Immune Imprinting Drives Human Norovirus Potential for Global Spread

ABSTRACT Understanding the complex interactions between virus and host that drive new strain evolution is key to predicting the emergence potential of variants and informing vaccine development. Under our hypothesis, future dominant human norovirus GII.4 variants with critical antigenic properties that allow them to spread are currently circulating undetected, having diverged years earlier. Through large-scale sequencing of GII.4 surveillance samples, we identified two variants with extensive divergence within domains that mediate neutralizing antibody binding. Subsequent serological characterization of these strains using temporally resolved adult and child sera suggests that neither candidate could spread globally in adults with multiple GII.4 exposures, yet young children with minimal GII.4 exposure appear susceptible. Antigenic cartography of surveillance and outbreak sera indicates that continued population exposure to GII.4 Sydney 2012 and antigenically related variants over a 6-year period resulted in a broadening of immunity to heterogeneous GII.4 variants, including those identified here. We show that the strongest antibody responses in adults exposed to GII.4 Sydney 2012 are directed to previously circulating GII.4 viruses. Our data suggest that the broadening of antibody responses compromises establishment of strong GII.4 Sydney 2012 immunity, thereby allowing the continued persistence of GII.4 Sydney 2012 and modulating the cycle of norovirus GII.4 variant replacement. Our results indicate a cycle of norovirus GII.4 variant replacement dependent upon population immunity. Young children are susceptible to divergent variants; therefore, emergence of these strains worldwide is driven proximally by changes in adult serological immunity and distally by viral evolution that confers fitness in the context of immunity.

ABSTRACT Understanding the complex interactions between virus and host that drive new strain evolution is key to predicting the emergence potential of variants and informing vaccine development. Under our hypothesis, future dominant human norovirus GII.4 variants with critical antigenic properties that allow them to spread are currently circulating undetected, having diverged years earlier. Through large-scale sequencing of GII.4 surveillance samples, we identified two variants with extensive divergence within domains that mediate neutralizing antibody binding. Subsequent serological characterization of these strains using temporally resolved adult and child sera suggests that neither candidate could spread globally in adults with multiple GII. 4 exposures, yet young children with minimal GII.4 exposure appear susceptible. Antigenic cartography of surveillance and outbreak sera indicates that continued population exposure to GII.4 Sydney 2012 and antigenically related variants over a 6-year period resulted in a broadening of immunity to heterogeneous GII.4 variants, including those identified here. We show that the strongest antibody responses in adults exposed to GII.4 Sydney 2012 are directed to previously circulating GII.4 viruses. Our data suggest that the broadening of antibody responses compromises establishment of strong GII.4 Sydney 2012 immunity, thereby allowing the continued persistence of GII.4 Sydney 2012 and modulating the cycle of norovirus GII.4 variant replacement. Our results indicate a cycle of norovirus GII.4 variant replacement dependent upon population immunity. Young children are susceptible to divergent variants; therefore, emergence of these strains worldwide is driven proximally by changes in adult serological immunity and distally by viral evolution that confers fitness in the context of immunity. IMPORTANCE In our model, preepidemic human norovirus variants harbor genetic diversification that translates into novel antigenic features without compromising viral fitness. Through surveillance, we identified two viruses fitting this profile, forming H uman norovirus is the causative agent of ;20% of all acute gastroenteritis episodes worldwide (1). Although all age groups are infected, the most significant impact of disease is among young children and the elderly, who together bear the highest mortality rates, estimated at more than 200,000 deaths/year (2). This global disease burden has motivated the development of norovirus vaccines, with two candidates currently in clinical trials (3)(4)(5). Vaccine development is hampered by extensive diversity within the .35 known norovirus genotypes that infect humans (6). Recent progress in identifying key correlates of protection such as histo-blood group antigen (HBGA)-blocking antibodies (Ab) and the development of cell culture systems for several norovirus strains has accelerated vaccine development. Antigenic diversity within the GII.4 genotype, serotype/epitope-level cross-immunity, lack of understanding of the drivers of strain emergence, and the absence of a robust small animal model remain significant hurdles. Of these, antigenic diversity within the GII.4 genotype is the most significant obstacle to vaccine development (7)(8)(9). Norovirus global resurgence events occur cyclically and correspond with emergence of novel GII.4 variants characterized by unique constellations of residues within neutralizing antibody epitopes encoded within the capsid protein (VP1) (7,(9)(10)(11). Virus recombination between the genes that encode the RNA-dependent RNA polymerase (ORF1-NS7) and viral capsid (ORF2) types may also contribute to variant circulation and persistence in human populations (12)(13)(14).
