USPatent applicationPatented

Compositions and methods for the modification of gene expression

Granted 22 Jul 2003 · 3 office actions

Application· this page
9598401
filed 20 Jun 2000
Publication
Not published
not published
Patent
US 6,596,925
granted 22 Jul 2003

Life of the application

22 dated events
⤢ drag to zoom20002002200420062008201020122014201620182020ProsecutionOwnershipTerm & fees
ProsecutionOwnershipTerm & feeshover for detail · click to open

Abstract

Novel isolated plant polynucleotide promoter sequences are provided, together with genetic constructs comprising such polynucleotides. Methods for using such constructs in modulating the transcription of DNA sequences of interest are also disclosed, together with transgenic plants comprising such constructs.

Description

142 parts
›CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to International Patent Application No. PCT/NZ00/00018, filed Feb. 24, 2000, and to U.S. patent application Ser. No. 60/146,591, filed Jul. 30, 1999, and is a continuation-in-part of U.S. patent application Ser. No. 09/276,599, filed Mar. 25, 1999 now U.S. Pat. No. 6,380,459.

›TECHNICAL FIELD OF THE INVENTION

This invention relates to the regulation of polynucleotide transcription and/or expression. More specifically, this invention relates to polynucleotide regulatory sequences isolated from plants that are capable of initiating and driving the transcription of polynucleotides, and the use of such regulatory sequences in the modification of transcription of endogenous and/or heterologous polynucleotides and production of polypeptides. Polypeptide sequences are also disclosed.

›BACKGROUND OF THE INVENTION

Gene expression is regulated, in part, by the cellular processes involved in transcription. During transcription, a single-stranded RNA complementary to the DNA sequence to be transcribed is formed by the action of RNA polymerases. Initiation of transcription in eucaryotic cells is regulated by complex interactions between cis-acting DNA motifs, located within the gene to be transcribed, and trans-acting protein factors. Among the cis-acting regulatory regions are sequences of DNA, termed promoters, to which RNA polymerase is first bound, either directly or indirectly. As used herein, the term “promoter” refers to the 5′ untranslated region of a gene that is associated with transcription and which generally includes a transcription start site. Other cis-acting DNA motifs, such as enhancers, may be situated further up- and/or down-stream from the initiation site.

Both promoters and enhancers are generally composed of several discrete, often redundant elements, each of which may be recognized by one or more trans-acting regulatory proteins, known as transcription factors. Promoters generally comprise both proximal and more distant elements. For example, the so-called TATA box, which is important for the binding of regulatory proteins, is generally found about 25 basepairs upstream from the initiation site. The so-called CAAT box is generally found about 75 basepairs upstream of the initiation site. Promoters generally contain between about 100 and 1000 nucleotides, although longer promoter sequences are possible.

For the development of transgenic plants, constitutive promoters that drive strong transgene expression are preferred. Currently, the only available constitutive plant promoter that is widely used is derived from Cauliflower Mosaic Virus. Furthermore, there exists a need for plant-derived promoters for use in transgenic food plants due to public conceptions regarding the use of viral promoters. Few gymnosperm promoters have been cloned and those derived from angiosperms have been found to function poorly in gymnosperms. There thus remains a need in the art for polynucleotide promoter regions isolated from plants for use in modulating transcription and expression of polynucleotides in transgenic plants.

›SUMMARY OF THE INVENTION · 1 of 2

Briefly, isolated polynucleotide regulatory sequences from eucalyptus and pine that are involved in the regulation of gene expression are disclosed, together with methods for the use of such polynucleotide regulatory regions in the modification of expression of endogenous and/or heterologous polynucleotides in transgenic plants. In particular, the present invention provides polynucleotide promoter sequences from 5′ untranslated, or non-coding, regions of plant genes that initiate and regulate transcription of polynucleotides placed under their control, together with isolated polynucleotides comprising such promoter sequences.

In a first aspect, the present invention provides isolated polynucleotide sequences comprising a polynucleotide selected from the group consisting of: (a) sequences recited in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; (b) complements of the sequences recited in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; (c) reverse complements of the sequences recited in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; (d) reverse sequences of the sequences recited. in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; (e) sequences having either 40%, 60%, 75or 90% identical nucleotides, as defined herein, to a sequence of (a)-(d); probes and primers corresponding to the sequences set out in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; polynucleotides comprising at least a specified number of contiguous residues of any of the polynucleotides identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; and extended sequences comprising portions of the sequences set out in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; all of which are referred to herein as “polynucleotides of the present invention.” The present invention also provides isolated polypeptide sequences identified in the attached Sequence Listing as SEQ ID NO: 63-80 and 87; polypeptide variants of those sequences; and polypeptides comprising the isolated polypeptide sequences and variants of those sequences.

In another aspect, the present invention provides genetic constructs comprising a polynucleotide of the present invention, either alone, or in combination with one or more additional polynucleotides of the present invention, or in combination with one or more known polynucleotides, together with cells and target organisms comprising such constructs.

In a related aspect, the present invention provides genetic constructs comprising, in the 5′-3′ direction, a polynucleotide promoter sequence of the present invention, a polynucleotide to be transcribed, and a gene termination sequence. The polynucleotide to be transcribed may comprise an open reading frame of a polynucleotide that encodes a polypeptide of interest, or it may be a non-coding, or untranslated, region of a polynucleotide of interest. The open reading frame may be orientated in either a sense or antisense direction. Preferably, the gene termination sequence is functional in a host plant. Most preferably, the gene termination sequence is that of the gene of interest, but others generally used in the art, such as the Agrobacterium tumefaciens nopalin synthase terminator may be usefully employed in the present invention. The genetic construct may further include a marker for the identification of transformed cells.

In a further aspect, transgenic plant cells comprising the genetic constructs of the present invention are provided, together with organisms, such as plants, comprising such transgenic cells, and fruits, seeds and other products, derivatives, or progeny of such plants. Propagules of the inventive transgenic plants are included in the present invention. As used herein, the word “propagule” means any part of a plant that may be used in reproduction or propagation, sexual or asexual, including cuttings.

Plant varieties, particularly registerable plant varieties according to Plant Breeders' Rights, may be excluded from the present invention. A plant need not be considered a “plant variety” simply because it contains stably within its genome a transgene, introduced into a cell of the plant or an ancestor thereof.

In yet another aspect, methods for modifying gene expression in a target organism, such as a plant, are provided, such methods including stably incorporating into the genome of the organism a genetic construct of the present invention. In a preferred embodiment, the target organism is a plant, more preferably a woody plant, most preferably selected from the group consisting of eucalyptus and pine species, most preferably from the group consisting of Eucalyptus grandis and Pinus radiata.

In another aspect, methods for producing a target organism, such as a plant, having modified polypeptide expression are provided, such methods comprising transforming a plant cell with a genetic construct of the present invention to provide a transgenic cell, and cultivating the transgenic cell under conditions conducive to regeneration and mature plant growth.

In other aspects, methods for identifying a gene responsible for a desired function or phenotype are provided, the methods comprising transforming a plant cell with a genetic construct comprising a polynucleotide promoter sequence of the present invention operably linked to a polynucleotide to be tested, cultivating the plant cell under conditions conducive to regeneration and mature plant growth to provide a transgenic plant; and comparing the phenotype of the transgenic plant with the phenotype of non-transformed, or wild-type, plants.

In yet a further aspect, the present invention provides isolated polynucleotides that encode ubiquitin. In specific embodiments, the isolated polynucleotides comprise a polynucleotide selected from the group consisting of: (a) sequences recited in SEQ ID NO: 1 and 34; (b) complements of the sequences recited in SEQ ID NO: 1 and 34; (c) reverse complements of the sequences recited in SEQ ID NO: 1 and 34; (d) reverse sequences of the sequence recited in SEQ ID NO: 1 and 34; and (e) sequences having either 40%, 60%, 75% or 90% identical nucleotides, as defined herein, to a sequence of (a)-(d). Polypeptides encoded by such polynucleotides are also provided, together with genetic constructs comprising such polynucleotides, and host cells and transgenic organisms, for example plants, transformed with such genetic constructs. In specific embodiments, such polypeptides comprise a sequence provided in SEQ ID NO: 80 or 67.

›SUMMARY OF THE INVENTION · 2 of 2

In yet further aspects, the present invention provides isolated polynucleotides comprising the DNA sequence of SEQ ID NO: 21, or a complement, reverse complement or variant of SEQ ID NO: 21, together with genetic constructs comprising such polynucleotides and cells transformed with such sequences. As discussed below, removal of the sequence of SEQ ID NO: 21 from a polynucleotide that comprises the sequence of SEQ ID NO: 21 may enhance expression of the polynucleotide. Conversely, the inclusion of the sequence of SEQ ID NO: 21 in a genetic construct comprising a polynucleotide of interest may decrease expression of the polynucleotide.

The above-mentioned and additional features of the present invention and the manner of obtaining them will become apparent, and the invention will be best understood by reference to the following more detailed description. All references disclosed herein are hereby incorporated by reference in their entirety as if each was incorporated individually.

›BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 shows the expression in A. thaliana of the GUS gene in promoter reporter constructs containing either the superubiquitin promoter with introns, the superubiquitin promoter without introns, or the CaMV 35S promoter. The GUS expression was measured by fluorimetric determination of 4-methyl-umbelliferone (MU) in protein extracts from these plants.

FIG. 2 shows the expression of the GUS gene in tobacco plant protoplasts by deletion constructs containing the superubiquitin promoter with or without the intron. The constructs contained 1,103; 753; 573; 446; 368; and 195 bp upstream of the TATA sequence (bp numbers 1,104-1,110 of SEQ ID NO: 2). The GUS expression was measured by fluorimetric determination of 4-methyl-umbelliferone (MU) in protein extracts from these protoplasts.

FIG. 3 shows the expression of the GUS gene in tobacco plant protoplasts by constructs containing P. radiata either the constitutive promoters Elongation factor-1 alpha, 5-adenosylmethionine synthetase or the superubiquitin promoter without the intron. The GUS expression was measured by fluorimetric determination of 4-methyl-umbelliferone (MU) in protein extracts from these protoplasts.

›DETAILED DESCRIPTION OF THE INVENTION · 1 of 7

The present invention provides isolated polynucleotide regulatory regions that may be employed in the manipulation of plant phenotypes, together with isolated polynucleotides comprising such regulatory regions. More specifically, polynucleotide promoter sequences isolated from pine and eucalyptus are disclosed. As discussed above, promoters are components of the cellular “transcription apparatus” and are involved in the regulation of gene expression. Both tissue- and temporal-specific gene expression patterns have been shown to be initiated and controlled by promoters during the natural development of a plant. The isolated polynucleotide promoter sequences of the present invention may thus be employed in the modification of growth and development of plants, and of cellular responses to external stimuli, such as environmental factors and disease pathogens.

Using the methods and materials of the present invention, the amount of a specific polypeptide of interest may be increased or reduced by incorporating additional copies of genes, or coding sequences, encoding the polypeptide, operably linked to an inventive promoter sequence, into the genome of a target organism, such as a plant. Similarly, an increase or decrease in the amount of the polypeptide may be obtained by transforming the target plant with antisense, copies of such genes.

The polynucleotides of the present invention were isolated from forestry plant sources, namely from Eucalyptus grandis and Pinus radiata , but they may alternatively be synthesized using conventional synthesis techniques. Specifically, isolated polynucleotides of the present invention include polynucleotides comprising a sequence selected from the group consisting of sequences identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; complements of the sequences identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; reverse complements of the sequences identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120; at least a specified number of contiguous residues (x-mers) of any of the above-mentioned polynucleotides; extended sequences corresponding to any of the above polynucleotides; antisense sequences corresponding to any of the above polynucleotides; and variants of any of the above polynucleotides, as that term is described in this specification.

In another embodiment, the present invention provides isolated polypeptides encoded by the polynucleotides of SEQ ID NO: 63-80 and 87.

The polynucleotides and polypeptides of the present invention were putatively identified by DNA and polypeptide similarity searches. In the attached Sequence Listing, SEQ ID NOS. 1-14, 20, 22-62, 81-86 and 88-120 are polynucleotide sequences, and SEQ ID NOS. 63-80 and 87 are polypeptide sequences. The polynucleotides and polypeptides of the present invention have demonstrated similarity to promoters that are known to be involved in regulation of transcription and/or expression in plants. The putative identity of each of the inventive polynucleotides is shown below in Table 1, together with the 5′ untranslated region (5′ UTR) or putative promoter region (identified by residue number).

In one embodiment, the present invention provides polynucleotide sequences isolated from Pinus radiata and Eucalyptus grandis that encode a ubiquitin polypeptide. The full-length sequence of the ubiquitin polynucleotide isolated from Pinus radiata is provided in SEQ ID NO: 1, with the sequence of the promoter region including an intron being provided in SEQ ID NO: 2 and the sequence of the promoter region excluding the intron being provided in SEQ ID NO: 3. The sequence of the ubiquitin polynucleotide isolated from Eucalyptus grandis is provided in SEQ ID NO: 34. In a related embodiment the present invention provides isolated polypeptides encoded by the isolated polynucleotides of SEQ ID NO: 1 and 34, including polypeptides comprising the sequences of SEQ ID NO: 80 and 67.

The term “polynucleotide(s),” as used herein, means a single or double-stranded polymer of deoxyribonucleotide or ribonucleotide bases and includes DNA and corresponding RNA molecules, including HnRNA and mRNA molecules, both sense and anti-sense strands, and comprehends cDNA, genomic DNA and recombinant DNA, as well as wholly or partially synthesized polynucleotides. An HnRNA molecule contains introris and corresponds to a DNA molecule in a generally one-to-one manner. An mRNA molecule corresponds to an HnRNA and DNA molecule from which the introns have been excised. A polynucleotide may consist of an entire gene, or any portion thereof. Operable anti-sense polynucleotides may comprise a fragment of the corresponding polynucleotide, and the definition of “polynucleotide” therefore includes all such operable anti-sense fragments. Anti-sense polynucleotides and techniques involving anti-sense polynucleotides are well known in the art and are described, for example, in Robinson-Benion et al. “Antisense techniques,” Methods in Enzymol . 254(23):363-375, 1995; and Kawasaki et al., in Artific. Organs 20(8):836-848, 1996.

All of the polynucleotides and polypeptides described herein are isolated and purified, as those terms are commonly used in the art. Preferably, the polypeptides and polynucleotides are at least about 80% pure, more preferably at least about 90% pure, and most preferably at least about 99% pure.

The definition of the terms “complement”, “reverse complement” and “reverse sequence”, as used herein, is best illustrated by the following example. For the sequence 5′ AGGACC 3′, the complement, reverse complement and reverse sequence are as follows:

Some of the polynucleotides of the present invention are “partial” sequences, in that they do not represent a full-length gene encoding a full-length polypeptide. Such partial sequences may be extended by analyzing and sequencing various DNA libraries using primers and/or probes and well known hybridization and/or PCR techniques. Partial sequences may be extended until an open reading frame encoding a polypeptide, a full-length polynucleotide and/or gene capable of expressing a polypeptide, or another useful portion of the genome is identified. Such extended sequences, including full-length polynucleotides and genes, are described as “corresponding to” a sequence identified as one of the sequences of SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or a variant thereof, or a portion of one of the sequences of SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or a variant thereof, when the extended polynucleotide comprises an identified sequence or its variant, or an identified contiguous portion (x-mer) of one of the sequences of SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or a variant thereof. Such extended polynucleotides may have a length of from about 50 to about 4,000 nucleic acids or base pairs, and preferably have a length of less than about 4,000 nucleic acids or base pairs, more preferably yet a length of less than about 3,000 nucleic acids or base pairs, more preferably yet a length of less than about 2,000 nucleic acids or base pairs. Under some circumstances, extended polynucleotides of the present invention may have a length of less than about 1,800 nucleic acids or base pairs, preferably less than about 1,600 nucleic acids or base pairs, more preferably less than about 1,400 nucleic acids or base pairs, more preferably yet less than about 1,200 nucleic acids or base pairs, and most preferably less than about 1,000 nucleic acids or base pairs.

›DETAILED DESCRIPTION OF THE INVENTION · 2 of 7

Similarly, RNA sequences, reverse sequences, complementary sequences, antisense sequences, and the like, corresponding to the polynucleotides of the present invention, may be routinely ascertained and obtained using the cDNA sequences identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120.

The polynucleotides identified as SEQ ID NO. 1-14, 20, 22-62, 81-86 and 88-120, may contain open reading frames (“ORFs”) or partial open reading frames encoding polypeptides. Additionally, open reading frames encoding polypeptides may be identified in extended or full length sequences corresponding to the sequences set out as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120. Open reading frames may be identified using techniques that are well known in the art. These techniques include, for example, analysis for the location of known start and stop codons, most likely reading frame identification based on codon frequencies, etc. Suitable tools and software for ORF analysis include, for example, “GeneWise”, available from The Sanger Center, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, United Kingdom; “Diogenes”, available from Computational Biology Centers, University of Minnesota, Academic Health Center, UMHG Box 43, Minneapolis Minn. 55455 and “GRAL”, available from the Informatics Group, Oak Ridge National Laboratories, Oak Ridge, Tenn. Open reading frames and portions of open reading frames may be identified in the polynucleotides of the present invention. Once a partial open reading frame is identified, the polynucleotide may be extended in the area of the partial open reading frame using techniques that are well known in the art until the polynucleotide for the full open reading frame is identified. Thus, open reading frames encoding polypeptides may be identified using the polynucleotides of the present invention.

Once open reading frames are identified in the polynucleotides of the present invention, the open reading frames may be isolated and/or synthesized. Expressible genetic constructs comprising the open reading frames and suitable promoters, initiators, terminators, etc., which are well known in the art, may then be constructed. Such genetic constructs may be introduced into a host cell to express the polypeptide encoded by the open reading frame. Suitable host cells may include various prokaryotic and eukaryotic cells, including plant cells, mammalian cells, bacterial cells, algae and the like.

Polypeptides encoded by the polynucleotides of the present invention may be expressed and used in various assays to determine their biological activity. Such polypeptides may be used to raise antibodies, to isolate corresponding interacting proteins or other compounds, and to quantitatively determine levels of interacting proteins or other compounds.

The term “polypeptide”, as used herein, encompasses amino acid chains of any length including full length proteins, wherein amino acid residues are linked by covalent peptide bonds. Polypeptides of the present invention may be isolated and purified natural products, or may be produced partially or wholly using recombinant techniques. The term “polypeptide encoded by a polynucleotide” as used herein, includes polypeptides encoded by a nucleotide sequence which includes the partial isolated DNA sequences of the present invention.

In a related aspect, polypeptides are provided that comprise at least a functional portion of a polypeptide having a sequence selected from the group consisting of sequences provided in SEQ ID NO: 63-80 and 87, and variants thereof. As used herein, the “functional portion” of a polypeptide is that portion which contains the active site essential for affecting the function of the polypeptide, for example, the portion of the molecule that is capable of binding one or more reactants. The active site may be made up of separate portions present on one or more polypeptide chains and will generally exhibit high binding affinity. Functional portions of a polypeptide may be identified by first preparing fragments of the polypeptide by either chemical or enzymatic digestion of the polypeptide, or by mutation analysis of the polynucleotide that encodes the polypeptide and subsequent expression of the resulting mutant polypeptides. The polypeptide fragments or mutant polypeptides are then tested to determine which portions retain biological activity, using, for example, the representative assays provided below. A functional portion comprising an active site may be made up of separate portions present on one or more polypeptide chains and generally exhibits high substrate specificity.

Portions and other variants of the inventive polypeptides may be generated by synthetic or recombinant means. Synthetic polypeptides having fewer than about 100 amino acids, and generally fewer than about 50 amino acids, may be generated using techniques well known to those of ordinary skill in the art. For example, such polypeptides may be synthesized using any of the commercially available solid-phase techniques, such as the Merrifield solid-phase synthesis method, wherein amino acids are sequentially added to a growing amino acid chain. (Merrifield, J. Am. Chem. Soc . 85: 2149-2154, 1963). Equipment for automated synthesis of polypeptides is commercially available from suppliers such as Perkin Elmer/Applied Biosystems, Inc. (Foster City, Calif.), and may be operated according to the manufacturer's instructions. Variants of a native polypeptide may be prepared using standard mutagenesis techniques, such as oligonucleotide-directed site-specific mutagensis (Kunkel, Proc. Natl. Acad. Sci . USA 82: 488-492, 1985). Sections of DNA sequences may also be removed using standard techniques to permit preparation of truncated polypeptides.

As used herein, the term “variant” comprehends nucleotide or amino acid sequences different from the specifically identified sequences, wherein one or more nucleotides or amino acid residues is deleted, substituted, or added. Variants may be naturally occurring allelic variants, or non-naturally occurring variants. Variant sequences (polynucleotide or polypeptide) preferably exhibit at least 50%, more preferably at least 75%, and most preferably at least 90% identity to a sequence of the present invention. The percentage identity is determined by aligning the two sequences to be compared as described below, determining the number of identical residues in the aligned portion, dividing that number by the total number of residues in the inventive (queried) sequence, and multiplying the result by 100.

›DETAILED DESCRIPTION OF THE INVENTION · 3 of 7

Polynucleotide and polypeptide sequences may be aligned, and percentage of identical residues in a specified region may be determined against another polynucleotide and polypeptide sequences, using computer algorithms that are publicly available. Two exemplary algorithms for aligning and identifying the similarity of polynucleotide sequences are the BLASTN and FASTA algorithms. Polynucleotides may also be analyzed using the BLASTX algorithm, which compares the six-frame conceptual translation products of a nucleotide query sequence (both strands) against a protein sequence database. The similarity of polypeptide sequences may be examined using the BLASTP algorithm. The BLASTN algorithm version 2.0.4 [Feb. 24, 1998] and version 2.0.6 [Sep. 16, 1998], set to the default parameters described in the documentation and distributed with the algorithm, are preferred for use in the determination of polynucleotide variants according to the present invention. The BLASTP algorithm, is preferred for use in the determination of polypeptide variants according to the present invention. The use of the BLAST family of algorithms, including BLASTN, BLASTP, and BLASTX, is described in the publication of Altschul, et al., “Gapped BLAST and PSI-BLAST: a new generation of protein database search programs,” Nucleic Acids Res . 25: 3389-3402, 1997.

The FASTA software package is available from the University of Virginia (University of Virginia, PO Box 9025, Charlottesville, Va. 22906-9025). Version 2.0u4, February 1996, set to the default parameters described in the documentation and distributed with the algorithm, may be used in the determination of variants according to the present invention. The use of the FASTA algorithm is described in Pearson and Lipman, “Improved Tools for Biological Sequence Analysis,” Proc. Natl. Acad. Sci . USA 85: 2444-2448, 1988; and Pearson, “Rapid and Sensitive Sequence Comparison with FASTP and FASTA,” Methods in Enzymol . 183: 63-98, 1990.

The following running parameters are preferred for determination of alignments and similarities using BLASTN that contribute to the E values and percentage identity for polynucleotide sequences: Unix running command: blastall -p blastn -d embldb -e 10 -G0 -E0 -r 1 -v 30 -b 30 -i, queryseq -o results; the parameters are: -p Program Name [String]; -d Database [String]; -e Expectation value (E) [Real]; -G Cost to open a gap (zero invokes default behavior) [Integer]; -E Cost to extend a gap (zero invokes default behavior) [Integer]; -r Reward for a nucleotide match (BLASTN only) [Integer]; -v Number of one-line descriptions (V) [Integer]; -b Number of alignments to show (B) [Integer]; -i Query File [File In]; and -o BLAST report Output File [File Out] Optional.

The following running parameters are preferred for determination of alignments and similarities using BLASTP that contribute to the E values and percentage identity of polypeptide sequences: blastall -p blastp -d swissprotdb -e 10 -G 0 -E 0 -v 30 -b 30 -i queryseq -o results; the parameters are: -p Program Name [String]; -d Database [String]; -e Expectation value (E) [Real]; -G Cost to open a gap (zero invokes default behavior) [Integer]; -E Cost to extend a gap (zero invokes default behavior) [Integer]; -v Number of one-line descriptions (v) [Integer]; -b Number of alignments to show (b) [Integer]; -I Query File [File In]; -o BLAST report Output File [File Out] Optional.

The “hits” to one or more database sequences by a queried sequence produced by BLASTN, FASTA, BLASTP or a similar algorithm, align and identify similar portions of sequences. The hits are arranged in order of the degree of similarity and the length of sequence overlap. Hits to a database sequence generally represent an overlap over only a fraction of the sequence length of the queried sequence.

The BLASTN, FASTA and BLASTP algorithms also produce “Expect” values for alignments. The Expect value (E) indicates the number of hits one can “expect” to see over a certain number of contiguous sequences by chance when searching a database of a certain size. The Expect value is used as a significance threshold for determining whether the hit to a database, such as the preferred EMBL database, indicates true similarity. For example, an E value of 0.1 assigned to a polynucleotide hit is interpreted as meaning that in a database of the size of the EMBL database, one might expect to see 0.1 matches over the aligned portion of the sequence with a similar score simply by chance. By this criterion, the aligned and matched portions of the polynucleotide sequences then have a probability of 90% of being the same. For sequences having an E value of0.01 or less over aligned and matched portions, the probability of finding a match by chance in the EMBL database is 1% or less using the BLASTN or FASTA algorithm.

According to one embodiment, “variant” polynucleotides and polypeptides, with reference to each of the polynucleotides and polypeptides of the present invention, preferably comprise sequences having the same number or fewer nucleic or amino acids than each of the polynucleotides or polypeptides of the present invention and producing an E value of 0.01 or less when compared to the polynucleotide or polypeptide of the present invention. That is, a variant polynucleotide or polypeptide is any sequence that has at least a 99% probability of being the same as the polynucleotide or polypeptide of the present invention, measured as having an E value of 0.01 or less using the BLASTN, FASTA, or BLASTP algorithms set at parameters described above. According to a preferred embodiment, a variant polynucleotide is a sequence having the same number or fewer nucleic acids than a polynucleotide of the present invention that has at least a 99% probability of being the same as the polynucleotide of the present invention, measured as having an E value of 0.01 or less using the BLASTN or FASTA algorithms set at parameters described above. Similarly, according to a preferred embodiment, a variant polypeptide is a sequence having the same number or fewer amino acids than a polypeptide of the present invention that has at least a 99% probability of being the same as a polypeptide of the present invention, measured as having an E value of 0.01 or less using the BLASTP algorithm set at the parameters described above.

›DETAILED DESCRIPTION OF THE INVENTION · 4 of 7

Alternatively, variant polynucleotides of the present invention hybridize to the polynucleotide sequences recited in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or complements, reverse sequences, or reverse complements of those sequences under stringent conditions. As used herein, “stringent conditions” refers to prewashing in a solution of 6×SSC, 0.2% SDS; hybridizing at 65° C., 6×SSC, 0.2% SDS overnight; followed by two washes of 30 minutes each in 1×SSC, 0.1% SDS at 65° C. and two washes of 30 minutes: each in 0.2×SSC, 0.1% SDS at 65° C.

The present invention also encompasses polynucleotides that differ from the disclosed sequences but that, as a consequence of the discrepancy of the genetic code, encode a polypeptide having similar activity to a polypeptide encoded by a polynucleotide of the present invention. Thus, polynucleotides comprising sequences that differ from the polynucleotide sequences recited in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or complements, reverse sequences, or reverse complements thereof, as a result of conservative substitutions are contemplated by and encompassed within the present invention. Additionally, polynucleotides comprising sequences that differ from the polynucleotide sequences recited in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or complements, reverse complements or reverse sequences thereof, as a result of deletions and/or insertions totaling less than 10% of the total sequence to length are also contemplated by and encompassed within the present invention. Similarly, polypeptides comprising sequences that differ from the polypeptide sequences recited in SEQ ID NO: 63-80 and 87, as a result of amino acid substitutions, insertions, and/or deletions totaling less than 10% of the total sequence length are contemplated by an encompassed within the present invention. In certain embodiments, variants of the inventive polypeptides and polynucleotides possess biological activities that are the same or similar to those of the inventive polypeptides or polynucleotides. Such variant polynucleotides function as promoter sequences and are thus capable of modifying gene expression in a plant.

The polynucleotides of the present invention may be isolated from various libraries, or may be synthesized using techniques that are well known in the art. The polynucleotides may be synthesized, for example, using automated oligonucleotide synthesizers (e.g., Beckman Oligo 1000M DNA Synthesizer) to obtain polynucleotide segments of up to 50 or more nucleic acids. A plurality of such polynucleotide segments may then be ligated using standard DNA manipulation techniques that are well known in the art of molecular biology. One conventional and exemplary polynucleotide synthesis technique involves synthesis of a single stranded polynucleotide segment having, for example, 80 nucleic acids, and hybridizing that segment to a synthesized complementary 85 nucleic acid segment to produce a 5 nucleotide overhang. The next segment may then be synthesized in a similar fashion, with, a 5 nucleotide overhang on the opposite strand. The “sticky” ends ensure proper ligation when the two portions are hybridized. In this way, a complete polynucleotide of the present invention may be synthesized entirely in vitro.

Polynucleotides of the present invention also comprehend polynucleotides comprising at least a specified number of contiguous residues (x-mers) of any of the polynucleotides identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, complements, reverse sequences, and reverse complements of such sequences, and their variants. Similarly, polypeptides of the present invention comprehend polypeptides comprising at least a specified number of contiguous residues (x-mers) of any of the polypeptides identified as SEQ ID NO: 63-80 and 87, and their variants. As used herein, the term “x-mer,” with reference to a specific value of “x,” refers to a sequence comprising at least a specified number (“x”) of contiguous residues of any of the polynucleotides identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or the polypeptides identified as SEQ ID NO: 63-80 and 87. According to preferred embodiments, the value of x is preferably at least 20, more preferably at least 40, more preferably yet at least 60, and most preferably at least 80. Thus, polynucleotides and polypeptides of the present invention comprise a 20-mer, a 40-mer, a 60-mer, an 80-mer, a 100-mer, a 120-mer, a 150-mer, a 180-mer, a 220-mer, a 250-mer, a 300-mer, 400-mer, 500-mer or 600-mer of a polynucleotide or polypeptide identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, and variants thereof.

As noted above, the inventive polynucleotide promoter sequences may be employed in genetic constructs to drive transcription and/or expression of a polynucleotide of interest. The polynucleotide of interest may be either endogenous or heterologous to an organism, for example a plant, to be transformed. The inventive genetic constructs may thus be employed to modulate levels of transcription and/or expression of a polynucleotide, for example gene, that is present in the wild-type plant, or may be employed to provide transcription and/or expression of a DNA sequence that is not found in the wild-type plant.

In certain embodiments, the polynucleotide of interest comprises an open reading frame that encodes a target polypeptide. The open reading frame is inserted in the genetic construct in either a sense or antisense orientation, such that transformation of a target plant with the genetic construct will lead to a change in the amount of polypeptide compared to the wild-type plant. Transformation with a genetic construct comprising an open reading frame in a sense orientation will generally result in over-expression of the selected polypeptide, while transformation with a genetic construct comprising an open reading frame in an antisense orientation will generally result in reduced expression of the selected polypeptide. A population of plants transformed with a genetic construct comprising an open reading frame in either a sense or antisense orientation may be screened for increased or reduced expression of the polypeptide in question using techniques well known to those of skill in the art, and plants having the desired phenotypes may thus be isolated.

›DETAILED DESCRIPTION OF THE INVENTION · 5 of 7

Alternatively, expression of a target polypeptide may be inhibited by inserting a portion of the open reading frame, in either sense or antisense orientation, in the genetic construct. Such portions need not be full-length but preferably comprise at least 25 and more preferably at least 50 residues of the open reading frame. A much longer portion or even the full length DNA corresponding to the complete open reading frame may be employed. The portion of the open reading frame does not need to be precisely the same as the endogenous sequence, provided that there is sufficient sequence similarity to achieve inhibition of the target gene. Thus a sequence derived from one species may be used to inhibit expression of a gene in a different species.

In further embodiments, the inventive genetic constructs comprise a polynucleotide including an untranslated, or non-coding, region of a gene coding for a target polypeptide, or a polynucleotide complementary to such an untranslated region. Examples of untranslated regions which may be usefully employed in such constructs include introns and 5′-untranslated leader sequences. Transformation of a target plant with such a genetic construct may lead to a reduction in the amount of the polypeptide expressed in the plant by the process of cosuppression, in a manner similar to that discussed, for example, by Napoli et al., Plant Cell 2:279-290, 1990 and de Carvalho Niebel et al., Plant Cell 7:347-358, 1995.

Alternatively, regulation of polypeptide expression can be achieved by inserting appropriate sequences or subsequences (e.g. DNA or RNA) in ribozyme constructs (McIntyre and Manners, Transgenic Res . 5(4):257-262, 1996). Ribozymes are synthetic RNA molecules that comprise a hybridizing region complementary to two regions, each of which comprises at least 5 contiguous nucleotides in a mRNA molecule encoded by one of the inventive polynucleotides. Ribozymes possess highly specific endonuclease activity, which autocatalytically cleaves the mRNA.

