Nasal and systemic immune responses correlate with viral shedding after influenza challenge in people with complex preexisting immunity

Abstract

Each year in the United States, approximately 50% of adults ≥18 years are vaccinated against influenza viruses, with protective efficacy averaging 40.5% over the last 20 years. To model annual seasonal influenza, a cohort of 74 adults, who were unscreened for pre-existing A/H1N1 immunity and half of whom were recently immunized with licensed QIV (mean 64 days), were challenged with A/H1N1 influenza virus. Transcriptomic, proteomic, and VDJ repertoire analyses were performed on nasal and peripheral blood samples from participants to identify nasal mucosal and systemic immune responses that correlated with viral shedding as well as immune correlates of protection. Viral shedding participants showed increased T cell, but not B cell, VDJ diversity with expansion of low frequency B cell clones post-challenge, including broadly neutralizing motifs. Non-shedding participants demonstrated decreased clonality and increased richness of B and T cell VDJ clones, increased pre-inoculation nasal mucosal immune gene and serum protein expression, and increased ex vivo peripheral blood mononuclear cell responses. Nasal mucosal responses in participants shedding virus for two or more days showed higher early viral loads and exhibited stronger induction of antiviral responses compared with participants who shed virus for one day. Finally, participants with a single day of viral shedding were three times more likely to be female. These data shed light on the complex immune responses in the nasal mucosa and the periphery following influenza vaccination and infection, which will be critical for next-generation vaccine development.

One sentence summary

B and T cell receptor repertoires as well as nasal and systemic immune responses correlate with length of viral shedding after influenza challenge.

Introduction

Seasonal influenza A virus (IAV) infections are a substantial medical and economic burden worldwide, leading to an estimated 9 to 41 million illnesses, 120,000 to 710,000 hospitalizations and 6300 to 52,000 deaths annually in the US between 2010-2024 (1) . (The World Health Organization (WHO) estimates seasonal influenza virus results in 290,000 to 650,000 deaths annually worldwide (2). Moreover, the emergence of past pandemic IAV strains resulted in higher mortality, raising concerns about future potential IAV pandemics (3, 4).

Despite the non-systemic nature of human IAV infections, investigation of immune responses to vaccination and infection have focused on peripheral correlates of circulating antibody titers, primarily directed towards the surface glycoprotein hemagglutinin (HA) (5, 6). Compared with systemic responses, limited data are available for nasal mucosal immune responses to IAV and their relationship to peripheral responses and infection outcome. Nasal mucosal IgA titers are predictive of protection (7), and non-neutralizing polymeric secretory IgA plays a role in intersubtypic cross-protective immunity (8). Nasal mucosal-associated invariant T (MAIT) cells activate dendritic cells to promote humoral immunity (9). Immune cell residency in the nasal mucosa can also contribute to age-associated disease severity in respiratory infections (10). However, many of these studies involved animal experiments or in vitro and ex vivo approaches. In the absence of subsequent viral challenge, it is difficult to identify which responses are key to protective immunity. Human IAV challenge studies provide a unique opportunity to systematically characterize nasal protective immune responses in “real world” people with complicated immune history from past infections and vaccinations, which are not reproducible in animal models. This study highlights the relationships between nasal mucosal and peripheral immune responses in a population of influenza challenge participants with complex immunity (11) using a systems biology approach and improves our understanding of the correlates of protection for IAV infection and contagiousness in humans.

Results

Study design, shedding outcomes, and Poisson regression models of baseline nasal mucosal secretory IgA (SIgA) and serum hemagglutination inhibition (HAI) antibody titers

