BMC Genomics BioMed Central

Putative cold acclimation pathways in Arabidopsis thaliana identified by a combined analysis of mRNA co-expression patterns, promoter motifs and transcription factors Abstract Background: With the advent of microarray technology, it has become feasible to identify virtually all genes in an organism that are induced by developmental or environmental changes. However, relying solely on gene expression data may be of limited value if the aim is to infer the underlying genetic networks. Development of computational methods to combine microarray data with other information sources is therefore necessary. Here we describe one such method.


Background
The potato (Solanum tuberosum) tuber is a specialized underground storage organ that develops from modified stems termed stolons. Although the tuber is primarily composed of starch, it also contains high levels of proteins and due to its importance as a food source, a prime focus in potato research has been tuber quality [1][2][3][4][5][6]. Another key focus in potato research is disease resistance as potato is susceptible to several pathogens including Phytophthora infestans, the causal agent of late blight of potato. Molecular and genomic approaches, coupled with initial genetic mapping data, have identified resistance genes in potato against this pathogen [7][8][9][10][11] including a potentially viable commercial form of resistance to late blight conferred by the RB gene identified in the wild potato species, Solanum bulbocastanum, which can confer resistance to a wide range of P. infestans isolates [10].
Genomic resources for potato have been developed including Expressed Sequence Tag (ESTs; [12][13][14]), bacterial artificial chromosome (BAC) clone libraries [15,16], microarray platforms [2,17], and a dense genetic map [18]. These resources have been utilized in studies on potato physiology, development, responses to abiotic and biotic stress, polyploidy, comparative genomics as well as enhancement of genetic maps [2,17,[19][20][21][22][23][24][25][26]. The potato genome is reported to be 798-931 Mb [27] and with the availability of improved sequencing technologies, coupled with decreased fiscal constraints on genome sequencing, an international consortium to sequence the potato genome has been established [28]. The Potato Genome Sequencing Consortium (PGSC) is focused on generating an initial draft sequence of the potato genome using a BAC-by-BAC approach followed by a finishing phase. The PGSC is enabled by the availability of two resources, a dense genetic map for potato [18] and an anchored Amplified Fragment Length Polymorphism-fingerprinted BAC library [28].
Collectively, the Solanaceae family is one of the world's most important vegetable families as species are grown for their tubers (potato), fruits (tomato, pepper, eggplant), leaves (tobacco), and ornamental features (petunia, Nicotiana species). In 2006 in the U.S., potato production was valued at $3.2 billion with tomato, tobacco, and pepper production valued at $1.6 billion, $1.2 billion, and $686 million, respectively [29]. While the cultivated species have been bred for these diverse agronomic traits, genome sequence analysis has indicated that these species share to a large extent not only genes [30] but also gene order (synteny) between their genomes [31][32][33][34][35]. While major classes of repetitive sequences are conserved among some Solanaceae species [36,37], lineage-specific repetitive sequences have been reported, suggesting divergence of this fraction of the genome has occurred through evolution [36][37][38][39][40][41][42]. With the availability of large genomic datasets for two Solanaceae species, tomato and potato, the extent of sequence conservation as well as synteny can be addressed in a more robust manner. In this study, we report on the generation of the first large set of genomic sequences from the potato genome along with characterization of these sequences with respect not only to the potato genome landscape but also in a comparative manner with genome sequences from tomato. We further compared our potato genome sequences with sequences from the collective Solanaceae transcriptome to determine the extent to which available solanaceous sequences can be used to cross-annotate the potato genome.

