Targeted genomic sequencing of avian influenza viruses in wetland sediment from wild bird habitats

Abstract

Diverse influenza A viruses (IAVs) circulate in wild birds, including highly pathogenic strains that infect poultry and humans. Consequently, surveillance of IAVs in wild birds is a cornerstone of agricultural biosecurity and pandemic preparedness. Surveillance is traditionally done by testing wild birds directly, but obtaining these specimens is labor intensive, detection rates can be low, and sampling is often biased toward certain avian species. As a result, local incursions of dangerous IAVs are rarely detected before outbreaks begin. Testing environmental specimens from wild bird habitats has been proposed as an alternative surveillance strategy. These specimens are thought to contain diverse IAVs deposited by a broad range of avian hosts, including species that are not typically sampled by surveillance programs. To enable this surveillance strategy, we developed a targeted genomic sequencing method for characterizing IAVs in these challenging environmental specimens. It combines custom hybridization probes, unique molecular index-based library construction, and purpose-built bioinformatic tools, allowing IAV genomic material to be enriched and analyzed with single-fragment resolution. We demonstrated our method on 90 sediment specimens from wetlands around Vancouver, Canada. We recovered 2,312 IAV genome fragments originating from all eight IAV genome segments. Eleven hemagglutinin subtypes and nine neuraminidase subtypes were detected, including H5, the current global surveillance priority. Our results demonstrate that targeted genomic sequencing of environmental specimens from wild bird habitats could become a valuable complement to avian influenza surveillance programs.IMPORTANCEIn this study, we developed genome sequencing tools for characterizing avian influenza viruses in sediment from wild bird habitats. These tools enable an environment-based approach to avian influenza surveillance. This could improve early detection of dangerous strains in local wild birds, allowing poultry producers to better protect their flocks and prevent human exposures to potential pandemic threats. Furthermore, we purposefully developed these methods to contend with viral genomic material that is diluted, fragmented, incomplete, and derived from multiple strains and hosts. These challenges are common to many environmental specimens, making these methods broadly applicable for genomic pathogen surveillance in diverse contexts.

Keywords: avian influenza; genomics; influenza; surveillance.

Fig 1

Detection of influenza A virus genome fragments in sediment by probe capture-based targeted genomic sequencing. IAV genome fragments were recovered using probe capture-based sequencing from 74 sediment specimens that had previously tested positive for IAV genomic material by RT-qPCR. (A) The number of IAV genome fragments recovered from all specimens was counted. In addition to the total count, the number of fragments originating from each of the 8 IAV genome segments (PB2, PB1, PA, HA, NP, NA, M, and NS) was also determined. (B) The sensitivity of probe capture-based targeted genomic sequencing was determined for specimens that tested positive by RT-qPCR. Overall sensitivity was calculated as the percentage of specimens positive by RT-qPCR where probe capture-based targeted genomic sequencing detected at least one IAV genome fragment from any genome segment. Sensitivity was also calculated for each of the IAV genome segments separately. (C) The number of different IAV genome segments detected in each specimen was determined.

Fig 2

Detection of influenza A virus genome fragments was limited by low abundance of viral genomic material in sediment specimens. A total of 2,312 fragments of IAV genome were recovered using probe capture-based sequencing from 74 sediment specimens that had previously tested positive for IAV genomic material by RT-qPCR. (A) The number of IAV genome fragments recovered per specimen was counted. This distribution includes specimens where no IAV fragments were recovered. The median and maximum are indicated. (B) Distribution of screening RT-qPCR Ct values for specimens, including specimens where no IAV fragments were recovered. The minimum, median, and maximum are indicated. (C) There was a moderate and statistically significant monotonic association between screening RT-qPCR Ct values and the number of IAV genome fragments detected by probe capture-based targeted genomic sequencing. Results of Spearman’s rank correlation are indicated above the upper-right corner of the scatterplot. (D) There was a moderate and statistically significant monotonic association between screening RT-qPCR Ct values and the number of different IAV genome segments detected by probe capture-based targeted genomic sequencing. Results of Spearman’s rank correlation are indicated above the upper-right corner of the scatterplot.

