
tct gag cac tgg cca tta gca gaa gta cga aag gga tat act ttt ttg 576
Ser Glu His Trp Pro Leu Ala Glu Val Arg Lys Gly Tyr Thr Phe Leu
180 185 190
gtg tta atc act caa ttt ttg ttt cct ttt gct acg atg gct ttt tgc 624
Val Leu Ile Thr Gln Phe Leu Phe Pro Phe Ala Thr Met Ala Phe Cys
195 200 205
tat tat aat att ttt tca aga ctc aga caa cga gtg gaa acg aaa ctg 672
Tyr Tyr Asn Ile Phe Ser Arg Leu Arg Gln Arg Val Glu Thr Lys Leu
210 215 220
aag aaa tta tca gaa aga tct caa ctt ttg gag aat agt aca aca tgt 720
Lys Lys Leu Ser Glu Arg Ser Gln Leu Leu Glu Asn Ser Thr Thr Cys
225 230 235 240
gga acc act aac cac att gtt agc att aac gcg gaa gca gtt caa aat 768
Gly Thr Thr Asn His Ile Val Ser Ile Asn Ala Glu Ala Val Gln Asn
245 250 255
ggt ctg gaa aac aaa caa aga tta gct gtt ctt gct caa caa aga agg 816
Gly Leu Glu Asn Lys Gln Arg Leu Ala Val Leu Ala Gln Gln Arg Arg
260 265 270
aca act act att tta tcg tgc atg gtt ctt ctt ttt gca ttt aca tgg 864
Thr Thr Thr Ile Leu Ser Cys Met Val Leu Leu Phe Ala Phe Thr Trp
275 280 285
ctt cca cat aat gtt gtc act ttg atg ata gaa tac gat gga ttc ttt 912
Leu Pro His Asn Val Val Thr Leu Met Ile Glu Tyr Asp Gly Phe Phe
290 295 300
ttt cat tct gat gag aca tcg gca acc agt acc gat cat aca tat att 960
Phe His Ser Asp Glu Thr Ser Ala Thr Ser Thr Asp His Thr Tyr Ile
305 310 315 320
gtt tca atg act gct cat tta ata tcc atg tta aca aat gta acc aac 1008
Val Ser Met Thr Ala His Leu Ile Ser Met Leu Thr Asn Val Thr Asn
325 330 335
cca ttt ctg tat gct tgg ctt aac cca atg ttc aaa gaa atg ctt att 1056
Pro Phe Leu Tyr Ala Trp Leu Asn Pro Met Phe Lys Glu Met Leu Ile
340 345 350
aaa act ctc aga ggt gga agt aaa tcc ccg aaa cca gct gac atc aag 1104
Lys Thr Leu Arg Gly Gly Ser Lys Ser Pro Lys Pro Ala Asp Ile Lys
355 360 365
caa act tca ttc att cga atg ccg aat agt ggt gcg cca agt caa tct 1152
Gln Thr Ser Phe Ile Arg Met Pro Asn Ser Gly Ala Pro Ser Gln Ser
370 375 380
tcc tac ctg tg 1163
Ser Tyr Leu
385
14
387
›PRT
Homo sapiens
14
Met Ile Asn Glu Thr Glu Glu Thr Cys Asp Arg Tyr Ile Asp Lys His
1 5 10 15
Pro Asp Met Thr Asn Glu Pro Thr Val Leu Val Thr Phe Ser Leu Leu
20 25 30
Tyr Leu His Ile Phe Leu Leu Gly Ile Leu Gly Asn Ser Ala Val Leu
35 40 45
Tyr Leu Thr Met Lys His Arg Gln Leu Gln Thr Val Gln Asn Ile Phe
50 55 60
Ile Leu Asn Leu Cys Ala Ser Asn Val Leu Met Cys Leu Thr Ser Leu
65 70 75 80
Pro Ile Thr Phe Ile Thr Asn Val Tyr Lys Gln Trp Phe Phe Ser Ser
85 90 95
Pro Val Cys Lys Leu Ile Pro Leu Val Gln Gly Ala Ser Ile Phe Val
100 105 110
Ser Thr Phe Ser Leu Ser Ala Ile Ala Leu Asp Arg Tyr Asn Leu Val
115 120 125
Val Arg Pro His Lys Gln Lys Leu Ser Ser Arg Ser Ala Met Met Ile
130 135 140
Ala Leu Leu Ile Trp Val Ile Ser Val Val Val Cys Met Pro Tyr Gly
145 150 155 160
Trp Tyr Met Asp Val Glu Lys Leu Asn Gly Leu Cys Gly Glu Tyr Cys
165 170 175
Ser Glu His Trp Pro Leu Ala Glu Val Arg Lys Gly Tyr Thr Phe Leu
180 185 190
Val Leu Ile Thr Gln Phe Leu Phe Pro Phe Ala Thr Met Ala Phe Cys
195 200 205
Tyr Tyr Asn Ile Phe Ser Arg Leu Arg Gln Arg Val Glu Thr Lys Leu
210 215 220
Lys Lys Leu Ser Glu Arg Ser Gln Leu Leu Glu Asn Ser Thr Thr Cys
225 230 235 240
Gly Thr Thr Asn His Ile Val Ser Ile Asn Ala Glu Ala Val Gln Asn
245 250 255
Gly Leu Glu Asn Lys Gln Arg Leu Ala Val Leu Ala Gln Gln Arg Arg
260 265 270
Thr Thr Thr Ile Leu Ser Cys Met Val Leu Leu Phe Ala Phe Thr Trp
275 280 285
Leu Pro His Asn Val Val Thr Leu Met Ile Glu Tyr Asp Gly Phe Phe
290 295 300
Phe His Ser Asp Glu Thr Ser Ala Thr Ser Thr Asp His Thr Tyr Ile
305 310 315 320
Val Ser Met Thr Ala His Leu Ile Ser Met Leu Thr Asn Val Thr Asn
325 330 335
Pro Phe Leu Tyr Ala Trp Leu Asn Pro Met Phe Lys Glu Met Leu Ile
340 345 350
Lys Thr Leu Arg Gly Gly Ser Lys Ser Pro Lys Pro Ala Asp Ile Lys
355 360 365
Gln Thr Ser Phe Ile Arg Met Pro Asn Ser Gly Ala Pro Ser Gln Ser
370 375 380
Ser Tyr Leu
385
15
atg ggt gat gct gaa tct cat cat tgt ata gat gtg aac gcc att ctt 48
Met Gly Asp Ala Glu Ser His His Cys Ile Asp Val Asn Ala Ile Leu
1 5 10 15
cag cag ttc aat gat tgg aca gtc ctc ttt gaa gtt cgg ctt gga tat 96
Gln Gln Phe Asn Asp Trp Thr Val Leu Phe Glu Val Arg Leu Gly Tyr
20 25 30
tca gta cta tac ttt ctc ata tta ata atc gga ttg gtt gga aat ggg 144
Ser Val Leu Tyr Phe Leu Ile Leu Ile Ile Gly Leu Val Gly Asn Gly
35 40 45
cta ttg atc act tca att tta atg cga aag aaa ctt tcc gtg gca aac 192
Leu Leu Ile Thr Ser Ile Leu Met Arg Lys Lys Leu Ser Val Ala Asn
50 55 60
ata ttc ttg ata aac ctg gca gtt tct gat ttg ctt ctt tgc atc acg 240
Ile Phe Leu Ile Asn Leu Ala Val Ser Asp Leu Leu Leu Cys Ile Thr
65 70 75 80
gcg gtg ccg atc act cca gta ttg gcg ttt atg aag cga tgg ata ttt 288
Ala Val Pro Ile Thr Pro Val Leu Ala Phe Met Lys Arg Trp Ile Phe
85 90 95
gga ata att atg tgt aaa ttg gtt cca act tgt cag gcg ttt tcg gtg 336
Gly Ile Ile Met Cys Lys Leu Val Pro Thr Cys Gln Ala Phe Ser Val
100 105 110
ctc att tct tca tgg tct ttg tgt tac atc gca att gat aga tat cga 384
Leu Ile Ser Ser Trp Ser Leu Cys Tyr Ile Ala Ile Asp Arg Tyr Arg
115 120 125
agt att gtg acg cca ctc cgg gaa cca tgg tct gat agg cat gca agg 432
Ser Ile Val Thr Pro Leu Arg Glu Pro Trp Ser Asp Arg His Ala Arg
130 135 140
tgg ctt ctg atg ttc aca tgg gtg gtc gcc ttc ctt gct agt tat cct 480
Trp Leu Leu Met Phe Thr Trp Val Val Ala Phe Leu Ala Ser Tyr Pro
145 150 155 160
cta tat tac tca cag aac ttg aaa aca atg gtt att gaa aat gtg aca 528
Leu Tyr Tyr Ser Gln Asn Leu Lys Thr Met Val Ile Glu Asn Val Thr
165 170 175
tta tgt gga gat ttt tgc ggc gag ttc aat tgg cag tcg gat gaa ata 576
Leu Cys Gly Asp Phe Cys Gly Glu Phe Asn Trp Gln Ser Asp Glu Ile
180 185 190
tcc aag ttg aca tat act acg agt tta ttg att att cag ctg att att 624
Ser Lys Leu Thr Tyr Thr Thr Ser Leu Leu Ile Ile Gln Leu Ile Ile
195 200 205
cca gca att atc atg tct ttt tgt tat tta atg att cta caa aag gta 672
Pro Ala Ile Ile Met Ser Phe Cys Tyr Leu Met Ile Leu Gln Lys Val
210 215 220
caa acc gac tgg ctt gtc gac gag gga tcc atg ttg act gcc gca caa 720
Gln Thr Asp Trp Leu Val Asp Glu Gly Ser Met Leu Thr Ala Ala Gln
225 230 235 240
cag gct caa aca gca gtt cga aag cga cga gtg atg tac gtg ttg att 768
Gln Ala Gln Thr Ala Val Arg Lys Arg Arg Val Met Tyr Val Leu Ile
245 250 255
cta atg gtt att gtt ttt atg gct tgc tgg ttc ccg ttg tcc gcc gtg 816
Leu Met Val Ile Val Phe Met Ala Cys Trp Phe Pro Leu Ser Ala Val
260 265 270
aat ttg ttc aga gat ctc gga atg cga ttc gag ttc tgt caa act gtt 864
Asn Leu Phe Arg Asp Leu Gly Met Arg Phe Glu Phe Cys Gln Thr Val
275 280 285
tac aag gtt tta atg atg gac caa atg tat ttc aag ttg ctc aat gtg 912
Tyr Lys Val Leu Met Met Asp Gln Met Tyr Phe Lys Leu Leu Asn Val
290 295 300
cac gtc atc gcg atg act tcg atc gta tgg aat ccg gtg ctc tat ttc 960
His Val Ile Ala Met Thr Ser Ile Val Trp Asn Pro Val Leu Tyr Phe
305 310 315 320
tgg atg agc aag cgt cat cga cga gcc ctg aaa gac gac atg acg tgg 1008
Trp Met Ser Lys Arg His Arg Arg Ala Leu Lys Asp Asp Met Thr Trp
325 330 335
ctc acc aat gct cgc cgt cat aca aac gtc ggc gtt ctg tct cgc ttc 1056
Leu Thr Asn Ala Arg Arg His Thr Asn Val Gly Val Leu Ser Arg Phe
340 345 350
aca cct tct cca tca gtt tca gtg gtt tac aga cga act ctg gag cga 1104
Thr Pro Ser Pro Ser Val Ser Val Val Tyr Arg Arg Thr Leu Glu Arg
355 360 365
cat cta ggt gtc aat cat ttc cgc cgt ggc aca ctt gcg gac ccg aca 1152
His Leu Gly Val Asn His Phe Arg Arg Gly Thr Leu Ala Asp Pro Thr
370 375 380
tgc act tca cgt gaa cga agt ctt ccg cga gaa ctt caa tca aat tgt 1200
Cys Thr Ser Arg Glu Arg Ser Leu Pro Arg Glu Leu Gln Ser Asn Cys
385 390 395 400
ttc ctt ctt gtt ccg ctt atg cca tta tgt caa tct gtg acg agg aag 1248
Phe Leu Leu Val Pro Leu Met Pro Leu Cys Gln Ser Val Thr Arg Lys
405 410 415
aat agt cat cta gca atc aat cga gac ggg atc gtc att cca caa gcc 1296
Asn Ser His Leu Ala Ile Asn Arg Asp Gly Ile Val Ile Pro Gln Ala
420 425 430
aat ggc tca agt cgt cgg ccg agc agc gtg aat acc aat tca act cga 1344
Asn Gly Ser Ser Arg Arg Pro Ser Ser Val Asn Thr Asn Ser Thr Arg
435 440 445
gac tgg tg 1352
Asp Trp
450
16
450
›PRT · 1 of 2
Homo sapiens
16
Met Gly Asp Ala Glu Ser His His Cys Ile Asp Val Asn Ala Ile Leu
1 5 10 15
Gln Gln Phe Asn Asp Trp Thr Val Leu Phe Glu Val Arg Leu Gly Tyr
20 25 30
Ser Val Leu Tyr Phe Leu Ile Leu Ile Ile Gly Leu Val Gly Asn Gly
35 40 45
Leu Leu Ile Thr Ser Ile Leu Met Arg Lys Lys Leu Ser Val Ala Asn
50 55 60
Ile Phe Leu Ile Asn Leu Ala Val Ser Asp Leu Leu Leu Cys Ile Thr
65 70 75 80
Ala Val Pro Ile Thr Pro Val Leu Ala Phe Met Lys Arg Trp Ile Phe
85 90 95
Gly Ile Ile Met Cys Lys Leu Val Pro Thr Cys Gln Ala Phe Ser Val
100 105 110
Leu Ile Ser Ser Trp Ser Leu Cys Tyr Ile Ala Ile Asp Arg Tyr Arg
115 120 125
Ser Ile Val Thr Pro Leu Arg Glu Pro Trp Ser Asp Arg His Ala Arg
130 135 140
Trp Leu Leu Met Phe Thr Trp Val Val Ala Phe Leu Ala Ser Tyr Pro
145 150 155 160
Leu Tyr Tyr Ser Gln Asn Leu Lys Thr Met Val Ile Glu Asn Val Thr
165 170 175
Leu Cys Gly Asp Phe Cys Gly Glu Phe Asn Trp Gln Ser Asp Glu Ile
180 185 190
Ser Lys Leu Thr Tyr Thr Thr Ser Leu Leu Ile Ile Gln Leu Ile Ile
195 200 205
Pro Ala Ile Ile Met Ser Phe Cys Tyr Leu Met Ile Leu Gln Lys Val
210 215 220
Gln Thr Asp Trp Leu Val Asp Glu Gly Ser Met Leu Thr Ala Ala Gln
225 230 235 240
Gln Ala Gln Thr Ala Val Arg Lys Arg Arg Val Met Tyr Val Leu Ile
245 250 255
Leu Met Val Ile Val Phe Met Ala Cys Trp Phe Pro Leu Ser Ala Val
260 265 270
Asn Leu Phe Arg Asp Leu Gly Met Arg Phe Glu Phe Cys Gln Thr Val
275 280 285
Tyr Lys Val Leu Met Met Asp Gln Met Tyr Phe Lys Leu Leu Asn Val
290 295 300
His Val Ile Ala Met Thr Ser Ile Val Trp Asn Pro Val Leu Tyr Phe
305 310 315 320
Trp Met Ser Lys Arg His Arg Arg Ala Leu Lys Asp Asp Met Thr Trp
325 330 335
Leu Thr Asn Ala Arg Arg His Thr Asn Val Gly Val Leu Ser Arg Phe
340 345 350
Thr Pro Ser Pro Ser Val Ser Val Val Tyr Arg Arg Thr Leu Glu Arg
355 360 365
His Leu Gly Val Asn His Phe Arg Arg Gly Thr Leu Ala Asp Pro Thr
370 375 380
Cys Thr Ser Arg Glu Arg Ser Leu Pro Arg Glu Leu Gln Ser Asn Cys
385 390 395 400
Phe Leu Leu Val Pro Leu Met Pro Leu Cys Gln Ser Val Thr Arg Lys
405 410 415
Asn Ser His Leu Ala Ile Asn Arg Asp Gly Ile Val Ile Pro Gln Ala
420 425 430
Asn Gly Ser Ser Arg Arg Pro Ser Ser Val Asn Thr Asn Ser Thr Arg
435 440 445
Asp Trp
450
17
atg acc gac gag gag ttg ttc gag atc gct ctt ccc ggc ttc tta tat 48
Met Thr Asp Glu Glu Leu Phe Glu Ile Ala Leu Pro Gly Phe Leu Tyr
1 5 10 15
ctc act gtt ttt cta gtt gga act att gga aac tct atg gta ata ttt 96
Leu Thr Val Phe Leu Val Gly Thr Ile Gly Asn Ser Met Val Ile Phe
20 25 30
gtt gta aac cga ttc aaa cga atg cgc aac gta acg aac atc ttt ctc 144
Val Val Asn Arg Phe Lys Arg Met Arg Asn Val Thr Asn Ile Phe Leu
35 40 45
gcc tcg ctc agc acc gcc gac ttg tgc ctt atc tgg ttt tgt gtg cct 192
Ala Ser Leu Ser Thr Ala Asp Leu Cys Leu Ile Trp Phe Cys Val Pro
50 55 60
ata atg ttc atg aag tat atg tcg cat acg tgg tcg atg ggc agg ttt 240
Ile Met Phe Met Lys Tyr Met Ser His Thr Trp Ser Met Gly Arg Phe
65 70 75 80
gcg tgc tat tct gtg cac tac att caa cag ttt aca tgc ttc tgc tcg 288
Ala Cys Tyr Ser Val His Tyr Ile Gln Gln Phe Thr Cys Phe Cys Ser
85 90 95
gtg ctc acg atg act atg att agt ttt gag agg ttt ctt gca ata gca 336
Val Leu Thr Met Thr Met Ile Ser Phe Glu Arg Phe Leu Ala Ile Ala
100 105 110
tat cct atg cga aat atc tgg ttt tca tcc ata gga cga gct aaa aaa 384
Tyr Pro Met Arg Asn Ile Trp Phe Ser Ser Ile Gly Arg Ala Lys Lys
115 120 125
gta atc cta ctc atc tgg atg tca tcc gcg gta ctg gcc gtc cca acc 432
Val Ile Leu Leu Ile Trp Met Ser Ser Ala Val Leu Ala Val Pro Thr
130 135 140
gct gtt cgg atg gat tat gag acc aat ctt tca ttg tcc ggg cag agg 480
Ala Val Arg Met Asp Tyr Glu Thr Asn Leu Ser Leu Ser Gly Gln Arg
145 150 155 160
gtg cat tgg tgt cga aga cgg ttt ccc gcg caa ttt tta gga tat cct 528
Val His Trp Cys Arg Arg Arg Phe Pro Ala Gln Phe Leu Gly Tyr Pro
165 170 175
agg aca tcg ttg aat aaa gca tat gcc atg tat cag ttg ttg cta ctt 576
Arg Thr Ser Leu Asn Lys Ala Tyr Ala Met Tyr Gln Leu Leu Leu Leu
180 185 190
atc ata ttc ccg gtg ctc aca atg tcc ata tgt tat gct cgt gta tct 624
Ile Ile Phe Pro Val Leu Thr Met Ser Ile Cys Tyr Ala Arg Val Ser
195 200 205
gca ata gtg tac aag tcg tcg aaa gat cgt gta ata ctt tcc caa gct 672
Ala Ile Val Tyr Lys Ser Ser Lys Asp Arg Val Ile Leu Ser Gln Ala
210 215 220
atg gtc gca ttt tca aag gcc gcc act gac gcc gtt acc ttc tcc ggc 720
Met Val Ala Phe Ser Lys Ala Ala Thr Asp Ala Val Thr Phe Ser Gly
225 230 235 240
tac tct gca atc ccc atg att aca act tcc cgg aac ttg aaa acc gcc 768
Tyr Ser Ala Ile Pro Met Ile Thr Thr Ser Arg Asn Leu Lys Thr Ala
245 250 255
aac acg acc atc aag tcc tac agt aat cac aga aat aat cgc gtg gca 816
Asn Thr Thr Ile Lys Ser Tyr Ser Asn His Arg Asn Asn Arg Val Ala
260 265 270
gag gcg aat aag aaa cag ata gtt caa atg tta atc tcc atc gtc tgc 864
Glu Ala Asn Lys Lys Gln Ile Val Gln Met Leu Ile Ser Ile Val Cys
275 280 285
atg tac acg gtt tgc tgg ctc cca acg atc gtc gac gag ctt ctg acg 912
Met Tyr Thr Val Cys Trp Leu Pro Thr Ile Val Asp Glu Leu Leu Thr
290 295 300
tcc ttc ggg tac att tgt cga aca tca aat acg cag aca ttg aag cat 960
Ser Phe Gly Tyr Ile Cys Arg Thr Ser Asn Thr Gln Thr Leu Lys His
305 310 315 320
atg cgg atg gga ttt aat gcg ctg acc tat tgc caa tca tgc att aat 1008
Met Arg Met Gly Phe Asn Ala Leu Thr Tyr Cys Gln Ser Cys Ile Asn
325 330 335
cct atc ctg tat gcc ttc att tct cag aat ttc cgg tcc acg ttc aag 1056
Pro Ile Leu Tyr Ala Phe Ile Ser Gln Asn Phe Arg Ser Thr Phe Lys
340 345 350
acc gcc tat tcg aga atg aaa agt cgt ctt cag ggc gtg gaa gaa atg 1104
Thr Ala Tyr Ser Arg Met Lys Ser Arg Leu Gln Gly Val Glu Glu Met
355 360 365
cgc tct cgt atg gga tcg tgc tct tcg gcg agc atg atg tcg acc cgc 1152
Arg Ser Arg Met Gly Ser Cys Ser Ser Ala Ser Met Met Ser Thr Arg
370 375 380
aac cag cac cga atc tac ggg tcc agc ttc aac aca ctc aca gtg ccc 1200
Asn Gln His Arg Ile Tyr Gly Ser Ser Phe Asn Thr Leu Thr Val Pro
385 390 395 400
ggt cgc tcc atg atc act ccg aac atg tca cgc gac gtc tca cag ttg 1248
Gly Arg Ser Met Ile Thr Pro Asn Met Ser Arg Asp Val Ser Gln Leu
405 410 415
tca ttg tgc agg ccg acg tcg caa atg tcg ttc tgc cga ccg aag agt 1296
Ser Leu Cys Arg Pro Thr Ser Gln Met Ser Phe Cys Arg Pro Lys Ser
420 425 430
cca atg gcc gga gac tct ccg tcg act ttc atg aga cat cgt cct aga 1344
Pro Met Ala Gly Asp Ser Pro Ser Thr Phe Met Arg His Arg Pro Arg
›PRT · 2 of 2
435 440 445
tcg cca acc gac gtg tcc cat gat tcc gga aga cca cgt agc ccc acg 1392
Ser Pro Thr Asp Val Ser His Asp Ser Gly Arg Pro Arg Ser Pro Thr
450 455 460
gat ctt tcg cag tcg aca aag cct tcg cga cga tcc tcc tct ata aga 1440
Asp Leu Ser Gln Ser Thr Lys Pro Ser Arg Arg Ser Ser Ser Ile Arg
465 470 475 480
ccc cgt agc cca aca tcg aca tct caa atg tct acc ata gtc cga tcg 1488
Pro Arg Ser Pro Thr Ser Thr Ser Gln Met Ser Thr Ile Val Arg Ser
485 490 495
aga agt cct aca ggc gca tcg gac acc tct tca ttg ttc cca tca aga 1536
Arg Ser Pro Thr Gly Ala Ser Asp Thr Ser Ser Leu Phe Pro Ser Arg
500 505 510
aca aga agt cca act ctc caa tca aac aca tct ggt caa tcc atg gaa 1584
Thr Arg Ser Pro Thr Leu Gln Ser Asn Thr Ser Gly Gln Ser Met Glu
515 520 525
cga acc acg gac cga ctg agt gtc cgt gac gct gtc cga cca aaa aca 1632
Arg Thr Thr Asp Arg Leu Ser Val Arg Asp Ala Val Arg Pro Lys Thr
530 535 540
ccg cca gca gtg gtt gtc ta 1652
Pro Pro Ala Val Val Val
545 550
18
550
›PRT
Homo sapiens
18
Met Thr Asp Glu Glu Leu Phe Glu Ile Ala Leu Pro Gly Phe Leu Tyr
1 5 10 15
Leu Thr Val Phe Leu Val Gly Thr Ile Gly Asn Ser Met Val Ile Phe
20 25 30
Val Val Asn Arg Phe Lys Arg Met Arg Asn Val Thr Asn Ile Phe Leu
35 40 45
Ala Ser Leu Ser Thr Ala Asp Leu Cys Leu Ile Trp Phe Cys Val Pro
50 55 60
Ile Met Phe Met Lys Tyr Met Ser His Thr Trp Ser Met Gly Arg Phe
65 70 75 80
Ala Cys Tyr Ser Val His Tyr Ile Gln Gln Phe Thr Cys Phe Cys Ser
85 90 95
Val Leu Thr Met Thr Met Ile Ser Phe Glu Arg Phe Leu Ala Ile Ala
100 105 110
Tyr Pro Met Arg Asn Ile Trp Phe Ser Ser Ile Gly Arg Ala Lys Lys
115 120 125
Val Ile Leu Leu Ile Trp Met Ser Ser Ala Val Leu Ala Val Pro Thr
130 135 140
Ala Val Arg Met Asp Tyr Glu Thr Asn Leu Ser Leu Ser Gly Gln Arg
145 150 155 160
Val His Trp Cys Arg Arg Arg Phe Pro Ala Gln Phe Leu Gly Tyr Pro
165 170 175
Arg Thr Ser Leu Asn Lys Ala Tyr Ala Met Tyr Gln Leu Leu Leu Leu
180 185 190
Ile Ile Phe Pro Val Leu Thr Met Ser Ile Cys Tyr Ala Arg Val Ser
195 200 205
Ala Ile Val Tyr Lys Ser Ser Lys Asp Arg Val Ile Leu Ser Gln Ala
210 215 220
Met Val Ala Phe Ser Lys Ala Ala Thr Asp Ala Val Thr Phe Ser Gly
225 230 235 240
Tyr Ser Ala Ile Pro Met Ile Thr Thr Ser Arg Asn Leu Lys Thr Ala
245 250 255
Asn Thr Thr Ile Lys Ser Tyr Ser Asn His Arg Asn Asn Arg Val Ala
260 265 270
Glu Ala Asn Lys Lys Gln Ile Val Gln Met Leu Ile Ser Ile Val Cys
275 280 285
Met Tyr Thr Val Cys Trp Leu Pro Thr Ile Val Asp Glu Leu Leu Thr
290 295 300
Ser Phe Gly Tyr Ile Cys Arg Thr Ser Asn Thr Gln Thr Leu Lys His
305 310 315 320
Met Arg Met Gly Phe Asn Ala Leu Thr Tyr Cys Gln Ser Cys Ile Asn
325 330 335
Pro Ile Leu Tyr Ala Phe Ile Ser Gln Asn Phe Arg Ser Thr Phe Lys
340 345 350
Thr Ala Tyr Ser Arg Met Lys Ser Arg Leu Gln Gly Val Glu Glu Met
355 360 365
Arg Ser Arg Met Gly Ser Cys Ser Ser Ala Ser Met Met Ser Thr Arg
370 375 380
Asn Gln His Arg Ile Tyr Gly Ser Ser Phe Asn Thr Leu Thr Val Pro
385 390 395 400
Gly Arg Ser Met Ile Thr Pro Asn Met Ser Arg Asp Val Ser Gln Leu
405 410 415
Ser Leu Cys Arg Pro Thr Ser Gln Met Ser Phe Cys Arg Pro Lys Ser
420 425 430
Pro Met Ala Gly Asp Ser Pro Ser Thr Phe Met Arg His Arg Pro Arg
435 440 445
Ser Pro Thr Asp Val Ser His Asp Ser Gly Arg Pro Arg Ser Pro Thr
450 455 460
Asp Leu Ser Gln Ser Thr Lys Pro Ser Arg Arg Ser Ser Ser Ile Arg
465 470 475 480
Pro Arg Ser Pro Thr Ser Thr Ser Gln Met Ser Thr Ile Val Arg Ser
485 490 495
Arg Ser Pro Thr Gly Ala Ser Asp Thr Ser Ser Leu Phe Pro Ser Arg
500 505 510
Thr Arg Ser Pro Thr Leu Gln Ser Asn Thr Ser Gly Gln Ser Met Glu
515 520 525
Arg Thr Thr Asp Arg Leu Ser Val Arg Asp Ala Val Arg Pro Lys Thr
530 535 540
Pro Pro Ala Val Val Val
545 550
19
0
›DNA
synthetic
19
000
20
0
›PRT
synthetic
20
000
21
atg aat ggc tcc gat tgt ctg aat ctc aac tca gaa tta tgg ttg tat 48
Met Asn Gly Ser Asp Cys Leu Asn Leu Asn Ser Glu Leu Trp Leu Tyr
1 5 10 15
cga gaa gat ttg tca tca agg tgg tac ata atg tta gtg ttt gca ttt 96
Arg Glu Asp Leu Ser Ser Arg Trp Tyr Ile Met Leu Val Phe Ala Phe
20 25 30
ctc tac ctg ata atc att gcc gcc gga ata att gga aac tca tgt gtg 144
Leu Tyr Leu Ile Ile Ile Ala Ala Gly Ile Ile Gly Asn Ser Cys Val
35 40 45
att ttg gca atc aca agg aac aaa tca ctt caa act gtt ccg aat ctg 192
Ile Leu Ala Ile Thr Arg Asn Lys Ser Leu Gln Thr Val Pro Asn Leu
50 55 60
ttt att ctt tct tta tca tgt tct gat att gtg gta tgc tgc aca tct 240
Phe Ile Leu Ser Leu Ser Cys Ser Asp Ile Val Val Cys Cys Thr Ser
65 70 75 80
gca aca atc act ccg att act gca ttc aag aaa gaa tgg atc ttt gga 288
Ala Thr Ile Thr Pro Ile Thr Ala Phe Lys Lys Glu Trp Ile Phe Gly
85 90 95
gag gct tta tgc cga att gca cca ttc att gct ggt atc agc ctt tgt 336
Glu Ala Leu Cys Arg Ile Ala Pro Phe Ile Ala Gly Ile Ser Leu Cys
100 105 110
ttc tca act ttc aca ttg act gca atc tcc atc gac aga tac atc ctg 384
Phe Ser Thr Phe Thr Leu Thr Ala Ile Ser Ile Asp Arg Tyr Ile Leu
115 120 125
att cga ttt ccg atg agg aaa cct att acg cat tat caa gcg gtt gga 432
Ile Arg Phe Pro Met Arg Lys Pro Ile Thr His Tyr Gln Ala Val Gly
130 135 140
gtg att gct att att tgc gct ttt gct gca acc atc aca tcc cca ata 480
Val Ile Ala Ile Ile Cys Ala Phe Ala Ala Thr Ile Thr Ser Pro Ile
145 150 155 160
atg ttc aag caa aag ctg gga gag ttt gag aat ttt tgt ggg cag tac 528
Met Phe Lys Gln Lys Leu Gly Glu Phe Glu Asn Phe Cys Gly Gln Tyr
165 170 175
tgc acg gaa aac tgg gga gcc aat gaa agc cag aga aaa att tat ggt 576
Cys Thr Glu Asn Trp Gly Ala Asn Glu Ser Gln Arg Lys Ile Tyr Gly
180 185 190
gca gct ctg atg ttt ctt cag ctc gtc att ccg ctt acc atc atc atc 624
Ala Ala Leu Met Phe Leu Gln Leu Val Ile Pro Leu Thr Ile Ile Ile
195 200 205
ata tcc tac act gcg att tct ttg aag atc gga caa agc atg att ctc 672
Ile Ser Tyr Thr Ala Ile Ser Leu Lys Ile Gly Gln Ser Met Ile Leu
210 215 220
aaa ggg gcg aaa aag caa aaa aca gac aat tgg gaa atg gaa tta agt 720
Lys Gly Ala Lys Lys Gln Lys Thr Asp Asn Trp Glu Met Glu Leu Ser
225 230 235 240
gat caa caa aga atc gct gtg aag aga aga caa aga act aat aga atg 768
Asp Gln Gln Arg Ile Ala Val Lys Arg Arg Gln Arg Thr Asn Arg Met
245 250 255
ctt att ggt atg gta gtc gca ttc gct tgc agc tgg att tgg tca gtg 816
Leu Ile Gly Met Val Val Ala Phe Ala Cys Ser Trp Ile Trp Ser Val
260 265 270
acg ttc aac att ctg agg gac tat gaa tat ttg cca gag ctc atc aaa 864
Thr Phe Asn Ile Leu Arg Asp Tyr Glu Tyr Leu Pro Glu Leu Ile Lys
275 280 285
act caa gaa tat atc ttt gga att gct aca cat tgc att gca atg acc 912
Thr Gln Glu Tyr Ile Phe Gly Ile Ala Thr His Cys Ile Ala Met Thr
290 295 300
tca acg gta tgg aac ccg tta ctc tac gca gtg ctc aac ctc caa ctg 960
Ser Thr Val Trp Asn Pro Leu Leu Tyr Ala Val Leu Asn Leu Gln Leu
305 310 315 320
cgt gca gca ttc att gac ctg atg cct cac tgg ctt cgt cgt cat tta 1008
Arg Ala Ala Phe Ile Asp Leu Met Pro His Trp Leu Arg Arg His Leu
325 330 335
aac ctg gaa gga gac aac agc tct cca ttg ctc aac cat ccg acg atg 1056
Asn Leu Glu Gly Asp Asn Ser Ser Pro Leu Leu Asn His Pro Thr Met
340 345 350
aca att aca aac aaa ccg tca aag caa cat aca tta ata cca gca atg 1104
Thr Ile Thr Asn Lys Pro Ser Lys Gln His Thr Leu Ile Pro Ala Met
