Norovirus Whole-Genome Sequencing by SureSelect Target Enrichment: a Robust and Sensitive Method

Abstract

Norovirus full-genome sequencing is challenging due to sequence heterogeneity among genomes. Previous methods have relied on PCR amplification, which is problematic due to primer design, and transcriptome sequencing (RNA-Seq), which nonspecifically sequences all RNA, including host and bacterial RNA, in stool specimens. Target enrichment uses a panel of custom-designed 120-mer RNA baits that are complementary to all publicly available norovirus sequences, with multiple baits targeting each position of the genome, which overcomes the challenge of primer design. Norovirus genomes are enriched from stool RNA extracts to minimize the sequencing of nontarget RNA. SureSelect target enrichment and Illumina sequencing were used to sequence full genomes from 507 norovirus-positive stool samples with reverse transcription-real-time PCR cycle threshold (CT) values of 10 to 43. Sequencing on an Illumina MiSeq system in batches of 48 generated, on average, 81% on-target reads per sample and 100% genome coverage with >12,000-fold read depth. Samples included genotypes GI.1, GI.2, GI.3, GI.6, GI.7, GII.1, GII.2, GII.3, GII.4, GII.5, GII.6, GII.7, GII.13, GII.14, and GII.17. When outliers were accounted for, we generated >80% genome coverage for all positive samples, regardless of CT values. A total of 164 samples were tested in parallel with conventional PCR genotyping of the capsid shell domain; 164/164 samples were successfully sequenced, compared to 158/164 samples that were amplified by PCR. Four of the samples that failed capsid PCR analysis had low titers, which suggests that target enrichment is more sensitive than gel-based PCR. Two samples failed PCR due to primer mismatches; target enrichment uses multiple baits targeting each position, thus accommodating sequence heterogeneity among norovirus genomes.

FIG 1

Schematic of the norovirus full-genome assembly pipeline.

FIG 2

Numbers of samples sequenced according to norovirus genotype, classified by sequencing outcome. Pass, >90% genome coverage and >100-fold read depth; suboptimal, >90% genome coverage or >100-fold read depth; fail, <90% genome coverage and <100-fold read depth. Genotype refers to capsid genotype only.

FIG 3

Norovirus full-genome sequencing outcome metrics according to norovirus genotype (a to c) and RT-qPCR C_T values (d to f). OTR, on-target reads; C_T, cycle threshold; NegEx, negative control. Red lines, median values.

FIG 4

RT-qPCR C_T values for all samples, excluding runs 30 and 31 (n = 413), sequenced with SureSelect target enrichment. Pass, >90% genome coverage and >100-fold read depth; suboptimal, >90% genome coverage or >100-fold read depth; fail, <90% genome coverage and <100-fold read depth. Red lines, median values.

FIG 5

Observed and predicted genome coverage values with 95% prediction intervals, excluding outliers identified in Fig. SA3 in the supplemental material. The fitted linear regression model is y = 7.432 − 0.059x, where the dependent variable y is the logit-transformed genome coverage proportion and the independent variable x is the PCR C_T value (n = 477).