Fig 3

Diverse hemagglutinin and neuraminidase subtypes were detected in wetland sediment using probe capture-based targeted genomic sequencing. Haemagglutinin (HA) and neuraminidase (NA) genome segment fragments were recovered using probe capture-based sequencing from 74 sediment specimens that previously tested positive for IAV genomic material by RT-qPCR. A total of 225 HA fragments were recovered from 35 specimens, and 278 NA fragments were recovered from 34 specimens. (A) The percentage of specimens containing each HA and NA subtype was determined. (B) The total number of HA and NA fragments recovered for each HA and NA subtype was counted. (C) The number of different HA subtypes detected in each HA-positive specimen was determined, and the number of different NA subtypes detected in each NA-positive specimen was determined.

Fig 4

Recovered fragments of influenza A virus genome were sequenced deeply. IAV genome fragments were recovered using probe capture-based sequencing from 74 sediment specimens that previously tested positive for IAV genomic material by RT-qPCR. Multiple copies of each IAV fragment were sequenced, increasing sequencing depth per fragment. The median and 10th percentile of copies sequenced per fragment are indicated overall and for each IAV genome segment.

Fig 5

Length of influenza A virus genome fragments recovered from sediment specimens by probe capture-based targeted genomic sequencing. IAV genome fragments were recovered using probe capture-based sequencing from 74 sediment specimens that previously tested positive for IAV genomic material by RT-qPCR. (A) The length of each IAV genome fragment was estimated by FindFlu, a tool that aligned fragment sequences to a database of 555,364 IAV reference sequences (collected globally from avian, swine, and human hosts). Fragment length estimates were calculated from the start and end coordinates of these alignments. (B) FindFlu also estimated how much each fragment covered of its segment of origin by dividing the estimated fragment length by the length of the reference sequences to which it aligned.

Fig 6

Lineage/clade, collection location, and host species of best-matching reference sequences for H5 genomic fragments detected in wetland sediment. Ninety-three fragments of H5 subtype HA genome segment were recovered using probe capture-based sequencing from 74 sediment specimens that previously tested positive for influenza A virus genomic material by RT-qPCR. Recovered H5 fragments were aligned against 6,041 H5 subtype HA segment reference sequences annotated with lineage/clade, collection location, and host species. Best matches were identified by alignment bitscores. (A) All H5 fragments had their best matches to reference sequences belonging to American non-goose/Guangdong (gs/Gd) lineages. (B) All H5 fragments had their best matches to reference sequences collected in the United States of America (USA). (C) All H5 fragments had their best matches to reference sequences collected from waterfowl and shorebird species.

Fig 7

Phylogenetic context of H5 subtype influenza A viruses detected in wetland sediment by probe capture-based targeted genomic sequencing. A proxy phylogenetic tree was constructed from 147 recent HA segment nucleotide reference sequences belonging to the H5 subtype. Reference sequences were collected globally since 2018 (the past 5 years, inclusive). The HA segment sequence from the prototypical goose/Guangdong/96 lineage (GenBank accession NC_007362) was also included to represent clade 0 of this lineage. Monophyletic groups of highly similar sequences (all leaves within 0.025 substitutions/site of their common ancestor) were collapsed into single leaves for visual clarity. Leaves were colored according to their H5 lineage and clade. Background shading was applied to Gs/Gd lineage clades. Ninety-three fragments of H5 subtype HA segments were recovered from sediment specimens. These H5 fragments were aligned to the reference sequences composing the proxy tree. For each tree leaf, the percentage of recovered H5 fragments whose best-matching reference sequences belonged to the leaf was calculated. These percentages were indicated beside each leaf and used to scale leaf sizes.