Genomic Epidemiology of Respiratory Syncytial Virus in a New England Hospital System, 2024

Abstract

Respiratory syncytial virus (RSV) is one of the main seasonal respiratory pathogens in the United States. Although several RSV vaccines were recently approved, vaccination rates remain low. We analyzed RSV-positive nasopharyngeal swabs from Boston Medical Center in 2024 using amplicon-based whole genome sequencing. We found that >80% of the samples were RSV-B, representing a major switch from 2022, when Boston RSV samples were approximately 90% RSV-A. Forty-five of 48 RSV-B samples mapped into a single clade (B.D.E.1), though not a single source within it, suggesting that the predominance of RSV-B is multifactorial. We also found examples of highly related genomes, suggesting clustered transmission. Mutations associated with vaccine escape were not observed. Our work highlights the importance of genomic surveillance for respiratory pathogens to monitor transmission dynamics, such as the unexpected switch from RSV-A to RSV-B dominance, and to understand the epidemiological changes that may be associated with RSV interventions.

Respiratory syncytial virus (RSV) is an ongoing threat to human health. An estimated 3.2 million infants and children under age 5 worldwide are hospitalized due to RSV infection each year (80 000 per year in the United States [US]) [1]. Adults aged >65 years, or those of any age with certain chronic medical conditions, have an elevated risk of severe disease and hospitalization following RSV infection. Estimates from a recent meta-analysis suggest that >150 000 adults over age 65 are hospitalized due to RSV infection each year in the US [2]. RSV transmission is generally seasonal, peaking over the winter in temperate regions. The 2 subgroups of RSV (RSV-A and RSV-B) are most often observed co-circulating during transmission seasons, and whichever subgroup dominates in a season has not been associated with temporal variation [3].

Many current clinical tests for RSV infection do not differentiate between the subgroups, as clinical care does not vary by subgroup. Disease severity has not been consistently linked to only 1 subgroup [4, 5]. Whole genome sequencing (WGS) has been employed for epidemiological surveillance of RSV in recent years. Amplicon-based WGS is well-suited to amplify target viral RNA within nasal samples containing competing RNA. Multiple tiled amplicon designs have been developed and used for RSV genomic surveillance [6–12].

Genomic analysis also allows for investigation of whether circulating RSV is accumulating resistance to therapeutics or vaccines. Three RSV vaccines were recently introduced to the US general public, for all adults ≥75 years, adults aged 60–74 years with increased risk of severe RSV, and pregnant individuals between weeks 32 and 36 of pregnancy [13]. Uptake for vaccination in these 3 groups in July 2024 was estimated at 30.8% (95% confidence interval [CI], 30.1%–31.6%), 20.7% (95% CI, 20.2%–23.2%), and 7.7% (denominator: 12 586), respectively [14]. Nirsevimab, a monoclonal antibody, was recently recommended in the US for RSV prophylaxis for infants aged <1 year [15]. The F (fusion) gene within RSV is the basis of all 3 current vaccines and contains the binding site for nirsevimab. Although much of the F gene sequence has historically been well conserved [16], RSV WGS allows us to uncover new possible mutations and enables the detection of changes that may occur in response to vaccination and monoclonal antibodies. Recently, a portion of RSV-B breakthrough infections contained mutations associated with nirsevimab resistance [17], highlighting the need for ongoing surveillance for resistance mutations.

We were interested in understanding RSV circulation in the greater Boston area during the 2024 transmission season. The most recent genomic surveillance for RSV in this area is from 2022, when 90% of sequenced genomes were part of the RSV-A sublineage [18]. To better understand the currently circulating RSV and how it compares to the 2022 genomes, we collected samples from individuals who tested positive for RSV at Boston Medical Center (BMC), a safety-net hospital, and its satellite clinics. Using amplicon-based WGS of both RSV-A and RSV-B, we analyzed the patterns of genetic diversity and signals of transmission.

