Enhanced virome sequencing using targeted sequence capture

Metagenomic shotgun sequencing (MSS) is an important tool for characterizing viral populations. It is culture independent, requires no a priori knowledge of the viruses in the sample, and may provide useful genomic information. However, MSS can lack sensitivity and may yield insufficient data for detailed analysis. We have created a targeted sequence capture panel, ViroCap, designed to enrich nucleic acid from DNA and RNA viruses from 34 families that infect vertebrate hosts. A computational approach condensed ∼1 billion bp of viral reference sequence into <200 million bp of unique, representative sequence suitable for targeted sequence capture. We compared the effectiveness of detecting viruses in standard MSS versus MSS following targeted sequence capture. First, we analyzed two sets of samples, one derived from samples submitted to a diagnostic virology laboratory and one derived from samples collected in a study of fever in children. We detected 14 and 18 viruses in the two sets, comprising 19 genera from 10 families, with dramatic enhancement of genome representation following capture enrichment. The median fold-increases in percentage viral reads post-capture were 674 and 296. Median breadth of coverage increased from 2.1% to 83.2% post-capture in the first set and from 2.0% to 75.6% in the second set. Next, we analyzed samples containing a set of diverse anellovirus sequences and demonstrated that ViroCap could be used to detect viral sequences with up to 58% variation from the references used to select capture probes. ViroCap substantially enhances MSS for a comprehensive set of viruses and has utility for research and clinical applications.High-throughput, massively parallel nucleotide sequence analysis has made in-depth studies of the human microbiome feasible. Thus far, most microbiome studies have focused on bacteria (Turnbaugh et al. 2009; Arumugam et al. 2011; Gajer et al. 2012; Human Microbiome Project Consortium 2012), although some include fungi (Paulino et al. 2006; Findley et al. 2013; Cleland et al. 2014; Willger et al. 2014) and viruses (Reyes et al. 2010; Minot et al. 2011; Wylie et al. 2012, 2014; De Vlaminck et al. 2013). Viruses are particularly understudied, in part due to the challenges of assessing their presence in clinical samples. Viruses as a group have highly variable genomes, with no gene shared among all viruses that can be surveyed by an amplicon-based sequencing strategy. Therefore, studies of viruses based on nucleotide sequencing require a metagenomic approach. Metagenomic shotgun sequencing (MSS) is a relatively unbiased, culture-independent method in which nucleic acid extracted from a sample is sequenced. Sequence reads are classified based on similarity to reference genomes. This approach allows comprehensive study of the viral component of the microbiome (the virome) and has led to the discovery of novel viruses (for review, see Chiu 2013) and the characterization of viruses present in healthy and sick people (Reyes et al. 2010; Minot et al. 2011; Lysholm et al. 2012; Wylie et al. 2012, 2014; Holtz et al. 2014; Oh et al. 2014; Young et al. 2014). When adequate numbers of sequence reads are generated, viruses can be characterized with regard to taxonomy and the presence of genes associated with virulence and resistance to antiviral drugs.A limitation of MSS as employed to date for virus detection is that the amount and proportion of viral nucleic acid in samples from humans may be very low, and in these cases, few viral sequences are generated. In our experience using MSS, we have detected fewer than 10 viral sequences per 25 million sequence reads generated for a virus that was detected in a sample by a molecular assay (Wylie et al. 2012). In other instances, we have failed to detect viruses known to be present based on molecular assays (Wylie et al. 2012). These difficulties may reflect the small genome size of some viruses and/or low levels of virus in the sample. This can be a particular problem for studies of the virome of healthy, asymptomatic individuals (Wylie et al. 2012, 2014), in whom virus levels may be low. In efforts to increase the sequence yield, purification or enrichment procedures have been employed, including low-speed centrifugation and/or filtration to remove bacterial and host cells, sample treatment with nucleases to digest nucleic acid not protected within virions (Allander et al. 2001), or concentration of viral particles by high-speed gradient centrifugation (for review, see Duhaime and Sullivan 2012). Each of these procedures may bias against detection of some viruses (Breitbart and Rohwer 2005; Young et al. 2014).An alternative method for enrichment of viral sequences in a metagenomic sample prior to sequencing is targeted sequence capture, a well-established approach for targeted enrichment of specific nucleic acids. Targeted sequence capture has been used extensively to assess the human exome, as well as specific gene targets (Lovett et al. 1991; Albert et al. 2007; Hodges et al. 2007; Okou et al. 2007). Sequence capture has also been applied to the study of specific viruses (Depledge et al. 2011; Duncavage et al. 2011; Koehler et al. 2014). Our aim was to develop a comprehensive viral targeted sequence capture panel that could be used to (1) assess all viruses known to infect vertebrate cells and (2) detect divergent viruses. To this end, we created ViroCap, a targeted sequence capture panel that enhances the detection of a comprehensive set of viruses with vertebrate hosts. Here we describe the first application of ViroCap to enrich a broad range of viruses from human clinical samples.