Phylotemporal analysis of norovirus sequences suggests that preemergent GII.4 viruses with novel epitopes critical to immune escape circulate at endemic levels before becoming dominant, specifically in young children with limited GII.4 cross-protective immunity (15)(16)(17). Prior to global spread, these preemergent viruses would appear on phylogenetic trees at the end of atypically long branches. Population immunity would then determine which variants would successfully emerge and become dominant. As the most frequent cause of viral acute gastroenteritis outbreaks in the 21st century, six GII.4 norovirus variants have predominated worldwide in the past 25 years (18)(19)(20), resulting in significant burdens on health care, education, and military units (21)(22)(23) (24). Why this variant has persisted in the population for a decade is a fundamental question the field has yet to answer.
Here, based on preemergence diversification, we identified two candidate viruses fitting the phylogenetic characteristics of preepidemic variants and evaluated their emergence potential by comparing their susceptibilities to neutralization by sera from children and adults. Sera from young children were largely variant specific across time, whereas neutralization potency of sera from adults was broader with evidence of backboosting to previously dominant GII.4 variants. These data support the idea that Sydney 2012 persistence may be driven by a lack of durable Sydney 2012-specific protective immunity in adults. Further, we found that continued exposure to GII.4 Sydney 2012 and antigenically related viruses results in a broadening of the immune response to include viruses characterized as potentially preemergent by phylogenetic analyses, potentially limiting new variant emergence. Consequently, we propose a model in which emerging variant selection is likely dependent upon the dynamics of the serological repertoire of adult populations with preexisting immunity, shaped and then reshaped by sequential GII.4 exposures (15,25,26).

RESULTS
Identification of highly divergent GII.4 variants by capsid gene sequence. To search for viruses with preemergent signatures, we analyzed 930 GII.4 norovirus ORF2 sequences from samples collected in the United Kingdom between 1994 and 2019; all possessed full complementary metadata (including age, gender, geography, and date of sample). Two taxa with atypically long branches were observed in the phylogenetic tree ( Fig. 1; see also Fig. S1 in the supplemental material). The first, GII.4 Hong Kong 2019, demonstrated a rate of evolution in accordance with the remainder of the data (Fig. S2). It is most closely related to the GII.4 Osaka 2007 variant, although it diverged from this lineage well before the Osaka 2007 epidemic, in keeping with characteristics observed in known preepidemic variants (15) (Fig. S3). The second, GII.4 Den Haag 2017, emerged from within the GII.4 Den Haag 2006 cluster (Fig. S3).
Den Haag 2017 exhibits few changes in characterized antigenic regions (Fig. S4). It has 5 unique residues (294P, 339R, 340S, 378R, and 413A) relative to the 155 Den Haag 2006 strains in our data set; however, only 378R in antigenic site C is unique relative to all other strains in our data set. In contrast, Hong Kong 2019 possesses 16 unique antigenic site residues relative to the 10 Osaka 2007 strains in our data set (Fig. S4). Of these, six in antigenic sites A (298Q), D (393E, 395P, and 397F), and G (355A and 364N) are unique relative to all other strains in our data set. Unique combinations of residues in these sites are known to correlate with immune escape and widespread GII.4 variant emergence. Monitoring changes at known antigenic sites alone is insufficient to Immune Imprinting Drives Norovirus Spread Potential mBio predict strain emergence. Indeed, multiple GII.4 variants with differences within known neutralizing/blockade Ab antigenic sites have been shown to cocirculate (15), yet only six became predominant and spread globally (18)(19)(20). Thus, we developed a phenotype-based testing pipeline to evaluate the impact of genotype changes. All GII.4 variants that have spread globally share two features: novel neutralizing/ligand binding blockade antibody antigenic site profiles and binding to multiple cellular ligands (8,20,27,28). Human noroviruses use ABO histo-blood group antigens (HBGAs) as cellular binding ligands (29,30). HBGA expression varies between and within human populations; thus, the ability to bind to a diverse set of ligands corresponds to a potentially large susceptible population. The HBGA ligand binding pocket that binds the common glycan among ABO antigens is highly conserved across GII.4 variants; however, variation within residues 391 to 397 (antigenic site D) provides stabilizing interactions with ligands that impact variant avidity for different ABO HBGAs as well as neutralizing antibodies (7,8,(31)(32)(33)(34).  (35)(36)(37). Retention of binding to B antigen is likely mediated by 393N as this residue has been previously shown to stabilize GII.4 binding to B antigen (8), although residues in antigenic site A may also contribute to HBGA binding affinity. No other tested GII.4 variants required bile for binding Evaluation of divergent variant emergence potential based on population serum antibody reactivity. Among individuals genetically susceptible to GII.4 infection (those who express ABO antigens on mucosal surfaces), human norovirus GII.4 global outbreak potential depends on the presence of key residues in neutralizing Ab (NAb) antigenic sites that allow the virus to escape prevailing population immunity (7,8,25,27,38,39). To evaluate these findings in the context of more contemporary population immunity, we made the same comparison with sera collected from healthy adults in 2017 to 2019 (n = 19) (Fig. 4C) and young children with a confirmed first GII.4 infection The breadth of antigenic divergence of the novel variants depends upon the exposure history of the population. To better understand and quantify the antigenic relationships between the GII.4 variants over time and between sets of sera, we used antigenic cartography to visualize the antigen distance (AD) of the viruses to each other compared to each serum set (43)(44)(45). One AD equals a 2-fold change in blockade antibody response. A 4-fold change is associated with GII.4 norovirus immune escape in immunized mice (9,46). We performed a combined analysis including all four sets of sera and all 5 variants. Figure 4E to H shows the antigenic relationships where we have illustrated each serum set separately for visual clarity. The serum sets are shown together in Fig. S7A. Representation of the data in two dimensions (2D) recapitulated the distance matrix generated from 50% inhibitory dilution (ID 50  As with the surveillance sera, we employed antigenic cartography to better understand and quantify the antigenic relationships between the viral variants relative to the three outbreak sera in 2D (Fig. 5D to F; Fig. S7B). 2D representation recapitulated the distance matrix generated from ID 50 values as well as, if not better than, the 3D mapping ( Fig. S8B; both R 2 = 0.65 and 0.60, respectively).
The variants cluster into two distinct groups: US95/96 and Farmington Hills 2002, and Den Haag 2006, New Orleans 2009, and Sydney 2012. The antigenic distance between the two clusters (AD range of 2 to 3) is larger than that within each group (AD range of 0 to 1). The antigenic dissimilarity between these two clusters is in line Taken together with the results from nonoutbreak sera, these data illustrate the complexities of population responses, suggesting that population immunity to human norovirus broadens by exposure and is shaped by preexposure histories with backboosting of previous immunity.

DISCUSSION
Many human norovirus GII.4 variants cocirculate globally without causing increased levels of disease. Global emergence and global predominance of GII.4 noroviruses likely require use of a broad selection of binding ligands with change in a combination of viral characteristics (novel neutralizing epitopes) and host factors (susceptible genetic background and ineffective population immunity). By applying the metrics described here for evaluating the potential for immune evasion and broad host range to viruses selected for sequence divergence, we have established a potential pipeline for predicting the potential of GII.4 human noroviruses to spread globally. Similar approaches may also be applicable to other acute viral infections with cocirculating variants that lack tractable laboratory systems for functional analyses (such as in vitro culture). Although both Den Haag 2017 and Hong Kong 2019 have unique residue changes in antigenic sites, the limited number of changes of Den Haag 2017 compared to other Den Haag 2006 strains, its emergence from within the Den Haag 2006 cluster, and its weak interaction with cellular ligands may hinder the potential for Den Haag 2017 to escape from population immunity and efficient transmission. While these data suggest that Hong Kong 2019 is better poised to infect a broad population, to emerge as a dominant variant Hong Kong 2019 would need to evade current population immunity. This variant is able to evade antibody in adult and child sera from 2013 to 2014 and child sera from 2018 to 2019. In contrast, adult sera from 2017 to 2019 contain effective blocking antibodies. This may explain why, although Hong Kong 2019like strains have occasionally been detected in Asia and Europe, including the United Kingdom, they have failed to increase in numbers. Den Haag 2017 is more effectively blocked by all sera, which may explain why we find no evidence of any similar sequences among our own collection or in any public database.