The goal of this study was to model annual seasonal influenza vaccination and H1N1 exposure in humans to characterize the immune corelates of viral shedding after influenza vaccination and challenge (11). Accordingly, clinical, nasal mucosal and peripheral blood responses were studied in a cohort of 74 healthy adults, enrolled independent of pre-existing A/H1N1 immunity (fig. S1). Half were recently vaccinated with licensed 2018 quadrivalent influenza vaccine (mean 64 days). Participants were then challenged with A/H1N1 virus (11) Vaccinated (n=37) and unvaccinated (n=37) participants were challenged with 10⁷ half-maximal tissue culture infectious dose (TCID₅₀) of a 2009 pandemic H1N1 strain, A/Bethesda/MM2/H1N1 (Fig. 1A). Following challenge, daily BioFire Respiratory Pathogen Panel (RPP) assays were performed on participant nasal wash samples, revealing 21 participants who always tested negative for H1N1 (termed Non-shedding; NS). Of this NS group, 14 were vaccinated (p=0.06) and 7 unvaccinated (p=0.98). For the 53 participants who tested positive by RPP on at least one day post-challenge, 23 were vaccinated (p=0.98) and 30 unvaccinated (p=0.06). The 53 shedders were broken down further into participants who were RPP(+) on a single day, termed One Day Shedding (1DS; n= 19), and participants who were RPP(+) for two or more days, termed Multiday Shedding (MDS; n=34). Geometric mean serum HAI titers and ranges were also measured (Fig. 1A). Vaccination and sex-linked correlated shedding outcomes were calculated for enrichment of the two compared classes using the corresponding one-sided Fisher exact test. Individual laboratory test results and clinical data are shown in data file S1. Analysis of odds ratio by Fisher’s exact test showed limited sex-linked differences in shedding outcome, with females were 3.1 times more likely to be 1DS than males (p=0.04) (Fig. 1A). To further understand the relationships between participant age, sex, vaccination status, and baseline (D-1) nasal mucosal SIgA and serum HAI titer with multimodal shedding outcomes, Poisson regression analysis was performed using age, sex, vaccination status, and antibody titer as variables with length of shedding as outcome (Fig. 1B). Both D-1 SIgA and HAI titer were significant predictors of multimodal shedding outcome, with coefficients of −0.1962 (p=0.033) and −0.0049 (p=0.0001), respectively. Although participant covariates of age and vaccination status were not predictive, sex was a significant predictor with a coefficient of 0.34 (p=0.048), indicating that males were likely to be MDS in this Poisson regression model.

Fig. 1: Study design with viral shedding outcomes and Poisson regression of baseline nasal SIgA and serum HAI titers.

(A) Healthy participants 18 to 45 years of age were enrolled independent of pre-existing immunity against challenge virus. Half the participants were previously vaccinated prior to challenge using licensed 2018 quadrivalent influenza vaccine (QIV) at the NIH Clinical Center (11). Nasal wash positivity for H1N1 challenge virus was determined by Biofire Respiratory Pathogens Panel (RPP). The p-values were calculated for the enrichment of the two compared classes using the corresponding one-sided Fisher exact test. (B) Poisson regression model of baseline (D-1) serum HAI and nasal mucosal SIgA titers. The models treat length of shedding as outcome, considers age, sex, vaccination status, and antibody titer as variables. For all models, antibody titers on D-1 were used. Reported estimated coefficients and p-values are shown; *p<0.05.

Baseline immune-related gene and protein expression status correlate with shedding outcomes

The effect of baseline (D-1) expression of nasal mucosal and peripheral blood immune-related genes on multimodal (NS, 1DS, MDS) shedding outcome was examined using Poisson regression models (p<0.01) (Fig. 1B) for baseline SIgA (fig. S2, data file S2) or baseline HAI (fig. S3, data file S3). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway classification was used to characterize the genes that correlated with shedding outcome. Baseline nasal mucosal gene expression pathways, including transcription and translation, metabolic, and T cell signaling and differentiation pathways correlated with shedding outcome, as did peripheral blood mononuclear cell (PBMC) gene expression pathways, including cell proliferation and metabolism pathways (fig. S4). Analysis of D-1 nasal mucosal and serum immune protein expression was performed to compare effects of vaccination, viral shedding, and viral shedding in vaccinated and unvaccinated participants (fig. S5). The majority of significantly (t-test p<0.05) differentially expressed nasal mucosal immune proteins were associated with vaccination, but not viral shedding (fig. S5A). In contrast, numerous differentially expressed proteins were identified in D-1 serum in NS participants. Many of these were involved in natural killer (NK) cell activation, including FLT3L and CXCl1, which were elevated in in vaccinated NS; in contrast, KLRD1, CLEC7A, FCRL6, DAPP1, CD6, GALNT3, STC1, CKAP4, uPA, and CD244 were elevated in unvaccinated NS (fig. S5B). No correlation between pre-challenge nasal SIgA or serum HAI titers and D1 nasal mucosal influenza virus M gene expression by Pearson or Spearman rank sum were observed, respectively (fig. S6A and B).

