Genomic Analysis of Viral Outbreaks

Abstract

Genomic analysis is a powerful tool for understanding viral disease outbreaks. Sequencing of viral samples is now easier and cheaper than ever before and can supplement epidemiological methods by providing nucleotide-level resolution of outbreak-causing pathogens. In this review, we describe methods used to answer crucial questions about outbreaks, such as how they began and how a disease is transmitted. More specifically, we explain current techniques for viral sequencing, phylogenetic analysis, transmission reconstruction, and evolutionary investigation of viral pathogens. By detailing the ways in which genomic data can help us understand viral disease outbreaks, we aim to provide a resource that will facilitate the response to future outbreaks.

Keywords: evolution; infectious disease; phylogeny; sequencing; transmission.

FIGURE 1

Assembly and alignment pipeline for viral reads in heterogeneous samples. High-throughput sequencing reads are ❶ demultiplexed and filtered for high-quality reads, and ❷ depleted of host reads (130) and mapped to a database of possible viruses. ❸ Reads from each sample are de novo assembled, and ❹ all reads from each sample are mapped onto their own assembly. ❺ The consensus sequence is determined for each sample and then ❻ aligned to all other samples using multiple sequence alignment. See Supplemental Table 1 for available software for each step.

FIGURE 2

Rooting phylogenetic trees. Ebola virus (EBOV) sequences illustrate the importance of correctly rooting trees. Each point represents one sequence from the outbreak indicated by its color (scale bar = nucleotide substitutions per site). (a) Maximum likelihood tree rooted on the Zaire 1976 branch (shown to be the more likely root in (25, 53)). (b) The same tree rooted on the Guinea 2014 branch. Interpretation of the ancestral relationships of a single set of samples changes dramatically with root selection.

FIGURE 3

Tree topology illuminates the nature of an outbreak. (a) EBOV tree. Sequences from previous outbreaks are colored in distinct shades of green (see Figure 2). Selected sequences from the Sierra Leone outbreak highlight the low diversity within the outbreak, and the development of new clades (SL1–4, defined in (18, 25)) from a single recent ancestor. This topology suggests that each outbreak began with a single zoonotic transmission but was subsequently sustained by human-to-human transmission. (b) Lassa virus (LASV) tree containing S segment sequences from both human (circle nodes) and M. natalensis (rodent nodes) hosts. Samples are from Sierra Leone (109), where LASV is endemic. Sequences do not cluster by time or by host, indicating frequent animal-to-human transmission and a lack of discrete outbreaks.

FIGURE 4

Determining virus transmission. Transmission during the 2003 SARS outbreak in Singapore, for which both sequence and contact tracing data are available. (a) Maximum likelihood tree rooted on TOR2, the earliest reported case. The four major branches (blue boxes) roughly correspond to geographic origin (top to bottom: China, Taiwan, Singapore, Singapore). (b) Transmission tree reconstruction from sequences and sample collection dates (created using outbreaker (85); generation time: gamma distribution with mean = 8.4 and sd = 3.8, based on values from (1)). Red and yellow circles correspond to the two Singapore clades identified in (a); lined red circles are samples with only sequence data (no contact tracing). Arrows are labeled with (number of SNPs between samples) / (posterior probability of transmission). (c) Transmission tree created during the SARS outbreak by contact tracing, as reported by (131). Gray circles are unreported cases assumed to be part of the transmission chain. Comparison of panels (a-c) shows that the three methods generate similar relationships between samples.

FIGURE 5

Rate-based tests for selection in viruses. Various tests identify signals of selection at different genomic scales. (a) d_N-d_S scores for every codon in the hemagglutinin (HA) gene of H5N1 influenza A (calculated using (104)). The highest d_N-d_S scores (red) indicate codons most likely under positive selection. (b) log(d_N/d_S) for each EBOV gene and for the mucin-like region of the glycoprotein (GP). (values from (18)). (c) Synonymous constraint for every codon position in the West Nile virus genome (sliding window = 20 nucleotides) (110). Red bars mark regions of excess constraint. Asterisks mark two known RNA structural elements (orange = structural proteins, yellow = non-structural proteins), a hairpin in the capsid gene and a pseudoknot element within non-structural protein 2A.