454 sequencing put to the test using the complex genome of barley

Abstract

Background: During the past decade, Sanger sequencing has been used to completely sequence hundreds of microbial and a few higher eukaryote genomes. In recent years, a number of alternative technologies became available, among them adaptations of the pyrosequencing procedure (i.e. "454 sequencing"), promising an approximately 100-fold increase in throughput over Sanger technology--an advancement which is needed to make large and complex genomes more amenable to full genome sequencing at affordable costs. Although several studies have demonstrated its potential usefulness for sequencing small and compact microbial genomes, it was unclear how the new technology would perform in large and highly repetitive genomes such as those of wheat or barley.

Results: To study its performance in complex genomes, we used 454 technology to sequence four barley Bacterial Artificial Chromosome (BAC) clones and compared the results to those from ABI-Sanger sequencing. All gene containing regions were covered efficiently and at high quality with 454 sequencing whereas repetitive sequences were more problematic with 454 sequencing than with ABI-Sanger sequencing. 454 sequencing provided a much more even coverage of the BAC clones than ABI-Sanger sequencing, resulting in almost complete assembly of all genic sequences even at only 9 to 10-fold coverage. To obtain highly advanced working draft sequences for the BACs, we developed a strategy to assemble large parts of the BAC sequences by combining comparative genomics, detailed repeat analysis and use of low-quality reads from 454 sequencing. Additionally, we describe an approach of including small numbers of ABI-Sanger sequences to produce hybrid assemblies to partly compensate the short read length of 454 sequences.

Conclusion: Our data indicate that 454 pyrosequencing allows rapid and cost-effective sequencing of the gene-containing portions of large and complex genomes and that its combination with ABI-Sanger sequencing and targeted sequence analysis can result in large regions of high-quality finished genomic sequences.

Figure 1

Coverage of four BAC clones with sequence contigs assembled from sequence reads produced by 454 sequencing technology. a. Relationship between coverage and number of sequence contigs from two independent sequencing experiments 1 (blue) and 2 (red) for all four BACs. Because the BACs have different sizes, the number of contigs is normalised. b. Numbers of sequence contigs in different size ranges from experiment 1. Assembly of 454 sequences resulted for all four BAC clones in a few large and many small sequence contigs. c. Percentage of the total size of the BACs covered by sequence contigs of different size ranges from experiment 1. The cumulative size of all contigs was in all four cases smaller than the actual size of the BAC clone (percentage in parentheses underneath the BAC name). This is due to pooling of repetitive sequences into consensus contigs. For BAC 604D5 and 509D2, the percentage was calculated based on size estimates from agarose gels.

Figure 2

Comparison of results from 454 sequencing with ABI-Sanger sequencing. a. Map of the previously published BAC 519J4. Genes are depicted by grey boxes with transcriptional orientations indicated by arrows. Transposable elements are depicted as coloured boxes with LTRs indicated as shaded areas. Nested transposable elements are raised above the ones into which they have inserted. Regions covered by 454 sequence contigs are depicted as blue and purple bars underneath the map. Note that single copy sequences are covered well whereas multicopy sequences such as transposons or tandem repeats contain a large number of gaps. Sequence contigs used for comparison of ABI-Sanger and 454 sequencing results are depicted in purple. b. Detailed map of the region of Gap1. Three tandem repeats were pooled into the consensus contig c68. c. Multiple sequence alignment of the three repeat units shown in (b.) and the resulting consensus contig. Differences between repeat units are highlighted. d. Sequence coverage provided by 454 sequencing (blue) and ABI-Sanger sequencing (black). Red lines indicate simulated coverages with the same number of sequences assuming a purely random distribution. Red arrows indicate gaps in the ABI-Sanger coverage. Grey lines indicate coverage with 454 sequences from an independent sequencing experiment with fewer reads. The region of clearly higher coverage with 454 sequences suggests the presence of a duplicated sequence that could not be resolved with ABI-Sanger sequencing. e. Map of BAC 773K14 with aligned 454 sequence contigs and coverage with individual 454 sequences (colours as in d).

Figure 3

Production of working drafts of BAC sequences from assemblies of 454 sequences. The relative order of sequence contigs can be inferred through (a.) identification of target site duplications (TSD) of transposable element sequences located at the edges of contigs or (b.) sequence alignment with a known reference transposable element. The latter only works reliably for elements that occur only once on the BAC analysed. c. For BAC 604D5, information from the order of genes in the orthologous region of the rice genome was used as well as the structure and organisation of transposable elements. d. Five contigs from BAC 509D2 could be arranged in two supercontigs whose linear orientaion to each other is unknown. Regions covered by 454 sequencing contigs are indicated as grey bars underneath the maps in c. and d.. Genes are depicted as black and transposable elements as white boxes. Transcriptional orientations of genes are indicated by arrows. TSD used to infer contig order are indicated. Gaps that were closed through alignment to reference transposon sequences are indicated by a curly bracket. Gaps that could be closed with low-quality 454 sequences are indicated by upward arrows. Question marks indicate a gap of unknown size between. Numbers above genes correspond to gene descriptions in Table 5.