Vaccination status is associated with viral shedding outcomes and B and T cell VDJ repertoire diversity

The nasal mucosal and systemic molecular correlates of immunity were analyzed from a group of 74 participants (fig. S1) (11). The duration of viral shedding (0 to 12 days) with relative breakdown of vaccinated versus unvaccinated participants is shown in Fig. 2A. VDJ regions of B and T cells were sequenced by Adaptive Immunosequencing using genomic DNA isolated from PBMCs collected on D0 and D28 post-vaccination, prior to (D-1) challenge, and on D7, D28, and D56 post-challenge (fig. S7A). VDJ repertoire diversification was noted in B and T cells from pre- to D7 post-challenge with decreasing clonality [Wilcoxon Signed Rank p <0.001 for T cell receptor (TRBV) and p=0.02 for immunoglobulin heavy chain (IGHV); Fig. 2B] and increasing richness (p<0.001 for TRBV and p=0.03 for IGHV (fig. S7B, data files S4 and S5). From D7 to D28 post-challenge there was a focusing of the repertoire, with increasing clonality and decreasing richness (all p<0.001), consistent with clonal expansion events. Patterns were similar in vaccinated and unvaccinated cohorts, with no evidence of changes in repertoire diversity due to vaccination alone (Fig. 2B). B and T cell diversity metrics also correlated with shedding outcomes. Directly following H1N1 challenge, decreases in B and T cell clonality with concomitant increases in richness were observed in participants who did not shed virus (Fig. 2C). A decrease in T cell clonality and corresponding increase in richness was observed in participants with evidence of viral shedding post-challenge (Fig. 2C). However, there was no B cell diversity changes in participants with viral shedding (Fig. 2C), possibly indicating a lag in this part of immune response. Although antigen specificity cannot be directly assessed from sequencing data alone, there is evidence to support B and T cell response to IAV epitopes through expansion analyses. First, expansion of known broadly neutralizing B cell motifs in 84.2% of post-challenge samples was observed (fig. S7C). Second, in the vaccinated cohort, T cells that were identified as expanding after vaccination were observed to re-expand after H1N1 challenge (Fig. 2D and data file S6). Participants with expansion of T cell TRBV VDJ clones post-vaccination and D7 post-challenge were 28.6% of vaccinated NS, 60% of vaccinated 1DS, and 7.7% of vaccinated MDS, with vaccinated RPP(+) participants with T cell clones expanding post-vaccination and challenge being 18 times more likely (p=0.02) to belong to 1DS shedding class by one-sided Fisher exact test (Fig. 2E).

Fig. 2: Vaccination and H1N1 challenge affected shedding outcome and B and T cell VDJ repertoire diversity.

(A) Shown are the number of participants shedding virus each day post-challenge. Numbers of vaccinated and unvaccinated participants in each group are indicated by blue and orange bars, respectively. (B) T and B cell VDJ repertoire immunosequencing showing changes in Simpson’s clonality on D0 and D28 post-vaccination (Phase 1) and on D-1, D7, D28, and D56 post-challenge (Phase 2) in unvaccinated (orange) and vaccinated participants (11). (C) B and T cell VDJ analysis of D7 post-challenge clonality and down-sampled richness in viral non-shedders (teal) versus shedders (pink). (D) Shown are the number of initially expanding vaccine-associated T cell clones over time following vaccination and challenge. Data in (B to D) are presented as box plots, which show the minimum, the first quartile (Q1), the median, the third quartile (Q3), and the maximum Simpson’s clonality or number of initially expanding TRBV clones, respectively, with lines connecting individual participants values. All unadjusted Wilcoxon signed rank p-values <0.10 are shown in (B to D). (E) Association of post-QIV and post-H1N1 challenge T cell TRBV clonal expansion and shedding outcomes, as determined by Fisher’s exact test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