Results and Discussion
Characteristics of the potato genome A total of 77,568 potato BACs were end sequenced from both ends resulting in 155,130 total sequences. For low quality and vector sequences, 140,259 sequences were generated with an average read length (after trimming) of 621 nucleotides representing a total of 87.14 Mb of potato genome sequence (Table 1). The average GC content of the potato BAC end sequence (BES) dataset was 35.6%, comparable to that of tomato BES dataset (36.2%) and the whole genome sequences of Arabidopsis (36.0%; [43]), poplar (33.7%; [44]), and grape (34.6%; [45] ( Table 1)). Not surprisingly, the potato genome GC con- Of the total 18 BACs in phase 2 and 3, seven BACs were selected randomly from the potato genome including chromosome 6. The other 11 BACs were identified as putatively syntenic with tomato contigs generated from the Tomato Genome Initiative (Additional Data File 2). Genes were annotated on all of the potato BACs using a semi-automated annotation pipeline; a total of 287 genes were annotated within the 18 potato BACs ( Figure 1). The numbers of genes annotated as encoding "known"/"putative", "expressed," and "hypothetical" proteins are 160, 21, and 106, respectively. Using the same annotation pipeline, 221 genes (139 known/putative, 17 expressed, and 65 hypothetical genes) were annotated within the tomato BACs/contigs. Overall, the length of genes, exons, and introns of syntenic potato BACs were similar to that observed in the syntenic tomato BACs/contigs ( Table 2). While GC content and exon/intron length were similar between syntenic tomato/potato BACs and randomly selected potato BACs, the average gene in randomly selected BACs had one less exon per gene and consequently were shorter (2.4 kb vs 3.1 kb). The Tomato Genome Initiative [48] is focused on the euchromatic region of the tomato genome which is highly enriched in genes in comparison to the whole genome. As a consequence, syntenic potato/tomato BACs have an increased gene density relative to random BACs. Even with data from a limited number of BACs sampled, this skew in gene density and repetitive sequence content is discernible and is illustrated on potato chromosome 6 in Figure 2.
The BACs from the euchromatic arms show a higher gene density and a lower repetitive sequence content compared to those BACs in the heterochromatin containing centromeric region estimated to be at bin 16-17 (Iovene and Jiang, unpubl.)

Sequence level conservation within the Solanaceae and its use in annotation of the potato genome
Representation of the respective transcriptome is variable among the set of 13 Solanaceae Transcript Assemblies used in this study [49]; sequences (Transcript Assemblies and singletons) ranged from 716 in the S. lycopersicum × S. pimpinellifolium Transcript Assembly to 81,072 sequences in the S. tuberosum (potato) Transcript Assembly (Additional Data File 3). Of the combined 251,274 Solanaceae Transcript assemblies and singletons, over half of the sequences are derived from potato or tomato reflecting the emphasis in EST sequencing for these two crop species. Within the 287 potato genes annotated in this study,  (Table 3). Moreover, 71 potato genes have support from at least three other Solanaceae species and as shown in Figure 3, the selected gene model (AC212552.18) is supported by transcripts from 10 different solanaceous species. Based on sequence similarity to annotated proteins, this gene model encodes a putative calcium-binding EF hand family protein. Clearly, annotation of the potato genome can be greatly enhanced by inclusion of not only cognate S. tuberosum transcripts, but also transcripts from other solanaceous species.