The polynucleotide of interest, such as a coding sequence, is operably linked to a polynucleotide promoter sequence of the present invention such that a host cell is able to transcribe an RNA from the promoter sequence linked to the polynucleotide of interest. The polynucleotide promoter sequence is generally positioned at the 5′ end of the polynucleotide to be transcribed. Use of a constitutive promoter, such as the Pinus radiata ubiquitin polynucleotide promoter sequence of SEQ ID NO: 2 and 3 or the Eucalyptus grandis ubiquitin polynucleotide promoter sequence contained within SEQ ID NO: 34, will affect transcription of the polynucleotide of interest in all parts of the transformed plant. Use of a tissue specific promoter, such as the leaf-specific promoters of SEQ ID NO: 9-11, the root-specific promoters of SEQ ID NO: 13 and 14, the flower-specific promoters of SEQ ID NO: 29-33, 59 and 89-90, the pollen-specific promoters of SEQ ID NO: 49-55 and 94, the bud-specific promoter of SEQ ID NO: 40 or the meristem-specific promoter of SEQ ID NO: 45, will result in production of the desired sense or antisense RNA only in the tissue of interest. Temporally regulated promoters, such as the xylogenesis-specific promoter of SEQ ID NO: 5, 41-44 and 92, can be employed to effect modulation of the rate of DNA transcription at a specific time during development of a transformed plant. With genetic constructs employing inducible gene promoter sequences, the rate of DNA transcription can be modulated by external stimuli, such as light, heat, anaerobic stress, alteration in nutrient conditions and the like.

The inventive genetic constructs further comprise a gene termination sequence which is located 3′ to the polynucleotide of interest. A variety of gene termination sequences which may be usefully employed in the genetic constructs of the present invention are.well known in the art. One example of such a gene termination sequence is the 3′ end of the Agrobacterium tumefaciens nopaline synthase gene. The gene termination sequence may be endogenous to the target plant or may be exogenous, provided the promoter is functional in the target plant. For example, the termination sequence may be from other plant species, plant viruses, bacterial plasmids and the like.

The genetic constructs of the present invention may also contain a selection marker that is effective in cells of the target organism, such as a plant, to allow for the detection of transformed cells containing the inventive construct. Such markers, which are well known in the art, typically confer resistance to one or more toxins. One example of such a marker is the NPTII gene whose expression results in resistance to kanamycin or hygromycin, antibiotics which are usually toxic to plant cells at a moderate concentration (Rogers et al., in Weissbach A and H, eds. Methods for Plant Molecular Biology , Academic Press Inc.: San Diego, Calif., 1988). Transformed cells can thus be identified by their ability to grow in media containing the antibiotic in question. Alternatively, the presence of the desired construct in transformed cells can be determined by means of other techniques well known in the art, such as Southern and Western blots.

Techniques for operatively linking the components of the inventive genetic constructs are well known in the art and include the use of synthetic linkers containing one or more restriction endonuclease sites as described, for example, by Sambrook et al., ( Molecular cloning: a laboratory manual , CSHL Press: Cold Spring Harbor, N.Y., 1989). The genetic construct of the present invention may be linked to a vector having at least one replication system, for example E. coli , whereby after each manipulation, the resulting construct can be cloned and sequenced and the correctness of the manipulation determined.

The genetic constructs of the present invention may be used to transform a variety of target organisms including, but not limited to, plants. Plants which may be transformed using the inventive constructs include both monocotyledonous angiosperms (e.g., grasses, corn, grains, oat, wheat and barley) and dicotyledonous angiosperms (e.g., Arabidopsis, tobacco, legumes, alfalfa, oaks, eucalyptus, maple), and Gymnosperms (e.g., Scots pine; see Aronen, Finnish Forest Res. Papers , Vol. 595, 1996), white spruce (Ellis et al., Biotechnology 11:84-89, 1993), and larch (Huang et al., In Vitro Cell 27:201-207, 1991). In a preferred embodiment, the inventive genetic constructs are employed to transform woody plants, herein defined as a tree or shrub whose stem lives for a number of years and increases in diameter each year by the addition of woody tissue. Preferably the target plant is selected from the group consisting of eucalyptus and pine species, most preferably from the group consisting of Eucalyptus grandis and Pinus radiata . Other species which may be usefully transformed with the genetic constructs of the present invention include, but are not limited to: pines such as Pinus banksiana, Pinus brutia, Pinus caribaea, Pinus clausa, Pinus contorta, Pinus coulteri, Pinus echinata, Pinus eldarica, Pinus ellioti, Pinus jeffreyi, Pinus lambertiana, Pinus monticola, Pinus nigra, Pinus palustrus, Pinus pinaster, Pinus ponderosa, Pinus resinosa, Pinus rigida, Pinus serotina, Pinus strobus, Pinus sylvestris, Pinus taeda, Pinus virginiana ; other gymnosperrns, such as Abies amabilis, Abies balsamea, Abies concolor, Abies grandis, Abies lasiocarpa, Abies magnifica, Abies procera, Chamaecyparis lawsoniona, Chamaecyparis nootkatensis, Chamaecyparis thyoides, Huniperus virginiana, Larix decidua, Larix laricina, Larix leptolepis, Larix occidentalis, Larix siberica, Libocedrus decurrens, Picea abies, Picea engelmanni, Picea glauca, Picea mariana, Picea punggens, Picea rubens, Picea sitchensis, Pseudotsuga menziesii, Sequoia gigantea, Sequoia sempervirens, Taxodium distichum, Tsuga canadensis, Tsuga heterophylla, Tsuga mertensiana, Thuja occidentalis, Thuja plicata ; and Eucalypts, such as Eucalyptus alba, Eucalyptus bancroftii, Eucalyptus botyroides, Eucalyptus bridgesiana, Eucalyptus calophylla, Eucalyptus camaldulensis, Eucalyptus citriodora, Eucalyptus cladocalyx, Eucalyptus coccifera, Eucalyptus curtisii, Eucalyptus dalrympleana, Eucalyptus deglupta, Eucalyptus delagatensis, Eucalyptus diversicolor, Eucalyptus dunnii, Eucalyptus ficifolia, Eucalyptus globulus, Eucalyptus gomphocephala, Eucalyptus gunnii, Eucalyptus henryi, Eucalyptus laevopinea, Eucalyptus macarthurii, Eucalyptus macrorhyncha, Eucalyptus maculata, Eucalyptus marginata, Eucalyptus megacarpa, Eucalyptus melliodora, Eucalyptus nicholii, Eucalyptus nitens, Eucalyptus nova - anglica, Eucalyptus obliqua, Eucalyptus obtusiflora, Eucalyptus oreades, Eucalyptus pauciflora, Eucalyptus polybractea, Eucalyptus regnans, Eucalyptus resinifera, Eucalyptus robusta, Eucalyptus rudis, Eucalyptus saligna, Eucalyptus sideroxylon, Eucalyptus stuartiana, Eucalyptus tereticornis, Eucalyptus torelliana, Eucalyptus urnigera, Eucalyptus urophylla, Eucalyptus viminalis, Eucalyptus viridis, Eucalyptus wandoo and Eucalyptus youmanni ; and hybrids of any of these species.

›DETAILED DESCRIPTION OF THE INVENTION · 6 of 7

Techniques for stably incorporating genetic constructs into the genome of target plants are well known in the art and include Agrobacterium tumefaciens mediated introduction, electroporation, protoplast fusion, injection into reproductive organs, injection into immature embryos, high velocity projectile introduction and the like. The choice of technique will depend upon the target plant to be transformed. For example, dicotyledonous plants and certain monocots and gymnosperms may be transformed by Agrobacterium Ti plasmid technology, as described, for example by Bevan, Nucleic Acids Res . 12:8711-8721, 1984. Targets for the introduction of the genetic constructs of the present invention include tissues, such as leaf tissue, dissociated cells, protoplasts, seeds, embryos, meristematic regions; cotyledons, hypocotyls, and the like. The preferred method for transforming eucalyptus and pine is a biolistic method using pollen (see, for exanple, Aronen, Finnish Forest Res. Papers , Vol. 595, 53pp, 1996) or easily regenerable embryonic tissues.

Once the cells are transformed, cells having the inventive genetic construct incorporated in their genome may be selected by means of a marker, such as the kanamycin resistance marker discussed above. Transgenic cells may then be cultured in an appropriate medium to regenerate whole plants, using techniques well known in the art. In the case of protoplasts, the cell wall is allowed to reform under appropriate osmotic conditions. In the case of seeds or embryos, an appropriate germination or callus initiation medium is employed. For explants, an appropriate regeneration medium is used. Regeneration of plants is well established for many species. For a review of regeneration of forest trees see Dunstan et al., “Somatic embryogenesis in woody plants,” in Thorpe T A, ed., In Vitro Embryogenesis of Plants ( Current Plant Science and Biotechnology in Agriculture Vol. 20), Chapter 12, pp. 471-540, 1995. Specific protocols for the regeneration of spruce are discussed by Roberts et al., “Somatic embryogenesis of spruce,” in Redenbaugh K, ed., Synseed: applications of synthetic seed to crop improvement , CRC Press: Chapter 23, pp. 427-449, 1993). Transformed plants having the desired phenotype may be selected using techniques well known in the art. The resulting transformed plants may be reproduced sexually or asexually, using methods well known in the art, to give successive generations of transgenic plants.

As discussed above, the production of RNA in target cells can be controlled by choice of the promoter sequence, or by selecting the number of functional copies or the site of integration of the polynucleotides incorporated into the genome of the target host. A target organism may be transformed with more than one genetic construct of the present invention, thereby modulating the activity of more than gene. Similarly, a genetic construct may be assembled containing more than one open reading frame coding for a polypeptide of interest or more than one untranslated region of a gene coding for such a polypeptide.

The isolated polynucleotides of the present invention also have utility in genome mapping, in physical mapping, and in positional cloning of genes. As detailed below, the polynucleotide sequences identified as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, and their variants, may be used to design oligonucleotide probes and primers. Oligonucleotide probes designed using the polynucleotides of the present invention may be used to detect the presence and examine the expression patterns of genes in any organism having sufficiently similar DNA and RNA sequences in their cells using techniques that are well known in the art, such as slot blot DNA hybridization techniques. Oligonucleotide primers designed using the polynucleotides of the present invention may be used for PCR amplifications. Oligonucleotide probes and primers designed using the polynucleotides of the present invention may also be used in connection with various microarray technologies, including the microarray technology of Synteni (Palo Alto, Calif.).

As used herein, the term “oligonucleotide” refers to a relatively short segment of a polynucleotide sequence, generally comprising between 6 and 60 nucleotides, and comprehends both probes for use in hybridization assays and primers for use in the amplification of DNA by polymerase chain reaction.

An oligonucleotide probe or primer is described as “corresponding to” a polynucleotide of the present invention, including one of the sequences set out as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or a variant, if the oligonucleotide probe or primer, or its complement, is contained within one of the sequences set out as SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120, or a variant of one of the specified sequences. Oligonucleotide probes and primers of the present invention are substantially complementary to a polynucleotide disclosed herein.

Two single stranded sequences are said to be substantially complementary when the nucleotides of one strand, optimally aligned and compared, with the appropriate nucleotide insertions and/or deletions, pair with at least 80%, preferably at least 90% to 95% and more preferably at least 98% to 100% of the nucleotides of the other strand. Alternatively, substantial complementarity exists when a first DNA strand will selectively hybridize to a second DNA strand under stringent hybridization conditions. Stringent hybridization conditions for determining complementarity include salt conditions of less than about 1 M, more usually less than about 500 mM, and preferably less than about 200 mM. Hybridization temperatures can be as low as 5° C., but are generally greater than about 22° C., more preferably greater than about 30° C., and most preferably greater than about 37° C. Longer DNA fragments may require higher hybridization temperatures for specific hybridization. Since the stringency of hybridization may be affected by other factors such as probe composition, presence of organic solvents and extent of base mismatching, the combination of parameters is more important than the absolute measure of any one alone.

›DETAILED DESCRIPTION OF THE INVENTION · 7 of 7

In specific embodiments, the oligonucleotide probes and/or primers comprise at least about 6 contiguous residues, more preferably at least about 10 contiguous residues, and most preferably at least about 20 contiguous residues complementary to a polynucleotide sequence of the present invention. Probes and primers of the present invention may be from about 8 to 100 base pairs in length or, preferably from about 10 to 50 base pairs in length or, more preferably from about 15 to 40 base pairs in length. The probes can be easily selected using procedures well known in the art, taking into account DNA-DNA hybridization stringencies, annealing and melting temperatures, and potential for formation of loops and other factors, which are well known in the art. Preferred techniques for designing PCR primers are disclosed in Dieffenbach, C W and Dyksler, G S. PCR Primer: a laboratory manual , CSHL Press: Cold Spring Harbor, N.Y., 1995. A software program suitable for designing probes, and especially for designing PCR primers, is available from Premier Biosoft International, 3786 Corina Way, Palo Alto, Calif. 94303-4504.

A plurality of oligonucleotide probes or primers corresponding to a polynucleotide of the present invention may be provided in a kit form. Such kits generally comprise multiple DNA or oligonucleotide probes, each probe being specific for a polynucleotide sequence. Kits of the present invention may comprise one or more probes or primers corresponding to a polynucleotide of the present invention, including a polynucleotide sequence identified in SEQ ID NO: 1-14, 20, 22-62, 81-86 and 88-120.

In one embodiment useful for high-throughput assays, the oligonucleotide probe kits of the present invention comprise multiple probes in an array format, wherein each probe is immobilized at a predefined, spatially addressable location on the surface of a solid substrate. Array formats which may be usefully employed in the present invention are disclosed, for example, in U.S. Pat. Nos. 5,412,087 and 5,545,451; and PCT Publication No. WO 95/00450, the disclosures of which are hereby incorporated by reference.

The polynucleotides of the present invention may also be used to tag or identify an organism or reproductive material therefrom. Such tagging may be accomplished, for example, by stably introducing a non-disruptive non-functional heterologous polynucleotide identifier into an organism, the polynucleotide comprising one of the polynucleotides of the present invention.

The following examples are offered by way of illustration and not by way of limitation.

›Examples19
›EXAMPLE 1 · 1 of 2

Isolation and Characterization of a Ubiquitin Gene Promoter from Pinus radiata

Pinus radiata cDNA expression libraries were constructed and screened as follows. mRNA was extracted from plant tissue using the protocol of Chang et al., Plant Molecular Biology Reporter 11:113-116, 1993 with minor modifications. Specifically, samples were dissolved in CPC-RNAXB (100 mM Tris-Cl, pH 8,0; 25 mM EDTA; 2.0 M NaCl; 2% CTAB; 2% PVP and 0.05% Spernidine*3HCl) and extracted with chloroform:isoamyl alcohol, 24:1. mRNA was precipitated with ethanol and the total RNA preparate was purified using a Poly(A) Quik mRNA Isolation Kit (Stratagene, La Jolla, Calif.). A cDNA expression library was constructed from the purified mRNA by reverse transcriptase synthesis followed by insertion of the resulting cDNA clones in Lambda ZAP using a ZAP Express cDNA Synthesis Kit (Stratagene), according to the manufacturer's protocol. The resulting cDNAs were packaged using a Gigapack II Packaging Extract (Stratagene) employing 1 μl of sample DNA from the 5 μl ligation mix. Mass excision of the library was done using XL1-Blue MRF' cells and XLOLR cells (Stratagene) with ExAssist helper phage (Stratagene). The excised phagemids were diluted with NZY broth (Gibco BRL, Gaithersburg, Md.) and plated out onto LB-kanamycin agar plates containing X-gal and isopropylthio-beta-galactoside (IPTG).

Of the colonies plated and picked for DNA miniprep, 99% contained an insert suitable for sequencing. Positive colonies were cultured in NZY broth with kanamycin and cDNA was purified by means of alkaline lysis and polyethylene glycol (PEG) precipitation. Agarose gel at 1% was used to screen sequencing templates for chromosomal contamination. Dye primer sequences were prepared using a Turbo Catalyst 800 machine (Perkin Elmer/Applied Biosystems Division, Foster City, Calif.) according to the manufacturer's protocol.

DNA sequence for positive clones was obtained using a Perkin Elmer/Applied Biosystems Division Prism 377 sequencer. cDNA clones were sequenced first from the 5′ end and, in some cases, also from the 3′ end. For some clones, internal sequence was obtained using subcloned fragments. Subcloning was performed using standard procedures of restriction mapping and subcloning to pBluescript II SK+vector.

As described below, one of the most abundant sequences identified was a ubiquitin gene, hereinafter referred to as the “Super-Ubiquitin” gene.

Isolation of cDNA Clones Containing the Ubiquitin Gene

Sequences of cDNA clones with homology to the ubiquitin gene were obtained from high-throughput cDNA sequencing as described above. Sequences from several independent clones were assembled in a contig and a consensus sequence was generated from overlapping clones. The determined nucleotide sequence of the isolated Super Ubiquitin clone, comprising the promoter region (including an intron), coding region and 3′ untranslated region (UTR) is provided in SEQ ID NO: 1. The 5′ UTR is represented by residues 1 to 2064, the intron by residues 1196 to 2033, and the coding region of the gene, which contains three direct repeats, by residues 2065 to 2751. The 3′ UTR is 328 residues long (residues 2755 to 3083). The nucleotide sequence of the Super Ubiquitin promoter region only, including the intron, is given in SEQ ID NO: 2. The nucleotide sequence of the Super Ubiquitin promoter region only, excluding the intron, is given in SEQ ID NO: 3. The predicted amino acid sequence for the Pinus radiata Super Ubiquitin is provided in SEQ ID NO: 80.

Ubiquitin proteins function as part of a protein degradation pathway, in which they covalently attach to proteins, thereby targeting them for degradation (for a review, see Belknap and Garbarino, Trends in Plant Sciences 1:331-335, 1996). The protein is produced from a precursor polypeptide, encoded by a single mRNA. The Super Ubiquitin mRNA contains three copies of the ubiquitin monomer.

Cloning of the Super Ubiquitin Promoter

Fragments of the Super Ubiquitin promoter were cloned by two different PCR-based approaches.

Method 1: Long Distance Gene Walking PCR

Using “Long Distance Gene Walking” PCR (Min and Powell, Biotechniques 24:398-400, 1998), a 2 kb fragment was obtained that contained the entire coding region of the ubiquitin gene, a 900 bp intron in the 5′ UTR and approximately 100 bp of the promoter.

To generate this fragment, 2 nested primers were designed from the 3′ UTR of the Super Ubiquitin cDNA sequence isolated from pine. Generally, the 5′ UTR is used for primer design to amplify upstream sequence. However, the available 5′ UTR of Super Ubiquitin was very short, and two initial primers derived from this region failed to amplify any fragments. Therefore, the primers of SEQ ID NO: 15 and 16 were designed from the 3′ UTR.

The method involved an initial, linear PCR step with pine genomic DNA as template using the primer of SEQ ID NO: 15, and subsequent C-tailing of the single stranded DNA product using terminal transferase. The second PCR-step used these fragments as template for amplification with the primer of SEQ ID NO: 16 and primer AP of SEQ ID NO: 17. The AP primer was designed to bind to the polyC tail generated by the terminal transferase. Both primers (SEQ ID NO: 16 and 17) contained a 5′-NotI restriction site for the cloning of products into the NotI site of a suitable vector. The final PCR product contained fragments of different sizes. These fragments were separated by electrophoresis and the largest were purified from the gel, digested with restriction endonuclease NotI and cloned in the NotI site of expression vector pBK-CMV (Stratagene, La Jolla, Calif.). The largest of these clones contained the complete coding region of the gene. (no introns were found in the coding sequence) and a 5′ UTR which contained a 900 bp intron.

Method 2: “Genome Walker” Kit

The Super Ubiquitin gene promoter was cloned using a “Genome Walker” kit (Clontech, Palo Alto, Calif.). This is also a PCR-based method, which requires 2 PCR primers to be constructed, one of which must be gene-specific. Although the ubiquitin coding region is highly conserved, the 5′ UTR from different ubiquitin genes is not conserved and could therefore be used to design a gene-specific primer. A 2.2 kb fragment was amplified and subcloned in pGEM-T-easy (Promega, Madison, Wis.). Analysis by PCR and DNA sequencing showed that the clone contained 5′ UTR sequence of the Super Ubiquitin gene, including the 900 bp intron and approximately 1 kb of putative promoter region. An intron in the 5′ UTR is a common feature of plant polyubiquitin genes and may be involved in determining gene expression levels.

›EXAMPLE 1 · 2 of 2

The gene specific primers used for these PCR reactions are provided in SEQ ID NO: 18 and 19.

Expression of Super Ubiquitin

Using primers derived from the gene-specific 5′ and 3′ UTR sequences, expression levels of Super Ubiquitin in different plant tissues was examined by means of RT-PCR. Super Ubiquitin was found to be expressed in all plant tissues examined, including branch phloem and xylem, feeder roots, fertilized cones, needles, one year old cones, pollen sacs, pollinated cones, root xylem, shoot buds, structural roots, trunk phloem and trunk. Expression of Super Ubiquitin in plant tissues was also demonstrated in a Northern blot assay using a PCR probe prepared from the 5′ UTR.

Functional Analysis of the Super Ubiquitin Promoter

To test the function of the Super Ubiquitin promoter in plants, Arabidopsis thaliana was transformed with constructs containing the reporter gene for Green Fluorescent Protein (GFP) operably linked to either the Super Ubiquitin promoter of SEQ ID NO: 2 or SEQ ID NO: 3 (i.e., either with or without the infron). Constructs lacking a promoter were used as a negative control, with a plant T-DNA vector carrying a CaMV 35S promoter cloned in front of GFP being used as a positive control. The constructs were introduced into Arabidopsis via Agrobacterium-mediated transformation.

All the plant culture media were according to the protocol of Valvekens and Van Montagu, Proc. Natl. Acad. Sci . USA 85:5536-5540, 1988 with minor modifications. For root transformation, sterilized seeds were placed in a line on the surface of germination medium, the plates were placed on their sides to facilitate root harvesting, and the seeds were grown for two weeks at 24° C. with a 16 h photoperiod.

Expression of the constructs was measured by determining expression levels of the reporter gene for Green Fluorescent Protein (GFP). Preliminary GFP expression (transient) was detected in early transgenic roots during T-DNA transfer. Transgenic roots that developed green callus, growing on shoot-inducing medium containing 50 μg/ml Kanamycin and 100 μg/ml Timentin, were further tested for GFP expression. After several weeks of stringent selection on Kanamycin medium, several independent transgenic Arabidopsis lines were engineered and tested for GFP expression.

Expression was seen both with the Super Ubiquitin promoter including intron and the Super Ubiquitin promoter without the intron. However, preliminary results indicated that the levels of expression obtained with the Super Ubiquitin intron-less promoter construct were significantly higher than those seen with the promoter including intron, suggesting that the intron may contain a repressor. The sequence of the intron is provided in SEQ ID NO: 21.

›EXAMPLE 2

Isolation of a CDC Promoter from Pinus radiata

Plant EST sequences homologous to the Cell Division Control (CDC) protein gene were isolated from a Pinus radiata cDNA expression library as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, 5′ UTR sequence containing the putative promoter of the P. radiata CDC gene was isolated from genomic DNA. The determined nucleotide sequence is given in SEQ ID NO: 4.

›EXAMPLE 3

Isolation of a Xylogenesis-Specific Promoter from Pinus radiata

Plant EST sequences specific for plant xylogenesis were isolated from Pinus radiata cDNA expression libraries prepared from xylem, essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, sequences containing putative Pinus radiata xylogenesis-specific promoters were isolated from genomic DNA. The determined nucleotide sequences are provided in SEQ ID NO: 5 and 41-44. An extended cDNA sequence for the clone of SEQ ID NO: 41-44 is provided in SEQ ID NO: 92.

›EXAMPLE 4

Isolation of a 4-Coumarate-CoA Ligase Promoter from Pinus radiata

Plant EST sequences homologous to the 4-Coumarate-CoA Ligase (4CL) gene were isolated from a Pinus radiata cDNA expression library as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, sequences containing the putative promoter of the P. radiata 4CL gene was isolated from genomic DNA. The determined nucleotide'sequence is given in SEQ ID NO: 6.

Genetic constructs comprising the reporter gene for Green Fluorescent Protein (GFP) or GUS reporter genes operably linked to the promoter of SEQ ID NO: 6 were prepared and used to transform Arabidopsis thaliana plants.

›EXAMPLE 5

Isolation of a Cellulose Synthase Promoter from Eucalyptus grandis

Plant EST sequences homologous to the cellulose synthase gene were isolated from a Eucalyptus grandis cDNA expression library essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, 5′ UTR sequences containing the putative promoter of the E. grandis cellulose synthase gene were isolated from genomic DNA. Independent PCR experiments using different DNA bands as templates yielded two sequences which contained a number of base differences. One band was 750 bp in length and the nucleotide sequence of this band is given in SEQ ID NO: 7. The other band was 3 kb in length. The sequence of the 3′ end of this band corresponded to the sequence given in SEQ ID NO: 7, with a number of base pair differences. The sequence of this 3′ end is given in SEQ ID NO: 8. The sequence of the 5′ end of this band is given in SEQ ID NO: 20.

›EXAMPLE 6

Isolation of a Leaf-Specific Promoter from Eucalyptus grandis

Plant EST sequences specific for leaf were isolated from Eucalyptus grandis cDNA expression libraries prepared from leaf tissue, essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, 5′ UTR sequence containing a leaf-specific promoter of a novel E. grandis gene (of unknown function) was isolated from genomic DNA. Independent PCR experiments using different DNA bands as templates yielded three sequences which contained a number of base differences and deletions. The determined nucleotide sequences of the three PCR fragments are given in SEQ ID NO: 9-11.

›EXAMPLE 7

Isolation of an O-Methyl Transferase Promoter from Eucalyptus grandis

Plant EST sequences homologous to an O-methyl transferase (OMT) gene were isolated from a Eucalyptus grandis cDNA expression library essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, 5′UTR sequences containing the putative promoter of the E. grandis OMT gene was isolated from genomic DNA. The determined nucleotide sequence is given in SEQ ID NO: 12. This promoter sequence was extended by further sequencing. The extended cDNA sequences are given in SEQ ID NO: 60 and 113.

Genetic constructs comprising the reporter gene for Green Fluorescent Protein (GFP) operably linked to the promoter of SEQ ID NO: 12 were prepared and used to transform Arabidopsis thaliana.

›EXAMPLE 8

Isolation of Root-Specific Promoters from Pinus radiata

Plant EST sequences homologous to the root-specific receptor-like kinase gene were isolated from a Pinus radiata cDNA expression library as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, 5′UTR sequence containing a putative P. radiata root-specific promoter was isolated from genomic DNA. Two independent PCR experiments yielded sequences that contained a number of base differences. The determined nucleotide sequences from the two experiments are given in SEQ ID NO: 13, 14, 110 and 111.

›EXAMPLE 9

Isolation of an EF1-alpha Promoter from Eucalyptus Grandis

Plant EST sequences homologous to the Eucalyptus Elongation Factor-alpha (EF1-alpha) gene were isolated from a Eucalyptus grandis cDNA expression library and used to screen a Eucalyptus grandis genomic DNA library as follows.

The Eucalyptus grandis genomic DNA library was constructed using genomic DNA extracted from Eucalyptus nitens x grandis plant tissue, according to the protocol of Doyle and Doyle, Focus 12:13-15, 1990, with minor modifications. Specifically, plant tissue was ground under liquid nitrogen and dissolved in 2×CTAB extraction buffer (2% CTAB, hexadecyltrimethylammonium bromide; 1.4 M NaCl, 20 mM EDTA pH 8.0, 100 mM Tris.HCl pH 8.0, 1% polyvinylpyrollidone). After extraction with chloroform: isoamylalcohol (24:1), 10% CTAB was added to the aqueous layer and the chloroform:isoamylalcohol extraction repeated. Genomic DNA was precipitated with isopropanol.

The resulting DNA was digested with restriction endonuclease Sau3Al following standard procedures, extracted once with phenol:chloroform:isoamylalcohol (25:24:1) and ethanol precipitated. The digested fragments were separated on a sucrose density gradient using ultracentrifugation. Fractions containing fragments of 9-23 kb were pooled and ethanol precipitated. The resulting fragments were cloned into the lambda DASH II/BamHI vector (Stratagene, La Jolla, Calif.) following the manufacturer's protocol and packaged using a Gigapack II Packaging Extract (Stratagene). The library was amplified once.

The library was screened with radio-labeled EST fragments isolated from a Eucalyptus grandis library (as, described in Example 1), that showed homology to the Eucalyptus EF1-alpha gene. Phage lysates were prepared from positive plaques and genomic DNA was extracted.

From this genomic DNA, the 5′UTR region containing the putative promoter of the Eucalyptus EF1-alpha gene was obtained using the ELONGASE Amplification System (Gibco BRL). A 10 kb fragment was amplified and restriction mapped. The putative promoter region of the Eucalyptus elongation factor A (EF1-alpha) gene was identified on a 4 kb fragment, which was subcloned into a pUC19 vector (Gibco BRL) containing an engineered NotI-site. The determined genomic DNA sequences of the isolated fragment containing the promoter region are provided in SEQ ID NO: 61 and 62, with the predicted amino acid encoded by SEQ ID NO: 61 being provided in SEQ ID NO: 79.

›EXAMPLE 10

Isolation of Flower-Specific Promoters from Eucalyptus grandis

Plant EST sequences specific for flower-derived tissue were isolated from Eucalyptus grandis cDNA expression libraries prepared from flower tissue, essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, several sequences, each containing a putative Eucalyptus grandis flower-specific promoter, were isolated from genomic DNA. The determined nucleotide sequences are given in SEQ ID NO: 29-33 and 59. An extended cDNA sequence of the clone of SEQ ID NO: 30-33 is provided in SEQ ID NO: 89. An extended cDNA sequence of the clone of SEQ ID NO: 29 is provided in SEQ ID NO: 90.

›EXAMPLE 11

Isolation of Pollen-Specific Promoters from Eucalyptus grandis and Pinus radiata

Plant EST sequences specific for pollen were isolated from Eucalyptus grandis and Pinus radiata cDNA expression libraries prepared from pollen, essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, several sequences, each containing a putative pollen-specific promoter, were isolated from genomic DNA. The determined nucleotide sequences isolated from Pinus radiata are given in SEQ ID NO: 49-53, with the predicted amino acid sequences encoded by SEQ ID NO: 51-53 being provided in SEQ ID NO: 73-75, respectively. An extended cDNA sequence for the clone of SEQ ID NO: 49 is provided in SEQ ID NO: 94.

›EXAMPLE 12

Isolation of Bud-Specific and Meristem-Specific Promoter from Pinus radiata

Plant EST sequences specific for bud and meristem were isolated from Pinus radiata cDNA expression libraries prepared from bud and meristem, essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, two sequences, one containing a putative bud-specific promoter and the other containing a putative meristem-specific promoter, were isolated from genomic DNA. The determined nucleotide sequences for these two promoters are given in SEQ ID NO: 40 and 45, respectively. The predicted amino acid sequences encoded by the DNA sequences of SEQ ID NO: 40 and 45 are provided in SEQ ID NO: 70 and 71, respectively.

›EXAMPLE 13

Isolation of Promoters from Eucalyptus grandis

Plant EST sequences showing some homology to various known genes were isolated from Eucalyptus grandis cDNA expression libraries essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, sequences containing the putative promoters for the following E. grandis genes were isolated from genomic DNA: auxin induced protein (SEQ ID NO: 26-28); carbonic anhydrase (SEQ ID NO: 36); isoflavone reductase (SEQ ID NO: 37 and 38); pollen allergen (SEQ ID NO: 23-25); pollen coat protein (SEQ ID NO: 22), sucrose synthase (SEQ ID NO: 56-58); ubiquitin (SEQ ID NO: 34); glyceraldehyde-3-phosphate dehydrogenase (SEQ ID NO: 35 and 39); O-methyl transferase (OMT; SEQ ID NO: 60); macrophage migration inhibition factor from mammals (MIF; SEQ ID NO: 81-86); UDP glucose 6-dehydrogenase (SEQ ID NO: 103); laccase 1 (SEQ ID: NO: 105, 106 and 116); arabinogalactan-like 1 (SEQ ID NO: 107); arabinogalactan-like 2 (SEQ. ID NO: 108, 109), a hypothetical protein (SEQ ID NO: 104), constans (SEQ ID NO: 118) and Flowering Promoting Factor 1 (FPF1; SEQ ID NO: 119). The predicted amino acid sequences encoded by the DNA sequences of SEQ ID NO: 22, 25, 26, 28, 34, 35, 36, 56, 57, 60 and 86 are provided in SEQ ID NO: 63, 64, 65, 66, 67, 68, 69, 76, 77, 78 and 87 respectively. Extended cDNA sequences for the clones of SEQ ID NO: 58, 35, 60, 103, 106 and 107 are provided in SEQ ID NO: 91, 93, 113, 115-117 respectively.