355 360 365
gac aac cat acg tgt caa caa gtt ta 1130
Asp Asn His Thr Cys Gln Gln Val
370 375
22
376
›PRT
Homo sapiens
22
Met Asn Gly Ser Asp Cys Leu Asn Leu Asn Ser Glu Leu Trp Leu Tyr
1 5 10 15
Arg Glu Asp Leu Ser Ser Arg Trp Tyr Ile Met Leu Val Phe Ala Phe
20 25 30
Leu Tyr Leu Ile Ile Ile Ala Ala Gly Ile Ile Gly Asn Ser Cys Val
35 40 45
Ile Leu Ala Ile Thr Arg Asn Lys Ser Leu Gln Thr Val Pro Asn Leu
50 55 60
Phe Ile Leu Ser Leu Ser Cys Ser Asp Ile Val Val Cys Cys Thr Ser
65 70 75 80
Ala Thr Ile Thr Pro Ile Thr Ala Phe Lys Lys Glu Trp Ile Phe Gly
85 90 95
Glu Ala Leu Cys Arg Ile Ala Pro Phe Ile Ala Gly Ile Ser Leu Cys
100 105 110
Phe Ser Thr Phe Thr Leu Thr Ala Ile Ser Ile Asp Arg Tyr Ile Leu
115 120 125
Ile Arg Phe Pro Met Arg Lys Pro Ile Thr His Tyr Gln Ala Val Gly
130 135 140
Val Ile Ala Ile Ile Cys Ala Phe Ala Ala Thr Ile Thr Ser Pro Ile
145 150 155 160
Met Phe Lys Gln Lys Leu Gly Glu Phe Glu Asn Phe Cys Gly Gln Tyr
165 170 175
Cys Thr Glu Asn Trp Gly Ala Asn Glu Ser Gln Arg Lys Ile Tyr Gly
180 185 190
Ala Ala Leu Met Phe Leu Gln Leu Val Ile Pro Leu Thr Ile Ile Ile
195 200 205
Ile Ser Tyr Thr Ala Ile Ser Leu Lys Ile Gly Gln Ser Met Ile Leu
210 215 220
Lys Gly Ala Lys Lys Gln Lys Thr Asp Asn Trp Glu Met Glu Leu Ser
225 230 235 240
Asp Gln Gln Arg Ile Ala Val Lys Arg Arg Gln Arg Thr Asn Arg Met
245 250 255
Leu Ile Gly Met Val Val Ala Phe Ala Cys Ser Trp Ile Trp Ser Val
260 265 270
Thr Phe Asn Ile Leu Arg Asp Tyr Glu Tyr Leu Pro Glu Leu Ile Lys
275 280 285
Thr Gln Glu Tyr Ile Phe Gly Ile Ala Thr His Cys Ile Ala Met Thr
290 295 300
Ser Thr Val Trp Asn Pro Leu Leu Tyr Ala Val Leu Asn Leu Gln Leu
305 310 315 320
Arg Ala Ala Phe Ile Asp Leu Met Pro His Trp Leu Arg Arg His Leu
325 330 335
Asn Leu Glu Gly Asp Asn Ser Ser Pro Leu Leu Asn His Pro Thr Met
340 345 350
Thr Ile Thr Asn Lys Pro Ser Lys Gln His Thr Leu Ile Pro Ala Met
355 360 365
Asp Asn His Thr Cys Gln Gln Val
370 375
23
atg att att ttt tat ctt tac gta gca act caa gta ttt gtg gca att 48
Met Ile Ile Phe Tyr Leu Tyr Val Ala Thr Gln Val Phe Val Ala Ile
1 5 10 15
gca ttt gtc ctt ctg atg gca act gca ata att gga aac tct gtc gtg 96
Ala Phe Val Leu Leu Met Ala Thr Ala Ile Ile Gly Asn Ser Val Val
20 25 30
atg tgg ata att tat caa cac aaa gta atg cac tac ggc ttc aac tat 144
Met Trp Ile Ile Tyr Gln His Lys Val Met His Tyr Gly Phe Asn Tyr
35 40 45
ttt ctt ttc aat atg gca ttc gct gat ctt ttg att gct ctc ttc aat 192
Phe Leu Phe Asn Met Ala Phe Ala Asp Leu Leu Ile Ala Leu Phe Asn
50 55 60
gtt ggt act tca tgg acc tat aat tta tac tat gac tgg tgg tat ggt 240
Val Gly Thr Ser Trp Thr Tyr Asn Leu Tyr Tyr Asp Trp Trp Tyr Gly
65 70 75 80
gat cta tgt aca ctt act tcc ttc ttc ggt att gca cca act acc gta 288
Asp Leu Cys Thr Leu Thr Ser Phe Phe Gly Ile Ala Pro Thr Thr Val
85 90 95
tcc gtg tgc tca atg atg gct ctc agc tgg gat aga tgt caa gca gtc 336
Ser Val Cys Ser Met Met Ala Leu Ser Trp Asp Arg Cys Gln Ala Val
100 105 110
gtg aat cct cta caa aaa cgt cca cta tct cga aaa aga tcc gtc att 384
Val Asn Pro Leu Gln Lys Arg Pro Leu Ser Arg Lys Arg Ser Val Ile
115 120 125
gct att ctc atc att tgg gtt gtt tca acg gtt act gca ctt ccg ttt 432
Ala Ile Leu Ile Ile Trp Val Val Ser Thr Val Thr Ala Leu Pro Phe
130 135 140
gca att gct gca tct gtc aac tct ctc tac aca tat gac gtg gtt aca 480
Ala Ile Ala Ala Ser Val Asn Ser Leu Tyr Thr Tyr Asp Val Val Thr
145 150 155 160
tca act gtt tca aaa gct cac gtt tgt tca gca cca gtt aat act ttc 528
Ser Thr Val Ser Lys Ala His Val Cys Ser Ala Pro Val Asn Thr Phe
165 170 175
ttt gaa aaa gtg ctc ttt ggg att caa tat gct cta cct ata atc att 576
Phe Glu Lys Val Leu Phe Gly Ile Gln Tyr Ala Leu Pro Ile Ile Ile
180 185 190
ttg gga tca acg ttc aca cgg att gcc gtt gca ttt cga gca aca aat 624
Leu Gly Ser Thr Phe Thr Arg Ile Ala Val Ala Phe Arg Ala Thr Asn
195 200 205
gaa gcc act gac agt agt ctg aag aac aat cac aca cgc gca aag agc 672
Glu Ala Thr Asp Ser Ser Leu Lys Asn Asn His Thr Arg Ala Lys Ser
210 215 220
aag gca gta aaa atg ctc ttc cta atg gtc gtt gca ttc gtt gtc tgc 720
Lys Ala Val Lys Met Leu Phe Leu Met Val Val Ala Phe Val Val Cys
225 230 235 240
tgg ctt cca tac cac att tat cac gca ttc gct ctt gaa gaa ttc ttc 768
Trp Leu Pro Tyr His Ile Tyr His Ala Phe Ala Leu Glu Glu Phe Phe
245 250 255
gat gca gct cgt gga aaa tat gcg tat ttg ttg att tac tgg att gca 816
Asp Ala Ala Arg Gly Lys Tyr Ala Tyr Leu Leu Ile Tyr Trp Ile Ala
260 265 270
atg tct agt tgt gca tac aat cca att att tat tgt ttt gcc aac gaa 864
Met Ser Ser Cys Ala Tyr Asn Pro Ile Ile Tyr Cys Phe Ala Asn Glu
275 280 285
aga ttc cgc att ggc ttc cgg tac gtc ttc cga tgg att cca gtg att 912
Arg Phe Arg Ile Gly Phe Arg Tyr Val Phe Arg Trp Ile Pro Val Ile
290 295 300
gac tgc aaa aag gaa caa tat gaa tat tca caa tta ttc ccg gat aaa 960
Asp Cys Lys Lys Glu Gln Tyr Glu Tyr Ser Gln Leu Phe Pro Asp Lys
305 310 315 320
atg aga agt atg gct att tct ttg caa aaa gga aga gtg aac tct tcc 1008
Met Arg Ser Met Ala Ile Ser Leu Gln Lys Gly Arg Val Asn Ser Ser
325 330 335
tgt ttg gat aaa aaa gtc aaa gag aat tct tct caa gat ttg gtt tgt 1056
Cys Leu Asp Lys Lys Val Lys Glu Asn Ser Ser Gln Asp Leu Val Cys
340 345 350
gtg atg cat tcc gag aaa aat acg aag aaa tat tca aaa gta cat ctt 1104
Val Met His Ser Glu Lys Asn Thr Lys Lys Tyr Ser Lys Val His Leu
355 360 365
ctg agt tgc cat gaa cgg tga 1125
Leu Ser Cys His Glu Arg
370
24
374
›PRT
Homo sapiens
24
Met Ile Ile Phe Tyr Leu Tyr Val Ala Thr Gln Val Phe Val Ala Ile
1 5 10 15
Ala Phe Val Leu Leu Met Ala Thr Ala Ile Ile Gly Asn Ser Val Val
20 25 30
Met Trp Ile Ile Tyr Gln His Lys Val Met His Tyr Gly Phe Asn Tyr
35 40 45
Phe Leu Phe Asn Met Ala Phe Ala Asp Leu Leu Ile Ala Leu Phe Asn
50 55 60
Val Gly Thr Ser Trp Thr Tyr Asn Leu Tyr Tyr Asp Trp Trp Tyr Gly
65 70 75 80
Asp Leu Cys Thr Leu Thr Ser Phe Phe Gly Ile Ala Pro Thr Thr Val
85 90 95
Ser Val Cys Ser Met Met Ala Leu Ser Trp Asp Arg Cys Gln Ala Val
100 105 110
Val Asn Pro Leu Gln Lys Arg Pro Leu Ser Arg Lys Arg Ser Val Ile
115 120 125
Ala Ile Leu Ile Ile Trp Val Val Ser Thr Val Thr Ala Leu Pro Phe
130 135 140
Ala Ile Ala Ala Ser Val Asn Ser Leu Tyr Thr Tyr Asp Val Val Thr
145 150 155 160
Ser Thr Val Ser Lys Ala His Val Cys Ser Ala Pro Val Asn Thr Phe
165 170 175
Phe Glu Lys Val Leu Phe Gly Ile Gln Tyr Ala Leu Pro Ile Ile Ile
180 185 190
Leu Gly Ser Thr Phe Thr Arg Ile Ala Val Ala Phe Arg Ala Thr Asn
195 200 205
Glu Ala Thr Asp Ser Ser Leu Lys Asn Asn His Thr Arg Ala Lys Ser
210 215 220
Lys Ala Val Lys Met Leu Phe Leu Met Val Val Ala Phe Val Val Cys
225 230 235 240
Trp Leu Pro Tyr His Ile Tyr His Ala Phe Ala Leu Glu Glu Phe Phe
245 250 255
Asp Ala Ala Arg Gly Lys Tyr Ala Tyr Leu Leu Ile Tyr Trp Ile Ala
260 265 270
Met Ser Ser Cys Ala Tyr Asn Pro Ile Ile Tyr Cys Phe Ala Asn Glu
275 280 285
Arg Phe Arg Ile Gly Phe Arg Tyr Val Phe Arg Trp Ile Pro Val Ile
290 295 300
Asp Cys Lys Lys Glu Gln Tyr Glu Tyr Ser Gln Leu Phe Pro Asp Lys
305 310 315 320
Met Arg Ser Met Ala Ile Ser Leu Gln Lys Gly Arg Val Asn Ser Ser
325 330 335
Cys Leu Asp Lys Lys Val Lys Glu Asn Ser Ser Gln Asp Leu Val Cys
340 345 350
Val Met His Ser Glu Lys Asn Thr Lys Lys Tyr Ser Lys Val His Leu
355 360 365
Leu Ser Cys His Glu Arg
370
25
atg gaa gtt gaa aat ttt acc gac tgt caa gta tat tgg aaa gtg tat 48
Met Glu Val Glu Asn Phe Thr Asp Cys Gln Val Tyr Trp Lys Val Tyr
1 5 10 15
cca gat cct tct caa agt ata tat gcg ata gtg cca ttt ctg acc gtg 96
Pro Asp Pro Ser Gln Ser Ile Tyr Ala Ile Val Pro Phe Leu Thr Val
20 25 30
tac ctt ttt ctc ttt ttt ctt gga ctc ttt gga aat gtg acc ttg att 144
Tyr Leu Phe Leu Phe Phe Leu Gly Leu Phe Gly Asn Val Thr Leu Ile
35 40 45
tac gta act tgc agc cat aaa gct tta ctg agc gtt caa aac ata ttc 192
Tyr Val Thr Cys Ser His Lys Ala Leu Leu Ser Val Gln Asn Ile Phe
50 55 60
att ctg aac ctg gcg gcg agc gat tgc atg atg tgc ata tta tcg ctt 240
Ile Leu Asn Leu Ala Ala Ser Asp Cys Met Met Cys Ile Leu Ser Leu
65 70 75 80
cca atc act cca atc aca aat gtg tac aaa aac tgg tac ttt gga aat 288
Pro Ile Thr Pro Ile Thr Asn Val Tyr Lys Asn Trp Tyr Phe Gly Asn
85 90 95
cta ctc tgc cat ttg ata cca tgc att caa ggt atc agc att ttc gta 336
Leu Leu Cys His Leu Ile Pro Cys Ile Gln Gly Ile Ser Ile Phe Val
100 105 110
tgc aca ttc agt ctc ggt gcg att gct ttg gat cgg tat atc ctt gta 384
Cys Thr Phe Ser Leu Gly Ala Ile Ala Leu Asp Arg Tyr Ile Leu Val
115 120 125
gta aga cca cat tct aca cca cta tcc caa aga gga gca ttt ctt act 432
Val Arg Pro His Ser Thr Pro Leu Ser Gln Arg Gly Ala Phe Leu Thr
130 135 140
act gtt cta ttg tgg atc ctc tct ttt gtt gta act cta ccc tat gcg 480
Thr Val Leu Leu Trp Ile Leu Ser Phe Val Val Thr Leu Pro Tyr Ala
145 150 155 160
ttc aat atg caa atg att gaa tac aca gaa gag aga ata tgc ggc tac 528
Phe Asn Met Gln Met Ile Glu Tyr Thr Glu Glu Arg Ile Cys Gly Tyr
165 170 175
ttt tgc act gaa aag tgg gaa tct gcc aag tct aga aga gcc tac aca 576
Phe Cys Thr Glu Lys Trp Glu Ser Ala Lys Ser Arg Arg Ala Tyr Thr
180 185 190
atg atc gtg atg ctc gcc caa ttc gtg gtc cca ttc gct gtc atg gcc 624
Met Ile Val Met Leu Ala Gln Phe Val Val Pro Phe Ala Val Met Ala
195 200 205
ttc tgc tat gca aat att gtt tcg gtg ctc agc aag cgt gct cag aca 672
Phe Cys Tyr Ala Asn Ile Val Ser Val Leu Ser Lys Arg Ala Gln Thr
210 215 220
aag ata cgc aaa atg gtg gag aga aca agc gcg ttg gag agc tct tgc 720
Lys Ile Arg Lys Met Val Glu Arg Thr Ser Ala Leu Glu Ser Ser Cys
225 230 235 240
gca ttc ccg tcg cat ggt ctt gaa cag tat gaa aat gag ttg aac gaa 768
Ala Phe Pro Ser His Gly Leu Glu Gln Tyr Glu Asn Glu Leu Asn Glu
245 250 255
ttt cta gac aaa cag gaa aag gag aaa caa cga gtt gta ctt cag aac 816
Phe Leu Asp Lys Gln Glu Lys Glu Lys Gln Arg Val Val Leu Gln Asn
260 265 270
aga aga aca acg tca atc cta gtt acc atg gtt gtc tgg ttt ggg ata 864
Arg Arg Thr Thr Ser Ile Leu Val Thr Met Val Val Trp Phe Gly Ile
275 280 285
act tgg ctg cca cat aac gtc att tct ttg att att gaa tat gat gac 912
Thr Trp Leu Pro His Asn Val Ile Ser Leu Ile Ile Glu Tyr Asp Asp
290 295 300
aca caa tcg ttt ttc cga ctt tat ggc aga gat gat tac gat atc agt 960
Thr Gln Ser Phe Phe Arg Leu Tyr Gly Arg Asp Asp Tyr Asp Ile Ser
305 310 315 320
tat tta ctg aac ctt ttc act cac agt att gcc atg tcg aac aat gtt 1008
Tyr Leu Leu Asn Leu Phe Thr His Ser Ile Ala Met Ser Asn Asn Val
325 330 335
tta aac ccg gta ctc tat gcg tgg ctg aac cca agt ttc cgt caa ctg 1056
Leu Asn Pro Val Leu Tyr Ala Trp Leu Asn Pro Ser Phe Arg Gln Leu
340 345 350
gtc ata aag aca tat ttt gga gac cgg cgt aaa agt gac aga ata atc 1104
Val Ile Lys Thr Tyr Phe Gly Asp Arg Arg Lys Ser Asp Arg Ile Ile
355 360 365
aat caa aca tca gtt tac aaa aca aag atc gtg cat gat acg aag cac 1152
Asn Gln Thr Ser Val Tyr Lys Thr Lys Ile Val His Asp Thr Lys His
370 375 380
ttg aat gga aga gcc aaa att ggc ggt ggt ggt agc cac gag gcg ttg 1200
Leu Asn Gly Arg Ala Lys Ile Gly Gly Gly Gly Ser His Glu Ala Leu
385 390 395 400
aag gag agg gag ctg aac tcg tgt tcg gaa aat tta agc tat cat gtg 1248
Lys Glu Arg Glu Leu Asn Ser Cys Ser Glu Asn Leu Ser Tyr His Val
405 410 415
aac ggt cat acg agg act cct aca ccg gaa gtt cag ttg aat gaa gtt 1296
Asn Gly His Thr Arg Thr Pro Thr Pro Glu Val Gln Leu Asn Glu Val
420 425 430
tca agt cct gaa ata agt aaa ctt gtt gct gag ccg gaa gaa ttg atc 1344
Ser Ser Pro Glu Ile Ser Lys Leu Val Ala Glu Pro Glu Glu Leu Ile
435 440 445
gag ttc agc gtc aac gac acg cta gtc tga 1374
Glu Phe Ser Val Asn Asp Thr Leu Val
450 455
26
457
›PRT
Homo sapiens
26
Met Glu Val Glu Asn Phe Thr Asp Cys Gln Val Tyr Trp Lys Val Tyr
1 5 10 15
Pro Asp Pro Ser Gln Ser Ile Tyr Ala Ile Val Pro Phe Leu Thr Val
20 25 30
Tyr Leu Phe Leu Phe Phe Leu Gly Leu Phe Gly Asn Val Thr Leu Ile
35 40 45
Tyr Val Thr Cys Ser His Lys Ala Leu Leu Ser Val Gln Asn Ile Phe
50 55 60
Ile Leu Asn Leu Ala Ala Ser Asp Cys Met Met Cys Ile Leu Ser Leu
65 70 75 80
Pro Ile Thr Pro Ile Thr Asn Val Tyr Lys Asn Trp Tyr Phe Gly Asn
85 90 95
Leu Leu Cys His Leu Ile Pro Cys Ile Gln Gly Ile Ser Ile Phe Val
100 105 110
Cys Thr Phe Ser Leu Gly Ala Ile Ala Leu Asp Arg Tyr Ile Leu Val
115 120 125
Val Arg Pro His Ser Thr Pro Leu Ser Gln Arg Gly Ala Phe Leu Thr
130 135 140
Thr Val Leu Leu Trp Ile Leu Ser Phe Val Val Thr Leu Pro Tyr Ala
145 150 155 160
Phe Asn Met Gln Met Ile Glu Tyr Thr Glu Glu Arg Ile Cys Gly Tyr
165 170 175
Phe Cys Thr Glu Lys Trp Glu Ser Ala Lys Ser Arg Arg Ala Tyr Thr
180 185 190
Met Ile Val Met Leu Ala Gln Phe Val Val Pro Phe Ala Val Met Ala
195 200 205
Phe Cys Tyr Ala Asn Ile Val Ser Val Leu Ser Lys Arg Ala Gln Thr
210 215 220
Lys Ile Arg Lys Met Val Glu Arg Thr Ser Ala Leu Glu Ser Ser Cys
225 230 235 240
Ala Phe Pro Ser His Gly Leu Glu Gln Tyr Glu Asn Glu Leu Asn Glu
245 250 255
Phe Leu Asp Lys Gln Glu Lys Glu Lys Gln Arg Val Val Leu Gln Asn
260 265 270
Arg Arg Thr Thr Ser Ile Leu Val Thr Met Val Val Trp Phe Gly Ile
275 280 285
Thr Trp Leu Pro His Asn Val Ile Ser Leu Ile Ile Glu Tyr Asp Asp
290 295 300
Thr Gln Ser Phe Phe Arg Leu Tyr Gly Arg Asp Asp Tyr Asp Ile Ser
305 310 315 320
Tyr Leu Leu Asn Leu Phe Thr His Ser Ile Ala Met Ser Asn Asn Val
325 330 335
Leu Asn Pro Val Leu Tyr Ala Trp Leu Asn Pro Ser Phe Arg Gln Leu
340 345 350
Val Ile Lys Thr Tyr Phe Gly Asp Arg Arg Lys Ser Asp Arg Ile Ile
355 360 365
Asn Gln Thr Ser Val Tyr Lys Thr Lys Ile Val His Asp Thr Lys His
370 375 380
Leu Asn Gly Arg Ala Lys Ile Gly Gly Gly Gly Ser His Glu Ala Leu
385 390 395 400
Lys Glu Arg Glu Leu Asn Ser Cys Ser Glu Asn Leu Ser Tyr His Val
405 410 415
Asn Gly His Thr Arg Thr Pro Thr Pro Glu Val Gln Leu Asn Glu Val
420 425 430
Ser Ser Pro Glu Ile Ser Lys Leu Val Ala Glu Pro Glu Glu Leu Ile
435 440 445
Glu Phe Ser Val Asn Asp Thr Leu Val
450 455
27
atg aat acg tca ttt gta gag cca tta tat gca gat gtt gaa caa ctt 48
Met Asn Thr Ser Phe Val Glu Pro Leu Tyr Ala Asp Val Glu Gln Leu
1 5 10 15
gaa cca gtc cct cta ctt aga cat tca tat caa cta act gtt ctt tat 96
Glu Pro Val Pro Leu Leu Arg His Ser Tyr Gln Leu Thr Val Leu Tyr
20 25 30
aca gtt gca tat gga gct gta ttc ttc act ggt gta ctt gga aac acg 144
Thr Val Ala Tyr Gly Ala Val Phe Phe Thr Gly Val Leu Gly Asn Thr
35 40 45
ttt gtt gtc tta gcg gtt tgg gct cat aag aat tta aat ata acc acg 192
Phe Val Val Leu Ala Val Trp Ala His Lys Asn Leu Asn Ile Thr Thr
50 55 60
gat tat cta att ttg tct ctt gca ctt gct gat ctg ttt att tta tgg 240
Asp Tyr Leu Ile Leu Ser Leu Ala Leu Ala Asp Leu Phe Ile Leu Trp
65 70 75 80
att tgt ctg cca act acg ttg att aat agc att ttc aca gaa tgg ctt 288
Ile Cys Leu Pro Thr Thr Leu Ile Asn Ser Ile Phe Thr Glu Trp Leu
85 90 95
tgg ggt caa ttt ttc tgc cga ttg tcc aca tgg gct aac gca tct acg 336
Trp Gly Gln Phe Phe Cys Arg Leu Ser Thr Trp Ala Asn Ala Ser Thr
100 105 110
tca ttt gca tca gtt tac acc ttg gtt gca gtg acg gct gat cgt tat 384
Ser Phe Ala Ser Val Tyr Thr Leu Val Ala Val Thr Ala Asp Arg Tyr
115 120 125
cta gct att tgc cat acg ttg aaa tac aac act agc tgg gat cga gaa 432
Leu Ala Ile Cys His Thr Leu Lys Tyr Asn Thr Ser Trp Asp Arg Glu
130 135 140
tat aca aaa tat gtt ata ttt gct gtt tgg cta gta gct gca ata ttc 480
Tyr Thr Lys Tyr Val Ile Phe Ala Val Trp Leu Val Ala Ala Ile Phe
145 150 155 160
gga ata cct aat tgg tat aac tat gat ttg ata gta tgg caa gaa ggc 528
Gly Ile Pro Asn Trp Tyr Asn Tyr Asp Leu Ile Val Trp Gln Glu Gly
165 170 175
agt tat ggt tac cga tta tgt acg tca caa acg gat caa aaa tta tat 576
Ser Tyr Gly Tyr Arg Leu Cys Thr Ser Gln Thr Asp Gln Lys Leu Tyr
180 185 190
ttt tta ttt gtt aac tta ttg ctg gct ttc ata gtt cca ttt ggt ttg 624
Phe Leu Phe Val Asn Leu Leu Leu Ala Phe Ile Val Pro Phe Gly Leu
195 200 205
att tcg ggt cta tac acg aga ata ttt atc act gta tca aca cat aga 672
Ile Ser Gly Leu Tyr Thr Arg Ile Phe Ile Thr Val Ser Thr His Arg
210 215 220
agt ctg gca gtt gat gca aga gcc cga gaa gat cga gta aaa cta cga 720
Ser Leu Ala Val Asp Ala Arg Ala Arg Glu Asp Arg Val Lys Leu Arg
225 230 235 240
gtt gcc aca atg atg tta aca gta ata att gta ttt gct tgt tgc tgg 768
Val Ala Thr Met Met Leu Thr Val Ile Ile Val Phe Ala Cys Cys Trp
245 250 255
tta cct ctc tat tgc atc ttc aca tat ttt ttc ttc ttt gct gat cag 816
Leu Pro Leu Tyr Cys Ile Phe Thr Tyr Phe Phe Phe Phe Ala Asp Gln
260 265 270
cga tcg gat ctt ttt caa att act tca atg ttg att cgg cca ata ttc 864
Arg Ser Asp Leu Phe Gln Ile Thr Ser Met Leu Ile Arg Pro Ile Phe
275 280 285
caa tgg atg tca tta cta tcg agt tct tta aat cca ata att tat att 912
Gln Trp Met Ser Leu Leu Ser Ser Ser Leu Asn Pro Ile Ile Tyr Ile
290 295 300
gca tac agt cac aag tac aga aga gct ttc aaa agt ata tta ctg atg 960
Ala Tyr Ser His Lys Tyr Arg Arg Ala Phe Lys Ser Ile Leu Leu Met
305 310 315 320
cca tgt aag aca aga tac gaa aga gtc cga agc aca ata ctc cgt cgt 1008
Pro Cys Lys Thr Arg Tyr Glu Arg Val Arg Ser Thr Ile Leu Arg Arg
325 330 335
cac tct cgt ggc ttc aaa tca act gca aca att tca atg tct aat ttt 1056
His Ser Arg Gly Phe Lys Ser Thr Ala Thr Ile Ser Met Ser Asn Phe
340 345 350
gga act gaa cca aca aat tta ggt gga gca tct agt tta ctt att gaa 1104
Gly Thr Glu Pro Thr Asn Leu Gly Gly Ala Ser Ser Leu Leu Ile Glu
355 360 365
cta gat gga aaa cag gtg gaa aga tct act tct gat tgt taa 1146
Leu Asp Gly Lys Gln Val Glu Arg Ser Thr Ser Asp Cys
370 375 380
28
381
›PRT · 1 of 2
Homo sapiens
28
Met Asn Thr Ser Phe Val Glu Pro Leu Tyr Ala Asp Val Glu Gln Leu
1 5 10 15
Glu Pro Val Pro Leu Leu Arg His Ser Tyr Gln Leu Thr Val Leu Tyr
20 25 30
Thr Val Ala Tyr Gly Ala Val Phe Phe Thr Gly Val Leu Gly Asn Thr
35 40 45
Phe Val Val Leu Ala Val Trp Ala His Lys Asn Leu Asn Ile Thr Thr
50 55 60
Asp Tyr Leu Ile Leu Ser Leu Ala Leu Ala Asp Leu Phe Ile Leu Trp
65 70 75 80
Ile Cys Leu Pro Thr Thr Leu Ile Asn Ser Ile Phe Thr Glu Trp Leu
85 90 95
Trp Gly Gln Phe Phe Cys Arg Leu Ser Thr Trp Ala Asn Ala Ser Thr
100 105 110
Ser Phe Ala Ser Val Tyr Thr Leu Val Ala Val Thr Ala Asp Arg Tyr
115 120 125
Leu Ala Ile Cys His Thr Leu Lys Tyr Asn Thr Ser Trp Asp Arg Glu
130 135 140
Tyr Thr Lys Tyr Val Ile Phe Ala Val Trp Leu Val Ala Ala Ile Phe
145 150 155 160
Gly Ile Pro Asn Trp Tyr Asn Tyr Asp Leu Ile Val Trp Gln Glu Gly
165 170 175
Ser Tyr Gly Tyr Arg Leu Cys Thr Ser Gln Thr Asp Gln Lys Leu Tyr
180 185 190
Phe Leu Phe Val Asn Leu Leu Leu Ala Phe Ile Val Pro Phe Gly Leu
195 200 205
Ile Ser Gly Leu Tyr Thr Arg Ile Phe Ile Thr Val Ser Thr His Arg
210 215 220
Ser Leu Ala Val Asp Ala Arg Ala Arg Glu Asp Arg Val Lys Leu Arg
225 230 235 240
Val Ala Thr Met Met Leu Thr Val Ile Ile Val Phe Ala Cys Cys Trp
245 250 255
Leu Pro Leu Tyr Cys Ile Phe Thr Tyr Phe Phe Phe Phe Ala Asp Gln
260 265 270
Arg Ser Asp Leu Phe Gln Ile Thr Ser Met Leu Ile Arg Pro Ile Phe
275 280 285
Gln Trp Met Ser Leu Leu Ser Ser Ser Leu Asn Pro Ile Ile Tyr Ile
290 295 300
Ala Tyr Ser His Lys Tyr Arg Arg Ala Phe Lys Ser Ile Leu Leu Met
305 310 315 320
Pro Cys Lys Thr Arg Tyr Glu Arg Val Arg Ser Thr Ile Leu Arg Arg
325 330 335
His Ser Arg Gly Phe Lys Ser Thr Ala Thr Ile Ser Met Ser Asn Phe
340 345 350
Gly Thr Glu Pro Thr Asn Leu Gly Gly Ala Ser Ser Leu Leu Ile Glu
355 360 365
Leu Asp Gly Lys Gln Val Glu Arg Ser Thr Ser Asp Cys
370 375 380
29
atg gaa gtg aaa gat ata gat aac tac tgt gat cgt gga atc agt ccg 48
Met Glu Val Lys Asp Ile Asp Asn Tyr Cys Asp Arg Gly Ile Ser Pro
1 5 10 15
aat gca tcc aat tat ctc acg tac cca ttt gac ggg ctc tgt cta cag 96
Asn Ala Ser Asn Tyr Leu Thr Tyr Pro Phe Asp Gly Leu Cys Leu Gln
20 25 30
aaa ttt ttt tat caa ctc caa act tct ttg cga agg ttc act cct tac 144
Lys Phe Phe Tyr Gln Leu Gln Thr Ser Leu Arg Arg Phe Thr Pro Tyr
35 40 45
gaa gaa atc att tac aca aca gtt tac atc att atc tct gta gca gct 192
Glu Glu Ile Ile Tyr Thr Thr Val Tyr Ile Ile Ile Ser Val Ala Ala
50 55 60
gtt att gga aat gga ttg gtg ata atg gct gta gta cgg aaa aag aca 240
Val Ile Gly Asn Gly Leu Val Ile Met Ala Val Val Arg Lys Lys Thr
65 70 75 80
atg aga aca aac aga aat gtt ttg att tta aat ctc gcg ctt tca aac 288
Met Arg Thr Asn Arg Asn Val Leu Ile Leu Asn Leu Ala Leu Ser Asn
85 90 95
ttg ata ctc gcc atc acc aac atc cca ttt cta tgg ctt ccg tca att 336
Leu Ile Leu Ala Ile Thr Asn Ile Pro Phe Leu Trp Leu Pro Ser Ile
100 105 110
gat ttc gaa ttt ccg tac tct cga ttt ttc tgc aaa ttt gcc aat gtg 384
Asp Phe Glu Phe Pro Tyr Ser Arg Phe Phe Cys Lys Phe Ala Asn Val
115 120 125
ctt ccg ggt agt aat atc tac tgc tca act tta acc atc tcg gtg atg 432
Leu Pro Gly Ser Asn Ile Tyr Cys Ser Thr Leu Thr Ile Ser Val Met
130 135 140
gca att gat aga tat tat tcg gtg aag aaa ttg aaa att gca tca aat 480
Ala Ile Asp Arg Tyr Tyr Ser Val Lys Lys Leu Lys Ile Ala Ser Asn
145 150 155 160
cgt aaa caa tgt ttc cat gct gtt ttg gtt tca ttg gct att tgg att 528
Arg Lys Gln Cys Phe His Ala Val Leu Val Ser Leu Ala Ile Trp Ile
165 170 175
gtg tca ttc atc ctc tcg tta cct ctg ctc ttg tac tat gaa acc tca 576
Val Ser Phe Ile Leu Ser Leu Pro Leu Leu Leu Tyr Tyr Glu Thr Ser
180 185 190
atg ctc tac gtc atg aga gaa att cga gtt gtt gat caa agc ggt cag 624
Met Leu Tyr Val Met Arg Glu Ile Arg Val Val Asp Gln Ser Gly Gln
195 200 205
gaa gtc atc cga agc tac gga tgg aga cag tgc cgt cta gtt tct gcc 672
Glu Val Ile Arg Ser Tyr Gly Trp Arg Gln Cys Arg Leu Val Ser Ala
210 215 220
gga cga tta ccg gac atc acc caa agc atc cag ttg ctc atg tct att 720
Gly Arg Leu Pro Asp Ile Thr Gln Ser Ile Gln Leu Leu Met Ser Ile
225 230 235 240
ctt caa gtc gct ttc cta tac atc gtt cca ctc ttt gtt ctt tca atc 768
Leu Gln Val Ala Phe Leu Tyr Ile Val Pro Leu Phe Val Leu Ser Ile
245 250 255
