Tracking the evolution of the SARS coronavirus using high-throughput, high-density resequencing arrays

Abstract

Mutations in the SARS-Coronavirus (SARS-CoV) can alter its clinical presentation, and the study of its mutation patterns in human populations can facilitate contact tracing. Here, we describe the development and validation of an oligonucleotide resequencing array for interrogating the entire 30-kb SARS-CoV genome in a rapid, cost-effective fashion. Using this platform, we sequenced SARS-CoV genomes from Vero cell culture isolates of 12 patients and directly from four patient tissues. The sequence obtained from the array is highly reproducible, accurate (>99.99% accuracy) and capable of identifying known and novel variants of SARS-CoV. Notably, we applied this technology to a field specimen of probable SARS and rapidly deduced its infectious source. We demonstrate that array-based resequencing-by-hybridization is a fast, reliable, and economical alternative to capillary sequencing for obtaining SARS-CoV genomic sequence on a population scale, making this an ideal platform for the global monitoring of SARS-CoV and other small-genome pathogens.

Figure 1

SARS Resequencing array. (A) Diagram of the different sequence variants that can be detected by the array. Specific probes were designed to screen for previously published insertion and deletion sequences. (B) Resequencing array hybridized with Cy-3-labeledSARS-CoV cDNA. (C) Close-up view of oligonucleotide probes synthesized on the array. The four possible nucleotides for each position are synthesized adjacent to each other. SARS cDNA bound to perfect-match (PM) probes (in red) fluoresce with higher intensity than those bound to mismatch (MM) probes (in black).

Figure 2

Distribution and frequency of ambiguous calls across the SARS-CoV genome. We observed N calls at a total of 1148 bases in this study, of which 580 occurred in more than one sample.

Figure 3

Stratification of probes according to %G/C and assessment of probe performance. All PM probes were binned according to %G/C, and average PM/MM ratios, call rates, and average feature intensities were calculated. G/C content <20% or >50% leads to lowest PM/MM ratios, resulting in increased rate of ambiguous calls.

Figure 4

Effects of secondary structure on probe annealing and ambiguous calls. The most stable structure as predicted using GeneRunner software, is illustrated for two sequences with recurrent Ns; (A) bases 25953–25959, (B) bases 22781–22796. In both cases, the frequency of ambiguous calls peaks at the bases within the predicted loop structure.