›EXAMPLE 14

Isolation of Promoters from Pinus radiata

Plant EST sequences showing some homology to various known genes were isolated from Pinus radiata cDNA expression libraries essentially as described in Example 1. Using the “Genome Walker” protocol described above and gene specific primers designed from these plant EST sequences, sequences containing the putative promoters for the following Pinus radiata genes were isolated from genomic DNA: senescence-like protein (SEQ ID NO: 46-48); nodulin homolog pollen specific (SEQ ID NO: 54 and 55); chalcone synthase (SEQ ID NO: 88); PrMALE1 (SEQ ID NO: 95, 96); UDP glucose glycosyltransferase (SEQ ID NO: 97); elogation factor 1 alpha (SEQ ID NO: 98, 99); S-adenosylmethionine synthase (SEQ ID NO: 100-102); Pinus radiata lipid transfer protein 2 (PrLTP2; SEQ ID NO: 112) and Pinus radiata agamous protein (SEQ ID NO: 120). The predicted amino acid sequence encoded by the sequence of SEQ ID NO: 46 is provided in SEQ ID NO: 72. An extended cDNA sequence for the clone of SEQ ID NO: 97 is provided in SEQ ID NO: 114.

›EXAMPLE 15

Polynucleotide and Amino Acid Analysis

The determined cDNA sequences described above were compared to and aligned with known sequences in the EMBL database (as updated to April 2000). Specifically, the polynucleotides identified in SEQ ID NO: 22-62 and 88-120 were compared to polynucleotides in the EMBL database using the BLASTN algorithm Version 2.0.6 [Sep. 16, 1998] set to the following running parameters: Unix running command: blastall -p blastn -d embldb -e 10 -G0 -E0 -r1 -v30 -b3.0 -i queryseq -o results. Multiple alignments of redundant sequences were used to build up reliable consensus sequences. Based on similarity to known sequences from other plant or non-plant species, the isolated polynucleotides of the present invention identified as SEQ ID NO: 22-62 and 88-120 were putatively identified as having the functions shown in Table 1, above.

The cDNA sequences of SEQ ID NO: 1-22, 23, 25-42, 45-49, 57-59, 62, 88-99, 101-112 and 114-120 were determined to have less than 40% identity to sequences in the EMBL database using the computer algorithm BLASTN, as described above. The cDNA sequences of SEQ ID NO: 56 and 113 was determined to have less than 60% identity to sequences in the EMBL database using BLASTN, as described above. The cDNA sequences of SEQ ID NO: 43, 52, 60 and 61 were determined to have less than 75% identity to sequences in the EMBL database using BLASTN, as described above. The cDNA sequences of SEQ ID NO: 24, 51 and 100 were determined to have less than 90% identity to sequences in the EMBL database using BLASTN, as described above.

›EXAMPLE 16

Modification of a Reporter Gene under Control of the Superubiquitin Promoter

Six independent Arabidopsis thaliana transgenic lines were transformed with Pinus radiata superubiquitin promoter constructs to demonstrate the relative expression of a GUS reporter gene under control of different superubiquitin promoter constructs. The reporter constructs in the plasmid pBI-101 contained the GUS (β-D-glucuronidase) reporter gene in frame with the superubiquitin promoter with the intron (SEQ ID NO: 2), the superubiquitin promoter without the intron (SEQ ID NO: 3), and the CaMV 35S promoter. A reporter gene construct without a promoter sequence was used as control.

Groups of six Arabidopsis thaliana plants were transformed with the reporter constructs described above, using Agrobacterium tumefaciens transformation protocols. A. tumefaciens was transformed with 100 ng of the plasmid DNA according to standard techniques, as described, for example, by Bevan ( Nucleic Acids Res . 12:8711-8721, 1984). Fresh plant material was collected from each plant, protein extracted from the whole plant, and the protein concentration determined (Bradford, Anal. Biochem . 72:248-254, 1976). The protein samples were diluted with carrier bovine serum albumin to 100 ng protein to maintain readings on the fluorimeter in the linear part of the standard curve using 4-methyl-umbelliferone (MU). GUS activity was quantified by fluorimetric analysis, using a Victor 1420 multi-label counter (Wallac, Turku, Finland) as described by Jefferson ( Plant Mol. Biol. Rep . 5:387-405, 1987). As shown in FIG. 1, the construct containing the superubiquitin promoter without the intron showed seven times more GUS activity than the CaMV 35S promoter and the construct containing the superubiquitin promoter with the intron showed sixty two times more GUS activity than the CaMV 35S promoter. No activity was detected for the promoter-less control construct.

›EXAMPLE 17 · 1 of 2

Determination of the Activity of Superubiquitin Promoter Constructs in Tobacco Plant Protoplasts

Isolation of Protoplasts

Protoplasts were isolated from sterile tobacco ( Nicotiana tabacum ) leaf tissue and transformed with superubiquitin promoter constructs. Mesophyll protoplasts were prepared according to the method of Bilang et al., Plant Molecular Biology Manual A1:1-16, 1994. A number of fully expanded leaves were removed from sterile wild type tobacco plants, sliced perpendicular to the midrib and submerged in a digestion enzyme solution containing 1.2% cellulase and 0.4% pectinase (Sigma, St. Louis Mo.). The leaves were left to incubate in the dark without agitation at 26° C. for approximately 18 hours. The leaf strips were then gently agitated for 30 min to release the protoplasts. Protoplasts were further purified by filtration through 100 μm nylon mesh. One ml of W5 solution (154 mM MgCl 2 , 125 mM CaCl 2 , 5 mM KCl, 5 mM glucose, pH 5.8-6) was carefully layered on top of the filtrate and centrifuged at 80×g for 10 min. The live protoplast layer was removed with a wide bore pipette, washed twice with 10 ml W5 solution using centrifugation at 70×g for 5 min, with final resuspension in 5 ml W5 solution. Protoplasts were counted in a hemocytometer and viability was determined under the microscope after staining with 5 mg/ml fluoroscein diacetate (FDA) in 100% acetone.

Transformation with Promoter Constructs

The isolated protoplasts were transformed with plasmid DNA using a polyethylene glycol protocol. After centrifugation of the purified protoplasts at 70×g for 5 min, they were resuspended in MMM solution (15 mM MgCl 2 , 0.1% w/v 2[N-morpholino]ethanesulfonic acid (MES), 0.5 M mannitol pH 5.8) to a density of 2×10 6 protoplasts/ml. Aliquots containing 5×10 5 protoplasts/ml in 250 μl were distributed to 15 ml tubes and mixed with 20 μg plasmid DNA. 250 μl polyethylene glycol-4000 (40%) was gently added and incubated for 5 minutes at room temperature. Ten ml W5 solution was slowly added, the protoplasts centrifuged at 70×g for 5 min and finally resuspended in 2 ml K3 medium (Bilang et al., Plant Molecular Biology Manual A1:1-16, 1994). The transformed protoplasts were incubated in the dark at 26° C. for 24 hours before protein was extracted for reporter enzyme assays using 4-methyl-umbelliferyl-glucuronide (MUG).

Protein was extracted from the protoplasts using the following protocol. Transformed protoplast suspensions were centrifuged at 70×g for 10 min, resuspended in 50 μl extraction buffer (Jefferson, Plant Mol. Biol. Rep . 5:387-405, 1987) and vigorously mixed using a vortex. The homogenate was cleared by centrifugation at 4,300 rpm for 5 min, the supernatant removed and used for protein assays (Bradford, Anal. Biochem . 72:248-254, 1976).

The results shown in FIG. 2 demonstrate the promoter activity of deletion constructs of the superubiquitin promoter without the intron (SEQ ID NO: 3) and the superubiquitin promoter with the intron (SEQ ID NO: 2) in tobacco plant protoplasts transformed as described above. The deletion constructs were made in plasmid pBI-101 that contained the GUS reporter gene, using Endonuclease III (Gibco BRL, Gaithersburg, Md.) according to the manufacturer's protocols. The deletion constructs contained 1,103; 753; 573; 446; 368 and 195 bp of superubiquitin promoter sequence, respectively, upstream of the TATA sequence (bp numbers 1,104-1,110 of SEQ ID NO: 2). A control construct containing no sequence upstream of the TATA sequence was also made. These results show that the construct containing the entire superubiquitin promoter with the intron had the highest MU activity in the protoplasts.

In FIG. 3, the tobacco protoplasts were transformed with four different promoter constructs in plasmid pBI-101 containing the GUS reporter gene. These included the superubiquitin promoter without the intron (SEQ ID NO: 3), an elongation factor lax promoter (SEQ ID NO: 99) and a 5-adenosylmethionine synthetase promoter (SEQ ID NO: 100). A promoterless control was included in the experiment, and is referred to in FIG. 3 as pBI-101.