ttc aac gtg aaa ctc acc cgg ttt tta aaa aca aat gcc aac aaa atg 816
Phe Asn Val Lys Leu Thr Arg Phe Leu Lys Thr Asn Ala Asn Lys Met
260 265 270
agc aaa act cgt gct ccg cca aaa cga ttt gac aga tcc gat agc cac 864
Ser Lys Thr Arg Ala Pro Pro Lys Arg Phe Asp Arg Ser Asp Ser His
275 280 285
cat aat tcg ttg aaa aat aac aac aac cac acg tct tct cta cgt tcg 912
His Asn Ser Leu Lys Asn Asn Asn Asn His Thr Ser Ser Leu Arg Ser
290 295 300
cca tca atg ccc tca atc aga agt tcg ata acg gag aga aac aag acg 960
Pro Ser Met Pro Ser Ile Arg Ser Ser Ile Thr Glu Arg Asn Lys Thr
305 310 315 320
aat cag aga aca aac aga act act tcg tta ctg att gca atg gcc gga 1008
Asn Gln Arg Thr Asn Arg Thr Thr Ser Leu Leu Ile Ala Met Ala Gly
325 330 335
agc tat gcg gct ctt tgg ttt cca ttc act ctc atc act ttt ttg ata 1056
Ser Tyr Ala Ala Leu Trp Phe Pro Phe Thr Leu Ile Thr Phe Leu Ile
340 345 350
gac ttc gag ttg atc atc aat caa gac tac gtg aac ttg gta gaa cga 1104
Asp Phe Glu Leu Ile Ile Asn Gln Asp Tyr Val Asn Leu Val Glu Arg
355 360 365
att gac caa acc tgc aaa atg gta tcc atg ctg tcc att tgt gtg aac 1152
Ile Asp Gln Thr Cys Lys Met Val Ser Met Leu Ser Ile Cys Val Asn
370 375 380
cca ttt ctc tac gga ttc cta aat acc aat ttt cga cac gaa ttt tcc 1200
Pro Phe Leu Tyr Gly Phe Leu Asn Thr Asn Phe Arg His Glu Phe Ser
385 390 395 400
gac atc tac tac cga tac att cgc tgt gaa aca aag agt cag cca gcc 1248
Asp Ile Tyr Tyr Arg Tyr Ile Arg Cys Glu Thr Lys Ser Gln Pro Ala
405 410 415
ggc cga ttc cat cat gat gta tca tca atc gct cac cat aga caa gac 1296
Gly Arg Phe His His Asp Val Ser Ser Ile Ala His His Arg Gln Asp
420 425 430
tct gtt tac aat gat gag gca aca ctt ttg act aca ggg cgt cag agt 1344
Ser Val Tyr Asn Asp Glu Ala Thr Leu Leu Thr Thr Gly Arg Gln Ser
435 440 445
aat ggt aaa gat ggg agc tcg tct cca ata gga ttc cgc tct agc gtc 1392
Asn Gly Lys Asp Gly Ser Ser Ser Pro Ile Gly Phe Arg Ser Ser Val
450 455 460
cgt gtt tgc tcg ggt caa aca aaa atg att gga gat cgc att gta ttg 1440
Arg Val Cys Ser Gly Gln Thr Lys Met Ile Gly Asp Arg Ile Val Leu
›PRT · 2 of 2
465 470 475 480
gac gat gat atc gag aaa gat agt ttt gtc taa 1473
Asp Asp Asp Ile Glu Lys Asp Ser Phe Val
485 490
30
490
›PRT
Homo sapiens
30
Met Glu Val Lys Asp Ile Asp Asn Tyr Cys Asp Arg Gly Ile Ser Pro
1 5 10 15
Asn Ala Ser Asn Tyr Leu Thr Tyr Pro Phe Asp Gly Leu Cys Leu Gln
20 25 30
Lys Phe Phe Tyr Gln Leu Gln Thr Ser Leu Arg Arg Phe Thr Pro Tyr
35 40 45
Glu Glu Ile Ile Tyr Thr Thr Val Tyr Ile Ile Ile Ser Val Ala Ala
50 55 60
Val Ile Gly Asn Gly Leu Val Ile Met Ala Val Val Arg Lys Lys Thr
65 70 75 80
Met Arg Thr Asn Arg Asn Val Leu Ile Leu Asn Leu Ala Leu Ser Asn
85 90 95
Leu Ile Leu Ala Ile Thr Asn Ile Pro Phe Leu Trp Leu Pro Ser Ile
100 105 110
Asp Phe Glu Phe Pro Tyr Ser Arg Phe Phe Cys Lys Phe Ala Asn Val
115 120 125
Leu Pro Gly Ser Asn Ile Tyr Cys Ser Thr Leu Thr Ile Ser Val Met
130 135 140
Ala Ile Asp Arg Tyr Tyr Ser Val Lys Lys Leu Lys Ile Ala Ser Asn
145 150 155 160
Arg Lys Gln Cys Phe His Ala Val Leu Val Ser Leu Ala Ile Trp Ile
165 170 175
Val Ser Phe Ile Leu Ser Leu Pro Leu Leu Leu Tyr Tyr Glu Thr Ser
180 185 190
Met Leu Tyr Val Met Arg Glu Ile Arg Val Val Asp Gln Ser Gly Gln
195 200 205
Glu Val Ile Arg Ser Tyr Gly Trp Arg Gln Cys Arg Leu Val Ser Ala
210 215 220
Gly Arg Leu Pro Asp Ile Thr Gln Ser Ile Gln Leu Leu Met Ser Ile
225 230 235 240
Leu Gln Val Ala Phe Leu Tyr Ile Val Pro Leu Phe Val Leu Ser Ile
245 250 255
Phe Asn Val Lys Leu Thr Arg Phe Leu Lys Thr Asn Ala Asn Lys Met
260 265 270
Ser Lys Thr Arg Ala Pro Pro Lys Arg Phe Asp Arg Ser Asp Ser His
275 280 285
His Asn Ser Leu Lys Asn Asn Asn Asn His Thr Ser Ser Leu Arg Ser
290 295 300
Pro Ser Met Pro Ser Ile Arg Ser Ser Ile Thr Glu Arg Asn Lys Thr
305 310 315 320
Asn Gln Arg Thr Asn Arg Thr Thr Ser Leu Leu Ile Ala Met Ala Gly
325 330 335
Ser Tyr Ala Ala Leu Trp Phe Pro Phe Thr Leu Ile Thr Phe Leu Ile
340 345 350
Asp Phe Glu Leu Ile Ile Asn Gln Asp Tyr Val Asn Leu Val Glu Arg
355 360 365
Ile Asp Gln Thr Cys Lys Met Val Ser Met Leu Ser Ile Cys Val Asn
370 375 380
Pro Phe Leu Tyr Gly Phe Leu Asn Thr Asn Phe Arg His Glu Phe Ser
385 390 395 400
Asp Ile Tyr Tyr Arg Tyr Ile Arg Cys Glu Thr Lys Ser Gln Pro Ala
405 410 415
Gly Arg Phe His His Asp Val Ser Ser Ile Ala His His Arg Gln Asp
420 425 430
Ser Val Tyr Asn Asp Glu Ala Thr Leu Leu Thr Thr Gly Arg Gln Ser
435 440 445
Asn Gly Lys Asp Gly Ser Ser Ser Pro Ile Gly Phe Arg Ser Ser Val
450 455 460
Arg Val Cys Ser Gly Gln Thr Lys Met Ile Gly Asp Arg Ile Val Leu
465 470 475 480
Asp Asp Asp Ile Glu Lys Asp Ser Phe Val
485 490
31
1130
›DNA
Homo sapiens
CeGPCR11 (F35G8.1)
31
atg aat atc gac tcg aat tta aca gat ttt caa tgc gag tac ccg gcg 48
Met Asn Ile Asp Ser Asn Leu Thr Asp Phe Gln Cys Glu Tyr Pro Ala
1 5 10 15
att gac ata gga att cat atg aaa acc ttg tta gca gtc ggg tat ggt 96
Ile Asp Ile Gly Ile His Met Lys Thr Leu Leu Ala Val Gly Tyr Gly
20 25 30
ttg gtg gga gcc ctg agc tta gtt gga aat ctt gct gta ttg ctc att 144
Leu Val Gly Ala Leu Ser Leu Val Gly Asn Leu Ala Val Leu Leu Ile
35 40 45
gtg ata tgc agg aga gaa atg caa aca gtc acc aac att ttt ata agt 192
Val Ile Cys Arg Arg Glu Met Gln Thr Val Thr Asn Ile Phe Ile Ser
50 55 60
agc gtc tcc gcg gct gac ttg gta ata aca agt ttc tcc tta tgg gcg 240
Ser Val Ser Ala Ala Asp Leu Val Ile Thr Ser Phe Ser Leu Trp Ala
65 70 75 80
acc cca tta gcc tat tat caa cgg gta tgg cat ttc gga aag tat atg 288
Thr Pro Leu Ala Tyr Tyr Gln Arg Val Trp His Phe Gly Lys Tyr Met
85 90 95
tgc tat atg gtt tcc atc ata caa ggc tta tcc ctc atg tgg gta cca 336
Cys Tyr Met Val Ser Ile Ile Gln Gly Leu Ser Leu Met Trp Val Pro
100 105 110
ctg acc cta gca gca gtt gca tta gat cgc tac tct cta gtt gct tct 384
Leu Thr Leu Ala Ala Val Ala Leu Asp Arg Tyr Ser Leu Val Ala Ser
115 120 125
ccg ttc cgt cag cca atg tct aaa aag act tgc ttg ctg atc atc gct 432
Pro Phe Arg Gln Pro Met Ser Lys Lys Thr Cys Leu Leu Ile Ile Ala
130 135 140
gga atc tgg atg gga ggg ttt gca gtt ttg tcg cca atg att cgg atg 480
Gly Ile Trp Met Gly Gly Phe Ala Val Leu Ser Pro Met Ile Arg Met
145 150 155 160
gta gat ttt gtt gac agc tat gga cca tgc cat ttt tgc ctg gaa tcc 528
Val Asp Phe Val Asp Ser Tyr Gly Pro Cys His Phe Cys Leu Glu Ser
165 170 175
tgg gac cac gac aaa caa cac tac cga ctt ttc tac gga ctc tcg gtg 576
Trp Asp His Asp Lys Gln His Tyr Arg Leu Phe Tyr Gly Leu Ser Val
180 185 190
ctc gtg atc cgc tcc gca att cca ctt gtt ctc att tcc ctg tgc cac 624
Leu Val Ile Arg Ser Ala Ile Pro Leu Val Leu Ile Ser Leu Cys His
195 200 205
tgg aga atc gca gtt att ttg aac acg cag acg gag aaa ttc caa aca 672
Trp Arg Ile Ala Val Ile Leu Asn Thr Gln Thr Glu Lys Phe Gln Thr
210 215 220
tta cgt agt gcc agc aca gtc acc caa tcg act gac atc cgc cgt aaa 720
Leu Arg Ser Ala Ser Thr Val Thr Gln Ser Thr Asp Ile Arg Arg Lys
225 230 235 240
caa cgt ctt cag act ctc tta ttg gca atg gtg gtc ata ttt gcc gtg 768
Gln Arg Leu Gln Thr Leu Leu Leu Ala Met Val Val Ile Phe Ala Val
245 250 255
tcc agt ctc cca ctg gac ctt tca aat gtt ctt caa gat ttg atc gta 816
Ser Ser Leu Pro Leu Asp Leu Ser Asn Val Leu Gln Asp Leu Ile Val
260 265 270
gtg tac cag gtt cga cct gtc ccc gac aac gtt cgc cat ttt atc ttc 864
Val Tyr Gln Val Arg Pro Val Pro Asp Asn Val Arg His Phe Ile Phe
275 280 285
ttc ttt tgc cat tgg acc gca atg gca gga aca ctt ctc aac ccg tta 912
Phe Phe Cys His Trp Thr Ala Met Ala Gly Thr Leu Leu Asn Pro Leu
290 295 300
gtc tac gcg tac tac aac gaa aac ttc cgg cgt cag att caa aca tgc 960
Val Tyr Ala Tyr Tyr Asn Glu Asn Phe Arg Arg Gln Ile Gln Thr Cys
305 310 315 320
ttt gga gaa atg cgt gga cag gga gag ttc aag cgg ggg cta tac tct 1008
Phe Gly Glu Met Arg Gly Gln Gly Glu Phe Lys Arg Gly Leu Tyr Ser
325 330 335
att gtc tcc ggc aga tac tcc tac cga gca gac gac gag gaa aat cat 1056
Ile Val Ser Gly Arg Tyr Ser Tyr Arg Ala Asp Asp Glu Glu Asn His
340 345 350
cac aga caa aac aca cgt atc gaa ctg gct aac aat caa act ggt gat 1104
His Arg Gln Asn Thr Arg Ile Glu Leu Ala Asn Asn Gln Thr Gly Asp
355 360 365
atc gaa gtt ctc aga acg gat ctt ta 1130
Ile Glu Val Leu Arg Thr Asp Leu
370 375
32
376
›PRT
Homo sapiens
32
Met Asn Ile Asp Ser Asn Leu Thr Asp Phe Gln Cys Glu Tyr Pro Ala
1 5 10 15
Ile Asp Ile Gly Ile His Met Lys Thr Leu Leu Ala Val Gly Tyr Gly
20 25 30
Leu Val Gly Ala Leu Ser Leu Val Gly Asn Leu Ala Val Leu Leu Ile
35 40 45
Val Ile Cys Arg Arg Glu Met Gln Thr Val Thr Asn Ile Phe Ile Ser
50 55 60
Ser Val Ser Ala Ala Asp Leu Val Ile Thr Ser Phe Ser Leu Trp Ala
65 70 75 80
Thr Pro Leu Ala Tyr Tyr Gln Arg Val Trp His Phe Gly Lys Tyr Met
85 90 95
Cys Tyr Met Val Ser Ile Ile Gln Gly Leu Ser Leu Met Trp Val Pro
100 105 110
Leu Thr Leu Ala Ala Val Ala Leu Asp Arg Tyr Ser Leu Val Ala Ser
115 120 125
Pro Phe Arg Gln Pro Met Ser Lys Lys Thr Cys Leu Leu Ile Ile Ala
130 135 140
Gly Ile Trp Met Gly Gly Phe Ala Val Leu Ser Pro Met Ile Arg Met
145 150 155 160
Val Asp Phe Val Asp Ser Tyr Gly Pro Cys His Phe Cys Leu Glu Ser
165 170 175
Trp Asp His Asp Lys Gln His Tyr Arg Leu Phe Tyr Gly Leu Ser Val
180 185 190
Leu Val Ile Arg Ser Ala Ile Pro Leu Val Leu Ile Ser Leu Cys His
195 200 205
Trp Arg Ile Ala Val Ile Leu Asn Thr Gln Thr Glu Lys Phe Gln Thr
210 215 220
Leu Arg Ser Ala Ser Thr Val Thr Gln Ser Thr Asp Ile Arg Arg Lys
225 230 235 240
Gln Arg Leu Gln Thr Leu Leu Leu Ala Met Val Val Ile Phe Ala Val
245 250 255
Ser Ser Leu Pro Leu Asp Leu Ser Asn Val Leu Gln Asp Leu Ile Val
260 265 270
Val Tyr Gln Val Arg Pro Val Pro Asp Asn Val Arg His Phe Ile Phe
275 280 285
Phe Phe Cys His Trp Thr Ala Met Ala Gly Thr Leu Leu Asn Pro Leu
290 295 300
Val Tyr Ala Tyr Tyr Asn Glu Asn Phe Arg Arg Gln Ile Gln Thr Cys
305 310 315 320
Phe Gly Glu Met Arg Gly Gln Gly Glu Phe Lys Arg Gly Leu Tyr Ser
325 330 335
Ile Val Ser Gly Arg Tyr Ser Tyr Arg Ala Asp Asp Glu Glu Asn His
340 345 350
His Arg Gln Asn Thr Arg Ile Glu Leu Ala Asn Asn Gln Thr Gly Asp
355 360 365
Ile Glu Val Leu Arg Thr Asp Leu
370 375
33
atg aca acg tgt ccc cta cca ccc agt tta gac gaa atg gat ctg cga 48
Met Thr Thr Cys Pro Leu Pro Pro Ser Leu Asp Glu Met Asp Leu Arg
1 5 10 15
tta gct gcc gat aaa gtt cta aac ggt tca ctg atc aac tgt aca ttc 96
Leu Ala Ala Asp Lys Val Leu Asn Gly Ser Leu Ile Asn Cys Thr Phe
20 25 30
caa tcg ttt tac gat caa atg tat caa aca cac gga gta tac ttc ata 144
Gln Ser Phe Tyr Asp Gln Met Tyr Gln Thr His Gly Val Tyr Phe Ile
35 40 45
ttt gaa cca act ccg ttc gtt cat cca att gtt tca cag att ttc tat 192
Phe Glu Pro Thr Pro Phe Val His Pro Ile Val Ser Gln Ile Phe Tyr
50 55 60
gga atc ctg ttc aca cta aca ata ttt ctg gcg tta atg ggc aat ttt 240
Gly Ile Leu Phe Thr Leu Thr Ile Phe Leu Ala Leu Met Gly Asn Phe
65 70 75 80
aca gtg atg tgg ata atc ctg tac cac cgt caa atg cga agc gtc aca 288
Thr Val Met Trp Ile Ile Leu Tyr His Arg Gln Met Arg Ser Val Thr
85 90 95
aat tac tat ctg ttc aac ttg gca gtg gcg gat gcc tcg att tca gtg 336
Asn Tyr Tyr Leu Phe Asn Leu Ala Val Ala Asp Ala Ser Ile Ser Val
100 105 110
ttc aac acg gga ttc tca tgg tct tat aat tat tat tat gtt tgg aaa 384
Phe Asn Thr Gly Phe Ser Trp Ser Tyr Asn Tyr Tyr Tyr Val Trp Lys
115 120 125
ttc gga agc ttt tac tgt cga ata aac aat ttg atg gga ata act ccg 432
Phe Gly Ser Phe Tyr Cys Arg Ile Asn Asn Leu Met Gly Ile Thr Pro
130 135 140
att tgt gca agt gta ttc aca atg att gtc atg agc att gaa aga tat 480
Ile Cys Ala Ser Val Phe Thr Met Ile Val Met Ser Ile Glu Arg Tyr
145 150 155 160
tat gcc ata atc cat cca ttg aaa aag cgt cca gga cga cga tca act 528
Tyr Ala Ile Ile His Pro Leu Lys Lys Arg Pro Gly Arg Arg Ser Thr
165 170 175
gtt acc atc atc ata atg att tgg ttc atg gca ttt ttg ttc ggg gtt 576
Val Thr Ile Ile Ile Met Ile Trp Phe Met Ala Phe Leu Phe Gly Val
180 185 190
cca gca ttt ctt gcg tca aag gtt gat gtc tat tac ttc tac gat ggt 624
Pro Ala Phe Leu Ala Ser Lys Val Asp Val Tyr Tyr Phe Tyr Asp Gly
195 200 205
tac aca tta tac gag aat cca ctg tgc ctt gca gac aat tat ccc ggt 672
Tyr Thr Leu Tyr Glu Asn Pro Leu Cys Leu Ala Asp Asn Tyr Pro Gly
210 215 220
gga aat gaa tca cta ctt gga cag gta tac aac aac gga ctg ata act 720
Gly Asn Glu Ser Leu Leu Gly Gln Val Tyr Asn Asn Gly Leu Ile Thr
225 230 235 240
gtt caa tac att ctt cca cta tgc att tta tca gct gct tat tat cga 768
Val Gln Tyr Ile Leu Pro Leu Cys Ile Leu Ser Ala Ala Tyr Tyr Arg
245 250 255
gtt ggt gtt gag ctg aga aag gat aaa acc gtc ggt gac gtg aga cat 816
Val Gly Val Glu Leu Arg Lys Asp Lys Thr Val Gly Asp Val Arg His
260 265 270
gca aaa tca gtg gcc gca aag aag aag gcg tca atc atg ttg gca gtg 864
Ala Lys Ser Val Ala Ala Lys Lys Lys Ala Ser Ile Met Leu Ala Val
275 280 285
gtt gtg ttc att ttc atg att gtc tgg ttc cca tac aat gcc tac tat 912
Val Val Phe Ile Phe Met Ile Val Trp Phe Pro Tyr Asn Ala Tyr Tyr
290 295 300
ctc aca ttg cat tta gtt gag cca att ggg aat aaa atg ttg agt ctg 960
Leu Thr Leu His Leu Val Glu Pro Ile Gly Asn Lys Met Leu Ser Leu
305 310 315 320
tac att tat atc aac atc tat tgg ttg gga atg tcg tca act gtc ttc 1008
Tyr Ile Tyr Ile Asn Ile Tyr Trp Leu Gly Met Ser Ser Thr Val Phe
325 330 335
aat cct gtc atc tat tat ttt atg aac aaa cga ttc cgt gtc gga ttc 1056
Asn Pro Val Ile Tyr Tyr Phe Met Asn Lys Arg Phe Arg Val Gly Phe
340 345 350
cat cac gca ttc cgt tgg ctt cca ttc gtt cgt tcc gat aaa gat gaa 1104
His His Ala Phe Arg Trp Leu Pro Phe Val Arg Ser Asp Lys Asp Glu
355 360 365
tat caa aca att ctt tca cag aca cgt cca tct ctg atg cca ccc aca 1152
Tyr Gln Thr Ile Leu Ser Gln Thr Arg Pro Ser Leu Met Pro Pro Thr
370 375 380
acg atg gcg cat acg gat ttc tga 1176
Thr Met Ala His Thr Asp Phe
385 390
34
391
›PRT
Homo sapiens
34
Met Thr Thr Cys Pro Leu Pro Pro Ser Leu Asp Glu Met Asp Leu Arg
1 5 10 15
Leu Ala Ala Asp Lys Val Leu Asn Gly Ser Leu Ile Asn Cys Thr Phe
20 25 30
Gln Ser Phe Tyr Asp Gln Met Tyr Gln Thr His Gly Val Tyr Phe Ile
35 40 45
Phe Glu Pro Thr Pro Phe Val His Pro Ile Val Ser Gln Ile Phe Tyr
50 55 60
Gly Ile Leu Phe Thr Leu Thr Ile Phe Leu Ala Leu Met Gly Asn Phe
65 70 75 80
Thr Val Met Trp Ile Ile Leu Tyr His Arg Gln Met Arg Ser Val Thr
85 90 95
Asn Tyr Tyr Leu Phe Asn Leu Ala Val Ala Asp Ala Ser Ile Ser Val
100 105 110
Phe Asn Thr Gly Phe Ser Trp Ser Tyr Asn Tyr Tyr Tyr Val Trp Lys
115 120 125
Phe Gly Ser Phe Tyr Cys Arg Ile Asn Asn Leu Met Gly Ile Thr Pro
130 135 140
Ile Cys Ala Ser Val Phe Thr Met Ile Val Met Ser Ile Glu Arg Tyr
145 150 155 160
Tyr Ala Ile Ile His Pro Leu Lys Lys Arg Pro Gly Arg Arg Ser Thr
165 170 175
Val Thr Ile Ile Ile Met Ile Trp Phe Met Ala Phe Leu Phe Gly Val
180 185 190
Pro Ala Phe Leu Ala Ser Lys Val Asp Val Tyr Tyr Phe Tyr Asp Gly
195 200 205
Tyr Thr Leu Tyr Glu Asn Pro Leu Cys Leu Ala Asp Asn Tyr Pro Gly
210 215 220
Gly Asn Glu Ser Leu Leu Gly Gln Val Tyr Asn Asn Gly Leu Ile Thr
225 230 235 240
Val Gln Tyr Ile Leu Pro Leu Cys Ile Leu Ser Ala Ala Tyr Tyr Arg
245 250 255
Val Gly Val Glu Leu Arg Lys Asp Lys Thr Val Gly Asp Val Arg His
260 265 270
Ala Lys Ser Val Ala Ala Lys Lys Lys Ala Ser Ile Met Leu Ala Val
275 280 285
Val Val Phe Ile Phe Met Ile Val Trp Phe Pro Tyr Asn Ala Tyr Tyr
290 295 300
Leu Thr Leu His Leu Val Glu Pro Ile Gly Asn Lys Met Leu Ser Leu
305 310 315 320
Tyr Ile Tyr Ile Asn Ile Tyr Trp Leu Gly Met Ser Ser Thr Val Phe
325 330 335
Asn Pro Val Ile Tyr Tyr Phe Met Asn Lys Arg Phe Arg Val Gly Phe
340 345 350
His His Ala Phe Arg Trp Leu Pro Phe Val Arg Ser Asp Lys Asp Glu
355 360 365
Tyr Gln Thr Ile Leu Ser Gln Thr Arg Pro Ser Leu Met Pro Pro Thr
370 375 380
Thr Met Ala His Thr Asp Phe
385 390
35
atg agt tct tcg aat cac tgc atc gac atc cgt gca tac ttg tgg cag 48
Met Ser Ser Ser Asn His Cys Ile Asp Ile Arg Ala Tyr Leu Trp Gln
1 5 10 15
aca aag cat gac ctg acg ctc cac ccg att ccc atc gcg att ctt gca 96
Thr Lys His Asp Leu Thr Leu His Pro Ile Pro Ile Ala Ile Leu Ala
20 25 30
acc atc tac act ata att gtc gta gtt ggc gta acc ggc aat ttg tta 144
Thr Ile Tyr Thr Ile Ile Val Val Val Gly Val Thr Gly Asn Leu Leu
35 40 45
gta gtg atg tcg gtg atg agg ttc aaa gtt ctt caa tca gtc agg aac 192
Val Val Met Ser Val Met Arg Phe Lys Val Leu Gln Ser Val Arg Asn
50 55 60
atg ttc atc gta tct ttg tca gtt tct gac att ttt gtg gcg att gtt 240
Met Phe Ile Val Ser Leu Ser Val Ser Asp Ile Phe Val Ala Ile Val
65 70 75 80
agt ggt tca gta acg ccg ata acc gca ttc tct aaa gtt tgg tta ttt 288
Ser Gly Ser Val Thr Pro Ile Thr Ala Phe Ser Lys Val Trp Leu Phe
85 90 95
ggt gga cca ttg tgt cat tta cta cct ttg tta cag ggt acc gcg ttg 336
Gly Gly Pro Leu Cys His Leu Leu Pro Leu Leu Gln Gly Thr Ala Leu
100 105 110
agt ttt tcc acg tta acg ctc acc gca att gca att gac aga tat att 384
Ser Phe Ser Thr Leu Thr Leu Thr Ala Ile Ala Ile Asp Arg Tyr Ile
115 120 125
ctc atc tgt cat ccg acg aaa gaa ccg ata cgc aaa gat caa gca ttg 432
Leu Ile Cys His Pro Thr Lys Glu Pro Ile Arg Lys Asp Gln Ala Leu
130 135 140
aaa atg ata agt ttc aac agc gcc atc tca gtt ggg ctt tcg gta cca 480
Lys Met Ile Ser Phe Asn Ser Ala Ile Ser Val Gly Leu Ser Val Pro
145 150 155 160
tta ttc atg aaa cag gaa ctt atg caa ttc cga aac tat tgc gga gaa 528
Leu Phe Met Lys Gln Glu Leu Met Gln Phe Arg Asn Tyr Cys Gly Glu
165 170 175
tat tgc tca gaa aac tgg gga cca gat gct tat ttg aga agc gtt tat 576
Tyr Cys Ser Glu Asn Trp Gly Pro Asp Ala Tyr Leu Arg Ser Val Tyr
180 185 190
gga aca gtg gtg ttc att att caa ttc gtg ttt cca ttg atc acc atc 624
Gly Thr Val Val Phe Ile Ile Gln Phe Val Phe Pro Leu Ile Thr Ile
195 200 205
aca ttt tgc tat gca tct att tct atc aaa cta cga cgt ggt gtc ttt 672
Thr Phe Cys Tyr Ala Ser Ile Ser Ile Lys Leu Arg Arg Gly Val Phe
210 215 220
gtg aga gga agc caa aaa gag ctg atg tct gag gca cgt cgt caa ttg 720
Val Arg Gly Ser Gln Lys Glu Leu Met Ser Glu Ala Arg Arg Gln Leu
225 230 235 240
acg caa cgt cga ctt cgc aca aat cgg atg ctt att atc atg aca gtc 768
Thr Gln Arg Arg Leu Arg Thr Asn Arg Met Leu Ile Ile Met Thr Val
245 250 255
aca ttc gct ctc tca tgg ctg cca tct gtt ggc ttc aac ttt ctc cgg 816
Thr Phe Ala Leu Ser Trp Leu Pro Ser Val Gly Phe Asn Phe Leu Arg
260 265 270
gat tac tca gcg ctt ccc ggc att att gat tca caa gat tac cta ttc 864
Asp Tyr Ser Ala Leu Pro Gly Ile Ile Asp Ser Gln Asp Tyr Leu Phe
275 280 285
gga att att ttc cat tgc att tca atg aca tcg gtg att gtg aat ccc 912
Gly Ile Ile Phe His Cys Ile Ser Met Thr Ser Val Ile Val Asn Pro
290 295 300
ttc ctt tac ggt tac tgt aat gaa cac ttt cgt gct gca ttt gcg gct 960
Phe Leu Tyr Gly Tyr Cys Asn Glu His Phe Arg Ala Ala Phe Ala Ala
305 310 315 320
ctt ctt gac acg gtg aag gca gct tgt gga atg aga cga gtg agt ccg 1008
Leu Leu Asp Thr Val Lys Ala Ala Cys Gly Met Arg Arg Val Ser Pro
325 330 335
ggc aat cca gcg tgc tcc cag cta ctc agt act cac ttt gaa agc acc 1056
Gly Asn Pro Ala Cys Ser Gln Leu Leu Ser Thr His Phe Glu Ser Thr
340 345 350
aca aga cga tcc gtg act acc acg att cca agt tca att taa 1098
Thr Arg Arg Ser Val Thr Thr Thr Ile Pro Ser Ser Ile
355 360 365
36
365
›PRT
Homo sapiens
36
Met Ser Ser Ser Asn His Cys Ile Asp Ile Arg Ala Tyr Leu Trp Gln
1 5 10 15
Thr Lys His Asp Leu Thr Leu His Pro Ile Pro Ile Ala Ile Leu Ala
20 25 30
Thr Ile Tyr Thr Ile Ile Val Val Val Gly Val Thr Gly Asn Leu Leu
35 40 45
Val Val Met Ser Val Met Arg Phe Lys Val Leu Gln Ser Val Arg Asn
50 55 60
Met Phe Ile Val Ser Leu Ser Val Ser Asp Ile Phe Val Ala Ile Val
65 70 75 80
Ser Gly Ser Val Thr Pro Ile Thr Ala Phe Ser Lys Val Trp Leu Phe
85 90 95
Gly Gly Pro Leu Cys His Leu Leu Pro Leu Leu Gln Gly Thr Ala Leu
100 105 110
Ser Phe Ser Thr Leu Thr Leu Thr Ala Ile Ala Ile Asp Arg Tyr Ile
115 120 125
Leu Ile Cys His Pro Thr Lys Glu Pro Ile Arg Lys Asp Gln Ala Leu
130 135 140
Lys Met Ile Ser Phe Asn Ser Ala Ile Ser Val Gly Leu Ser Val Pro
145 150 155 160
Leu Phe Met Lys Gln Glu Leu Met Gln Phe Arg Asn Tyr Cys Gly Glu
165 170 175
Tyr Cys Ser Glu Asn Trp Gly Pro Asp Ala Tyr Leu Arg Ser Val Tyr
180 185 190
Gly Thr Val Val Phe Ile Ile Gln Phe Val Phe Pro Leu Ile Thr Ile
195 200 205
Thr Phe Cys Tyr Ala Ser Ile Ser Ile Lys Leu Arg Arg Gly Val Phe
210 215 220
Val Arg Gly Ser Gln Lys Glu Leu Met Ser Glu Ala Arg Arg Gln Leu
225 230 235 240
Thr Gln Arg Arg Leu Arg Thr Asn Arg Met Leu Ile Ile Met Thr Val
245 250 255
Thr Phe Ala Leu Ser Trp Leu Pro Ser Val Gly Phe Asn Phe Leu Arg
260 265 270
Asp Tyr Ser Ala Leu Pro Gly Ile Ile Asp Ser Gln Asp Tyr Leu Phe
275 280 285
Gly Ile Ile Phe His Cys Ile Ser Met Thr Ser Val Ile Val Asn Pro
290 295 300
Phe Leu Tyr Gly Tyr Cys Asn Glu His Phe Arg Ala Ala Phe Ala Ala
305 310 315 320
Leu Leu Asp Thr Val Lys Ala Ala Cys Gly Met Arg Arg Val Ser Pro
325 330 335
Gly Asn Pro Ala Cys Ser Gln Leu Leu Ser Thr His Phe Glu Ser Thr
340 345 350
Thr Arg Arg Ser Val Thr Thr Thr Ile Pro Ser Ser Ile
355 360 365
37
atg caa ccg cca aac tct acg gga caa tgg aca gaa att cac tgc gat 48
Met Gln Pro Pro Asn Ser Thr Gly Gln Trp Thr Glu Ile His Cys Asp
1 5 10 15
tgg ttt gca gaa atg caa aat tat cag aac agt ctc tac aaa ata aat 96
Trp Phe Ala Glu Met Gln Asn Tyr Gln Asn Ser Leu Tyr Lys Ile Asn
20 25 30
ttt gat gga tcc cca ata gtg cca atg tat ggg ttg gtg tgc tca ttc 144
Phe Asp Gly Ser Pro Ile Val Pro Met Tyr Gly Leu Val Cys Ser Phe
35 40 45
gga gct ctt gcc aac ttt atc gtt ctc ctc gca ttt gtc aga act gca 192
Gly Ala Leu Ala Asn Phe Ile Val Leu Leu Ala Phe Val Arg Thr Ala
50 55 60
aat ctt cga aac ttg aga aac agt ttc att gtg aat ttg gca ttt tcc 240
Asn Leu Arg Asn Leu Arg Asn Ser Phe Ile Val Asn Leu Ala Phe Ser
65 70 75 80
gat ttg att ctg tgt gtg gtc aca gcg cca gtc act ctc tat acg agt 288
Asp Leu Ile Leu Cys Val Val Thr Ala Pro Val Thr Leu Tyr Thr Ser