Synteny between potato and tomato
Previous studies with the Solanaceae [31,33,34,50,51] identified synteny between a number of Solanaceae species including potato and tomato. These studies utilized genetic markers and showed, albeit at a low resolution, conservation of gene order between potato and tomato.
With the pending availability of the tomato genome sequence, we were interested in determining the extent of synteny between tomato and potato to assess whether tomato genome sequences can be used 1) to identify syntenic potato BACs for the potato genome sequencing initiative, 2) to provide contig order and orientation information for potato BACs sequenced to draft level, and 3) to provide as a "reference genome" for structural annotation of the potato genome.
In total, we selected 11 potato BACs that were putatively syntenic with tomato; eight candidate syntenic BACs were identified using BES analyses (Set I, see Methods) and Gene and repetitive sequence content in the candidate syntenic potato AC212552 Figure 1 Gene and repetitive sequence content in the candidate syntenic potato AC212552. Shown in the figure is ~130 kb of sequence with loci, gene models, and repetitive sequences annotated. For the Loci track, genes encoding hypothetical proteins are colored in red, genes encoding expressed proteins in yellow, and genes encoding known/putative proteins colored in green. Gene models were generated using the annotation pipeline described in the Materials. Repetitive sequences are shown on the bottom track with retrotransposons colored brown, transposons colored orange, miniature inverted-repeat transposable elements colored green, centromeric-related sequences colored blue, telomeric-related sequences colored purple, rDNA sequences colored pink, and unclassified repetitive sequences colored yellow.
three candidate syntenic potato BACs were selected using gene model analyses (Set II, see Methods; Additional Data File 1). The tomato BACs, identified as syntenic to potato BACs, were downloaded either from Genbank or SGN [52] and 14 overlapping tomato BACs were merged into 6 contigs (Additional Data File 1). A total of 1.69 Mb of potato sequence and 2.2 Mb of tomato sequence were used to determine the extent of synteny present between these two Solanum species. To assess the sequence similar-ity at the nucleotide level, we used the program NUCMER with the default settings (i.e., NUCMER command-line options "--minmatch 20 --maxgap 90 --mincluster 65 -breaklen 200") [53], to align these syntenic blocks. Collectively, 515 kb of the tomato and potato sequence could be aligned (Table 4) with high levels of nucleotide identity (89-91%). Although dependent on the portions of the genome the respective BACs represented, up to 73% coverage between syntenic clones was observed. At the gene level, alignment of the potato and tomato protein sequences revealed a high degree of synteny; a total of 98 annotated proteins within the contiguous regions were identified as syntenic. This synteny was sufficient to enable ordering and orientation of contigs within HTG phase 1 potato BACs (Additional Data File 1). However, synteny was not absolute between tomato and potato. In the absence of large insertions/deletions, syntenic regions should have similar length and the large difference in length between some potato and corresponding tomato syntenic regions suggested the existence a bulk insertion or deletion event ( Table 4). As shown in Figure 4, an 86 kb insertion/deletion is apparent between potato BAC AC212316 and tomato contig 29. This insertion involves not only repetitive sequences but also non-transposable element-related genes. In addition, micro-scale inversions were observed as shown in Figure 5.  Table 5; Additional Data File 4). In both potato and tomato, more than half of the repeats identified (19.28% and 25.64%, respectively) fell into the unclassified repeat class, most likely due to the lack of characterization of the Solanum repeats. Retrotransposon sequences were the most abundant in both potato and tomato BES datasets (11.77% and 14.54%, respectively). However, while there were more Ty3-gypsy type retrotransposons than Ty1-copia retrotransposons (4.60% vs. 1.82%) in the potato BES dataset, the margin between these two types in tomato BES dataset was much smaller (Ty3-gypsy 4.61% vs. Ty1-copia 3.75%). Exami-Distribution of gene density and repetitive sequences along potato chromosome 6 Figure 2 Distribution of gene density and repetitive sequences along potato chromosome 6. Bins are plotted on the xaxis as reported by van Os [18] with gene density (gene per kb) and repetitive sequences plotted on the y-axes. For BACs reported in a bin range, the average of the distance between the two bins was used. The centromere is located near bins 16-17 (Iovene and Jiang, unpub.). nation of individual libraries within the potato and tomato BES datasets (Additional Data File 4) indicated that this was not due to an over-representation of Ty1copia elements in tomato in the MboI library but rather due to a decreased representation of Ty1-copia elements in both the potato HindIII and EcoRI datasets compared to tomato. Overall, the percent of transposon sequences found in potato versus tomato BES dataset was comparable (1.32% vs. 1.39%). This is due to the similar represen-tation of unclassified transposons in both tomato and potato. Representation of Ac/Ds transposons was skewed between potato and tomato (0.13% vs 0.05%), which was attributable to a much higher representation of Ac/Ds elements in the potato HindIII library compared to the tomato HindIII library (Additional Data File 4).