1

aaaacccctc acaaatacat aaaaaaaatt ctttatttaa ttatcaaact ctccactacc 60

tttcccacca accgttacaa tcctgaatgt tggaaaaaac taactacatt gatataaaaa 120

aactacatta cttcctaaat catatcaaaa ttgtataaat atatccactc aaaggagtct 180

agaagatcca cttggacaaa ttgcccatag ttggaaagat gttcaccaag tcaacaagat 240

ttatcaatgg aaaaatccat ctaccaaact tactttcaag aaaatccaag gattatagag 300

taaaaaatct atgtattatt aagtcaaaaa gaaaaccaaa gtgaacaaat attgatgtac 360

aagtttgaga ggataagaca ttggaatcgt ctaaccagga ggcggaggaa ttccctagac 420

agttaaaagt ggccggaatc ccggtaaaaa agattaaaat ttttttgtag agggagtgct 480

tgaatcatgt tttttatgat ggaaatagat tcagcaccat caaaaacatt caggacacct 540

aaaattttga agtttaacaa aaataacttg gatctacaaa aatccgtatc ggattttctc 600

taaatataac tagaattttc ataactttca aagcaactcc tcccctaacc gtaaaacttt 660

tcctacttca ccgttaatta cattccttaa gagtagataa agaaataaag taaataaaag 720

tattcacaaa ccaacaattt atttctttta tttacttaaa aaaacaaaaa gtttatttat 780

tttacttaaa tggcataatg acatatcgga gatccctcga acgagaatct tttatctccc 840

tggttttgta ttaaaaagta atttattgtg gggtccacgc ggagttggaa tcctacagac 900

gcgctttaca tacgtctcga gaagcgtgac ggatgtgcga ccggatgacc ctgtataacc 960

caccgacaca gccagcgcac agtatacacg tgtcatttct ctattggaaa atgtcgttgt 1020

tatccccgct ggtacgcaac caccgatggt gacaggtcgt ctgttgtcgt gtcgcgtagc 1080

gggagaaggg tctcatccaa cgctattaaa tactcgcctt caccgcgtta cttctcatct 1140

tttctcttgc gttgtataat cagtgcgata ttctcagaga gcttttcatt caaaggtatg 1200

gagttttgaa gggctttact cttaacattt gtttttcttt gtaaattgtt aatggtggtt 1260

tctgtggggg aagaatcttt tgccaggtcc ttttgggttt cgcatgttta tttgggttat 1320

ttttctcgac tatggctgac attactaggg ctttcgtgct ttcatctgtg ttttcttccc 1380

ttaataggtc tgtctctctg gaatatttaa ttttcgtatg taagttatga gtagtcgctg 1440

tttgtaatag gctcttgtct gtaaaggttt cagcaggtgt ttgcgtttta ttgcgtcatg 1500

tgtttcagaa ggcctttgca gattattgcg ttgtacttta atattttgtc tccaaccttg 1560

ttatagtttc cctcctttga tctcacagga accctttctt ctttgagcat tttcttgtgg 1620

cgttctgtag taatatttta attttgggcc cgggttctga gggtaggtga ttattccagt 1680

›EXAMPLE 17 · 2 of 2

gatgtgcttt ccctataagg tcctctatgt gtaagctgtt agggtttgtg cgttactatt 1740

gacatgtcac atgtcacata ttttcttcct cttatccttc gaactgatgg ttctttttct 1800

aattcgtgga ttgctggtgc catattttat ttctattgca actgtatttt agggtgtctc 1860

tttctttttg atttcttgtt aatatttgtg ttcaggttgt aactatgggt tgctagggtg 1920

tctgccctct tcttttgtgc ttctttcgca gaatctgtcc gttggtctgt atttgggtga 1980

tgaattattt attccttgaa gtatctgtct aattagcttg tgatgatgtg caggtatatt 2040

cgttagtcat atttcaattt caag atg cag atc ttt gtc aag act ctc acc 2091

Met Gln Ile Phe Val Lys Thr Leu Thr

1 5

ggt aag acc atc act ctc gag gtc gag agc tct gac acc att gac aat 2139

Gly Lys Thr Ile Thr Leu Glu Val Glu Ser Ser Asp Thr Ile Asp Asn

10 15 20 25

gtt aaa gct aag atc cag gac aag gaa ggg att ccc ccc gac cag cag 2187

Val Lys Ala Lys Ile Gln Asp Lys Glu Gly Ile Pro Pro Asp Gln Gln

30 35 40

cgt ctg atc ttc gca gga aag cag ctt gag gac ggc cga acc ctt gcc 2235

Arg Leu Ile Phe Ala Gly Lys Gln Leu Glu Asp Gly Arg Thr Leu Ala

45 50 55

gat tac aac atc cag aaa gaa tct acc ctc cac ctt gtt ctc cgt ttg 2283

Asp Tyr Asn Ile Gln Lys Glu Ser Thr Leu His Leu Val Leu Arg Leu

60 65 70

agg ggt ggc atg caa atc ttt gta aaa aca cta act gga aag aca att 2331

Arg Gly Gly Met Gln Ile Phe Val Lys Thr Leu Thr Gly Lys Thr Ile

75 80 85

aca ttg gaa gtt gag agc tcg gac acc att gac aac gtc aag gcc aag 2379

Thr Leu Glu Val Glu Ser Ser Asp Thr Ile Asp Asn Val Lys Ala Lys

90 95 100 105

atc cag gac aag gaa gga att ccc cct gac cag cag agg ctt atc ttc 2427

Ile Gln Asp Lys Glu Gly Ile Pro Pro Asp Gln Gln Arg Leu Ile Phe

110 115 120

gct ggt aag cag ctg gag gat ggc agg acc ttg gct gat tac aat att 2475

Ala Gly Lys Gln Leu Glu Asp Gly Arg Thr Leu Ala Asp Tyr Asn Ile

125 130 135

caa aag gaa tcg acc ctg cat ttg gtg ctt cgt cta aga gga ggc atg 2523

Gln Lys Glu Ser Thr Leu His Leu Val Leu Arg Leu Arg Gly Gly Met

140 145 150

caa atc ttt gtg aaa acc ctt aca ggt aaa acc att act ctg gaa gtg 2571

Gln Ile Phe Val Lys Thr Leu Thr Gly Lys Thr Ile Thr Leu Glu Val

155 160 165

gaa agc tcg gac acc att gac aat gtg aag gct aag atc cag gac aag 2619

Glu Ser Ser Asp Thr Ile Asp Asn Val Lys Ala Lys Ile Gln Asp Lys

170 175 180 185

gag gga att cca cct gac cag cag agg ttg atc ttt gcc ggt aag cag 2667

Glu Gly Ile Pro Pro Asp Gln Gln Arg Leu Ile Phe Ala Gly Lys Gln

190 195 200

ctg gaa gat ggt cgt act ctc gcc gat tac aat att cag aag gaa tcg 2715

Leu Glu Asp Gly Arg Thr Leu Ala Asp Tyr Asn Ile Gln Lys Glu Ser

205 210 215

acc ctt cac ctg gtg ctc cgt ctc cgc ggt ggc ttt taggtttggg 2761

Thr Leu His Leu Val Leu Arg Leu Arg Gly Gly Phe

220 225

tgttatttgt ggataataaa ttcgggtgat gttcagtgtt tgtcgtattt ctcacgaata 2821

aattgtgttt atgtatgtgt tagtgttgtt tgtctgtttc agaccctctt atgttatatt 2881

tttcttttcg tcggtcagtt gaagccaata ctggtgtcct ggccggcact gcaataccat 2941

ttcgtttaat ataaagactc tgttatccgt tatgtaattc catgttatgt ggtgaaatgt 3001

ggatgaaatt cttagaaatt attattgtaa tttgaaactt ccttcgtcaa taatctgcac 3061

aacacattta ccaaaaaaaa aa 3083

2

aaaacccctc acaaatacat aaaaaaaatt ctttatttaa ttatcaaact ctccactacc 60

tttcccacca accgttacaa tcctgaatgt tggaaaaaac taactacatt gatataaaaa 120

aactacatta cttcctaaat catatcaaaa ttgtataaat atatccactc aaaggagtct 180

agaagatcca cttggacaaa ttgcccatag ttggaaagat gttcaccaag tcaacaagat 240

ttatcaatgg aaaaatccat ctaccaaact tactttcaag aaaatccaag gattatagag 300

taaaaaatct atgtattatt aagtcaaaaa gaaaaccaaa gtgaacaaat attgatgtac 360

aagtttgaga ggataagaca ttggaatcgt ctaaccagga ggcggaggaa ttccctagac 420

agttaaaagt ggccggaatc ccggtaaaaa agattaaaat ttttttgtag agggagtgct 480

tgaatcatgt tttttatgat ggaaatagat tcagcaccat caaaaacatt caggacacct 540

aaaattttga agtttaacaa aaataacttg gatctacaaa aatccgtatc ggattttctc 600

taaatataac tagaattttc ataactttca aagcaactcc tcccctaacc gtaaaacttt 660

tcctacttca ccgttaatta cattccttaa gagtagataa agaaataaag taaataaaag 720

tattcacaaa ccaacaattt atttctttta tttacttaaa aaaacaaaaa gtttatttat 780

tttacttaaa tggcataatg acatatcgga gatccctcga acgagaatct tttatctccc 840

tggttttgta ttaaaaagta atttattgtg gggtccacgc ggagttggaa tcctacagac 900

gcgctttaca tacgtctcga gaagcgtgac ggatgtgcga ccggatgacc ctgtataacc 960

caccgacaca gccagcgcac agtatacacg tgtcatttct ctattggaaa atgtcgttgt 1020

tatccccgct ggtacgcaac caccgatggt gacaggtcgt ctgttgtcgt gtcgcgtagc 1080

gggagaaggg tctcatccaa cgctattaaa tactcgcctt caccgcgtta cttctcatct 1140

tttctcttgc gttgtataat cagtgcgata ttctcagaga gcttttcatt caaaggtatg 1200

gagttttgaa gggctttact cttaacattt gtttttcttt gtaaattgtt aatggtggtt 1260

tctgtggggg aagaatcttt tgccaggtcc ttttgggttt cgcatgttta tttgggttat 1320

ttttctcgac tatggctgac attactaggg ctttcgtgct ttcatctgtg ttttcttccc 1380

ttaataggtc tgtctctctg gaatatttaa ttttcgtatg taagttatga gtagtcgctg 1440

tttgtaatag gctcttgtct gtaaaggttt cagcaggtgt ttgcgtttta ttgcgtcatg 1500

tgtttcagaa ggcctttgca gattattgcg ttgtacttta atattttgtc tccaaccttg 1560

ttatagtttc cctcctttga tctcacagga accctttctt ctttgagcat tttcttgtgg 1620

cgttctgtag taatatttta attttgggcc cgggttctga gggtaggtga ttattccagt 1680

gatgtgcttt ccctataagg tcctctatgt gtaagctgtt agggtttgtg cgttactatt 1740

gacatgtcac atgtcacata ttttcttcct cttatccttc gaactgatgg ttctttttct 1800

aattcgtgga ttgctggtgc catattttat ttctattgca actgtatttt agggtgtctc 1860

tttctttttg atttcttgtt aatatttgtg ttcaggttgt aactatgggt tgctagggtg 1920

tctgccctct tcttttgtgc ttctttcgca gaatctgtcc gttggtctgt atttgggtga 1980

tgaattattt attccttgaa gtatctgtct aattagcttg tgatgatgtg caggtatatt 2040

cgttagtcat atttcaattt caag 2064

3

1226

›DNA

Pinus radiata

3

aaaacccctc acaaatacat aaaaaaaatt ctttatttaa ttatcaaact ctccactacc 60

tttcccacca accgttacaa tcctgaatgt tggaaaaaac taactacatt gatataaaaa 120

aactacatta cttcctaaat catatcaaaa ttgtataaat atatccactc aaaggagtct 180

agaagatcca cttggacaaa ttgcccatag ttggaaagat gttcaccaag tcaacaagat 240

ttatcaatgg aaaaatccat ctaccaaact tactttcaag aaaatccaag gattatagag 300

taaaaaatct atgtattatt aagtcaaaaa gaaaaccaaa gtgaacaaat attgatgtac 360

aagtttgaga ggataagaca ttggaatcgt ctaaccagga ggcggaggaa ttccctagac 420

agttaaaagt ggccggaatc ccggtaaaaa agattaaaat ttttttgtag agggagtgct 480

tgaatcatgt tttttatgat ggaaatagat tcagcaccat caaaaacatt caggacacct 540

aaaattttga agtttaacaa aaataacttg gatctacaaa aatccgtatc ggattttctc 600

taaatataac tagaattttc ataactttca aagcaactcc tcccctaacc gtaaaacttt 660

tcctacttca ccgttaatta cattccttaa gagtagataa agaaataaag taaataaaag 720

tattcacaaa ccaacaattt atttctttta tttacttaaa aaaacaaaaa gtttatttat 780

tttacttaaa tggcataatg acatatcgga gatccctcga acgagaatct tttatctccc 840

tggttttgta ttaaaaagta atttattgtg gggtccacgc ggagttggaa tcctacagac 900

gcgctttaca tacgtctcga gaagcgtgac ggatgtgcga ccggatgacc ctgtataacc 960

caccgacaca gccagcgcac agtatacacg tgtcatttct ctattggaaa atgtcgttgt 1020

tatccccgct ggtacgcaac caccgatggt gacaggtcgt ctgttgtcgt gtcgcgtagc 1080

gggagaaggg tctcatccaa cgctattaaa tactcgcctt caccgcgtta cttctcatct 1140

tttctcttgc gttgtataat cagtgcgata ttctcagaga gcttttcatt caaaggtata 1200

ttcgttagtc atatttcaat ttcaag 1226

4

agtaaaattg gcccatgtag gactaagtca aaatcaaaat tccatctcta aaagcggaac 60

tttgtcccct gaaaattttg actaatttcc aaccaaaaaa aagtggggga aaatataaaa 120

ctctaactaa taaaacaata atcaccaaaa atctatcacc aaaaatgaaa aaagattttg 180

aatactaggc catatgagct acacaaattt caaaagtatc ttacacttat tacgcacccg 240

gatgtcccca ctttcgaaaa acccgtttca agcctttcac gaaagtccaa cggtcagaaa 300

attcaaaatg actgtttgag gcagagccaa tctaggacca cgctccattt atatatggcc 360

tctgcttctc tcgaccctta gagtcctctg ctctgcgaat cttgttgtta gttactgtgt 420

acgctgtaac aatggatgcc tatgagaagt tggagaaggt gggagaagga acctatggga 480

aggtg 485

5

tgagaacatg ataagctgtg taaattcatg ctagtcacca taacttttct cattgctttt 60

catccacact gttgattcat tcattatata agatcagatt cgtatgatat acaggcaacc 120

atagaaacaa ccagcaaagt tactagcagg aaatccaact aggtatcatg aagactacca 180

acgcaggctc gataatgttg gtgctcatta tttttgggtg ctgtttcatt ggggtcatag 240

ctacat 246

6

caccaattta atgggatttc agatttgtat cccatgctat tggctaagcc atttttctta 60

ttgtaatcta accaattcca atttccaccc tggtgtgaac tgactgacaa atgcggcccg 120

aaaacagcga atgaaatgtc tgggtgatcg gtcaaacaag cggtgggcga gagaacgcgg 180

gtgttggcct agccgggatg ggggtaggta gacggcgtat taccggcgag ttgtccgaat 240

ggagttttcg gggtaggtag taacgtagac gtcaatggaa aaagtcataa tctccgtcaa 300

aaatccaacc gctccttcac accgcagagt tggtggccac gggaccctcc acccactcac 360

tcaatcgatc gcctgccgtg gttgcccatt attcaaccat acgccacttg actcttcacc 420

aacaattcca ggccggcttt cgagacaatg tactgcacag gaaaatccaa tataaaaggc 480

cggcctccgc ttccttctca gtagccccca gctcattcaa ttcttcccac tgcaggctac 540

atttgtcaga cacgttttcc gccatttttc gcctgtttct gcggagaatt tgatcaggtt 600

7

agtttggaat gtgttgtgtg tgatgtgatg gagagtatca gcattccaaa catgacatgg 60

ttttaactta tctgcaatgg tttctttttt attcagcgaa ctcgatggct gatgctgaga 120

gaaatgaatt gggaagtcga tcgacaatgg cagctcaact caatgatcct caggtataag 180

catttttttg gcagctctgg tcattgtgtc ttcaactttt agatgagagc aaatcaaatt 240

gactctaata ccggttatgt gatgagtgaa tcatttgctt ttagtagctt taatttatgc 300

ccccatctta gttgggtata aaggttcaga gtgcgaagat tacatctatt ttggttcttg 360

caggacacag ggattcatgc tagacacatc agcagtgttt ctacgttgga tagtggtatg 420

tacttagcta ctataaagga aattttgata gatatgtttg atatggtgct tgtacagatc 480

tatttaatgt caatgtattt gaaactatct tgtctcataa ctttcttgaa gaatacaatg 540

atgagactgg gaaccctatc tggaagaata gagtggagag ctggaaggac a 591

8

atgctgagag aaatgaattg ggaagtcgat cgacaatggc agctcaactc aatgatcctc 60

aggtataagc atttttttgg cagctctggt cattgtgtct tcaactttta gatgagagca 120

aatcaaattg actctaatac cagttatgtg atgagtgaat catttgcttt tagtagcttt 180

aatttatgcc cccatcttag ttgggtataa aggttcagag tgcgaagatt acatctattt 240

tggttcttgc aggacacagg gattcatgct agacacatca gcagtgtttc tacgttggat 300

agtggtatgt acttagctac tataaaggaa attttgatag atatgtttga tatggtgctt 360

gtacagatct atttaatgcc aatgtatttg aaactatctt gtctcataac tttcttgaag 420

aatacaatga tgagactggg aaccctatct ggaagaatag agtggagagc tggaaggaca 480

9

gcccatctca ggtgcaacgg tttaactgat gtttactaca cgcaaggggg aggtatccgg 60

aaagcttgca aatcgggtaa aaacgaaaat gggcgacgtg gactcagcct gcccatgttt 120

tcggtctctc tcctggactt ccatgcccga taagggccgc caactctctc tctctctctc 180

tttttctctc acatctctct gcctgttcat gtcgcctgca agtgaagatt cgtcggagca 240

agaaggacga accgggcaca tggcggggtc ggcggtcgcg acggttctaa agggtctctt 300

cctggtgt 308

10

gcccatctca ggtgcaacgg tttaactgat gtttactaca cgcaaggggg aggtatccgg 60

aaagcttgca aatcgggtaa aaacgaaaat gggcgacgtg gactcagcct gcccatgttt 120

tcggtccctc tcctggactt ccatgcccga taaaggccgc caactctctc tctttttctc 180

tcacatctct ctgcctgttc atgtcgcctg caagtgaaga ttcgtcggag caagaaggac 240

gaactgggca tatggcgggg tcggcggtcg cgacggttct aaagggtctc ttcctggtgt 300

11

297

›DNA

Eucalyptus grandis

11

gtgcaacggt ttaactgatg tttactacac gcaaggggga ggtatccgga aagcttgcaa 60

atcgggtaaa aacgaaaatg ggcgacgtgg actcagcctg cccatgtttt cggtctctct 120

cctggacttc catgcccgat aagggccgcc aactctctct ctctctctct ttttctctca 180

catctctctg cctgttcatg tcgcctgcaa gtgaagattc gtcggagcaa gaaggacgaa 240

ctgggcatat ggcggggtcg gcggtcgcga cggttctaaa gggtctcttc ctggtgt 297

12

661

›DNA

Eucalyptus grandis

12

ctgagccatt taattcgaga gcacatcgcc caaaattatt cttcttgctg ccataactgt 60

cgaattttct cttttaggta agtaaccaat gatgcatcat gttgacaaaa aggctgatta 120

gtatgatctt ggagttgttg gtgcaaattt gcaagctgac gatggcccct cagggaaatt 180

aaggcgccaa cccagattgc aaagagcaca aagagcacga tccaaccttt ccttaacaag 240

atcatcacca gatcggccag taagggtaat attaatttaa caaatagctc ttgtaccggg 300

aactccgtat ttctctcact tccataaacc cctgattaat ttggtgggaa agcgacagcc 360

aacccacaaa aggtcagatg tcatcccacg agagagagag agagagagag agagagagag 420

agagttttct ctctatattc tggttcaccg gttggagtca atggcatgcg tgacgaatgt 480

acatattggt gtagggtcca atattttgcg ggagggttgg tgaaccgcaa agttcctata 540

tatcgaacct ccaccaccat acctcacttc aatccccacc atttatccgt tttatttcct 600

ctgctttcct ttgctcgagt ctcgcggaag agagagaaga gaggagagga gagaatgggt 660

t 661

13

336

›DNA

Pinus radiata

13

actagtgatt tgttgagaat gagtaggcat tgctacaccc atcatcacaa gcatcatcat 60

gaggagaaga agatccattt ctcactctat tactcgaact tccttcagat taggctgtgt 120

atttctcact ctaccactcc aacttccttc aaatgctgtg agtttttgtt gtaattgccc 180

cgtctattta taatcgcagc agcactcgtc atataaagac ccgtgtgtgt gaacaacaac 240

caagtgattt gaattggaaa tgaagagcga gaatggcggt gtcatgaccg ggagcaacca 300

gcccgggccg tcgaccacgc gtgccctata gtaatc 336

14

763

›DNA

Pinus radiata

14

actagtgatt tgttgagaat gagtaggcat tgctacaccc atcatcacaa gcatcaacat 60

gaagagaaga agacgatcca tttctcactc tatcactcca acttccttca gattaggctg 120

tgtatttctc actctaccac tccaactacc actccaactt attgccgcaa aagagagagg 180

ttcccaaact ctgtcggaat tctcccactc aaagcattaa aggaaagatc taattgctgc 240

aaaaaagaga gattcccaat atatttctca actcccttca aatgatttct cactctacca 300

ctccaactcc cttcaaatga tttctcactc taccactcca acttccttca aatgctgtga 360

gtttttgttg taattgcccc gtctatttat aatcgcagca gcactcgtca tataaagacc 420

cgtgcgtgtg aacaacaatg gcggtgtctt gactgggagc aaccgcataa agaaagtggg 480

cttcatacat taaaaaaatc tgtaaatttt acggatttgg aaaaaggaag agcaggaggg 540

acctcccgac ttgacccgag aatggcggtg tcttgaccgc gtaaagaaag tggtcttctg 600

tacccgactt gacccgaaaa aagaggaaac gttgaacgag acaatctctg ggaacttcat 660

cgaaatgaac ctcacgactt gactctttcg attgtactgt tttcattgtt cccgcgtaaa 720

acgaccagcc cgggccgtcg accacgcgtg ccctatagta atc 763

15

40

›DNA

Artificial Sequence

Made in a lab

15

acggataaca gagtctttat attaaacgaa atggtattgc 40

16

51

›DNA

Artificial Sequence

Made in a lab

16

tgacgcggcc gcgaccgacg aaaagaaaaa tataacataa gagagtctga a 51

17

27

›DNA

Artificial Sequence

Made in a lab

17

tatagcggcc gcgggggggg ggggggg 27

18

30

›DNA

Artificial Sequence

Made in a lab

18

cggagaacaa ggtggagggt agattctttc 30

19

31

›DNA

Artificial Sequence

Made in a lab

19

tctgcatctt gaaattgaaa tatgactaac g 31

20

363

›DNA

Eucalyptus grandis

20

aatcgggtga aaatagggcc gccctaaatt agaattgaca acatttcttg ggcaaagtta 60

atgtaagtta catgaaaaaa aaaaaaaagg atagtttgtt ggaagtaatg gagcatttgt 120

attgtgaaat tcacgataga gctaacaaaa ataaaggtag ttggtgggtt aacccagtta 180

aaaaagaaca ataatttgaa gagaggagag agagagagag gagggggaga gcatttcgat 240

aaattcacta gaaaaaatgc gtgttttagt ataaatgaga gtggaaatag ggccatctag 300

ggaacgatcg atcgcccctg cacccggcca tctggagagt ctgtttatac ttctctccgg 360

ctt 363

21

gtatggagtt ttgaagggct ttactcttaa catttgtttt tctttgtaaa ttgttaatgg 60

tggtttctgt gggggaagaa tcttttgcca ggtccttttg ggtttcgcat gtttatttgg 120

gttatttttc tcgactatgg ctgacattac tagggctttc gtgctttcat ctgtgttttc 180

ttcccttaat aggtctgtct ctctggaata tttaattttc gtatgtaagt tatgagtagt 240

cgctgtttgt aataggctct tgtctgtaaa ggtttcagca ggtgtttgcg ttttattgcg 300

tcatgtgttt cagaaggcct ttgcagatta ttgcgttgta ctttaatatt ttgtctccaa 360

ccttgttata gtttccctcc tttgatctca caggaaccct ttcttctttg agcattttct 420

tgtggcgttc tgtagtaata ttttaatttt gggcccgggt tctgagggta ggtgattatt 480

cncagtgatg tgctttccct ataaggtcct ctatgtgtaa gctgttaggg tttgtgcgtt 540

actattgaca tgtcacatgt cacatatttt cttcctctta tccttcgaac tgatggttct 600

ttttctaatt cgtggattgc tggtgccata ttttatttct attgcaactg tattttaggg 660

tgtctctttc tttttgattt cttgttaata tttgtgttca ggttgtaact atgggttgct 720

agggtgtctg ccctcttctt ttgtgcttct ttcgcagaat ctgtccgttg gtctgtattt 780

gggtgatgaa ttatttattc cttgaagtat ctgtctaatt agcttgtgat gatgtgcag 839

22

881

›DNA

Eucalyptus grandis

22

acgtgacgat gctcgagtct cgcgttctcc tctctcttgt tctgcaaaac agaaaagaga 60

gaatggaggt tggcctctct caattacgtg gacgccaatg agataactca ggtgggcgac 120

aaaacaaacg cctcttgatt tcctcaaacc ccaaaccgaa tccctcgtca aggggcaagg 180

cttttggtcc cgcggcccca cggatcgctc gttcccgtct cgccacgtcg cgtcgcagcg 240

tgtcgagcaa acagaggggt ccgagcgact ataaaatccc gacgccatcg acaccacagt 300

ccatcgaaaa ccttgttcaa ttcccaagtg aaagtgagta actgtgaacg aagagttgaa 360

ctttgcatct cggcgtgtgg attcaagagg aagcagcaaa gtggaaatgg acaactccaa 420

gatgggcttc aatgcagggc aggccaaggg ccagactcag gagaagagca accagatgat 480

ggataaggca tccaacactg ctcaatctgc aagggattcc atgcaagaga ctggtcagca 540

gatgaaggcc aaagcccagg gtgctgctga tgcagtgaag aatgccaccg ggatgaacaa 600

atgaagagct caagacatga atgaataaat aattaagctc tggttatcat ttgcttttcc 660

ggtcgtttgt tgtcctgttt ttccttgtca agagcttatt atgagggtcc ttttgctctt 720

tccttagttc tttttgtttc ttggttgttc catgaagaga gcaactctct gtgtttgaga 780

gtactcatct cgcttcataa ggtctcagta tgtagttgcc tttcgagaat gttatgttct 840

ctctcataat gctattctga ttttataaaa aaaaaaaaaa a 881

23

350

›DNA

Eucalyptus grandis

23

ctatagggca cgcgtggtcg acggcccggg ctggtccttt cttacaaaaa gcaaaattct 60

tataattttt tttgatataa taaaaatgat ccataaactt ttgcttaatg tgcaacgtaa 120

accataatat attcaacgtg atgcttaaac tttaatcgag tatgcaatgt agtccataat 180

atattcaata tgatccttca atccaattga agtgtgcaat gtggtcgcta gattttttta 240

tgtattcaac ttagtcttta agctaccaac cttccaataa tttatgtttt agaaataata 300

tcgaacatct tttatattat tcaaggaata aaacgaacat gcatcaaaag 350

24

49

›DNA

Eucalyptus grandis

24

actatagggc acgcgtggtc gacggcccgg gctggtactt tttttttct 49

25

909

›DNA

Eucalyptus grandis

25

cagggtaaag aaaatggaat atttgcttgg ccccccagct ttgaaagttg ctgtaagaac 60

acactcacct tgcatttata cgatggttgt gagcagtgca ggctggtggt gctgcaaatt 120

tatgatgctg atgtgatagg cagatgaatg gcagttgagc taagttaaag ccctcataca 180

tagatcagag caggaggagt agtatatata ggcatcttgg caagtcccta aaagagcggc 240

ttcgtgtatt cccacatatt cctctctcgt tagaacgttc agaaatgggt ggccctttga 300

ctcttgatgc agaggttgag gttaagtctc ctgcagacaa gttctgggtg agcgtgagag 360

actccaccaa actgttccca aagatcttcc cggaccagta caagaatatt gaagtccttg 420

agggagatgg gaaggctcct ggctcagttc gcctcttcac gtatggtgaa ggttctccac 480

ttgttaaagt atcaaaggag aagattgatg gtgtggacga agcagacaag gtcgtgacct 540

acagcgttat agacggtgat ctcctgaagt actacaagaa tttcaatggc agcatcaagg 600

taattcctaa aggagacgga agcttggtga aatggtcgtg tgggtttgag aaggcaagcg 660

atgaaattcc tgatccccac gtaatcaagg acttcgcaat ccagaatttc aaagagcttg 720

atgagttcat cctcaaggca tagatgccgc caatcgtcta tccggatttg cactaaatat 780

caataaaata atgcggagct ggactccgca cttctatatg catctagtat gagagtcccc 840

tgctgtctct gtttgtattc acttgaaggg ttttctatta agctctcttt actgcctccg 900

aaaaaaaaa 909

26

430

›DNA

Eucalyptus grandis

26

tggagcttga gatagatcga ccgagagatc ccagcggaaa tagaagattt cctgatacca 60

tcgatccttc ttctccaatg gctgcgaatt tcgtcattcc gaccaaaatg aaggcttggg 120

tgtaccgtga gcacggaaac gtcgccgacg tattgggatt ggacccggaa ctcaaggtcc 180

ctgaattgca agaaggccaa gtgctggtta aagttcttgc cgcagcgctc aatccagtcg 240

acgccgcgag aatgaagggg gttatcaagc tcccgggctt ttctctaccg gccgtgccag 300

gttacgatct cgccggcgtt gtggtaaagg tgggccgcga agtgaaggag ctcaagatcg 360

gggacgaggt atatggattt atgtttcacg ccaagaaaga cgggacgctg gctgagtacg 420

cagccgtgga 430

27

1253

›DNA

Eucalyptus grandis

27

gcttgagata gatcgactga gagatcctag tggaaataga agatttcctg ataccatcga 60

tccattcttc tccaatggct gcgaatttcg tcattccaac caaaatgaag gcttgggtgt 120

accgtgagca cggagacgtc gccaacgtat tgggattgga cccggaactc aaggtccctg 180

aattgcaaga aggccaagtg ctggttaaag ttcttgccgc ggcgctcaat ccaatcgaca 240

ccgcgagagt gaagggggtt atcaagctcc cgggcttttc tctaccggcc gtgccaggtt 300

acgatctcgc cggcgttgtg gtgaaggtgg gccgcgaagt gaaggagctc aaggtcgggg 360

acgaggtata tggatttatg tttcacgcca agaaagacgg gacgctggct gagtacgcag 420

ccgtggaaga gtcgttcttg gctttgaagc ccaagaagct gcgtttcggg gaggctgctt 480

ctctgccggt ggtcattcag accgcctatg gaggccttga aagagctggc ctctctcatg 540

gcaagtccct cctcgtctta ggtggtgctg gtggcgtcgg cacactcata atacagctag 600

ctaaggaagt ttttggtgca tcaagagtag cagctacatc cagcactggg aagctagagt 660

tgttgaagag cttgggtgct gatctggcca ttgactacac caaagtcaac tttgaagacc 720

tcccagaaaa gtttgatgtt gtctacgata cagttgggga aattgagcgg gcagcgaagg 780

ctgtgaagcc aggagggagc atcgtgacga tcgtaaaaca aaacaagaca ttacccccgc 840

ctgctttctt ttttgcagta acttcgaacc gttcgacctt ggagaagttg aagcccttct 900

tggagagcgg gaaggtgaag ccggtgatcg accccaagag cccgttccca ttttcgcaag 960

ccattgaggc cttctcgtat cttcaaaccc gccgggcaac tggaaaactc gtgattcacc 1020

ccgtcccatg atacacaaac gagaaagaaa taaagcgtcc acatggatct gccttaatca 1080

cgagtcctta attagtagtc gatggtgctt gctgtttgtc tccgtacatt cagcttctct 1140

ttgcatagta gtttctacat agtgcgtgta gagaagcaag tggatgtaca agtaaaataa 1200

ttactttttc tataaacaat attacaaact caaaaaaaaa aaaaaaaaaa aaa 1253

28

99

›DNA

Eucalyptus grandis

28

gatagatcga ccgagagatc ccagcggaaa tagaagattt cctgatacca tcgatccatt 60

cttctccaat ggctgcgaat ttcgtcattc cgaccaaaa 99

29

927

›DNA

Eucalyptus grandis

29

cgacgtcgca tgctcccggc cgccatgcgg ccgcgggaat tcgattacta tagggcacgc 60

gtggtcgacg gcccgggctg gtactctcac taattcttta gttttccaat ttagcccctt 120

ctgtaattgc tcatcttctt taccaaattc tctaatttgg ccggcgaagg gctgacaagg 180

gattggtcat gtcaccctca ccaaaggttg ccgaaggtcc ggtgacctca gctgacggcc 240

acctacacca aatctagctc actagcagcc taagcccttc atcaactcta gtgaaaggtt 300

ttgagtattt tttaataaaa aatatttaaa aaatatatag cgagagctca ttacaaaaaa 360

attttaaaaa aaaatctaaa cattacttga actcaaagtg actttataaa gagtttttac 420

caaaggatct tggtttcatc atttgcacta cacccaaaac ccaatttcta agttaaatca 480

aacccactgt ctaatagaga taaggtaaat gttataaacc aaattccaaa attccgaagc 540

actaaatata tttgctgatc ttataatcgc caattgagag ggtctcattc tccaagggat 600

tgtgacatat tagtaattga tagggtctca tccgtaggac tccgactcag ccgcgccacg 660

tgactggatc gctgaacggc gcggaaccag aggagcgtga ttacctaata ttttctccta 720

ccttggcctt gagattgaat ttcagaaaaa gaaaaagaaa aaggaacaac ttcgccgact 780

gttctataaa atgcatgcgc caccccgacc cccacccacg catcacatcc atccagcctc 840

cacgacagac gcataaacac aacacacgtc ggttagagag agagagagag agagagagag 900

agagagagag atgcttggac agttgtc 927

30

411

›DNA

Eucalyptus grandis

30

actatagggc acgcgtggtc gacggcccgg gctggtctga aactgtcgct cggcgatgca 60

taccaaaggc tgaaggtatc agaatctaat gcagcttatg taaaagcgcg atcaatttat 120

tgaccccgac gaccttgact ccatacttca cgcctcagct ttgtgttgga tggtcttgac 180

ctctctcacc ctaaaaggta gctcaaaaga atgagacttt ccgtcatact tataaaccga 240

ccaccagcct ctttcacaac cgacatggga caacctcaaa tagaattttt aacaacaccc 300

ttgcacgctc tttctatcca ctttattatg ccatcacatg agcgttttcc acgcgtaaat 360

cggctaccac ccactttcac acggcggcga aacgagaaaa aggtcctacc t 411

31

178

›DNA

Eucalyptus grandis

31

cgagtcagca gaaacccagt tacactccgc ccaaacggaa gctaaacctg atgggccata 60

cgatttcttt cactgagcct cttgcttttc ctccggaatc tcacggcacc ggaatgccgg 120

aggaacttgg gaagaaccaa tgatgcctgg tcactgagtg atcgatgaat gcaatagt 178

32

178

›DNA

Eucalyptus grandis

32

gtccaatgtc ctgtcaaagg aggaaagatg actatggccc cggcgccggc ggggactgca 60

tgggatttag tatgttgatt gagtacccgt cgccaccacc ttcaagtaaa tcaggagtca 120

gcagaaaccc agtacactcg ccaaacggag ctaaacctga tggccatacg atttcttt 178

33

178

›DNA

Eucalyptus grandis

33

gcatgggatt tagtatgttg attgagtacc cgtcgccacc accttcaagt aaatcaggag 60

tcagcagaaa cccagtacac tcgccaaacg gagctaaacc tgatggccat acgatttctt 120

tcactgagcc tcttgctttt cctccggaat ctcacggcac cggaatgccg gaggcaac 178

34

1274

›DNA

Eucalyptus grandis

34

ctatagggca cgcgtggtcg acggcccggg ctggtccttt cttacaaaaa gcaaaattct 60

tataattttt tttgatataa taaaaatgat ccataaactt ttgcttaatg tgcaacgtaa 120

accataatat attcaacgtg atgcttaaac tttaatcgag tatgcaatgt agtccataat 180

atattcaata tgatccttca attttaattg aatgtgcaat gtggtcgcta gattttttta 240

tgtattcaac ttagtcttta agctaccaac cttccaataa tttatgttta gaaataatat 300

cgaacatctt ttatattatt caaggaataa aacgaacatg catcaaaagt ttaaatatat 360

caaataaaat aaaattttaa gaattatatt acatattaaa attaaagttc atgattaaat 420

tgaaataaaa taaaaattta aaaatcacgt tgtatgttgt gccgaaacaa aattcagtga 480

cttgtggtgt caattttctt aggtggagct ccacaagcat tgagatggag tgttccttcc 540

gccgaggttt tcattgcgtg gctcaaaacg gtggcgcgtt ttgcacgaca cgagatgcct 600

cgattgccgc atcgtgtagg cgacgcaacg gaaaaacgcg ttgccgtggc gtctatccgg 660

ggtttcgtct ccgatgcggc acgtagccta taaatgcgca cgatctcccg gtctgccaat 720

tcgctatcga ttgcagaaga aaactcaaac cctaggcgct ctctctccgt tcgacctctc 780

