Substantial deletion overlap among divergent Arabidopsis genomes revealed by intersection of short reads and tiling arrays
© Santuari et al.; licensee BioMed Central Ltd. 2010
Received: 30 November 2009
Accepted: 12 January 2010
Published: 12 January 2010
Identification of small polymorphisms from next generation sequencing short read data is relatively easy, but detection of larger deletions is less straightforward. Here, we analyzed four divergent Arabidopsis accessions and found that intersection of absent short read coverage with weak tiling array hybridization signal reliably flags deletions. Interestingly, individual deletions were frequently observed in two or more of the accessions examined, suggesting that variation in gene content partly reflects a common history of deletion events.
Ultra-high throughput sequencing (UHTS) has become affordable to re-sequence genomes of model organisms, such as Arabidopsis thaliana [1–5]. While identification of single nucleotide polymorphisms (SNPs) and small indels from UHTS short reads is relatively easy, detection of structural variation, such as larger deletions, is less straightforward [2, 3, 6, 7]. This is particularly true for analysis of divergent genomes, such as those of Arabidopsis strains that are not closely related to the reference accession, Columbia-0 (Col-0). For instance, the accuracy of short read mapping depends on the number of polymorphic sites permitted per read . If it is set too high, it can result in read mapping to false locations; if it is set too low, it can prevent mapping to the correct location. Moreover, local accumulation of polymorphisms with respect to the reference genome can occur and such reads could only be correctly mapped with unrealistically relaxed settings that would interfere with overall correct annotation. Consequently, the corresponding reference genome regions would not be covered in standard mapping protocols, and whether or not these regions reflect excess polymorphism or deletions would remain ambiguous. Novel technologies, such as paired end read sequencing, combined with novel instruments, might eventually enable precise mapping of larger deletions. However, to date bioinformatic tools to exploit such data are still scarce , and whether the available algorithms deliver comprehensive analyses has not been experimentally verified.
Another tool to predict deletions are genome tiling array hybridizations, either through statistical analysis of hybridization signals [9–11] or empirically determined thresholds [12, 13]. In these approaches, signal ratios from hybridizations with DNA from a divergent strain versus DNA from the reference strain used for array design are analyzed to infer absence of the sequence homologous to a given tile. However, experimental verification suggests that deletions predicted in this manner contain a high number of false positives (approximately 47%) .
Finally, although inherently difficult and non-comprehensive [14, 15], contig-building from UHTS could identify larger deletions in genome variants with some success . Interestingly, these correlated with reduced hybridization signal in corresponding re-sequencing arrays [3, 7, 16]. Thus, intersection of UHTS with tiling array hybridization could be a powerful tool to pinpoint deletions. Here we applied this procedure to investigate genomic variation in four divergent, isogenized Arabidopsis strains (so-called accessions): Eilenburg-0 (Eil-0), Loch Ness-0 (Lc-0), Slavice-0 (Sav-0) and Tsushima-1 (Tsu-1).
Results and discussion
Novel UHTS data were generated for Eil-0, Lc-0 and Sav-0 using an Illumina Genome Analyzer II platform, while published data for Tsu-1  served as comparison. To estimate the quality of our data, we mapped the Eil-0 and Lc-0 short reads onto previously established approximately 94 kb (Eil-0) and approximately 96 kb (Lc-0) of high quality genomic DNA sequence obtained from 144 loci by dideoxy sequencing . Mapping with three mismatches allowed in the 5', 28 bp of each 35- to 36-bp read to account for sequencing errors using MAQ (Mapping and Assembly Quality software)  failed to cover approximately 1.3% (Eil-0) and 5.0% (Lc-0) of sequence, which thus appeared to be absent. Such missing sequence is not unusual and could reflect insufficient coverage (17.1 for Eil-0, 6.4 for Lc-0), the stochastic nature of the sequencing process, or technical biases [3, 5, 18–20].