Understanding the host and virologic factors that facilitate the transition of a GII.4 variant from causing low-level disease in children/asymptomatic infection in adults to a variant that spreads globally, causing symptomatic disease in adults, would increase the reliability of predicting which variants will spread globally and inform vaccine development. GII.4 variants cocirculate and evolve for years, specifically in children (15)(16)(17), and it is host factors that appear to be the major driving force that discriminates between variants that cause endemic disease versus those that spread globally. Adults likely become susceptible to symptomatic infection when either antibody neutralization titers wane at mucosal sites or infection reshapes the serological repertoire of Immune Imprinting Drives Norovirus Spread Potential mBio circulating neutralizing/blockade antibodies and provides an avenue for the emergence of a preexisting, epitope-divergent variant with good fitness. Thus, GII.4 variant threat prioritization lists will vary over time and depend on immunity in adults with extensive preexposure history and back-boosted memory antibody responses to human norovirus. Analyzing sera from exposed and recently infected children and adults demonstrates the complex immune profile that develops from single and sequential exposures to norovirus GII.4 variants. Following the first GII.  (7,28,52). Notably, independent serological and cartographic analyses of GII.4 outbreak sera with known exposure history verify the pattern of broadening immunity and back boosting described in surveillance sera, suggesting that the above observations do not result from selective waning of antibody responses postinfection but instead reflect the complex dynamics of changing population immunity to human norovirus. Young children should be the primary target for norovirus vaccination if the aim is to prevent infection and decrease burden of disease (53)(54)(55). To be most effective, infants with limited GII.4 exposure will need to be immunized. Here, we show that sera from children are less cross-reactive than sera from adults, particularly following a first GII.4 infection. This finding fits with data showing that younger infants and single-variant-immunized mice have been shown to respond primarily to the infecting variant (9,46,56,57). Greater breadth of immunity likely develops over time as a consequence of multiple infections. How the order or frequency of these infections impacts antibody responses and subsequently serologic and mucosal immunity is unknown but will affect the cross-protection patterns and vaccine formulation requirements as new variants emerge and population immunity shifts. The correlation between serological and mucosal immunity for norovirus and the impact of each on protection against infection and disease is not well studied, especially considering the disease burden.
Incorporating population immunity into a conceptual model of GII.4 variant global emergence likely improves the accuracy of the results but also has challenges. Accurate prediction will require the establishment of a large panel of reagents for the testing pipeline, which is especially complex within the context of continually evolving population immunity and necessitates a continued requirement for contemporary human samples or complex immunization schemes for animals. Here, we assembled a unique resource of sera comprised of both year-matched archived and year-matched contemporary sera. However, we were unable to match the serum sets demographically. Specifically, the first GII.4 infection child sera were collected in a low-to middle-income country with less genetic diversity than the source of the other sera, the United Kingdom and United States. While these features likely impact the likelihood of GII.4 exposure or infection for a child, they are unlikely to directly explain the differences in serum breadth discussed here, as only individuals with serological evidence of GII.4 infection (genetically susceptible) were included in the analyses and the first infection breadth of blockade antibody responses in the children is similar to those reported after single GII.4 immunization of animals and those seen in adults after ancestral GII.4 infection (8,27).
It also remains uncertain for how long variants might circulate before emergence. It is likely that regular and repeat modeling of adult and child sera will be necessary to differentiate between antibody cross-reactivity and preexposure with defined molecular diagnostics techniques. The roles that adult parents of young children play in accelerating or mitigating transitions of new emergent variants to global outbreaks will also need to be defined. Further, we cannot exclude at this time the possibility that the increase in the numbers of children in the population who are susceptible to a particular GII.4 variant (e.g., Hong Kong 2019), relative to adults with immunity, may itself increase the chance of a global spread of a variant. How these factors interplay with the fact that children initially exposed to Sydney 2012 will have different antibody set points for back-boosting and cross-reactivity has yet to be determined but will guide both the spread of new GII.4 variants and vaccine component requirements in the future. Several of these issues could be addressed by implementing national surveillance systems not only for outbreaks but also for sporadic norovirus cases as well as environmental samples that may better reflect the diversity of infection. These questions apply not only to human norovirus but also to other RNA viruses under selective pressure.
