Introduction

Wheat is probably one of the most important crops in the world, yet it has one of the most complex genomes, confounding many genomic applications for wheat crop improvement. Bread wheat is hexaploid, being derived from a combination of three diploid donor grass species each with seven pairs of chromosomes, resulting in a total of 21 pairs of chromosomes. The donor species are proposed to have diverged from an ancestral diploid species between 2.5 and 6 MYA (Huang et al. 2002; Chantret et al. 2005), and subsequently underwent two interspecies hybridisations which were each followed by a chromosome doubling event that produced a series of allopolyploid genomes. The first polyploidisation event between 0.5 and 3 MYA combined the genomes of Triticum urartu (AuAu) and an unidentified species (BB), which bears high similarity to Aegilops speltoides, to produce the allotetraploid genome of wild emmer wheat, Triticum turgidum (AuAu BB) (Chantret et al. 2005; Eckardt 2001; Huang et al. 2002). A second polyploidisation event occurred following domestication, with the hybridisation of T. turgidum (AuAu BB) and Ae. tauschii (DD), to produce the allohexaploid genome of T. aestivum (AuAu BBDD). Each of these diploid progenitor genomes is between 4,000 and 6,000 million base pairs, almost twice the size of the human genome (Šafář et al. 2010). The bread wheat genome consists predominantly of repetitive elements which make up between 80 and 90% of the genome (Flavell et al. 1977; Wanjugi et al. 2009). The size, abundance of repeats and hexaploid nature of the wheat genome make it one of the most challenging crop genome sequences to assemble.

Reference genome sequences are valuable tools, allowing researchers to relate the heritable variation in agronomic traits with underlying variation in the genome (Edwards and Batley 2010; Duran et al. 2010; Buell and Last 2010). The sequence content of the genes or specifically the allelic variants of the sequenced genes are responsible for many of the heritable differences between crop varieties. An increasing number of crop genomes are becoming available, however, elucidating the larger genomes of some cereal crops such as wheat and barley has been hampered by their size and complexity. Recently, second generation sequencing (2GS) methods have been applied to characterise isolated fractions of these genomes, leading to a greater understanding of gene content and genome structure (Mayer et al. 2009; Berkman et al. 2011).

2GS platforms produce large amounts of short DNA sequence reads of length typically between 25 and 500 bp. Illumina produce some of the leading technologies for second generation sequencing. These systems use reversible terminator chemistry, and their current flagship system, the HiSeq 2000, can generate more than 200 thousand million bases of usable data per run. Illumina sequence reads are relatively short (100–150 bp), but can be produced as read pairs where the ends of a DNA fragment of known length are sequenced which provides additional information about the context of the sequence reads. The use of paired read information has greatly improved the applicability of short read sequence data for genome assembly, as reviewed by Imelfort and Edwards (2009).

While it is unlikely that a large eukaryote genome, including repeats, could be completely assembled using current second generation sequencing technology alone, the sequencing and assembly of the gene-rich non-repetitive regions is becoming relatively routine. In addition to focussing on assembling unique and low copy regions, it is possible to dissect the wheat genome into isolated chromosome arms using flow sorting (Doležel et al. 2004). The power of this approach, when applied to the complex hexaploid wheat genome, is that it provides a means to differentiate between homoeologous sequences, with chromosome arms separated prior to DNA sequencing (Šafář et al. 2010). We have previously tested the utility of this strategy in wheat by assembling shotgun sequence data from isolated chromosome arm 7DS (Berkman et al. 2011). Sequences assembled into contigs representing the unique and low copy regions, including all known 7DS genes, the majority of which could be placed within a sequence scaffold build based on synteny with a close relative (syntenic build).