Nasal mucosal gene and protein responses correlate with viral shedding after IAV challenge

Comparisons of participants (n=74) without shedding and participants shedding virus for ≥1 days. Nasal mucosal RNA was analyzed by quantitative real-time polymerase chain reaction (qRT-PCR) to measure IAV matrix (M ) gene RNA and expression microarrays to measure gene expression responses prior to challenge (D-1) and at multiple time points (D1, 3 and 7) during the acute phase of IAV infection. Higher M gene RNA amounts were observed in MDS participants compared with NS participants on D1 and D3 post-challenge as determine by ANOVA (Fig. 3A).

Fig. 3: Nasal mucosal gene and protein expression correlate with viral shedding.

(A) Plot of IAV M gene RNA in nasal mucosal samples collected on D1 and D3 post-challenge; asterisks denote differences with p<0.05 by t-test. n=69 participants. (B) Heatmap showing kinetics of expression of shedding-associated genes on D-1, D1, D3, and D7 post-challenge and relationship to viral load and FluPro symptom scores (11) in a subset of vaccinated (blue) and unvaccinated (orange) RPP(+) and RPP(−) participants. Genes shown in yellow and blue are increased or decreased, respectively, relative to each participant’s D-1 baseline. n=15 participants. Increasing blue color indicates increasing FluPro score and increasing pink color indicates increasing IAV M gene expression. (C) Plots of proteins with increased expression in RPP(+) compared with RPP(−) on D3 or D5 post-challenge. (D) Plots of proteins with higher expression in RPP(−) compared with RPP(+) participants. Values in (C and D) were statistically significant (p<0.05) between the groups using standard t-test with Benjamini Hochberg correction. Scale of Olink Normalized Protein eXpression (NPX) is log₍₂₎. Box plots show the minimum, the first quartile (Q1), the median, the third quartile (Q3), and the maximum expression values. For (C and D), ⁺p<0.05 D3 only, ⁺⁺p<0.05 D5 only, ⁺⁺⁺p<0.05 D3 and D5, n=66 participants. (E) Kinetics of nasal mucosal protein expression. The intensity of the red color indicates the number of times the protein's NPX values are significantly differentially expressed (t-test p<0.01) at that timepoint in up to twelve pairwise comparisons of shedding and vaccination classes. n=66 participants. Cell type expression was determined using proteinatlas.org. MAIT, mucosal-associated invariant T cell.

Sequences significantly associated with viral shedding were identified by t-test (p<0.05) comparing participants that tested positive or negative for IAV in nasal washes at each timepoint post-challenge or by Pearson correlation with viral load on D3 post-challenge. No sequences were identified between all shedders and NS at D1 post-challenge, consistent with results from a previous study analyzing PBMC responses to acute IAV infection (12). Separate comparisons of MDS or 1DS participants to NS participants also failed to identify any sequences associated with viral shedding at D1 post-challenge. However, 49 sequences were identified as differentially expressed (2-fold difference in median expression value, p<0.05) between all shedders (1DS and MDS) and NS at D3 (data file S7). Expression of these viral shedding-associated sequences generally correlated with peak M gene viral load and daily FluPro symptom scores (Fig. 3B). Pearson correlation (ρ>0.6) also identified 81 sequences that correlated with nasal mucosal viral load (data file S8), some of which (n=16) overlapped with shedding sequences identified by t-test. Although the overlap between the two statistical tests was relatively small, the gene lists mapped to similar pathways (fig. S8). Most genes associated with viral shedding, identified by either t-test or Pearson correlation with viral load, mapped to antiviral processes, including interferon (IFN) signaling (including IFI6, IFI35, IFI44, IFIT1, IFIT2, IFIT3, Mx2, OASL, IL28A, DDX58, and ISG15). VS genes correlating with viral load also mapped to recruitment and activation of immune cells (including CXCL9, CX3CL1CXCL10, CXCL11, CCL3, CXCL9, TNF, CCL4, and CCL19).

