Metagenomic analysis of human diarrhea: viral detection and discovery

Abstract

Worldwide, approximately 1.8 million children die from diarrhea annually, and millions more suffer multiple episodes of nonfatal diarrhea. On average, in up to 40% of cases, no etiologic agent can be identified. The advent of metagenomic sequencing has enabled systematic and unbiased characterization of microbial populations; thus, metagenomic approaches have the potential to define the spectrum of viruses, including novel viruses, present in stool during episodes of acute diarrhea. The detection of novel or unexpected viruses would then enable investigations to assess whether these agents play a causal role in human diarrhea. In this study, we characterized the eukaryotic viral communities present in diarrhea specimens from 12 children by employing a strategy of "micro-mass sequencing" that entails minimal starting sample quantity (<100 mg stool), minimal sample purification, and limited sequencing (384 reads per sample). Using this methodology we detected known enteric viruses as well as multiple sequences from putatively novel viruses with only limited sequence similarity to viruses in GenBank.

Figure 1. Categorization of sequence reads based on best tBLASTX scores (E-value: <10⁻⁵).

Pies on the left side of each box depict the categorization of sequences from individual samples by phylotype: viral (V); phage (P); bacterial (B); human (H); fungal (F); other (O); and unassigned (U). Pies on the right side of each box depict further characterization of viral sequences by viral families/taxa: Reoviridae (Reo); Caliciviridae (Calici); Astroviridae (Astro); anellovirus (Anello); picobirnavirus (Picobirna); Picornaviridae (Picorna); mimivirus (Mimi); Nodaviridae (Noda); Adenoviridae (Adeno); Parvoviridae (Parvo). Numbers in parentheses indicate the number of sequence reads in each category.

Figure 2. Phylogenetic analysis of highly divergent astrovirus-like sequence reads.

Maximum parsimony phylogenetic trees were generated by comparing the translated amino acid sequence of individual sequence reads to the corresponding sequences from known astroviruses. 1,000 replicates were generated with bootstrap values over 700 shown. A) Representative sequence read mapping to astrovirus serine protease ORF (Accession number ET065575); B) Representative sequence read mapping to astrovirus RNA polymerase (Accession number ET065582).

Figure 3. Phylogenetic analysis of a highly divergent nodavirus-like sequence read.

Maximum parsimony phylogenetic trees were generated by comparing the translated amino acid sequence of one sequence read (Accession number ET066010) to the corresponding RNA polymerase sequences of nodaviruses. 1,000 replicates were generated with bootstrap values over 700 shown.