A number of translocations are known to have occurred in the bread wheat genome and its donor species. The best characterised of these occurred as a series of translocation and inversion events between chromosomes 4A, 5A and 7B (Naranjo et al. 1987). Consequently, chromosome arm 5AL contains a region which originated from 4AL, chromosome arm 4AL contains regions from 5AL and 7BS, and chromosome arm 7BS also contains a small region from 5AL (Devos et al. 1995). A translocation has been also identified by genetic mapping between chromosomes 2B and 6B (Conley et al. 2004; Devos et al. 1993) and while other minor rearrangements may have occurred on other chromosomes, it is understood that syntenic blocks in wheat are largely intact (Akhunov et al. 2003).

Identifying genome rearrangement events allows a detailed analysis of the syntenic relationships between crop species and assists our understanding of the timeline of wheat genome evolution associated with human cultivation over the last few 1,000 years. Understanding genome rearrangement in this important crop is of particular value in that it can assist our understanding of the genomic basis for phenotypic differences between closely related species and varieties.

We have applied the techniques developed for the assembly of 7DS to assemble Illumina paired read sequence data for isolated chromosome arm 7BS. Comparison of the 7DS and 7BS assemblies reveals the previously reported 7BS–4AL translocation to have relocated the region of 7BS between a T. aestivum orthologue of the Bradi1g49550 gene and the 7BS telomere to 4AL. The comparison of gene sequences from wheat homoeologous chromosome arms provides a basis for distinguishing between gene homoeologues, which will in turn help our understanding of genome evolution following polyploidisation and differential expression of gene homoeologues.

Materials and methods

Data generation and validation

Seeds of double ditelosomic lines 7B and 4A of Triticum aestivum cv. Chinese Spring were provided by Professor Bikram Gill (Kansas State University, Manhattan, USA). The seeds were germinated and root tips of young seedlings were used for the preparation of liquid suspensions of intact chromosomes as previously described (Vrána et al. 2000). Chromosome arms 7BS were flow-sorted as telocentric chromosomes in two batches of 28,000 chromosomes representing 20 ng DNA whereas 4AL arms were sorted in one batch of 50,000 chromosomes corresponding to 54 ng DNA. In order to estimate contamination with other chromosomes, 1,000 chromosomes were sorted onto a microscope slide in three replicates and used for fluorescence in situ hybridization (FISH) with probes for Afa family and telomeric repeats. The average purity in sorted fractions was 93.4 and 89% for 7BS and 4AL, respectively. Chromosomal DNA was purified and subsequently amplified using Illustra GenomiPhi V2 DNA Amplification Kit (GE Healthcare, Chalfont St. Giles, United Kingdom) as previously described (Šimková et al. 2008). Two and three independent amplifications were performed for 7BS and 4AL, respectively. A total of 200 ng of pooled, amplified DNA from 7BS was used to prepare an Illumina paired-end library which was sequenced on the Illumina GAIIx and HiSeq platforms using standard protocols. This single 7BS library, which had an expected insert size of ~350 bp, was sequenced across four lanes in three separate runs. Amplified DNA from 4AL was used to prepare a Roche shotgun library, which was sequenced on the GSFLX platform using the Titanium chemistry (three full sequencing runs) at the Lifesequencing S.L. facilities (Valencia, Spain). B. distachyon chromosome sequences were downloaded from the Bd21 8× assembly databank hosted at Brachypodium.org (Vogel et al. 2010).

A custom ‘double-barrelled BLAST’ script based on the TAGdb algorithm (Marshall et al. 2010) was used to compare query sequences with the 7BS sequence data. Paired-end coverage was defined as the number of read-pair regions aligned at each nucleotide of the query sequence. This numeric coverage data was converted into a blue–red colour scale (blue = 0, red = 18) and plotted as heat maps.

Syntenic build and assembly annotation

Several trimming, filtering and assembly parameters were assessed prior to the production of the final assembly. The 7BS sequence data was filtered and trimmed using an in-house script, trimConverter.py, to produce reads with a quality score of at least 15 at each nucleotide position and a minimum length of 30 bp. The resulting read-set was filtered to remove any reads containing k-mers of length 35 bp that occurred only once. If a read was discarded, its respective read pair was passed into a single-read file for inclusion in the assembly. The trimmed and filtered read-set was assembled using Velvet version 1.0.09 (Zerbino and Birney 2008) on a DELL R905 server with 128 GB RAM. The final assembly using used a k-mer of size 33 bp and an expected coverage of 21.0, which represents the read depth after filtering. A comparative genomics approach was applied to order and orientate the wheat contigs into a draft syntenic build by identifying reciprocal best blast (RBB) hits of assembled contigs and Brachypodium genes using MEGABLAST (Zhang et al. 2000) with default e-value cut-off, as previously described (Mayer et al. 2009; Berkman et al. 2011).

