Within-Host Rhinovirus Evolution in Upper and Lower Respiratory Tract Highlights Capsid Variability and Mutation-Independent Compartmentalization

Abstract

Background

Rhinovirus (RV) infections can progress from the upper (URT) to lower (LRT) respiratory tract in immunocompromised individuals, causing high rates of fatal pneumonia. Little is known about how RV evolves within hosts during infection.

Methods

We sequenced RV complete genomes from 12 hematopoietic cell transplant patients with infection for up to 190 days from both URT (nasal wash, NW) and LRT (bronchoalveolar lavage, BAL). Metagenomic and amplicon next-generation sequencing were used to track the emergence and evolution of intrahost single nucleotide variants (iSNVs).

Results

Identical RV intrahost populations in matched NW and BAL specimens indicated no genetic adaptation is required for RV to progress from URT to LRT. Coding iSNVs were 2.3-fold more prevalent in capsid over nonstructural genes. iSNVs modeled were significantly more likely to be found in capsid surface residues, but were not preferentially located in known RV-neutralizing antibody epitopes. Newly emergent, genotype-matched iSNV haplotypes from immunocompromised individuals in 2008–2010 could be detected in Seattle-area community RV sequences in 2020–2021.

Conclusions

RV infections in immunocompromised hosts can progress from URT to LRT with no specific evolutionary requirement. Capsid proteins carry the highest variability and emergent mutations can be detected in other, including future, RV sequences.

We examine within-host RV evolution in immunocompromised patients and show a lack of specific evolutionary requirement for RV to progress from the upper to lower respiratory tract, along with a higher variability of the capsid proteins compared to nonstructural proteins.

Human rhinoviruses (RVs) constitute the most prevalent cause of the common cold but are also associated with exacerbations of preexisting airway diseases such as asthma and chronic obstructive pulmonary disease [1–3]. RV infection progression from the upper (URT) to lower respiratory tract (LRT) is well described in hematopoietic cell transplant (HCT) recipients [4–6]. Once LRT infection has developed, rates of RV-associated respiratory failure and fatal pneumonia are higher among HCT recipients and are comparable to those seen with other respiratory viruses, including respiratory syncytial virus, parainfluenza virus, and influenza virus [7, 8].

RVs belong to the Picornaviridae family and are divided into RV-A, RV-B, and RV-C species, which are themselves divided into genotypes based on either neutralization assays or phylogenetic classification. RVs consist of a positive-sense single-stranded RNA genome of approximately 7000 nt length that is organized in a single open reading frame that generates 4 capsid proteins (VP1–4) and 7 nonstructural proteins (2A–2C and 3A–3D). The VP1 nucleotide region serves as the basis for phylogenetic classification of RVs [1, 9]. A total of 169 RV genotypes have been identified (80 RV-A, 32 RV-B, and 57 RV-C types) which can cocirculate throughout the year, complicating both vaccine and drug development [2, 10, 11].

While studies have focused on RV species and genotype surveillance, there is scarce knowledge about how RV diversity is triggered within hosts and how the virus progresses from the URT to LRT in immunocompromised individuals [12–15]. In this study, we analyzed within-host RV evolution dynamics in 12 HCT recipients with prolonged infection by different RV species, comparing both URT and LRT specimens.

METHODS

Patient and Viral Load Testing

URT (nasal wash, NW) and LRT (bronchioalveolar lavage, BAL) samples were collected from allogeneic HCT recipients after being screened by reverse transcription quantitative polymerase chain reaction (RT-qPCR) during the routine follow-up at the University of Washington Medicine-Fred Hutchinson Cancer Center from 2006 to 2014. This study was approved by the Institutional Review Board at the Fred Hutchinson Cancer Center. Subjects signed informed consent permitting the use of data and samples for research.

Genome Sequencing and Phylogenetic Analysis

Genome sequencing and analysis methods are more extensively discussed in the Supplementary Material. Briefly, metagenomic next-generation sequencing (mNGS) libraries were prepared as described previously [16, 17]. In addition, 1 kb tiling amplicon-based sequencing libraries were used to obtain coverage in specimens where mNGS failed. Consensus genomes were called using the Revica pipeline (https://github.com/greninger-lab/revica) using a database containing all existing rhinovirus complete genomes from the International Committee on Taxonomy of Viruses [18]. Maximum likelihood trees were inferred with IQ-TREE version 2.1 using SH-aLRT test (1000 replicates) and UFBoot2 method (10 000 replicates) to evaluate reliability of sequence clusters [19–21].