Using a subset of nasal mucosal protein samples from participants (n=66) for which a complete time course was available, we identified inflammatory and immune response-related proteins that were significantly associated with shedding by t-test (p<0.05), comparing participants that tested positive or negative for IAV in nasal washes on D1, D3, and D5 post-challenge. Comparison of all shedders (1DS and MDS) to NS at D1 post-challenge failed to identify proteins associated with viral shedding. Comparisons of only 1DS or MDS participants to NS also failed to identify any proteins associated with viral shedding on D1 post-challenge. However, 29 proteins had different values (p<0.05) in all shedders (1DS and MDS) versus NS at D3 or D5 post-challenge (data file S9). CD8A, interleukin (IL)-6, CXCL11, TRAIL, PD-L1, CXCL10, IFN-γ, MCP-2, CX3CL1, and GLB1) were often more abundant in shedders at D3 and D5 (Fig. 3C) suggesting they were induced in response to viral replication. Others (IL-8, VEGF-A, OPG, CXCL1, TNFSF14, CXCL5, EN-RAGE, CXCL6, PRDX5 and ST1A1) were lower in shedders compared to NS, mainly at D5 (Fig. 3D). Higher amounts of these proteins in NS compared with shedders suggests a protective role against active viral replication.

Kinetic analysis of nasal mucosal protein expression and viral shedding was conducted by first ranking proteins significant associations (t-test p<0.01) between the expression at each timepoint with different shedding-related group comparisons. The expression kinetics of the top 10 ranked shedding-correlated proteins are shown (Fig. 3E and data file S10). Cell type expression of these top-ranked shedding-correlated nasal mucosal proteins and their secretome profile was determined using proteinatlas.org. These results showed increased expression of mucosal epithelial and innate immune cells, and adaptive immune cells. Many of these top ranked chemokines and cytokines (data file S10) are known to be secreted to blood, including IL-6, IFNG, MCP-2, CXCL10, CXCL11, TRAIL, CXCL9, CX2CL1, and IL-7 (Fig. 3E).

Shedding-correlated gene and serum protein expression responses differed between nasal mucosal and PBMC samples

Gene expression profiles were generated in PBMCs collected from participants (n=74), with 74 sequences differentially expressed (1.5-fold difference in median expression value, p<0.05) between all shedders (1DS and MDS) and NS on D3 post-challenge (data file S11). Comparison of the magnitude and kinetics of the respective shedding signatures in nasal mucosal and PBMC samples generally showed peak response at D3 for most MDS in both compartments (fig. S9). Gene ontology analysis of the respective nasal mucosal (n=49 sequences) and PBMC (n=74 sequences) shedding signatures identified by t-test (p-value < 0.05) showed enrichment of pathways associated with antiviral responses and macrophage activation in both compartments (Fig. 4A). Additional processes including the inflammasome pathway, pyroptosis signaling, and necroptosis signaling were uniquely enriched in the PBMC shedding-associated sequences; in contrast, granulocyte adhesion and diapedesis, granzyme B, STAT3, and immunogenic cell death signaling were enriched in nasal mucosa samples (Fig. 4A).

Fig. 4: Distinct nasal mucosal and peripheral blood responses correlate with viral shedding outcomes.