MATERIALS AND METHODS

Samples

All nasopharyngeal swabs positive for RSV within January to June 2024 were identified from the electronic medical record and requested from the BMC clinical microbiology laboratory. Some requested remnant clinical samples could not be obtained because they were either already discarded, missing, or had inadequate quantity. Remnant clinical samples were stored at −80°C in 300 µL of viral transport medium. All sample and data collection was approved by the Boston University institutional review board (H-42887).

Sample Processing

Each sample was batch-processed through a 96-well plate, beginning with RNA extraction following the Zymo Quick-RNA Viral 96 kit (no DNase I treatment) (catalog number: R1040). Prior to the final elution of RNA into nuclease-free water, samples were subjected to a dry spin (no solution added). Complementary DNA was then generated from the resulting RNA using the New England Biolabs (NEB) LunaScript RT SuperMix Kit (catalog number: E3010). We used a 400 bp tiled amplicon scheme described previously (original methods available in Supplement 1 and primer sequences available in Supplement 2, both in [19]). In brief, we used Q5 Hot Start High-Fidelity DNA Polymerase (NEB, catalog: M0494) to amplify each sample with the 4 distinct primer pools. All pools used the same thermocycler protocol: 30 seconds at 98°C; 40 cycles of 15 seconds at 95°C and 5 minutes at 63°C; then held at 4°C. Each pool was run on a 1.5% agarose gel to confirm the presence of appropriate amplicon fragments. Amplicons from pools 1 and 2 for successfully amplified samples were pooled together and placed through library clean-up using AMPure XP beads. These samples underwent Illumina Nextera XT DNA Library Preparation (catalog number: FC-131-1096) and indexing with the IDT for Illumina UD Indexes Plate A/Set 1. The prepared libraries were sequenced with 2% PhiX in paired-end, 200-cycle format with a P2 flow cell on an Illumina NextSeq 2000.

Sequencing Data Processing

For each sample, we aligned reads to both the RSV-A and RSV-B reference genomes (hRSV/A/England/397/2017 [Global Initiative on Sharing All Influenza Data {GISAID}: EPI_ISL_412866] and hRSV/B/Australia/VIC-RCH056/2019 [GISAID: EPI_ISL_1653999], respectively) using minimap2 [20]. We trimmed primer-binding sites and assessed depth of the aligned reads using samtools [21]. We called consensus sequences using samtools, requiring a minimum read depth of 20 and a call fraction above 50%. We called variants using LoFreq [22], again requiring a minimum read depth of 20. We annotated variants and gene positions with a custom R script, using the tidyverse [23] and Biostrings [24] packages. Variants used in all analyses were at consensus within a given sample, with the exception of the single-nucleotide variants (SNVs) from genome BMC_RSV_145, included in Supplementary Table 1, which included subconsensus SNVs above 10% within-sample frequency.

For all downstream analyses, we required genomes to have a minimum of 20× coverage over at least 90% of the genome, as well as at least 90% of each of the F and G genes. We did not observe any indels or frameshift mutations within the genomes passing coverage filters.

Data Analysis

We used Nextclade [25] to generate coverage statistics and assign G-protein [26] and full-genome clades to each sample that passed coverage checks described above, following the standardized phylogenetic classification defined recently [27]. We performed all multiple sequence alignment using Clustal Omega [28]. We generated maximum likelihood trees using ModelFinder [29] within IQTree v2.3.6 [30], as well as time-resolved trees using the least square dating method [31] within IQTree. We used a custom R script to make pairwise comparisons between genomes, adapted from previous scripts [32], using R 4.4.1 [28].

We compared consensus sequences from the newly sequenced genomes to genomes from GISAID. On 8 November 2024, we filtered all GISAID RSV genomes available to those flagged as both “complete” and “high coverage.” We manually reviewed alignments of the RSV-A and RSV-B genomes and removed 13 RSV-B genomes from the analysis, resulting in final comparator datasets of 600 RSV-A and 796 RSV-B genomes. The full list of genomes with GISAID IDs is shown in Supplementary Table 2. We also included 7 RSV-B genomes from the surveillance of the Boston 2022 RSV surge [18]; the GenBank IDs for these genomes are also shown in Supplementary Table 2.