Intrahost Single Nucleotide Variant Analysis

Intrahost single nucleotide variants (iSNVs) were called using the LAVA pipeline [22] using bwa-mem [23] with a minimum allele frequency (MAF) of 10% in the positions with a minimum sequence depth of coverage of 30 ×. All iSNVs are listed in Supplementary Table 1. Microbiome composition for URT and LRT specimens with mNGS data was assessed using Kraken2 software compared to the standard prebuilt genomes database (v20221209) [24] and the phyloseq library in RStudio [25] and plotted with ggplot2 library [26]. Capsid iSNVs were mapped onto the 3-dimensional viral capsid structure of RV-A16 (Protein Data Bank [PDB], 1AYM) using HHPred and visualized using Pymol (Supplementary Table 2) [27, 28]. HHPred was also used to map iSNVs onto the coxsackievirus B3 capsid (PDB, 4GB3) and correlated with published mutational fitness effect (MFE) measures using Mann-Whitney-Wilcoxon statistical test [29].

RESULTS

RV Genome Sequencing and Viral Genotype Assignment

A total of 12 HCT recipients with RV-positive specimens from 2006 to 2014 were included (Table 1). Except for a 1 year old, all subjects were adults (age range, 24–67 years; median, 46 years). Ten of the 12 individuals had low absolute lymphocyte count (<1000 cells/µL) and all had respiratory symptoms at the time of the initial RV infection detection (Table 1). In addition, all individuals developed proven or probable LRT infection, considered as RV detected in BAL despite the presence or absence of radiographic abnormalities. Ten of the 12 individuals died in the 5 months following the last RV-positive detection (except individuals S01 and S06).

Table 1.

Epidemiological Information Summary

Individual	Age, y	Sex	RV Genotype	Time From HCT to RV Infection, d	Absolute Lymphocyte Count, k/µL
S01	46	Male	C17	88	0.68
S02	1	Male	C36	273	0.55
S03	50	Male	B97	0	0.05
S04	50	Female	B06	246	0.57
S05	64	Male	A57	36	0.18
S06	34	Female	A105	5	1.49
S07	67	Female	A58	414	2.34
S08	37	Male	A78	162	0.50
S09	24	Male	A102	18	0.11
S10	41	Male	A39	181	0.11
S11	63	Male	A82	14	0.59
S12	44	Male	C28	108	0.00

The age and sex of each individual is detailed as well as the RV genotype detected. The time from the HCT to the first RV-positive sample detected is informed in days. The absolute lymphocyte count was evaluated at the time of the first RV-positive sample detected.

Abbreviations: HCT, hematopoietic cell transplantation; RV, human rhinovirus.

RT-qPCR cycle threshold (Ct) values were variable among samples and individuals (range, 17.5–34.3). mNGS failed to recover genomes in 7 specimens, which required amplicon-based sequencing for genome recovery (Table 2). Comparison of mNGS and amplicon-based sequencing methods in select samples showed high levels of consistency in variant allele frequencies recovered (r² > 0.96; Supplementary Figure 1). mNGS analysis revealed a coinfection with human coronavirus 229E (HCoV 229E) and metapneumovirus in specimen BAL10b.

Table 2.