Repetitive content of the potato genome
Surprisingly, there were nearly twice as many telomererelated repeat sequences identified in the potato BES data-Alignment of a selected potato gene model with transcripts from diverse solanaceous species set compared to that of the tomato BES (1.31% in potato vs 0.73% in tomato) with the major difference occurring in the telomere repeat representation (0.82% in potato, 0.01% in tomato, Table 5). While telomeric sequences are enriched in the telomeres, they can be found in centromeric and pericentromeric regions [38] and clearly, based on their abundance in the potato BES dataset, are preva-lent in the potato genome. We selected two potato BAC clones which contain telomeric repeats on both end sequences and six BAC clones in which only one of the ends contained the telomeric-repetitive sequences (Additional Data File 5). These BACs were used in fluorescent in situ hybridization (FISH) studies to assess where on the chromosomes these repetitive sequences localized. As Each syntenic region consists of a number of matches and gaps The Match Length and Average Identity columns represent the total length and the average sequence identity of the matches in each syntenic block. The length of the syntenic region (i.e., the sum of the lengths of the matches and gaps) and the percentage of the matches over the syntenic blocks in the potato BACs is shown in the columns "Synteny Length" and "Coverage", respectively. The Synteny Length Difference column represents the length difference between the potato and the corresponding tomato syntenic regions. A positive number means that the tomato syntenic region is longer than the corresponding potato syntenic region (i.e., there is an insertion in the tomato BAC or a deletion in the potato BAC) and vice versa. Columns Match Length, Average Identity, Synteny Length, Coverage and Synteny Length Difference are statistics summarized from the sequence alignments using the program NUCMER [53] at the nucleotide level. The No. Gene Pairs column shows the number of gene pairs syntenic identified through the DAGchainer program [68].
Insertion/deletions between the potato and the tomato sequences Figure 4 Insertion/deletions between the potato and the tomato sequences. Syntenic potato and tomato sequences, along with predicted genes (shown in grey highlights), are shown. Tomato contig 29 is shown at the top of the figure with loci (color coded as described in Figure 1) and gene models. Repetitive sequences in tomato were identified. Potato BAC AC212316 is syntenic with the tomato contig; five potato gene models are conserved in sequence and transcription order with tomato. A large region of non-colinearity (86 kb) is present.
shown in Figure 6, these clones did not generate unambiguous signals at the telomeres of potato chromosomes but produced major signals in the centromeric and pericentromeric regions of several potato chromosomes. These results showed that these potato BACs were most likely derived from centromeric rather than telomeric regions of potato chromosomes.
A significant amount of rDNA sequences (3.99%) were detected in the tomato BES dataset while rDNA sequences found in potato BAC ends (0.50%) were minimal in comparison. The tomato BES were derived from three libraries constructed with EcoRI, HindIII and MboI restriction enzymes while the potato BES were derived from two libraries constructed with EcoRI and HindIII. Multiple EcoRI and MboI restriction sites are present in both the tomato and potato rDNA sequence (data not shown) and for the potato BES dataset, the ratio between EcoRI BES and HindIII BES is 0.70 (57,778/82,481). Therefore, there should be ample detection of rDNA sequences in the potato BES datasets suggesting that there may be a bias in overall rDNA content between potato and tomato. Analysis of individual libraries for potato and tomato confirmed this finding (Additional Data File 4). The rDNA sequences in potato are reported to be on chromosomes 1 (5S; [54]) and 2 (45S; nucleolar organizing region, [54,55]). It has been reported that rDNA content differs between potato and tomato with tomato having more rDNA than potato [36,56]. Thus, it is likely that the sampling of rDNA sequences, as reflected by BES survey sequencing, is reflective of a true rDNA content difference in the nuclear genomes of tomato and potato.
To contrast with the short BES-derived genome sequence, a total of 18 phase 2 and 3 potato BACs (2.20 Mb) and 16 tomato BACs (1.69 Mb, in 8 contigs/BACs) were analyzed for repetitive sequence content. Overall, the repetitive sequence fractions identified were comparable between potato and tomato BACs (25.90% vs. 22.30%). Similar to that observed with the BES datasets, more than half of the repeats identified in both the potato and tomato BAC sequences were unclassified (13.51% vs. 11.61%, respectively) while retrotransposon sequences were the most abundant characterized repetitive element in both potato and tomato BACs (9.58% vs. 8.32%, respectively). As observed with the potato BES dataset, there were more Ty3-gypsy type retrotransposons than Ty1-copia retrotransposons (3.34% vs. 1.92%) in the potato BACs. However, in contrast to that observed in the tomato BES dataset, more Ty1-copia than Ty3-gypsy type retrotransposons were present in tomato BAC sequences (0.99% Ty3-gypsy vs. 3.48% Ty1-copia). Interestingly, more transposon sequences were found in potato and tomato BAC sequences (2.69% vs 2.35%, respectively) than in the BES datasets (1.32% potato vs 1.39% tomato BES). Not surprisingly, there were almost no telomeric-related repeti- For potato, the overall percentages of repetitive sequences identified in the BAC and BES datasets were relatively comparable (total repeats: BES 34.18% vs. BAC 25.90%) and is consistent with the fact that nearly half of the potato BACs sequenced in this study (7/19 BACs) were randomly selected and reflect characteristics of the entire potato BAC library and genome. However, for tomato, the overall composition of repetitive sequences in the BES and BAC datasets differed significantly (total repeats: BES 46.29% vs. BAC 22.30%). This is attributable to the fact that the BACs sequenced by the Tomato Genome Initiative [52] are preferentially selected from the euchromatic regions which contain less repetitive sequences than the heterochromatin regions of the genome [48].