85 90 95
tta aat tta ttt tgg cca ttc gga gat tgg tcg tgc aag ttt ctc gct 336
Leu Asn Leu Phe Trp Pro Phe Gly Asp Trp Ser Cys Lys Phe Leu Ala
100 105 110
ggt gtt caa gca gtg aac aca ttt gtt tca tca ctt act ctt gca ttc 384
Gly Val Gln Ala Val Asn Thr Phe Val Ser Ser Leu Thr Leu Ala Phe
115 120 125
att gca atg gat cgt gtt ctt cta aca ctt tgt cca gtt cga tgg aga 432
Ile Ala Met Asp Arg Val Leu Leu Thr Leu Cys Pro Val Arg Trp Arg
130 135 140
tta gca gca act gca cca ctt tta tgc tat ggt gtc gta tgg atc att 480
Leu Ala Ala Thr Ala Pro Leu Leu Cys Tyr Gly Val Val Trp Ile Ile
145 150 155 160
tcg atc ata gtt gct ctt cca tat gca ttg gca gta agc tca aag ttg 528
Ser Ile Ile Val Ala Leu Pro Tyr Ala Leu Ala Val Ser Ser Lys Leu
165 170 175
gct cca ttt gat cca tgg agt gat aga gca act ccc aaa atg cta acc 576
Ala Pro Phe Asp Pro Trp Ser Asp Arg Ala Thr Pro Lys Met Leu Thr
180 185 190
tat tgc aac cgt caa gtg cca gaa ata tgt gca gag atg caa gaa gcc 624
Tyr Cys Asn Arg Gln Val Pro Glu Ile Cys Ala Glu Met Gln Glu Ala
195 200 205
tgg gat aat gca att gtc tca aaa acc acg tac aca ttt gta gta ctt 672
Trp Asp Asn Ala Ile Val Ser Lys Thr Thr Tyr Thr Phe Val Val Leu
210 215 220
ggt att caa tat att cta cca ctt gct gcc ttg gca tat gca tat ttc 720
Gly Ile Gln Tyr Ile Leu Pro Leu Ala Ala Leu Ala Tyr Ala Tyr Phe
225 230 235 240
caa atc ggg tca aca att caa aaa cga tca aaa gta tca cgt aca gtg 768
Gln Ile Gly Ser Thr Ile Gln Lys Arg Ser Lys Val Ser Arg Thr Val
245 250 255
gat aca aca aga aga atg caa atg caa aat aga aat cga aga gct ctt 816
Asp Thr Thr Arg Arg Met Gln Met Gln Asn Arg Asn Arg Arg Ala Leu
260 265 270
ttg ttg ctc ttt ttg tta gta ctc aca tat gcc gtg tgt tgg gct cca 864
Leu Leu Leu Phe Leu Leu Val Leu Thr Tyr Ala Val Cys Trp Ala Pro
275 280 285
atg aat att tat cat gtt ttg aat ggt ctc gag att att aac tat tca 912
Met Asn Ile Tyr His Val Leu Asn Gly Leu Glu Ile Ile Asn Tyr Ser
290 295 300
caa aat atg tac atc ttc tgt cat ttg gtc gga atc tct tca acg tgt 960
Gln Asn Met Tyr Ile Phe Cys His Leu Val Gly Ile Ser Ser Thr Cys
305 310 315 320
gtt aat cca ata gta tat gca ctg gtc aat gaa tca ttt aga aat gct 1008
Val Asn Pro Ile Val Tyr Ala Leu Val Asn Glu Ser Phe Arg Asn Ala
325 330 335
tta caa tca atg att ctt caa ttc cgc ccg tgt tac gtc aca act acc 1056
Leu Gln Ser Met Ile Leu Gln Phe Arg Pro Cys Tyr Val Thr Thr Thr
340 345 350
ggt aca gca gcc acc aac gtg tat gca tat tct gcg act tcg aag gcg 1104
Gly Thr Ala Ala Thr Asn Val Tyr Ala Tyr Ser Ala Thr Ser Lys Ala
355 360 365
gaa aat gtc act ttg atg cgt gat cct ttc tca aca act cct cgt cca 1152
Glu Asn Val Thr Leu Met Arg Asp Pro Phe Ser Thr Thr Pro Arg Pro
370 375 380
aat gaa cgc cca gac agt gtt tga 1176
Asn Glu Arg Pro Asp Ser Val
385 390
38
391
›PRT · 1 of 2
Homo sapiens
38
Met Gln Pro Pro Asn Ser Thr Gly Gln Trp Thr Glu Ile His Cys Asp
1 5 10 15
Trp Phe Ala Glu Met Gln Asn Tyr Gln Asn Ser Leu Tyr Lys Ile Asn
20 25 30
Phe Asp Gly Ser Pro Ile Val Pro Met Tyr Gly Leu Val Cys Ser Phe
35 40 45
Gly Ala Leu Ala Asn Phe Ile Val Leu Leu Ala Phe Val Arg Thr Ala
50 55 60
Asn Leu Arg Asn Leu Arg Asn Ser Phe Ile Val Asn Leu Ala Phe Ser
65 70 75 80
Asp Leu Ile Leu Cys Val Val Thr Ala Pro Val Thr Leu Tyr Thr Ser
85 90 95
Leu Asn Leu Phe Trp Pro Phe Gly Asp Trp Ser Cys Lys Phe Leu Ala
100 105 110
Gly Val Gln Ala Val Asn Thr Phe Val Ser Ser Leu Thr Leu Ala Phe
115 120 125
Ile Ala Met Asp Arg Val Leu Leu Thr Leu Cys Pro Val Arg Trp Arg
130 135 140
Leu Ala Ala Thr Ala Pro Leu Leu Cys Tyr Gly Val Val Trp Ile Ile
145 150 155 160
Ser Ile Ile Val Ala Leu Pro Tyr Ala Leu Ala Val Ser Ser Lys Leu
165 170 175
Ala Pro Phe Asp Pro Trp Ser Asp Arg Ala Thr Pro Lys Met Leu Thr
180 185 190
Tyr Cys Asn Arg Gln Val Pro Glu Ile Cys Ala Glu Met Gln Glu Ala
195 200 205
Trp Asp Asn Ala Ile Val Ser Lys Thr Thr Tyr Thr Phe Val Val Leu
210 215 220
Gly Ile Gln Tyr Ile Leu Pro Leu Ala Ala Leu Ala Tyr Ala Tyr Phe
225 230 235 240
Gln Ile Gly Ser Thr Ile Gln Lys Arg Ser Lys Val Ser Arg Thr Val
245 250 255
Asp Thr Thr Arg Arg Met Gln Met Gln Asn Arg Asn Arg Arg Ala Leu
260 265 270
Leu Leu Leu Phe Leu Leu Val Leu Thr Tyr Ala Val Cys Trp Ala Pro
275 280 285
Met Asn Ile Tyr His Val Leu Asn Gly Leu Glu Ile Ile Asn Tyr Ser
290 295 300
Gln Asn Met Tyr Ile Phe Cys His Leu Val Gly Ile Ser Ser Thr Cys
305 310 315 320
Val Asn Pro Ile Val Tyr Ala Leu Val Asn Glu Ser Phe Arg Asn Ala
325 330 335
Leu Gln Ser Met Ile Leu Gln Phe Arg Pro Cys Tyr Val Thr Thr Thr
340 345 350
Gly Thr Ala Ala Thr Asn Val Tyr Ala Tyr Ser Ala Thr Ser Lys Ala
355 360 365
Glu Asn Val Thr Leu Met Arg Asp Pro Phe Ser Thr Thr Pro Arg Pro
370 375 380
Asn Glu Arg Pro Asp Ser Val
385 390
39
atg gag att att ctt tgt ttt ttg aac tac ctg gca cat ttt aaa gcg 48
Met Glu Ile Ile Leu Cys Phe Leu Asn Tyr Leu Ala His Phe Lys Ala
1 5 10 15
ggt ttc gag att gat tta ttt aac ttc att aac ata aca att tat aag 96
Gly Phe Glu Ile Asp Leu Phe Asn Phe Ile Asn Ile Thr Ile Tyr Lys
20 25 30
aaa ttt cct ttt ttt ttt cat cga ctg aaa ttt gat aga aaa ttg gat 144
Lys Phe Pro Phe Phe Phe His Arg Leu Lys Phe Asp Arg Lys Leu Asp
35 40 45
ttt atg att aat ttc cag aac acg ctt ttc gat ttt ccc att tta gaa 192
Phe Met Ile Asn Phe Gln Asn Thr Leu Phe Asp Phe Pro Ile Leu Glu
50 55 60
aaa aac att ttt cga ttt tcg aat tta aaa aat tta aaa act cta aaa 240
Lys Asn Ile Phe Arg Phe Ser Asn Leu Lys Asn Leu Lys Thr Leu Lys
65 70 75 80
ttt tca acg ata ttt tta ttt tat gga ttt tat gca tgt aat ttt ctt 288
Phe Ser Thr Ile Phe Leu Phe Tyr Gly Phe Tyr Ala Cys Asn Phe Leu
85 90 95
caa aac aac ctt cga tgc atg gtt agt tcg gcg gcc acc att tcg acc 336
Gln Asn Asn Leu Arg Cys Met Val Ser Ser Ala Ala Thr Ile Ser Thr
100 105 110
att tca acc aca acg act ccc tcc acc atc agc aac gtt atc aca agt 384
Ile Ser Thr Thr Thr Thr Pro Ser Thr Ile Ser Asn Val Ile Thr Ser
115 120 125
cat tcg aac aat ggc tcg tgc att cag atc gct gag gcg att gcg gca 432
His Ser Asn Asn Gly Ser Cys Ile Gln Ile Ala Glu Ala Ile Ala Ala
130 135 140
caa ggc atc gat gat att act gta gac ttt tac atc cga tca atc ttc 480
Gln Gly Ile Asp Asp Ile Thr Val Asp Phe Tyr Ile Arg Ser Ile Phe
145 150 155 160
aca ttc ctc tac ggg ttc ctg ttt gta tta ggc att ttt gga aac ggc 528
Thr Phe Leu Tyr Gly Phe Leu Phe Val Leu Gly Ile Phe Gly Asn Gly
165 170 175
ggc gta cta tgg gcg gtg gcg aga aac aag cgg ctc caa tcg gct cgc 576
Gly Val Leu Trp Ala Val Ala Arg Asn Lys Arg Leu Gln Ser Ala Arg
180 185 190
aac gta ttt ctg ctc aac ttg atc ttc acc gat ttg ata ttg gtg ttc 624
Asn Val Phe Leu Leu Asn Leu Ile Phe Thr Asp Leu Ile Leu Val Phe
195 200 205
aca gcg att cca gtc aca cca tgg tac gcg atg acc aaa gac tgg gca 672
Thr Ala Ile Pro Val Thr Pro Trp Tyr Ala Met Thr Lys Asp Trp Ala
210 215 220
ttc ggg tca gtg atg tgc cat tta gtt cct ttg tca aat tcg tgt tcg 720
Phe Gly Ser Val Met Cys His Leu Val Pro Leu Ser Asn Ser Cys Ser
225 230 235 240
gtg ttt gtg acg agt tgg agc ctc act gca atc tcc tta gat aaa ttt 768
Val Phe Val Thr Ser Trp Ser Leu Thr Ala Ile Ser Leu Asp Lys Phe
245 250 255
ctg cat atc aac gat ccc acc aaa caa cca gtt tct att cgt caa gcg 816
Leu His Ile Asn Asp Pro Thr Lys Gln Pro Val Ser Ile Arg Gln Ala
260 265 270
ttg gca ata aca ttt ctt atc tgg ata gtc tca aca ctg ata aat cta 864
Leu Ala Ile Thr Phe Leu Ile Trp Ile Val Ser Thr Leu Ile Asn Leu
275 280 285
ccg tat ctt atg tct ttc gag cac gtc gat gga agc ttt tac gtt cag 912
Pro Tyr Leu Met Ser Phe Glu His Val Asp Gly Ser Phe Tyr Val Gln
290 295 300
ccc gga gaa act cca tac tgc ggg cac ttt tgc gac gag gcg aat tgg 960
Pro Gly Glu Thr Pro Tyr Cys Gly His Phe Cys Asp Glu Ala Asn Trp
305 310 315 320
cag agc gaa aat agt cga aag att tac gga act acg gtt atg ttg tta 1008
Gln Ser Glu Asn Ser Arg Lys Ile Tyr Gly Thr Thr Val Met Leu Leu
325 330 335
cag ttc gtc gtg ccg atg gca gtg atc acg tat tgc tac ttc aaa atc 1056
Gln Phe Val Val Pro Met Ala Val Ile Thr Tyr Cys Tyr Phe Lys Ile
340 345 350
ttg caa aaa gtg tca aaa gac atg atc atc caa aat gct caa ttc tgt 1104
Leu Gln Lys Val Ser Lys Asp Met Ile Ile Gln Asn Ala Gln Phe Cys
355 360 365
caa tca ctg aca caa aag cag aga agt gat gcg acg tca cga aag aag 1152
Gln Ser Leu Thr Gln Lys Gln Arg Ser Asp Ala Thr Ser Arg Lys Lys
370 375 380
aaa gtg aat tat att cta att gca atg gtt gtc aca ttt atc ggg tgt 1200
Lys Val Asn Tyr Ile Leu Ile Ala Met Val Val Thr Phe Ile Gly Cys
385 390 395 400
tgg ttg cct tta aca tta ctc aat ttg gtc aaa gat ttt aaa aaa gag 1248
Trp Leu Pro Leu Thr Leu Leu Asn Leu Val Lys Asp Phe Lys Lys Glu
405 410 415
ccc gaa tgg cta aaa cgt cag ccg ttc ttc tgg gca ata aat gct cac 1296
Pro Glu Trp Leu Lys Arg Gln Pro Phe Phe Trp Ala Ile Asn Ala His
420 425 430
gtc ata gcc atg tcc tta gtc gtc tgg aac cct ctg cta ttc ttt tgg 1344
Val Ile Ala Met Ser Leu Val Val Trp Asn Pro Leu Leu Phe Phe Trp
435 440 445
ctg aca cga aaa caa aaa cgt tcc gga ctg tca aaa ata ctc aac tca 1392
Leu Thr Arg Lys Gln Lys Arg Ser Gly Leu Ser Lys Ile Leu Asn Ser
450 455 460
aca gag ggt tcg aaa aaa gca ggt ggt tct gga ttg cga ggg atc cag 1440
Thr Glu Gly Ser Lys Lys Ala Gly Gly Ser Gly Leu Arg Gly Ile Gln
›PRT · 2 of 2
465 470 475 480
cta cac gac ctc ctc ccg acc tct act cat tcg gac aga tgt gca ggc 1488
Leu His Asp Leu Leu Pro Thr Ser Thr His Ser Asp Arg Cys Ala Gly
485 490 495
aac tct ttc taa 1500
Asn Ser Phe
40
499
›PRT · 1 of 2
Homo sapiens
40
Met Glu Ile Ile Leu Cys Phe Leu Asn Tyr Leu Ala His Phe Lys Ala
1 5 10 15
Gly Phe Glu Ile Asp Leu Phe Asn Phe Ile Asn Ile Thr Ile Tyr Lys
20 25 30
Lys Phe Pro Phe Phe Phe His Arg Leu Lys Phe Asp Arg Lys Leu Asp
35 40 45
Phe Met Ile Asn Phe Gln Asn Thr Leu Phe Asp Phe Pro Ile Leu Glu
50 55 60
Lys Asn Ile Phe Arg Phe Ser Asn Leu Lys Asn Leu Lys Thr Leu Lys
65 70 75 80
Phe Ser Thr Ile Phe Leu Phe Tyr Gly Phe Tyr Ala Cys Asn Phe Leu
85 90 95
Gln Asn Asn Leu Arg Cys Met Val Ser Ser Ala Ala Thr Ile Ser Thr
100 105 110
Ile Ser Thr Thr Thr Thr Pro Ser Thr Ile Ser Asn Val Ile Thr Ser
115 120 125
His Ser Asn Asn Gly Ser Cys Ile Gln Ile Ala Glu Ala Ile Ala Ala
130 135 140
Gln Gly Ile Asp Asp Ile Thr Val Asp Phe Tyr Ile Arg Ser Ile Phe
145 150 155 160
Thr Phe Leu Tyr Gly Phe Leu Phe Val Leu Gly Ile Phe Gly Asn Gly
165 170 175
Gly Val Leu Trp Ala Val Ala Arg Asn Lys Arg Leu Gln Ser Ala Arg
180 185 190
Asn Val Phe Leu Leu Asn Leu Ile Phe Thr Asp Leu Ile Leu Val Phe
195 200 205
Thr Ala Ile Pro Val Thr Pro Trp Tyr Ala Met Thr Lys Asp Trp Ala
210 215 220
Phe Gly Ser Val Met Cys His Leu Val Pro Leu Ser Asn Ser Cys Ser
225 230 235 240
Val Phe Val Thr Ser Trp Ser Leu Thr Ala Ile Ser Leu Asp Lys Phe
245 250 255
Leu His Ile Asn Asp Pro Thr Lys Gln Pro Val Ser Ile Arg Gln Ala
260 265 270
Leu Ala Ile Thr Phe Leu Ile Trp Ile Val Ser Thr Leu Ile Asn Leu
275 280 285
Pro Tyr Leu Met Ser Phe Glu His Val Asp Gly Ser Phe Tyr Val Gln
290 295 300
Pro Gly Glu Thr Pro Tyr Cys Gly His Phe Cys Asp Glu Ala Asn Trp
305 310 315 320
Gln Ser Glu Asn Ser Arg Lys Ile Tyr Gly Thr Thr Val Met Leu Leu
325 330 335
Gln Phe Val Val Pro Met Ala Val Ile Thr Tyr Cys Tyr Phe Lys Ile
340 345 350
Leu Gln Lys Val Ser Lys Asp Met Ile Ile Gln Asn Ala Gln Phe Cys
355 360 365
Gln Ser Leu Thr Gln Lys Gln Arg Ser Asp Ala Thr Ser Arg Lys Lys
370 375 380
Lys Val Asn Tyr Ile Leu Ile Ala Met Val Val Thr Phe Ile Gly Cys
385 390 395 400
Trp Leu Pro Leu Thr Leu Leu Asn Leu Val Lys Asp Phe Lys Lys Glu
405 410 415
Pro Glu Trp Leu Lys Arg Gln Pro Phe Phe Trp Ala Ile Asn Ala His
420 425 430
Val Ile Ala Met Ser Leu Val Val Trp Asn Pro Leu Leu Phe Phe Trp
435 440 445
Leu Thr Arg Lys Gln Lys Arg Ser Gly Leu Ser Lys Ile Leu Asn Ser
450 455 460
Thr Glu Gly Ser Lys Lys Ala Gly Gly Ser Gly Leu Arg Gly Ile Gln
465 470 475 480
Leu His Asp Leu Leu Pro Thr Ser Thr His Ser Asp Arg Cys Ala Gly
485 490 495
Asn Ser Phe
41
atg tcg ttg gga aat gtc atc att gcc tct tca gta aat aca cgt tca 48
Met Ser Leu Gly Asn Val Ile Ile Ala Ser Ser Val Asn Thr Arg Ser
1 5 10 15
ccc ata gat tcc cac ttg ttt aga aaa att ttg tgt ttt aga att tat 96
Pro Ile Asp Ser His Leu Phe Arg Lys Ile Leu Cys Phe Arg Ile Tyr
20 25 30
gac gac act atc agt ttt gag aga aaa att gga ata atc att cca aca 144
Asp Asp Thr Ile Ser Phe Glu Arg Lys Ile Gly Ile Ile Ile Pro Thr
35 40 45
att ttt gcc gtt atc att ttg gtc gga ctt gtt ggt aat gct ctg gtg 192
Ile Phe Ala Val Ile Ile Leu Val Gly Leu Val Gly Asn Ala Leu Val
50 55 60
gtg att gtt gct ttt gga cga caa atg aga aat tca acc aac act ctt 240
Val Ile Val Ala Phe Gly Arg Gln Met Arg Asn Ser Thr Asn Thr Leu
65 70 75 80
att att gga ctt gcc ata tct gac ttg atg ttc ctt ttg ctt tgc gtt 288
Ile Ile Gly Leu Ala Ile Ser Asp Leu Met Phe Leu Leu Leu Cys Val
85 90 95
cct ttc aca gct gtg gat tac gca gct ccg aca tgg atc ttt ccc gag 336
Pro Phe Thr Ala Val Asp Tyr Ala Ala Pro Thr Trp Ile Phe Pro Glu
100 105 110
tgg aca tgc tct atg atc aat ttc ttc cag cat acg tct gca tac tgt 384
Trp Thr Cys Ser Met Ile Asn Phe Phe Gln His Thr Ser Ala Tyr Cys
115 120 125
agt gtt tgg aca ttg acc ctt atg gcc ctc gat aga tat cta gca gtt 432
Ser Val Trp Thr Leu Thr Leu Met Ala Leu Asp Arg Tyr Leu Ala Val
130 135 140
gta tat cca gtt gaa tca atg acg tta cga aca cca agg aac act gtg 480
Val Tyr Pro Val Glu Ser Met Thr Leu Arg Thr Pro Arg Asn Thr Val
145 150 155 160
att gcg tta tgc ttc att tac atc att atc att gct tct cag att cct 528
Ile Ala Leu Cys Phe Ile Tyr Ile Ile Ile Ile Ala Ser Gln Ile Pro
165 170 175
gtg gga cga atg cac ggg att tat gtg tat gac ttt att atg gaa aaa 576
Val Gly Arg Met His Gly Ile Tyr Val Tyr Asp Phe Ile Met Glu Lys
180 185 190
cgg tca acg tgc gca att ctg aca att gcc act gca gag gct act ccg 624
Arg Ser Thr Cys Ala Ile Leu Thr Ile Ala Thr Ala Glu Ala Thr Pro
195 200 205
acg atg gct cga aca tat ttt atg aca ttc aat gtg ttt ggt tac gtg 672
Thr Met Ala Arg Thr Tyr Phe Met Thr Phe Asn Val Phe Gly Tyr Val
210 215 220
ctt cca ctg ggc att tct gtt gtc ctg tat ggt ctt atg ttg aga aaa 720
Leu Pro Leu Gly Ile Ser Val Val Leu Tyr Gly Leu Met Leu Arg Lys
225 230 235 240
ctt tgg gac atg cca agg ccg ggt aac tct caa tct gtt ggt ggt aga 768
Leu Trp Asp Met Pro Arg Pro Gly Asn Ser Gln Ser Val Gly Gly Arg
245 250 255
aat ctt aca aat cgt gat agt ggt tca agt att cgt cgg agg ccg gaa 816
Asn Leu Thr Asn Arg Asp Ser Gly Ser Ser Ile Arg Arg Arg Pro Glu
260 265 270
gca act gct gcc aaa aga aaa gtt acc cgc ctt gta ctg tgt gtt ttg 864
Ala Thr Ala Ala Lys Arg Lys Val Thr Arg Leu Val Leu Cys Val Leu
275 280 285
atc acc tgg gca ctt tgc tgg ctt cca cta aac gtg tgc ttt ttc atg 912
Ile Thr Trp Ala Leu Cys Trp Leu Pro Leu Asn Val Cys Phe Phe Met
290 295 300
tct ggt ctc gcc tac ccg gaa cca ctt gtt ata tct cat ggt gtc atc 960
Ser Gly Leu Ala Tyr Pro Glu Pro Leu Val Ile Ser His Gly Val Ile
305 310 315 320
atg gta att gtt cag att gct agt caa gtt ctt gcc tac acc aat tcg 1008
Met Val Ile Val Gln Ile Ala Ser Gln Val Leu Ala Tyr Thr Asn Ser
325 330 335
tgt ctt aat cca att cta tac gct ttg atg tcg cag agt ttc cgg gaa 1056
Cys Leu Asn Pro Ile Leu Tyr Ala Leu Met Ser Gln Ser Phe Arg Glu
340 345 350
gga ttc att cga gtc atg aaa atg cta ata aat aag cta tcg cgt gga 1104
Gly Phe Ile Arg Val Met Lys Met Leu Ile Asn Lys Leu Ser Arg Gly
355 360 365
agg ttc tgc act aat tac cga cga agt gct ctt cgg aca gag ctc aca 1152
Arg Phe Cys Thr Asn Tyr Arg Arg Ser Ala Leu Arg Thr Glu Leu Thr
370 375 380
cac tat aac caa act cca gca cat cca gca aac act gtg gtt caa gtg 1200
His Tyr Asn Gln Thr Pro Ala His Pro Ala Asn Thr Val Val Gln Val
385 390 395 400
tca aac gga gaa cga tcc tcg ctt ctg aaa gat aac tcg agc tct gcg 1248
Ser Asn Gly Glu Arg Ser Ser Leu Leu Lys Asp Asn Ser Ser Ser Ala
405 410 415
acg tct gta caa cca ctt cgg acg tca att caa gcg aag aaa aca aag 1296
›PRT · 2 of 2
Thr Ser Val Gln Pro Leu Arg Thr Ser Ile Gln Ala Lys Lys Thr Lys
420 425 430
aat gtc tgt gat acg atg ctc ttt gtg tgt ttc ttc taa 1335
Asn Val Cys Asp Thr Met Leu Phe Val Cys Phe Phe
435 440
42
444
›PRT
Homo sapiens
42
Met Ser Leu Gly Asn Val Ile Ile Ala Ser Ser Val Asn Thr Arg Ser
1 5 10 15
Pro Ile Asp Ser His Leu Phe Arg Lys Ile Leu Cys Phe Arg Ile Tyr
20 25 30
Asp Asp Thr Ile Ser Phe Glu Arg Lys Ile Gly Ile Ile Ile Pro Thr
35 40 45
Ile Phe Ala Val Ile Ile Leu Val Gly Leu Val Gly Asn Ala Leu Val
50 55 60
Val Ile Val Ala Phe Gly Arg Gln Met Arg Asn Ser Thr Asn Thr Leu
65 70 75 80
Ile Ile Gly Leu Ala Ile Ser Asp Leu Met Phe Leu Leu Leu Cys Val
85 90 95
Pro Phe Thr Ala Val Asp Tyr Ala Ala Pro Thr Trp Ile Phe Pro Glu
100 105 110
Trp Thr Cys Ser Met Ile Asn Phe Phe Gln His Thr Ser Ala Tyr Cys
115 120 125
Ser Val Trp Thr Leu Thr Leu Met Ala Leu Asp Arg Tyr Leu Ala Val
130 135 140
Val Tyr Pro Val Glu Ser Met Thr Leu Arg Thr Pro Arg Asn Thr Val
145 150 155 160
Ile Ala Leu Cys Phe Ile Tyr Ile Ile Ile Ile Ala Ser Gln Ile Pro
165 170 175
Val Gly Arg Met His Gly Ile Tyr Val Tyr Asp Phe Ile Met Glu Lys
180 185 190
Arg Ser Thr Cys Ala Ile Leu Thr Ile Ala Thr Ala Glu Ala Thr Pro
195 200 205
Thr Met Ala Arg Thr Tyr Phe Met Thr Phe Asn Val Phe Gly Tyr Val
210 215 220
Leu Pro Leu Gly Ile Ser Val Val Leu Tyr Gly Leu Met Leu Arg Lys
225 230 235 240
Leu Trp Asp Met Pro Arg Pro Gly Asn Ser Gln Ser Val Gly Gly Arg
245 250 255
Asn Leu Thr Asn Arg Asp Ser Gly Ser Ser Ile Arg Arg Arg Pro Glu
260 265 270
Ala Thr Ala Ala Lys Arg Lys Val Thr Arg Leu Val Leu Cys Val Leu
275 280 285
Ile Thr Trp Ala Leu Cys Trp Leu Pro Leu Asn Val Cys Phe Phe Met
290 295 300
Ser Gly Leu Ala Tyr Pro Glu Pro Leu Val Ile Ser His Gly Val Ile
305 310 315 320
Met Val Ile Val Gln Ile Ala Ser Gln Val Leu Ala Tyr Thr Asn Ser
325 330 335
Cys Leu Asn Pro Ile Leu Tyr Ala Leu Met Ser Gln Ser Phe Arg Glu
340 345 350
Gly Phe Ile Arg Val Met Lys Met Leu Ile Asn Lys Leu Ser Arg Gly
355 360 365
Arg Phe Cys Thr Asn Tyr Arg Arg Ser Ala Leu Arg Thr Glu Leu Thr
370 375 380
His Tyr Asn Gln Thr Pro Ala His Pro Ala Asn Thr Val Val Gln Val
385 390 395 400
Ser Asn Gly Glu Arg Ser Ser Leu Leu Lys Asp Asn Ser Ser Ser Ala
405 410 415
Thr Ser Val Gln Pro Leu Arg Thr Ser Ile Gln Ala Lys Lys Thr Lys
420 425 430
Asn Val Cys Asp Thr Met Leu Phe Val Cys Phe Phe
435 440
43
atg gag ggt ggt cga aac tgt gta atg aca gta caa cag tgg caa cct 48
Met Glu Gly Gly Arg Asn Cys Val Met Thr Val Gln Gln Trp Gln Pro
1 5 10 15
gaa tac aat gat atg aac cag ata aga gca ata ttc tcg tta ctg tac 96
Glu Tyr Asn Asp Met Asn Gln Ile Arg Ala Ile Phe Ser Leu Leu Tyr
20 25 30
ctt ctt gtt tgg gtt gga gct att gtt ggc aac act cta gta ctt tat 144
Leu Leu Val Trp Val Gly Ala Ile Val Gly Asn Thr Leu Val Leu Tyr
35 40 45
gta ctc act ttt aat cag gta tca ctg tca gtt agg acg gta ttt gtt 192
Val Leu Thr Phe Asn Gln Val Ser Leu Ser Val Arg Thr Val Phe Val
50 55 60
gga tgt cta gct ggt tca gac ctt cta atg tgc ctt ttc tcg ttg cca 240
Gly Cys Leu Ala Gly Ser Asp Leu Leu Met Cys Leu Phe Ser Leu Pro
65 70 75 80
att acc gcg att tct ata ttt tca aga gtt tgg gta ttt cct gct ata 288
Ile Thr Ala Ile Ser Ile Phe Ser Arg Val Trp Val Phe Pro Ala Ile
85 90 95
ttt tgc aag ttg atc gga gtt ttc cag ggc ggt acg att ttt gtc tca 336
Phe Cys Lys Leu Ile Gly Val Phe Gln Gly Gly Thr Ile Phe Val Ser
100 105 110
tca ttc aca tta aca gtt att gct ctg gac aga tgt gta cta att cta 384
Ser Phe Thr Leu Thr Val Ile Ala Leu Asp Arg Cys Val Leu Ile Leu
115 120 125
cgt cca aat cag gag ata gta aat ttc ccg aga gct gtc ttc att gtt 432
Arg Pro Asn Gln Glu Ile Val Asn Phe Pro Arg Ala Val Phe Ile Val
130 135 140
ttc tgc att tgg ctt ctc gga tac tct cta gca ctt cct gta ggc atc 480
Phe Cys Ile Trp Leu Leu Gly Tyr Ser Leu Ala Leu Pro Val Gly Ile
145 150 155 160
tac agc gac att gca gta tac gac gaa att tgc ggc aca ttc tgt gaa 528
Tyr Ser Asp Ile Ala Val Tyr Asp Glu Ile Cys Gly Thr Phe Cys Glu
165 170 175
gag aat tgg cct gat ttc aat ccg gat act gga aga tcg gga att cga 576
Glu Asn Trp Pro Asp Phe Asn Pro Asp Thr Gly Arg Ser Gly Ile Arg
180 185 190
aga gct tat gga ctt tct gtg ttg gta ctt caa ttt ggt att cct gca 624
Arg Ala Tyr Gly Leu Ser Val Leu Val Leu Gln Phe Gly Ile Pro Ala
195 200 205
ttg ata agt tca att tgt tac tgg atg att agt cga gtg atg tca gat 672
Leu Ile Ser Ser Ile Cys Tyr Trp Met Ile Ser Arg Val Met Ser Asp
210 215 220
caa tta gca aga aga aga ggg cac aat att cga ccg gaa tct gaa aca 720
Gln Leu Ala Arg Arg Arg Gly His Asn Ile Arg Pro Glu Ser Glu Thr
225 230 235 240
aag ctg gtg aat cga aag aca aga gct aat cga atg atg atc gta atg 768
Lys Leu Val Asn Arg Lys Thr Arg Ala Asn Arg Met Met Ile Val Met
245 250 255
gtt gtt gga ttc gtt ctc gcg tgg atg cca ttc aat gca gtc aat ctc 816
Val Val Gly Phe Val Leu Ala Trp Met Pro Phe Asn Ala Val Asn Leu
260 265 270
tac cgt gac cta ttt gga att tct aaa tgg tat tct aca gtc ttt gcg 864
Tyr Arg Asp Leu Phe Gly Ile Ser Lys Trp Tyr Ser Thr Val Phe Ala
275 280 285
ctt tgc cac gta tgc gca atg tgc tcc gct gtg ctc aac cca atc atc 912
Leu Cys His Val Cys Ala Met Cys Ser Ala Val Leu Asn Pro Ile Ile
290 295 300
tat tcc tgg ttc aat cct caa ttc cga caa agt atc acc act ttg ttc 960
Tyr Ser Trp Phe Asn Pro Gln Phe Arg Gln Ser Ile Thr Thr Leu Phe
305 310 315 320
aag ggt act gat gaa gct aga ttg atc aag aag aaa cca caa tct acc 1008
Lys Gly Thr Asp Glu Ala Arg Leu Ile Lys Lys Lys Pro Gln Ser Thr
325 330 335
agt aaa atg gtt tcg tat ccg act aat ttc tcc gag atc cga aaa gaa 1056
Ser Lys Met Val Ser Tyr Pro Thr Asn Phe Ser Glu Ile Arg Lys Glu
340 345 350
aca gaa att gca tcg aca aag aca aaa atc aca att gct gaa aac gac 1104
Thr Glu Ile Ala Ser Thr Lys Thr Lys Ile Thr Ile Ala Glu Asn Asp
355 360 365
tat cga gct gga gat caa ctt tta taa 1131