RV Sequencing and Intra-host Single Nucleotide Variant Characterization

Individual	Sample ID	Collection Date	Days Since First Positive Sample	qRT-PCR Ct	iSNVs	All iSNVs Ratio	NS iSNVs Ratio	Sequencing Method	Coverage, %	NCBI SRA Accession Number
S01	NW1	24 Aug 2009	0	17.5	0	NA	NA	mNGS	99.2	SRX5167679
	BAL1	24 Aug 2009	0	22.3	0	NA	NA	mNGS	98.7	SRX5167676
S02	NW2	17 Jun 2008	0	31.8	2	NA	NA	mNGS	100	SRX5167677
	BAL2	18 Jun 2008	1	34.3	0	NA	NA	mNGS + PCR-tiling	97.2	SRX5167685
S03	BAL3	8 Mar 2010	28	29.5	16	3.4	2.1	PCR-tiling	98.7	SRX5167684
	NW3	9 Mar 2010	29	22.7	19	2.1	NA	mNGS + PCR-tiling	99.8	SRX5167680
S04	NW4	10 Jan 2008	8	30	32	0.9	1.4	PCR-tiling	100	SRX5167678
	BAL4	12 Jan 2008	9	24.1	18	1.9	3.1	mNGS + PCR-tiling	100	SRX5167689
S05	NW5	22 Sep 2006	15	27.4	4	1.5	NA	mNGS	100	SRX5167667
	BAL5	27 Sep 2006	20	23.3	11	1.3	1.5	mNGS	99.9	SRX5167690
S06	NW6	22 Sep 2006	71	33	9	3.0	6.1	mNGS + PCR-tiling	97.8	SRX5167668
	BAL6	27 Sep 2006	76	23	3	0.8	0.0	mNGS	99.8	SRX5167693
S07	BAL7a	28 Oct 2008	0	21.5	5	1.1	0.8	mNGS + PCR-tiling	96.3	SRX5167688
	NW7	10 Nov 2008	13	21.8	10	1.0	0.3	mNGS	100	SRX5167683
	BAL7b	4 Jan 2009	68	NA	19	1.8	1.5	mNGS + PCR-tiling	99.9	SRX5167687
S08	NW8	10 Jan 2008	0	20	3	NA	NA	mNGS	100	SRX5167681
	BAL8	23 Jan 2008	13	24	13	3.4	12.1	mNGS	99.9	SRX5167692
S09	BAL9a	23 May 2014	0	23.5	10	2.3	1.0	mNGS	100	SRX5167671
	BAL9b	4 Jun 2014	12	26.2	7	9.0	4.5	mNGS	100	SRX5167672
S10	BAL10a	3 Nov 2009	0	21.4	0	NA	NA	mNGS	100	SRX5167686
	BAL10b	12 May 2010	190	23	19	2.1	3.8	mNGS	99.8	SRX5167673
S11	BAL11a	19 Feb 2010	30	20.9	3	1.5	NA	mNGS + PCR-tiling	97.3	SRX5167691
	BAL11b	3 Apr 2010	73	NA	18	1.5	4.1	PCR-tiling	97.3	SRX5167675
	NW11	13 Apr 2010	83	24.7	28	3.8	25.9	mNGS + PCR-tiling	97.7	SRX5167682
	BAL11c	20 Apr 2010	90	NA	14	3.4	2.5	PCR-tiling	97.7	SRX5167674
S12	BAL12a	11 Jan 2013	72	23.8	12	1.5	NA	mNGS	99.7	SRX5167669
	BAL12b	16 Feb 2013	108	25.3	9	0.8	NA	mNGS	99.7	SRX5167670

For each individual, the upper (NW) and lower (BAL) sample is identified and the collection date, the days elapsed since the first RV positive and the RT-qPCR Ct value are described. For each sample, the total number of iSNVs detected is shown together with the calculation of the ratio of all iSNVs between the viral capsid region versus nonstructural proteins (all iSNVs ratio) as well as for nonsynonymous iSNVs only (NS iSNVs ratio). The sequencing methods used in each sample, the consensus genome coverage, and the NCBI SRA accession number to the FASTQ file are shown. Individuals S01–S06 were used for URT vs LRT analysis while individuals S07–S12 were used for longitudinal infection analysis.

Abbreviations: BAL, bronchioalveolar lavage; Ct, cycle threshold; iSNV, intrahost single nucleotide variant; mNGS, metagenomic next-generation sequencing; NA, not applicable; NS, nonsynonymous; NW, nasal wash; PCR-tiling, amplicon-based next-generation sequencing; RT-qPCR reverse transcription quantitative polymerase chain reaction.

Complete and near complete RV genomes obtained included 7 individuals with RV-A, 3 with RV-C, and 2 with RV-B species (Table 1). Phylogenetic analysis showed statistically supported clusters for the consensus sequences from each individual, yielding the same RV genotype across matched and longitudinal specimens with each individual having a different RV genotype (Table 2 and Supplementary Figure 2).