Micro-inversion within a syntenic region of potato and tomato
Certainly, identification of differences in relative composition of repetitive sequences between potato and tomato is not novel, however, the large difference in total repetitive sequence content between tomato and potato is surprising considering that these two Solanum species diverged less than 12 Million Years Ago [57]. The haploid genome size of tomato and potato differ with tomato reported to be 950 Mb while that of potato is 865 Mb (range 798-931 Mb; [27]). Thus, with 34.2% and 46.3% repetitive sequences in potato and tomato, respectively, the total repetitive sequence space within the whole genome would be 296 Mb (potato) and 440 Mb (tomato) leaving a comparable non-repetitive fraction of their genomes of 569 Mb in potato and 510 Mb in tomato. This higher level of repetitive sequence is consistent with our finding of a higher frequency of matches within the potato BES to a Solanaceae EST compared to the tomato BES (5.5% vs 3.8%, respectively). Thus, the repetitive sequences within their respective genomes not only diverged in terms of classes of sequences but also in number leading to a biased amplification of repetitive sequences in tomato compared to potato.

Conclusion
We report on a large set of genomic sequences representing 10.2% of the potato genome. Using comparative analyses with solanaceous species we were able to demonstrate the utility and power of comparative genomics to not only annotate potato genomic sequences but also to assist in genome sequencing efforts among the Solanaceae. While we were able to confirm synteny on a genome scale with segments of the tomato and potato genome > 100 kb, we have also demonstrated that syn- teny is not absolute and that insertions/deletions as well as micro-inversions have occurred since the divergence of potato and tomato. More strikingly, the repetitive sequence content and composition of potato and tomato have diverged with impacts seen on genome architecture at both the macro-and the micro-level as evidenced through differences in telomeric-repetitive sequences and rDNA content and in interruption of synteny through transposition of retrotransposons. Our data are consistent with previous reports on repetitive sequences [36][37][38][39][40][41][42] which show divergence of this fraction of the genome within the Solanaceae. These data clearly suggest that while these two solanaceous genomes can be cross-leveraged for analysis of gene content and order, they are not interchangeable with respect to all genomic features.