(A) Venn diagram of shedding-associated genes identified by t-test (p< 0.05) with Benjamini Hochberg correction) in expression microarray D3 post-challenge in nasal mucosal (n=74 participants) and PBMC samples (n=74 participants). Gene ontology shows pathway enrichment of shedding-associated sequences in either nasal mucosal or PBMC compartments. (B) Venn diagram of shedding-associated proteins identified by t-test (p< 0.05 with Benjamini Hochberg correction) in nasal (n=66 participants) and serum (n=74 participants) samples collected on D3 or D5 post-challenge. *Protein that is increased in mucosal samples and decreased in serum samples from shedders; **Protein decreased in mucosal samples and increased in serum samples from shedders. (C) Plots represent NPX expression values (log₍₂₎ of shedding-associated proteins in NS (teal) and shedders (pink) that are common to serum and nasal mucosal samples. Box plots show the minimum, the first quartile (Q1), the median, the third quartile (Q3), and the maximum expression values. ⁺p<0.05 D3 only, ⁺⁺p<0.05 D5 only, ⁺⁺⁺p<0.05 D3 and D5, n=66 participants. Scale of Olink NPX is log₍₂₎.

Comparison of nasal mucosal and PBMC shedding-associated genes revealed sequences that were common (n=19), or unique to either nasal mucosal (n=30) or PBMC (n=56) (Fig. 4A, data file S12). Many of the common were antiviral response genes, some with higher expression in the nasal mucosa (IFIT2, IF6, OAS2, IFIT3, ISG15, DD58, and OASL), whereas expression of others were comparable (fig. S10). Different sequences associated with IFN signaling were induced specifically in PBMC or nasal mucosa samples, reflecting differences in cell type-specific signaling and immune cell heterogeneity in blood. Unique sequences in nasal mucosal shedding signature included T cell recruitment (CXCL10, CXCL11), proliferation/survival (BATF3, IL15RA) and activation (GZMB), possibly representing the early stages of cell-mediated immune responses to IAV infection (data file S12). Other sequences involved in cell-adhesion (NEXN, HAPLN3) and proliferation (SAMD9, FBXO39) were also selectively induced in the nasal mucosa. Increased expression of negative regulators of IFN signaling (ETV7, GBP4) and cytokine signaling (SOCS1) at D3 suggest resolution of antiviral responses was initiated by this time, consistent with reduced expression of the general shedding signature by D7 (fig. S8). VS sequences unique to PBMC included those involved in macrophage function (DHRS9, MARCO, MS4A4A, SIGLEC1, IL15, PARP9, STAT1, and TNFSF10) and inflammasome signaling (AIM2, C2, and CASP1) (data file S12). Like the nasal mucosal VS signature, sequences associated with anti-inflammatory signaling (ATF3 and MT2A) and T and B cell function (IL15, EAF2, LY6E, MT1B, and TNFSF10) were also elevated in PBMC.

Inflammatory and immune response-related proteins in the serum (n=66 participants) significantly associated with shedding were identified by t-test (p<0.05) comparing participants that tested positive or negative for IAV in nasal washes on D1, D3 and D5 post-challenge. Comparison of all shedders (1DS and MDS) to NS at D1 post-challenge failed to identify proteins associated with shedding. Comparisons of only 1DS or MDS participants to NS also failed to identify any proteins associated with viral shedding on D1 post-challenge. However, 52 proteins had significantly (p<0.05) different expression in all shedders (1DS and MDS) versus NS at D3 or D5 (data file S13 and Fig. 4B).

Comparison of nasal mucosa and serum samples revealed common shedding-associated proteins (n=11) (Fig. 4B) Whereas the maximum values for TNFSF10 and CXCL10 were similar, the maximum values for IL-6, CXCL11, PDL1, CXCL6, MCP-2, CX3CL1 and GLB1 trended higher in serum (Fig. 4C). Conversely, the expression of EIF4G1 and IFN-γ trended higher in nasal mucosa. Expression of some proteins peaked earlier in the nasal mucosa, including IL-6, GLB1 and MCP-2, which were significantly elevated (p<0.05) in nasal mucosa at D3 but not until D5 (p<0.05) in serum (Fig. 4C). Proteins that either peaked earlier or exhibited higher expression in the nasal mucosa were likely from virally infected cells, which would result in higher concentrations of these proteins in that compartment. They may also reflect early nasal mucosal responses to infection. Although CXCL6 and EIF4G1 expression was not significantly increased with viral shedding at D3 in either compartment, they decreased in shedders at D5 in nasal mucosa (p<0.05) but increased in serum at D5 (p<0.05) (Fig. 4C).