Tyr Arg Ala Gly Asp Gln Leu Leu
370 375
44
376
›PRT
Homo sapiens
44
Met Glu Gly Gly Arg Asn Cys Val Met Thr Val Gln Gln Trp Gln Pro
1 5 10 15
Glu Tyr Asn Asp Met Asn Gln Ile Arg Ala Ile Phe Ser Leu Leu Tyr
20 25 30
Leu Leu Val Trp Val Gly Ala Ile Val Gly Asn Thr Leu Val Leu Tyr
35 40 45
Val Leu Thr Phe Asn Gln Val Ser Leu Ser Val Arg Thr Val Phe Val
50 55 60
Gly Cys Leu Ala Gly Ser Asp Leu Leu Met Cys Leu Phe Ser Leu Pro
65 70 75 80
Ile Thr Ala Ile Ser Ile Phe Ser Arg Val Trp Val Phe Pro Ala Ile
85 90 95
Phe Cys Lys Leu Ile Gly Val Phe Gln Gly Gly Thr Ile Phe Val Ser
100 105 110
Ser Phe Thr Leu Thr Val Ile Ala Leu Asp Arg Cys Val Leu Ile Leu
115 120 125
Arg Pro Asn Gln Glu Ile Val Asn Phe Pro Arg Ala Val Phe Ile Val
130 135 140
Phe Cys Ile Trp Leu Leu Gly Tyr Ser Leu Ala Leu Pro Val Gly Ile
145 150 155 160
Tyr Ser Asp Ile Ala Val Tyr Asp Glu Ile Cys Gly Thr Phe Cys Glu
165 170 175
Glu Asn Trp Pro Asp Phe Asn Pro Asp Thr Gly Arg Ser Gly Ile Arg
180 185 190
Arg Ala Tyr Gly Leu Ser Val Leu Val Leu Gln Phe Gly Ile Pro Ala
195 200 205
Leu Ile Ser Ser Ile Cys Tyr Trp Met Ile Ser Arg Val Met Ser Asp
210 215 220
Gln Leu Ala Arg Arg Arg Gly His Asn Ile Arg Pro Glu Ser Glu Thr
225 230 235 240
Lys Leu Val Asn Arg Lys Thr Arg Ala Asn Arg Met Met Ile Val Met
245 250 255
Val Val Gly Phe Val Leu Ala Trp Met Pro Phe Asn Ala Val Asn Leu
260 265 270
Tyr Arg Asp Leu Phe Gly Ile Ser Lys Trp Tyr Ser Thr Val Phe Ala
275 280 285
Leu Cys His Val Cys Ala Met Cys Ser Ala Val Leu Asn Pro Ile Ile
290 295 300
Tyr Ser Trp Phe Asn Pro Gln Phe Arg Gln Ser Ile Thr Thr Leu Phe
305 310 315 320
Lys Gly Thr Asp Glu Ala Arg Leu Ile Lys Lys Lys Pro Gln Ser Thr
325 330 335
Ser Lys Met Val Ser Tyr Pro Thr Asn Phe Ser Glu Ile Arg Lys Glu
340 345 350
Thr Glu Ile Ala Ser Thr Lys Thr Lys Ile Thr Ile Ala Glu Asn Asp
355 360 365
Tyr Arg Ala Gly Asp Gln Leu Leu
370 375
45
atg gga tcg gtg aat gaa tca tgt gac aat tat gta gaa att ttc aac 48
Met Gly Ser Val Asn Glu Ser Cys Asp Asn Tyr Val Glu Ile Phe Asn
1 5 10 15
aaa atc aac tac ttt ttt cga gat gat cag gtt atc aat ggg act gag 96
Lys Ile Asn Tyr Phe Phe Arg Asp Asp Gln Val Ile Asn Gly Thr Glu
20 25 30
tat tca cca aaa gag ttc gga tat ttc atc aca ttc gca tac atg ctg 144
Tyr Ser Pro Lys Glu Phe Gly Tyr Phe Ile Thr Phe Ala Tyr Met Leu
35 40 45
atc att ttg ttt gga gca ata ggc aac ttt ttg aca atc atc gta gtc 192
Ile Ile Leu Phe Gly Ala Ile Gly Asn Phe Leu Thr Ile Ile Val Val
50 55 60
ata ctc aac cca gca atg cgg aca aca agg aac ttt ttc att tta aac 240
Ile Leu Asn Pro Ala Met Arg Thr Thr Arg Asn Phe Phe Ile Leu Asn
65 70 75 80
ttg gcg ctg tcg gac ttt ttt gtt tgt att gtg aca gcg ccg acc aca 288
Leu Ala Leu Ser Asp Phe Phe Val Cys Ile Val Thr Ala Pro Thr Thr
85 90 95
tta tac acg gtt ctc tac atg ttc tgg cca ttt agc agg aca tta tgc 336
Leu Tyr Thr Val Leu Tyr Met Phe Trp Pro Phe Ser Arg Thr Leu Cys
100 105 110
aaa att gcg ggt tcg ctg caa ggc ttt aac ata ttt tta tcc aca ttc 384
Lys Ile Ala Gly Ser Leu Gln Gly Phe Asn Ile Phe Leu Ser Thr Phe
115 120 125
tcg ata gcc tca att gct gtt gat aga tac gtg ctc att atc ttc cca 432
Ser Ile Ala Ser Ile Ala Val Asp Arg Tyr Val Leu Ile Ile Phe Pro
130 135 140
acg aag cga gaa cga caa caa aat ctg tcg ttc tgc ttt ttt atc atg 480
Thr Lys Arg Glu Arg Gln Gln Asn Leu Ser Phe Cys Phe Phe Ile Met
145 150 155 160
atc tgg gtg att tcc cta atc ctt gcg gtt cca ctt ctg cag gct tct 528
Ile Trp Val Ile Ser Leu Ile Leu Ala Val Pro Leu Leu Gln Ala Ser
165 170 175
gat ttg aca ccg gtt ttc gtt gag cca tcg tgc gat ttg gct ctt tac 576
Asp Leu Thr Pro Val Phe Val Glu Pro Ser Cys Asp Leu Ala Leu Tyr
180 185 190
att tgc cat gag caa aat gag ata tgg gaa aag atg atc ata tca aaa 624
Ile Cys His Glu Gln Asn Glu Ile Trp Glu Lys Met Ile Ile Ser Lys
195 200 205
ggc acc tac acg ttg gca gtt ctt atc acc caa tac gca ttt ccc ctg 672
Gly Thr Tyr Thr Leu Ala Val Leu Ile Thr Gln Tyr Ala Phe Pro Leu
210 215 220
ttt tca cta gtc ttc gcc tac tcc cgg ata gca cat cgg atg aag ctg 720
Phe Ser Leu Val Phe Ala Tyr Ser Arg Ile Ala His Arg Met Lys Leu
225 230 235 240
aga ttt gca aac cga aat cag aat gtg aca aca aat acc aat acg agt 768
Arg Phe Ala Asn Arg Asn Gln Asn Val Thr Thr Asn Thr Asn Thr Ser
245 250 255
cag agg aga cga tcc gtc gtg gaa cgt caa cga cgc acc cac ctc ctt 816
Gln Arg Arg Arg Ser Val Val Glu Arg Gln Arg Arg Thr His Leu Leu
260 265 270
ctt gta tgc gtt gta gct gta ttc gcc gtc gcc tgg ctg cca ctc aac 864
Leu Val Cys Val Val Ala Val Phe Ala Val Ala Trp Leu Pro Leu Asn
275 280 285
gtt ttt cat atc ttc aac aca ttc gag ctg gtc aac agt ttt tcc gtt 912
Val Phe His Ile Phe Asn Thr Phe Glu Leu Val Asn Ser Phe Ser Val
290 295 300
aca acg ttc agc atc tgt cac tgc ttg gca atg tgc tcg gcg tgc ttg 960
Thr Thr Phe Ser Ile Cys His Cys Leu Ala Met Cys Ser Ala Cys Leu
305 310 315 320
aat ccg ctg atc tat gca ttt ttc aat cat aac ttc cgg atc gaa ttt 1008
Asn Pro Leu Ile Tyr Ala Phe Phe Asn His Asn Phe Arg Ile Glu Phe
325 330 335
atg cat ctt ttc gat aga gtt ggg ttg aga tcc ctt cga gtc gtg ata 1056
Met His Leu Phe Asp Arg Val Gly Leu Arg Ser Leu Arg Val Val Ile
340 345 350
ttt gga gag caa gag tcg ctc aag aaa agt atg cgc act gaa ttc aga 1104
Phe Gly Glu Gln Glu Ser Leu Lys Lys Ser Met Arg Thr Glu Phe Arg
355 360 365
agc cgt gga ggc tgt aag acg gta act act gca gag cca gca acc ttc 1152
Ser Arg Gly Gly Cys Lys Thr Val Thr Thr Ala Glu Pro Ala Thr Phe
370 375 380
cag aga atg aat gaa agt atg att ctg agt gcg atg gag caa gat gag 1200
Gln Arg Met Asn Glu Ser Met Ile Leu Ser Ala Met Glu Gln Asp Glu
385 390 395 400
cag ctg tga 1209
Gln Leu
46
402
›PRT
Homo sapiens
46
Met Gly Ser Val Asn Glu Ser Cys Asp Asn Tyr Val Glu Ile Phe Asn
1 5 10 15
Lys Ile Asn Tyr Phe Phe Arg Asp Asp Gln Val Ile Asn Gly Thr Glu
20 25 30
Tyr Ser Pro Lys Glu Phe Gly Tyr Phe Ile Thr Phe Ala Tyr Met Leu
35 40 45
Ile Ile Leu Phe Gly Ala Ile Gly Asn Phe Leu Thr Ile Ile Val Val
50 55 60
Ile Leu Asn Pro Ala Met Arg Thr Thr Arg Asn Phe Phe Ile Leu Asn
65 70 75 80
Leu Ala Leu Ser Asp Phe Phe Val Cys Ile Val Thr Ala Pro Thr Thr
85 90 95
Leu Tyr Thr Val Leu Tyr Met Phe Trp Pro Phe Ser Arg Thr Leu Cys
100 105 110
Lys Ile Ala Gly Ser Leu Gln Gly Phe Asn Ile Phe Leu Ser Thr Phe
115 120 125
Ser Ile Ala Ser Ile Ala Val Asp Arg Tyr Val Leu Ile Ile Phe Pro
130 135 140
Thr Lys Arg Glu Arg Gln Gln Asn Leu Ser Phe Cys Phe Phe Ile Met
145 150 155 160
Ile Trp Val Ile Ser Leu Ile Leu Ala Val Pro Leu Leu Gln Ala Ser
165 170 175
Asp Leu Thr Pro Val Phe Val Glu Pro Ser Cys Asp Leu Ala Leu Tyr
180 185 190
Ile Cys His Glu Gln Asn Glu Ile Trp Glu Lys Met Ile Ile Ser Lys
195 200 205
Gly Thr Tyr Thr Leu Ala Val Leu Ile Thr Gln Tyr Ala Phe Pro Leu
210 215 220
Phe Ser Leu Val Phe Ala Tyr Ser Arg Ile Ala His Arg Met Lys Leu
225 230 235 240
Arg Phe Ala Asn Arg Asn Gln Asn Val Thr Thr Asn Thr Asn Thr Ser
245 250 255
Gln Arg Arg Arg Ser Val Val Glu Arg Gln Arg Arg Thr His Leu Leu
260 265 270
Leu Val Cys Val Val Ala Val Phe Ala Val Ala Trp Leu Pro Leu Asn
275 280 285
Val Phe His Ile Phe Asn Thr Phe Glu Leu Val Asn Ser Phe Ser Val
290 295 300
Thr Thr Phe Ser Ile Cys His Cys Leu Ala Met Cys Ser Ala Cys Leu
305 310 315 320
Asn Pro Leu Ile Tyr Ala Phe Phe Asn His Asn Phe Arg Ile Glu Phe
325 330 335
Met His Leu Phe Asp Arg Val Gly Leu Arg Ser Leu Arg Val Val Ile
340 345 350
Phe Gly Glu Gln Glu Ser Leu Lys Lys Ser Met Arg Thr Glu Phe Arg
355 360 365
Ser Arg Gly Gly Cys Lys Thr Val Thr Thr Ala Glu Pro Ala Thr Phe
370 375 380
Gln Arg Met Asn Glu Ser Met Ile Leu Ser Ala Met Glu Gln Asp Glu
385 390 395 400
Gln Leu
47
31
›DNA
Homo sapiens
Description of Artificial Sequence PCR primer
47
gccgccatgc tgtatgcaat tatagccttc a 31
48
31
›DNA
Homo sapiens
Description of Artificial Sequence PCR primer
48
tctagatcag ggagaaagat tatttgtaga a 31
49
31
›DNA
Homo sapiens
Description of Artificial Sequence PCR primer
49
gccgccatgg cccgtgcagt caacatatct c 31
50
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
50
tctagattat gaataggtga aagtgttgaa t 31
51
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
51
gccgccatgg agggtggtcg aaactgtgta a 31
52
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
52
tctagattat aaaagttgat ctccagctcg a 31
53
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
53
gccgccatga atggctccga ttgtctgaat c 31
54
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
54
tctagatcac gtgcatagtg tcgtcgaacg g 31
55
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
55
gccgccatgg gatcggtgaa tgaatcatgt g 31
56
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
56
tctagatcaa ctcagtaccc cttcaactat a 31
57
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
57
gccgccatga ttatttttta tctttacgta g 31
58
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
58
tctagatcac cgttcatggc aactcagaag a 31
59
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
59
gccgccatgg aagttgaaaa ttttaccgac t 31
60
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
60
ctcgagtcag actagcgtgt cgttgacgct g 31
61
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
61
gccgccatga atacgtcatt tgtagagcca t 31
62
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
62
tctagattaa caatcagaag tagatctttc c 31
63
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
63
gccgccatgg aagtgaaaga tatagataac t 31
64
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
64
tctagattag acaaaactat ctttctcgat a 31
65
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
65
gccgccatgt gcttcgcaga aaaaggagaa g 31
66
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
66
tctagattag acacgagaag ttgagctggg t 31
67
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
67
gccgccatgg gactccaagg gcgtgcaatt a 31
68
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
68
tctagattaa agatccgttc tgagaacttc g 31
69
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
69
gccgccatgg aaatggaata tttccgccca t 31
70
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
70
tctagattaa agcagaagcg tctgacaact t 31
71
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
71
gccgccatga tcaacgaaac agaagagaca t 31
72
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
72
ctcgagtcat tggcaaacaa aataagtata a 31
73
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
73
gccgccatgg gtgatgctga atctcatcat t 31
74
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
74
tctagattat tgtttcgttt cgaaacagac a 31
75
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
75
gccgccatga caacgtgtcc cctaccaccc a 31
76
31
›DNA
Artificial Sequence
Description of Artificial Sequence PCR primer
76
tctagatcag aaatccgtat gcgccatcgt t 31
77
Ala Ser Glu Asp Ala Leu Phe Gly Thr Met Arg Phe
1 5 10
78
Ala Ser Pro Ser Phe Ile Arg Phe
1 5
79
8
›PRT
A. suum
AF3
79
Ala Val Pro Gly Val Leu Arg Phe
1 5
80
Asp Val Pro Gly Val Leu Arg Phe
1 5
81
Glu Ala Glu Glu Pro Leu Gly Thr Met Arg Phe
1 5 10
82
Glu Asp Gly Asn Ala Pro Phe Gly Thr Met Arg Phe
1 5 10
83
Glu Ile Val Phe His Gln Ile Ser Pro Ile Phe Phe Arg Phe
1 5 10
84
9
›PRT
A. suum
AF4
84
Gly Asp Val Pro Gly Val Leu Arg Phe
1 5
85
Gly Gly Pro Gln Gly Pro Leu Arg Phe
1 5
86
9
›PRT
A. suum/C.elegans (flp 21)
AF9
86
Gly Leu Gly Pro Arg Pro Leu Arg Phe
1 5
87
8
›PRT
A. suum
AF20
87
Gly Met Pro Gly Val Leu Arg Phe
1 5
88
Gly Asn Ser Phe Leu Arg Phe
1 5
89
Gly Pro Ser Gly Pro Leu Arg Phe
1 5
90
Ile Leu Xaa Arg Phe
1 5
91
Lys Pro Asn Phe Leu Arg Phe
1 5
92
Lys Ser Val Pro Gly Val Leu Arg Phe
1 5
93
Asn Gly Ala Pro Gln Pro Phe Val Arg Phe
1 5 10
94
10
›PRT
Locust
SchistoFLRFa
94
Pro Asp Val Asp His Val Phe Leu Arg Phe
1 5 10
95
11
›PRT
Locust
leucomyosuppressin
95
Xaa Glu Asp Val Asp His Val Phe Leu Arg Phe
1 5 10
96
Ser Ala Asp Asp Ser Ala Pro Phe Gly Thr Met Arg Phe
1 5 10
97
Ser Ala Glu Pro Phe Gly Thr Met Arg Phe
1 5 10
98
Ser Glu Val Pro Gly Val Leu Arg Phe
1 5
99
Ser Pro Leu Gly Thr Met Arg Phe
1 5
100
Ser Val Pro Gly Val Leu Arg Phe
1 5
101
Thr Asp Val Asp His Val Phe Leu Arg Phe
1 5 10
102
8
›PRT
lobster
Lobster peptide II
102
Thr Asn Arg Asn Phe Leu Arg Phe
1 5
103
Xaa Pro Asp Val Asp His Val Phe Leu Arg Phe
1 5 10
104
atg gtt agt tcg gcg gcc acc att tcg acc att tca acc aca acg act 48
Met Val Ser Ser Ala Ala Thr Ile Ser Thr Ile Ser Thr Thr Thr Thr
1 5 10 15
ccc tcc acc atc agc aac gtt atc aca agt cat tcg aac aat ggc tcg 96
Pro Ser Thr Ile Ser Asn Val Ile Thr Ser His Ser Asn Asn Gly Ser
20 25 30
tgc att cag atc gct gag gcg att gcg gca caa ggc atc gat gat att 144
Cys Ile Gln Ile Ala Glu Ala Ile Ala Ala Gln Gly Ile Asp Asp Ile
35 40 45
act gta gac ttt tac atc cga tca atc ttc aca ttc ctc tac ggg ttc 192
Thr Val Asp Phe Tyr Ile Arg Ser Ile Phe Thr Phe Leu Tyr Gly Phe
50 55 60
ctg ttt gta tta ggc att ttt gga aac ggc ggc gta cta tgg gcg gtg 240
Leu Phe Val Leu Gly Ile Phe Gly Asn Gly Gly Val Leu Trp Ala Val
65 70 75 80
gcg aga aac aag cgg ctc caa tcg gct cgc aac gta ttt ctg ctc aac 288
Ala Arg Asn Lys Arg Leu Gln Ser Ala Arg Asn Val Phe Leu Leu Asn
85 90 95
ttg atc ttc acc gat ttg ata ttg gtg ttc aca gcg att cca gtc aca 336
Leu Ile Phe Thr Asp Leu Ile Leu Val Phe Thr Ala Ile Pro Val Thr
100 105 110
cca tgg tac gcg atg acc aaa gac tgg gca ttc ggg tca gtg atg tgc 384
Pro Trp Tyr Ala Met Thr Lys Asp Trp Ala Phe Gly Ser Val Met Cys
115 120 125
cat tta gtt cct ttg tca aat tcg tgt tcg gtg ttt gtg acg agt tgg 432
His Leu Val Pro Leu Ser Asn Ser Cys Ser Val Phe Val Thr Ser Trp
130 135 140
agc ctc act gca atc tcc tta gat aaa ttt ctg cat atc aac gat ccc 480
Ser Leu Thr Ala Ile Ser Leu Asp Lys Phe Leu His Ile Asn Asp Pro
145 150 155 160
acc aaa caa cca gtt tct att cgt caa gcg ttg gca ata aca ttt ctt 528
Thr Lys Gln Pro Val Ser Ile Arg Gln Ala Leu Ala Ile Thr Phe Leu
165 170 175
atc tgg ata gtc tca aca ctg ata aat cta ccg tat ctt atg tct ttc 576
Ile Trp Ile Val Ser Thr Leu Ile Asn Leu Pro Tyr Leu Met Ser Phe
180 185 190
gag cac gtc gat gga agc ttt tac gtt cag ccc gga gaa act cca tac 624
Glu His Val Asp Gly Ser Phe Tyr Val Gln Pro Gly Glu Thr Pro Tyr
195 200 205
tgc ggg cac ttt tgc gac gag gcg aat tgg cag agc gaa aat agt cga 672
Cys Gly His Phe Cys Asp Glu Ala Asn Trp Gln Ser Glu Asn Ser Arg
210 215 220
aag att tac gga act acg gtt atg ttg tta cag ttc gtc gtg ccg atg 720
Lys Ile Tyr Gly Thr Thr Val Met Leu Leu Gln Phe Val Val Pro Met
225 230 235 240
gca gtg atc acg tat tgc tac ttc aaa atc ttg caa aaa gtg tca aaa 768
Ala Val Ile Thr Tyr Cys Tyr Phe Lys Ile Leu Gln Lys Val Ser Lys
245 250 255
gac atg atc atc caa aat gct caa ttc tgt caa tca ctg aca caa aag 816
Asp Met Ile Ile Gln Asn Ala Gln Phe Cys Gln Ser Leu Thr Gln Lys
260 265 270
cag aga agt gat gcg acg tca cga aag aag aaa gtg aat tat att cta 864
Gln Arg Ser Asp Ala Thr Ser Arg Lys Lys Lys Val Asn Tyr Ile Leu
275 280 285
att gca atg gtt gtc aca ttt atc ggg tgt tgg ttg cct tta aca tta 912
Ile Ala Met Val Val Thr Phe Ile Gly Cys Trp Leu Pro Leu Thr Leu
290 295 300
ctc aat ttg gtc aaa gat ttt aaa aaa gag ccc gaa tgg cta aaa cgt 960
Leu Asn Leu Val Lys Asp Phe Lys Lys Glu Pro Glu Trp Leu Lys Arg
305 310 315 320
cag ccg ttc ttc tgg gca ata aat gct cac gtc ata gcc atg tcc tta 1008
Gln Pro Phe Phe Trp Ala Ile Asn Ala His Val Ile Ala Met Ser Leu
325 330 335
gtc gtc tgg aac cct ctg cta ttc ttt tgg ctg aca cga aaa caa aaa 1056
Val Val Trp Asn Pro Leu Leu Phe Phe Trp Leu Thr Arg Lys Gln Lys
340 345 350
cgt tcc gga ctg tca aaa ata ctc aac tca aca gag ggt tcg aaa aaa 1104
Arg Ser Gly Leu Ser Lys Ile Leu Asn Ser Thr Glu Gly Ser Lys Lys
355 360 365
gca ggt ggt tct gga ttg cga ggg atc cag cta cac gac ctc ctc ccg 1152
Ala Gly Gly Ser Gly Leu Arg Gly Ile Gln Leu His Asp Leu Leu Pro
370 375 380
acc tct act cat tcg gac aga tgt gca ggc aac tct ttc ta 1193
Thr Ser Thr His Ser Asp Arg Cys Ala Gly Asn Ser Phe
385 390 395
105
397
›PRT
Homo sapiens
105
Met Val Ser Ser Ala Ala Thr Ile Ser Thr Ile Ser Thr Thr Thr Thr
1 5 10 15
Pro Ser Thr Ile Ser Asn Val Ile Thr Ser His Ser Asn Asn Gly Ser
20 25 30
Cys Ile Gln Ile Ala Glu Ala Ile Ala Ala Gln Gly Ile Asp Asp Ile
35 40 45
Thr Val Asp Phe Tyr Ile Arg Ser Ile Phe Thr Phe Leu Tyr Gly Phe
50 55 60
Leu Phe Val Leu Gly Ile Phe Gly Asn Gly Gly Val Leu Trp Ala Val
65 70 75 80
Ala Arg Asn Lys Arg Leu Gln Ser Ala Arg Asn Val Phe Leu Leu Asn
85 90 95
Leu Ile Phe Thr Asp Leu Ile Leu Val Phe Thr Ala Ile Pro Val Thr
100 105 110
Pro Trp Tyr Ala Met Thr Lys Asp Trp Ala Phe Gly Ser Val Met Cys
115 120 125
His Leu Val Pro Leu Ser Asn Ser Cys Ser Val Phe Val Thr Ser Trp
130 135 140
Ser Leu Thr Ala Ile Ser Leu Asp Lys Phe Leu His Ile Asn Asp Pro
145 150 155 160
Thr Lys Gln Pro Val Ser Ile Arg Gln Ala Leu Ala Ile Thr Phe Leu
165 170 175
Ile Trp Ile Val Ser Thr Leu Ile Asn Leu Pro Tyr Leu Met Ser Phe
180 185 190
Glu His Val Asp Gly Ser Phe Tyr Val Gln Pro Gly Glu Thr Pro Tyr
195 200 205
Cys Gly His Phe Cys Asp Glu Ala Asn Trp Gln Ser Glu Asn Ser Arg
210 215 220
Lys Ile Tyr Gly Thr Thr Val Met Leu Leu Gln Phe Val Val Pro Met
225 230 235 240
Ala Val Ile Thr Tyr Cys Tyr Phe Lys Ile Leu Gln Lys Val Ser Lys
245 250 255
Asp Met Ile Ile Gln Asn Ala Gln Phe Cys Gln Ser Leu Thr Gln Lys
260 265 270
Gln Arg Ser Asp Ala Thr Ser Arg Lys Lys Lys Val Asn Tyr Ile Leu
275 280 285
Ile Ala Met Val Val Thr Phe Ile Gly Cys Trp Leu Pro Leu Thr Leu
290 295 300
Leu Asn Leu Val Lys Asp Phe Lys Lys Glu Pro Glu Trp Leu Lys Arg
305 310 315 320
Gln Pro Phe Phe Trp Ala Ile Asn Ala His Val Ile Ala Met Ser Leu
325 330 335
Val Val Trp Asn Pro Leu Leu Phe Phe Trp Leu Thr Arg Lys Gln Lys
340 345 350
Arg Ser Gly Leu Ser Lys Ile Leu Asn Ser Thr Glu Gly Ser Lys Lys
355 360 365
Ala Gly Gly Ser Gly Leu Arg Gly Ile Gln Leu His Asp Leu Leu Pro
370 375 380
Thr Ser Thr His Ser Asp Arg Cys Ala Gly Asn Ser Phe
385 390 395
106
atg gtt agt tcg gcg gcc acc att tcg acc att tca acc aca acg act 48
Met Val Ser Ser Ala Ala Thr Ile Ser Thr Ile Ser Thr Thr Thr Thr
1 5 10 15
ccc tcc acc atc agc aac gtt atc aca agt cat tcg aac aat ggc tcg 96
Pro Ser Thr Ile Ser Asn Val Ile Thr Ser His Ser Asn Asn Gly Ser
20 25 30
tgc att cag atc gct gag gcg att gcg gca caa ggc atc gat gat att 144
Cys Ile Gln Ile Ala Glu Ala Ile Ala Ala Gln Gly Ile Asp Asp Ile
35 40 45
act gta gac ttt tac atc cga tca atc ttc aca ttc ctc tac ggg ttc 192
Thr Val Asp Phe Tyr Ile Arg Ser Ile Phe Thr Phe Leu Tyr Gly Phe
50 55 60
ctg ttt gta tta ggc att ttt gga aac ggc ggc gta cta tgg gcg gtg 240
Leu Phe Val Leu Gly Ile Phe Gly Asn Gly Gly Val Leu Trp Ala Val
65 70 75 80
gcg aga aac aag cgg ctc caa tcg gct cgc aac gta ttt ctg ctc aac 288
Ala Arg Asn Lys Arg Leu Gln Ser Ala Arg Asn Val Phe Leu Leu Asn
85 90 95
ttg atc ttc acc gat ttg ata ttg gtg ttc aca gcg att cca gtc aca 336
Leu Ile Phe Thr Asp Leu Ile Leu Val Phe Thr Ala Ile Pro Val Thr
100 105 110
cca tgg tac gcg atg acc aaa gac tgg gca ttc ggg tca gtg atg tgc 384
Pro Trp Tyr Ala Met Thr Lys Asp Trp Ala Phe Gly Ser Val Met Cys
115 120 125
cat tta gtt cct ttg tca aat tcg tgt tcg gtg ttt gtg acg agt tgg 432
His Leu Val Pro Leu Ser Asn Ser Cys Ser Val Phe Val Thr Ser Trp
130 135 140
agc ctc act gca atc tcc tta gat aaa ttt ctg cat atc aac gat ccc 480
Ser Leu Thr Ala Ile Ser Leu Asp Lys Phe Leu His Ile Asn Asp Pro
145 150 155 160
acc aaa caa cca gtt tct att cgt caa gcg ttg gca ata aca ttt ctt 528
Thr Lys Gln Pro Val Ser Ile Arg Gln Ala Leu Ala Ile Thr Phe Leu
165 170 175
atc tgg ata gtc tca aca ctg ata aat cta ccg tat ctt atg tct ttc 576
Ile Trp Ile Val Ser Thr Leu Ile Asn Leu Pro Tyr Leu Met Ser Phe
180 185 190
gag cac gtc gat gga agc ttt tac gtt cag ccc gga gaa act cca tac 624
Glu His Val Asp Gly Ser Phe Tyr Val Gln Pro Gly Glu Thr Pro Tyr
195 200 205
tgc ggg cac ttt tgc gac gag gcg aat tgg cag agc gaa aat agt cga 672
Cys Gly His Phe Cys Asp Glu Ala Asn Trp Gln Ser Glu Asn Ser Arg
210 215 220
aag att tac gga act acg gtt atg ttg tta cag ttc gtc gtg ccg atg 720
Lys Ile Tyr Gly Thr Thr Val Met Leu Leu Gln Phe Val Val Pro Met
225 230 235 240
gca gtg atc acg tat tgc tac ttc aaa atc ttg caa aaa gtg tca aaa 768
Ala Val Ile Thr Tyr Cys Tyr Phe Lys Ile Leu Gln Lys Val Ser Lys
245 250 255
gac atg atc atc caa aat gct caa ttc tgt caa tca ctg aca caa aag 816
Asp Met Ile Ile Gln Asn Ala Gln Phe Cys Gln Ser Leu Thr Gln Lys
260 265 270
cag aga agt gat gcg acg tca cga aag aag aaa gtg aat tat att cta 864
Gln Arg Ser Asp Ala Thr Ser Arg Lys Lys Lys Val Asn Tyr Ile Leu
275 280 285
att gca atg gtt gtc aca ttt atc ggg tgt tgg ttg cct tta aca tta 912
Ile Ala Met Val Val Thr Phe Ile Gly Cys Trp Leu Pro Leu Thr Leu
290 295 300
ctc aat ttg gtc aaa gat ttt aaa aaa gag ccc gaa tgg cta aaa cgt 960
Leu Asn Leu Val Lys Asp Phe Lys Lys Glu Pro Glu Trp Leu Lys Arg
305 310 315 320
cag ccg ttc ttc tgg gca ata aat gct cac gtc ata gcc atg tcc tta 1008
Gln Pro Phe Phe Trp Ala Ile Asn Ala His Val Ile Ala Met Ser Leu
325 330 335
gtc gtc tgg aac cct ctg cta ttc ttt tgg ctg aca cga aaa caa aaa 1056
Val Val Trp Asn Pro Leu Leu Phe Phe Trp Leu Thr Arg Lys Gln Lys
340 345 350
cgt tcc gga ctg tca aaa ata ctc aac tca aca gag att gtg tcc tcg 1104
Arg Ser Gly Leu Ser Lys Ile Leu Asn Ser Thr Glu Ile Val Ser Ser
355 360 365
ttt gcc agt aga gtg agt aac tcg att cgg cgg tca acg ttt cgg aga 1152
Phe Ala Ser Arg Val Ser Asn Ser Ile Arg Arg Ser Thr Phe Arg Arg
370 375 380
aac aat att gac agg gtt cga aaa aag cag gtg gtt ctg gat tgc gag 1200
Asn Asn Ile Asp Arg Val Arg Lys Lys Gln Val Val Leu Asp Cys Glu
385 390 395 400
gga tcc agc tac acg acc tcc tcc cga cct cta ctc att cgg aca gat 1248