Using available tiling array data  and additional hybridizations, we determined the hybridization signal ratio (that is, log2 of mean signal from two arrays hybridized with divergent DNA divided by mean signal from two arrays hybridized with Col-0 DNA) of all 25-bp tiles (Affymetrix Arabidopsis Tiling 1.0R Arrays) for each accession. To avoid ambiguities due to cross-hybridization, we concentrated on tiles that are unique in the Col-0 genome . Next, we determined the tiles' UHTS coverage based on our MAQ mappings. Tiles that were not at all covered were considered candidates for missing sequence and analyzed further. We first applied an empirically determined threshold  and selected tiles with a signal ratio less than -1.5. To detect major deletions, we focused on consecutive tiles that covered ≥300 bp (taking into account spacing between tiles, typically 10 bp). For experimental verification, we chose 47 deletions predicted on chromosome 1 of Eil-0 (26) or Tsu-1 (21) and designed flanking primers (Additional file 2). In replicate PCR experiments with independent genomic DNA template preparations, we then observed a consistent pattern: nearly all (46) loci could be amplified from Col-0 DNA as expected; by contrast, loci presumptively deleted in Eil-0 could not be amplified from Eil-0 DNA, but could be amplified from Tsu-1 DNA, and vice versa; loci presumptively missing from both Eil-0 and Tsu-1 could not be amplified from either background. Inspection of the tiles flanking the loci, up to and beyond primer locations, revealed that they were often not covered and had negative, although not <-1.5, signal ratios (average -0.85 for Eil-0, -1.22 for Tsu-1). Thus, our criteria were apparently overly stringent. The particular threshold used should be driven by the goals of particular researchers and the cost associated with false positive or false negative inferences (Additional file 3). In the following, we focused on an empirical threshold of less than -1.0 derived from the signal ratios describe above and the average ratios from polymorphic tiles in the Eil-0 and Lc-0 dideoxy reference sequences. This simple threshold identified a set of putative deletions with high confidence.
To estimate technical variability, we first intersected Col-0 tiling array hybridizations and UHTS data . Out of 2.88 million tiles considered, 62,720 displayed a signal ratio <-1.0, and 4,711 could not be covered by UHTS reads (Figure 1c). The intersection of the two groups was only 212 tiles. Considering the range of intersection in the four accessions (46,008 to 61,798 tiles), false positives due to technical variability thus appeared to be relatively low. In the divergent genomes, a significant fraction of intersection tiles might represent SNP hotspots that could not be mapped . Interestingly, such hotspots have been preferentially found around confirmed deletions in rice strains . However, the fraction of such tiles should be relatively low, as even high levels of polymorphism (5 to 10 SNPs in 25 bp) resulted in rather mild negative signal ratio as determined from the dideoxy data (average -0.21). Moreover, based on the <-1.0 threshold, we selected 21 predicted deletions ≥100 bp from Lc-0, the accession with lowest UHTS coverage. For PCR verification, primers were this time designed to anneal in well covered flanking regions (Additional file 2). All 21 loci could be amplified and 17 displayed deletions in Lc-0. Thus, our method performed well even with limited UHTS data.
Approximately 57% of Eil-0 reads and all Lc-0 reads originated from paired end sequencing runs, which would principally enable direct prediction of deletions from paired end map positions. To estimate the performance of our approach, we thus re-analyzed the Eil-0 and Lc-0 reads using the Breakdancer algorithm , an extension of MAQ that takes into account spacing between mapped paired end reads to predict deletions. Interestingly, this approach generally predicted fewer deletions (Additional file 4) and failed to identify 2 out of 17 experimentally confirmed deletions in Lc-0, and 17 out of 26 in Eil-0 (Additional file 2). Importantly, this was true for repeated analyses that explored the Breakdancer parameter range. Thus, with our data, intersection of UHTS with tiling arrays yielded more comprehensive information, particularly with respect to larger deletions, such as those experimentally verified for Eil-0.
Our study suggests that combination of UHTS with tiling array analysis is a valid and economical approach to reliably flag deletions in divergent genomes. Analysis of the four divergent genomes suggests that deletions preferentially affect transposable element genes, but also significant numbers of protein coding genes. Our observation that many predicted deletions are shared between two or more of the accessions examined suggests that variation in gene content to some degree reflects a common history of deletion events, which has been partly shaped by transposable element activity.