We plotted all data in R 4.4.1 [22], with the tidyverse [33], ggprism [34], and ggtree suite of packages [34–38]. We arranged final figures in Adobe Illustrator.

RESULTS

From January through June 2024, the BMC clinical microbiology laboratory confirmed RSV in 108 nasal swabs using the bioMérieux Biofire comprehensive respiratory panel polymerase chain reaction (21 targets), bioMérieux Spotfire R/ST mini panel (5 targets), or Cobas-liat (Roche). All nasal swabs with detected RSV were requested from the BMC clinical laboratory. We received 59 of the remnant swabs. The remaining swabs were not received because they had already been discarded or were not available for other reasons. The age distribution of individuals whose samples were sequenced was significantly different from individuals whose samples were not received for sequencing (Mann-Whitney test, W = 906, P < .001; see Table 1). Distributions of individual sex did not vary significantly between groups (Pearson χ² test, χ²(1) = 0.54; P = .46). The median patient age of sequenced samples was 42 years (range, 0–89 years), while the median patient age of all RSV-positive samples was 9 years (range, 0–93 years) (Figure 1A). Temporally, the majority of samples were collected within the conventional RSV season that overlapped with our sampling period, peaking in January (Figure 1B).

Table 1.

Demographics of Respiratory Syncytial Virus–Positive Samples

Variable	Level	No. Sequenceda	No. Not Sequencedb	No. Totalc
Age, y	Median (range)	42 (0–89)	1 (0–93)	9 (0–93)
	Mean (SD)	37 (29)	18 (29)	28 (30)
	0–2	17	28	45
	60–74	13	2	15
	≥75	4	4	8
Sex	Female	33	23	56
	Male	26	26	52
Collection month, 2024	January	26	13	39
	February	22	12	34
	March	3	18	21
	April	5	6	11
	May	2	0	2
	June	1	0	1

Abbreviation: SD, standard deviation.

^aFifty-nine samples that were processed for whole-genome sequencing.

^bForty-nine samples that were identified as RSV positive from the electronic medical record but could not be obtained for sequencing.

^cAll 108 samples. The age distributions of samples sequenced and samples not sequenced were significantly different (Mann-Whitney test, W = 906, P < .001). Distributions of individual sex did not vary significantly between those groups (Pearson χ² test, χ²(1) = 0.54; P = .46).

Figure 1.

Patient ages and collection dates of respiratory syncytial virus samples received from Boston Medical Center in 2024. A, Number of samples, stratified by patient age group. B, Number of samples received per month (January–June 2024). N = 59.

We used amplicon-based Illumina sequencing to selectively amplify RSV genomes (Figure 2A). We extracted total RNA from the viral transport medium and amplified each sample separately with tiled amplicon primers specific for RSV-A or RSV-B [19]. We then identified samples with robust amplification for an individual subgroup (A or B) for next-generation library preparation and sequencing. We aligned all processed sequencing data to both RSV-A and RSV-B reference genomes. Samples that amplified with RSV-A primers aligned well to the RSV-A genome (median read depth >1000) and did not align well to the RSV-B genome (Figure 2B); the reverse was true for samples that amplified with RSV-B primers. We examined coverage across each full genome, as well as across the F and G genes individually (Figure 2C). All samples showed effective amplification and proceeded through library construction (Figure 2D). One of the 59 samples did not generate usable sequencing data after the demultiplexing step, suggesting an issue during sample processing or library construction. Of the 58 successfully sequenced samples, 54 had a minimum of 20× genome coverage across >90% of the full genome and of each the F and G genes, and these 54 genomes were used in subsequent analyses. Of these 54 genomes, 6 were subgroup RSV-A and 48 were subgroup RSV-B. The RSV-A genomes originated entirely from children ages 6 and under, while the RSV-B samples originated from individuals of all ages (Supplementary Figure 1A). RSV-B samples were detected throughout the sample collection period, and RSV-A samples were detected in January, March, and April (Supplementary Figure 1B).

Figure 2.