The physical isolation brought on by the COVID-19 pandemic has decreased hospitalizations for influenza, respiratory syncytial virus (RSV), rotavirus, and norovirus infections (58). Norovirus reports decreased by an estimated 40 to 79% during the COVID-19 pandemic (59)(60)(61). These stark declines support the potential effectiveness of nonpharmaceutical interventions at controlling norovirus transmission and emphasize the impact that effective surveillance and control measures (e.g., alert systems linked to public health messaging) could have in the absence of a licensed vaccine. Fewer human norovirus infections during the COVID-19 pandemic imply that population-wide antibody titers have likely waned, potentially setting the stage for emergence of a new GII.4 variant (62).

MATERIALS AND METHODS
Variant data set. Nine hundred thirty GII.4 norovirus samples, with full metadata, collected and sequenced as part of the NoroPatrol Surveillance Project were analyzed here.
Fecal sample collection. Fecal specimens from norovirus RNA-positive patients were referred from outbreak events detected by local and regional National Health Service (NHS) and Public Health England (PHE) laboratories to the National Enteric Virus Reference Laboratory at PHE Colindale. All samples were confirmed norovirus RNA positive by real-time reverse transcription-PCR (RT-PCR) (63) and genotyped by partial capsid sequencing, targeting either region C (64) or the hypervariable P2 domain (65) in open reading frame 2 (ORF2) of the norovirus genome, in accordance with local standard operating procedures.
Fecal specimens were prepared as 10% (wt/vol) suspensions in balanced salt solution (minimal essential medium; Life Technologies), and viral RNA was purified from the supernatants of clarified suspensions using the QIAsymphony, operating the Complex200_v4 protocol with the DSP virus/pathogen kit (Qiagen), eluting RNA in a final volume of 60 mL. Quantity of norovirus RNA in each sample was estimated by real-time RT-PCR to establish a cycle threshold (C T ) value, and samples with a C T value of ,35 were selected for whole-genome sequencing.
Library preparation and whole-genome sequencing. SuperScript IV (Life Technologies) was used to synthesize single-stranded cDNA from RNA samples. This was followed immediately by second-strand synthesis using the NEBNext mRNA second-strand synthesis module (New England Biolabs) to generate double-stranded cDNA which was purified using AMPure XP magnetic beads (Beckman Coulter). Libraries were prepared using the SureSelect XT low-input reagent kit (Agilent). Briefly, the cDNA was sheared into fragments of ;300 bp using the E220 focused ultrasonicator (Covaris), after which ends were repaired, adenosine tails were added, and adapters were ligated. Adapter-ligated libraries were amplified, incorporating a unique index into each library to allow multiplexed pooling. Libraries were then hybridized with custom-designed biotinylated norovirus RNA baits, and baits were captured using streptavidin-coated magnetic beads. After a second amplification step, the final libraries were quantified, pooled in equimolar amounts, and sequenced on a MiSeq sequencer (Illumina) using a V2 500-cycle kit for 250-bp paired-end reads.
Mapping. Raw read data in paired-end Fastq files were first adapter trimmed, followed by removal of low-quality (,Q30) reads. The reads that passed quality control were then mapped to a large custom data set of norovirus genotype references. From the multimapping BAM file we selected the best reference for the data set based on the number of reads mapping to it and the completeness of the genome. In cases of mixed infection, multiple top references were selected. The reads were then remapped to the best reference or the best set of references (in cases of mixed infection), duplicate reads were removed, and then consensus sequence was called at a minimum depth of 10 reads. Complete analysis was carried out using CLC Genomics Workbench v 11.01. The consensus sequences (Fasta format) are then genotyped using the norovirus genotyping tool hosted at https://www.rivm.nl/mpf/typingtool/norovirus/.