Of the 18 shedding-associated proteins unique to the nasal mucosa, most (n= 11) showed no change in values in shedders at D3 but decreased significantly (p-value < 0.05) at D5 (Fig. 4B). Many, including IL-8, CCL1, TNFSF14, EN-RAGE, CXCL5 are inflammatory mediators involved in immune cell recruitment and activation or protection from inflammation-associated oxidative damage (PRDX5). Reduced values in shedders may reflect a lack of protective response in these individuals. Alternatively, decreased values at D5 in the nasal mucosa may represent activation of anti-inflammatory pathways and reduced expression of immune cell recruitment as part of the infection resolution process in shedders. ZBTB16, which showed increased expression in shedders on D3 in the nasal mucosa only, is a transcription factor that modifies chromatin to restrain inflammatory signaling responses (13). Other proteins with increased expression unique to nasal mucosa at D3 included PIK3AP1, DAPP1, and TANK. The T cell marker CD8A showed increased expression at D3 and D5, consistent with higher expression of IFN-γ in the nasal mucosa relative to serum of shedders. The expression of SH2D1A were also increased at D3 and D5 (p<0.05) and is a negative regulator of lymphocyte activation (14), suggesting simultaneous activation and modulation of cell-mediated responses.

Shedding-associated proteins (p<0.05) unique to serum (n=41) mapped to processes predominantly associated with cell-mediated responses and were all increased in shedders compared to NS (Fig. 4B). Proteins at D3 (p<0.05) included LAG3, a marker of activated lymphocytes and NK cells, CXCL9/CXCL19, involved in T and B cell trafficking and recruitment to thymus and lymphoid organs, respectively and LAMP3, a protein produced by dendritic cells with functions in antigen presentation. Proteins increased in shedders at D5 (p<0.05) included NK, T, and B cell markers (SLAMF1, CDCP1, IL-18R1, CD5, CD6, TNFRSF9, and CD244) and lymphocyte differentiation and activation markers (NFAT3C, NT-3, IL-17A, Sirt2, and IL-15RA). Tissue damage and repair proteins (uPA, FGF-5) and attenuation of inflammation and cell-mediated responses (GDNF, IL-10RB, IL-10, LIF-R, LAP-transforming growth factor (TGF)-β1, STAMBP, Flt3L, and LILRB4) were also elevated at D5 (p<0.05), suggesting activation was associated with resolution of infection. Proteins associated with macrophage recruitment and function (TNF, DNER, CLEC6A), which participate in tissue repair, were similarly elevated (p<0.05). Finally, expression of TRIM21, a cytosolic ubiquitin ligase and antibody receptor, was increased in shedders at D5.

Vaccinated and unvaccinated viral shedding participants exhibit distinct gene and protein expression profiles

Differences in nasal mucosal gene expression between vaccinated and unvaccinated shedders (1DS and MDS) on D3 post-challenge were identified by t-test (p<0.01) and showed distinct grouping of vaccinated and unvaccinated participants by principal component analysis (PCA) (Fig. 5A). Pathway analysis of the respective vaccinated and unvaccinated shedding gene signatures showed that, although both groups induced antiviral and type I IFN responses D3 post-challenge, vaccinated shedders also showed increased enrichment of lymphocyte mediated responses, including NK cell signaling, cytotoxic T lymphocyte-mediated apoptosis, T cell exhaustion, and STAT3 signaling, suggesting increased cell-mediated responses during acute infection in vaccinated participants. The D3 shedding gene signature in unvaccinated participants was dominated by pathways related to IFN signaling, cytokine signaling, and neutrophil infiltration (Fig. 5B). Comparison of nasal mucosal protein expression by t-test (p<0.05) in vaccinated and unvaccinated shedders at D3 identified differences suggesting increased T cell activity in vaccinated shedders. Expression of CD8A, IFN-γ and PD-L1 in the nasal mucosa were higher in vaccinated shedders compared with unvaccinated shedders (p<0.05). Expression of the pro-inflammatory proteins IL-33, IL-20RA and TNF-β were also higher in vaccinated shedders compared with unvaccinated shedders (p<0.05) (Fig. 5C). These proteins did not show differences in expression between vaccinated and unvaccinated shedders in the serum.