Gly Ser Ser Tyr Thr Thr Ser Ser Arg Pro Leu Leu Ile Arg Thr Asp
405 410 415
gtg cag gca act ctt tct aat ggc tcg acg agt act acc cgc gag atg 1296
Val Gln Ala Thr Leu Ser Asn Gly Ser Thr Ser Thr Thr Arg Glu Met
420 425 430
ctg ta 1301
Leu
107
433
›PRT
Homo sapiens
107
Met Val Ser Ser Ala Ala Thr Ile Ser Thr Ile Ser Thr Thr Thr Thr
1 5 10 15
Pro Ser Thr Ile Ser Asn Val Ile Thr Ser His Ser Asn Asn Gly Ser
20 25 30
Cys Ile Gln Ile Ala Glu Ala Ile Ala Ala Gln Gly Ile Asp Asp Ile
35 40 45
Thr Val Asp Phe Tyr Ile Arg Ser Ile Phe Thr Phe Leu Tyr Gly Phe
50 55 60
Leu Phe Val Leu Gly Ile Phe Gly Asn Gly Gly Val Leu Trp Ala Val
65 70 75 80
Ala Arg Asn Lys Arg Leu Gln Ser Ala Arg Asn Val Phe Leu Leu Asn
85 90 95
Leu Ile Phe Thr Asp Leu Ile Leu Val Phe Thr Ala Ile Pro Val Thr
100 105 110
Pro Trp Tyr Ala Met Thr Lys Asp Trp Ala Phe Gly Ser Val Met Cys
115 120 125
His Leu Val Pro Leu Ser Asn Ser Cys Ser Val Phe Val Thr Ser Trp
130 135 140
Ser Leu Thr Ala Ile Ser Leu Asp Lys Phe Leu His Ile Asn Asp Pro
145 150 155 160
Thr Lys Gln Pro Val Ser Ile Arg Gln Ala Leu Ala Ile Thr Phe Leu
165 170 175
Ile Trp Ile Val Ser Thr Leu Ile Asn Leu Pro Tyr Leu Met Ser Phe
180 185 190
Glu His Val Asp Gly Ser Phe Tyr Val Gln Pro Gly Glu Thr Pro Tyr
195 200 205
Cys Gly His Phe Cys Asp Glu Ala Asn Trp Gln Ser Glu Asn Ser Arg
210 215 220
Lys Ile Tyr Gly Thr Thr Val Met Leu Leu Gln Phe Val Val Pro Met
225 230 235 240
Ala Val Ile Thr Tyr Cys Tyr Phe Lys Ile Leu Gln Lys Val Ser Lys
245 250 255
Asp Met Ile Ile Gln Asn Ala Gln Phe Cys Gln Ser Leu Thr Gln Lys
260 265 270
Gln Arg Ser Asp Ala Thr Ser Arg Lys Lys Lys Val Asn Tyr Ile Leu
275 280 285
Ile Ala Met Val Val Thr Phe Ile Gly Cys Trp Leu Pro Leu Thr Leu
290 295 300
Leu Asn Leu Val Lys Asp Phe Lys Lys Glu Pro Glu Trp Leu Lys Arg
305 310 315 320
Gln Pro Phe Phe Trp Ala Ile Asn Ala His Val Ile Ala Met Ser Leu
325 330 335
Val Val Trp Asn Pro Leu Leu Phe Phe Trp Leu Thr Arg Lys Gln Lys
340 345 350
Arg Ser Gly Leu Ser Lys Ile Leu Asn Ser Thr Glu Ile Val Ser Ser
355 360 365
Phe Ala Ser Arg Val Ser Asn Ser Ile Arg Arg Ser Thr Phe Arg Arg
370 375 380
Asn Asn Ile Asp Arg Val Arg Lys Lys Gln Val Val Leu Asp Cys Glu
385 390 395 400
Gly Ser Ser Tyr Thr Thr Ser Ser Arg Pro Leu Leu Ile Arg Thr Asp
405 410 415
Val Gln Ala Thr Leu Ser Asn Gly Ser Thr Ser Thr Thr Arg Glu Met
420 425 430
Leu
108
1277
›DNA
Homo sapiens
Clone identifier CEGPCR21 (F02E8.2)
108
atg aca cat gac aat gaa tat caa tct gca ctt gta ttc gat aga att 48
Met Thr His Asp Asn Glu Tyr Gln Ser Ala Leu Val Phe Asp Arg Ile
1 5 10 15
aga cga tta atc tta aac aac tcg aat acc cca ata acg aca aaa gag 96
Arg Arg Leu Ile Leu Asn Asn Ser Asn Thr Pro Ile Thr Thr Lys Glu
20 25 30
ctt ata aag cat tgc ttt cat cca gat gtt cag ttt ctt gct cac atg 144
Leu Ile Lys His Cys Phe His Pro Asp Val Gln Phe Leu Ala His Met
35 40 45
acc ctg aaa tcg ttg gat cat ccc ggt gaa cag ttt aat aat tca tca 192
Thr Leu Lys Ser Leu Asp His Pro Gly Glu Gln Phe Asn Asn Ser Ser
50 55 60
aat caa ctt tta tta gat cgc gct ttt cta aac cca ata ctc aca ctc 240
Asn Gln Leu Leu Leu Asp Arg Ala Phe Leu Asn Pro Ile Leu Thr Leu
65 70 75 80
att ttt gtt ttt gtc ggg ctt gtg ggt ctc att ggc aat ttg ctc aca 288
Ile Phe Val Phe Val Gly Leu Val Gly Leu Ile Gly Asn Leu Leu Thr
85 90 95
gtt att gtg att ttc aag aca aat tcc ctg cac tcg cac aca aac tat 336
Val Ile Val Ile Phe Lys Thr Asn Ser Leu His Ser His Thr Asn Tyr
100 105 110
ttc ttg gca aat ctt gct acc ggc gac ttc tgt ctc atc gtt gtt gga 384
Phe Leu Ala Asn Leu Ala Thr Gly Asp Phe Cys Leu Ile Val Val Gly
115 120 125
gtt tcg ttt gac ttg gtg aat atc tgg aac gat gag gaa ccg ctg gac 432
Val Ser Phe Asp Leu Val Asn Ile Trp Asn Asp Glu Glu Pro Leu Asp
130 135 140
att ttt gga tat tgc tct ctt aca agc act ttt ata tct ctg ttc aca 480
Ile Phe Gly Tyr Cys Ser Leu Thr Ser Thr Phe Ile Ser Leu Phe Thr
145 150 155 160
ttt gct tca att ctc aca att gtt ctt ttg acg gca gaa aga ttt aca 528
Phe Ala Ser Ile Leu Thr Ile Val Leu Leu Thr Ala Glu Arg Phe Thr
165 170 175
gcg att tgt tat cct ttt tcc cat aga aca att ttc gac gag aag cgt 576
Ala Ile Cys Tyr Pro Phe Ser His Arg Thr Ile Phe Asp Glu Lys Arg
180 185 190
gtt aaa agg ttt ata ctg ctt att tgg ttc gta gca ctt ctt cca tca 624
Val Lys Arg Phe Ile Leu Leu Ile Trp Phe Val Ala Leu Leu Pro Ser
195 200 205
att ttc att ggc tct atg ttc aaa cga gtg tct caa gac ttt tgt ggc 672
Ile Phe Ile Gly Ser Met Phe Lys Arg Val Ser Gln Asp Phe Cys Gly
210 215 220
ttc aat cgc caa atg act tat atc ggt cga tgt gat ttg gtg aca tct 720
Phe Asn Arg Gln Met Thr Tyr Ile Gly Arg Cys Asp Leu Val Thr Ser
225 230 235 240
cca gat agt ttc ttc cgg tac cca ttt gaa tca gca ata acc atc act 768
Pro Asp Ser Phe Phe Arg Tyr Pro Phe Glu Ser Ala Ile Thr Ile Thr
245 250 255
ttt gta ctt ccc ttg ttc ttc att att tat tgt tac ttc cgg att cta 816
Phe Val Leu Pro Leu Phe Phe Ile Ile Tyr Cys Tyr Phe Arg Ile Leu
260 265 270
gtc aca ttg aac gag atg tcc aat tcg act cat gtg cac act cca gtc 864
Val Thr Leu Asn Glu Met Ser Asn Ser Thr His Val His Thr Pro Val
275 280 285
gga act gct cgt agt gat agt gga gct ttt cct ttt ccc cat aca tct 912
Gly Thr Ala Arg Ser Asp Ser Gly Ala Phe Pro Phe Pro His Thr Ser
290 295 300
aat aac tct aac act caa agt ttc ccg cta aca gtc cat act aaa aat 960
Asn Asn Ser Asn Thr Gln Ser Phe Pro Leu Thr Val His Thr Lys Asn
305 310 315 320
gtc caa ccg ccg cga agt caa caa gcc cag aaa atg gtc att aag atg 1008
Val Gln Pro Pro Arg Ser Gln Gln Ala Gln Lys Met Val Ile Lys Met
325 330 335
ctt gtt act gta acc gct gta ttt ttt gtc tgc tat ctt cca tat cac 1056
Leu Val Thr Val Thr Ala Val Phe Phe Val Cys Tyr Leu Pro Tyr His
340 345 350
gct caa cgt ctc atc gtc aaa tac aac agc aag gat tgt tcc aac tca 1104
Ala Gln Arg Leu Ile Val Lys Tyr Asn Ser Lys Asp Cys Ser Asn Ser
355 360 365
gac ttc tgc aag ctt ctc tac cca ata gca gga atc ctc caa tac atc 1152
Asp Phe Cys Lys Leu Leu Tyr Pro Ile Ala Gly Ile Leu Gln Tyr Ile
370 375 380
tcc gct tct ttg aat cca att ttc tac aac ctc atg tct gta cga ttt 1200
Ser Ala Ser Leu Asn Pro Ile Phe Tyr Asn Leu Met Ser Val Arg Phe
385 390 395 400
cgc aac gga ttc aag aag ctc atc aaa gac gtc tgg gca cat cga gcc 1248
Arg Asn Gly Phe Lys Lys Leu Ile Lys Asp Val Trp Ala His Arg Ala
405 410 415
cgt agt tac agc aat cta gct cgt gtc ta 1277
Arg Ser Tyr Ser Asn Leu Ala Arg Val
420 425
109
425
›PRT
Homo sapiens
109
Met Thr His Asp Asn Glu Tyr Gln Ser Ala Leu Val Phe Asp Arg Ile
1 5 10 15
Arg Arg Leu Ile Leu Asn Asn Ser Asn Thr Pro Ile Thr Thr Lys Glu
20 25 30
Leu Ile Lys His Cys Phe His Pro Asp Val Gln Phe Leu Ala His Met
35 40 45
Thr Leu Lys Ser Leu Asp His Pro Gly Glu Gln Phe Asn Asn Ser Ser
50 55 60
Asn Gln Leu Leu Leu Asp Arg Ala Phe Leu Asn Pro Ile Leu Thr Leu
65 70 75 80
Ile Phe Val Phe Val Gly Leu Val Gly Leu Ile Gly Asn Leu Leu Thr
85 90 95
Val Ile Val Ile Phe Lys Thr Asn Ser Leu His Ser His Thr Asn Tyr
100 105 110
Phe Leu Ala Asn Leu Ala Thr Gly Asp Phe Cys Leu Ile Val Val Gly
115 120 125
Val Ser Phe Asp Leu Val Asn Ile Trp Asn Asp Glu Glu Pro Leu Asp
130 135 140
Ile Phe Gly Tyr Cys Ser Leu Thr Ser Thr Phe Ile Ser Leu Phe Thr
145 150 155 160
Phe Ala Ser Ile Leu Thr Ile Val Leu Leu Thr Ala Glu Arg Phe Thr
165 170 175
Ala Ile Cys Tyr Pro Phe Ser His Arg Thr Ile Phe Asp Glu Lys Arg
180 185 190
Val Lys Arg Phe Ile Leu Leu Ile Trp Phe Val Ala Leu Leu Pro Ser
195 200 205
Ile Phe Ile Gly Ser Met Phe Lys Arg Val Ser Gln Asp Phe Cys Gly
210 215 220
Phe Asn Arg Gln Met Thr Tyr Ile Gly Arg Cys Asp Leu Val Thr Ser
225 230 235 240
Pro Asp Ser Phe Phe Arg Tyr Pro Phe Glu Ser Ala Ile Thr Ile Thr
245 250 255
Phe Val Leu Pro Leu Phe Phe Ile Ile Tyr Cys Tyr Phe Arg Ile Leu
260 265 270
Val Thr Leu Asn Glu Met Ser Asn Ser Thr His Val His Thr Pro Val
275 280 285
Gly Thr Ala Arg Ser Asp Ser Gly Ala Phe Pro Phe Pro His Thr Ser
290 295 300
Asn Asn Ser Asn Thr Gln Ser Phe Pro Leu Thr Val His Thr Lys Asn
305 310 315 320
Val Gln Pro Pro Arg Ser Gln Gln Ala Gln Lys Met Val Ile Lys Met
325 330 335
Leu Val Thr Val Thr Ala Val Phe Phe Val Cys Tyr Leu Pro Tyr His
340 345 350
Ala Gln Arg Leu Ile Val Lys Tyr Asn Ser Lys Asp Cys Ser Asn Ser
355 360 365
Asp Phe Cys Lys Leu Leu Tyr Pro Ile Ala Gly Ile Leu Gln Tyr Ile
370 375 380
Ser Ala Ser Leu Asn Pro Ile Phe Tyr Asn Leu Met Ser Val Arg Phe
385 390 395 400
Arg Asn Gly Phe Lys Lys Leu Ile Lys Asp Val Trp Ala His Arg Ala
405 410 415
Arg Ser Tyr Ser Asn Leu Ala Arg Val
420 425
110
1310
›DNA
Homo sapiens
Clone identifier CeGPCR22 (C06G4.5)
110
atg tct aca aat ttg gtg gac tat gtc gat gat tcg tat ttg aat caa 48
Met Ser Thr Asn Leu Val Asp Tyr Val Asp Asp Ser Tyr Leu Asn Gln
1 5 10 15
tca atg aat tca gaa aat gga ttg gat tca gtc aca cag att atg tat 96
Ser Met Asn Ser Glu Asn Gly Leu Asp Ser Val Thr Gln Ile Met Tyr
20 25 30
gat atg aaa aag tac aat ata gtg aat gat gtt cta cct cct ccc aat 144
Asp Met Lys Lys Tyr Asn Ile Val Asn Asp Val Leu Pro Pro Pro Asn
35 40 45
cac gaa gat cta cat gtt gta ata atg gca gtt tca tac ctt ctt tta 192
His Glu Asp Leu His Val Val Ile Met Ala Val Ser Tyr Leu Leu Leu
50 55 60
ttt tta tta ggt act tgt gga aat gtg gca gtg tta act aca ata tac 240
Phe Leu Leu Gly Thr Cys Gly Asn Val Ala Val Leu Thr Thr Ile Tyr
65 70 75 80
cat gtt att cga tcg tct cga gcc acg ttg gat aac aca tta ata tat 288
His Val Ile Arg Ser Ser Arg Ala Thr Leu Asp Asn Thr Leu Ile Tyr
85 90 95
gtc att gtg ctt tct tgt gtc gac ttc gga gtt tgt ttg tca ctt cca 336
Val Ile Val Leu Ser Cys Val Asp Phe Gly Val Cys Leu Ser Leu Pro
100 105 110
att acg gtt att gat cag att ctc ggt ttc tgg atg ttt ggc aaa ata 384
Ile Thr Val Ile Asp Gln Ile Leu Gly Phe Trp Met Phe Gly Lys Ile
115 120 125
cca tgt aaa ctt cat gca gta ttc gag aat ttc ggt aaa atc ctg agt 432
Pro Cys Lys Leu His Ala Val Phe Glu Asn Phe Gly Lys Ile Leu Ser
130 135 140
gct ctc att ctg acg gca atg agt ttt gat cgg tac gcc gga gtt tgc 480
Ala Leu Ile Leu Thr Ala Met Ser Phe Asp Arg Tyr Ala Gly Val Cys
145 150 155 160
cat cca cag cga aaa cgg ttg aga tca agg aat ttt gca att act ata 528
His Pro Gln Arg Lys Arg Leu Arg Ser Arg Asn Phe Ala Ile Thr Ile
165 170 175
ctt tta gtt ctt gcc gta tat gca ttc atc aca ctc tgc ccg tta tta 576
Leu Leu Val Leu Ala Val Tyr Ala Phe Ile Thr Leu Cys Pro Leu Leu
180 185 190
tgg tct ttt act gca cgg gaa att ata ctt tat gcg aag gaa aca gca 624
Trp Ser Phe Thr Ala Arg Glu Ile Ile Leu Tyr Ala Lys Glu Thr Ala
195 200 205
ccc gga atg ctg aca aga atg aaa att gag aaa tgc aca gtg gat atc 672
Pro Gly Met Leu Thr Arg Met Lys Ile Glu Lys Cys Thr Val Asp Ile
210 215 220
gac tca caa atg ttc aca gct ttc acg att tat cag ttc att ctc tgt 720
Asp Ser Gln Met Phe Thr Ala Phe Thr Ile Tyr Gln Phe Ile Leu Cys
225 230 235 240
tat tgt act cca ttg gtt ctc atc gcc ttt ttc tat acg aaa ctt ctc 768
Tyr Cys Thr Pro Leu Val Leu Ile Ala Phe Phe Tyr Thr Lys Leu Leu
245 250 255
agc aaa ctc cgt gaa cac aca agg acg ttt aag agc tct caa atc cca 816
Ser Lys Leu Arg Glu His Thr Arg Thr Phe Lys Ser Ser Gln Ile Pro
260 265 270
ttt ttg cac att tcg ttg tac act cta gca gtt gca tgt ttc tat ttt 864
Phe Leu His Ile Ser Leu Tyr Thr Leu Ala Val Ala Cys Phe Tyr Phe
275 280 285
tta tgc tgg acc cca ttc tgg atg gct aca tta ttc gca gtt tat ctc 912
Leu Cys Trp Thr Pro Phe Trp Met Ala Thr Leu Phe Ala Val Tyr Leu
290 295 300
gaa aac tca gca aat tcg agt agt gtt cca cca gtt ttt gta tat att 960
Glu Asn Ser Ala Asn Ser Ser Ser Val Pro Pro Val Phe Val Tyr Ile
305 310 315 320
atg tat ttt att cat gct cta ccg ttc acc aac tct gcc att aat tgg 1008
Met Tyr Phe Ile His Ala Leu Pro Phe Thr Asn Ser Ala Ile Asn Trp
325 330 335
att tta tac ggt gca cta aat ggc caa ctt caa caa aga tat cga tca 1056
Ile Leu Tyr Gly Ala Leu Asn Gly Gln Leu Gln Gln Arg Tyr Arg Ser
340 345 350
aat cgt tca aat tct aca aaa aag acg aca aca aca aca gct tca aca 1104
Asn Arg Ser Asn Ser Thr Lys Lys Thr Thr Thr Thr Thr Ala Ser Thr
355 360 365
gct tta ttg gaa aag aaa atc aca aat ttg aat act aac tct aat tat 1152
Ala Leu Leu Glu Lys Lys Ile Thr Asn Leu Asn Thr Asn Ser Asn Tyr
370 375 380
cag gta aat ggc tca atg aac tca ata gcc act gca gct cca aca aaa 1200
Gln Val Asn Gly Ser Met Asn Ser Ile Ala Thr Ala Ala Pro Thr Lys
385 390 395 400
acg att gga aat aat gaa gta ctt gtt gcc acg tca aca att gat gat 1248
Thr Ile Gly Asn Asn Glu Val Leu Val Ala Thr Ser Thr Ile Asp Asp
405 410 415
gat gtt gca act gat gtt gta gat gtt cga ctt ttg agt aat cat aat 1296
Asp Val Ala Thr Asp Val Val Asp Val Arg Leu Leu Ser Asn His Asn
420 425 430
cca act ttt ctt tg 1310
Pro Thr Phe Leu
435
111
436
›PRT
Homo sapiens
111
Met Ser Thr Asn Leu Val Asp Tyr Val Asp Asp Ser Tyr Leu Asn Gln
1 5 10 15
Ser Met Asn Ser Glu Asn Gly Leu Asp Ser Val Thr Gln Ile Met Tyr
20 25 30
Asp Met Lys Lys Tyr Asn Ile Val Asn Asp Val Leu Pro Pro Pro Asn
35 40 45
His Glu Asp Leu His Val Val Ile Met Ala Val Ser Tyr Leu Leu Leu
50 55 60
Phe Leu Leu Gly Thr Cys Gly Asn Val Ala Val Leu Thr Thr Ile Tyr
65 70 75 80
His Val Ile Arg Ser Ser Arg Ala Thr Leu Asp Asn Thr Leu Ile Tyr
85 90 95
Val Ile Val Leu Ser Cys Val Asp Phe Gly Val Cys Leu Ser Leu Pro
100 105 110
Ile Thr Val Ile Asp Gln Ile Leu Gly Phe Trp Met Phe Gly Lys Ile
115 120 125
Pro Cys Lys Leu His Ala Val Phe Glu Asn Phe Gly Lys Ile Leu Ser
130 135 140
Ala Leu Ile Leu Thr Ala Met Ser Phe Asp Arg Tyr Ala Gly Val Cys
145 150 155 160
His Pro Gln Arg Lys Arg Leu Arg Ser Arg Asn Phe Ala Ile Thr Ile
165 170 175
Leu Leu Val Leu Ala Val Tyr Ala Phe Ile Thr Leu Cys Pro Leu Leu
180 185 190
Trp Ser Phe Thr Ala Arg Glu Ile Ile Leu Tyr Ala Lys Glu Thr Ala
195 200 205
Pro Gly Met Leu Thr Arg Met Lys Ile Glu Lys Cys Thr Val Asp Ile
210 215 220
Asp Ser Gln Met Phe Thr Ala Phe Thr Ile Tyr Gln Phe Ile Leu Cys
225 230 235 240
Tyr Cys Thr Pro Leu Val Leu Ile Ala Phe Phe Tyr Thr Lys Leu Leu
245 250 255
Ser Lys Leu Arg Glu His Thr Arg Thr Phe Lys Ser Ser Gln Ile Pro
260 265 270
Phe Leu His Ile Ser Leu Tyr Thr Leu Ala Val Ala Cys Phe Tyr Phe
275 280 285
Leu Cys Trp Thr Pro Phe Trp Met Ala Thr Leu Phe Ala Val Tyr Leu
290 295 300
Glu Asn Ser Ala Asn Ser Ser Ser Val Pro Pro Val Phe Val Tyr Ile
305 310 315 320
Met Tyr Phe Ile His Ala Leu Pro Phe Thr Asn Ser Ala Ile Asn Trp
325 330 335
Ile Leu Tyr Gly Ala Leu Asn Gly Gln Leu Gln Gln Arg Tyr Arg Ser
340 345 350
Asn Arg Ser Asn Ser Thr Lys Lys Thr Thr Thr Thr Thr Ala Ser Thr
355 360 365
Ala Leu Leu Glu Lys Lys Ile Thr Asn Leu Asn Thr Asn Ser Asn Tyr
370 375 380
Gln Val Asn Gly Ser Met Asn Ser Ile Ala Thr Ala Ala Pro Thr Lys
385 390 395 400
Thr Ile Gly Asn Asn Glu Val Leu Val Ala Thr Ser Thr Ile Asp Asp
405 410 415
Asp Val Ala Thr Asp Val Val Asp Val Arg Leu Leu Ser Asn His Asn
420 425 430
Pro Thr Phe Leu
435
112
1130
›DNA
Homo sapiens
Clone identifier CEGPCR23 (T02E9.1)
112
atg aat atc aca aat tca agc atc ggg tac cca cct cca cca ctg gta 48
Met Asn Ile Thr Asn Ser Ser Ile Gly Tyr Pro Pro Pro Pro Leu Val
1 5 10 15
ggt gca tca ttt gca aaa aca gct ata cca tat tca ata tgt ttt gtt 96
Gly Ala Ser Phe Ala Lys Thr Ala Ile Pro Tyr Ser Ile Cys Phe Val
20 25 30
ttt gga act ctc ggc aac acg gcc gtt ctc tcc tac gtt ttc ttc att 144
Phe Gly Thr Leu Gly Asn Thr Ala Val Leu Ser Tyr Val Phe Phe Ile
35 40 45
aca cga tcc cta aaa tca tct gta act gca ctt ggg aac aca ttt att 192
Thr Arg Ser Leu Lys Ser Ser Val Thr Ala Leu Gly Asn Thr Phe Ile
50 55 60
tac atc gtt gca tta tgc gca gtc gat ctt ctt gtc acc gtt tcg att 240
Tyr Ile Val Ala Leu Cys Ala Val Asp Leu Leu Val Thr Val Ser Ile
65 70 75 80
cca ttc tcc cta tcg tac atg att ctc aac aac tgg gta ttc gga gag 288
Pro Phe Ser Leu Ser Tyr Met Ile Leu Asn Asn Trp Val Phe Gly Glu
85 90 95
ctg gtg tgc aag atc cac ttt atg ctt gag ctt tct aat aag atg tgc 336
Leu Val Cys Lys Ile His Phe Met Leu Glu Leu Ser Asn Lys Met Cys
100 105 110
tcc aca ttc att cta acc gca ctg gca ttt gat cgt tac atg gca ata 384
Ser Thr Phe Ile Leu Thr Ala Leu Ala Phe Asp Arg Tyr Met Ala Ile
115 120 125
tgt cat ccg gaa ata aaa cga atc cat gag atg cgt cac acg att tat 432
Cys His Pro Glu Ile Lys Arg Ile His Glu Met Arg His Thr Ile Tyr
130 135 140
atc acg aca att ctt gca aca ttg tca ctt ttt ctc ata tct cca gtc 480
Ile Thr Thr Ile Leu Ala Thr Leu Ser Leu Phe Leu Ile Ser Pro Val
145 150 155 160
gtt ttg tct gcc aga gtg acg agt ttc aaa agt gga caa tat ttt gtg 528
Val Leu Ser Ala Arg Val Thr Ser Phe Lys Ser Gly Gln Tyr Phe Val
165 170 175
agc gca aag aat gaa cgg cac gaa gtt atc cga caa atg tgt att gac 576
Ser Ala Lys Asn Glu Arg His Glu Val Ile Arg Gln Met Cys Ile Asp
180 185 190
gga atg gca tta gag tgg aag gtt tgg gtg tct gca ttt ctg atc ttc 624
Gly Met Ala Leu Glu Trp Lys Val Trp Val Ser Ala Phe Leu Ile Phe
195 200 205
ttt gca ttc tta ctt cca tgc act ctt cta acc tac ttt tac gcg aag 672
Phe Ala Phe Leu Leu Pro Cys Thr Leu Leu Thr Tyr Phe Tyr Ala Lys
210 215 220
atc gtc ctt cgt ctg agg aga caa aga aga aca atg ctc caa tct cga 720
Ile Val Leu Arg Leu Arg Arg Gln Arg Arg Thr Met Leu Gln Ser Arg
225 230 235 240
att ccc ctc cgc cgc att aca ata tac aca atg gca gcc acg ttc ttc 768
Ile Pro Leu Arg Arg Ile Thr Ile Tyr Thr Met Ala Ala Thr Phe Phe
245 250 255
tat ctt tcg tgt cat att cca ttt tgg ctc cca cag atc tac aac att 816
Tyr Leu Ser Cys His Ile Pro Phe Trp Leu Pro Gln Ile Tyr Asn Ile
260 265 270
ttc tcg aca gtt ctt ggg cac aaa atg aat cca aaa gtc atg aca ttc 864
Phe Ser Thr Val Leu Gly His Lys Met Asn Pro Lys Val Met Thr Phe
275 280 285
acc tat tat tca cat ctt ctt ccg ttc ata tcg gcg gca ttt aac tgg 912
Thr Tyr Tyr Ser His Leu Leu Pro Phe Ile Ser Ala Ala Phe Asn Trp
290 295 300
ata ttc tat gct cga ctt aat agt caa ttc aaa aaa gga ttg gtt cta 960
Ile Phe Tyr Ala Arg Leu Asn Ser Gln Phe Lys Lys Gly Leu Val Leu
305 310 315 320
gtt acg gaa aga atg atc aga aag cga aca aaa tcg atg cat gag aaa 1008
Val Thr Glu Arg Met Ile Arg Lys Arg Thr Lys Ser Met His Glu Lys
325 330 335
gga tac agt gag gca gct gtt gag ctt acg agc aag ttc gat gat gtc 1056
Gly Tyr Ser Glu Ala Ala Val Glu Leu Thr Ser Lys Phe Asp Asp Val
340 345 350
cca ttg atg tgt cca cac tgt gaa gct caa ctt tct att cgt tca agt 1104
Pro Leu Met Cys Pro His Cys Glu Ala Gln Leu Ser Ile Arg Ser Ser
355 360 365
agt aat gga aag aag aat tca aga ta 1130
Ser Asn Gly Lys Lys Asn Ser Arg
370 375
113
376
›PRT
Homo sapiens
113
Met Asn Ile Thr Asn Ser Ser Ile Gly Tyr Pro Pro Pro Pro Leu Val
1 5 10 15
Gly Ala Ser Phe Ala Lys Thr Ala Ile Pro Tyr Ser Ile Cys Phe Val
20 25 30
Phe Gly Thr Leu Gly Asn Thr Ala Val Leu Ser Tyr Val Phe Phe Ile
35 40 45
Thr Arg Ser Leu Lys Ser Ser Val Thr Ala Leu Gly Asn Thr Phe Ile
50 55 60
Tyr Ile Val Ala Leu Cys Ala Val Asp Leu Leu Val Thr Val Ser Ile
65 70 75 80
Pro Phe Ser Leu Ser Tyr Met Ile Leu Asn Asn Trp Val Phe Gly Glu
85 90 95
Leu Val Cys Lys Ile His Phe Met Leu Glu Leu Ser Asn Lys Met Cys
100 105 110
Ser Thr Phe Ile Leu Thr Ala Leu Ala Phe Asp Arg Tyr Met Ala Ile
115 120 125
Cys His Pro Glu Ile Lys Arg Ile His Glu Met Arg His Thr Ile Tyr
130 135 140
Ile Thr Thr Ile Leu Ala Thr Leu Ser Leu Phe Leu Ile Ser Pro Val
145 150 155 160
Val Leu Ser Ala Arg Val Thr Ser Phe Lys Ser Gly Gln Tyr Phe Val
165 170 175
Ser Ala Lys Asn Glu Arg His Glu Val Ile Arg Gln Met Cys Ile Asp
180 185 190
Gly Met Ala Leu Glu Trp Lys Val Trp Val Ser Ala Phe Leu Ile Phe
195 200 205
Phe Ala Phe Leu Leu Pro Cys Thr Leu Leu Thr Tyr Phe Tyr Ala Lys
210 215 220
Ile Val Leu Arg Leu Arg Arg Gln Arg Arg Thr Met Leu Gln Ser Arg
225 230 235 240
Ile Pro Leu Arg Arg Ile Thr Ile Tyr Thr Met Ala Ala Thr Phe Phe
245 250 255
Tyr Leu Ser Cys His Ile Pro Phe Trp Leu Pro Gln Ile Tyr Asn Ile
260 265 270
Phe Ser Thr Val Leu Gly His Lys Met Asn Pro Lys Val Met Thr Phe
275 280 285
Thr Tyr Tyr Ser His Leu Leu Pro Phe Ile Ser Ala Ala Phe Asn Trp
290 295 300
Ile Phe Tyr Ala Arg Leu Asn Ser Gln Phe Lys Lys Gly Leu Val Leu
305 310 315 320
Val Thr Glu Arg Met Ile Arg Lys Arg Thr Lys Ser Met His Glu Lys
325 330 335
Gly Tyr Ser Glu Ala Ala Val Glu Leu Thr Ser Lys Phe Asp Asp Val
340 345 350
Pro Leu Met Cys Pro His Cys Glu Ala Gln Leu Ser Ile Arg Ser Ser
355 360 365
Ser Asn Gly Lys Lys Asn Ser Arg
370 375
114
1284
›DNA
Homo sapiens
Clone identifier CEGPCR24a (Y59H11AL.a-01)
114
atg gac gaa gga ggg ggt att gga agc agt ttg ctc tcc aga atc acg 48
Met Asp Glu Gly Gly Gly Ile Gly Ser Ser Leu Leu Ser Arg Ile Thr
1 5 10 15
acg aca gcc tct gaa att atg atg cga aac gaa ccg acg aca act gaa 96
Thr Thr Ala Ser Glu Ile Met Met Arg Asn Glu Pro Thr Thr Thr Glu
20 25 30
aat cca gct gtt caa gaa atg aat cat att tat cat ttg aca cca agt 144
Asn Pro Ala Val Gln Glu Met Asn His Ile Tyr His Leu Thr Pro Ser
35 40 45
atg aag atg tta tgt att ctt ttc tat agt ata ctt tgc gta tgc tgt 192
Met Lys Met Leu Cys Ile Leu Phe Tyr Ser Ile Leu Cys Val Cys Cys
50 55 60
gtc tat gga aat gtg ctc gtt att ctg gtt att gtt tat ttt aaa cga 240
Val Tyr Gly Asn Val Leu Val Ile Leu Val Ile Val Tyr Phe Lys Arg
65 70 75 80
ctt cga acg gcg act aat att ttg ata ttg aac ttg gca gtt gct gat 288
Leu Arg Thr Ala Thr Asn Ile Leu Ile Leu Asn Leu Ala Val Ala Asp
85 90 95
ctt ctc ata tca gta ttc tgc att ccg ttc agc tat tgg caa gta ttg 336
Leu Leu Ile Ser Val Phe Cys Ile Pro Phe Ser Tyr Trp Gln Val Leu
100 105 110
att tat gat gat caa cgt tgg ctc ttc ggc tca atg atg tgc tct tta 384
Ile Tyr Asp Asp Gln Arg Trp Leu Phe Gly Ser Met Met Cys Ser Leu
115 120 125
tta gca ttc ctt caa gca atg gct gta ttt tta tca gct tgg aca ctt 432
Leu Ala Phe Leu Gln Ala Met Ala Val Phe Leu Ser Ala Trp Thr Leu