The assembled contigs from chromosome arm 7BS were divided into two groups, the first containing contigs corresponding to genes that were identified to be present on 7BS via the syntenic build process and the second group containing contigs that did not. All ESTs from the organism T. aestivum available in the NCBI database as of 4 August 2011, were downloaded. The two groups of 7BS contigs were compared with all ESTs from the dataset from NCBI using TBLASTX (Altschul et al. 1990; Karlin and Altschul 1993) with an e-value cut-off of 1e−10, and the number of contigs from each group with a hit against the NCBI EST dataset was calculated. ESTs corresponding to 18,785 loci bin-mapped to specific chromosomal regions of T. aestivum, 345 of which had been predicted to be located on 7BS, were downloaded from GrainGenes (Carollo et al. 2005; Matthews et al. 2003; Qi et al. 2004). The ESTs for the 345 loci bin-mapped to 7BS were compared to the assembled 7BS contigs using BLAST and an e-value cut-off of 1e−10. All bin-mapped ESTs were compared separately to both the assembled 7BS and 7DS (Berkman et al. 2011) contigs using BLAST and an e-value cut-off of 1e−50.

454 data generated from chromosome arm 4AL was assembled with gsAssembler using default parameters (minimum overlap between reads 40 nt, minimum overlap identity 90%, alignment score identity +2 and alignment difference score −3.) (Margulies et al. 2005). All assembled contigs were compared to the predicted genes from B. distachyon using MEGABLAST (Zhang et al. 2000). RBB hits were identified from the results and RBB hits of B. distachyon predicted genes against 7BS, as well as 4AL and 7DS (Berkman et al. 2011), were plotted with respect to their position on B. distachyon chromosome 1 using in-house scripts. Code can be made available for academic purposes upon request.

Results

Data generation and validation

Two lanes of paired-end data were generated on the Illumina GAIIx with a read length of 100 bp, as well as an additional two lanes of paired-end data generated on the Illumina HiSeq with a read length of 35 bp. 173,915,402 reads were generated in total, representing 10.8 Gbp of sequence data. The mean insert size of the library was identified to be ~360 bp with a standard deviation of ~30 bp. Based on the estimated 360 Mbp size of this chromosome arm (Šafář et al. 2010), total coverage of 7BS was calculated to be 30.0×. All data has been submitted to the NCBI short read archive, reference SRA028115.1.

The short reads from 7BS were compared with the genome of B. distachyon to validate the sequence data and define the syntenic regions. Regions identified as syntenic to 7BS or 7DS are displayed as heat maps in Fig. 1. The 7BS regions are consistent with those identified as syntenic to 7DS (Berkman et al. 2011), with the exception of a short region of B. distachyon chromosome 1 which was absent in 7BS.

Fig. 1
figure 1

Heatmaps of 7BS and 7DS read pair coverage against B. distachyon genome; Wheat 7BS read pairs (a) and 7DS read pairs (b) were mapped onto the genomes of B. distachyon. The heatmap depicts read density across each of the chromosomes, with a blue–red colour scale (blue = 0, red = 18). Regions on chromosomes one and three of B. distachyon showing the highest density of both 7BS and 7DS reads are the known syntenic regions for these wheat chromosome arms. Notably, the green syntenic region identified by 7BS read pairs mapped to B. distachyon chromosome 1 does not extend beyond the 50 Mbp measure, while 7DS read pairs mapped to this same region do extend beyond the 50 Mbp measure (highlighted by brackets)

Full size image

Data assembly and bin-mapped cDNA comparison

Following data pre-processing, the 7BS sequences were assembled using Velvet (Zerbino and Birney 2008) with a k-mer size of 33 and estimated coverage of 21.0× representing the coverage level after filtering. The assembly contained 1,038,681 contigs, with an N50 of 472 bp and maximum contig length of 29,196 bp. The total assembly length was 176,154,889 bp, approximately 49% of the predicted size of this chromosome arm (Šafář et al. 2010).