The number of iSNVs per sample did not show a statistical association with RV RT-qPCR Ct value nor with the number of days since the first positive sample (P > .05, Pearson correlation test). In addition, the number of iSNVs detected showed no association with URT versus LRT samples (ANOVA, P > .05). However, specimens containing RV-B genomes had significantly more iSNVs than those containing RV-A or RV-C sequences (ANOVA P = .00325, Fisher-least significant difference test).

For detailed iSNV analysis, individuals were divided into 2 study categories: (1) spatial niche genome evolution of URT and LRT respiratory specimens collected less than 5 days from each other (S01–S06; Table 2); and (2) longitudinal within-host RV genomic evolution from respiratory specimens collected across 10 to 190 days (median, 13 days) (S07–S12; Table 2).

No Consensus Genomic Evolution Required for RV Progression From Upper to Lower Respiratory Tract

To examine whether progression of RV infection from URT to LRT was associated with specific RV genomic evolution, we examined 6 cases where RV-positive URT and LRT specimens were obtained within 5 days of each other, consisting of 2 RV-A, 2 RV-B, and 2 RV-C infections (individuals S01–S06; Figure 1 and Supplementary Figure 3). RV consensus genomes had no coding changes between URT and LRT specimens for 5 of the 6 individuals. Moreover, the pair of RV-C17 genomes (S01; Figure 1A) had no iSNVs in samples taken on the same day and the pair of RV-C39 genomes contained only 2 iSNVs at 10% MAF in samples taken 1 day apart (S02; Supplementary Figure 3A). Individual S04 had 7 consensus changes including 4 coding mutations in the RV-B06 genome present in a BAL specimen taken 2 days after NW. All these mutations were present as iSNVs in the NW at MAFs between 20% and 45% (Figure 1C). Microbiome analysis of mNGS-sequenced specimens showed distinct microbial populations from URT and LRT specimens and a lack of oral flora present in LRT specimens, consistent with a lack of contamination between specimens that would affect URT versus LRT RV sequence analysis (Supplementary Figure 4). Taken together, these data are consistent with the lack of requirement for specific RV genomic evolution to allow infection of the URT and LRT.

Figure 1.

RV evolution in upper and lower respiratory tract samples in immunocompromised individuals. Allele frequency (in percentage) of each iSNV across the RV genome in individuals S01 (A), S03 (B), S04 (C), and S05 (D) is depicted. iSNV causing synonymous mutation are shown in blue, while nonsynonymous mutations are shown in red and include the amino acid change description. Upper (NW) and lower (BAL) respiratory tract samples are detailed in the upper right of each plot. In addition, the day when the sample was collected relative to the first RV positive and the maximum sequencing depth of coverage for each sample is shown. RV mature peptides are annotated at the bottom of each panel. The grey background plot illustrates the sequencing coverage profile for each sample. Regions with low sequencing coverage (<30 ×) are shown in pink. Abbreviations: BAL, bronchioalveolar lavage; iSNV, intrahost single nucleotide variants; NW, nasal wash; RV, rhinovirus.

Preferential Accumulation of Nonsynonymous iSNVs in RV Capsid Proteins

We next examined longitudinal evolution of RV genomes across 6 individuals (5 RV-A and 1 RV-C; individuals S07–S12 in Table 2). Given the lack of mutational changes present in closely matched URT and LRT specimens above, we utilized both URT and LRT specimens for this analysis. Samples were collected a median of 30 days after the first detection of RV in each individual (range, 0–190 days). iSNVs present in the first collected specimen increased in allele frequency to near fixation in 5 of the 6 individuals (Figure 2 and Supplementary Figure 5), with the sole exception being individual S10 who had no iSNVs in the day 0 sample and only a day 190 specimen available.

Figure 2.

Long-term RV infection dynamics in immunocompromised individuals. Minor allele frequency (in percentage) of each iSNV across the RV genome in individuals S07 (A), S10 (B), and S11 (C) is represented by dots. Blue dots indicate iSNV causing synonymous mutation and red dots indicate nonsynonymous mutation, which also includes the amino acid change description. Upper (NW) and lower (BAL) respiratory tract samples are detailed in the upper right of each plot. In addition, the day when the sample was collected relative to the first RV positive and the maximum sequencing depth of coverage for each sample is shown. RV mature peptides are annotated at the bottom of each panel. The grey background plot illustrates the sequencing coverage profile for each sample. Regions with low sequencing coverage (<30 ×) are shown in pink. Abbreviations: BAL, bronchioalveolar lavage; iSNV, intrahost single nucleotide variants; NW, nasal wash; RV, rhinovirus.

