High-Throughput Resequencing of Maize Landraces at Genomic Regions Associated with Flowering Time

Abstract

Despite the reduction in the price of sequencing, it remains expensive to sequence and assemble whole, complex genomes of multiple samples for population studies, particularly for large genomes like those of many crop species. Enrichment of target genome regions coupled with next generation sequencing is a cost-effective strategy to obtain sequence information for loci of interest across many individuals, providing a less expensive approach to evaluating sequence variation at the population scale. Here we evaluate amplicon-based enrichment coupled with semiconductor sequencing on a validation set consisting of three maize inbred lines, two hybrids and 19 landrace accessions. We report the use of a multiplexed panel of 319 PCR assays that target 20 candidate loci associated with photoperiod sensitivity in maize while requiring 25 ng or less of starting DNA per sample. Enriched regions had an average on-target sequence read depth of 105 with 98% of the sequence data mapping to the maize 'B73' reference and 80% of the reads mapping to the target interval. Sequence reads were aligned to B73 and 1,486 and 1,244 variants were called using SAMtools and GATK, respectively. Of the variants called by both SAMtools and GATK, 30% were not previously reported in maize. Due to the high sequence read depth, heterozygote genotypes could be called with at least 92.5% accuracy in hybrid materials using GATK. The genetic data are congruent with previous reports of high total genetic diversity and substantial population differentiation among maize landraces. In conclusion, semiconductor sequencing of highly multiplexed PCR reactions is a cost-effective strategy for resequencing targeted genomic loci in diverse maize materials.

Fig 1. Ampliseq workflow.

Target regions are selected and amplicons are designed to cover the region. Amplicons are then partially digested, and adapters and barcodes are ligated onto amplicons. Samples are then equalized and pooled. Sequencing was completed on an Ion Torrent PGM™.

Fig 2. Number of reads per amplicon versus amplicon length.

The number of reads per amplicon has little relationship with amplicon length. Long amplicons are represented.

Fig 3. Venn diagram comparing different SNP datasets.

Venn diagram representing the relationships among SNPs called using GATK and SAMtools, and the HapMap3 SNPs in the same genomic regions. Novel variants are those unique to the GATK and SAMtools datasets.

Fig 4. Variant classification by method.

The distribution of variants classified according to genomic region or type of variation. GATK and SAMtools variant calls were compared to the HapMap3 variants in the region of interest.

Fig 5. Multidimensional scaling applied to all samples at all resequenced loci.

The plot includes five samples per accession and control samples.