Materials and methods
Tiling arrays: mapping and pre-processing
DNA samples (extracted with Qiagen [Hilden, Germany] DNeasy Plant kits according to the manufacturer's instructions) from the four accessions were hybridized to Affymetrix GeneChip® Arabidopsis Tiling 1.0R arrays in duplicate as described . Probe sequences from the BPMAP specification of the array (At35b_MR_v04-2_TIGRv5) were mapped on the Col-0 TAIR8.0 genome release downloaded from The Arabidopsis Information Resource (TAIR) , using BioConductor . Only probes with a perfect match and single occurrence in the genome were retrieved. Approximately 15% of reads in each sample represented contamination from organelle DNA. For each accession, probe intensities from two tiling array hybridizations were normalized by quantile normalization along with the intensity values of the two reference array hybridizations of Col-0 DNA. The log2 ratio of the mean of the two intensities from the accession arrays over the mean of the two intensities from the control arrays was taken as the reference signal for each tile.
UHTS: genome-wide mapping of short reads and coverage analysis
The genomes of Eil-0, Lc-0 and Sav-0 were re-sequenced using the Illumina Genome Analyzer II platform according to the manufacturer's instructions. Several lanes of either single end runs (Eil-0 and Sav-0) or paired end runs (Eil-0 and Lc-0) were produced for each accession. Short reads from single end runs for Col-0 and Tsu-1 were retrieved from published data . For each accession, short reads were filtered by quality (MAQ standard settings) and mapped on the TAIR 8.0 Col-0 genome using the MAQ algorithm . We allowed up to three mismatches in the first (5') 28 bp of the read. The number of reads mapped on each base-pair was considered in all subsequent analyses and we defined it as the read coverage. For each tile, we computed the mean coverage across the 25 bp interval on the genome relative to the probe sequence and we used this information in the comparison of the tiling array signal with the short read coverage. Purely bioinformatic deletion mapping taking into account the information from paired end data was performed using the Breakdancer algorithm , an extension to MAQ.
Mapping of short reads to the Eil-0 and Lc-0 dideoxy reference sequence
Short reads from the Eil-0 accession were mapped onto 94,076 bp of dideoxy sequence obtained from 144 loci of the Eil-0 genome, onto 95,980 bp of dideoxy sequence obtained from 144 loci of the Lc-0 genome using the MAQ software. We allowed from zero up to three mismatches in the mapping process to take into account possible sequencing errors. We performed five repetitions in order to see how much the reads with several possible mapping positions on the reference sequence affect the coverage.
Identification of deletions and gene level analysis
For each of the four accessions, we analyzed the mean coverage and the signal relative to the genomic positions of the probe sequence of each tile. We identified regions where the probe sequences are spaced by typically 10 bp, but always less than 40 bp, and are characterized by having no short read coverage and a tile signal ratio below an arbitrary threshold. According to the analysis of the distribution of the signal in each array, at first we decided to set this threshold to be <-1.5. After PCR validation of major deletions in Eil-0 and Tsu-1, we were able to determine a less stringent threshold, <-1.0, and we repeated the above analysis to annotate putative deletions for each strain. To understand how these deletions affect functional gene content in the accessions, we considered the base-pair positions of the deleted regions that span the genes, based on the TAIR 8.0 GFF gene annotation. We first analyzed untranslated regions, exons and introns, according to the 'mRNA' feature in the GFF annotation file, and then focused on the coding sequences of the genes, the 'CDS' (coding sequence) feature in the GFF file.
Molecular biology and plant materials
Plant tissue culture and molecular biology procedures followed routine protocols as described [12, 27]. Tiling array source files are available from ArrayExpress [ArrayExpress:E-MEXP-2220], all short reads generated in this study are available from the NCBI-GEO short read archive [NCBI-GEO:SRA009330]. The scripts used for the bioinformatics analyses of our data are provided in Additional files 7 and 8.
A. thaliana accession Columbia-0
A. thaliana accession Eilenburg-0
A. thaliana accession Loch Ness-0
Mapping and Assembly Quality software
A. thaliana accession Slavice-0
single nucleotide polymorphism
The Arabidopsis Information Resource
A. thaliana accession Tsushima-1
ultra-high throughput sequencing.
We would like to thank O Hagenbüchle and A Paillusson for Affymetrix tiling array hybridizations and S Plantegenet for DNA samples and help with primer design. The computations were performed at the Vital-IT Center for high-performance computing of the Swiss Institute of Bioinformatics. This study was supported by the University of Lausanne, by a Marie-Curie post-doctoral fellowship awarded to ED, by National Science Foundation grant DEB 0823305 awarded to TEJ, and by SystemsX 'Plant growth in a changing environment' funding for CSH, IX and LS.