Of the 2,807 contigs which were predicted to contain genes, 2,161 (76.99%) matched wheat ESTs, with an average of 45.5 ESTs per contig. A total of 94,214 (9.1%) of the contigs which were predicted not to contain genes matched wheat ESTs, with an average of 1.00 EST per contig.

We obtained 18,785 loci from the GrainGenes database (Carollo et al. 2005; Matthews et al. 2003), which had been mapped by hybridisation of ESTs with DNA from wheat deletion lines missing defined chromosomal regions (Hossain et al. 2004; Qi et al. 2004). Of the 345 loci which had been mapped to 7BS, 307 (89.0%) were identified within the 7BS assembly, 227 of which also had a match in the B. distachyon genome, with 148 of these matching within the syntenic regions. Of the 38 7BS bin-mapped loci without a match in the 7BS assembly, none matched the B. distachyon syntenic regions. These results are consistent with our previous results for 7DS and the expected error rate for bin mapping ESTs in wheat (Berkman et al. 2011) (M. Sorrells, personal communication) and suggest that the assembly represents all or nearly all the genes on 7BS.

Analysis of all 18,785 bin-mapped wheat loci (Fig. 2) showed that loci which had been bin-mapped to 7AS, 7BS or 7DS, matched contigs in both 7BS and 7DS assemblies. In addition, a greater number of 7BS bin-mapped loci had a match to a 7BS contig than 7DS mapped cDNAs. Almost twice as many 4AL mapped loci matched 7DS contigs than matched 7BS contigs. These results are consistent with a major translocation of genes from 7BS to 4AL.

Fig. 2
figure 2

Bin-mapped loci RBB hits with wheat 7BS and 7DS assembled contigs; The histogram depicts the number of RBB hits of loci which had previously been bin-mapped to individual chromosome arms, with 7BS assembled contigs (blue) or 7DS assembled contigs (red). The arrow highlights a significant difference in 4AL bin-mapped loci found in these assemblies

Full size image

Producing a syntenic build

Comparing the 7BS contigs with B. distachyon genes in the syntenic region, 780 genes were found to have a reciprocal best BLAST (RBB) hit, with an additional 187 genes found to be assembled into contigs hosting a neighbouring gene with a RBB hit, resulting in a predicted 967 wheat 7BS genes which could be placed within a syntenic build. In total, 2,471 genes are predicted to be on the Brachypodium genome in these syntenic regions, of which only 967 (39.1%) have been retained on 7BS. The orthologue containing 7BS contigs were ordered and orientated with respect to the B. distachyon genome, as described previously (Berkman et al. 2011), to produce a syntenic build of 6.5 Mbp. The majority of 7BS contigs did not match any B. distachyon genes. Reviewing the annotation of these contigs suggests that they are predominantly made up of nested transposable element insertions (data not shown).

Following the generation of the 7BS syntenic build, alignment to the 7DS syntenic build (Berkman et al. 2011) revealed 69.03% of the 1,072 genes from 7DS were conserved on 7BS, with 227 genes (23.5%) included in the syntenic build of 7BS not found in the 7DS assembly (Fig. 3). We calculate the number of genes predicted on 7BS and 7DS (Berkman et al. 2011) to be 967 and 1,072, respectively, within the syntenic builds and 665 and 663, respectively, outside the syntenic builds. By dividing these gene counts by the total size of chromosome arms 7BS (360 Mbp) and 7DS (381 Mbp) (Šafář et al. 2010) and then multiplying by the overall size of the wheat genome (17 Gbp), we estimate that there are around 77,000 genes in the whole wheat genome, with between 45,000 and 50,000 of these genes likely to be found in blocks syntenic to regions of the Brachypodium genome.