Examination of these 6 longitudinal infections revealed a significant accumulation of nonsynonymous iSNVs in RV capsid proteins compared to nonstructural proteins (Supplementary Figure 6). Across the 6 infections, a total of 65 nonsynonymous iSNVs in capsid proteins were detected compared to 23 in nonstructural proteins. When iSNVs from all 12 individuals were included, the median capsid to nonstructural ratio for nonsynonymous iSNVs was 2.3 (range, 0–25.9) and for total iSNVs was 1.8 (range, 0.8–9.0), after adjusting for locus length (Table 2).

Given the high proportion of RV-A samples in our study, we specifically modeled iSNVs located in the RV-A16 capsid structure (PDB, 1AYM) over 3 time periods relative to the first RV-A–positive sample: day 0, 12–30 days, and >68 days (Figure 3A). Capsid iSNVs accumulated in both number and allele frequency across time, appearing first in VP2 and VP3 proteins and later in VP1 (Figure 3A). Among RV-As, surface residues on the capsid were 4.0 times more likely to contain iSNVs compared to nonsurface residues (P = 4.7e-6, Pearson χ² test), and surface residues were 3.1 times more likely to contain iSNVs across RV-A/B/C (P = 1.3e-5). Allele frequency of capsid iSNVs showed a nonsignificant positive association with residue solvent accessibility surface area for VP1 and VP3, but not for VP2 (Supplementary Figure 7).

Figure 3.

Rhinovirus evolution dynamics during prolonged within-host infection. A, RV-A capsid variability. iSNVs modeled in the 3-dimensional crystal structure of the RV-A16 capsid (PDB accession number 1AYM). Proteins VP1, VP2, and VP3 are differentiated in the 3-dimensional model with different gray degrees. iSNV allele frequency is depicted on a scale from blue to red (from 10% to 100%). iSNV detection was categorized in day zero (first sample positive detected), after 12 to 30 days from the first RV positive and after more than 68 days from the first RV positive. B, Evaluation of the mutational fitness effect in a CVB3 capsid (PDB, 4GB3) RV-A, RV-B, and RV-C iSNVs. Violin plot of the average MFE in sites with detection of nonsynonymous iSNVs in the VP1, VP2, VP3, and VP4 capsid proteins across all 12 RV infections. P value of the ranked Wilcoxon test is displayed. C, Analysis of fixed within-host nonsynonymous iSNVs found in the immunocompetent community. Partial figures showed phylogenetic clades from maximum likelihood tree of RV-A complete genome alignment containing publicly available sequences sharing the high-frequency haplotype (denoted by a black circle). Complete phylogenetic tree is available in Supplementary Figure 8. Abbreviations: iSNV, intrahost single nucleotide variants; MFE, mutational fitness effect; PDB, Protein Data Bank; RV, rhinovirus.

iSNVs Are Enriched in Sites of Higher In Vitro Mutational Fitness Effect in VP2 and not Located in RV Epitope Sites

To understand potential causes or effects of emergent iSNVs, we compared iSNV presence with estimations of MFE from deep mutational scanning of the Coxsackievirus B3 capsid [29]. iSNVs sites from all RV species had a higher average MFE per residue VP2 (P value = .004) but not VP1, VP3, and VP4 (P value > .05, Mann-Whitney Wilcoxon) (Figure 3B). iSNVs were not preferentially located in neutralizing antibody epitopes previously mapped in RV-A2 and RV-B14 (P = .204, χ² test) [30, 31].