130 135 140
gtc gtt atc agt ttt gat cgg tgg atg gct ata atg ttc ctt tta act 480
Val Val Ile Ser Phe Asp Arg Trp Met Ala Ile Met Phe Leu Leu Thr
145 150 155 160
cca aat att cga att aca aga cga aga gct ctt tat ctg gta gct gcc 528
Pro Asn Ile Arg Ile Thr Arg Arg Arg Ala Leu Tyr Leu Val Ala Ala
165 170 175
acg tgg ata ttc agt atc cta atg gcg ttg ccg tta ctc ttc aca acg 576
Thr Trp Ile Phe Ser Ile Leu Met Ala Leu Pro Leu Leu Phe Thr Thr
180 185 190
aga ttt ttc gaa gac caa gac ggt tta ccg aat tgt gga gaa aat tgg 624
Arg Phe Phe Glu Asp Gln Asp Gly Leu Pro Asn Cys Gly Glu Asn Trp
195 200 205
acg tat ttt gga gat tct gga gaa caa gtg aga aaa gtg tat tct tca 672
Thr Tyr Phe Gly Asp Ser Gly Glu Gln Val Arg Lys Val Tyr Ser Ser
210 215 220
atg gtc tta att cta caa tat gtt gta cct caa gca gtt tta ata ata 720
Met Val Leu Ile Leu Gln Tyr Val Val Pro Gln Ala Val Leu Ile Ile
225 230 235 240
act tac aca cat att gga att aaa atg tgg aat agt cga gta cca gga 768
Thr Tyr Thr His Ile Gly Ile Lys Met Trp Asn Ser Arg Val Pro Gly
245 250 255
atg cag aat gga gca aca aag aaa atg atc gtt gat cga cat gaa agt 816
Met Gln Asn Gly Ala Thr Lys Lys Met Ile Val Asp Arg His Glu Ser
260 265 270
gtc aaa aag ctg gtc cca atg gtg att ctc att tcg gca ctc ttc gca 864
Val Lys Lys Leu Val Pro Met Val Ile Leu Ile Ser Ala Leu Phe Ala
275 280 285
ctt tgt tgg ctt cct tta ctt ata ctg atc aac gtc att cca gaa ttc 912
Leu Cys Trp Leu Pro Leu Leu Ile Leu Ile Asn Val Ile Pro Glu Phe
290 295 300
tat cca gat atc aac agt tgg gga tat att ctg tat ttg tgg tgg ttt 960
Tyr Pro Asp Ile Asn Ser Trp Gly Tyr Ile Leu Tyr Leu Trp Trp Phe
305 310 315 320
gct cat gga ctt gcc atg tct cat tca atg gtc aac cca att atc tat 1008
Ala His Gly Leu Ala Met Ser His Ser Met Val Asn Pro Ile Ile Tyr
325 330 335
ttc att cga aat gcc cgt ttc cgt gaa gga ttc tgt ttc ttc tct tca 1056
Phe Ile Arg Asn Ala Arg Phe Arg Glu Gly Phe Cys Phe Phe Ser Ser
340 345 350
aaa ctt ctt cca tgt ata tca ttt aaa gaa ctt cgt ctt tta act gat 1104
Lys Leu Leu Pro Cys Ile Ser Phe Lys Glu Leu Arg Leu Leu Thr Asp
355 360 365
aat act agc aga aga cat cgc tta cga gat att cac gaa gtg gag tcc 1152
Asn Thr Ser Arg Arg His Arg Leu Arg Asp Ile His Glu Val Glu Ser
370 375 380
ttg aca ggc aaa cat gtc gtt cgg cac gtt tct tcg aag ccc gac cac 1200
Leu Thr Gly Lys His Val Val Arg His Val Ser Ser Lys Pro Asp His
385 390 395 400
tcg tcg tcg tcc gaa aca act ctg cca att ctc tcg cgt agc ttt tcc 1248
Ser Ser Ser Ser Glu Thr Thr Leu Pro Ile Leu Ser Arg Ser Phe Ser
405 410 415
cgt atc att aag aaa att gat cta cca tgt act tga 1284
Arg Ile Ile Lys Lys Ile Asp Leu Pro Cys Thr
420 425
115
427
›PRT
Homo sapiens
115
Met Asp Glu Gly Gly Gly Ile Gly Ser Ser Leu Leu Ser Arg Ile Thr
1 5 10 15
Thr Thr Ala Ser Glu Ile Met Met Arg Asn Glu Pro Thr Thr Thr Glu
20 25 30
Asn Pro Ala Val Gln Glu Met Asn His Ile Tyr His Leu Thr Pro Ser
35 40 45
Met Lys Met Leu Cys Ile Leu Phe Tyr Ser Ile Leu Cys Val Cys Cys
50 55 60
Val Tyr Gly Asn Val Leu Val Ile Leu Val Ile Val Tyr Phe Lys Arg
65 70 75 80
Leu Arg Thr Ala Thr Asn Ile Leu Ile Leu Asn Leu Ala Val Ala Asp
85 90 95
Leu Leu Ile Ser Val Phe Cys Ile Pro Phe Ser Tyr Trp Gln Val Leu
100 105 110
Ile Tyr Asp Asp Gln Arg Trp Leu Phe Gly Ser Met Met Cys Ser Leu
115 120 125
Leu Ala Phe Leu Gln Ala Met Ala Val Phe Leu Ser Ala Trp Thr Leu
130 135 140
Val Val Ile Ser Phe Asp Arg Trp Met Ala Ile Met Phe Leu Leu Thr
145 150 155 160
Pro Asn Ile Arg Ile Thr Arg Arg Arg Ala Leu Tyr Leu Val Ala Ala
165 170 175
Thr Trp Ile Phe Ser Ile Leu Met Ala Leu Pro Leu Leu Phe Thr Thr
180 185 190
Arg Phe Phe Glu Asp Gln Asp Gly Leu Pro Asn Cys Gly Glu Asn Trp
195 200 205
Thr Tyr Phe Gly Asp Ser Gly Glu Gln Val Arg Lys Val Tyr Ser Ser
210 215 220
Met Val Leu Ile Leu Gln Tyr Val Val Pro Gln Ala Val Leu Ile Ile
225 230 235 240
Thr Tyr Thr His Ile Gly Ile Lys Met Trp Asn Ser Arg Val Pro Gly
245 250 255
Met Gln Asn Gly Ala Thr Lys Lys Met Ile Val Asp Arg His Glu Ser
260 265 270
Val Lys Lys Leu Val Pro Met Val Ile Leu Ile Ser Ala Leu Phe Ala
275 280 285
Leu Cys Trp Leu Pro Leu Leu Ile Leu Ile Asn Val Ile Pro Glu Phe
290 295 300
Tyr Pro Asp Ile Asn Ser Trp Gly Tyr Ile Leu Tyr Leu Trp Trp Phe
305 310 315 320
Ala His Gly Leu Ala Met Ser His Ser Met Val Asn Pro Ile Ile Tyr
325 330 335
Phe Ile Arg Asn Ala Arg Phe Arg Glu Gly Phe Cys Phe Phe Ser Ser
340 345 350
Lys Leu Leu Pro Cys Ile Ser Phe Lys Glu Leu Arg Leu Leu Thr Asp
355 360 365
Asn Thr Ser Arg Arg His Arg Leu Arg Asp Ile His Glu Val Glu Ser
370 375 380
Leu Thr Gly Lys His Val Val Arg His Val Ser Ser Lys Pro Asp His
385 390 395 400
Ser Ser Ser Ser Glu Thr Thr Leu Pro Ile Leu Ser Arg Ser Phe Ser
405 410 415
Arg Ile Ile Lys Lys Ile Asp Leu Pro Cys Thr
420 425
116
1305
›DNA
Homo sapiens
Clone identifier CEGPCR24b (Y59H11A1.a-02)
116
atg gac gaa gga ggg ggt att gga agc agt ttg ctc tcc aga atc acg 48
Met Asp Glu Gly Gly Gly Ile Gly Ser Ser Leu Leu Ser Arg Ile Thr
1 5 10 15
acg aca gcc tct gaa att atg atg cga aac gaa ccg acg aca act gaa 96
Thr Thr Ala Ser Glu Ile Met Met Arg Asn Glu Pro Thr Thr Thr Glu
20 25 30
aat cca gct gtt caa gaa atg aat cat att tat cat ttg aca cca agt 144
Asn Pro Ala Val Gln Glu Met Asn His Ile Tyr His Leu Thr Pro Ser
35 40 45
atg aag atg tta tgt att ctt ttc tat agt ata ctt tgc gta tgc tgt 192
Met Lys Met Leu Cys Ile Leu Phe Tyr Ser Ile Leu Cys Val Cys Cys
50 55 60
gtc tat gga aat gtg ctc gtt att ctg gtt att gtt tat ttt aaa cga 240
Val Tyr Gly Asn Val Leu Val Ile Leu Val Ile Val Tyr Phe Lys Arg
65 70 75 80
ctt cga acg gcg act aat att ttg ata ttg aac ttg gca gtt gct gat 288
Leu Arg Thr Ala Thr Asn Ile Leu Ile Leu Asn Leu Ala Val Ala Asp
85 90 95
ctt ctc ata tca gta ttc tgc att ccg ttc agc tat tgg caa gta ttg 336
Leu Leu Ile Ser Val Phe Cys Ile Pro Phe Ser Tyr Trp Gln Val Leu
100 105 110
att tat gat gat caa cgt tgg ctc ttc ggc tca atg atg tgc tct tta 384
Ile Tyr Asp Asp Gln Arg Trp Leu Phe Gly Ser Met Met Cys Ser Leu
115 120 125
tta gca ttc ctt caa gca atg gct gta ttt tta tca gct tgg aca ctt 432
Leu Ala Phe Leu Gln Ala Met Ala Val Phe Leu Ser Ala Trp Thr Leu
130 135 140
gtc gtt atc agt ttt gat cgg tgg atg gct ata atg ttc ctt tta act 480
Val Val Ile Ser Phe Asp Arg Trp Met Ala Ile Met Phe Leu Leu Thr
145 150 155 160
cca aat att cga att aca aga cga aga gct ctt tat ctg gta gct gcc 528
Pro Asn Ile Arg Ile Thr Arg Arg Arg Ala Leu Tyr Leu Val Ala Ala
165 170 175
acg tgg ata ttc agt atc cta atg gcg ttg ccg tta ctc ttc aca acg 576
Thr Trp Ile Phe Ser Ile Leu Met Ala Leu Pro Leu Leu Phe Thr Thr
180 185 190
aga ttt ttc gaa gac caa gac ggt tta ccg aat tgt gga gaa aat tgg 624
Arg Phe Phe Glu Asp Gln Asp Gly Leu Pro Asn Cys Gly Glu Asn Trp
195 200 205
acg tat ttt gga gat tct gga gaa caa gtg aga aaa gtg tat tct tca 672
Thr Tyr Phe Gly Asp Ser Gly Glu Gln Val Arg Lys Val Tyr Ser Ser
210 215 220
atg gtc tta att cta caa tat gtt gta cct caa gca gtt tta ata ata 720
Met Val Leu Ile Leu Gln Tyr Val Val Pro Gln Ala Val Leu Ile Ile
225 230 235 240
act tac aca cat att gga att aaa atg tgg aat agt cga gta cca gga 768
Thr Tyr Thr His Ile Gly Ile Lys Met Trp Asn Ser Arg Val Pro Gly
245 250 255
atg cag aat gga gca aca aag aaa atg atc gtt gat cga cat gaa agt 816
Met Gln Asn Gly Ala Thr Lys Lys Met Ile Val Asp Arg His Glu Ser
260 265 270
gtc aaa aag ctg gtc cca atg gtg att ctc att tcg gca ctc ttc gca 864
Val Lys Lys Leu Val Pro Met Val Ile Leu Ile Ser Ala Leu Phe Ala
275 280 285
ctt tgt tgg ctt cct tta ctt ata ctg atc aac gtc att cca gaa ttc 912
Leu Cys Trp Leu Pro Leu Leu Ile Leu Ile Asn Val Ile Pro Glu Phe
290 295 300
tat cca gat atc aac agt tgg gga tat att ctg tat ttg tgg tgg ttt 960
Tyr Pro Asp Ile Asn Ser Trp Gly Tyr Ile Leu Tyr Leu Trp Trp Phe
305 310 315 320
gct cat gga ctt gcc atg tct cat tca atg gtc aac cca att atc tat 1008
Ala His Gly Leu Ala Met Ser His Ser Met Val Asn Pro Ile Ile Tyr
325 330 335
ttc att cga aat gcc cgt ttc cgt gaa gga ttc tgt ttc ttc tct tca 1056
Phe Ile Arg Asn Ala Arg Phe Arg Glu Gly Phe Cys Phe Phe Ser Ser
340 345 350
aaa ctt ctt cca tgt ata tca ttt aaa gaa ctt cgt ctt tta act gat 1104
Lys Leu Leu Pro Cys Ile Ser Phe Lys Glu Leu Arg Leu Leu Thr Asp
355 360 365
aat act agc aga agt ttt cga aac cgt tca cgg ttt tct ggt gtc ata 1152
Asn Thr Ser Arg Ser Phe Arg Asn Arg Ser Arg Phe Ser Gly Val Ile
370 375 380
aat ccg act tca agc gat gaa aaa cct gct aca tcg ctt acg aga tat 1200
Asn Pro Thr Ser Ser Asp Glu Lys Pro Ala Thr Ser Leu Thr Arg Tyr
385 390 395 400
tca cga agt gga gtc ctt gac agg caa aca tgt cgt tcg gca cgt ttc 1248
Ser Arg Ser Gly Val Leu Asp Arg Gln Thr Cys Arg Ser Ala Arg Phe
405 410 415
ttc gaa gcc cga cca ctc gtc gtc gtc cga aac aac tct gcc aat tct 1296
Phe Glu Ala Arg Pro Leu Val Val Val Arg Asn Asn Ser Ala Asn Ser
420 425 430
ctc gcg tag 1305
Leu Ala
117
434
›PRT
Homo sapiens
117
Met Asp Glu Gly Gly Gly Ile Gly Ser Ser Leu Leu Ser Arg Ile Thr
1 5 10 15
Thr Thr Ala Ser Glu Ile Met Met Arg Asn Glu Pro Thr Thr Thr Glu
20 25 30
Asn Pro Ala Val Gln Glu Met Asn His Ile Tyr His Leu Thr Pro Ser
35 40 45
Met Lys Met Leu Cys Ile Leu Phe Tyr Ser Ile Leu Cys Val Cys Cys
50 55 60
Val Tyr Gly Asn Val Leu Val Ile Leu Val Ile Val Tyr Phe Lys Arg
65 70 75 80
Leu Arg Thr Ala Thr Asn Ile Leu Ile Leu Asn Leu Ala Val Ala Asp
85 90 95
Leu Leu Ile Ser Val Phe Cys Ile Pro Phe Ser Tyr Trp Gln Val Leu
100 105 110
Ile Tyr Asp Asp Gln Arg Trp Leu Phe Gly Ser Met Met Cys Ser Leu
115 120 125
Leu Ala Phe Leu Gln Ala Met Ala Val Phe Leu Ser Ala Trp Thr Leu
130 135 140
Val Val Ile Ser Phe Asp Arg Trp Met Ala Ile Met Phe Leu Leu Thr
145 150 155 160
Pro Asn Ile Arg Ile Thr Arg Arg Arg Ala Leu Tyr Leu Val Ala Ala
165 170 175
Thr Trp Ile Phe Ser Ile Leu Met Ala Leu Pro Leu Leu Phe Thr Thr
180 185 190
Arg Phe Phe Glu Asp Gln Asp Gly Leu Pro Asn Cys Gly Glu Asn Trp
195 200 205
Thr Tyr Phe Gly Asp Ser Gly Glu Gln Val Arg Lys Val Tyr Ser Ser
210 215 220
Met Val Leu Ile Leu Gln Tyr Val Val Pro Gln Ala Val Leu Ile Ile
225 230 235 240
Thr Tyr Thr His Ile Gly Ile Lys Met Trp Asn Ser Arg Val Pro Gly
245 250 255
Met Gln Asn Gly Ala Thr Lys Lys Met Ile Val Asp Arg His Glu Ser
260 265 270
Val Lys Lys Leu Val Pro Met Val Ile Leu Ile Ser Ala Leu Phe Ala
275 280 285
Leu Cys Trp Leu Pro Leu Leu Ile Leu Ile Asn Val Ile Pro Glu Phe
290 295 300
Tyr Pro Asp Ile Asn Ser Trp Gly Tyr Ile Leu Tyr Leu Trp Trp Phe
305 310 315 320
Ala His Gly Leu Ala Met Ser His Ser Met Val Asn Pro Ile Ile Tyr
325 330 335
Phe Ile Arg Asn Ala Arg Phe Arg Glu Gly Phe Cys Phe Phe Ser Ser
340 345 350
Lys Leu Leu Pro Cys Ile Ser Phe Lys Glu Leu Arg Leu Leu Thr Asp
355 360 365
Asn Thr Ser Arg Ser Phe Arg Asn Arg Ser Arg Phe Ser Gly Val Ile
370 375 380
Asn Pro Thr Ser Ser Asp Glu Lys Pro Ala Thr Ser Leu Thr Arg Tyr
385 390 395 400
Ser Arg Ser Gly Val Leu Asp Arg Gln Thr Cys Arg Ser Ala Arg Phe
405 410 415
Phe Glu Ala Arg Pro Leu Val Val Val Arg Asn Asn Ser Ala Asn Ser
420 425 430
Leu Ala
118
Xaa Gly Leu Gly Pro Arg Pro Leu Arg Phe
1 5 10
119
9
›PRT
Artificial Sequence
Description of Artificial Sequence Synthetic
analog
119
Xaa Asp Val Pro Gly Val Leu Arg Phe
1 5
120
Asp Phe Asp Gly Ala Met Pro Gly Val Leu Arg Phe
1 5 10
121
Glu Ile Pro Gly Val Leu Arg Phe
1 5
122
Ser Asp Asn Phe Met Arg Phe
1 5
123
Pro Asp Asn Phe Met Arg Phe
1 5
124
0
›PRT
synthetic
124
000
125
Tyr Gly Leu Gly Pro Arg Pro Leu Arg Phe
1 5 10
126
Tyr Leu Arg Phe
1
127
Ser Asp Arg Asn Phe Leu Arg Phe
1 5
128
Phe Leu Arg Phe
1
129
Phe Met Arg Phe
1
130
0
›PRT
synthetic
130
000
131
Asp Pro Ser Phe Leu Arg Phe
1 5
132
0
›PRT
synthetic
132
000
133
Ala Gly Pro Arg Phe Ile Arg Phe
1 5
134
0
›PRT
synthetic
134
000
135
Lys Pro Asn Phe Leu Arg Tyr
1 5
136
Ala Arg Gly Pro Gln Leu Arg Leu Arg Phe
1 5 10
137
Tyr Ile Arg Phe
1
138
0
›PRT
synthetic
138
000
139
Ala Gly Ala Lys Phe Ile Arg Phe
1 5
140
0
›PRT
synthetic
140
000
141
Lys His Glu Tyr Leu Arg Phe
1 5
142
0
›PRT
synthetic
142
000
143
Ser Pro Leu Gly Thr Met Arg Phe
1 5
144
0
›PRT
synthetic
144
000
145
0
›PRT
synthetic
145
000
146
0
›PRT
synthetic
146
000
147
0
›PRT
synthetic
147
000
148
0
›PRT
synthetic
148
000
149
0
›PRT
synthetic
149
000
150
Ala Glu Gly Leu Ser Ser Pro Leu Ile Arg Phe
1 5 10
151
Phe Asp Arg Asp Phe Met His Phe
1 5
152
Val Leu Arg Phe
1
153
Gly Val Leu Arg Phe
1 5
154
Gly Val Leu Arg Phe
1 5
155
Pro Gly Val Leu Arg Phe
1 5
156
Val Pro Gly Val Leu Arg Phe
1 5
157
Val Pro Gly Val Leu Arg Phe
1 5
158
Tyr Asp Val Pro Gly Val Leu Arg Phe
1 5
159
Tyr Val Pro Gly Val Leu Arg Phe
1 5
160
Xaa Val Pro Gly Val Leu Arg Phe
1 5
161
Tyr Pro Gly Val Leu Arg Phe
1 5
162
Xaa Pro Gly Val Leu Arg Phe
1 5
163
Tyr Gly Val Leu Arg Phe
1 5
164
Xaa Gly Val Leu Arg Phe
1 5
165
Lys Pro Asn Phe Ile Arg Phe
1 5
166
Ser Gln Pro Asn Phe Leu Arg Phe
1 5
167
Leu Gln Pro Asn Phe Leu Arg Phe
1 5
168
0
›PRT
Artificial Sequence
Description of Artificial Sequence synthetic
analog
168
000
169
0
›PRT
Artificial Sequence
Description of Artificial Sequence synthetic
analog
169
000
170
0
›PRT
Artificial Sequence
Description of Artificial Sequence synthetic
analog
170
000
171
25
›DNA
Artificial Sequence
SCN162
171
acgtttaaga gctctcaaat cccat 25
172
25
›DNA
Artificial Sequence
SCN160
172
ccagagctca tcaaaactca agaat 25
173
31
›DNA
Artificial Sequence
DEL-1850 (CEGPCR12c_F.2)
173
gccgccatgt cgaatgatct cgtgccttca g 31
174
25
›DNA
Artificial Sequence
SCN189
174
ttacaattta aaactaggtg cttct 25
175
31
›DNA
Artificial Sequence
SCN199
175
gccgccatga acttttcggc caccgattcg a 31
176
atg gcc cgt gca gtc aac ata tct cca ttc gct tct tat aca gtc gta 48
Met Ala Arg Ala Val Asn Ile Ser Pro Phe Ala Ser Tyr Thr Val Val
1 5 10 15
cca att acc tcg gca tgg cca ccg gac gac ttg aaa gcc gat cga gtt 96
Pro Ile Thr Ser Ala Trp Pro Pro Asp Asp Leu Lys Ala Asp Arg Val
20 25 30
caa ctt gcc agt ttc acg act gga aaa gct ctg cta gcc att gcc att 144
Gln Leu Ala Ser Phe Thr Thr Gly Lys Ala Leu Leu Ala Ile Ala Ile
35 40 45
ctt gca atg att att atg aca acg gtc ggc aat gcg ctg gtt tgc tta 192
Leu Ala Met Ile Ile Met Thr Thr Val Gly Asn Ala Leu Val Cys Leu
50 55 60
gca gtt tta ctg gta cga aag ttg aaa cat cct caa aat ttt ctt ttg 240
Ala Val Leu Leu Val Arg Lys Leu Lys His Pro Gln Asn Phe Leu Leu
65 70 75 80
gtc tca ctg gca gtg gcg gac ttt ttc gtg ggg ctc gtt gtg atg cca 288
Val Ser Leu Ala Val Ala Asp Phe Phe Val Gly Leu Val Val Met Pro
85 90 95
ctt gca ctg att gac ctt ctc ttc gat aaa tgg cca ctt gga agc acc 336
Leu Ala Leu Ile Asp Leu Leu Phe Asp Lys Trp Pro Leu Gly Ser Thr
100 105 110
atg tgt tcg gtg tac aca aca tcc gac cta acg cta tgc act gct tca 384
Met Cys Ser Val Tyr Thr Thr Ser Asp Leu Thr Leu Cys Thr Ala Ser
115 120 125
att gtg aat ctg tgt gca ata tca gtt gat aga tat ctg gtg att tca 432
Ile Val Asn Leu Cys Ala Ile Ser Val Asp Arg Tyr Leu Val Ile Ser
130 135 140
agt cca ctt cgg tat tct gca aaa cga aca act aaa cgt atc atg atg 480
Ser Pro Leu Arg Tyr Ser Ala Lys Arg Thr Thr Lys Arg Ile Met Met
145 150 155 160
tat att gcc tgt gta tgg atc atc gct gcc att gtg agc att tcg tcg 528
Tyr Ile Ala Cys Val Trp Ile Ile Ala Ala Ile Val Ser Ile Ser Ser
165 170 175
cac att atc gca aat ctt ctg aac gac ggc act tat gtc gat gac acg 576
His Ile Ile Ala Asn Leu Leu Asn Asp Gly Thr Tyr Val Asp Asp Thr
180 185 190
gga act tgt cag gtc atc cct cat ttc atc tat caa agc tat gct aca 624
Gly Thr Cys Gln Val Ile Pro His Phe Ile Tyr Gln Ser Tyr Ala Thr
195 200 205
att atc tca ttc tat gcg cca aca ttt atc atg gtc atc ttg aac atc 672
Ile Ile Ser Phe Tyr Ala Pro Thr Phe Ile Met Val Ile Leu Asn Ile
210 215 220
aaa att tgg cga gca gca aaa cga ttg gca gct caa gac agg ttg atg 720
Lys Ile Trp Arg Ala Ala Lys Arg Leu Ala Ala Gln Asp Arg Leu Met
225 230 235 240
tct cac tgt aac tct gtg gat gcc tcc gaa cgg cca cga aat gga tca 768
Ser His Cys Asn Ser Val Asp Ala Ser Glu Arg Pro Arg Asn Gly Ser
245 250 255
gct gag aca aag gat ttc ctg aat gag aaa gaa aca att gat gtt ccg 816
Ala Glu Thr Lys Asp Phe Leu Asn Glu Lys Glu Thr Ile Asp Val Pro
260 265 270
aaa aaa gag cgg gcc aat tca aca aac tcg agg ctt ttc aag ttg gaa 864
Lys Lys Glu Arg Ala Asn Ser Thr Asn Ser Arg Leu Phe Lys Leu Glu
275 280 285
cgt aaa tat ctt cat cga ccc agt gca ttc ttt tcg gct gca aaa gga 912
Arg Lys Tyr Leu His Arg Pro Ser Ala Phe Phe Ser Ala Ala Lys Gly
290 295 300
cca ttg att cgg cag acg gag aaa agt gaa tgc aaa gcc cgg aaa aca 960
Pro Leu Ile Arg Gln Thr Glu Lys Ser Glu Cys Lys Ala Arg Lys Thr
305 310 315 320
ttg ggg gtc atc atg tca gtg ttc att att tgc tgg ctc cca ttc ttc 1008
Leu Gly Val Ile Met Ser Val Phe Ile Ile Cys Trp Leu Pro Phe Phe
325 330 335
att ctt gcc ata ttc aag tct ttc ggc atg tgg atc ccg gat tgg ttg 1056
Ile Leu Ala Ile Phe Lys Ser Phe Gly Met Trp Ile Pro Asp Trp Leu
340 345 350
gat ctg cta gca ctg tgg tta ggc tac tca aat agc aca tta aat cca 1104
Asp Leu Leu Ala Leu Trp Leu Gly Tyr Ser Asn Ser Thr Leu Asn Pro
355 360 365
ttg atc tac tgc aag tac aac aaa gag ttt cga ata cca ttc cgt gaa 1152
Leu Ile Tyr Cys Lys Tyr Asn Lys Glu Phe Arg Ile Pro Phe Arg Glu
370 375 380
atg ctt gct tgt cgc tgt gct act tta cag act gta atg cga caa caa 1200
Met Leu Ala Cys Arg Cys Ala Thr Leu Gln Thr Val Met Arg Gln Gln
385 390 395 400
agc ttc acc agc cgc tac gga cct cct gtg tga 1233
Ser Phe Thr Ser Arg Tyr Gly Pro Pro Val
405 410
177
410
›PRT
C. elegans
177
Met Ala Arg Ala Val Asn Ile Ser Pro Phe Ala Ser Tyr Thr Val Val
1 5 10 15
Pro Ile Thr Ser Ala Trp Pro Pro Asp Asp Leu Lys Ala Asp Arg Val
20 25 30
Gln Leu Ala Ser Phe Thr Thr Gly Lys Ala Leu Leu Ala Ile Ala Ile
35 40 45
Leu Ala Met Ile Ile Met Thr Thr Val Gly Asn Ala Leu Val Cys Leu
50 55 60
Ala Val Leu Leu Val Arg Lys Leu Lys His Pro Gln Asn Phe Leu Leu
65 70 75 80
Val Ser Leu Ala Val Ala Asp Phe Phe Val Gly Leu Val Val Met Pro
85 90 95
Leu Ala Leu Ile Asp Leu Leu Phe Asp Lys Trp Pro Leu Gly Ser Thr
100 105 110
Met Cys Ser Val Tyr Thr Thr Ser Asp Leu Thr Leu Cys Thr Ala Ser
115 120 125
Ile Val Asn Leu Cys Ala Ile Ser Val Asp Arg Tyr Leu Val Ile Ser
130 135 140
Ser Pro Leu Arg Tyr Ser Ala Lys Arg Thr Thr Lys Arg Ile Met Met
145 150 155 160
Tyr Ile Ala Cys Val Trp Ile Ile Ala Ala Ile Val Ser Ile Ser Ser
165 170 175
His Ile Ile Ala Asn Leu Leu Asn Asp Gly Thr Tyr Val Asp Asp Thr
180 185 190
Gly Thr Cys Gln Val Ile Pro His Phe Ile Tyr Gln Ser Tyr Ala Thr
195 200 205
Ile Ile Ser Phe Tyr Ala Pro Thr Phe Ile Met Val Ile Leu Asn Ile
210 215 220
Lys Ile Trp Arg Ala Ala Lys Arg Leu Ala Ala Gln Asp Arg Leu Met
225 230 235 240
Ser His Cys Asn Ser Val Asp Ala Ser Glu Arg Pro Arg Asn Gly Ser
245 250 255
Ala Glu Thr Lys Asp Phe Leu Asn Glu Lys Glu Thr Ile Asp Val Pro
260 265 270
Lys Lys Glu Arg Ala Asn Ser Thr Asn Ser Arg Leu Phe Lys Leu Glu
275 280 285
Arg Lys Tyr Leu His Arg Pro Ser Ala Phe Phe Ser Ala Ala Lys Gly
290 295 300
Pro Leu Ile Arg Gln Thr Glu Lys Ser Glu Cys Lys Ala Arg Lys Thr
305 310 315 320
Leu Gly Val Ile Met Ser Val Phe Ile Ile Cys Trp Leu Pro Phe Phe
325 330 335
Ile Leu Ala Ile Phe Lys Ser Phe Gly Met Trp Ile Pro Asp Trp Leu
340 345 350
Asp Leu Leu Ala Leu Trp Leu Gly Tyr Ser Asn Ser Thr Leu Asn Pro
355 360 365
Leu Ile Tyr Cys Lys Tyr Asn Lys Glu Phe Arg Ile Pro Phe Arg Glu
370 375 380
Met Leu Ala Cys Arg Cys Ala Thr Leu Gln Thr Val Met Arg Gln Gln
385 390 395 400
Ser Phe Thr Ser Arg Tyr Gly Pro Pro Val
405 410
178
1749
›DNA
C. elegans
CEGPCR25 (Y39A3B.5)
178
atg tca acg gta cat cag gag gaa ata tgc cgg ttt cgt ggt aca aca 48
Met Ser Thr Val His Gln Glu Glu Ile Cys Arg Phe Arg Gly Thr Thr
1 5 10 15
gaa aat tac aca atc gcc gtg acg ttt ttc atg att ttt ttg cta tca 96
Glu Asn Tyr Thr Ile Ala Val Thr Phe Phe Met Ile Phe Leu Leu Ser
20 25 30
gtc gtc gga aat tcg gtg gtt ttg ata gtg att att aag caa cgt gca 144
Val Val Gly Asn Ser Val Val Leu Ile Val Ile Ile Lys Gln Arg Ala
35 40 45
atg cga tca ata acc aat ata tat tta atg aat ctg gcg gca tcg gat 192
Met Arg Ser Ile Thr Asn Ile Tyr Leu Met Asn Leu Ala Ala Ser Asp
50 55 60
atg atg tta tca gtg gtt tgt atg ccg ccg aca ctt gtc tcc atg gtt 240
Met Met Leu Ser Val Val Cys Met Pro Pro Thr Leu Val Ser Met Val
65 70 75 80
atg aat tgt tgg atg ttc ggc aat tat atg tgc aaa att ttg gct tat 288
Met Asn Cys Trp Met Phe Gly Asn Tyr Met Cys Lys Ile Leu Ala Tyr
85 90 95
ttg caa ccc gta gta gtc act gct tca gcg tac acc cta gca gta atc 336
Leu Gln Pro Val Val Val Thr Ala Ser Ala Tyr Thr Leu Ala Val Ile
100 105 110
gca ttc gag agg tac ttt gca att tgc aag ccg ctt cac tca agg att 384
Ala Phe Glu Arg Tyr Phe Ala Ile Cys Lys Pro Leu His Ser Arg Ile
115 120 125
tgg caa acc aga tcc cac gcc tac gca atg atc aca ctt gtg tgg gtg 432
Trp Gln Thr Arg Ser His Ala Tyr Ala Met Ile Thr Leu Val Trp Val
130 135 140
atc gcc att gcc gcc aat att ctg atg ctt ttt atg tac gaa caa caa 480
Ile Ala Ile Ala Ala Asn Ile Leu Met Leu Phe Met Tyr Glu Gln Gln
145 150 155 160
acg tat agc tcg aat gga tac acg tgt gct cca att cat ccg ccg att 528
Thr Tyr Ser Ser Asn Gly Tyr Thr Cys Ala Pro Ile His Pro Pro Ile
165 170 175
tat cat ttt gct tat cag gaa aat cgg aaa ttt cac tta aaa tcg ata 576
Tyr His Phe Ala Tyr Gln Glu Asn Arg Lys Phe His Leu Lys Ser Ile
180 185 190
ata tat ttt cag gta tac atg acc gtc gtc cta ctt gta atc ccg cta 624
Ile Tyr Phe Gln Val Tyr Met Thr Val Val Leu Leu Val Ile Pro Leu
195 200 205
gtt gtg atg gcg ggt ttg tac ggc aac gtc att acc tcc cta aaa tcc 672
Val Val Met Ala Gly Leu Tyr Gly Asn Val Ile Thr Ser Leu Lys Ser
210 215 220
ggc atc aaa ctg gaa atc gcc tct gtg gat ccg ccg ctc gcc acc gcc 720