Fig. 3
figure 3

Venn diagram displaying the common genes between 7BS (grey) and 7DS (blue) within the syntenic build (a) and genes outside of the syntenic build (b)

Full size image

Delimiting the translocation between 4AL and 7BS

The similarity between the syntenic build of 7BS and 7DS dropped significantly between the genes orthologous to Bradi1g49510 and Bradi1g52510. Therefore two distinct regions can be described on 7DS, with the region from the centromere to the T. aestivum orthologue of the Bradi1g49550 displaying a high degree of similarity with 7BS, and the region from this orthologue to the telomere displaying very little similarity with 7BS. This 7DS region which is missing from 7BS likely represents the translocation from 7BS to 4AL. In the region of high similarity between 7DS and 7BS, 740 7DS genes (84.6%) were found to be conserved in 7BS.

Recently, Hernandez et al. (2011) obtained a total of 2,987,532 reads of Roche 454 sequence data for chromosome arm 4AL, representing 900,594,638 bp and approximately 1.7× coverage (NCBI short read archive reference SRA034928.1). We assembled this data using Newbler (Margulies et al. 2005) with a minimum read length of 20, overlap seed length of 16, overlap seed step of 12 and overlap minimum match identity of 90%. This assembly obtained an N50 of 451 bp, a longest contig of 7,350 bp and a total assembly size of 70,326,673 bp, approximately 13% of the predicted the size of the chromosome arm. Comparison of the 7DS region corresponding to the predicted translocated region of 7BS revealed 166 7DS gene scaffolds (83.7%) having reciprocal best blast hits with the 4AL assembly. Comparison of the 4AL assembly with the genome of B. distachyon also revealed similarity between B. distachyon genes Bradi1g49470 and Bradi1g52330, consistent with the predicted 7BS translocation (Fig. 4).

Fig. 4
figure 4

Comparison of B. distachyon chromosome 1 with a 7DS, b 7BS, and c 4AL assemblies; Reciprocal best blast hits of genes between the wheat arms and B. distachyon chromosome 1 are presented as black lines. The plots in this figure are based on chromosomal position from the short arm telomere on the y-axis and therefore display the short arm at the bottom and long arm at the top (i.e. reversed orientation from standard chromosomal display). Arrows indicate the translocated regions on 4AL and 7BS

Full size image

In addition to the region identified to have translocated from 7BS to 4AL, a small region syntenic to the telomeric region of B. distachyon 1S, containing 13 genes and likely to have originated from 5AL, appears to have transferred from 4AL to 7BS (Fig. 4). Given the predicted position of the 7BS–4AL translocation, these genes were able to be positioned at the telomeric region of the 7BS syntenic build, increasing the number of genes in the 7BS syntenic build to 980 with final size of 6,508,016 bp.

Discussion

Based on the Lander–Waterman coverage probability (Lander and Waterman 1988) the probability that a nucleotide is missing is 9 × 10−14, which suggests that every nucleotide of the chromosome arm has been included in the sequence data. The observation that the total length of the assembled sequence corresponds to only 49% of the chromosome arm can be explained by the collapsing of repetitive regions into single assemblies, while unique regions can be reasonably expected to assembly into unique, longer contigs. This preferential assembly of unique/genic regions provides the basis for the syntenic build methodology (Berkman et al. 2011). This conclusion is supported by the identification of all known 7DS and 7BS genes within the respective assemblies (Berkman et al. 2011).

Wheat 7DS contains a large number of genes which are not found on 7BS, however, the majority of these genes can be found in the 4AL assembly. Alignment of the 7BS and 7DS syntenic builds clearly places the position of the translocation between Taestivum|7BS|Bradi1g49500 and Taestivum|7BS|Bradi1g49550, with the four intervening genes missing from both 7DS and 7BS assemblies. Given the absence of these genes from 7DS as well as 7BS, it is likely that they are missing due to gene movement/loss events rather than the translocation. Sequencing of the remaining chromosome arms of wheat will identify if they are present elsewhere in the genome.