Newly Increasing Nonsynonymous iSNV Haplotypes Detected in Future Consensus RV Sequences

We next examined whether iSNVs that increased in allele frequency in our 2006–2014 specimens could be detected in other RV consensus sequences present in the NCBI GenBank. Sequence haplotypes that nearly fixed in 2 longitudinal RV-A infections from 2006–2014 were found in NCBI GenBank sequences identified in future years (Figure 3C and Supplementary Figure 8). For example, the RV-A58 haplotype (VP2, S76T; 2B, T85A; 3D, E143K) that nearly fixed in individual S07 from November 2008 was detected in 5 RV-A58 sequences from Seattle in 2021 as well as 1 pediatric infection from Wisconsin in 2009 [32]. Similarly, the RV-A82 haplotype (VP2, I28V; VP1, T188I) that fixed in individual S11 in April 2010 was also found in RV-A82 sequences taken in 2009 and 2010 in Wisconsin, and 2016 and 2020 in Seattle [32, 33]. Although iSNV VP2:N145K from S10 could also be detected in a 2021 RV-A39 sequence taken from Seattle, it did not have a close phylogenetic relationship with specimen BAL9b and failed to contain the full iSNV haplotype.

DISCUSSION

//Here, we compared the within-host genomic evolution of RV in URT and LRT specimens from 12 immunocompromised individuals after HCT. Overall, we found a lack of specific genetic changes associated with compartmentalization in URT versus LRT and a strong genetic plasticity in capsid proteins compared to nonstructural proteins. Capsid iSNVs were preferentially located in surface residues as opposed to core residues, but intriguingly these residues were not necessarily associated with neutralizing antibody epitopes or had particularly strong fitness effects in deep mutational scanning data from CVB3 capsid.

Global epidemiology of RV has shown that RV-A is the most frequently detected species in both immunocompetent and immunocompromised individuals [12, 15, 34, 35], matching our preferential recovery of RV-A in more than half of the infections profiled. In addition, no prevalence of a specific RV genotype was found, albeit in only 12 infections.

Intrahost evolution of RV infections has been less studied compared to infections with enveloped human respiratory viruses. Cordey et al (2010) examined intrahost evolution after experimental infection of RV-A39 over 5 days in fewer than 5 immunocompetent adults [36]. They found most low-level mutations occurred in VP2, VP1, and 2C regions and most consensus-level mutations occurred in VP1, VP2, and VP3 structural proteins. Tapparel and colleagues (2011) also found that structural proteins underwent a higher mutation rate during chronic RV-A and RV-B infections in 5 lung transplant recipients [37]. In addition, they found no ecological association between RV species and detection in upper and LRT specimens. Our results match these studies in finding preferential accumulation of mutations in the capsid region compared to nonstructural region in longitudinal infections. Our results also support the lack of association of LRT infection with any specific mutation or genotype, using RV sequencing data from matched LRT and URT combined with metagenomic analysis to ensure specimen integrity. Unlike Cordey et al [36], we detected mutations in the capsid drug-binding pocket—a hydrophobic cavity in the capsid conferring antiviral drug susceptibility—despite the patients having not received any antiviral treatment [38]. Specifically, for both individuals with RV-B, N198S in VP1 protein was detected. In addition, individual S04 had the mutation D200H in VP1 and individual S03 had the mutation I224M at VP1, which is a confirmed residue that interacts with an antiviral capsid inhibitor called pleconaril [11]. In RV-A, the emergent iSNV T260K located in the drug-binding pocket was detected in an RV-A57 sample of individual S05 [39]. It was reported that the amino acid identity in the drug-binding pocket is above 80% in RV-A and RV-B in comparison with the overall VP1 identity (70%–74% identity) [39]. We hypothesized that nonsynonymous iSNVs in the hydrophobic pocket may confer a variant advantage in the context of within-host viral competition causing structural changes that improve binding and/or uncoating of the genome during viral entry.

The lack of a specific evolutionary requirement for RV to change LRT/URT niche strongly indicates RV progression to LRT infection is largely due to nonviral factors, which is consistent with prior associations of LRT progression with intravenous immunoglobulin administration, steroid use, and low monocyte count [4]. While LRT samples could reflect URT secretions that were aspirated during the sampling procedure, we evaluated this bias by assessing a comparative microbiome analysis in the NW and BAL with metagenomic NGS. Microbiome profiles in BAL at advanced infection times showed high abundance of both commensal and opportunistic pathogens in the immunocompromised host (Stutzerimonas stutzeri, Brevundimonas, Alcaligenes, Ralstonia, and Rothia mucilaginosa) [40, 41].