Genomic data generation and quality metrics of sequenced respiratory syncytial virus (RSV) samples. A, The sample processing pipeline from nasopharyngeal sample input to whole genome sequence data output. B, We aligned processed sequence data from each sample to both RSV-A and RSV-B reference genomes. “RSV-A Samples” describes samples amplified with the RSV-A primer sets, and “RSV-B” Samples” describes samples amplified with the RSV-B primer sets. The y-axis shows read depth + 1, to aid in plotting on a log scale. The dashed line is at 100 read-pairs. C, The y-axis represents the percent of nucleotides within the F or G gene that had >20× read depth for a given sample. The dashed line at 90% represents the threshold used to retain genomes for further analysis. D, Sample success tracker for the 59 original RSV samples received that proceeded through library preparation and sequencing. The y-axis represents the number of samples and their result at each step of the process. The x-axis shows samples as they proceeded through aliquoting, amplification, library construction, initial sequencing quality control, and subgroup confirmation (through alignment to reference genomes, as shown in panel B).

We built maximum likelihood phylogenetic trees for the RSV-A and RSV-B genomes and all high coverage genomes available from GISAID [39], as of 8 November 2024. We also assigned G-protein genotypes [26] and whole genome clades [27] to genomes in the trees using Nextclade [25]. The RSV-B genomes were genetically similar to each other (Figure 3A). All RSV-B genomes had the GB5.0.5a G-protein genotype, and 45 of 58 RSV-B samples were in the B.D.E.1 clade. These B.D.E.1 genomes did not form a single cluster; instead, these genomes were representative of much of the global diversity within the B.D.E.1 clade.

Figure 3.

Respiratory syncytial virus subgroup B (RSV-B) genomes from this study compared to global RSV-B genomes. A, Maximum likelihood phylogenetic tree of all RSV-B genomes sequenced in this study and all high-coverage, complete RSV-B genomes available in Global Initiative on Sharing All Influenza Data (GISAID) as of 8 November 2024 (n = 796). The inner ring denotes by each genome's clade (assigned by Nextclade), and the outer ring denotes genome source (genomes from this study, from United States samples in GISAID, or other GISAID samples). The dashed lines on the inside of the clade ring denote the regions of the tree that are expanded in (B) and (C). B, Expansion of the B.D.4 genomes within the larger tree in (A). All genomes are colored by location, and the 2 genomes labeled “Boston” and colored in orange are from this study. C, Expansion of a portion of the B.D.4.1.1 genomes within the larger tree in (A). All genomes are colored by location; the genome labeled “Boston” and colored in orange is from this study.

The non-B.D.E.1 genomes were split between clades B.D.4 (n = 2) and B.D.4.1.1 (n = 1). When comparing these sequences to other sequences within the GISAID database, we found that the 2 B.D.4 genomes are most similar to each other, and they fell within a cluster of genomes almost entirely from Morocco (Figure 3B). The single BMC genome within the B.D.4.1.1 clade (Figure 3C) was most similar to a single genome from Australia. The other samples closest to these B.D.4.1.1 genomes were from disparate geographies, suggesting that this genome is from an undersampled area of the RSV-B phylogeny.

To look for potential signs of transmission links, we examined the BMC genomes within the B.D.E.1 clade and their similarities to each other (Figure 4A). We performed pairwise comparisons between genomes in this clade. We found that many of these genomes are unique, but others formed small clusters of highly related genomes (Figure 4B). The first cluster consisted of 2 samples originating from the same person, taken 1 day apart, and resulted in identical genomes (BMC_RSV_11 and BMC_RSV_20). Another highly related cluster of genomes were from samples taken within a 2-week period of time in late January (BMC_RSV_48, BMC_RSV_26, BMC_RSV_50, BMC_RSV_09), suggesting potential transmission links.

Figure 4.