This study clearly demonstrates the value of 2GS technology in understanding the structure of complex genomes. While the translocation between 7BS and 4AL was originally identified using cytological and marker-based techniques, we have been able to pinpoint the position of this translocation at a far higher resolution than has previously been possible. By generating a syntenic build for individual chromosome arms in wheat, a sequence-based template can be produced which not only provides a detailed view of these genomic regions to assist in understanding past genomic rearrangement events, but also provides a sequence-based foundation for further analysis of genetically mapped traits. While the sequence of syntenic builds does not give the complete picture of the wheat genome, the assembly of genic and low copy regions from 2GS data provides sufficient high-resolution detail for comparative genomics, evolutionary studies, gene and genetic marker discovery, assisting researchers in applied crop improvement programs. There are likely to be several rearrangements which cannot be resolved at the resolution of a syntenic build, and these will be identified following further advances in wheat genome sequencing.

Previous studies have demonstrated the loss of gene colinearity between genomes of the grass family with increasing evolutionary distance (Wicker et al. 2010). The assembled syntenic builds enable a comparison of the level of this “gene movement” between the translocated and non-translocated regions. In this case, the level of gene conservation between 7BS/7DS and 4AL/7DS appears to be consistent (84.6% compared with 83.7%). In contrast, the level of gene conservation between the wheat chromosome arms and syntenic regions of Brachypodium is far lower than is found between the homoeologues (39.1%). This result is consistent with the analysis of wheat 7DS, where 1,072 genes (38.6%) were conserved in the Brachypodium region syntenic to 7DS (Berkman et al. 2011). While the level of gene conservation from Brachypodium to wheat is ~39% in both samples, the number of genes placed into the syntenic build compared with the number of genes in total identified in each chromosome arm (Fig. 3) indicate colinearity with Brachypodium of 59.3% for 7BS (967/1632 genes) and 61.8% for 7DS (1072/1735 genes). The observation that gene colinearity between wheat homoeologues (~84%) is greater than gene colinearity between wheat and Brachypodium (~60%) reflects the divergence of these grasses 35–40 MYA, compared with only 2.5–6 million years since the divergence within the Triticeae (Huang et al. 2002). While levels of colinearity between wheat and Brachypodium appear lower than the reported 69% conservation of colinearity between rice, sorghum, and Brachypodium (Wicker et al. 2010), our rate of colinearity between wheat and Brachypodium is consistent with results recently reported by Massa et al. (2011). This could be explained by a high rate of transposable element activity in hexaploid wheat accelerating the erosion of colinearity, which is consistent with the hypothesis of TE-driven gene movement (Vogel et al. 2010; Wicker et al. 2010) and has recently been further supported by evidence of gene erosion being lineage-dependant with an accelerated erosion of colinearity in wheat (Massa et al. 2011). While it has been suggested that wheat genes are not affected by major structural rearrangements (Choulet et al. 2010; Massa et al. 2011), the full impact of polyploidy on colinearity is not yet known. Further analysis of the gene content in wheat may provide additional insight into this.