gaagttctcc tctcttcgcg tcaagatgca aatctttgtg aaaaccctta ctggcaagac 840

aatcaccctc gaggtggaaa gctcggacac agtcgataat gtgaaagcaa aaatccagga 900

caaggaaggg atccctccgg accagcagag gcttatcttt gctggcaagc agctggaaga 960

tggccgaacc ttggccgatt ataacattca gaaggagtcc accctccact tggtgctccg 1020

tctcagggga ggcatgcaaa tttttgtgaa gactcttact ggcaagacaa tcaccctcga 1080

ggtggaaagc tccgacacag ttgataatgt gaaagcaaaa atccaggaca aggaagggat 1140

ccctccggac cagcagaggc ttatctttgc tggcaagcag ctggaagatg gccgaacctt 1200

ggccgattat aacattcaga aggagtccac cctccacttg gtgctccgtc tcaagggagg 1260

catgcaaatc tttg 1274

35

795

›DNA

Eucalyptus grandis

35

aaaaatacag gctttcgaaa gctagtgcgg tataaataac ctgggaaaag caagccgctt 60

gagctttagt ttcagtcagc catggccact cacgcagctc ttgctccctc aaccctcccc 120

gccaatgcca agttctctag caagagctcc tctcactcct tccccactca atgcttctct 180

aagaggctcg aggtggcgga attctcaggc cttcgtgctg gatcgtgtgt gacttatgcg 240

aagaatgccg gggagggatc cttcttcgat gctgtggctg ctcagctcac tcccaagact 300

tcagcaccag ctccagctaa gggagagact gtcgctaaac tgaaggtggc aatcaatggt 360

ttcggtcgca ttggtcggaa cttccttaga tgctggcacg ggagaaagaa ctcgcccctt 420

gatgtcattg ttgtcaatga cagcggtggt gtcaaaaatg cttcacattt gctgaagtat 480

gattccatgc tggggacttt caaagctgat gtgaaaattg tggacaatga gaccatcagc 540

gtcgatggga agcccgttaa ggtcgtctct aaccgggacc ctctcaagct cccctgggct 600

gagctcggca tcgacattgt cattgaggga actggagtct tcgtggatgg ccctggtgct 660

ggaaaacata ttcaagctgg tgccaagaaa gttatcatca ctgcaccagc aaaaggcgct 720

gatataccca cctacgtcta tggtgtgaat gagacagatt attcgcatga agttgctaac 780

ataatcagca atgct 795

36

1200

›DNA

Eucalyptus grandis

36

aaaatatcca tcgacagcat caccccgctt agagaacggt gtctcggctt ctcacaatgt 60

ctatagccga atgtacaaaa tcggcataat gttctataat atagcggact ttacagatga 120

gcattcaaat acgtacgccg tactcgattc ccattcgatt gttcattcat ccgcatgcaa 180

atttcataga gataatatct gtgcacgtcc ttagattaag aacaaccaaa gagtatctgg 240

tggaagtttg aagcatgacc accgaagtca gatggaacaa acaaggtggg tggtggggat 300

atagtggaca aaggaacgag aggtgaatag gaaaaggaga aggcaagatg cgggagatag 360

gatttacgtg gcgagcggcg attgcacgca tggtccaccc caccctcaac ctcaaacttt 420

cgaaaatgca acgggcatca gggtggcgat gaaggagacg atggagatat tgttgctttc 480

tccccccaaa aaacatcatc caatccatcc ccattcctca tcttcaccac aaggagtctg 540

aagctctcct tcaccggtcc gtcgctttct ctcttatctt cttcttctcc ctcctcttct 600

cgttcttcct tcgaccgttc tctcggtatc gtgaatttat tgcggggtgg ttcgcatgct 660

ataaattcca cagcaacgag ggccccttgc cacaatgtcg acgtctccgg ttagcagctg 720

gtgcgccacc tccttctccc ctgcccattc ctcgctcaag agagccgccg gcctacggcc 780

ctctctctcc gcccgcctcg gcccttcctc ctcctcctcc tccgtctctc ctccgaccct 840

catccgtaac gagcccgttt tcgccgcccc cgcccctgtc atcaacccca cttggacaga 900

agagatgggc aaggactatg acgaggccat tgaggctctc aagaaactcc tcagtgagaa 960

gggggacctg aaagccacag cagccgcaaa agtggagcaa ataactgcgg agttgcaaac 1020

tgcttcccca gacatcaagc catccagctc cgttgacaga atcaaaactg gcttcacctt 1080

cttcaagaag gagaaatacg acaagaaccc tgctttatat ggtgaactgg caaagcagag 1140

tccaaagttc atggtgtttg cttgctcgga ctcgagagtg tgcccatctc atgtgctgga 1200

37

648

›DNA

Eucalyptus grandis

37

cgacggactc ctttcacgat atcgaaacga ggaaacggag gagaagcaga agaaagaaga 60

tgaagaaagg cagatggttg gtgatggatg aaactgtcgg gaagctggga gcttcaggga 120

gttctattta tggggcgaaa caggggaggg gaaaccgaat ttaccaagat gcccttcttg 180

gtgggattgg acatggagct gcacgaccgt cgtcccatca cgaagagtct tgctcttcgg 240

tacacatgca atcgtcggcg aaccgacctt atccgaccgg ttccaagctt gtcctggtaa 300

aaggtttcga accttggaaa aggcttaaga gatgtatcgg tgccttaacc attattccat 360

gttcacataa tatttggccc ggttttcagg tcaattttgg agtagcccgg ttcggttcta 420

gtcccgctcc cgattcaaaa attcattggg aacaaatttt gacactgtct ggtatttttg 480

gtctaagacc ctacccaatt ttagaactgt acacccttgc tttatcccaa aataaaattg 540

tcaattagtc aacttttcac acttgatgat cgattaagta gatggatgac atggtctttt 600

accagcccgg gccgtcgacc acgcgtgccc tatagtgagt cgtattac 648

38

288

›DNA

Eucalyptus grandis

38

gattgtaata cgactcacta tagggcacgc gtggtcgacg gcccgggctg gtatcgtgaa 60

agaagtccgt cgacgacaat ggccgagaag agcaaggtcc tgatcatcgg agagaagagc 120

aaggtcctga tcatcggaga gaagagcaag gtcctgatca tcggagagaa gagcagggtc 180

cttatcatcg gagaatcgaa ttcccgcggc cgccatggcg gccgggagca tgcgacgtcg 240

ggcccaattc gccctatagt gagtcgtatt acaattcact ggccgtcg 288

39

382

›DNA

Eucalyptus grandis

39

acagcaatct catctgatga ttcttcagtt cggagctcag aggatacatc atctatagct 60

gaattgagct gtgcaatctt ctcggcaagc accttcctcg ttttctgaaa atcatcagat 120

tttaaggtga atccatattt cgcagatggc catgttactg ctacactctc ttcacagcat 180

acatgaagga ggtcacatag caagcataca taggacctca tatacaaata tgacagcaga 240

ccagcccggg ccgtcgacca cgcgtgccct atagtagtag tggggaagga gtgagaggag 300

ctcttgatga ggaatgtcgg cttttcttcc atcagttgat gttccgggtt cctagtcatt 360

atgccgatgg tggccactcc ag 382

40

986

›DNA

Eucalyptus grandis

40

aaatacaaac tggtttaata ttcaactcag ataattacat gacaccacct aaataatgga 60

aagtcaagca aatagacata ttatccccac acataatcaa ctatattcat gactggagag 120

gtgctagatg gtatagagtc cctagttatt atttattttt ttgggcccga gaagatcctg 180

atggatctat gctgtttgat actttcagat ttgttttgtc tacagctcaa ataaattagt 240

gcttgggttt tgatatatta tctaatctga tacaagtctt tgtcctggcc aatttttgca 300

gagtttcctg caaaacagtg cactaaagct tccagaggac ctcatgccat gcccaagggc 360

accacctatg atggaacgga gaatcaaacc acagactgaa caggcgttga aatgccccag 420

atgtgattct acaaacacaa aattctgtta ctataacaac tacaatcttt cacaacctcg 480

ccatttctgc aagacctgca ggcgatactg gaccaaagga ggtgccttac gtaacgttcc 540

tgttggtggg ggttgcagaa agaataaacg agccaagcga gcagtagacc atcctgtctc 600

tgctcagaat gaagcatcca cctctgcagc cccaggcaac gaagtacctg accggtctcc 660

ctttgagcca ccatcttcaa aatccattta ctatggggga gaaaacatga acttaaccgg 720

tctccccttt agcagaattc agcaggaccg agctgcattg gcccactgca actcttcttc 780

ctttctagga atgtcatgtg gcacccaatc ggcctctctg gaaccacatc tttcggcttt 840

aaatacattt aattcattca agtctaacaa tcctggtctg gattttccta gcttaagcac 900

agaccagaat tcactgtttg agaccagcca gccacaactg tcaagagcaa tggcatctgc 960

ccttttttct atgccaatgg ctcctg 986

41

313

›DNA

Pinus radiata

41

aaaggaaaat tcaaagatct ttagccaatt tttgttgttg tgaccttgaa tttctaaaaa 60

atttaatgga ttcgttttct aaattcctga ttcgtcaaag gctgaagggc acgatagtaa 120

tagaaaatgg acggcagttt atcctttcat ggctggacac acagaatttg tggagggact 180

ctccattctg gtttatccgc cgttagttct ctctgtactc cacccttagt tctctttgta 240

ctcgagacct ttaatgatta gccctgctta tgctgtcatt actgaactca cttccagagc 300

cccaaaaatc tct 313

42

713

›DNA

Pinus radiata

42

taattcacaa gtagaaaatg agatttttgc aattttgtaa ctaacatttc ccggtctcct 60

ctgtatgttt tcacccctta atgtaattga aatttgcacc cgggttagat tcaaagcgga 120

gaataacatc ggggccttgt tctagacaga gatttttcac aaataacagg ttcgaaggta 180

tgtgtagaca tctgggtagt tgtagaataa agacggagcc cattaggtga tccaatcgaa 240

gagctcagat gggaaaacag ataaaaatta tcgggtggac cttccttcac atgttaatta 300

tatatcaagt gtcgccaatc cttatgtgaa acatttagta aagcttcgcc agagcacttc 360

ttataggcat tctgtgggct ctgttgttgt ggttggaagt actcctttaa gggaggtatc 420

tgaatatttg caacagaagt cagttaaaca agtggttgac tgtctgtttg tacaagatgt 480

tactggcata cctgtgggct tgatagagac ttccaggcgc attgtgcatg taaatcattt 540

ggtgatgcag aagctagccg gagtagagtc tatagagccc actgaagcaa ttggtgtaat 600

caagcttcct agcagcttct acaacttgga atctcttgaa attcactcta gttcccagat 660

atggtgctcg tcgccacatc gtctgcttgt acttgatggc attcaggatc ctg 713

43

28

›DNA

Pinus radiata

43

ccacctcaca tcaataaatt ttatacga 28

44

35

›DNA

Pinus radiata

44

gctgtttcat tggggtcata gctacgtggt gctga 35

45

1729

›DNA

Pinus radiata

45

cttattgaca tataaaagca aagttggatc catctgttat tttgggtccc ctccagaagc 60

cttactaaat gcggacaaaa aatccacgta aagaacttct gaatttaccg tcatctgggc 120

tctgtaatta cgaatttagg gtttcctctg tcaatatctg gtagtgacaa acaaggttta 180

atggcagcct tagcaacaac tgaagtttgt gatacatatc cacgccttgt ggagaatggt 240

gagcttcgtg tcttgcaacc aattttccag atatatggtc gacgtcgagc tttctctgga 300

cctatagtta cactgaaggt ctttgaggac aatgtccttt tgcgggaatt ccttgaggag 360

agaggtaatg gaagagtttt ggtagttgat ggaggaggaa gccttagatg tgccatactg 420

gggggcaatg tagttgtatc tgcccaaaac aatggttggt ctggaataat tgtcactggc 480

tgcataaggg acgttgatga aataaacaga tgtgacattg gtataagagc actgacatct 540

aacccactga aggccaacaa gaagggtgtg ggtgaaaaac atgcgcctat ttacattgct 600

ggtacccgca ttcttccggg ggaatggtgt tatgctgaca gtgatggtat tcttgtttca 660

cagcaagagt tatcactgtg agataataaa attcataagt ttcagattgt gactttcatg 720

tcctgtggaa catatatttg actcgagtta gattctaata ggattaattg atagattctg 780

aaaattgagg aatatctctg gtcatgaaaa tcttcttctc atgtgatctt ttatgctcag 840

ctttgagtac aggatgataa gaagtttgtg catgtttgtc taaaggttta gcaagtatta 900

tcggaccatc ataagagata gattatggaa ctcagggact tgctattttt aatccaaaat 960

aacatttatt ctttgtgttt ttgccaaatt aacttttatt tcccttggca ccactagtga 1020

tttgcaatat ccagttgctg agaacataga agtgggcaac ggtgagagtt gcaacagtat 1080

ctagcataga tttaacaagt attgttggat cattataaga aaataaacta cagaaccaag 1140

ggaatctagt tgacaacata gttaaagtag gcatggtgct actgtatcga tacatcttca 1200

taaacagaaa aatatgaaca agctctaatg atgggagaaa ctccagcttg gtgttttgat 1260

taagcatcca tattcacacc taaaaggtta caagttccaa aataaaaatt ccaatgaatt 1320

tagccaatct aatcagacct tataagaaat acactaggca tctggggatc aaaatccagt 1380

agtttagaaa gtagttgtaa ataacccaga gacaaaaatc tcaatgatag cttgcttggg 1440

tcataggttt gataataatt gaaaacatag ttgaaaggag aatcctagca atggctagct 1500

tgaataatag atgtacagca aaattacagt agttgagaac aaagatggaa ggataatccc 1560

aacgatagct agcttggaca gtaggatgat tacatcaaaa tcatagcagt tgagaacata 1620

gttggaagga gaatccttat gatggctacg ttggataata ggcgtgatta tcgtaggtag 1680

attagagcac aagatcaaac taatagctgg cgcagctatc gactatttt 1729

46

1038

›DNA

Pinus radiata

46

tgattactat agggcacgcg tggtcgacgg cccgggctgg taaatgagaa catgataagc 60

tgtgtaaatt catgctagtc accataactt ttctcattgc ttttcatcca cactgttgat 120

tcattcatta tataagatca gattcgtatg atatacaggc aaccatagaa acaaccagca 180

aagttactag caggaaatcc aactaggtat catgaagact accaacgcag gctcgataat 240

gttggtgctc attatttttg ggtgctgttt cattggggtc atagctacat cttttgattt 300

ctattacttc gttcaacagt ggcctggttc atactgcgat actcgtagag gatgctgtta 360

ccctcgcacg ggaaggcctg cttccgaatt ttccattcat ggcctctggc ccaactacaa 420

gaccggtaaa tggccacagt tctgtggttc ctccgaagaa ttcgactact caaagatctc 480

agatctggag gaggagctga acaggtattg gggttcgtta agctgtccaa gcagcgatgg 540

acaggaattt tggggacacg agtgggagaa acatggcact tgctctctca atcttgatga 600

gcattcatac tttgagaagg ctctctcctt gagacaaaat atagacattc ttggggctct 660

taaaactgca ggtattaaac ccgatggaag ccaatacagt ttgagcgata tcaaggaagc 720

cattaaacaa aacactgggc agctcccagg aatcgattgc aacacgagcg cagagggaga 780

gcatcaacta tatcaggtgt atgtgtgtgt tgataaatcc gatgcttcca ctgttattga 840

atgccccatt tatccacaca gcaattgccc atccatggtt gtgtttcctc cttttgggga 900

ggatcaggag gaccgagatg gttacacaga aggaatgtac gagctgtaga tctggacaaa 960

cagcatttct tctctccgca tttgattttt atcaatgaaa tttccgattc caacattttg 1020

taaaaaaaaa aaaaaaaa 1038

47

91

›DNA

Pinus radiata

47

aattttccat tcatgcctct gcccaactac aagaccggta aatggccaca gttctgtggt 60

tcctccgaag aattcgatat caagcttatc g 91

48

91

›DNA

Pinus radiata

48

gcttttcatc cacactggtg cctcattcat tatataagat cagattcgtg tgatatacag 60

gcaaccatag aaacaaccgg caaagttact a 91

49

809

›DNA

Pinus radiata

49

tgatatatat aacttctagc agaatgacac gcgacttgta tatcttttca ttttttaacc 60

catgaaaacc gattagggta ttgcaaatta gggcattgcc attcaaataa ttctcagatg 120

aaagattctc tctaacaatt acaaatgatt atttttttcc atgagtgttg catgttcgaa 180

cggtctgccc agtctgtgag agagcataga gaaccctccc tgcccaattt gttagagcat 240

agagaaccct actgcatgag tagtaagaaa aatattcggt ctcaattcgg caaagaccac 300

ctcgaatgga tgacttcaac gacaatctca tgatagtgtt ctgatcagca ccagttcacc 360

tatatatttt atctagggtt tagtttgcat gtatcaatcc tctggtgcac taggtaattc 420

tttcccagta tcatatatcc ttaatactgt tttgtctttt aatccatggc taccatcaga 480

acaagctcaa agcagaataa gggagcatca gccatcctct tgcttatcgc gattgcaggg 540

ttagtaaatg cgtgcaacgc tgtgggtatt gagccaatgt gcgacactgt ggtgtcgagt 600

cttctgaggc ttctgccatg caggacggct gttgatccct caattgccgc cattccactt 660

ccaagctgct gcaacgcggt tgagtcagct gggcttcaat gcctctgtct cgtcgttaac 720

ggccctcctt ttccaggggt cgaccgcggc ctcgcaatgc agctgcctgc caaatgccat 780

ctcacccttc ctccctgtaa cagttagtt 809

50

428

›DNA

Eucalyptus grandis

50

tttcttgtga ctattcattt tcctcctgat tatccattca agcccccgaa ggttgcattt 60

aggactaaag ttttccaccc aaatataaat aacaatggaa gtatctgcct tgacatcttg 120

aaggaacagt ggagtcctgc tttgacaatc tccaaggttt tgctctcaat ttgctctttg 180

ttgacggatc caaacccaga tgatcctctt gtaccagaga ttgctcatat gtacaagact 240

gataggggca aatatgagtc cactgcacgg agttggactc agaaatatgc aatgggttaa 300

ctttaaaaac tatatatcag tgatggaact ttatccctaa gttggaatct cttcgaatca 360

atgacttgtt tgcttgtaag aaatgtttcc ttaagataag tggctttcct caaaacttga 420

ttgaagtg 428

51

525

›DNA

Pinus radiata

51

cccttctttg ccttcaacta atcctgctca tcctctcctg cccccattcc caaagatggc 60

tgcacccaga tcatccgcta aattgggtgc acttttggca atactgctca tagttgcggc 120

agcgcaggct caagattgct caaatgccat ggacaaattg gctccatgca cttcagcagt 180

gggactgtct agcaatggag tgaagccctc atctgagtgc tgtgatgccc tcaaaggaac 240

cagtactggc tgcgtctgca agtctgtgag agcagtgata tcacttcctg ctaagtgcaa 300

tctcccagcc ataacctgct ctggatctcg ctgaaggctc tctgttatgg cgattctcag 360

atcgtggatc tctttaagat tttcagcaag caagtgatag aataaattct cagattttga 420

gatatctata tagcgatttt cagtatcaga ttgtctatag tactcatata tttaagtgat 480

tgaatagcat tctccgattc cgagttggaa acacagacac aatga 525

52

1126

›DNA

Pinus radiata

52

actagtgatt actatagggc acgcgtggtc gacggcccgg gctggtaaat acccaactta 60

atttaattgt tattgagcca gagagatgcg tagtcgctca tgtcacttgt gtttaccaaa 120

aagacataca taaacacctg cacctaaaag ttataatgat aacatgcata caaccctaca 180

acgtacgtag tcacatgcgg ctagaactta aacccctacc acaaacatag ccacctgcac 240

ccagaagtta taataataac atacatagaa cccttacaat aaaaaaagtt atctccaatg 300

attattaatc tactgcaggc cagccatact cagcttgaac gtgaaaattc gcattgtaag 360

catggcgcca cattaaaata acctcggcaa tattttcatg tccaagtggc cggccagcca 420

cgctcctcgc actctgagaa tactctattc atccacttgt ctctgccccg caactcatat 480

aaatgtggcc aacccaagca ccatatccat gttcattaat cccctctttg ccttcaacta 540

atcctgctca tcccctcttg ccccaattcc caaagatggc tgcacccaga tcatccgcta 600

aatcggctgc acttttcgca atactgctca tagttgcggc agtacaggct gaagattgct 660

caaatgccat ggacaaattg gctccatgca cttcagcagt gggactgtct agcaatggag 720

tgaagccctc atctgagtgc tgtgatgccc tcaaaggaac cagtactggc tgcgtctgca 780

aatctgtgag agcagtgata tcacttcctg ctaagtgcaa tctcccagcc ttaacctgct 840

ctggatctcg ctgaaggctc tctgttatgg cgattctcag atcgtggatc tctttaagat 900

tttcaggaag caagtgatag aataaattct cagatgttga gatatctata tagcgatttt 960

cagtatcaga ttgtctacag taccaatata tttaagtgat tgaatggaat tctcggattc 1020

tgagatagaa atataggcac agaatgtggc cggaggaatg ttcgaattcg agaatgataa 1080

taaataataa atgattgatt tctctctgca aaaaaaaaaa aaaaaa 1126

53

454

›DNA

Pinus radiata

53

atcctgctca tcctctcctg cccccattcc caaagatggc tgcacccaga tcatccgcta 60

aattgggtgc acttttggca atactgctca tagttgcggc agcgcaggct caagattgct 120

caaatgccat ggacaaattg gctccatgca cttcagcagt gggactgtct agcaatggag 180

tgaagccctc atctgagtgc tgtgatgccc tcaaaggaac cagtactggc tgcgtctgca 240

agtctgtgag agcagtgata tcacttcctg ctaagtgcaa tctcccagcc ataacctgct 300

ctggatctcg ctgaaggctc tctgttatgg cgattctcag atcgtggata tctttaagat 360

tttcagcaag tgatagaata aattctcaga ttttgagata tctatatagc gattttcagt 420

atcagattgt ctatagtact catatattta agtg 454

54

335

›DNA

Pinus radiata

54

agaagcacct gttaaaaagg aggcctgctc tttgttcatg agcttataga taagccctag 60

tctgcaagga ttattgccct gtagttattt ggaagtagat cattttcaca ggcccagatg 120

cattatattc taatgcagtt gtttgttaat tgaagtgcaa atagttccaa aatgtttaca 180

tgaatcaata gtgaacaaat ccctctgttt tatatcatat tgatggatta ttcgattttt 240

tggtgacgtg gcgcgaaact gcttttcgaa ctcatggaaa tagtaattgt tataatccat 300

aggcatgaga ttcttgttaa tcgtgcacaa ggttt 335

55

336

›DNA

Pinus radiata

55

aaaccttgtg cacgattaac aagaatctca tgcctatgga ttataacaat tactatttcc 60

atgagttcga aaagcagttt cgcgccacgt caccaaaaaa tcgaataatc catcaatatg 120

atataaaaca gagggatttg ttcactattg attcatgtaa acattttgga actatttgca 180

cttcaattaa caaacaactg cattagaata taatgcatct ggtgcctgtg aaaatgatct 240

acttccaaat aactacaggg caataatcct tgcagactag ggcttatcta taagctcatg 300

aacaaagagc aggcctcctt tttaacaggt gcttct 336

56

532

›DNA

Pinus radiata

56

cgttcgttcc cttccctttc cattgttgcg tttaagccct ccaattttct tttggcgtcc 60

cgtttttggg gctcccttga agatctcctc ttcatttcgg gatttcctgc cttcgccgcg 120

ccatttgaag ttctttttct gagagaagaa tttagacatg gctgatcgca tgttgactcg 180

aagccacagc cttcgcgagc gtttggacga gaccctctct gctcaccgca acgatattgt 240

ggccttcctt tcaagggttg aagccaaggg caaaggcatc ttgcagcgcc accagatttt 300

tgctgagttt gaggccatct ctgaggagag cagagcaaag cttcttgatg gggcctttgg 360

tgaagtcctc aaatccactc aggaagcgat tgtgtcgcct ccatgggttg ctcttgctgt 420

tcgtccaagg ccgggcgtgt gggagcacat ccgtgtgaac gtccatgcgc ttgttcttga 480

gcaattggag gttgctgagt atctgcactt caaagaagag cttgctgatg ga 532

57

3103

›DNA

Eucalyptus grandis

57

gggtgaaaac aattaatgag atcatttgaa ttaaggaaag tggaaaggcg gttttctgat 60

tggtacactg aaacaacagg aaggtggtgg aggccgcaat gatggaattt atccacttta 120

atcattttat gaaatcgata cactaacctt tgtttctcct aaacccaaag gcattaatcc 180

ctgtcctcct cactcgatct cgaaggccag aagggggagg ccgagcctct tgcttttttt 240

cgtgtataaa agggcctccc ccattcctca tttttcacca tcctccgttc gttcgttccc 300

ttccctttcc attgttgcgt ttaagccctc caattttctt ttggcgtccc gtttttgggg 360

ctcccttgaa gatctcctct tcatttcggg atttcctgcc ttcgccgcgc catttgaagt 420

tctttttctg agagaagaat ttagacatgg ctgatcgcat gttgactcga agccacagcc 480

ttcgcgagcg tttggacgag accctctctg ctcaccgcaa cgatattgtg gccttccttt 540

caagggttga agccaagggc aaaggcatct tgcagcgcca ccagattttt gctgagtttg 600

aggccatctc tgaggagagc agagcaaagc ttcttgatgg ggcctttggt gaagtcctca 660

aatccactca ggaagcgatt gtgtcgcctc catgggttgc tcttgctgtt cgtccaaggc 720

cgggcgtgtg ggagcacatc cgtgtgaacg tccatgcgct tgttcttgag caattggagg 780

ttgctgagta tctgcacttc aaagaagagc ttgctgatgg aagcttgaat ggtaactttg 840

tgcttgagct tgactttgag ccattcactg cctcttttcc gcgcccgact ctttccaagt 900

ctattggcaa tggcgtcgag tttctcaatc gccatctctc cgctaagctc ttccatgaca 960

aggaaagctt gcaccctctg cttgaattcc tccaagtcca ctgctacaag gggaagaaca 1020

tgatggtgaa tgccagaatc cagaatgtgt tctccctcca acatgtcctg aggaaggcgg 1080

aggagtatct gacctcgctc aaacccgaga ccccgtactc ccagttcgag cacaagttcc 1140

aggagatcgg gctcgagcgg gggtggggtg acacggctga gcgcgtcctc gagatgatcc 1200

agctcctgtt ggatctcctt gaggctcccg acccgtgcac tctcgagaag ttcttggata 1260

gggttcccat ggtcttcaac gtcgtgatca tgtctcccca cggatacttt gctcaggacg 1320

acgtccttgg ttatccggat accggtggcc aggttgttta catcctggat caagttcgtg 1380

ccctagagga agaaatgctt caccgcatta agcaacaagg actggatatt actcctcgga 1440

ttctcattat cactcggctt cttccagacg cggttggaac cacctgtggc cagcgccttg 1500

agaaagtttt tgggaccgag tactcccaca ttcttcgcgt ccccttcaga aatgagaagg 1560

gagtcgtccg caagtggatt tcccggttcg aggtgtggcc ctatttggaa agatacactg 1620

aggatgtcgc gagcgaactt gctggagagt tgcagggcaa gcctgatctg atcatcggaa 1680

actacagtga tggaaacatt gttgcttcct tgttagcaca taaattaggt gttacacagt 1740

gtacaatagc ccatgccctc gagaagacga agtacccaga gtcagacata tactggaaga 1800

aatttgagga aaagtaccac ttctcttgcc agttcactgc tgatctcatc gccatgaacc 1860

acaccgactt cattatcacc agcaccttcc aagaaattgc tggaagcaag gatacagtgg 1920

ggcagtatga gagtcacatg aacttcactc ttcctggact ctaccgagtt gtccacggga 1980

tcgacgtctt cgacccgaag ttcaacattg tttcaccagg tgctgacatg agcatctact 2040

ttgcttacac cgaacaggag cggcggttga aatccttcca ccctgagatc gaggaactcc 2100

tcttcagcga tgttgagaac aaggaacact tgtgtgtgtt gaaagataag aagaagccta 2160

ttattttcac catggcaagg ctggaccgtg tcaagaactt gacagggctt gttgagtggt 2220

atggcaagaa ctccaagttg agggaactcg ccaacttggt cgtggttgga ggtgacagga 2280

ggaaggattc gaaggacttg gaagagcagt ctgagatgaa gaaaatgtac gacctcatcg 2340

aaaagtacaa gctgaatggc cagttcaggt ggatttcctc ccagatgaac cgggtgagga 2400

atggagagct ctaccgctac atctgtgaca cgaagggagt cttcgttcaa ccggctatct 2460

atgaagcttt cgggttgacc gtggttgagg ccatgacttg tggattgcca acctttgcca 2520

cttgcaatgg tggaccagct gagatcattg tgcatggcaa atcgggctac cacattgatc 2580

cttaccatgg tgaccaggcg gccgagcttc ttgtagactt cttcaacaag tgcaagattg 2640

accagtccca ctgggacgag atctcaaagg gtgccatgca gagaattgaa gagaagtata 2700

catggaaaat atattctgag aggctgttga acctgactgc cgtgtatggc ttctggaagc 2760

atgtgactaa ccttgatcgg cgcgagagtc gccggtacct tgaaatgttc tatgccctca 2820

agtatcgccc actggcacag tctgttcctc cggctgtcga gtaaacaaag agacagattg 2880

ttaccagaag acggaagcat tggacttttg aagttttcaa ggaataaaca ttggaaattg 2940

tttgaatttg ggattgccaa gagcgatctt tttcgtttcc tttttttggt cctttttctc 3000

ttctttgttt ccattccgcg aatgtttgca ttttggggtt tgtacccatc aattcagtaa 3060

atggttcatt ttcttttcaa aaaaaaaaaa aaaaaaaaaa aaa 3103

58

326

›DNA

Eucalyptus grandis

58

ctcgaaaccg agacgctgac tgtgggttga gctctaacca atgggagtga tgtctctctt 60

acgtgcctgc cgtgggcccc agtgacgggc cccaaaagtg taaacgaagg aagctcccgg 120

ggatctgatt ggccgcgacg tccgcctctg acgtggcacc accgacgatt tttttttaat 180

atcttggtca agtcctaatt taactatggg gtccagatta gaagcttatc cactatggat 240

taaattaaat caaatgggaa ttaaattaaa ttaaaatcat cgtgcggagg tgcacgagat 300

gcacgagatc cgacggcgca gagcag 326

59

311

›DNA

Eucalyptus grandis

59

attactatag ggcacgcgtg gtcgacggcc cgggctggta ctctcactaa ttctttagtt 60

ttccaattta gccccttctg taattgctca tcttctttac caaattctct aatttggccg 120

gcgaagggct gacaagggat tggtcatgtc accctcacca aaggttgccg aaggtccggt 180

gacctcagct gacggccacc tacaccaaat ctagctcact agcagcctaa gcccttcatc 240

aactctagtg aaaggttttg agtatttttt aataaaaaat atttaaaaaa tatatagcga 300

gagctcatta c 311

60

2096

›DNA

Eucalyptus grandis

60

gattactata gggcacgcgt ggtcgacggc ccgggctggt ctgagccatt taattcgaga 60

gcacatcgcc caaaattatt cttcttgctg ccataactgt cgaattttct cttttaggta 120

agtaaccaat gatgcatcat gttgacaaaa aggctgatta gtatgatctt ggagttgttg 180

gtgcaaattt gcaagctgac gatggcccct cagggaaatt aaggcgccaa cccagattgc 240

aaagagcaca aagagcacga tccaaccttt ccttaacaag atcatcacca gatcggccag 300

taagggtaat attaatttaa caaatagctc ttgtaccggg aactccgtat ttctctcact 360

tccataaacc cctgattaat ttggtgggaa agcgacagcc aacccacaaa aggtcagatg 420

tcatcccacg agagagagag agagagagag agagagagag agagttttct ctctatattc 480

tggttcaccg gttggagtca atggcatgcg tgacgaatgt acatattggt gtagggtcca 540

atattttgcg ggagggttgg tgaaccgcaa agttcctata tatcgaacct ccaccaccat 600

acctcacttc aatccccacc atttatccgt tttatttcct ctgctttcct ttgctcgagt 660

ctcgcggaag agagagaaga gaggagagga gagaatgggt tcgaccggat ccgagaccca 720

gatgaccccg acccaagtct cggacgagga ggcgaacctc ttcgccatgc agctggcgag 780

cgcctccgtg ctccccatgg tcctcaaggc cgccatcgag ctcgacctcc tcgagatcat 840

ggccaaggcc gggccgggcg cgttcctctc cccgggggaa gtcgcggccc agctcccgac 900

ccagaacccc gaggcacccg tcatgctcga ccggatcttc cggctgctgg ccagctactc 960

cgtgctcacg tgcaccctcc gcgacctccc cgatggcaag gtcgagcggc tctacggctt 1020

agcgccggtg tgcaagttct tggtcaagaa cgaggacggg gtctccatcg ccgcactcaa 1080

cttgatgaac caggacaaaa tcctcatgga aagctggtat tacctgaaag atgcggtcct 1140

tgaaggcgga atcccattca acaaggcgta cgggatgacc gcgttcgagt atcatggcac 1200

cgacccgcga ttcaacaaga tctttaaccg gggaatgtct gatcactcca ccattactat 1260

gaagaagata ctggaaacat acaagggctt cgagggcctc gagaccgtgg tcgatgtcgg 1320

aggcggcact ggggccgtgc tcagcatgat cgttgccaaa tacccatcaa tgaaagggat 1380

caacttcgac cgccccaacg gattgaagac gccccacccc ttcctggtgt caagcacgtc 1440

ggaggcgaca tgttcgtcag cgttccaaag ggagatgcca ttttcatgaa gtggatatgc 1500

catgactgga gtgacgacca ttgcgcgaag ttcctcaaga actgctacga tgcgcttccc 1560

aacaatggaa aggtgatcgt tgcagagtgc gtactccctg tgtacccaga cacgagccta 1620

gcgaccaaga atgtgatcca catcgactgc atcatgttgg cccacaaccc aggcgggaaa 1680

gagaggacac agaaggagtt cgaggcattg gccaaagggg ccggatttca gggcttccaa 1740

gtcatgtgct gcgctttcgg cactcacgtc atggagttcc tgaagaccgc ttgatctgct 1800

cctctgtggt gatgttcatg gttcttggat ttgaaaggtc gtgaaggagc ccttttctca 1860

cagttggctt cggcatacca agttcttctc ataaaaggaa acaataagaa gcgactgtat 1920

gatggcgcaa gtggaagtta caagatttgt tgttttatgt ctataaagtt ttgagtcttc 1980

tgcatactga tttcacagaa tgtgtaacga aacggcgtat atggatgtgc ctgaatgatg 2040

gaaattgtga tattctgtct tctttttcag taaatcactt cgaacaaaaa aaaaaa 2096

61

522

›DNA

Eucalyptus grandis

61

ctaaaacgct aatcctgccc tgcccttccc ttctgctgct gctgctcgtc acctctctct 60

ccctctcgcg gccagctgcg agatctgccg agtttaagcc tcgtacatca aaatgggtaa 120

ggagaagatt cacatcagca ttgtggtcat tggccatgtc gattctggga agtcaaccac 180

aactggccac ttgatataca agctcggagg aatcgacaag cgtgtgattg agagattcga 240

gaaggaagct gctgagatga acaagagatc gttcaagtat gcttgggtgc ttgacaagct 300

caaggccgag cgcgagcgcg gtattaccat tgatattgcc ttgtggaagt tcgagaccac 360

caagtactac tgcactgtca ttgatgctcc tggacatcgt gactttatta agaatatgat 420

tactggaacc tcccaggccg actgtgctgt ccttatcatt gattccacca ctggtggttt 480

cgaagctggt atttccaagg atggccagac ccgtgaacat gc 522

62

420

›DNA

Eucalyptus grandis

62

tttgatacgc taacaaacaa aacatgtgaa aagcttaatt atggcaatta tcataaatag 60

aaaaaaatta gaaaaaaaga gaggaaatgg gccattattt aaattgcaat cgaaagattg 120

agggcaattc tgtttctcta gtgtaaataa gggtgtattt aataattgag ggatggaaat 180

agcatggtca ctcggtaatt atcaaggaaa gcaagaataa aaatggaaaa aaaaaaaaaa 240

aaagcttgaa gaggccaatg tcgaaattat gagcgcgaga tgaggacact cctgggaaac 300

gaaaaatggc attcgcgggg ggtgctatat aaagcctcgt gtaagggtgc gttcctcact 360

ctcaaaccct aatcctgccc ttcccttctg ctgctgctgc tcgtcacctc tctcctccct 420

63

65

›PRT

Eucalyptus grandis

63

Met Asp Asn Ser Lys Met Gly Phe Asn Ala Gly Gln Ala Lys Gly Gln

1 5 10 15

Thr Gln Glu Lys Ser Asn Gln Met Met Asp Lys Ala Ser Asn Thr Ala

20 25 30

Gln Ser Ala Arg Asp Ser Met Gln Glu Thr Gly Gln Gln Met Lys Ala

35 40 45

Lys Ala Gln Gly Ala Ala Asp Ala Val Lys Asn Ala Thr Gly Met Asn

50 55 60

Lys

65

64

152

›PRT

Eucalyptus grandis

64