Examination of respiratory syncytial virus subgroup B (RSV-B) genomes from this study alone. A, Maximum likelihood phylogenetic tree of all RSV-B genomes sequenced in this study, aligned to reference genome (EPI_ISL_1653999, labeled as “Reference” in the figure). Genomes clades were assigned by Nextclade. Genomes labeled by sample ID are specifically mentioned in the text. C1 and C2 denote genomes that are part of clusters 1 and 2, respectively. B, Pairwise comparisons of consensus sequences of all RSV-B genomes sequenced in this study. Each genome is represented by a row and column, and each tile is colored by the number of differences between the genomes intersecting at it. Only genome pairs with 0 differences between them are shown. Boxes denote the comparisons between genomes in C1 and C2, matching those indicated in (A).

The RSV-A samples from BMC were a minority of sequenced samples (n = 6). They were evenly distributed across the collection period, and the genomes were more genetically diverse than the BMC RSV-B genomes (Supplementary Figure 2). While they all shared the GA2.3.5 G-protein genotype, the 6 genomes were from 5 distinct clades (A.D.1, A.D.5.1, A.D.1.4, A.D.1.5, and A.D.1.6). This indicated that there were multiple introductions of RSV-A to the BMC community, but no onward transmission was detected.

All 3 of the RSV vaccines currently available in the US contain protein or messenger RNA of the prefusion form of the F gene. Nirsevimab, a monoclonal antibody used as prophylaxis for infants before the winter transmission season, targets antigenic regions of the F protein. With these potential selection pressures in mind, we examined all nonsynonymous SNVs within the F gene in both the BMC genomes and all 2024 genomes from GISAID (Supplementary Table 1). We identified 15 SNVs in the BMC samples, 7 of which were present in >1 genome. Three SNVs (S190N, S211N, S389P) were present in >90% of both the BMC and GISAID samples, representing almost all of the B.D.E.1 genomes in each sample set. Of the identified nonsynonymous SNVs, 3 (S190N, R191K, R209Q) are within antigenic site Ø, the binding site of nirsevimab [40]. One SNV (I432V) is present within antigenic site IV [41]. No nonsynonymous SNVs within antigenic site II [41] were identified in the BMC genomes.

One individual in the BMC cohort (BMC_RSV_14) received prophylactic nirsevimab 40 days prior to their RSV-positive sample. We examined subconsensus SNVs within the genome isolated from this sample (Supplementary Table 1). This genome contained 3 subconsensus nonsynonymous SNVs within the F gene (L467F, A529V, T558A); however, none of these have been previously associated with nirsevimab resistance.

DISCUSSION

In this study, we report on RSV genomic diversity in the greater Boston area from January to June 2024. We found an unexpected shift in subgroup dominance in this set of genomes compared to what was previously seen in the area in 2022 and compared to other recent RSV surveillance elsewhere in the US. The majority of genomes sequenced in this study were RSV-B (48/54), while the most recent previous RSV genomes from the Boston area were primarily RSV-A (70/77 in 2022) [18].

Recent surveillance in Arizona [42], Ontario [10], and France [43] reported that RSV-A genomes predominate in their sequencing studies. This is analogous to the 2022 RSV surge described in Boston [18]. Surveillance in Minnesota from July 2023 to February 2024 [44], overlapping partially with our sampling period, reported nearly equal numbers of RSV-A and RSV-B genomes, the highest recent proportion of RSV-B genomes in the US reported before this study. At this time, it is unclear whether the predominance of RSV-B in our study presages a convergence to RSV-B nationwide, or if this predominance is a more geographically restricted phenomenon.

Within the RSV-B genomes sequenced in this study, we found clade representation similar to other recent studies [10, 12, 19, 42–45]. Nearly all of the Boston RSV-B genomes were within clade B.D.E.1, replicating the B.D.E.1 dominance within RSV-B genomes sequenced in recent surveillance in Minnesota [44] and Ontario [10]. The B.D.E.1 clade also contains over half of the high-coverage, complete genomes available on GISAID as of 8 November 2024. Interestingly, of the 7 RSV-B genomes sequenced in the previous Boston dataset [18], 6 are in a cluster closely related to a cluster of 14 of the 2024 genomes sequenced in this study (Supplementary Figure 3) [18]. This suggests that the RSV-B dominance in our study was not solely from expansion of virus populations from 2022; instead, additional RSV-B strains were introduced that have continued to circulate during the 2024 season.