Gly Ile Lys Leu Glu Ile Ala Ser Val Asp Pro Pro Leu Ala Thr Ala
225 230 235 240
aca acc aca ggt gct aaa aat tta ggc tcc cac tct gac tcg gct cac 768
Thr Thr Thr Gly Ala Lys Asn Leu Gly Ser His Ser Asp Ser Ala His
245 250 255
ctt ttg ttg aac aat gtg ctc gtc ggt agc tcg caa tta gct cgt gcc 816
Leu Leu Leu Asn Asn Val Leu Val Gly Ser Ser Gln Leu Ala Arg Ala
260 265 270
acg tcg tgc atc gcg ttg aac acg ttt tcc tgc aac aac acc act aac 864
Thr Ser Cys Ile Ala Leu Asn Thr Phe Ser Cys Asn Asn Thr Thr Asn
275 280 285
ccc ttc agt acc ctg aca ccg ccg ccg ccg att cag cag aaa aat cga 912
Pro Phe Ser Thr Leu Thr Pro Pro Pro Pro Ile Gln Gln Lys Asn Arg
290 295 300
tcg aag ccg caa ctt ctg cag ctg ccc gga aaa atc gat aat ttc gaa 960
Ser Lys Pro Gln Leu Leu Gln Leu Pro Gly Lys Ile Asp Asn Phe Glu
305 310 315 320
gag ttc cgc ctt cag tgc ctg tcc gac tgt cgt agt gat gga gtg ttg 1008
Glu Phe Arg Leu Gln Cys Leu Ser Asp Cys Arg Ser Asp Gly Val Leu
325 330 335
ttt ccc ccg cca gca att gtt gcg tcg atg acc gac gag caa aag tta 1056
Phe Pro Pro Pro Ala Ile Val Ala Ser Met Thr Asp Glu Gln Lys Leu
340 345 350
tcg ttc tgg aat aag ctg tcg aac aag ctg act ttt agt cag caa gat 1104
Ser Phe Trp Asn Lys Leu Ser Asn Lys Leu Thr Phe Ser Gln Gln Asp
355 360 365
aaa acc gtg caa cac cca aac ttt ggg cat cgc aaa tcc gac aca tca 1152
Lys Thr Val Gln His Pro Asn Phe Gly His Arg Lys Ser Asp Thr Ser
370 375 380
att tgc ctc gaa aat cct agt tta agg tct acg cac act cag aaa agt 1200
Ile Cys Leu Glu Asn Pro Ser Leu Arg Ser Thr His Thr Gln Lys Ser
385 390 395 400
gca atg gca aag caa aga gtg att aaa atg ctc att gtt gtc gta att 1248
Ala Met Ala Lys Gln Arg Val Ile Lys Met Leu Ile Val Val Val Ile
405 410 415
atc ttt ttc tgt tgt tgg acg cct tcc tac atc tgg tgg ttg tta ttg 1296
Ile Phe Phe Cys Cys Trp Thr Pro Ser Tyr Ile Trp Trp Leu Leu Leu
420 425 430
atc gcg gga gac tcg ttt caa agc ctc aac tta tct gtc tgg aac agc 1344
Ile Ala Gly Asp Ser Phe Gln Ser Leu Asn Leu Ser Val Trp Asn Ser
435 440 445
gat atc aac aca ttc atc act ctt ctt acc tat att tct tca tgc acc 1392
Asp Ile Asn Thr Phe Ile Thr Leu Leu Thr Tyr Ile Ser Ser Cys Thr
450 455 460
aat ccc atc aca tat tgc ttc ctg aac aaa aaa ttt agg aat gcg gtg 1440
Asn Pro Ile Thr Tyr Cys Phe Leu Asn Lys Lys Phe Arg Asn Ala Val
465 470 475 480
tat gca aca ttt ggc agg aag aaa aat atg cga cat cat ttt cag aag 1488
Tyr Ala Thr Phe Gly Arg Lys Lys Asn Met Arg His His Phe Gln Lys
485 490 495
gta tat ttt gat att ttt gaa tta ttg ata ttt ttt aaa atc aat tat 1536
Val Tyr Phe Asp Ile Phe Glu Leu Leu Ile Phe Phe Lys Ile Asn Tyr
500 505 510
caa aaa tca ata aaa tct ccg gtt tca gtt ttt tct aac tat ctg ttt 1584
Gln Lys Ser Ile Lys Ser Pro Val Ser Val Phe Ser Asn Tyr Leu Phe
515 520 525
ttc ggc ttt tcg gcg ccg cta att tct gtt aac ctt agt ttg att ttg 1632
Phe Gly Phe Ser Ala Pro Leu Ile Ser Val Asn Leu Ser Leu Ile Leu
530 535 540
ttc agg tct aca ttc cgg tca acg gcg gaa cta att aca cta aca aag 1680
Phe Arg Ser Thr Phe Arg Ser Thr Ala Glu Leu Ile Thr Leu Thr Lys
545 550 555 560
gag agc cga gac aaa tta tta aca ggt act aac aca cac aaa ttt gta 1728
Glu Ser Arg Asp Lys Leu Leu Thr Gly Thr Asn Thr His Lys Phe Val
565 570 575
gtg att tcg ctt ttt gtc tag 1749
Val Ile Ser Leu Phe Val
580
179
582
›PRT
C. elegans
179
Met Ser Thr Val His Gln Glu Glu Ile Cys Arg Phe Arg Gly Thr Thr
1 5 10 15
Glu Asn Tyr Thr Ile Ala Val Thr Phe Phe Met Ile Phe Leu Leu Ser
20 25 30
Val Val Gly Asn Ser Val Val Leu Ile Val Ile Ile Lys Gln Arg Ala
35 40 45
Met Arg Ser Ile Thr Asn Ile Tyr Leu Met Asn Leu Ala Ala Ser Asp
50 55 60
Met Met Leu Ser Val Val Cys Met Pro Pro Thr Leu Val Ser Met Val
65 70 75 80
Met Asn Cys Trp Met Phe Gly Asn Tyr Met Cys Lys Ile Leu Ala Tyr
85 90 95
Leu Gln Pro Val Val Val Thr Ala Ser Ala Tyr Thr Leu Ala Val Ile
100 105 110
Ala Phe Glu Arg Tyr Phe Ala Ile Cys Lys Pro Leu His Ser Arg Ile
115 120 125
Trp Gln Thr Arg Ser His Ala Tyr Ala Met Ile Thr Leu Val Trp Val
130 135 140
Ile Ala Ile Ala Ala Asn Ile Leu Met Leu Phe Met Tyr Glu Gln Gln
145 150 155 160
Thr Tyr Ser Ser Asn Gly Tyr Thr Cys Ala Pro Ile His Pro Pro Ile
165 170 175
Tyr His Phe Ala Tyr Gln Glu Asn Arg Lys Phe His Leu Lys Ser Ile
180 185 190
Ile Tyr Phe Gln Val Tyr Met Thr Val Val Leu Leu Val Ile Pro Leu
195 200 205
Val Val Met Ala Gly Leu Tyr Gly Asn Val Ile Thr Ser Leu Lys Ser
210 215 220
Gly Ile Lys Leu Glu Ile Ala Ser Val Asp Pro Pro Leu Ala Thr Ala
225 230 235 240
Thr Thr Thr Gly Ala Lys Asn Leu Gly Ser His Ser Asp Ser Ala His
245 250 255
Leu Leu Leu Asn Asn Val Leu Val Gly Ser Ser Gln Leu Ala Arg Ala
260 265 270
Thr Ser Cys Ile Ala Leu Asn Thr Phe Ser Cys Asn Asn Thr Thr Asn
275 280 285
Pro Phe Ser Thr Leu Thr Pro Pro Pro Pro Ile Gln Gln Lys Asn Arg
290 295 300
Ser Lys Pro Gln Leu Leu Gln Leu Pro Gly Lys Ile Asp Asn Phe Glu
305 310 315 320
Glu Phe Arg Leu Gln Cys Leu Ser Asp Cys Arg Ser Asp Gly Val Leu
325 330 335
Phe Pro Pro Pro Ala Ile Val Ala Ser Met Thr Asp Glu Gln Lys Leu
340 345 350
Ser Phe Trp Asn Lys Leu Ser Asn Lys Leu Thr Phe Ser Gln Gln Asp
355 360 365
Lys Thr Val Gln His Pro Asn Phe Gly His Arg Lys Ser Asp Thr Ser
370 375 380
Ile Cys Leu Glu Asn Pro Ser Leu Arg Ser Thr His Thr Gln Lys Ser
385 390 395 400
Ala Met Ala Lys Gln Arg Val Ile Lys Met Leu Ile Val Val Val Ile
405 410 415
Ile Phe Phe Cys Cys Trp Thr Pro Ser Tyr Ile Trp Trp Leu Leu Leu
420 425 430
Ile Ala Gly Asp Ser Phe Gln Ser Leu Asn Leu Ser Val Trp Asn Ser
435 440 445
Asp Ile Asn Thr Phe Ile Thr Leu Leu Thr Tyr Ile Ser Ser Cys Thr
450 455 460
Asn Pro Ile Thr Tyr Cys Phe Leu Asn Lys Lys Phe Arg Asn Ala Val
465 470 475 480
Tyr Ala Thr Phe Gly Arg Lys Lys Asn Met Arg His His Phe Gln Lys
485 490 495
Val Tyr Phe Asp Ile Phe Glu Leu Leu Ile Phe Phe Lys Ile Asn Tyr
500 505 510
Gln Lys Ser Ile Lys Ser Pro Val Ser Val Phe Ser Asn Tyr Leu Phe
515 520 525
Phe Gly Phe Ser Ala Pro Leu Ile Ser Val Asn Leu Ser Leu Ile Leu
530 535 540
Phe Arg Ser Thr Phe Arg Ser Thr Ala Glu Leu Ile Thr Leu Thr Lys
545 550 555 560
Glu Ser Arg Asp Lys Leu Leu Thr Gly Thr Asn Thr His Lys Phe Val
565 570 575
Val Ile Ser Leu Phe Val
580
180
9
›PRT
Artificial Sequence
Description of Artificial Sequence synthetic
180
Phe Asp Asp Tyr Gly His Leu Arg Pro
1 5
›Tables in the description — 65
| ID | Wormpep | ||
| NO | Name | sequence | Difference(s) |
| 1 | CEGPCR1a 1 | AC7.1 | 94 bp deletion between nucleotides 594-595 of SEQ |
| ID NO:1 | |||
| 3 | CEGPCR1f | AC7.1 | 94 bp deletion between nucleotides 594-595 of SEQ |
| ID NO:3; addition of nucleotides 868-912 of SEQ ID | |||
| NO:3 | |||
| 5 | CEGPCR12c | F41E7.3 | 90 bp addition at 5′ end of SEQ ID NO:5; deletion of |
| 52 bp between nucleotides 173-174 of SEQ ID NO:5; | |||
| deletion of 6 bp between nucleotides 1115-1116 of | |||
| SEQ ID NO:5 | |||
| 7 | CEGPCR12h | F41E7.3 | 90 bp addition at 5′ end of SEQ ID NO:7; deletion of |
| 52 bp between nucleotides 173-174 of SEQ ID NO:7; | |||
| deletion of 12 bp between nucleotides 1115-1116 of | |||
| SEQ ID NO:7 | |||
| 9 | CEGPCR12u | F41E7.3 | 90 bp addition at 5′ end of SEQ ID NO:9; deletion of |
| 52 bp between nucleotides 173-174 of SEQ ID NO:9 | |||
| 11 | CEGPCR12v | F41E7.3 | 90 bp addition at 5′ end of SEQ ID NO:11; deletion |
| of 52 bp between nucleotides 173-174 of SEQ ID | |||
| NO:11; addition of 97 bp between nucleotides 1115- | |||
| 1212 of SEQ ID NO:11 | |||
| 1 The “CEGPCR” labeling system used herein corresponds to the “PNU” labeling system used in related USSN 60/162,523 (e.g., CEGPCR1a corresponds to PNU1a). | |||
| SEQ | |||
| ID | Wormpep | ||
| NO | Name | sequence | Difference(s) |
| 17 | CEGPCR18a | Y54E2A.1 | deletion of 58 bp between nucleotides 1089-1090 of |
| SEQ ID NO: 17; deletion of 32 bp between | |||
| nucleotides 1331-1332 of SEQ ID NO:17 | |||
| 106 | CEGPCR19.1 | Y58G8a.1 | |
| 104 | CEGPCR19.2 | Y58G8a.2 | |
| 114 | CECPCR24a | Y59H11A | |
| 1.a-01 | |||
| 116 | CEGPCR24b | Y59H11A | |
| 1.a-02 |
| ID | Wormpep | ||
| NO | Name | sequence | Difference(s) |
| CEGPCR1a | |||
| cegpcr1(a,f), | |||
| cegpcr4, | |||
| cegpcr5, | |||
| cegpcr11, | |||
| cegpcr12(c,h, | |||
| u,v), | |||
| cegpcr13, | |||
| cegpcr14, | |||
| cegpcr18, | |||
| cegpcr19(L,S), | |||
| cegpcr22, | |||
| cegpcr24(a,b) | |||
| 21 | CEGPCR4 | C16D6.2 | deletion of 55 bp between nucleotides 1070- |
| 1071 of SEQ ID NO:21 | |||
| CEGPCR5 | |||
| CEGPCR11 | |||
| CEGPCR12c | |||
| CEGPCR12h | |||
| CEGPCR12u | |||
| CEGPCR12v | |||
| 13 | CEGPCR13 | T05A1.1 | deletion of 85 bp between nucleotides 753-754 |
| of SEQ ID NO:13; addition of 226 bp between | |||
| nucleotides 926-1152 of SEQ ID NO:13 | |||
| 15 | CEGPCR14 | ZC412.1 | deletion of 92 bp between nucleotides 1277- |
| 1278 of SEQ ID NO:15 | |||
| 17 | CEGPCR18a | Y54E2A.1a | |
| 106 | CEGPCR19.1 | Y58G8a.1 | |
| 104 | CEGPCR19.2 | Y58C8a.2 | |
| 110 | CEGPCR22 | C06G4.5 | |
| 114 | CEGPCR24a | Y59H11A | |
| L.a | |||
| 116 | CEGPCR24b | Y59H11A | |
| L.b |
| SIDE CHAIN CHARACTERISTIC | AMINO ACID | ||||||
| Aliphatic | Non-polar | G | A | P | I | L | V |
| Polar-uncharged | C | S | T | M | N | Q | |
| Polar-charged | D | E | K | R | |||
| Aromatic | H | F | W | Y | |||
| Other | N | Q | D | E |
| CHARACTERISTIC | AMINO ACID |
| Non-polar (hydrophobic) | |
| A. Aliphatic: | A L I V P |
| B. Aromatic: | F W Y |
| C. Sulfur-containing: | M |
| D. Borderline: | G |
| Uncharged-polar | |
| A. Hydroxyl: | S T Y |
| B. Amides: | N Q |
| C. Sulfhydryl: | C |
| D. Borderline: | G |
| Positively Charged (Basic): | K R H |
| Negatively Charged (Acidic): | D E |
| CHROMOSOME | STRUCTURALLY | PEPTIDES PREDICTED |
| NUMBER | CHARACTERIZED FaRPs | FROM GENE |
| FLP-1 | SDPNFLRFa | KPNFMRYa |
| SADPNFLRFa | AGSDPNFLRFa | |
| SQPNFLRFa | ||
| SQPNFLRFa | ASGDPNFLRFa | |
| ASGDPNFLRFa | SDPNFLRFa | |
| AAADPNFLRFa | AAADPNFLRFa | |
| SADPNFLRFa | ||
| KPNFLRFa | ||
| FLP-2 | LRGEPIRFa | |
| SPREPIRFa | ||
| FLP-3 | SPLGTMRFa | |
| TPLGTMRFa | ||
| SAEPFGTMRFa | ||
| NPENDTPFGTMRFa | ||
| ASEDALFGTMRFa | ||
| EDGNAPFGTMRFa | ||
| EAEEPLGTMRFa | ||
| SADDSAPFGTMRFa | ||
| NPLGTMRFa | ||
| FLP-4 | PTFIRFa | |
| ASPSFIRFa | ||
| FLP-5 | APKPKFIRFa | |
| AGAKFRFa | ||
| GAKFIRFa | ||
| FLP-6 | KSAYMRFa | (6 copies) KSAYMRFa |
| FLP-7 | (2 copies) | |
| TPMQRSSMVRFa | ||
| (3 copies) | ||
| SPMQRSSMVRFa | ||
| SPMERSAMVRFa | ||
| SPMDRSKMVRFa | ||
| FLP-8 | (3 copies) KNEFIRFa | |
| FLP-9 | KPSFVRFa | (2 copies) KPSFVRFa |
| FLP-10 | QPKARSGYIRFa | |
| FLP-11 | AMRNALVRFa | |
| ASGGMRNALVRFa | ||
| NGAPQPFVRFa | ||
| FLP-12 | RNKFEFTRFa | |
| FLP-13 | SDRPTRAMDSPLIRFa | |
| (2 copies) | ||
| AADGAPLIRFa | ||
| (2 copies) | ||
| APEASPFIRFa | ||
| ASPSAPLIRFa | ||
| SPSAVPLIRFa | ||
| SAAAPLIRFa | ||
| ASSAPLIRFa | ||
| FLP-14 | (4 copies) KHEYLRFa | |
| FLP-15 | GGPQGPLRFa | |
| GPSGPLRFa | ||
| FLP-16 | AQTFVRFa | |
| GQTFVRFa | ||
| FLP-17 | (2 copies) KSAFVRFa | |
| KSQYIRFa | ||
| FLP-18 | DFDGAMPGVLRFa | |
| DMPGVLRFa | ||
| KSVPGVLRFa | ||
| SVPGVLRFa | ||
| EIPGVLRFa | ||
| SEVPGVLRFa | ||
| DVPGVLRFa | ||
| SVPGVLRFa | ||
| OTHER | TKFQDFLRFa | |
| PUTATIVE | AMRNSLVRFa | |
| FLP GENES | DYDFVRFa | |
| DGFVRFa | ||
| AFFKNVLRFa |
| TM1 | TM2 | TM3 | TM4 | TMS | TM6 | TM7 | ||||||||
| SIN | start | end | start | end | start | end | start | end | start | end | start | end | start | end |
| CEGPCR1a | 47 | 71 | 83 | 109 | 121 | 142 | 161 | 181 | 213 | 233 | 262 | 282 | 298 | 321 |
| CEGPCR1f | 47 | 71 | 83 | 109 | 121 | 142 | 161 | 181 | 213 | 233 | 277 | 297 | 313 | 336 |
| CEGPCR2 | 42 | 66 | 78 | 104 | 116 | 137 | 158 | 178 | 204 | 224 | 320 | 340 | 351 | 374 |
| CEGPCR3 | 26 | 50 | 62 | 88 | 100 | 121 | 140 | 160 | 196 | 216 | 252 | 272 | 285 | 308 |
| CEGPCR4 | 28 | 52 | 64 | 90 | 102 | 123 | 142 | 162 | 192 | 212 | 256 | 276 | 293 | 316 |
| CEGPCR5 | 40 | 64 | 76 | 102 | 114 | 135 | 154 | 174 | 212 | 232 | 222 | 292 | 305 | 328 |
| CEGPCR6 | 12 | 36 | 48 | 74 | 85 | 106 | 126 | 146 | 178 | 198 | 229 | 249 | 263 | 286 |
| CEGPCR7 | 27 | 51 | 63 | 89 | 101 | 122 | 141 | 161 | 192 | 212 | 278 | 298 | 322 | 345 |
| CEGPCR8 | 30 | 54 | 66 | 92 | 104 | 125 | 146 | 166 | 194 | 214 | 244 | 264 | 283 | 306 |
| CEGPCR9 | 51 | 75 | 87 | 113 | 125 | 146 | 167 | 187 | 239 | 259 | 330 | 350 | 368 | 391 |
| CEGPCR11 | 25 | 49 | 61 | 87 | 99 | 120 | 139 | 159 | 189 | 209 | 246 | 266 | 286 | 309 |
| CEGPCR12c | 38 | 62 | 74 | 100 | 112 | 133 | 157 | 177 | 207 | 227 | 255 | 275 | 291 | 314 |
| CEGPCR12h | 38 | 62 | 74 | 100 | 112 | 133 | 157 | 177 | 207 | 227 | 255 | 275 | 291 | 314 |
| CEGPCR12u | 38 | 62 | 74 | 100 | 112 | 133 | 157 | 177 | 207 | 227 | 255 | 275 | 291 | 314 |
| CEGPCR12v | 38 | 62 | 74 | 100 | 112 | 133 | 157 | 177 | 207 | 227 | 255 | 275 | 291 | 314 |
| CEGPCR13 | 27 | 51 | 63 | 89 | 101 | 122 | 141 | 161 | 190 | 210 | 276 | 296 | 320 | 343 |
| CEGPCR14 | 30 | 54 | 65 | 91 | 103 | 124 | 143 | 163 | 198 | 218 | 254 | 274 | 299 | 322 |
| CEGPCR15 | 62 | 86 | 98 | 124 | 136 | 157 | 176 | 196 | 235 | 255 | 285 | 305 | 321 | 344 |
| CEGPCR16 | 29 | 53 | 65 | 91 | 103 | 124 | 143 | 163 | 193 | 213 | 250 | 270 | 287 | 310 |
| CEGPCR17 | 36 | 60 | 72 | 98 | 110 | 131 | 150 | 170 | 220 | 246 | 274 | 294 | 306 | 329 |
| CEGPCR18a | 10 | 34 | 46 | 72 | 84 | 105 | 126 | 146 | 185 | 205 | 282 | 302 | 320 | 343 |
| CEGPCR19.1 | 56 | 80 | 92 | 118 | 130 | 151 | 170 | 190 | 228 | 248 | 287 | 307 | 324 | 347 |
| CEGPCR19.2 | 56 | 80 | 92 | 118 | 130 | 151 | 170 | 190 | 228 | 248 | 287 | 307 | 324 | 347 |
| CEGPCR20 | 44 | 68 | 79 | 105 | 117 | 138 | 159 | 179 | 215 | 235 | 283 | 303 | 323 | 346 |
| CEGPCR21 | 76 | 100 | 112 | 138 | 151 | 172 | 193 | 213 | 249 | 269 | 336 | 356 | 373 | 396 |
| CEGPCR22 | 55 | 79 | 95 | 121 | 132 | 153 | 173 | 193 | 233 | 253 | 279 | 299 | 319 | 342 |
| CEGPCR23 | 21 | 45 | 64 | 90 | 101 | 122 | 142 | 162 | 203 | 223 | 249 | 269 | 287 | 310 |
| CEGPCR24a | 51 | 75 | 87 | 113 | 128 | 149 | 170 | 189 | 223 | 243 | 279 | 299 | 316 | 339 |
| CEGPCR24b | 51 | 75 | 87 | 113 | 128 | 149 | 170 | 189 | 223 | 243 | 279 | 299 | 316 | 339 |
| Receptor | Peptide ligand | SEQ ID NO | Species of origin | EC 50 range |
| CEGPCR3 | GGPQGPLRF-NH 2 | 85 | C. elegans /flp15 | 150-400 nM |
| (SEQ ID | GPSGPLRF-NH 2 | 89 | C. elegans /flp15 | |
| NO:44) | GNSFLRF-NH 2 | 88 | Manduca | ˜8 μM |
| CEGPCR7 | GLGPRPLRF-NH 2 | 86 | A. suum (AF9)/ | ˜200-250 nM |
| (SEQ ID | C. elegans (flp21) | |||
| NO:26) | [I]Y 0 -GLGPRPLRF-NH 2 | 118 | Synthetic AF9 | |
| analog |
| Receptor | Peptide | NO | Species of origin | EC 50 range |
| CEGPCR5 | GLGPRPLRF-NH 2 (AF9) | 86 | C. elegans (flp21), | ˜1-10 nM |
| (SEQ ID | ||||
| A. suum | ||||
| NO:46) | ||||
| CEGPCR5 | YGLGPRPLRF-NH 2 | 125 | [Y 0 ]AF9 analog | ˜80-800 nM |
| (SEQ ID | [iodo Y]GLGPRPLRF-NH 2 | 118 | [iodo Y 0 AF9 analog | |
| NO:46) | GNSFLRF-NH 2 | 88 | Manducca | |
| YLRF-NH 2 | 126 | Leech | ||
| SDRNFLRF-NH 2 | 127 | Lobster | ||
| FLRF-NH 2 | 128 | Trematoda | ||
| FMRF-NH 2 | 129 | Mollusca | ||
| TNRNFLRF-NH 2 | 102 | Lobster | ||
| DPSFLRF-NH 2 | 131 | Manducca | ||
| SVPGVLRF-NH 2 | 100 | C. elegans (flp18) | ||
| CEGPCR5 | AGPRFIRF-NH 2 | 133 | ||
| A. suum | ||||
| ˜1-8 μM | ||||
| (SEQ ID | KSVPGVLRF-NH 2 | 92 | C. elegans (flp18) | |
| NO:46) | KPNFLRY-NH 2 | 135 | C. elegans (flp18) | |
| ARGPQLRLRF-NH 2 | 136 | |||
| L. decemlineata | ||||
| YIRF-NH 2 | 137 | Thematoda | ||
| GMPGVLRF-NH 2 | 87 | |||
| A. suum | ||||
| AGAKFIRF-NH 2 | 139 | C. elegans (flp5) | ||
| KPNFLRF-NH 2 | 91 | C. elegans (flp1) | ||
| KHEYLRF-NH 2 | 141 | C. elegans (flp14) | ||
| EIVFHQISPIFFRF-NH 2 | 83 | C. elegans (flp14) |
| SEQ | Species of | |||
| Receptor | Sequence | ID NO | origin | EC 50 range |
| CEGPCR16 | SPLGTMRF-NH 2 | 99 | C. elegans flp3 | ˜60-300 nM |
| (SEQ ID | SAFPFGTMRF-NH 2 | 97 | C. elegans flp3 | |
| NO:36) | SADDSAPFGTMRF-NH 2 | 96 | C. elegans flp3 | |
| ASEDALFGTMRF-NH 2 | 77 | C. elegans flp3 | ||
| EDGNAPFGTMRF-NH 2 | 82 | C. elegans flp3 | ||
| EAEEPLGTMRF-NH 2 | 81 | C. elegans flp3 | ||
| GMPGVLRF-NH 2 (AF20) | 87 | |||
| A. suum | ||||
| SVPGVLRF-NH 2 | 100 | C. elegans flp18 | ||
| CEGPCR16 | KSVPGVLRF-NH 2 | 92 | C. elegans flp18 | ˜0.7-4.6 μM |
| (SEQ ID | DVPGVLRF-NH 2 | 80 | C. elegans flp18 | |
| NO:36) | SEVPGVLRF-NH 2 | 98 | C. elegans flp18 | |
| DFDGAMPGVLRF-NH 2 | 120 | C. elegans flp18 | ||
| EIPGVLRF-NH 2 | 121 | C. elegans flp18 | ||
| AVPGVLRF-NH 2 (AF3) | 79 | |||
| A. suum | ||||
| GDVPGVLRF-NH 2 (AF4) | 84 | |||
| A. suum |
| 77 | 12 | PRT | Caenorhabditis elegans/flp 3 |
| MOD_RES | (12) | AMIDATION |
| 78 | 8 | PRT | Caenorhabditis elegans/flp 4 |
| MOD_RES | (8) | AMIDATION |
| 80 | 8 | PRT | Caenorhabditis elegans/flp 18 |
| MOD_RES | (8) | AMIDATION |
| 81 | 11 | PRT | Caenorhabditis elegans/flp 3 |
| MOD_RES | (11) | AMIDATION |
| 82 | 12 | PRT | Caenorhabditis elegans/flp 3 |
| MOD_RES | (12) | AMIDATION |
| 83 | 14 | PRT | Caenorhabditis elegans/flp 14 |
| MOD_RES | (14) | AMIDATION |
| 85 | 9 | PRT | Caenorhabditis elegans/flp15 |
| MOD_RES | (9) | AMIDATION |
| 88 | 7 | PRT | Manduca |
| MOD_RES | (7) | AMIDATION |
| 89 | 8 | PRT | Caenorhabditis elegans/flp 15 |
| MOD_RES | (8) | AMIDATION |
| 90 | 5 | PRT | Artificial Sequence |
| SITE | (3) | Xaa=Norleucine |
| 91 | 7 | PRT | Caenorhabditis elegans/flp 1 |
| MOD_RES | (7) | AMIDATION |
| 92 | 9 | PRT | Caenorhabditis elegans/flp 18 |
| MOD_RES | (9) | AMIDATION |
| 93 | 10 | PRT | Caenorhabditis elegans/flp11 |
| MOD_RES | (10) | AMIDATION |
| 96 | 13 | PRT | Caenorhabditis elegans/flp3 |
| MOD_RES | (13) | AMIDATION |
| 97 | 10 | PRT | Caenorhabditis elegans/flp 3 |
| MOD_RES | (10) | AMIDATION |
| 98 | 9 | PRT | Caenorhabditis elegans/flp 18 |
| MOD_RES | (9) | AMIDATION |
| 99 | 8 | PRT | Caenorhabditis elegans/flp 3 |
| MOD_RES | (8) | AMIDATION |
| 100 | 8 | PRT | Caenorhabditis elegans/flp 18 |
| MOD_RES | (8) | AMIDATION |
| 101 | 10 | PRT | Drosophila |
| MOD_RES | (10) | AMIDATION |
| 103 | 11 | PRT | locust |
| SITE | (1) | Xaa=IodoY |
| 118 | 10 | PRT | A. suum/C. elegans (flp21 modified) |
| SITE | (1) | Xaa=IodoY |
| 120 | 12 | PRT | C. elegans/flp 18 |
| MOD_RES | (12) | AMIDATION |
| 121 | 8 | PRT | C. elegans/flp 18 |
| MOD_RES | (8) | AMIDATION |
| 122 | 7 | PRT | Drosophila |
| MOD_RES | (7) | AMIDATION |
| 123 | 7 | PRT | Drosophila |
| MOD_RES | (7) | AMIDATION |
| 125 | 10 | PRT | [Y] AF9 analog |
| MOD_RES | (10) | AMIDATION |
| 126 | 4 | PRT | Leech |
| MOD_RES | (4) | AMIDATION |
| 127 | 8 | PRT | Lobster |
| MOD_RES | (8) | AMIDATION |
| 128 | 4 | PRT | Trematoda |
| MOD_RES | (4) | AMIDATION |
| 129 | 4 | PRT | Mollusca |
| MOD_RES | (4) | AMIDATION |
| 131 | 7 | PRT | Manduca |
| MOD_RES | (7) | AMIDATION |
| 133 | 8 | PRT | A. suum |
| MOD_RES | (8) | AMIDATION |
| 135 | 7 | PRT | C. elegans (flp 1) |
| MOD_RES | (7) | AMIDATION |
| 136 | 10 | PRT | L. decemlineata |
| MOD_RES | (10) | AMIDATION |
| 137 | 4 | PRT | Trematoda |
| MOD_RES | (4) | AMIDATION |
| 139 | 8 | PRT | C. elegans (flp 5) |
| MOD_RES | (8) | AMIDATION |
| 141 | 7 | PRT | C. elegans (flp 14) |
| MOD_RES | (7) | AMIDATION |
| 143 | 8 | PRT | C. elegans flp3 |
| MOD_RES | (8) | AMIDATION |
| 150 | 11 | PRT | A. suum |
| MOD_RES | (11) | AMIDATION |
| 151 | 8 | PRT | A. suum |
| MOD_RES | (8) | AMIDATION |
| 152 | 4 | PRT | Artificial Sequence |
| MOD_RES | (4) | AMIDATION |
| 153 | 5 | PRT | Artificial Sequence |
| MOD_RES | (5) | AMIDATION |
| 154 | 5 | PRT | Artificial Sequence |
| MOD_RES | (5) | AMIDATION |
| 155 | 6 | PRT | Artificial Sequence |
| MOD_RES | (6) | AMIDATION |
| 156 | 7 | PRT | Artificial Sequence |
| MOD_RES | (7) | AMIDATION |
| 157 | 7 | PRT | Artificial Sequence |
| MOD_RES | (1) | ACETYLATION |
| 158 | 9 | PRT | Artificial Sequence |
| MOD_RES | (9) | AMIDATION |
| 159 | 8 | PRT | Artificial Sequence |
| MOD_RES | (8) | AMIDATION |
| 160 | 8 | PRT | Artificial Sequence |
| MOD_RES | (8) | AMIDATION |
| 161 | 7 | PRT | Artificial Sequence |
| MOD_RES | (7) | AMIDATION |
| 162 | 7 | PRT | Artificial Sequence |
| MOD_RES | (7) | AMIDATION |
| 163 | 6 | PRT | Artificial Sequence |
| MOD_RES | (6) | AMIDATION |
| 164 | 6 | PRT | Artificial Sequence |
| SITE | (1) | Xaa=3-Iodo-Y |
| 165 | 7 | PRT | P. redivivus |
| MOD_RES | (7) | AMIDATION |
| 166 | 8 | PRT | C. elegans flp1 |
| MOD_RES | (8) | AMIDATION |
| 167 | 8 | PRT | H. contortus |
| MOD_RES | (8) | AMIDATION |
Claims as granted
21 claimsLog in to read the claims of this application.
Log in to unlockClassifications
9 codes- C12N15/12
- C07K14/705
Claim changes
SoonSee which claims were amended, added or cancelled during examination, with every added and removed word marked.
The published claims of this application are not paired with the granted ones in what we hold.
File wrapper
See the full prosecution history — every USPTO and applicant action on this file, in order.
Log in to unlockDocuments
Log in to open the documents of this file: the application as filed, every office action and response, the notice of allowance.
Log in to unlockChain of title
See the full assignment history — every owner this patent has passed through, with recordation dates and reel/frame numbers.
Log in to unlock