Met Gly Gly Pro Leu Thr Leu Asp Ala Glu Val Glu Val Lys Ser Pro

1 5 10 15

Ala Asp Lys Phe Trp Val Ser Val Arg Asp Ser Thr Lys Leu Phe Pro

20 25 30

Lys Ile Phe Pro Asp Gln Tyr Lys Asn Ile Glu Val Leu Glu Gly Asp

35 40 45

Gly Lys Ala Pro Gly Ser Val Arg Leu Phe Thr Tyr Gly Glu Gly Ser

50 55 60

Pro Leu Val Lys Val Ser Lys Glu Lys Ile Asp Gly Val Asp Glu Ala

65 70 75 80

Asp Lys Val Val Thr Tyr Ser Val Ile Asp Gly Asp Leu Leu Lys Tyr

85 90 95

Tyr Lys Asn Phe Asn Gly Ser Ile Lys Val Ile Pro Lys Gly Asp Gly

100 105 110

Ser Leu Val Lys Trp Ser Cys Gly Phe Glu Lys Ala Ser Asp Glu Ile

115 120 125

Pro Asp Pro His Val Ile Lys Asp Phe Ala Ile Gln Asn Phe Lys Glu

130 135 140

Leu Asp Glu Phe Ile Leu Lys Ala

145 150

65

117

›PRT

Eucalyptus grandis

65

Met Ala Ala Asn Phe Val Ile Pro Thr Lys Met Lys Ala Trp Val Tyr

1 5 10 15

Arg Glu His Gly Asn Val Ala Asp Val Leu Gly Leu Asp Pro Glu Leu

20 25 30

Lys Val Pro Glu Leu Gln Glu Gly Gln Val Leu Val Lys Val Leu Ala

35 40 45

Ala Ala Leu Asn Pro Val Asp Ala Ala Arg Met Lys Gly Val Ile Lys

50 55 60

Leu Pro Gly Phe Ser Leu Pro Ala Val Pro Gly Tyr Asp Leu Ala Gly

65 70 75 80

Val Val Val Lys Val Gly Arg Glu Val Lys Glu Leu Lys Ile Gly Asp

85 90 95

Glu Val Tyr Gly Phe Met Phe His Ala Lys Lys Asp Gly Thr Leu Ala

100 105 110

Glu Tyr Ala Ala Val

115

66

318

›PRT

Eucalyptus grandis

66

Met Ala Ala Asn Phe Val Ile Pro Thr Lys Met Lys Ala Trp Val Tyr

1 5 10 15

Arg Glu His Gly Asp Val Ala Asn Val Leu Gly Leu Asp Pro Glu Leu

20 25 30

Lys Val Pro Glu Leu Gln Glu Gly Gln Val Leu Val Lys Val Leu Ala

35 40 45

Ala Ala Leu Asn Pro Ile Asp Thr Ala Arg Val Lys Gly Val Ile Lys

50 55 60

Leu Pro Gly Phe Ser Leu Pro Ala Val Pro Gly Tyr Asp Leu Ala Gly

65 70 75 80

Val Val Val Lys Val Gly Arg Glu Val Lys Glu Leu Lys Val Gly Asp

85 90 95

Glu Val Tyr Gly Phe Met Phe His Ala Lys Lys Asp Gly Thr Leu Ala

100 105 110

Glu Tyr Ala Ala Val Glu Glu Ser Phe Leu Ala Leu Lys Pro Lys Lys

115 120 125

Leu Arg Phe Gly Glu Ala Ala Ser Leu Pro Val Val Ile Gln Thr Ala

130 135 140

Tyr Gly Gly Leu Glu Arg Ala Gly Leu Ser His Gly Lys Ser Leu Leu

145 150 155 160

Val Leu Gly Gly Ala Gly Gly Val Gly Thr Leu Ile Ile Gln Leu Ala

165 170 175

Lys Glu Val Phe Gly Ala Ser Arg Val Ala Ala Thr Ser Ser Thr Gly

180 185 190

Lys Leu Glu Leu Leu Lys Ser Leu Gly Ala Asp Leu Ala Ile Asp Tyr

195 200 205

Thr Lys Val Asn Phe Glu Asp Leu Pro Glu Lys Phe Asp Val Val Tyr

210 215 220

Asp Thr Val Gly Glu Ile Glu Arg Ala Ala Lys Ala Val Lys Pro Gly

225 230 235 240

Gly Ser Ile Val Thr Ile Val Lys Gln Asn Lys Thr Leu Pro Pro Pro

245 250 255

Ala Phe Phe Phe Ala Val Thr Ser Asn Arg Ser Thr Leu Glu Lys Leu

260 265 270

Lys Pro Phe Leu Glu Ser Gly Lys Val Lys Pro Val Ile Asp Pro Lys

275 280 285

Ser Pro Phe Pro Phe Ser Gln Ala Ile Glu Ala Phe Ser Tyr Leu Gln

290 295 300

Thr Arg Arg Ala Thr Gly Lys Leu Val Ile His Pro Val Pro

305 310 315

67

156

›PRT

Eucalyptus grandis

67

Met Gln Ile Phe Val Lys Thr Leu Thr Gly Lys Thr Ile Thr Leu Glu

1 5 10 15

Val Glu Ser Ser Asp Thr Val Asp Asn Val Lys Ala Lys Ile Gln Asp

20 25 30

Lys Glu Gly Ile Pro Pro Asp Gln Gln Arg Leu Ile Phe Ala Gly Lys

35 40 45

Gln Leu Glu Asp Gly Arg Thr Leu Ala Asp Tyr Asn Ile Gln Lys Glu

50 55 60

Ser Thr Leu His Leu Val Leu Arg Leu Arg Gly Gly Met Gln Ile Phe

65 70 75 80

Val Lys Thr Leu Thr Gly Lys Thr Ile Thr Leu Glu Val Glu Ser Ser

85 90 95

Asp Thr Val Asp Asn Val Lys Ala Lys Ile Gln Asp Lys Glu Gly Ile

100 105 110

Pro Pro Asp Gln Gln Arg Leu Ile Phe Ala Gly Lys Gln Leu Glu Asp

115 120 125

Gly Arg Thr Leu Ala Asp Tyr Asn Ile Gln Lys Glu Ser Thr Leu His

130 135 140

Leu Val Leu Arg Leu Lys Gly Gly Met Gln Ile Phe

145 150 155

68

238

›PRT

Eucalyptus grandis

68

Met Ala Thr His Ala Ala Leu Ala Pro Ser Thr Leu Pro Ala Asn Ala

1 5 10 15

Lys Phe Ser Ser Lys Ser Ser Ser His Ser Phe Pro Thr Gln Cys Phe

20 25 30

Ser Lys Arg Leu Glu Val Ala Glu Phe Ser Gly Leu Arg Ala Gly Ser

35 40 45

Cys Val Thr Tyr Ala Lys Asn Ala Gly Glu Gly Ser Phe Phe Asp Ala

50 55 60

Val Ala Ala Gln Leu Thr Pro Lys Thr Ser Ala Pro Ala Pro Ala Lys

65 70 75 80

Gly Glu Thr Val Ala Lys Leu Lys Val Ala Ile Asn Gly Phe Gly Arg

85 90 95

Ile Gly Arg Asn Phe Leu Arg Cys Trp His Gly Arg Lys Asn Ser Pro

100 105 110

Leu Asp Val Ile Val Val Asn Asp Ser Gly Gly Val Lys Asn Ala Ser

115 120 125

His Leu Leu Lys Tyr Asp Ser Met Leu Gly Thr Phe Lys Ala Asp Val

130 135 140

Lys Ile Val Asp Asn Glu Thr Ile Ser Val Asp Gly Lys Pro Val Lys

145 150 155 160

Val Val Ser Asn Arg Asp Pro Leu Lys Leu Pro Trp Ala Glu Leu Gly

165 170 175

Ile Asp Ile Val Ile Glu Gly Thr Gly Val Phe Val Asp Gly Pro Gly

180 185 190

Ala Gly Lys His Ile Gln Ala Gly Ala Lys Lys Val Ile Ile Thr Ala

195 200 205

Pro Ala Lys Gly Ala Asp Ile Pro Thr Tyr Val Tyr Gly Val Asn Glu

210 215 220

Thr Asp Tyr Ser His Glu Val Ala Asn Ile Ile Ser Asn Ala

225 230 235

69

168

›PRT

Eucalyptus grandis

69

Met Ser Thr Ser Pro Val Ser Ser Trp Cys Ala Thr Ser Phe Ser Pro

1 5 10 15

Ala His Ser Ser Leu Lys Arg Ala Ala Gly Leu Arg Pro Ser Leu Ser

20 25 30

Ala Arg Leu Gly Pro Ser Ser Ser Ser Ser Ser Val Ser Pro Pro Thr

35 40 45

Leu Ile Arg Asn Glu Pro Val Phe Ala Ala Pro Ala Pro Val Ile Asn

50 55 60

Pro Thr Trp Thr Glu Glu Met Gly Lys Asp Tyr Asp Glu Ala Ile Glu

65 70 75 80

Ala Leu Lys Lys Leu Leu Ser Glu Lys Gly Asp Leu Lys Ala Thr Ala

85 90 95

Ala Ala Lys Val Glu Gln Ile Thr Ala Glu Leu Gln Thr Ala Ser Pro

100 105 110

Asp Ile Lys Pro Ser Ser Ser Val Asp Arg Ile Lys Thr Gly Phe Thr

115 120 125

Phe Phe Lys Lys Glu Lys Tyr Asp Lys Asn Pro Ala Leu Tyr Gly Glu

130 135 140

Leu Ala Lys Gln Ser Pro Lys Phe Met Val Phe Ala Cys Ser Asp Ser

145 150 155 160

Arg Val Cys Pro Ser His Val Leu

165

70

214

›PRT

Eucalyptus grandis

70

Met Pro Cys Pro Arg Ala Pro Pro Met Met Glu Arg Arg Ile Lys Pro

1 5 10 15

Gln Thr Glu Gln Ala Leu Lys Cys Pro Arg Cys Asp Ser Thr Asn Thr

20 25 30

Lys Phe Cys Tyr Tyr Asn Asn Tyr Asn Leu Ser Gln Pro Arg His Phe

35 40 45

Cys Lys Thr Cys Arg Arg Tyr Trp Thr Lys Gly Gly Ala Leu Arg Asn

50 55 60

Val Pro Val Gly Gly Gly Cys Arg Lys Asn Lys Arg Ala Lys Arg Ala

65 70 75 80

Val Asp His Pro Val Ser Ala Gln Asn Glu Ala Ser Thr Ser Ala Ala

85 90 95

Pro Gly Asn Glu Val Pro Asp Arg Ser Pro Phe Glu Pro Pro Ser Ser

100 105 110

Lys Ser Ile Tyr Tyr Gly Gly Glu Asn Met Asn Leu Thr Gly Leu Pro

115 120 125

Phe Ser Arg Ile Gln Gln Asp Arg Ala Ala Leu Ala His Cys Asn Ser

130 135 140

Ser Ser Phe Leu Gly Met Ser Cys Gly Thr Gln Ser Ala Ser Leu Glu

145 150 155 160

Pro His Leu Ser Ala Leu Asn Thr Phe Asn Ser Phe Lys Ser Asn Asn

165 170 175

Pro Gly Leu Asp Phe Pro Ser Leu Ser Thr Asp Gln Asn Ser Leu Phe

180 185 190

Glu Thr Ser Gln Pro Gln Leu Ser Arg Ala Met Ala Ser Ala Leu Phe

195 200 205

Ser Met Pro Met Ala Pro

210

71

166

›PRT

Pinus radiata

71

Met Ala Ala Leu Ala Thr Thr Glu Val Cys Asp Thr Tyr Pro Arg Leu

1 5 10 15

Val Glu Asn Gly Glu Leu Arg Val Leu Gln Pro Ile Phe Gln Ile Tyr

20 25 30

Gly Arg Arg Arg Ala Phe Ser Gly Pro Ile Val Thr Leu Lys Val Phe

35 40 45

Glu Asp Asn Val Leu Leu Arg Glu Phe Leu Glu Glu Arg Gly Asn Gly

50 55 60

Arg Val Leu Val Val Asp Gly Gly Gly Ser Leu Arg Cys Ala Ile Leu

65 70 75 80

Gly Gly Asn Val Val Val Ser Ala Gln Asn Asn Gly Trp Ser Gly Ile

85 90 95

Ile Val Thr Gly Cys Ile Arg Asp Val Asp Glu Ile Asn Arg Cys Asp

100 105 110

Ile Gly Ile Arg Ala Leu Thr Ser Asn Pro Leu Lys Ala Asn Lys Lys

115 120 125

Gly Val Gly Glu Lys His Ala Pro Ile Tyr Ile Ala Gly Thr Arg Ile

130 135 140

Leu Pro Gly Glu Trp Cys Tyr Ala Asp Ser Asp Gly Ile Leu Val Ser

145 150 155 160

Gln Gln Glu Leu Ser Leu

165

72

236

›PRT

Pinus radiata

72

Met Leu Val Leu Ile Ile Phe Gly Cys Cys Phe Ile Gly Val Ile Ala

1 5 10 15

Thr Ser Phe Asp Phe Tyr Tyr Phe Val Gln Gln Trp Pro Gly Ser Tyr

20 25 30

Cys Asp Thr Arg Arg Gly Cys Cys Tyr Pro Arg Thr Gly Arg Pro Ala

35 40 45

Ser Glu Phe Ser Ile His Gly Leu Trp Pro Asn Tyr Lys Thr Gly Lys

50 55 60

Trp Pro Gln Phe Cys Gly Ser Ser Glu Glu Phe Asp Tyr Ser Lys Ile

65 70 75 80

Ser Asp Leu Glu Glu Glu Leu Asn Arg Tyr Trp Gly Ser Leu Ser Cys

85 90 95

Pro Ser Ser Asp Gly Gln Glu Phe Trp Gly His Glu Trp Glu Lys His

100 105 110

Gly Thr Cys Ser Leu Asn Leu Asp Glu His Ser Tyr Phe Glu Lys Ala

115 120 125

Leu Ser Leu Arg Gln Asn Ile Asp Ile Leu Gly Ala Leu Lys Thr Ala

130 135 140

Gly Ile Lys Pro Asp Gly Ser Gln Tyr Ser Leu Ser Asp Ile Lys Glu

145 150 155 160

Ala Ile Lys Gln Asn Thr Gly Gln Leu Pro Gly Ile Asp Cys Asn Thr

165 170 175

Ser Ala Glu Gly Glu His Gln Leu Tyr Gln Val Tyr Val Cys Val Asp

180 185 190

Lys Ser Asp Ala Ser Thr Val Ile Glu Cys Pro Ile Tyr Pro His Ser

195 200 205

Asn Cys Pro Ser Met Val Val Phe Pro Pro Phe Gly Glu Asp Gln Glu

210 215 220

Asp Arg Asp Gly Tyr Thr Glu Gly Met Tyr Glu Leu

225 230 235

73

92

›PRT

Pinus radiata

73

Met Ala Ala Pro Arg Ser Ser Ala Lys Leu Gly Ala Leu Leu Ala Ile

1 5 10 15

Leu Leu Ile Val Ala Ala Ala Gln Ala Gln Asp Cys Ser Asn Ala Met

20 25 30

Asp Lys Leu Ala Pro Cys Thr Ser Ala Val Gly Leu Ser Ser Asn Gly

35 40 45

Val Lys Pro Ser Ser Glu Cys Cys Asp Ala Leu Lys Gly Thr Ser Thr

50 55 60

Gly Cys Val Cys Lys Ser Val Arg Ala Val Ile Ser Leu Pro Ala Lys

65 70 75 80

Cys Asn Leu Pro Ala Ile Thr Cys Ser Gly Ser Arg

85 90

74

92

›PRT

Pinus radiata

74

Met Ala Ala Pro Arg Ser Ser Ala Lys Ser Ala Ala Leu Phe Ala Ile

1 5 10 15

Leu Leu Ile Val Ala Ala Val Gln Ala Glu Asp Cys Ser Asn Ala Met

20 25 30

Asp Lys Leu Ala Pro Cys Thr Ser Ala Val Gly Leu Ser Ser Asn Gly

35 40 45

Val Lys Pro Ser Ser Glu Cys Cys Asp Ala Leu Lys Gly Thr Ser Thr

50 55 60

Gly Cys Val Cys Lys Ser Val Arg Ala Val Ile Ser Leu Pro Ala Lys

65 70 75 80

Cys Asn Leu Pro Ala Leu Thr Cys Ser Gly Ser Arg

85 90

75

92

›PRT

Pinus radiata

75

Met Ala Ala Pro Arg Ser Ser Ala Lys Leu Gly Ala Leu Leu Ala Ile

1 5 10 15

Leu Leu Ile Val Ala Ala Ala Gln Ala Gln Asp Cys Ser Asn Ala Met

20 25 30

Asp Lys Leu Ala Pro Cys Thr Ser Ala Val Gly Leu Ser Ser Asn Gly

35 40 45

Val Lys Pro Ser Ser Glu Cys Cys Asp Ala Leu Lys Gly Thr Ser Thr

50 55 60

Gly Cys Val Cys Lys Ser Val Arg Ala Val Ile Ser Leu Pro Ala Lys

65 70 75 80

Cys Asn Leu Pro Ala Ile Thr Cys Ser Gly Ser Arg

85 90

76

125

›PRT

Eucalyptus grandis

76

Met Ala Asp Arg Met Leu Thr Arg Ser His Ser Leu Arg Glu Arg Leu

1 5 10 15

Asp Glu Thr Leu Ser Ala His Arg Asn Asp Ile Val Ala Phe Leu Ser

20 25 30

Arg Val Glu Ala Lys Gly Lys Gly Ile Leu Gln Arg His Gln Ile Phe

35 40 45

Ala Glu Phe Glu Ala Ile Ser Glu Glu Ser Arg Ala Lys Leu Leu Asp

50 55 60

Gly Ala Phe Gly Glu Val Leu Lys Ser Thr Gln Glu Ala Ile Val Ser

65 70 75 80

Pro Pro Trp Val Ala Leu Ala Val Arg Pro Arg Pro Gly Val Trp Glu

85 90 95

His Ile Arg Val Asn Val His Ala Leu Val Leu Glu Gln Leu Glu Val

100 105 110

Ala Glu Tyr Leu His Phe Lys Glu Glu Leu Ala Asp Gly

115 120 125

77

805

›PRT

Eucalyptus grandis

77

Met Ala Asp Arg Met Leu Thr Arg Ser His Ser Leu Arg Glu Arg Leu

1 5 10 15

Asp Glu Thr Leu Ser Ala His Arg Asn Asp Ile Val Ala Phe Leu Ser

20 25 30

Arg Val Glu Ala Lys Gly Lys Gly Ile Leu Gln Arg His Gln Ile Phe

35 40 45

Ala Glu Phe Glu Ala Ile Ser Glu Glu Ser Arg Ala Lys Leu Leu Asp

50 55 60

Gly Ala Phe Gly Glu Val Leu Lys Ser Thr Gln Glu Ala Ile Val Ser

65 70 75 80

Pro Pro Trp Val Ala Leu Ala Val Arg Pro Arg Pro Gly Val Trp Glu

85 90 95

His Ile Arg Val Asn Val His Ala Leu Val Leu Glu Gln Leu Glu Val

100 105 110

Ala Glu Tyr Leu His Phe Lys Glu Glu Leu Ala Asp Gly Ser Leu Asn

115 120 125

Gly Asn Phe Val Leu Glu Leu Asp Phe Glu Pro Phe Thr Ala Ser Phe

130 135 140

Pro Arg Pro Thr Leu Ser Lys Ser Ile Gly Asn Gly Val Glu Phe Leu

145 150 155 160

Asn Arg His Leu Ser Ala Lys Leu Phe His Asp Lys Glu Ser Leu His

165 170 175

Pro Leu Leu Glu Phe Leu Gln Val His Cys Tyr Lys Gly Lys Asn Met

180 185 190

Met Val Asn Ala Arg Ile Gln Asn Val Phe Ser Leu Gln His Val Leu

195 200 205

Arg Lys Ala Glu Glu Tyr Leu Thr Ser Leu Lys Pro Glu Thr Pro Tyr

210 215 220

Ser Gln Phe Glu His Lys Phe Gln Glu Ile Gly Leu Glu Arg Gly Trp

225 230 235 240

Gly Asp Thr Ala Glu Arg Val Leu Glu Met Ile Gln Leu Leu Leu Asp

245 250 255

Leu Leu Glu Ala Pro Asp Pro Cys Thr Leu Glu Lys Phe Leu Asp Arg

260 265 270

Val Pro Met Val Phe Asn Val Val Ile Met Ser Pro His Gly Tyr Phe

275 280 285

Ala Gln Asp Asp Val Leu Gly Tyr Pro Asp Thr Gly Gly Gln Val Val

290 295 300

Tyr Ile Leu Asp Gln Val Arg Ala Leu Glu Glu Glu Met Leu His Arg

305 310 315 320

Ile Lys Gln Gln Gly Leu Asp Ile Thr Pro Arg Ile Leu Ile Ile Thr

325 330 335

Arg Leu Leu Pro Asp Ala Val Gly Thr Thr Cys Gly Gln Arg Leu Glu

340 345 350

Lys Val Phe Gly Thr Glu Tyr Ser His Ile Leu Arg Val Pro Phe Arg

355 360 365

Asn Glu Lys Gly Val Val Arg Lys Trp Ile Ser Arg Phe Glu Val Trp

370 375 380

Pro Tyr Leu Glu Arg Tyr Thr Glu Asp Val Ala Ser Glu Leu Ala Gly

385 390 395 400

Glu Leu Gln Gly Lys Pro Asp Leu Ile Ile Gly Asn Tyr Ser Asp Gly

405 410 415

Asn Ile Val Ala Ser Leu Leu Ala His Lys Leu Gly Val Thr Gln Cys

420 425 430

Thr Ile Ala His Ala Leu Glu Lys Thr Lys Tyr Pro Glu Ser Asp Ile

435 440 445

Tyr Trp Lys Lys Phe Glu Glu Lys Tyr His Phe Ser Cys Gln Phe Thr

450 455 460

Ala Asp Leu Ile Ala Met Asn His Thr Asp Phe Ile Ile Thr Ser Thr

465 470 475 480

Phe Gln Glu Ile Ala Gly Ser Lys Asp Thr Val Gly Gln Tyr Glu Ser

485 490 495

His Met Asn Phe Thr Leu Pro Gly Leu Tyr Arg Val Val His Gly Ile

500 505 510

Asp Val Phe Asp Pro Lys Phe Asn Ile Val Ser Pro Gly Ala Asp Met

515 520 525

Ser Ile Tyr Phe Ala Tyr Thr Glu Gln Glu Arg Arg Leu Lys Ser Phe

530 535 540

His Pro Glu Ile Glu Glu Leu Leu Phe Ser Asp Val Glu Asn Lys Glu

545 550 555 560

His Leu Cys Val Leu Lys Asp Lys Lys Lys Pro Ile Ile Phe Thr Met

565 570 575

Ala Arg Leu Asp Arg Val Lys Asn Leu Thr Gly Leu Val Glu Trp Tyr

580 585 590

Gly Lys Asn Ser Lys Leu Arg Glu Leu Ala Asn Leu Val Val Val Gly

595 600 605

Gly Asp Arg Arg Lys Asp Ser Lys Asp Leu Glu Glu Gln Ser Glu Met

610 615 620

Lys Lys Met Tyr Asp Leu Ile Glu Lys Tyr Lys Leu Asn Gly Gln Phe

625 630 635 640

Arg Trp Ile Ser Ser Gln Met Asn Arg Val Arg Asn Gly Glu Leu Tyr

645 650 655

Arg Tyr Ile Cys Asp Thr Lys Gly Val Phe Val Gln Pro Ala Ile Tyr

660 665 670

Glu Ala Phe Gly Leu Thr Val Val Glu Ala Met Thr Cys Gly Leu Pro

675 680 685

Thr Phe Ala Thr Cys Asn Gly Gly Pro Ala Glu Ile Ile Val His Gly

690 695 700

Lys Ser Gly Tyr His Ile Asp Pro Tyr His Gly Asp Gln Ala Ala Glu

705 710 715 720

Leu Leu Val Asp Phe Phe Asn Lys Cys Lys Ile Asp Gln Ser His Trp

725 730 735

Asp Glu Ile Ser Lys Gly Ala Met Gln Arg Ile Glu Glu Lys Tyr Thr

740 745 750

Trp Lys Ile Tyr Ser Glu Arg Leu Leu Asn Leu Thr Ala Val Tyr Gly

755 760 765

Phe Trp Lys His Val Thr Asn Leu Asp Arg Arg Glu Ser Arg Arg Tyr

770 775 780

Leu Glu Met Phe Tyr Ala Leu Lys Tyr Arg Pro Leu Ala Gln Ser Val

785 790 795 800

Pro Pro Ala Val Glu

805

78

264

›PRT

Eucalyptus grandis

78

Met Gly Ser Thr Gly Ser Glu Thr Gln Met Thr Pro Thr Gln Val Ser

1 5 10 15

Asp Glu Glu Ala Asn Leu Phe Ala Met Gln Leu Ala Ser Ala Ser Val

20 25 30

Leu Pro Met Val Leu Lys Ala Ala Ile Glu Leu Asp Leu Leu Glu Ile

35 40 45

Met Ala Lys Ala Gly Pro Gly Ala Phe Leu Ser Pro Gly Glu Val Ala

50 55 60

Ala Gln Leu Pro Thr Gln Asn Pro Glu Ala Pro Val Met Leu Asp Arg

65 70 75 80

Ile Phe Arg Leu Leu Ala Ser Tyr Ser Val Leu Thr Cys Thr Leu Arg

85 90 95

Asp Leu Pro Asp Gly Lys Val Glu Arg Leu Tyr Gly Leu Ala Pro Val

100 105 110

Cys Lys Phe Leu Val Lys Asn Glu Asp Gly Val Ser Ile Ala Ala Leu

115 120 125

Asn Leu Met Asn Gln Asp Lys Ile Leu Met Glu Ser Trp Tyr Tyr Leu

130 135 140

Lys Asp Ala Val Leu Glu Gly Gly Ile Pro Phe Asn Lys Ala Tyr Gly

145 150 155 160

Met Thr Ala Phe Glu Tyr His Gly Thr Asp Pro Arg Phe Asn Lys Ile

165 170 175

Phe Asn Arg Gly Met Ser Asp His Ser Thr Ile Thr Met Lys Lys Ile

180 185 190

Leu Glu Thr Tyr Lys Gly Phe Glu Gly Leu Glu Thr Val Val Asp Val

195 200 205

Gly Gly Gly Thr Gly Ala Val Leu Ser Met Ile Val Ala Lys Tyr Pro

210 215 220

Ser Met Lys Gly Ile Asn Phe Asp Arg Pro Asn Gly Leu Lys Thr Pro

225 230 235 240

His Pro Phe Leu Val Ser Ser Thr Ser Glu Ala Thr Cys Ser Ser Ala

245 250 255

Phe Gln Arg Glu Met Pro Phe Ser

260

79

136

›PRT

Eucalyptus grandis

79

Met Gly Lys Glu Lys Ile His Ile Ser Ile Val Val Ile Gly His Val

1 5 10 15

Asp Ser Gly Lys Ser Thr Thr Thr Gly His Leu Ile Tyr Lys Leu Gly

20 25 30

Gly Ile Asp Lys Arg Val Ile Glu Arg Phe Glu Lys Glu Ala Ala Glu

35 40 45

Met Asn Lys Arg Ser Phe Lys Tyr Ala Trp Val Leu Asp Lys Leu Lys

50 55 60

Ala Glu Arg Glu Arg Gly Ile Thr Ile Asp Ile Ala Leu Trp Lys Phe

65 70 75 80

Glu Thr Thr Lys Tyr Tyr Cys Thr Val Ile Asp Ala Pro Gly His Arg

85 90 95

Asp Phe Ile Lys Asn Met Ile Thr Gly Thr Ser Gln Ala Asp Cys Ala

100 105 110

Val Leu Ile Ile Asp Ser Thr Thr Gly Gly Phe Glu Ala Gly Ile Ser

115 120 125

Lys Asp Gly Gln Thr Arg Glu His

130 135

80

229

›PRT

Eucalyptus grandis

80

Met Gln Ile Phe Val Lys Thr Leu Thr Gly Lys Thr Ile Thr Leu Glu

1 5 10 15

Val Glu Ser Ser Asp Thr Ile Asp Asn Val Lys Ala Lys Ile Gln Asp

20 25 30

Lys Glu Gly Ile Pro Pro Asp Gln Gln Arg Leu Ile Phe Ala Gly Lys

35 40 45

Gln Leu Glu Asp Gly Arg Thr Leu Ala Asp Tyr Asn Ile Gln Lys Glu

50 55 60

Ser Thr Leu His Leu Val Leu Arg Leu Arg Gly Gly Met Gln Ile Phe

65 70 75 80

Val Lys Thr Leu Thr Gly Lys Thr Ile Thr Leu Glu Val Glu Ser Ser

85 90 95

Asp Thr Ile Asp Asn Val Lys Ala Lys Ile Gln Asp Lys Glu Gly Ile

100 105 110

Pro Pro Asp Gln Gln Arg Leu Ile Phe Ala Gly Lys Gln Leu Glu Asp

115 120 125

Gly Arg Thr Leu Ala Asp Tyr Asn Ile Gln Lys Glu Ser Thr Leu His

130 135 140

Leu Val Leu Arg Leu Arg Gly Gly Met Gln Ile Phe Val Lys Thr Leu

145 150 155 160

Thr Gly Lys Thr Ile Thr Leu Glu Val Glu Ser Ser Asp Thr Ile Asp

165 170 175

Asn Val Lys Ala Lys Ile Gln Asp Lys Glu Gly Ile Pro Pro Asp Gln

180 185 190

Gln Arg Leu Ile Phe Ala Gly Lys Gln Leu Glu Asp Gly Arg Thr Leu

195 200 205

Ala Asp Tyr Asn Ile Gln Lys Glu Ser Thr Leu His Leu Val Leu Arg

210 215 220

Leu Arg Gly Gly Phe

225

81

345

›DNA

Eucalyptus grandis

81

taataaatga tgaatttatt ataaacgtat ccgtttgaga tttttgtggg tcataggtgt 60

atcaatttga aatctttgat agtaacaaaa ataattttag gtagtttatg tttttcatga 120

tataaacctt gaaagttaat gctactaaat tgttatatat atattaggca aattacaacc 180

ttaatgcaac agttaatgac gtgatactgt tcagattata gatacaatgg ttatccttga 240

atgaataaga agaagtccta agggcaagtg ctatgagctt gcacgactgc ttttgcgcca 300

tttttgttta ccagcccggg ccgtcgacca cgcgtgccct atagt 345

82

72

›DNA

Eucalyptus grandis

82

cagtagggga cttgttcccc caagggcacg tgtcgttggt gaagctctgg cggtggatga 60

accgcgtggg cc 72

83

544

›DNA

Eucalyptus grandis

83

actagtgatt tcgtcgtctt cgtcttcttc gtcttctgga acttcgttgc tccgagcttt 60

atcagaaccg gcgatggaaa tgaaaccctc gttctctctc cctcgctcct ctctttcttc 120

tatccaggag cgtttgtaca ctgggagtac agagcttctt gcgataccga aactaccctt 180

ggacgactgg cctttttgcc tcgcgccccc tctctgagcc ggggcgcaat ttgtcccttt 240

cccagagcga agtgtcgatt ttgtccttcc acgaggcttt acctactccc atcgcccgag 300

ccccaagccc aggcccaaat gcctgttcct tgtggccctg ccaacattcc ctttgaaatt 360

aaaaaattaa aaaaaaactc tctgccaggc aaaagtaaag attaacacca ccaaaattta 420

taacaaattt atcattcatt aattttcgtt aaattttatt ttcaaattac tgagtcgaat 480

tacatgtata aattcacgga tgtatcggtt cgagatttta tcctctaatt atcattagtg 540

tatg 544

84

515

›DNA

Eucalyptus grandis

84

gattactata gggcacgcgt ggtcgacggc ccgggctggt ctgccttcct ttaactcccc 60

ttttttgtaa ctttttaaaa tgtagtttta aatttaattt aattactttt tatattaatt 120

atttaccaca tcagagacaa aacaatgtct tttttgtatt ttctagtcac gtcaacatgc 180

aaaacaacgc cattttgcac tcaccttgcc ggaaaattgc cacgtcaaca atttggctag 240

agtggcgctt aagtgatcta ttttgctcca attttggcac ttaagtgtca ttttcctaaa 300

ttttagcact taaagtattc ctctatgtca agttttgaca cttggggtgt actttgtcca 360

atcataaacc gtataagttc actttaaaca aaaatggcgc aaaagcagtc gtgcaagctc 420

atagcacttg cccttaggac ttcttcttat tcattcaagg ataaccattg tatctataat 480

ctgaacagta tcacgtcatt aactgttgca ttaag 515

85

515

›DNA

Eucalyptus grandis

85

actagtgatt tcgtcgtctt cgtcttcttc gtcttctgga acttcgttgc tccgagcttt 60

atcagaaccg gcgatggaaa tgaaaccctc gttctctctc cctcgctcct ctctttcttc 120

tatccaggag cgtttgtaca ctgggagtac agagcttctt gcgataccga aactaccctt 180

ggacgactgg cctttttgcc tcgtgccccc tctctgagcc ggggcgcaat ttgtcccttt 240

cccagagcga agtgtcgatt ttgtccttcc acgaggcttt acctactccc atcgcccgag 300

ccccaagccc aggcccaaat gcctgttcct tgtggccctg ccaacattcc ctttgaaatt 360

aaaaaattaa aaaaaaactc tctgccaggc aaaagtaaag attaacacca ccaaaattta 420

taacaaattt atcattcatt aattttcgtt aaattttatt ttcaaattac tgagtcgaat 480

tacatgtata aattcacgga tgtatcggtt cgaga 515

86

782

›DNA

Eucalyptus grandis

86

gagggtttca tttccatcgc cggttctgat aaagctcgga gcaacgaagt tccagaagac 60

gaagaagacg aagacgacga cggcgacatg ccttgcttga acatctccac caacgtcagc 120

ctcgacggcc tcgacacctc cgccattctc tccgagacca cctccggcgt cgccaagctc 180

atcggcaagc ccgaggccta tgtgatgatt gtgttgaagg ggtcagtccc catggctttt 240

ggtgggactg agcaacctgc tgcctatggc gagttggtgt caatcggcgg tttgaacccc 300

gatgtgaaca agaagctgag tgctgcaatt gcttcaatcc tcgaaaccaa gctgtccatc 360

cccaagtcgc ggttcttcct gaaattttat gataccaagg gttccttctt tggatggaat 420

ggatccacct tctgagctgt tggtcgcatt ctcctcagtg tttaccatgt atttcggccc 480

taaactctac ttctaggcct gttaaaagtg tcttttttaa ggtaattctg ctattacccc 540

tcttaagtgc atcttatcag taaacatgga atatcctgaa ctttgattat atgccggctc 600

gtggctgtgg aagcacttct ttatgttacc accagcttct caggtgaata taagctttgc 660

ccagtctgtt ctctggggga tttgcttggt gggtagtggc aatcagatgg ttttgtcact 720

tttgtgcata tttaagtagt aaatgtccac gacagcccaa agagtagcaa tccgggtgca 780

ct 782

87

115

›PRT

Eucalyptus grandis

87

Met Pro Cys Leu Asn Ile Ser Thr Asn Val Ser Leu Asp Gly Leu Asp

1 5 10 15

Thr Ser Ala Ile Leu Ser Glu Thr Thr Ser Gly Val Ala Lys Leu Ile

20 25 30

Gly Lys Pro Glu Ala Tyr Val Met Ile Val Leu Lys Gly Ser Val Pro

35 40 45

Met Ala Phe Gly Gly Thr Glu Gln Pro Ala Ala Tyr Gly Glu Leu Val

50 55 60

Ser Ile Gly Gly Leu Asn Pro Asp Val Asn Lys Lys Leu Ser Ala Ala

65 70 75 80

Ile Ala Ser Ile Leu Glu Thr Lys Leu Ser Ile Pro Lys Ser Arg Phe

85 90 95

Phe Leu Lys Phe Tyr Asp Thr Lys Gly Ser Phe Phe Gly Trp Asn Gly

100 105 110

Ser Thr Phe

115

88

1521

›DNA

Pinus radiata

88

ccttcaaaga caacagagaa agttatgcaa tatgctggca gctagctctt gggataatct 60

atttagcgat gggtttgtcg agaagttggg agcatttatt gtgaagcttc acagaaaaaa 120

tgtcgaatac atcaagcaca tgaagaagca atttgtgcca taggctatct ttagcctcat 180

ggatgttaaa ataatttctt tctttccttc cttcttcttt cttacccacc aaaacacaaa 240

ataatagttt caaattttga attttcaccc aattttatga gaggacaaaa ttacttagag 300

tctttcactc tttaatttat attctacata agtacctaaa gaggctctcc gacaatcata 360

tgataccata aaagtaacct cgattagaga gcgcctctcc atacaatcat ttgattttcg 420

agttaaatca aaattatagg ctatttccaa atcaatctat cgtccaactg aaaatttcaa 480

atgaatggaa ccagcacgga gtttcgtagg aaatagaagt aataggtgaa aagaagcatt 540

gtcgaatttg aaagaatacc ctacgttttc atttcaaaaa ccatggtttt ttgtaagagg 600

gattaagttg actcaaggtt gtagaaggtt gacataacaa tagcatgcag gcacaggatg 660

catgtagtgc ccgtaatttg gaccaaccta gtaagattgt cacccgtttc aaatgactgc 720

ctacaagtgc atgcaaaggc catggaagtt gatggttagt gaaaagatcc ggagagacga 780

ttattccatc atgcaatgca catcgcacgc ttgctttatt actcacacga ccaacgttcc 840

cttcatccac ggaattaatt tctctaatcg atccaataaa ccgccttcga tgtcgatttc 900

caaatgaatt aaatcgttac atgcccaccc gacttcacac atgctccctg cacgtgcaac 960

caaatccatt acgcccaccg ggcccggccc tgctcacaca tcttgcatcg cccaactact 1020

ctgattttac atgaatatca atactattcc ctccacttat aaaatggcca aacgccctgc 1080

ttagttctca aagcagatca gagcctttca agagcttccg caaagatttt ctttgcgagt 1140

aatttgatcg agaaggatgt ctgcatcgaa cggaactaat ggtgttgtcg cagtcaagtc 1200

tcgccgacag cacagacctg ggaaaacgac agccatggcg ttcgggaggg cgtttccaga 1260

tcagctggtg atgcaggagt tcctcgtcga tggatatttc cgcaacacga attgccagga 1320

ccccgtcctc cgccagaagc tcgaaaggct ttgcaagacg acgacggtga agacgcgata 1380

cgtggtgatg tcggatgaaa tattggcgca gcatcctgag ctggcagtgg aaggttcggc 1440

caccgtccga cagcgactcg agatctcgaa cgtggccgtg accgacatgg cggtggacgc 1500

gtgccgtgac tgcctcaaag a 1521

89

2590

›DNA

Eucalyptus grandis

89

ctgaaactgt cgctcggcga tgcataccaa aggctgaagg tatcagaatc taatgcagct 60

tatgtaaaag cgcgatcaat ttattgaccc cgacgacctt gactccatac ttcacgcctc 120

agctttgtgt tggatggtct tgacctctct caccctaaaa ggtagctcaa aagaatgaga 180

ctttccgtca tacttataaa ccgaccacca gcctctttca caaccgacat gggacaacct 240

caaatagaat ttttaacaac acccttgcac gctctttcta tccactttat tatgccatca 300

catgagcgtt ttccacgcgt aaatcggcta ccacccactt tcacacggcg gcgaaacgag 360

aaaaaggtcc tacctttgac tccccccgcg tcccaaattc tcactcccga ccggtaaccg 420

agctcacaag tttcagcctt tcatcatcat cactcgaagg cagagagaag gacatacact 480

aaagacaacg aaacagtctc tccatcccgc catccgacac gatccacatt acggtacgga 540

acacatcccg cggagcaacc cgacgtccca aactcttcgc tgatcaaaac cagtccggtc 600

gactccgttt cgcgcggacg caacgtgaga gagggagaga gagagagaga gtaccggcga 660

ggggatgatg ctgtgcggaa gcgtcgtcgg gcgctctccc ggcgaacgcg tctctacatt 720

ccggcgacgg cgacggcgac gaaggcgggg aggggaatgc cgcggggttt ctgcaacgac 780

ggaagctcac ggcatttttc agagagagag agagagatgg cacgtcagag cgccattccc 840

ccacgcgacg ttccgccttc cggtattcct tccgggagaa aaagtgggca aattgcaata 900

gacaaaaaaa aaaagaaaaa aaagacggtc acccaaatta tttcttataa cacaaaaaat 960

cgtacctata taatatatct atcactaact tgtgcagtat gacaaattta cacatttacc 1020

tgaaactgtt tttataacat aaaaaattta aacatttttc tgtgacaata aatgttcaca 1080

caaatataaa actgggattt ttatttcaat tacaaattta gaataaatgc gcaacataaa 1140

tacaaattta tgatttttcg tgttggcaag aaagtttgag ataaatgtat cattgtaggt 1200