The 3 RSV-B genomes that did not map into the B.D.E.1 clade had recent common ancestors that suggested an origin outside of the US. This indicates but does not prove that these cases could be travel-related. Importantly, the introduction of these distinct RSV-B clade genomes did not lead to further onward transmission within the BMC community. The predominance of B.D.E.1 in our dataset and the minimal clade dispersion is consistent with transmission of RSV-B being driven by local and not imported transmission. The predominance of B.D.E.1 was different from what we and others observe with the RSV-A genome. We identified multiple RSV-A clades in circulation, similar to the diversity of clades present in recent RSV-A surveillance in North America [10, 44] and globally [12].

In addition to these patterns of genetic diversity, we examined polymorphisms within the F gene to see if there was evidence of purifying selection for mutations that would provide resistance to US Food and Drug Administration–approved monoclonal antibodies or RSV vaccines. However, the consistency in common nonsynonymous SNVs in the samples sequenced here, in recent surveillance for nirsevimab resistance-associated mutations [17, 43], and in the wider set of samples from GISAID does not indicate any SNVs restricted to these Boston genomes alone. Thus, we see no evidence of treatment-resistant mutations in Boston-circulating RSV.

There are limitations to our study. First, data come from a single center. Further surveillance data from other surrounding and not contiguous medical centers are needed to confirm the generalizability of our observations. Second, we were unable to sequence all RSV detected within the BMC community, and our sampling period may not have captured the earliest RSV cases of this transmission season, those in late 2023. The samples not sequenced could have a different proportion of RSV-A and RSV-B especially because of the demographic differences. Finally, we can speculate but cannot determine the reasons for the predominance of RSV-B over RSV-A in 2024 as compared to 2022.

This study highlights the importance of regular genomic surveillance to monitor pathogen dynamics. Just 2 years after the most recent RSV surveillance in Boston [18], we identified a drastic shift in subgroup prevalence in RSV circulating in this area. In future surveillance studies, comparisons could be done to search for any linkage between specific mutations and disease severity, as well as to track the spread of mutations of interest within the population over time. In this study, we also identified differences in transmission dynamics between the subgroups. We saw a variety of RSV-A strains introduced but not expanded over time, whereas we observed clustered transmission of RSV-B and suggestions of sustained RSV-B transmission over the same period. It is tempting to speculate that a previous period of strong RSV-A transmission could create an immunological backdrop that would favor RSV-B transmission. Genomic surveillance of RSV will remain important as new interventions become more common, to monitor for selective pressure and changes in transmission dynamics.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. Conceptualization: E. L., S. B., and J. H. C. Data curation: E. L., S. B., and C. S.-J. Formal analysis: E. L. Funding acquisition: M. S. and J. H. C. Investigation: S. B., C. S.-J., and Y. M. L. Resources: Y. M. L. and M. S.. Supervision: M. S. and J. H. C. Visualization: E. L. and S. B. Writing–original draft: E. L., S. B., and J. H. C. Writing–review and editing: All authors.

Acknowledgments. We thank Joseline Velasquez-Reyes and Michelle Nguyen for helpful comments and discussion. We gratefully acknowledge all data contributors (ie, the authors and their originating laboratories responsible for obtaining the specimens, and their submitting laboratories) for generating the genetic sequence and metadata and sharing via GISAID, on which the global comparison and phylogeny of this research are based.

Data availability. Raw sequencing data are available from the Sequencing Read Archive within BioProject PRJNA1223358. Genome consensus sequences are available from GISAID (accession numbers EPI_ISL_19725019 to EPI_ISL_19725072). Code used for analysis can be found at github.com/neidl-connor-lab/rsv-2024.

Disclaimer. The contents of this work are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention (CDC).

Financial support. Sample and data collection was funded by the Massachusetts Consortium for Pathogen Readiness (Mass-CPR) (to M. S. and J. H. C.). This publication was supported by the Office of Advanced Molecular Detection, CDC (cooperative agreement number CK22-2204 to J. H. C.).