Specific capture and whole-genome sequencing of viruses from clinical samples

Abstract

Whole genome sequencing of viruses directly from clinical samples is integral for understanding the genetics of host-virus interactions. Here, we report the use of sample sparing target enrichment (by hybridisation) for viral nucleic acid separation and deep-sequencing of herpesvirus genomes directly from a range of clinical samples including saliva, blood, virus vesicles, cerebrospinal fluid, and tumour cell lines. We demonstrate the effectiveness of the method by deep-sequencing 13 highly cell-associated human herpesvirus genomes and generating full length genome alignments at high read depth. Moreover, we show the specificity of the method enables the study of viral population structures and their diversity within a range of clinical samples types.

Figure 1. Coverage across sequenced genomes is highest using the target enrichment methods.

Proportions of assembled genomes at which read depth per base falls below 100 fold (lightest grey), 50 fold, 20 fold, 5 fold, 1 fold and 0 (indicated by increasing darkness).

Figure 2. Total numbers of minority variant positions in all sequenced VZV samples.

Each bar indicates the number of genome positions at which multiple alleles are present (minor allele frequency 5–49.9%). Datasets are normalised (corrected for the total number of mapped reads per sample) and showed no evidence that minority reads map to specific regions of the genome or that any bias between the proportions occurring in coding and non–coding regions of the genomes is present. Viral genome copies, post-target enrichment could not be determined for some samples (nd).

Figure 3. Mutational spectra of minority variants occurring within clinical samples.

Each bar indicates the number of genome positions at which specific allele combinations (see graphic) are present (minor allele frequency 1–10%). Datasets are normalised (corrected for the total number of mapped reads per sample) and show a clear bias toward A to G and T to C substitutions in samples prepared by long PCR. No bias was observed in samples prepared using target enrichment methods.