Sequencing methods
The RHPOTKEY BAC library was constructed from RH parent Solanum tuberosum var. RH89-039-16 using HindIII and EcoRI restriction enzymes (C. Bachem, Pers. Comm., [28]). Templates were prepared using a high throughput alkaline lysis method, sequenced on ABI 3730 × l sequencers using TF and TR primers using standard high throughput sequencing methods, and processed with Paracel Trace Tuner [59]. All sequences were trimmed to remove vector, low-quality, and E. coli sequences using Lucy [60] and iterative runs of the TIGR Seqclean Tool [61]. All potato BAC end sequences have been submitted to the GSS division of Genbank with accession numbers EI367122-EI91525, EI812397-EI846477, and ER788642-ER870415.
Potato BAC DNA was isolated using the Sigma Phase Prep BAC DNA kit (Sigma, St Louis, MO) according to manu-FISH analysis of BACs that contain telomeric repeat sequences Figure 6 FISH analysis of BACs that contain telomeric repeat sequences RHPOTKEY083D09 and RHPOTKEY013E08, which contain telomeric repeat at one end, and RHPOTKEY017N11, which contains telomeric repeats at both ends. (a, d, g) Chromosomes prepared from USW1 (a, d) and Katahdin (g), respectively. (b, e, h) FISH signals derived from the BAC clones. (c, f, i) Images merged from chromosomes and FISH signals. Bars = 5 μm.
facturer's protocol. Approximately 7.5 ug was used for library construction. Samples were treated overnight with 100 U of Plasmid-Safe ATP-Dependent Dnase (Epicenter, Madison, WI) to remove contaminating bacterial chromosomal DNA and nebulized. Sheared DNA was precipitated and polished using the DNATerminator End Repair Kit following the manufacturer's protocol (Lucigen, Middleton, WI). Samples were electrophoresed on a 1.0% low melting point agarose and fragments in the range of 3-6 kb were selected for ligation into the pSMART-HCKan vector (Lucigen, Middleton, WI). Templates from the shotgun libraries were sequenced using TX and TY primers as described by Lucigen using standard high throughput sequencing methods on ABI 3730 × l sequencers. Sequences were trimmed as described above for the BAC end sequences and assembled with Celera Assembler [62]. Potato BAC sequences have been deposited in the HTG division of Genbank under accession numbers AC204499, AC204500, AC206931-AC206936, AC209514-AC209520, AC212037, AC212316, AC212552, and AC212966.

Fluorescent in situ hybridization
Potato variety Katahdin (2n = 48) and a haploid clone USW1 (2n = 24) derived from Katahdin were used in FISH analysis. The FISH procedure followed published protocols [63]. Briefly, BAC DNA was isolated and labeled with Biotin-UTP. Hybridization signals were detected FITCconjugated avidin. Chromosomes were counterstained by 4', 6-diamidino-2phenylindole (DAPI) and were pseudocolored in red. Images were captured digitally using a Sen-Sys CCD (charge coupled device) camera attached to an Olympus BX60 epifluorescence microscope. The CCD camera was controlled using IPLab Spectrum v3.1 software (Signal Analytics, Vienna, VA) on a Macintosh computer.

Annotation
The potato BACs and the syntenic tomato contigs were annotated in parallel. First, the potato and tomato BACs were masked for repetitive sequences using RepeatMasker with a modified TIGR Solanum Repeat Database v3.3 in which miniature inverted repeat transposable elements (MITEs) and non-transposable element-related repeats were excluded. Second, gene models were predicted using the ab initio gene finder FGENESH (dicot matrix; [64]) and were updated using transcript evidence (ESTs, cDNAs) with the Program to Assemble Spliced Alignments [65]. Moreover, the gene structures were manually inspected and some aberrant models, e.g., overlapping/nested or short (< 50 amino acids) genes, were removed. Third, gene function was assigned based on sequence identity to proteins within an in-house non-redundant protein database and/or the presence of Pfam domain(s), in a similar manner as reported previously for annotation of the rice genome [66]. Gene functions were classified into three categories: "known/putative", "expressed" or "hypothetical". Genes in which functional assignments could be assigned based on sequence similarity to a known protein or the presence of a Pfam domain above the trusted cutoff score (unique for each Pfam domain) were annotated as encoding either a known or putative protein; the remaining gene models for which no sequence similarity or Pfam domain evidence was available were annotated as encoding an "expressed protein" if cognate transcript support was available or "hypothetical protein" if cognate transcript support was absent.
The solanaceous transcript assemblies (downloaded from [58]) were searched against the potato BACs using the program GAP2 [67]. High quality alignments were defined as having sequence identity ≥ 80% and coverage ≥ 70% of the length of the Transcript Assembly. Only alignments meeting these cutoff criteria were used in downstream analyses and a solanaceous transcript was considered to support the ab initio-based annotation if the spliced alignment of the transcript overlapped a minimum of 100 bp with the gene model.