aaagtttaga gttttttttt atggctttta accaaaatgc acattttagt tccgagttct 1260

aaaagaaaaa ttactatttt cctttacatt tacttatgta ggtgtgtaat tataaatatt 1320

aattctcttt aggatttgta acaattcttt gagcttttgt tttgccttta ggccattaga 1380

attactaaaa agttaataat ataaacattt tttcgaccac ggtcaccatt catacctaac 1440

ttctaattat tgaaagattc tcgcatttga tcgaaatcca tttactctca taaatttgag 1500

gttttgaacg gtatctacca taagatcatg gtttattaca aaacacttat ggcgggtggc 1560

gcggacctgg cgagaatgtg gctactttaa tgatgaggat ttgagatatt ataccacgat 1620

ccataataat aaaggagcgc ggcaatcata tcttttttca tataaaggac gatttatttt 1680

ctatgctgtg agtatttgct cttggaatta taagatatta gagatcaaac ctatcaccaa 1740

cggtgatttg aaattaaaga agtccttgta tcacttacaa aaataaatat ataaaaaaag 1800

ctttcattgt gcacttgaat atttaaacat aaattattag tagtagataa ttttttaatt 1860

taactaataa tgagcactca tttttagaaa aatagttttc aaatcattca ttttctactt 1920

aaaaaaacca attgaccaac taaattagta tctctcattc agttggtgaa tgaatgactc 1980

gcactctaac ccttcacttg gcgagtcatt ctgtgtagac cagtctctgc aaatctagcc 2040

atgctcatct agcaactacc ttcaagcgca agtactttgt catgtagacc aaacgttgag 2100

caacacggaa tgaatcctaa cgcacttgga aaacaatcaa tccacgctac gcaagctaat 2160

gctcacacaa gcatcatgat acccgaagcc gaaaatacat gagtcgaaag acatcgaact 2220

ccgccgtcct cgcgaatcat ccgaatcgca tgtcacgccg ctcgacttgg tagcttaacg 2280

agccttccag tacctgctgt ttaaatgctt tgtcaatgtg attcgaatcc tttcaaagat 2340

cctgaaagtg cagcttcaaa aatggcgtcg accaaatggg cttgcgttgc tgcaatctcg 2400

ctcctactga gcctaggatc gagcgctgct cagaggtctc tccttatgag cagcgccaac 2460

tggcaagagg ccggtgagcc gacggatctg gacttacgtg gaggaattgc cggaaccctg 2520

gggtcatcaa gtgagggcgg caccatggcc agctccgaca tgggcggttt tggccaggac 2580

atgcctggtg 2590

90

1172

›DNA

Eucalyptus grandis

90

actctcacta attctttagt tttccaattt agccccttct gtaattgctc atcttcttta 60

ccaaattctc taatttggcc ggcgaagggc tgacaaggga ttggtcatgt caccctcacc 120

aaaggttgcc gaaggtccgg tgacctcagc tgacggccac ctacaccaaa tctagctcac 180

tagcagccta agcccttcat caactctagt gaaaggtttt gagtattttt taataaaaaa 240

tatttaaaaa atatatagcg agagctcatt acaaaaaaat tttaaaaaaa aatctaaaca 300

ttacttgaac tcaaagtgac tttataaaga gtttttacca aaggatcttg gtttcatcat 360

ttgcactaca cccaaaaccc aatttctaag ttaaatcaaa cccactgtct aatagagata 420

aggtaaatgt tataaaccaa attccaaaat tccgaagcac taaatatatt tgctgatctt 480

ataatcgcca attgagaggg tctcattctc caagggattg tgacatatta gtaattgata 540

gggtctcatc cgtaggactc cgactcagcc gcgccacgtg actggatcgc tgaacggcgc 600

ggaaccagag gagcgtgatt acctaatatt ttctcctacc ttggccttga gattgaattt 660

cagaaaaaga aaaagaaaaa ggaacaactt cgccgactgt tctataaaat gcatgcgcca 720

ccccgacccc cacccacgca tcacatccat ccagcctcca cgacagacgc ataaacacaa 780

cacacgtcgg ttagagagag agagagagag agagagagag agagagagat gcttggacag 840

ttgtcgcacg agacggaaat gaaggtggga gcaggcaaag catgggagct gtatggcacg 900

ctcaagctgg tcctgctggc caagcaggaa ttctctaata ccatctgcga cgtcttggaa 960

ggtgatggcg gcgttggcac cgtcatcaag ctcaattttg gaagtttatc ctatacagag 1020

aagtacacaa aggtggacca cgagcgccgc gtgaaagaaa cggaggcgat cgaaggtggg 1080

ttcctggaca tggggtctcg ctgtatcgat tgcgattcga agtgataggc aaggacgagg 1140

aggagtcgtt ccgttattaa agcccccccc cc 1172

91

446

›DNA

Eucalyptus grandis

91

gggtgaaaac aattaatgag atcatttgaa ttaaggaaag tggaaaggcg gttttctgat 60

tggtacactg aaacaacagg aaggtggtgg aggccgcaat gatggaattt atccacttta 120

atcattttat gaaatcgata cactaacctt tgtttctcct aaacccaaag gcattaatcc 180

ctgtcctcct cactcgatct cgaaggccag aagggggagg ccgagcctct tgcttttttt 240

cgtgtataaa agggcctccc ccattcctca tttttcacca tcctccgttc gttcgttccc 300

ttccctttcc attgttgcgt ttaagccctc caattttctt ttggcgtccc gtttttgggg 360

ctcccttgaa gatctcctct tcatttcggg atttcctgcc ttcgccgcgc catttgaagt 420

tctttttctg agagaagaat ttagac 446

92

2119

›DNA

Pinus radiata

92

atcttattcc cacctcacat caataaattt tatacgattt taacatcttt aaaattaaaa 60

gaatcaagaa ggcatccagg tgataaagcc acgtccaata taaaatctcc tcggtggatc 120

ctttcaatcc agctacccaa tgcggcgaaa ataacgctga ttggactggg ctacactgta 180

atcacaaatt cccttccgtt tagatttcaa ctcgttgacc tacgagtatt ttatcgattt 240

aaaattatac aaaaaattgt ggaatgtttt acataagcaa aacttaaata atgtaaatag 300

cgatgatgct ttacttgtac ctaaaaattt cttccaaatt aaaccaaata tcaaatccta 360

gattgatgag ttccagtgga gtctgccatt ttatttcttt ctctctttca ttctttgcaa 420

cgaaaggaga aaatccttaa cacaattcga aaacgataat gattctggca aaagagaaaa 480

aaaacgtgaa gattagacac ttgttttgtt ttaaatgagc aatcacatgt gaatagagag 540

ggttttatgg gcctggtttt gtgtgcataa tttcttatga aagcgatgtg cctggagcgt 600

tgaagctcat agaacattgc aacaagagat cgagagtgtg ggttagaaaa ccgcaacaat 660

agtttgtgtc gtgtttttct atattcagag gtgttgtgtg gtaaatatct ctggatttat 720

ctcgaatgcg tcacttttac agacacagaa gctcagcgga aaccctcaac gctttaaggg 780

ccataaattt gctcagtttt aaaaattgtt tgatttccca ggtttgaata ttttcttttt 840

gttatcggaa gtggctctgc cttatgagta tcatgttctt ggttttgtgt tgggcgctta 900

ttgattcagg tatgtattat ttctagtcct ttttatcagc ataggtggaa tgttctgtat 960

tttatatttt ggggccatac acatggaacc gttgtcatta ccatgcttta tagataatgt 1020

ctctctgaat ttgtttttat aggcttttgc ctcctacgca gatttttaaa ggaaaataca 1080

aagatattta gccaattttt gttgttgtga ccttgaattt ctaaaaaatt taatggattc 1140

gttttctaaa ttcctgattc gtcaaaggct gaagggcgcg atagtaatag aaaatggacg 1200

agagtttatc ttttcatggc tggacacaca gaatttgtgg aggggattct ccattctggt 1260

ttatccaccg ttagttctct ctgtactcca cccttagttc tctttgtact cgagaccttt 1320

aatgattaac cctgcttatg ctgtcagtac tgaactcact tccagagccc caaaaatctc 1380

tcccaagttt gccttatttc ttaaaataat tcacaagtag aaaatgagat ttttgcaatt 1440

ttgtaactaa catttcccgg tctcctctgt atgttttcac cccttaatgt aattgaaatt 1500

tgcacccggg ttagattcaa agcggagaat aacatcgggg ccttgttcta gacagagatt 1560

tttcacaaat aacaggttcg aaggtatgtg tagacatctg ggtagttgta gaataaagac 1620

ggagcccatt aggtggatcc aatcgaagaa ctcagatggg aaaacagata aaaattatcg 1680

ggtggacctt cctccacatg ttaattatat atcaagtgtc gccaatcctt atgtgaaaca 1740

tttagtaaag cttcgccaga gcacttctta taggcattct gtgggctctg ttgttgtggt 1800

tggaagtact cctttaaggg aggtatctga atatttgcaa cagaagtcag taaaacaagt 1860

ggttgactgt ctgtttgtac aagatgttac tggcatacct gtgggcttga tagagacttc 1920

caggcgcatt gtgcatgtaa atcatttggt gatgcagaag ctagccggag tagagtctat 1980

agagcccact gaagcaattg gtgtaatcaa gcttcctagc agcttctaca acttggaatc 2040

tcttgaaatc actctagttc ccagatatgg tgctcgtcgc cacatcgtct gcttgtactt 2100

gatggcattc aggatcctg 2119

93

2571

›DNA

Eucalyptus grandis

93

aaggtaactg gttcagcaga gcgcagatac caaatacttg ttcttctagt gtagccgtag 60

ttaggccacc acttcaagaa ctctgtagca ccgcctacat acctcgctct gctaatcctg 120

ttaccagtgg ctgctgccag tggcgataag tcgtgtctta ccgggttgga ctcaagacga 180

tagttaccgg ataaggcgca gcggtcgggc tgaacggggg gttcgtgcac acagcccagc 240

ttggagcgaa cgacctacac cgaactgaga tacctacagc gtgagctatg agaaagcgcc 300

acgcttcccg aagggagaaa ggcggacagg tatccggtaa gcggcagggt cggaacagga 360

gagcgcacga gggagcttcc agggggaaac gcctggtatc tttatagtcc tgtcgggttt 420

cgccacctct gacttgagcg tcgatttttg tgatgctcgt caggggggcg gagcctatgg 480

aaaaacgcca gcaacgcggc ctttttacgg ttcctggcct tttgctggcc ttttgctcac 540

atgttctttc ctgcgttatc ccctgattct gtggataacc gtattaccgc ctttgagtga 600

gctgataccg ctcgccgcag ccgaacgacc gagcgcagcg agtcagtgag cgaggaagcg 660

gaagagcgcc caatacgcaa accgcctctc cccgcgcgtt ggccgattca ttaatgcagc 720

tggcacgaca ggtttcccga ctggaaagcg ggcagtgagc gcaacgcaat taatgtgagt 780

tagctcactc attaggcacc ccaggcttta cactttatgc ttccggctcg tatgttgtgt 840

ggaattgtga gcggataaca atttcacaca ggaaacagct atgaccatga ttacgccaag 900

ctatttaggt gacactatag aatactcaag ctatgcatcc aacgcgttgg gagctctccc 960

atatggtcga cctgcaggcg gccgcgaatt cactagtgat tggcccgggc tggtctggag 1020

tggccaccat cggcataatg actaggaacc cggaacatca actgatggaa gaaaagccga 1080

cattcctcat caagagctcc tctcactcct tccccactac tactataggg cacgcgtggt 1140

cgacggcccg ggctggtctg ctgtcatatt tgtatatgag gtcctatgta tgcttgctat 1200

gtgacctcct tcatgtatgc tgtgaagaga gtgtagcagt aacatggcca tctgcgaaat 1260

atggattcac cttaaaatct gatgattttc agaaaacgag gaaggtgctt gccgagaaga 1320

ttgcacagct caattcagct atagatgatg tatcctctga gctccgaact gaagaatcat 1380

cagatgagat tgctgttgcc cctgatgaaa ttgaagctgc tgtttgatgg cccaaacctc 1440

ccaggcctac gatcatggtc atcttctgtt ttggtgcaat tggctctacc tttttggtgg 1500

cctccatata acagaataat ggttcatatt gtaaaatctt ctgtttattt ctaaagacca 1560

atgcactcag tttcttttga tatgattgtc tcgattgagg aagtgcatca ttcgtggtat 1620

gattatgcag aataccattt aactcagcag actttgtacc gtatcatcgc agcttttccc 1680

ttcttgtgta tgcataaatc tagtccttca ttgaaggtga tcgccgttac agtctggata 1740

gtgtgtgcca tcagatggca ctacgattag tgtggttgac atggtgtcaa cttgaaagcc 1800

aattggtgac gatggtactt aatgtaagat tggcagatgg tgagaacgag attttgctcc 1860

agaatggcaa agcaaggcta agttgtagcg aatcaaatga tctacgaacc atcctagctg 1920

gctgtgtgac cacacactga agttctattg aactaagcca gttatggatg atatgggagg 1980

agaaaattga gaaatccatc agatggagtg ttggccgtgt tgggcttttg tcgcaggccg 2040

atacttcgaa ttcaggcgta tttttattcc tgactgccgc ctctcccgga aagggaaggc 2100

ggatattatt ctctgaacga tttccaccat caactccaca tcgatctcca agccagaaat 2160

atacacaccc caattttctt ttaaatatat gggacatata tggtgtaggc tctcgcgcat 2220

gttaacacat aagctctctc aacaaaaatc tggctcgtgc ttttaaccga gaagttcacg 2280

agtcattgaa ggagtggcct ttaggggagg gagagagatg gattggtggt taaaatcagt 2340

ctgtggctca catttatacc gtggagatcc cccaacagca accttatccc attatatatc 2400

cccacaacac catattcacc actcgttcct tctaattggc ttccaaccat aattcacaga 2460

cacacatgta gtgaccaatg agaaaggaag aaaaatacag gctttcgaaa gctagtgcgg 2520

tataaataac ctgggaaaag caagccgctt gagctttagt ttcagtcagc c 2571

94

1406

›DNA

Pinus radiata

94

aaagaggcgg aggaattgtc tagatggtca aaagtgaccg gaatctaagc aaaaaatttc 60

aaaaaatgtt gtaaaggtag cgtttgaatt gtgtttttga tggtggaaat ggattcaacg 120

ccatcaaaaa cgtctaagac acctaaaatt ttgaatttta acaactatat cttggattta 180

caaaaatcct tgccggattt tctctaaact ccttcacctt acgcaaaaga tatatatttt 240

tttgtgtgat gttgtgcatt ataagtttga tagtgaagta atgatatata tcctttatgt 300

gatggatgat tgaataatga atatattaaa tgaaataaat aatgatggga taatgaatat 360

attatatgaa ataaatataa agtaaaatgc tattttttaa tggtgttaat gatgaattag 420

tatcatcctt aaataatttg ttagtgaatt attaaaatga tgagttagca tggtcgttaa 480

ataaattgtt agtgaattat tatatttata tatttcctta ttagaaagtt ttttttttgt 540

aaaagttttc cttgaacttc acccatattt aattatcaat aatttatatt taataaatga 600

tatatataac ttctagcaga atgacacgcg acttgtatat cttttcattt tttaacccat 660

gaaaaccgat tagggtattg caaattaggg cattgccatt caaataattc tcagatgaaa 720

gattctctct aacaattaca aatgattatt tttttccatg agtgttgcat gttcgaacgg 780

tctgcccagt ctgtgagaga gcatagagaa ccctccctgc ccaatttgtt agagcataga 840

gaaccctact gcatgagtag taagaaaaat attcggtctc aattcggcaa agaccacctc 900

gaatggatga cttcaacgac aatctcatga tagtgttctg atcagcacca gttcacctat 960

atattttatc tagggtttag tttgcatgta tcaatcctct ggtgcactag gtaattcttt 1020

cctagtatca tatatcctta atactgtttt gtcttttaat ccatggctac catcagaaca 1080

agctcaaagc agaaatcggg agcatcagcc atcctcttgc ttatcgcgct tgcagggtta 1140

gtaaatgcgt gcaacgctgt gggtattgag ccaatgtgcg acactgtggt gtcgagtctt 1200

ctgaggcttc tgccatgcag gacggctgtt gatccctcaa ttgccgccat tccacttcca 1260

agctgctgca acgcggttga gtcagctggg cttcaatgcc tctgtctcgt cgttaacggc 1320

cctccttttc caggggtcga ccgcggcctc gcaatgcagc tgcctgccaa atgccatctc 1380

acccttcctc cctgtaacag ttagtt 1406

95

2546

›DNA

Pinus radiata

95

ctggtagaac aagcagctca aggagcacca aggcacgagc ccactttgca tgttgtagac 60

taacgaattt tacattagaa taaaatatgt cgacaatatc gaggagatct tctccaaaat 120

ccaactcatt aatctctatt atgcacaaac gagtgatgtg tcgagactca tctgccaaca 180

agccatcaac atcaagaagg gaacggaata gagccaaagg gaaccctaga gaccctcatc 240

cacataataa tgaaatattc cacgtgtgtt tttcaaaatt tgaaaatttc atgtattttt 300

tggttgattg gttgtggtct ggttttttcc aaattcaatc tagttcaagt ttttggagtc 360

gaccagttgg gtaaccagtc taattctggt aacattgcat tgtacttgat ctcaataaaa 420

gcatatagga tagaattatc ttctgtcttg atggtttcca tgagaaccaa ctgctatact 480

atgaaaaata tcaatgttcc acaatatttt tgggacaagg gaacacaaga ttgagtcaac 540

agttcaggac cccagaaaaa ttattcctga gttcgcagat tattttccta aaagtgaaca 600

attcaagacc ctagccaaat cattcccaag tccaagttat gtgacactgc gactaacaag 660

gcaagttgga agaaaccatc aatcaatctc ctagttaatg acagtccttg taagaagttc 720

aagaagatta acaccagaag aggtcatgct gactgctttt atccaattct ctctgctctt 780

caccaacaga aatagccaag atggttgtac ccattcccta atctaattta ttatatgaat 840

ttctctttat ttttctacat ataaaaaaca aaaacttttc ttgatggtca aacagaaaag 900

gcagttcgat tggatttaaa catccaaata cctcccacag attgagaagg ccaagcccca 960

atccaacagt ccatgatata atatttattc aatcacactc aagataatgc aatgaaggtg 1020

caccacgcta ttagattctg cacagaactc agatgactgt aattatcaac tttaaccagg 1080

agtaatttaa aaactcaatt gtgcttcagc tatgtggaaa aactttggca ctggaaatgg 1140

tataaatgtt gttgaataag caaacatttt tcaagcactg aattcaaagt caagtcaaag 1200

gaacatctta cttgggctgt acaggaaatc tgaagtacaa aattagcgaa aaaacaggag 1260

aaagagagta gtcattacat gttataacat taccatatag gattttgtaa tacttcttga 1320

tatttcaact tcccgactga tgaaatgtat gccactacag aacaggtcag tcatgtatgt 1380

gagcaattag ccaaactagg tcctaaggtt caaccagtgc agacaacgct gtaactgaaa 1440

caaatttgtg ggacaattaa aaattctcta ccaggatagt tgtaccagta ggtgcccttt 1500

tcaaaccatg atttaaaaca caagggtggc ttaccacttg accaaatcat ttaataacca 1560

acccctcgaa catatcaaga aagaaaacat ctgcatataa gtaaattgaa agatgatatt 1620

taagaggcac tgccttaaat tttccatttg gacaaatcca cattgcttga taagcataaa 1680

accttggtta agagcaagtt tagggaacca tcaaatattt ctacatactt tacaatagtg 1740

tgtttataaa gctaatcaaa tgcttctatt taaatatata gcaacctaca caagaaattc 1800

actaggacag caatcacttg gccaatgtga ttaccaatat aaccatactt gaagagcata 1860

cataaatcac aaataatgat tcaattagaa atatcttaaa gataaactat tattcaatgt 1920

acatgttaca aagaacctca cctgtccgcc tttgaggagc aagtagacaa ctaaaagcgg 1980

aggttacatc ctgaactgaa cttgttctcc tctgttccaa gaacttgcat tgtattttga 2040

gtaacttcac tcgtgccgaa ttcggcacga gaaaacactt tgattgcttc cgcgggtggg 2100

ttttactttc tctggaatag ttagttccgc cgtttttgga agatttatca gaatggccaa 2160

aattcaggtg tcaaacggga gcgtcgtggt ggtggcggcg atgatattta tggtggcggt 2220

ggccatgcaa aaccatcacg tcgccgccca aagtgctgac tgcgcaccac cgcggagttg 2280

ctgagcccct gcgcctcggc ggtgggaaac aacccgcaga ccccactccc gaatgctgtg 2340

ctgttctcca gaccgccgat gtcgactgca tctgcgccct cgtcgaatca accataaaat 2400

tgccttccga atgtggtctt gacacccccc agtgcccaag cgactagatt ctcaagaccg 2460

tgactgagtg ttggtttcag agccagtaaa cattcattct gctaataaat gagtgtatgg 2520

agctttaata ttggaaaatg cttcat 2546

96

4726

›DNA

Pinus radiata

96

gattactata gggcacgcgt ggtcgacggc cctggctggt cctaggacac cgtaatatat 60

aacctcgaca tggcttacaa agctttgact tgcattctca ttgggcttac aatggtgctg 120

ccaaaaatga aaaagtacat atgtacccct gttgaaatga gcagtaatag gcttgaacaa 180

tagtgaattg ctacaaaatt atgaatgcct ttctttgctt gaatgtgggc taaggagaag 240

tgggatttac atttgacttg caaatcctaa gacttgtcta gagctaagcc tccagaggag 300

gaaccatctt acatagtctt gagtctagag cggagaagat agccaaattt gaaaggaaac 360

ttttatttat ggggagaagg caaacaactt gagggggaag gatgatcaat aagtagggta 420

agggaatcca caacagaggg cactaaggaa atgggggtgt tagaattggc aactagggcc 480

aaattccacc ttgggatagc tctctggatg gagatgatga ttgcattaga ttcctctttt 540

cgagaggacc aagattgata taaagatcat ctcatttgga caagcatagg tatgattttg 600

aatttatacc cactcatgca caattttttt aggtccgcca catcatcatg taggctcatg 660

aagcccaacg gacatgactc ttcgccctta tcgtcttgta taaatacaag tgtcctccca 720

cctcatttgg catcttcatc tcttacagat tctctcttct tccctcattg gttcttgcat 780

cattgggcat tctctctctc ccacgtgtgg cacaaggagg atgaaattac aagaccgaaa 840

ataatagaaa ttttgcaatt tgaccagcat tgaccatgac cttccaagca tcattcgact 900

tcaatttttt tgggttattt ttgtctcaac aagccgcata ttttggcaaa aaaatcgagg 960

cattctgggc acttcgacta caaaccaaaa ttgtaggttg actgcaaatt tcaaatagtt 1020

tgactattga cattgtcact gttttcgatt gactttgacc tcctaattag gccgagtttg 1080

actaggggag gctgatttgt tttaaggaca tttgattgat gctttgacta gcattgactt 1140

ttatagttaa ggttgaagtt tgactacagt tgactgcata aatttgcaga gatgttttga 1200

ctttgaattg ggcaagtcaa tttgaatttt gtactatctc tctattttga acatttgata 1260

taataataag aagattcgat caaagggttt tccccgcatt gggttttttc cctggcatcc 1320

gccaaatctg gtgttctctt gtctttgctt gtcttatgca ttttgtttca ttttctatct 1380

acttttactg tcaatgtgat tattgtcagt gttattggaa attggaaatt gtgattgggc 1440

tgctaaggaa cattgaagta aattgtgcta aacaaagaac ataccattgt taacgaaaat 1500

taacaaaggg gaaacacaga ggaatggttg caattgcaag attgtcattg attttgactt 1560

caagtgagga aggtcgcgtg gaggtcgcaa ggggagagga ataggagaga aggccctatc 1620

aacttgttca aggagagggg caatacaagg aatggaggaa ccctcaccaa tgaataatcc 1680

atgcacaaaa gtaatagaat gaacaaactt accacacgga agagcttcct tgttgccaaa 1740

agccttgcct ccgagacctg aatcctccaa tgcatcaaaa ttattgatca ttgaatcaac 1800

cacgattagg gccacttcct tggctaataa agcaattagt gtagcaaatt ctaaagctaa 1860

cttcaaagaa accttagctt tccaaaaaac aattgaaggg aggcaatgaa gatggcttat 1920

cacactaagc ctaaacatgc cccaccctat ggcatctaaa acatctaaaa gggattcact 1980

agtaatcgat cttttgtact tatgaaaaat tcccatgaac caattcgatc tcttccaaaa 2040

agccatctat gaggtcaacc tcaacctggc tctaatgttg attgagcttg taatcctagc 2100

cctactccaa tcttaagaac caaccaattt tatttccaat tgattcaagg acccctacac 2160

tccaaaagaa gcaagggaag gccaaggaga atggcccaaa cttgagcaga gaataaggat 2220

tctctgtgag ggtcgaaact aacatcccat tcacgtaaaa tcaaaccaga gagacctcaa 2280

ctccaactct tcttaatgat gaagcacaaa tattattttg agtgaaattt gaaaccaaga 2340

aaacctctca ctaatatatg gaagaggggc aatattcaac cattggtacc caaatcgcct 2400

caagacactt accaagggag ccaaccaaac aatcttacca caaaaccaac caacagtgtt 2460

tttacccaca agctcttgga tggaatccag gataatgtct tcaccaacaa ccatcttatg 2520

tctatccttg caagcacaaa tgcattgagc tttagatttg gagtgcataa atacaggggg 2580

gtatccaggg gggggagggg gtttgctaga accccagact caccaaggca tgaagacaaa 2640

atgaggagag agggatctag attgggggat gcaagttgat gaagcatgaa aaggcaatcc 2700

atcaccctgc atggcatatt tacgaaggtt gttcagagga atgagaacta atggatgaac 2760

aacagctggt agaacaagca gctcaaggag cgccaaggca cgagcccact ttgcatgttg 2820

tagactaacg aattttacat tagaataaaa tatgtcgaca atatcgagga gatcttctcc 2880

aaaatccaac tcattaatct ctattatgca caaacgagtg atgtgtcgag actcatctgc 2940

caacaagcca tcaacatcaa gaagggaacg gaatagagcc aaagggaacc ctagagaccc 3000

tcatccacat aataatgaaa tattccacgt gtgtttttca aaatttggaa atttcatgta 3060

ttttttggtt gattgttgtg gtctggtttt ttccaaattc aatctagttc aagtttttgg 3120

agtcgaccag ttgggtaacc agtctaattc tggtaacatt gcattgtact tgatctcaat 3180

aaaagcatat aggatagaat tatcttctgt cttgatggtt gccatgagaa ccaactgcta 3240

tactatgaaa aatatcaatg ttccacaata tttttgggac aagggaacac aagattgagt 3300

caacagttca ggaccccaga aaaattattc ctgagtttgc agattatttt cctaaaagtg 3360

aacaattcaa gaccctagcc aaatcattcc caagtccaag ttatgtgaca ctgcgactaa 3420

caaggcaagt tggaagaaac catcaatcaa tctcctagtt aatgacagtc cttgtaagaa 3480

gttcaagaag attaacacca gaagaggtca tgctgactgc ttttatccaa ttctctctgc 3540

tcttcaccaa cagaaatagc caagatggtt gtacccattc cctaatctaa tttattatat 3600

gaatttctct ttatttttct acatataaaa aacaaaaact tttcttgatg gtgaaacaga 3660

aaaggcagtt cgattggatt taaacatcca aatacctccc acagattgag aaggccaagc 3720

cccaatccaa cagtccatga tataatattt attcaatcac actcaagata atgcaatgaa 3780

ggtgcaccac gctattagat tctgcacaga actcagatga ctgtaattat caactttaac 3840

caggagtaat ttaaaaactc aattgtgctt cagctatgtg gaaaaacttt ggcactggaa 3900

atggtataaa tgttgttgaa taagcaaaca ttttagaaca tttttcaagc actgaattca 3960

aagtcaagtc aaaggaacat cttacttggg ctgtacagga aatctgaagt acaaaattag 4020

tgaaaaaaca ggagaaagag agtagtcatt acatgttata acattaccat ataggatttt 4080

gtaatacttc ttgatatttc aacttcccga ctgatgaaat gtataccact acagaacagg 4140

tcagtcatgt atgtgagcaa ttagccaaac taggtcctaa ggttcaacca gtgcagacaa 4200

cgctgtaact gaaacaaatt tgtgggacaa ttaaaaattc tctaccagga tagttgtgcc 4260

agtaggtgcc cttttcaaac catgatttaa aacacaaggg tggcttacca cttgaccaaa 4320

tcatttaata accaacccct cgaacatatc aagaaagaaa acatctgcat ataagtaaat 4380

tgaaagatga tatttaagag gcactgcctt aaattttcca tttggcaaat ccacattgct 4440

tgataagcat aaaaccttgg ttaagagcaa gtttagggaa ccatcaaata tttctacata 4500

ctttacaata gtgtgtttat aaagctaatc aaatgcttct atttaaatat atagcaacct 4560

acacaagaaa ttcactagga cagcaatcac ttggccaatg tgattaccaa tataaccata 4620

cttgaagagc atacataaat cacaaataat gattcaatta gaaatatctt aaagataaac 4680

tattattcaa tgtacatgtt acaaagaacc tcacctgtcc gccttt 4726

97

635

›DNA

Pinus radiata

97

aaattctatg aaaaaaatcc aatcatatta aaagtccaat tgattagcaa ttttatgaga 60

aaaatccaat tatgttaaaa gtcactgagt gtggccgaaa ttgtgaccga aattgaatgc 120

aataaccgag ggtttttcaa accaaggtta agcctctcat cattggggtg tgtatgaaaa 180

tgtaatgggc atcgataacc ttttattaca acttcacgaa aattgcctct attcaatggg 240

tgtggatgaa aatgtaagtg cgcatcgata atggaaagcg atatgcagca aaatcaataa 300

acctgacttc ccatgtgagt gatgatttga tcgtacaact gatggtgtga agttactttc 360

agcttcacct tcgggcataa tcagggaagt agggccaagt ttgcttagta tcactctaat 420

ccccaacacc gtgattacta tcttcatcaa caatggccac cttcgtcatt actttaactg 480

gtgggataca gctactttac aactgtaaat ttgttgaggc agcctatcct cagcctatac 540

atactaatta ttgcagctcg attaggtatc tgctgtgaga atagctgtgt atctctgcgc 600

tggttgcagg atccaagttc ctctcagagc cctcc 635

98

468

›DNA

Pinus radiata

98

ctggtaaatt gagattccaa attattgatg cgaagcttcc tcgtggctgg tcggtgctgc 60

tggcatccaa accctaaatg aaaaagaaaa aggtgtccgg acggattttt ttagtatttt 120

tttcttattt tttttatgaa ccgtcggatt cgagatcgga cggcgatccg aaactgcaag 180

cgtcggccgt cggatgcagc atcggacggc aaagaaggaa ccctaaaacg cattgcaacg 240

tgcttggtgg gtggagggtc tatggccagt atatgttgat aacaagggag aggaagtagt 300

cctcttcatc tagtgcgagt ctctctgctt ttctacgccg ctgcgaagct gttctgtggt 360

gtttctgatt ctccagactc aggcagtcgt ttttgtaaga gaatttagtt catcatggga 420

aaggagaaaa cccatatcaa cattgtggtt attggccatg tcgactcc 468

99

222

›DNA

Pinus radiata

99

atccaaaccc taaatgaaaa agaaaaaggt gtccggacgg atttttttag tatttttttt 60

tcttattttt tttttatgaa ccgtcggatt cgagatcgga cggcgatccg aaactgcaag 120

cgtcggccgt cggatgcagc atcggacggc aaagaaggaa ccctaaaacg cattgcaacg 180

tgcttggtgg gtggagggtc tatggccaga tatgttgtaa tc 222

100

597

›DNA

Pinus radiata

100

aaatgaggca gctaactatt tatttggttt tggcttcact gacttgttcc ttagtgtatt 60

aatgaacaat ctctttagac tcagagatgg tgagaaagat tctatgagaa atattcttgt 120

tattgcttcg actcatatcc cccaaagagt ggatccagct ctaatagctc caaatcgatt 180

agatagatcg atcaatattc gaatgcttgt tatcccacaa cgacaaaggg aatttcctat 240

tcttttatgt agcaaaggat tatactcggg aaaatgtccc gatgaatttg gatctataac 300

catagattat gatgcacgag ctctattagc tcaggcctct ctgctgctcc ttggattgca 360

atctcattct ctgatttgcc gtgctgtttg ctctgctcac ttcagcccag atggagacct 420

tcttgttcac atcggagtct gtaaatgagg gacacccaga caaactctgt gaccagattt 480

ctgatgcagt gttggatgca tgcctcaccc aggaccccga cagcaaggta gcatgcgaga 540

cttgcactaa aacgaacatg gtcatggttt ttggtgaaat caccaccaag gccgatg 597

101

669

›DNA

Pinus radiata

101

cctggaaatg ctatattaac tcaacaaagg attttcagcc aatcacaatt tgacaggttt 60

gaaatgaaag attacaggca tttccaatgg aacagaatat aattacttta ttccctcaaa 120

gtatcgtata aaataaatct tttgctccac acactttgga aaatacattt tcaacaatgc 180

accgacaaac tttttctacc acgttatgga accatacaag ttaaatttaa acacgaatta 240

cgcgtatatt tctaataaat cgatggttga gattgaatgc cgtgggcgat tctcacgcgt 300

ccgattggga tcactagtcc atcactcatg gtctgcattg cctttaaatt ggcggggcga 360

ggaaagacca atgcgtcatt ggtgtagacg agctctatta gctcaggcct ctctgctgct 420

ccttggattg caatctcatt ctctgatttg ccgtgctgtt tgctctgctc acttcagccc 480

agatggagac cttcttgttc acatcggagt ctgtaaatga gggacaccca gacaaactct 540

gtgaccagat ttctgatgca gtgttggatg catgcctcac ccaggacccc gacagcaagg 600

tagcatgcga gacttgcact aaaacgaaca tggtcatggt ttttggtgaa atcaccacca 660

aggccgatg 669

102

230

›DNA

Pinus radiata

102

atccacctcg gaatgaaatc actatgcaca ctccaccttt tttttggctt cttttctcgt 60

tgcctttacc atcagaatca agcacgaaga gtaaatatca cccatgcttt acaagtgggt 120

tggtagcatt agcgattccc ttcaccaaat gaaccctttg ctggtgatga gtggacaacc 180

taaagttgtt tgctggtgat gagtggacaa ccagagtggg ggttggggaa 230

103

596

›DNA

Eucalyptus grandis

103

actttgaaag ggtctcgagt caaagtgctc aaattgagag ggagaatttt agaacaaaat 60

cagatttgga gaatacatgc cattttaggg ggattttggg gatttcgcat atggcgtcgc 120

gtcgtcggcg ccttcttctt tacagattgt atcctcccat taaccgcgtg gacctgcact 180

gtaaccccga aacggtgggg gccaatttcg tctttccgcc tcctccactc agcttcgtgg 240

aagattaaaa tcctcaccgt ccgtgcaaac gccacgtggc gcgttagttt gcgcgtggaa 300

aggtcctcac gaaccgtaaa gggcaaaaaa aagggaaaat aaaaaaggag gaggaggagg 360

gaggaggaag aattgtccga ttgaaaataa gagtgcggtg gtgtggtgtg ggtagatctt 420

gaattgaacg agctcaatcc gcgtatttaa acccgccccg cttcctcatt cttccttgtc 480

catttcaact ctccctctct ccctctcttc tgcccctcga tcgatccagc gatcttccta 540

tttccggacg cggggagcag ctcctcttgt cgaaggttct aaattagtgt ggagag 596

104

653

›DNA

Eucalyptus grandis

104

aaaattttcc tttattttct tttcattaaa aagataaata aataaaaaaa aaaaagaagg 60

aaaacacatc gaggtgaggc ttaaaggtgc taggcaagga ccaccaagcc tacacaaggg 120

tcggcgaccc tcaccaatgc tggggcgagg gtgagcaacc ctcatccaaa tctggagagg 180

gttgtcactc gagaaagggt cactggccct cccctaaccg ctactaacat cgttggcctt 240

cgtcaccacc gcactaacaa tgggccacta attttatatt tttcgtgata ttaatcctat 300

taaaaatgaa aatatctcct taattaatta agcttgtcag gaccgatgta aacaaaatta 360

atgtaaatgg acgcgccttt gacttgccaa caaactcgaa acgacgtttc ctccgtctga 420

taactatctc gcgacctccg acgacatccg acggtgcaga tcgggtcccg gtcaaccatc 480

cagatccacc cgattttctc ccggccctcg acaactccca ccaccacctc tttcctccct 540

ctttccttcc ttcctttctc accagatttt cccgagaaaa tcacagagag agaaagaaaa 600

acctcaccgc ctagagagag aaagagagaa agagggaaga gagagagaga gag 653

105

342

›DNA

Eucalyptus grandis

105

agttgggtaa ccagtctaat tctggtaaca ttgcattgta cttgatctca ataaaagcat 60

ataggataga attatcttct gtcttgatgg tttccatgag aaccaactgc tatactatga 120

aaaatatcaa tgttccacaa tatttttggg acaagggaac acaagattga gtcaacagtt 180

caggacccca gaaaaattat tcctgagttc gcagattatt ttcctaaaag tgaacaattc 240

aagaccctag ccaaatcatt cccaagtcca agttatgtga cactgcgact aacaaggcaa 300

gttggaagaa accatcaatc aatctcctag ttaatgacag tc 342

106

342

›DNA

Eucalyptus grandis

106