Based on the number of genes we have identified to be present and conserved on the 7BS and 7DS chromosome arms (Berkman et al. 2011) in both syntenic and non-syntenic regions of B. distachyon, the estimated size of these regions (Šafář et al. 2010) and the size of the entire wheat genome, we estimate that there are approximately 77,000 genes in the wheat genome. Based on the number of genes included in the 7BS and 7DS syntenic builds, we estimate that between 45,000 and 50,000 wheat genes remain in syntenic blocks relative to B. distachyon. This prediction of overall gene content in wheat is substantially lower than previous estimates, with recent estimates suggesting that there may be between 100,000 and 350,000 genes in wheat (Paux et al. 2006; Rabinowicz et al. 2005; Devos et al. 2008). One previous estimate of 108,000 genes in hexaploid bread wheat was based on an 11 Mbp sample of BAC-end sequence data from chromosome 3B (Paux et al. 2006). Another estimate of 295,000 genes was based on sequence analysis of less than 1 Mbp of sequence data (Rabinowicz et al. 2005), though the authors suggest that the majority of the predicted genes are likely to represent pseudogenes. Choulet et al. (2010) provide several estimates of gene content using different methods. Using BAC sequences, they predict a weighted total of 50,000 genes per diploid genome, and around 40,000 genes on the B genome using low coverage Illumina sequence data. They conclude with an estimate of around 36,000–50,000 genes in the B genome of wheat, though gene annotation was by comparison with ESTs and cDNA sequences which may lead to a larger number compared with our more conservative estimate. All of these methods extrapolate gene presence based on relatively small samples. Extrapolating results from small samples which may not reflect the overall genome structure may bias the results, a risk that is substantially reduced by increasing the sample size (Devos et al. 2008). By sequencing 360 Mbp of 7BS and 381 Mbp of 7DS, we base our estimate on a significantly larger sample (Šafář et al. 2010; Berkman et al. 2011). In a recent paper by Massa et al. (2011) the authors suggest a total of around 36,000 genes in Ae. tauschii, the D genome donor of hexaploid wheat. This is higher than our very conservative estimate of gene content in hexaploid wheat. In comparing our assembled contigs with wheat ESTs, we identified many additional contigs which may be predicted to contain genes. While some of these contigs are likely to contain expressed genes, they may also represent pseudogenes or transposon related expressed genes. De Brujin graph based assemblers such as velvet are often confounded by repetitive regions and so produce longer contigs for low copy and unique regions compared to repeat regions (Pop 2009). This is reflected in our results by an average size of 3,521 bp for predicted gene containing contigs compared to an overall average of 161 bp. Genes which contain repetitive DNA sequence may be assembled as fragments using our approach and contribute to an overall reduction in estimated gene number. Our predicted gene count in wheat may also be an under-estimate due to the possible exclusion of novel genes, though there is a limited likelihood that the inclusion of novel genes would substantially increase the total gene count in wheat.

Our estimate of gene number does not take into account the emergence of unique, wheat-specific genes that may be annotated using alternate methods; however, it has also been noted that early estimates of gene content can inflate the actual number of genes due to mis-annotations (Bennetzen et al. 2004). The method of applying syntenic builds to estimate gene content in wheat may overcome the issue of mis-annotation by basing estimates on previously identified genes from related species, and our estimate is likely to suggest conservative minimum gene content in the genome of T. aestivum.

Conclusion

The application of 2GS data to chromosome arm 7BS of wheat has yielded results consistent with those we have previously described for 7DS, indicating that we have assembled all or nearly all genes on this chromosome arm, providing a strong basis for comparison of the homoeologues (Berkman et al. 2011).

Comparison of assemblies of two homoeologous arms 7DS and 7BS, together with a third assembly of 4AL using Roche 454 sequence data, has enabled the delimitation of the translocation between 7BS and 4AL. The previous identification of this translocation was based on genetic mapping and therefore provided limited resolution of the translocation position (Devos et al. 1995). Our method has allowed us to delimit the position of the translocation to the gene level. This high-resolution depiction of genomic rearrangements in T. aestivum provides the foundation required to undertake finer genomic analysis in wheat, particularly in deconvoluting relationships between homoeologous chromosome arms, by identifying the presence or absence of genes from specific chromosomal locations. In turn, this provides a detailed gene-rich reference enabling wheat crop improvement researchers to more effectively conduct their research.

We have provided an accurate measure of gene colinearity between homoeologues of 84% and between the homoeologues and B. distachyon of~60%. Our estimate of wheat’s gene content overcomes some of the sequencing bias, small sample size, and annotation errors, inherent in earlier estimates of gene content in the wheat genome (Paux et al. 2006; Rabinowicz et al. 2005). By applying a similar approach to the remaining chromosome arms of wheat, a much more accurate estimate could be provided on the gene content of wheat.