Although limited to infections in only 2 individuals, our sequencing data supported a higher intrahost diversity of RV-B infections. Intriguingly, multiple studies have found RV-B infections to have lower viral loads and/or reduced symptom severity compared to RV-A or RV-C infections [42–44]. Our RV-B infections had higher Ct values compared to RV-A or RV-C infections (Table 2), which may result in higher allele frequencies in mixed infections [45]. The lower severity of RV-B infections may bias or complicate ascertainment of the initiation of infection and potentially lead to longer infections that allow greater intrahost diversity. Given the reduced frequency of RV-B infections, comparatively less effort has gone into understanding evolutionary rates and factors of RV-B and more work is required to understand interspecies differences in rhinovirus evolution.

Analysis of RV Ct values showed no association with the number of iSNVs, and therefore no relationship with a given population diversity pattern. In addition, no association was found between the number of iSNVs and the sample type or the days since the first positive sample, highlighting a complex infection dynamic without compartmentalization restriction during immunocompromised state. We also examined whether newly derived capsid mutations and sequencing haplotypes could be found in other RV genomes based on searches of NCBI GenBank. Many of these RV consensus genomes in NCBI GenBank are derived from our laboratory from 2021 to 2022 after screening the high number of coronavirus disease 2019 (COVID-19)–negative nasal swab specimens sent to our laboratory and sequencing the RV positives. Indeed, several newly derived RV mutations from our study patients could be found in future community sequences. While on the surface this matches results seen with sequencing longitudinal influenza A virus infections in immunocompromised individuals [46], given the lack of antigenic drift seen in RVs [47], these data could just be indicative of the limited number of sequences available and viable mutations for the capsid of a given RV genotype. Indeed, new intrahost mutations arising in 1 of our profiled individuals from 2010 were also seen in RV from Wisconsin in 2009–2010. Certainly, the amount of sequencing data on RV pales in comparison to that available for influenza virus, complicating confident determination of global allele frequencies of RVs, especially after dividing across all RV genotypes. Nevertheless, these findings show that in the context of lack or diminished immune pressure the within-host RV genetic drift can cause the emergence and fixation of viral haplotypes that may also occur during the immunocompetent community evolution. Indeed, profiling of intrahost evolution of oral poliovirus vaccine strains and interhost evolution of vaccine-derived poliovirus has shown antigenic evolution in both cases [48].

Our examination of the association between iSNV presence and mutational fitness effect in CVB3 deep mutational scanning data only recovered a modest relationship with iSNVs present in the VP2 protein and failed to find an association with iSNV allele frequency [48, 49]. Although these results suggest an important role for VP2 in the RV-A intrahost evolution, in vitro experiments are needed to confirm this hypothesis. The lack of association between mutational fitness effect and iSNV presence in other capsid proteins, as well as the lack of association with iSNV allele frequency, may be the result of competition between equally fit viral populations in an environment with a lower selection pressure, or simply due to the inability to accurately model and relate residues across different enterovirus species.

Our study was chiefly limited by the number of individuals and specimens examined. Individuals were not regularly sampled over space and time and not all remnant specimens were stored from the clinical care. Given the incredible diversity of RV genotype, our limited sampling constrains our ability to make generalizable conclusions about RV evolution. In addition, capsid structures and deep mutational scanning data are only available for select genotype or species [28, 29]. Although the first RV-positive sample was considered representative of the beginning of the infection, it is not possible to confirm this with certainty. However, clinical practice guidelines recommend patients report symptoms urgently and then undergo PCR testing. We also did not perform 2 independent library preparations for each specimen sequenced, although we controlled for this by demonstrating robustness of our library preparation and sequencing protocols for select specimens and using a relatively high iSNV allele frequency of 10% [45].

Overall, our work provides valuable insights into the within-host RV evolution, including the higher ratio of variability of the capsid proteins compared to nonstructural proteins, the mutation-independent compartmentalization of URT and LRT infections, and the identification of emerging within-host mutations in the immunocompetent community. Our data highlight the unique evolutionary constraints seen in RVs that are generally not present in infections with enveloped human respiratory viruses.

Supplementary Data

Supplementary materials are available at The Journal of 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

Financial support . This work was supported by the Department of Laboratory Medicine and Pathology of the University of Washington; and the National Institutes of Health (grant number K23 AI114844-02 to A. W.).

Data availability . Sequencing data obtained in this study are available at GenBank in NCBI BioProject PRJNA907865.