ggtctggaag ctcatctctc caatttggtg aagattacag ctataagagg tagctatgat 60

gtgctggcca aatgcaagtg atgaaatacg tggaccacca agtgcgaagg cattcgaaga 120

acgagggtcg aatttatagt gggcgaagga tgattaggtg gaatatgaca agaaaatagg 180

tttgaaagag aaataaatat tatgatagtg aagggtcttc acatggttag tttgatctgt 240

ccgagggtgt ccacccttgt ctgatccgca attgctcttg gtcgtgctga attttagagt 300

gtagccaaag taagaatttt cctttcactg tccggacatt tc 342

107

948

›DNA

Eucalyptus grandis

107

ctgacaaatg caaatatcta aaaccattgg ttgtttggtg cttgcaagtc tggattaccc 60

cactttatgt ttcacctttc aataatgaat aacaaggtac tcgggaaaaa aaggaaaggg 120

aaattcgcac aaccaaagtt gctatgcaga agtcaactca atcctaatca agttgatgag 180

agtgttgggc cctattttct gcagcaaaca tgaatctcga ttcatctccc tcgcaaaaga 240

taaggaagct gcaaaagctt tcctcctaag tttgttggca agcaaattga ttttgtacca 300

gaaataaata caaagtgaaa cccaagcaat cacgcatggc ctgatttgtg ccatgtccat 360

ttgatctccc tctactattt ttcctgcttt ctcaagcaaa ctagttgctg taacagtgaa 420

tgatcccccg gctctctctc tctctctctc tctctctctc catttattcc atccatgttt 480

ttgcttttcg cacaacactt atcattgagg tgctaactac tgaattcccc taactaaaaa 540

ttggaacctc tcacctaatt tcattttctc ccactttgat gagcaccact ctctttccca 600

gatttcaaat aaattgccac tctctccctc ctctttcctc acacaaccaa aagccttctt 660

caagtaccac ttcttcactg tcctctcttc acaatccccc tcttaccaag agcaaagcaa 720

aaaacatgat gaagagactg tcatttctgc tcctactggt cctgctcttc caatgctcta 780

ccaccttggc tcagcctgcg gccgccccag ctccgcctgt gatagccccg gctgcacctg 840

ctacgcctgc cttaggcccg gctcctcctg tcttaggccc agctcctgca ggcccaaccg 900

acatcacgaa ggtcctcaag aaggtgagcc aatttacggt gctgctca 948

108

362

›DNA

Eucalyptus grandis

108

ccatcactca taatcaacaa ggatatctca tcatgtcttc caaccaaatt aaaccccaga 60

catctctaaa gcagtatgga aaagaaaaca gtccggaagt ctctagctca aaaactgtaa 120

ccccgaccta attccggttg tctctgatta catcaattct tatgtcttaa cactccattc 180

gcacctccac aataaataga tcggcccttc atctcccctt accatcgaat ccaatcccaa 240

aaacacttgc tcagacacca tcaaatcctt cgcaaagtct ttttcttaca aaaaacaaac 300

gaaagcaacc atgaagcacc agttcattgt tctggctctc ttattcctca tcaacacagc 360

cc 362

109

326

›DNA

Eucalyptus grandis

109

aaaaattaca atcaatggtt atcaatggat gttacaaagg gaggttacat atagaggtta 60

taaaagaggg ttacaaatag atgtctcaaa caattaccaa gcggttagat tgactccact 120

attttgacgg ttctcttgac tttactatct caacgattac tttatttcat catgttgacg 180

gttgcatcca tgattgttga cttcactttt tgtcgattcc ttcaagctgc tgattcttca 240

agttgccaat aattttattc ataaatgacg aaactctagc ctcatccatt aagtttgtta 300

cttgtccaca ataattaaat tcggta 326

110

296

›DNA

Pinus radiata

110

tgctcccggt catgacaccg ccattctcgc tcttcatttc caattcaaat cacttggttg 60

ttgttcacac acacgggtct ttatatgacg agtgctgctg cgattataaa tagacggggc 120

aattacaaca aaaactcaca gcatttgaag gaagttggag tggtagagtg agaaatacac 180

agcctaatct gaaggaagtt cgagtaatag agtgagaaat ggatcttctt ctcctcatga 240

tgatgcttgt gatgatgggt gtagcaatgc ctactcattc tcaacaaatc actagt 296

111

723

›DNA

Pinus radiata

111

cgttttacgc gggaacaatg aaaacagtac aatcgaaaga gtcaagtcgt gaggttcatt 60

tcgatgaagt tcccagagat tgtctcgttc aacgtttcct cttttttcgg gtcaagtcgg 120

gtacagaaga ccactttctt tacgcggtca agacaccgcc attctcgggt caagtcggga 180

ggtccctcct gctcttcctt tttccaaatc cgtaaaattt acagattttt ttaatgtatg 240

aagcccactt tctttatgcg gttgctccca gtcaagacac cgccattgtt gttcacacgc 300

acgggtcttt atatgacgag tgctgctgcg attataaata gacggggcaa ttacaacaaa 360

aactcacagc atttgaagga agttggagtg gtagagtgag aaatcatttg aagggagttg 420

gagtggtaga gtgagaaatc atttgaaggg agttgagaaa tatattggga atctctcttt 480

tttgcagcaa ttagatcttt cctttaatgc tttgagtggg agaattccga cagagtttgg 540

gaacctctct cttttgcggc aataagttgg agtggtagtt ggagtggtag agtgagaaat 600

acacagccta atctgaagga agttggagtg atagagtgag aaatggatcg tcttcttctc 660

ttcatgttga tgcttgtgat gatgggtgta gcaatgccta ctcattctca acaaatcact 720

agt 723

112

1301

›DNA

Pinus radiata

112

actatagggc acgcgtggtc gacggccctg gctggtagcg acagagctgg ttcagtgacc 60

gttcgtgatt agccgcagta aaacaaaacc ctaaccgtaa ccctttcgcg cagattccat 120

ccttccccgt cctaccaaaa cccaaacttc ttgcccgaac tcaccttcta tgtattaatt 180

cttattatta tttaataata ataaatagtt aaacataaat ttataaatta attaattttt 240

atgattttta ttttagttta aaaatgtgac attgttatag attaatgctt atgaacgttt 300

attggccata attaccctaa ttaattataa ttaaaatata tagttataat taaaaaattg 360

tatattttat aaattgaatt aagaatttct gatgatattt catcattcaa ttccatctta 420

tcaaagttag agggaatagt taaccatgta ctagatctat tcatagctaa catttgccaa 480

gttcgtacta ggagacttgg atttttttta aaacataatt ttggcagtaa aaagtgaatt 540

ctattgtttt gaaaacaaaa caaaatacag gaagcgtgat tgtggggttg ttgttgaact 600

tgcccgggca aaagaagaat gattagcggt agaggagtta gtagttacgt tcaactaaat 660

gcgtgactaa attatttatc ctccgccatg gaagcaggtg attcacacac aacttgctgc 720

acacattgct ctcaaacctt tcctataaat atccgtagca ggggctgcga tgatacacaa 780

cgcatttaat caaactactt tgattacttt ctgtgggttc tactttcttt gaatagtcag 840

ttctgctgtt tttagaagat ttataagaat ggccaaaatt caggtatcaa acgggaacgt 900

cgtggtggtg gctgcgatgt tatttatggt ggtggtggcc atgcaaaacc atcacgtcgc 960

cgcccaaagt gctgactgcg ccgccaccgc ggagtccctg agcccctgcg cctcggcggt 1020

gggaaacaac ccacaggatc ccactcccga atgctgtgct gttcttcaga ccgctaatgt 1080

cgactgcatc tgcgccctcg tccaatcaac catgcaattg ccttccgaat gcggtcttga 1140

gactcctcag tgcccaagcg actagggtct caagaccgtg actgagtgct ggtttcagag 1200

acagtagaca ttctgcctaa taaatgattg tatgagagct tttatatatg gaattgctca 1260

tatgctttcc tagatatgaa attattaaat tccatatgct t 1301

113

3070

›DNA

Eucalyptus grandis

113

agcaccatca gcaaaaaata gatgggatag agtgggacac cacctgttca gtttgattcc 60

cttgagatga cctacagtga tagcttgatg aataagatgg gataatagat tcaccagagg 120

gataaaaagg tagggagata ggggatctcc ccgtctgatg cctcgggtag gttgaaaata 180

aggcaaaagt tcgccgttga atttgacagc aaaagacacc gtcgttatgc attgcatgat 240

ccattgtacc catgtagggt gaaatcctag agtgaggaga tagtccttta gaaagtccca 300

ttccacccta tcataggctt tctgcatatc cattttaaga acagcccgga attgacgtct 360

acattttctg actttaaatt gatgtagaac ctcttagact attaaaatat tgtcctgaat 420

ttgacgtcca ctgacaaaag cgctttgctc ctggaaaata agtacaggca ggtagggctt 480

aaggcgattg gcaatcacct tagaaatgat cttatatgcg taattacaaa gactgatggg 540

gcggtattgg tctaattgtt caggatgtgg taccttgggt attagggcta tgatggttcg 600

attgagattc ggtggtatga tgccagaatt aaaaaagtgc tgcactgatg agaatagttc 660

atcctggagt atatcccaat gatgctggta gaagagtcca ttcaagccat ctggaccggg 720

ggccttggta agtcccagtt ggaaagtagc ctctctaact tccttcttgg taacaggagc 780

tattagggac atattcatct cattagtaac aacctaagga cactggttca gaataggcaa 840

gtagtctcga tgtcccactg tctgaaatag atgtgaaaag taacctatcg tcatcatctt 900

caaaatttca ggatcgcgca cccaagcttg attgtcatcc tgcaacatac taatcttgtt 960

tcgttgttgt ctttgtatag ttgttgcatg aaaaaattta gtatttttgt ccccccagct 1020

gagccattta attcgagagc acatcgccca aaattattct tcttgctgcc ataactgtcg 1080

aattttctct tttaggtaag taaccaatga tgcgccatgt tgacaaaaag gctgattagt 1140

atgatcttgg agttgttggt gcaaatttgc aagctgacga tggcccctca gggaaattaa 1200

ggcgccaacc cagattgcaa agagcacaaa gagcacgacc caacctttcc ttaacaagat 1260

catcaccaga tcggccagta agggtaatat taatttaaca aatagctctt gtaccgggaa 1320

ctccgtattt ctctcacttc cataaacccc tgattaattt ggtgggaaag cgacagccaa 1380

cccacaaaag gtcagatgtc atcccacgag agagagagag agagagagag agagagagtt 1440

ttctctctat attctggttc accggttgga gtcaatggca tgcgtgacga atgtacatat 1500

tggtgtaggg tccaatattt tgcgggaggg ttggtgaacc gcaaagttcc tatatatcga 1560

acctccacca ccatacctca cttcaatccc caccatttat ccgttttatt tcctctgctt 1620

tcctttgctc gagtctcgcg gaagagagag aagagaggag aggagagaat gggttcgacc 1680

ggctccgaga cccagatgac cccgacccaa gtctcggacg acgaggcgaa cctcttcgcc 1740

atgcagctgg cgagcgcctc cgtgctcccc atggtcctaa aggccgccat cgagatcgac 1800

ctcctcgaga tcatggccaa ggacgggccg ggcgcgttcc tctccacggg ggaaatcgcg 1860

gcacagctcc cgacccagaa ccccgaggca cccgtcatgc tcgaccggat cttccggctg 1920

ctggccagct actccgtgct cacgtgcacc ctccgcgacc tccccgatgg caaggtcgag 1980

cggctctacg gcttagcgcc ggtgtgcaag ttcttggtca agaacgagga cggggtctcc 2040

atcgccgcac tcaacttgat gaaccaggac aaaatcctca tggaaagctg gtattacctg 2100

aaagatgcgg tccttgaagg cggaatccca ttcaacaagg cgtacgggat gaccgcgttc 2160

gagtatcatg gcaccgaccc gcgattcaac aagatcttta accggggaat gtctgatcac 2220

tccaccatta ctatgaagaa gatactggaa acatacaagg gcttcgaggg cctcgagacc 2280

gtggtcgatg tcggaggcgg cactggggcc gtgctcagca tgatcgttgc caaataccca 2340

tcaatgaaag ggatcaactt cgaccgcccc aacggattga agacgcccca ccccttcctg 2400

gtgtcaagca cgtcggaggc gacatgttcg tcagcgttcc aaagggagat gccattttca 2460

tgaagtggat atgccatgac tggagtgacg accattgcgc gaagttcctc aagaactgct 2520

acgatgcgct tcccaacaat ggaaaggtga tcgttgcaga gtgcgtactc cctgtgtacc 2580

cagacacgag cctagcgacc aagaatgtga tccacatcga ctgcatcatg ttggcccaca 2640

acccaggcgg gaaagagagg acacagaagg agttcgaggc attggccaaa ggggccggat 2700

ttcagggctt ccaagtcatg tgctgcgctt tcggcactca cgtcatggag ttcctgaaga 2760

ccgcttgatc tgctcctctg tggtgatgtt catggttctt ggatttgaaa ggtcgtgaag 2820

gagccctttt ctcacagttg gcttcggcat accaagttct tctcataaaa ggaaacaata 2880

agaagcgact gtatgatggc gcaagtggaa gttacaagat ttgttgtttt atgtctataa 2940

agttttgagt cttctgcata ctgatttcac agaatgtgta acgaaacggc gtatatggat 3000

gtgcctgaat gatggaaatt gtgatattct gtcttctttt tcagtaaatc acttcgaaca 3060

aaaaaaaaaa 3070

114

1227

›DNA

Pinus radiata

114

aaatttcaag aggaagagat taattctttt aatttataaa attatataat aaaatattta 60

tatttaattt agatgataag tttatgaggt gtagaataga tagtgatggg tgtattattg 120

agttattccc ctaatgtgga gacaattgat tagaagttct atgagaaaaa tccaatcatg 180

ttaaagtgac ccctaatgtg aagacaattg attagaaatt ctatgaaaaa aatccaatca 240

tattaaaagt ccaattgatt agcaatttta tgagaaaaat ccaattatgt taaaagtcac 300

tgagtgtggc cgaaattgtg accgaaattg aatgcaataa ccgagggttt ttcaaaccaa 360

ggttaagcct ctcatcattg gggtgtgtat gaaaatgtaa tgggcatcga taacctttta 420

ttacaacttc acgaaaattg cctctattca atgggtgtgg atgaaaatgt aagtgcgcat 480

cgataatgga aagcgatatg cagcaaaatc aataaacctg acttcccatg tgagtgatga 540

tttgatcgta caactgatgg tgtgaagtta ctttcagctt caccttcggg cataatcagg 600

gaagtagggc caagtttgct tagtatcact ctaatcccca acaccgtgat tactatcttc 660

atcaacaatg gccaccttcg tcattacttt aactggtggg atacagctac tttacaactg 720

taaatttgtt gaggcagcct atcctcagcc tatacatact aattattgca gctcgattag 780

gtatctgctg tgagaatagc tgtgtatctc tgcgctggtt gcaggatcca agttcctctc 840

agagccctcc atggaagcgc agtcagtttc agttgttgag cagcgccccc atgccctact 900

attttcattt ccgttacagg gccacatcaa gcctttcatg aacttggcca agattttgtc 960

cagccggggc ttctatgtca cttttgccag taccgaattt gttgtaaagc gcctcgcaga 1020

atgtggtgaa agtatcgccc atcgtgattc gatggtgtgc agcgagaacg atgatgtatg 1080

taacataaaa tttgaaacag tgcccgacgg actgcctccc caccacgatc gcagtactca 1140

gaatcttgcg gagctcttcc aatccatgga agagaacgct catattcact tccacaagtt 1200

gatggagaag ctccagaatc ttcggga 1227

115

1169

›DNA

Eucalyptus grandis

115

ttcattatat gattattacg tcataatgat cgatttctag aaatttggag acatatgtaa 60

attcaggagg aatttcaaga aacgcgcgtt actttgaaag ggtctcgagt caaagtgctc 120

aaattgagag ggagaatttt agaacaaaat cagatttgga gaatacatgc cattttaggg 180

ggattttggg gatttcgcat atggcgtcgc gtcgtcggcg ccttcttctt tacagattgt 240

atcctcccat taaccgcgtg gacctgcata gggcacgcgt ggtcgacggc ccgggctggt 300

ttcattatat gattattacg tcataatgat cgatttctag aaatttggag acatatgtaa 360

attcaggagg aatttcaaga aacgcgcgtt actttgaaag ggtctcgagt caaagtgctc 420

aaattgagag ggagaatttt agaacaaaat cagatttgga gaatacatgc cattttaggg 480

ggattttggg gatttcgcat atggcgtcgc gtcgtcggcg ccttcttctt tacagattgt 540

atcctcccat taaccgcgtg gacctgcact gtaaccccga aacggtgggg gccaatttcg 600

tctttccgcc tcctccactc agcttcgtgg aagattaaaa tcctcaccgt ccgtgcaaac 660

gccacgtggc gcgttagttt gcgcgtggaa aggtcctcac gaaccgtaaa gggcaaaaaa 720

aagggaaaat aaaaaaggag gaggaggagg gaggaggaag aattgtccga ttgaaaataa 780

gagtgcggtg gtgtggtgtg ggtagatctt gaattgaacg agctcaattc gcgtatttaa 840

acccgccccg cttcctcatt cttccttgtc catttcaact ctccctctct ccctctcttc 900

tgcccctcga tcgatccagc gatcttccta tttccggacg cggggagcag ctcctcttgt 960

cgaaggttct aaattagtgt ggagagatgg tgaagatctg ctgcattggt gctggctatg 1020

tcggcgggcc tactatggcc gtgattgctc tcaagtgccc gtcagtagaa gttgcggtcg 1080

ttgatatttc tgtctctcgc atacaagcct ggaacagcga acagctccct atctatgaac 1140

caggccttga tgcggtggtg aagcaatgc 1169

116

947

›DNA

Eucalyptus grandis

116

ggtctggaag ctcatctctc caatttggtg aagattacag ctataagagg tagctatgat 60

gtgctggcca aatgcaagtg atgaaatacg tggaccacca agtgcgaagg cattcgaaga 120

acgagggtcg aatttatagt gggcgaagga tgattaggtg gaatatgaca agaaaatagg 180

tttgaaagag aaataaatat tatgatagtg aagggtcttc acatggttag tttgatctgt 240

ccgagggtgt ccacccttgt ctgatccgca attgctcttg gtcgtgctga attttagagt 300

gtagccaaag taagaatttt cctttcactg tccggacatt tcgattgcta catggaccat 360

cccgtgtcta cccattcttg agaaccttcg agtggaaagc atgaataacc caccttgtac 420

tatataggtt gccgaatatg cctagggcgc gaccatcatt gagacggagt tggggtgctc 480

cgctcggttc accaccacca ccaccaccac caccaccacc accaccattg ggcactgata 540

tagcgactcc accactaccc caaccgaggt tggcaaactc tagattgtac atgggatata 600

tcggagtagt tgaacatgat cagatcaatg gtagtggtta agactctaga aattattgaa 660

gcaatatgtt aaatcagata cgtgtgagaa agtgacttac taattgctat ggctttcatg 720

atacttaaac ttcaatgaat tggtaatgtg aagagcaatg tgatctccac aaatactact 780

agaaggccaa gtccttttct ttatgccgaa gtcctaaagt ttaatatttc aactctacct 840

atatcaaatt tgtatgcaaa ttgcataatc gcactgattt ctatggtttt attaatctag 900

ataagaactc tctccaagac attaactaat taagattgac cccattt 947

117

1766

›DNA

Eucalyptus grandis

117

atccagatcc ctacgaactg gattcacaca gtcactgctg taagctctgg ttttttttag 60

cttaggaagc aggttatgat caaacatgat taaaccatcg cgtgttcgcc agccatcaga 120

aatggaaagg caaatgttgt tatagtgatg gacagatcat gctgagatga ttgattatga 180

atcttactga tgactgtcat ttatgttatc gcactctgtg tgtgtgggtg tgtgtaatga 240

gtaatatcaa attaaccaga cgataggtgt tgaagattag ctgttgggcc accgtggcga 300

aaggtgtctt atacaagcca tcggcagtga cgcagaactg tagagaaccg ctgtaacaag 360

tcttcgaatg cattctttta atgtacagca cgacatgaag ggggttcgag tgtagcgaac 420

agttcgtgcg agaaagatca ttttcaatag cataaaagag tctgctttct gctgcaaaca 480

tggaaagaac ttacatttca atcattgagg agaagattat aacaaatcct aaatggttga 540

gattttagtt agtccattcg aactaaagtg gcgaagatgt cagtttttca agtggatgat 600

atttctcatg tatgttccgc agaggcaatc accttgtttg taactagaca tctagagaac 660

ctaacaagga ttgatggggg tgaggtgaaa tgtctgtttc ctctttaata tggatccagc 720

gatgccttac agagcggatg gatggcactg gcaagtctta atccttagct cgaatgtttg 780

attggtaaca gatgcctttt ctttcttttc aatcacagct gacaaatgca aatatctaaa 840

accattggtt gtttggtgct tgcaagtctg gattacccca ctttatgttt cacctttcaa 900

taatgaataa caaggtactc gggaaaaaaa ggaaagggaa attcgcacaa ccaaagttgc 960

tatgcagaag tcaactcaat cctaatcaag ctgatgagag tgttgggccc tattttctgc 1020

agcaaacatg aatctcgatt catctccctc gcaaaagata aggaagctgc aaaagctttc 1080

ctcctaagtt tgttggcaag caaattgatt ttgtaccaga aataaataca aagtgaaacc 1140

caagcaatca cgcatggcct gatttgtgcc atgtccattt gatctccctc tactattttt 1200

cctgctttct caagcaaact agttgctgta acagtgaatg atcccccggc tctccccctc 1260

tctctctctc tctctctcca tttattccat ccatgttttt gcttttcgca caacacttat 1320

cattgaggtg ctaactactg aattccccta actaaaaatt ggaacctctc gcctaatttc 1380

attttctccc actttgatga gcaccactct ctttcccaga tttcaaataa attgccactc 1440

tctccctcct ctttcctcac acaaccaaaa gccttcttca agtaccactt cttcactgtc 1500

ctctcttcac aatccccctc ttaccaagag caaagcaaaa aacatgatga agagactgtc 1560

atttctgctc ctactggtcc tgctcttcca atgctctacc accttggctc agcctgcggc 1620

cgccccagct ccgcctgtga tagccccggc tgcacctgct acgcctgcct taggcccggc 1680

tcctcctgtc ttaggcccag ctcctgcagg cccaaccgac atcacgaagg tcctcaagaa 1740

ggtgagccaa tttacggtgc tgctca 1766

118

1928

›DNA

Eucalyptus grandis

118

ctggttccac gtcaagcacc tcctggagtg acaaggaaat gccaccggaa aatcaagatt 60

gctgttttag gctcactttt ttcctgagct aagtgggtcg catttcaaga aacagtagaa 120

gttacgttct ccatggaaac tcgaaaggat aaaaattaag aaacggaagc tccatgagaa 180

cgatgggggt cagcatcact cctattgtat tgtgctctca ttatctctgg cctacttgag 240

aagtgatctg ggattcgcta ttagtgaaaa caatcgcagg ctaactaaga tcttttatgc 300

taatcatatg gagaaatatc cctcttaagg gaagcatatg agttttttct taggatgact 360

acgcttattc aaaacctatc atacacgtca tgccaataat acccacttgt tgttccttta 420

ctcaggatcc tcgatagcca atactaattg gcaagaacct tgagtaacaa gctgaggtat 480

acataggcct atcattcatt tactagactc gattgcaagc acacatgatg cacatttata 540

tcagcaatca gcaatcatat ttccgaaaat tgtctctcag agaaaaagag agagagagag 600

agtccatagt atgtcatagc caaaagaaaa attagcaaca agatctcgag gtattgttga 660

aaggtagggc aatatcaaga attccattgt aattaatgtg tctagacaac atctaagaaa 720

aaaaagtgaa agaaaagagc tatatagtta ataatattta tacatgttgg agataaactt 780

gagttagagg tttatgacct cctagattga ttaaacagac caaatagtag taatcagggc 840

acttcttaaa tctactaata tattgttcaa acatgacttt taacctatct tgattagaaa 900

tgagtgttca aagaaaacta atcatgcata tattttgtcg cccaatcacc ctagggtgga 960

aaaaaggcta tctactcaac aaatgctaaa attttacggc tacacgtggc cacagttgca 1020

gtacaattca tctcaaggaa ggactaaaac tgcaaagaga agaagactac ataggaaaaa 1080

ggaaaacaaa gaagccttga agtaaagagg agcataactc actcaactga gtgtgttcgc 1140

caatgtggca aagaaaaagc ctctaagatc ctcacaaatg gccacgtgga ctcacacggc 1200

accctataca agtactacta ctactacagg actatgccag aaggagaagt gttagcgtga 1260

gtaccacgtg cgcacgcaga atctaagcct agcaaaaact atgctgagtc aagcagctcc 1320

cccacccatg aagatagtac tgtaatgtga ctcttgacag cgaaaccaaa cagtactcca 1380

agagaaaagc caaagcagca aaaatggggc ccgcagcaag aacctctgac tcgacctgga 1440

cccaccaaga acaacagcca gccacaaaat aacgtaaaga ctttttgcgg ccactaactc 1500

ctcgacaagt ggcactgctt ggattccctt catcttgcct tcacttaacc cccaccctcc 1560

ctcacactgc attcacttca aacactcccc agtttcagag tttcattgag aaatatgttg 1620

aaggaagaca cgagtggcag cggcggcagc agcggcagcg gcagcggtgg taatagctgg 1680

gcacgtgtgt gtgacacttg ccgctcggca gcatgcaccg tgtactgccg tgccgacttg 1740

gcttacctat gctccagctg tgacgctcgt attcacgcag ccaccgtgtg gcctcgcgcc 1800

atgagcgcgt gtgggtgtgc gaagcgtgcg agcgcgcccc ggctgccttc ctctgcaagg 1860

ctgatgcagc atcactgtgc accgcctgcg atgcagacat acactcagcc aacccgcttg 1920

cgcgccgc 1928

119

602

›DNA

Eucalyptus grandis

119

attgggagga agtagagtgt gctgtgtgag attggtcgat gagctggctc ttgtggagat 60

ggcaagtgat tgtggcttct gtgatgcata tatataggca agggacgtga tgcggaggaa 120

gtatgtatca tcagcttata ataatgattg gtcagtttgt aagtgaatat taagggcctc 180

atgggtgttg gttcacggcc caaggcgggg cccactcacc gggggattta tcgtgtaagg 240

atacatccag ggtcagggtg tttggggaca cactttgcca tcttatgtgg gcatgatcag 300

attgagaaga atccgatcct tctttttcct aaaccattga acccaccatg agaatctttg 360

tttggaggga aaaataaaaa aatagattga gacgtattct aggagaggat agcaaaagaa 420

tgtgactttg tttgtttgtg tatcggattg atctaaggaa aaaagacact aaccgttcta 480

caattttcat acaactcttt catttaagca ccgtgacttc caaaaatcga tcatccttat 540

acggttggaa atcacacgtg gcattgctgt aaaagaaata gttgatgggt ctcattgaag 600

at 602

120

1326

›DNA

Pinus radiata

120

aaaaaaggga aacattatac caaattttat gatatctttc aacaacatac tcttctatat 60

atggtgcctc ctctgatgga cccttgtcaa ctttctcttt ttatgtgtaa tgcctcaaga 120

gcccccactc acaagataat atcttttcca taatataata tatattccta ttgaagcagt 180

cttttgatgt accgagtaca ctactcatgg tgaaggccgt gtcttgcagc ttttcccatg 240

gtttattttg aaagtaatag tactggacct catttgcaac gacacataat attcttactg 300

acgacacttt gtttgatttc ttatagaaaa atgcaaggtg gcacaaaaag atggaaagcc 360

cgacctatca agcatacgaa gggtcatgtt cacaccctct gaaatcttca gagtctcacc 420

ctatgttgga cgctaatcaa tgggatcacg ctgaaacata tcgtaaatga cgaatcaatc 480

aatcaatcat tgaaaaatat accagataac tcctacgatg gaggggatta tttgcgtacc 540

ctccgcgtgg gtgggcacat tgggcaggtc ctttggtaag tcttggagac agagtcacgt 600

ttccataatt gaagtggaca tttatgaatc tttcgaaagt tgtagaactc ttaattttcg 660

acggaatagt ttgacacgtt ttgtacgatc tggtttttcc ggggaacgcc aattttggtt 720

tctgaaggac agcatttaca atattgtctg tcgttgacca ggacagctgg ctcggaactc 780

gggtttccga tgcgcaggaa gcgcattgaa atgagaatat aatctagttc tacctgtgga 840

gctatcacaa aatactaaaa ctggtggaca tacctcttgt ctgttctcga aatcggccaa 900

aatgggaaag aagagggtag agctgaaacg cattcaaaac cctagcagtc gacatgctac 960

tttctctaaa cgcaagaatg gattgctaaa aaaggcgttc gagctttctg tcctctgtga 1020

tgctgaagtc gctctcatca ttttctctga aactggcaag atttacgaat ttgcgagcaa 1080

taacgatatg gcagcaattc tgggaaaata ccgagtacac gaagaaggca ctgaaacgtc 1140

cagtccaaca tcgcttcaaa acgtaaagta tcatgaatca gggcttgaga aattgcaaga 1200

gaagttgacc gctttgcaaa agaaggaaaa gaacttgatt ggtgaagact tggaggtatt 1260

aacaatgaaa gaactgcaac ggcttgaaaa acagttacaa attggcataa aaaggttagt 1320

gataga 1326

›Tables in the description — 12
TABLE 1 — Poly-
nucleotidePolypeptide
SEQ ID NO:SEQ ID NO:5′ UTRIDENTITY
1801-2064Super Ubiquitin coding region
and UTR
2—1-2064Super Ubiquitin promoter with
intron
3—1-1226Super Ubiquitin promoter without
intron
4—1-431Cell division control
5—1-167Xylogenesis - specific
6—1-6004-Coumarate-CoA Ligase (4CL)
7—1-591Cellulose synthase
8—1-4803′ end, cellulose synthase
20—1-3635′ end, cellulose synthase
9—1-259Leaf specific
10—1-251Leaf specific
11—1-248Leaf specific
12—1-654O-methyl transferase
13—1-396Root specific
14—1-763Root specific
22631-406Pollen coat protein
23—1-350Pollen allergen
24—1-49Pollen allergen
25641-284Pollen allergen
26651-77Auxin-induced protein
27—1-74Auxin-induced protein
28661-99Auxin-induced protein
29—1-927Flower specific
30—1-411Flower specific
31—1-178Flower specific
32—1-178Flower specific
33—1-178Flower specific
34671-805Ubiquitin
35681-81Glyceraldehyde-3-phosphate
dehydrogenase
36691-694Carbonic anhydrase
37—1-648Isoflavone reductase
38—1-288Isoflavone reductase
39—1-382Glyceraldehyde-3-phosphate
dehydrogenase
40701-343Bud specific
41—1-313Xylem-specific
42—1-713Xylem-specific
43—1-28Xylem-specific
44—1-35Xylem-specific
45711-180Meristem-specific
46721-238Senescence-like protein
47—1-91Senescence-like protein
48—1-91Senescence-like protein
49—1-809Pollen-specific
50—1-428Pollen-specific
51731-55Pollen-specific
52741-575Pollen-specific
53751-35Pollen-specific
54—1-335Nodulin homolog pollen specific
55—1-336Nodulin homolog pollen specific
56761-157Sucrose synthase
57771-446Sucrose synthase
58—1-326Sucrose synthase
59—1-311Flower specific
60781-694O-methyl transferase
61791-112Elongation factor A
62—1-420Elongation factor A
81——MIF homologue
82——MIF homologue
83——MIF homologue
84——MIF homologue
85——MIF homologue
86871-87MIF homologue
88—1-1156Chalcone synthase
89—1-2590Unknown flower specific
90—1-1172Unknown flower specific
91—1-446Sucrose synthase
92—1-2119Unknown xylem specific
93—1-2571Glyceraldehyde-3-Phosphate
dehydrogenase
94—1-1406Unknown pollen specific
95—1-2546Pinus radiata male-specific protein
(PrMALE1)
96—1-4726Pinus radiata male-specific protein
(PrMALE1)
97—1-635UDP glucose glycosyltransferase
98—1-468Elongation Factor A1
99—1-222Elongation Factor A1
100—1-410S-adenosylmethionine synthetase
101—1-482S-adenosylmethionine synthetase
102—1-230S-adenosylmethionine synthetase
103—1-596UDP glucose 6 dehydrogenase
104—1-653Hypothetical protein
105—1-342Laccase 1
106—1-342Laccase 1
106—1-948Arabinogalactan-like 1
108—I-362Arabinogalactan-like 2
109—1-326Arabinogaiactan like-2
110—1-296Root Receptor-like kinase
111—1-723Root Receptor-Iike kinase
112—1-1301Pinus radiata Lipid Transfer
Protein 2 (PrLTP2)
113—1-1668Caffeic acid O-methyltransferase
114—1-850UDP glucose glycosyltransferase
115—1-986UDP glucose 6 dehydrogenase
116—1-947Laccase 1
117—1-1766Arabinogalactan like-1
118—1-1614Constans
119—1-602Flowering Promoting Factor 1
(FPF1)
120—1-901Agamous
Complement3′ TCCTGG 5′
Reverse complement3′ GGTCCT 5′
Reverse sequence5′ CCAGGA 3′
120
13083DNAPinus radiata
5′UTR(1)...(2064)
intron(1196)...(2033)
CDS(2065)...(2751)
3′UTR(2755)...(3083)
22064DNAPinus radiata
5′UTR(1)...(2064)
intron(1196)...(2033)
4485DNAPinus radiata
5′UTR(1)...(431)
TATA_signal(350)...(356)
CAAT_signal(326)...(333)
5246DNAPinus radiata
5′UTR(1)...(167)
TATA_signal(185)...(191)
6600DNAPinus radiata
5′UTR(1)...(167)
TATA_signal(471)...(477)
CAAT_signal(444)...(451)
7591DNAEucalyptus grandis
5′UTR(1)...(591)
TATA_signal(432)...(437)
8480DNAEucalyptus grandis
5′UTR(1)...(480)
9308DNAEucalyptus grandis
5′UTR(1)...(259)
10300DNAEucalyptus grandis
5′UTR(1)...(251)
21839DNAPinus radiata
misc_feature(1)...(839)n = A,T,C or G

Claims as granted

14 claims

Log in to read the claims of this application.

Log in to unlock

Classifications

20 codes
IPC · International Patent Classification
Section A — Human necessities
  • A01H5/00
Section C — Chemistry; metallurgy
  • C12Q1/02
  • C12N15/09
  • C07K14/415
  • C12N9/00
  • C12Q1/68
  • C12N15/82
  • C12N5/10
  • C12N15/29
USPC · US Patent Classification
800/278435/69.1800/295435/419435/468800/298536/23.1800/287536/23.6435/320.1536/24.5

Claim changes

Soon
Coming soonHow the claims changed between publication and grant

See which claims were amended, added or cancelled during examination, with every added and removed word marked.

AmendedAddedCancelledUnchanged

The published claims of this application are not paired with the granted ones in what we hold.

File wrapper

⤢ drag to zoomJul 2000Jan 2001Jul 2001Jan 2002Jul 2002Jan 2003Jul 2003USPTOApplicantRestriction requirementNon-final rejectionResponse after non-finalResponse after final
USPTOApplicanthover for detail · click to open
Pendency
3.1 y
1,127 days filing → grant
Office actions
2
after a restriction
Responses
2
no RCE
Examiner
Elizabeth F. McElwain
art unit 1638 · TC 1600
Citations: 47 back · 22 forward

See the full prosecution history — every USPTO and applicant action on this file, in order.

Log in to unlock

Documents

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 unlock

Chain of title

⤢ drag to zoom20002002200420062008201020122014201620182020Owner 5Owner 6Owner 7liens, releases & corrections
TitleLienhover for detail · click to open

See the full assignment history — every owner this patent has passed through, with recordation dates and reel/frame numbers.

Log in to unlock