- Hillier LW, Marth GT, Quinlan AR, Dooling D, Fewell G, Barnett D, Fox P, Glasscock JI, Hickenbotham M, Huang W, Magrini VJ, Richt RJ, Sander SN, Stewart DA, Stromberg M, Tsung EF, Wylie T, Schedl T, Wilson RK, Mardis ER: Whole-genome sequencing and variant discovery in C. elegans. Nat Methods. 2008, 5: 183-188. 10.1038/nmeth.1179.PubMedView ArticleGoogle Scholar
- Lister R, Gregory BD, Ecker JR: Next is now: new technologies for sequencing of genomes, transcriptomes, and beyond. Curr Opin Plant Biol. 2009, 12: 107-118. 10.1016/j.pbi.2008.11.004.PubMedPubMed CentralView ArticleGoogle Scholar
- Ossowski S, Schneeberger K, Clark RM, Lanz C, Warthmann N, Weigel D: Sequencing of natural strains of Arabidopsis thaliana with short reads. Genome Res. 2008, 18: 2024-2033. 10.1101/gr.080200.108.PubMedPubMed CentralView ArticleGoogle Scholar
- Srivatsan A, Han Y, Peng J, Tehranchi AK, Gibbs R, Wang JD, Chen R: High-precision, whole-genome sequencing of laboratory strains facilitates genetic studies. PLoS Genet. 2008, 4: e1000139-10.1371/journal.pgen.1000139.PubMedPubMed CentralView ArticleGoogle Scholar
- Wei W, McCusker JH, Hyman RW, Jones T, Ning Y, Cao Z, Gu Z, Bruno D, Miranda M, Nguyen M, Wilhelmy J, Komp C, Tamse R, Wang X, Jia P, Luedi P, Oefner PJ, David L, Dietrich FS, Li Y, Davis RW, Steinmetz LM: Genome sequencing and comparative analysis of Saccharomyces cerevisiae strain YJM789. Proc Natl Acad Sci USA. 2007, 104: 12825-12830. 10.1073/pnas.0701291104.PubMedPubMed CentralView ArticleGoogle Scholar
- Chen K, Wallis JW, McLellan MD, Larson DE, Kalicki JM, Pohl CS, McGrath SD, Wendl MC, Zhang Q, Locke DP, Shi X, Fulton RS, Ley TJ, Wilson RK, Ding L, Mardis ER: BreakDancer: an algorithm for high-resolution mapping of genomic structural variation. Nat Methods. 2009, 6: 677-681. 10.1038/nmeth.1363.PubMedPubMed CentralView ArticleGoogle Scholar
- Clark RM, Schweikert G, Toomajian C, Ossowski S, Zeller G, Shinn P, Warthmann N, Hu TT, Fu G, Hinds DA, Chen H, Frazer KA, Huson DH, Scholkopf B, Nordborg M, Ratsch G, Ecker JR, Weigel D: Common sequence polymorphisms shaping genetic diversity in Arabidopsis thaliana. Science. 2007, 317: 338-342. 10.1126/science.1138632.PubMedView ArticleGoogle Scholar
- Gilad Y, Pritchard JK, Thornton K: Characterizing natural variation using next-generation sequencing technologies. Trends Genet. 2009, 25: 463-471. 10.1016/j.tig.2009.09.003.PubMedPubMed CentralView ArticleGoogle Scholar
- Zhang X, Byrnes JK, Gal TS, Li WH, Borevitz JO: Whole genome transcriptome polymorphisms in Arabidopsis thaliana. Genome Biol. 2008, 9: R165-10.1186/gb-2008-9-11-r165.PubMedPubMed CentralView ArticleGoogle Scholar
- Borevitz JO, Liang D, Plouffe D, Chang HS, Zhu T, Weigel D, Berry CC, Winzeler E, Chory J: Large-scale identification of single-feature polymorphisms in complex genomes. Genome Res. 2003, 13: 513-523. 10.1101/gr.541303.PubMedPubMed CentralView ArticleGoogle Scholar
- Vaughn MW, Tanurdzic M, Lippman Z, Jiang H, Carrasquillo R, Rabinowicz PD, Dedhia N, McCombie WR, Agier N, Bulski A, Colot V, Doerge RW, Martienssen RA: Epigenetic natural variation in Arabidopsis thaliana. PLoS Biol. 2007, 5: e174-10.1371/journal.pbio.0050174.PubMedPubMed CentralView ArticleGoogle Scholar