Phylogenetic analysis. The maximum likelihood tree of the alignment was constructed using RAxML (66) with the GTR model and 750 bootstrap replicates. The temporal signal of the data set was explored using TempEst (67). The optimum tree root was identified by maximizing the R squared correlation coefficient. Trees were visualized using ggtree (68)(69)(70) Serum samples. Sera collected in the United Kingdom were obtained from two sources: sera from children 1 to 2 and 15 years of age were obtained from the Public Health England Seroepidemiology Unit (PHE SEU), and sera from adults collected in the 2012-2014 FluWatch were obtained from the Health Survey for England Biobank (41). The use of coded serum samples was approved by the National Health Service Research Ethics Committee (reference 17/EE/0269), London School of Hygiene and Tropical Medicine (reference LEO12196), University of North Carolina at Chapel Hill (18-0214), and CDC (IRB no. 5051). Sera collected in 2017 to 2019 from adults were purchased from BioIVT (Hicksville, NY). Sera were obtained from a birth cohort study in León, Nicaragua, following the subjects' first reverse transcription-quantitative PCR (RT-qPCR)-confirmed symptomatic Sydney 2012 infection. The study was approved by the Institutional Review Boards of the National Autonomous University of Nicaragua, León (UNAN-León, Acta Number 45, 2017) and the University of North Carolina at Chapel Hill (study number . Children experiencing their first GII.4 infection were a median of 12 months old (IQR, 7 to 13 months), and the convalescent-phase sera were collected less than 7 months after GII.4 infection. Outbreak sera were selected from GII.4 outbreaks between 1988 and 2014 based on year and norovirus strain. Convalescent-phase sera were selected from norovirus RT-qPCR-confirmed cases. The epidemiological and demographic data available are summarized in Table S1 in the supplemental material. All sera were received coded with no link back to donor identification and were heat inactivated for 30 min at 56°C before use.
Structural homology modeling. Blockade/neutralizing antibody antigenic sites were mapped onto the surface of Sydney 2012 P domain dimer (PDB 4wzt) and with the PyMOL Molecular Graphics System, version 2.0 (Schrödinger, LLC).
Blockade of VLP-ligand binding assays. Ligand binding blockade antibody assays were performed as described previously (73,74). VLPs were pretreated with decreasing concentrations of sera for 1 h and then transferred to ligand-coated plates for 1 h, and bound VLPs were detected as for ligand binding assays. PGM (10 mg/mL) was the ligand source for the Den Haag 2006, Osaka 2007, and Hong Kong 2019 assays. B saliva supplemented with 1% bile was the ligand source for the Den Haag 2017 assay (35,36). The percent control binding was compared to that for no serum pretreatment. Mean 50% inhibitory dilution (ID 50 ) titers and 95% confidence intervals (95% CIs) were determined from log(inhibitor) versus normalized response-variable slope curve fit in GraphPad Prism 9.1.2 (26,51). Sera that did not block at least 50% of VLP binding to ligand at the lowest dilution tested were assigned a titer equal to 0.5 times the lowest tested dilution for statistical analysis.
Antigenic cartography. All ID 50 titer measurements were used in the process of antigenic cartography. The antigenic map in 2D and 3D was made with Racmac software implemented in RStudio, which implements the methodology as in reference 43. Shaded polygons are defined using function geom_encircle (https://CRAN.R-project.org/package=ggalt). The data visualizations were created using tidyverse, ggplot2, ggalt, and ggforce packages (https://CRAN.R-project.org/package=ggalt, https://ggplot2.tidyverse.org/). Statistical analysis. Blockade antibody titer and ligand binding statistical analyses were performed using GraphPad Prism 9.1.2 (7,26). ID 50 values were determined by nonlinear curve fit of normalized data with variable slope (7,26). ID 50 values were log transformed for analysis and compared by Wilcoxon matched-pairs signed-rank test when comparing VLPs for a serum set or by Mann-Whitney test when comparing serum sets for the same VLP. Ligand binding data were fit with log(inhibitor) versus response-variable slope (four parameters) for EC 50 calculation or maximum binding and dissociation constants by single-site binding curve analysis in GraphPad Prism 9.1.2 (75). A difference was considered significant if P was ,0.05.
Data availability. The sequence of Den Haag 2017 and Hong Kong 2019 are available on GenBank with accession codes OK376714.1 and MT742777.1 respectively. ID50 values, coordinates in antigenic space and code used to generate figures are available on GitHub (https://github.com/ftettamanti/ norovirus-antigenic-cartography).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank the Microscopy Services Laboratory, Department of Pathology and Laboratory Medicine, University of North Carolina Chapel Hill, and Shilpi Sheth, LSHTM, for her support of the genomics work through retrieval of specimens and preparation and testing of nucleic acid extracts.
The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.