Identification of candidate syntenic tomato sequences
We utilized two methods to identify potential syntenic tomato-potato sequences. For Set I, tomato BACs were downloaded either from Genbank or SGN [52] and 14 overlapping tomato BACs were merged into 6 contigs to facilitate alignment and mapping to the potato BES. The potato BES were repeat masked and mapped to the tomato contigs using the program BLASTN with an E value cutoff of ≤ 1e-5. Paired potato BES were selected if they mapped to the same tomato contig in the correct orientation and within an expected intervening distance (50~200 kb). In total, 52 potato BACs were identified as candidate syntenic clones; eight potato BAC clones were sequenced. It is possible that BACs either from chromosome 6 or other chromosomes in the potato genome are syntenic with tomato BACs available in the public domain. To address this issue, we utilized the ab initio gene finder, FGENESH [64] to predict genes in the 18 phase 2 and 3 potato BACs and the 518 tomato BACs and searched these gene models against each other using BLASTP. The DAGchainer program [68] was employed to identify syntenic gene blocks between the potato and tomato contigs; putative syntenic potato-tomato BACs identified with this approach were termed Set II.
Synteny between tomato and potato was examined at the nucleotide and the protein level. Genomic comparisons at the nucleotide level utilized the NUCMER program [53]. Syntenic gene blocks between potato and tomato contigs were generated by the BLASTP/DAGchainer [68] pipeline using the predicted protein sequences from the semi-auto-mated annotation pipeline with improved gene structures/models rather than the ab initio FGENESH predictions.

Repeat database construction
Publicly available sequences were searched to expand our existing TIGR Solanum Repeat Database [69,70]. New Solanaceae repetitive sequences were first collected from Genbank and used to update the TIGR Solanaceae Repeat Database. The TIGR Solanaceae Repeat Database was then searched against Solanum BAC sequences (41 non-tomato Solanum BACs and 301 tomato BACs, 40.05 Mb total sequence) from GenBank and the SGN [52] using Repeat-Masker ([71] with a cut-off score of 225 which should not yield false positives). Sequences within the BACs that matched a repetitive sequence in the TIGR Solanaceae Repeat Database with ≥ 75% identity and ≥ 95% overall length were excised, coded [72], and combined with other Solanum repetitive sequences in the TIGR Solanaceae Repeat Database. Lastly, the same set of Solanum BAC sequences was searched with the de novo repetitive sequence finding algorithm, RepeatScout [73]. Low-complexity sequences in the RepeatScout-generated fasta-formated sequence output were filtered out. To prevent inclusion of paralogous protein coding genes, all Repeat-Scout-generated sequences with similarity to known proteins or Pfam domains were identified and removed. All remaining repetitive sequences were coded based on the similarity with known repetitive sequences and added to the Solanum repetitive sequences to create the TIGR Solanum Repeat Database v3.3.

Repetitive sequence identification
Potato and tomato BAC end sequences (BES, 87.14 Mb and 273.99 Mb, respectively) and BAC sequences used in this study (2.20 Mb and 1.69 Mb, respectively) were searched against the TIGR Solanum Repeat Database v3.3 using RepeatMasker with a cut-off score of 225. Genomic sequences were quantified based RepeatMasker matches to the TIGR Solanum Repeat Database v3.3 sequences and quantitated at the sub-class level [70].