- Plantegenet S, Weber J, Goldstein DR, Zeller G, Nussbaumer C, Thomas J, Weigel D, Harshman K, Hardtke CS: Comprehensive analysis of Arabidopsis expression level polymorphisms with simple inheritance. Mol Syst Biol. 2009, 5: 242-10.1038/msb.2008.79.PubMedPubMed CentralView ArticleGoogle Scholar
- Emerson JJ, Cardoso-Moreira M, Borevitz JO, Long M: Natural selection shapes genome-wide patterns of copy-number polymorphism in Drosophila melanogaster. Science. 2008, 320: 1629-1631. 10.1126/science.1158078.PubMedView ArticleGoogle Scholar
- Butler J, MacCallum I, Kleber M, Shlyakhter IA, Belmonte MK, Lander ES, Nusbaum C, Jaffe DB: ALLPATHS: de novo assembly of whole-genome shotgun microreads. Genome Res. 2008, 18: 810-820. 10.1101/gr.7337908.PubMedPubMed CentralView ArticleGoogle Scholar
- Simpson JT, Wong K, Jackman SD, Schein JE, Jones SJ, Birol I: ABySS: a parallel assembler for short read sequence data. Genome Res. 2009, 19: 1117-1123. 10.1101/gr.089532.108.PubMedPubMed CentralView ArticleGoogle Scholar
- Zeller G, Clark RM, Schneeberger K, Bohlen A, Weigel D, Ratsch G: Detecting polymorphic regions in Arabidopsis thaliana with resequencing microarrays. Genome Res. 2008, 18: 918-929. 10.1101/gr.070169.107.PubMedPubMed CentralView ArticleGoogle Scholar
- Maq: Mapping and Assembly with Qualities. [http://maq.sourceforge.net]
- Dohm JC, Lottaz C, Borodina T, Himmelbauer H: Substantial biases in ultra-short read data sets from high-throughput DNA sequencing. Nucleic Acids Res. 2008, 36: e105-10.1093/nar/gkn425.PubMedPubMed CentralView ArticleGoogle Scholar
- Pinard R, de Winter A, Sarkis GJ, Gerstein MB, Tartaro KR, Plant RN, Egholm M, Rothberg JM, Leamon JH: Assessment of whole genome amplification-induced bias through high-throughput, massively parallel whole genome sequencing. BMC Genomics. 2006, 7: 216-10.1186/1471-2164-7-216.PubMedPubMed CentralView ArticleGoogle Scholar
- Harismendy O, Ng PC, Strausberg RL, Wang X, Stockwell TB, Beeson KY, Schork NJ, Murray SS, Topol EJ, Levy S, Frazer KA: Evaluation of next generation sequencing platforms for population targeted sequencing studies. Genome Biol. 2009, 10: R32-10.1186/gb-2009-10-3-r32.PubMedPubMed CentralView ArticleGoogle Scholar
- The Arabidopsis Genome Initiative: Analysis of the genome sequence of the flowering plant Arabidopsis thaliana. Nature. 2000, 408: 796-815. 10.1038/35048692.View ArticleGoogle Scholar
- Shen Y, Sarin S, Liu Y, Hobert O, Pe'er I: Comparing platforms for C. elegans mutant identification using high-throughput whole-genome sequencing. PLoS ONE. 2008, 3: e4012-10.1371/journal.pone.0004012.PubMedPubMed CentralView ArticleGoogle Scholar
- Bentley DR: Whole-genome re-sequencing. Curr Opin Genet Dev. 2006, 16: 545-552. 10.1016/j.gde.2006.10.009.PubMedView ArticleGoogle Scholar
- Tian D, Wang Q, Zhang P, Araki H, Yang S, Kreitman M, Nagylaki T, Hudson R, Bergelson J, Chen JQ: Single-nucleotide mutation rate increases close to insertions/deletions in eukaryotes. Nature. 2008, 455: 105-108. 10.1038/nature07175.PubMedView ArticleGoogle Scholar
- The Arabidopsis Information Resource. [http://www.arabidopsis.org]
- Bioconductor. [http://www.bioconductor.org]
- Sibout R, Plantegenet S, Hardtke CS: Flowering as a condition for xylem expansion in Arabidopsis hypocotyl and root. Curr Biol. 2008, 18: 458-463. 10.1016/j.cub.2008.02.070.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.