Fig. 5: Changes in nasal mucosal gene and protein expression associated with shedding in vaccinated and unvaccinated participants.

(A) Principal component analysis of shedding-associated sequences on D3 identified by ANOVA with Tukey’s Honest Significant Difference (HSD; p<0.05) comparing vaccinated and unvaccinated (orange) shedders (n=23 participants). PC1, principal component 1; PC2, principal component 2. (B) Pathway analysis comparing D3 shedding-associated sequences in vaccinated and unvaccinated participants (n=23 participants). (C) Plots of nasal mucosal proteins with different expression in vaccinated compared to unvaccinated shedders (n=23 participants) on D3. NPX values (log₍₂₎) were statistically significant (p<0.05) between vaccinated and unvaccinated shedders, but not between non-shedders using standard t-test with Benjamini Hochberg correction. Plots show data in NS (teal) and shedders (pink). Box plots show the minimum, the first quartile (Q1), the median, the third quartile (Q3), and the maximum expression values. (D) Two-dimensional clustering of Kolmogorov-Smirnov (KS) test values showing related KEGG pathways that were differentially regulated between vaccinated and unvaccinated shedders on D3. Pathways shown had minimum significance proportion of 0.05. (E) KEGG pathways identified in clusters A, B, and C.

Differences in shedding-correlated host pathways in vaccinated and unvaccinated MDS participants were further identified using Kolmogorov-Smirnov (KS) testing and test statistics distribution-based clustering of KEGG pathways using nasal mucosal gene expression data. A standard screening test (e.g., t-test) on each gene was performed individually to calculate corresponding p-values, which were mapped to KEGG pathways. Each pathway has a collection of differentially expressed shedding-correlated genes that form a distribution using the corresponding p-values. To calculate the distance between pathways, a KS test was used to compare p-value distributions and identify shedding-correlated dissimilarity in pathway activation. Finally, hierarchical clustering was performed on this matrix to obtain 12 clusters KEGG pathways with similar KS test and proportional significance values and identify the most significant differences between vaccinated and unvaccinated 1DS and MDS participants (Fig. 5D). The top 3 pathway clusters (Fig. 5D and E) consisted of metabolism, carbohydrate, and lipid biosynthesis pathways (cluster A), innate and adaptive immune, antiviral, and cellular defense responses (cluster B), and cell proliferation, steroid hormone, and endocrine pathways (cluster C). Additional ranked pathway clusters are shown in data file S14.

Ex vivo PBMC responses to H1N1 virus stimulation differ between NS, 1DS, and MDS shedding groups and between vaccination status

PBMCs isolated from a subset of individuals with sufficient viable cells (n=23 unvaccinated, n=21 vaccinated) at D-1 (pre-challenge) and D1, D3, D5, and D7 post-challenge were exposed ex vivo to infectious A/Bethesda/MM2/H1N1 for 48 hours. Concentrations of inflammation-related proteins (n=184) secreted into the supernatant were measured using Olink proteomics. Positive and negative controls included stimulation with Cytostim or RPMI-1640 complete media, respectively. One-way ANOVA, using the multi-model shedding groups (NS, 1DS, and MDS) as the variable, identified 33 and 19 proteins that were significantly different (p<0.05) between at least 2 shedding groups on D1 and D3, respectively (data file S15). Multiple proteins with different abundance (p<0.05) in PBMC supernatants on D1 were linked to T cell activation and recruitment (IL-2, EIF4G1, SPRY2, CD8A, IL-12B, CASP8, CCL20, TNFSF14, HCLS), oxidative stress (PRDX5, PSIP1, PPP1R9B) and antiviral responses (IRF9, TRIM5, DCTN1) (Fig. 6A). Most showed higher values in NS participants (p<0.05), suggesting that early activation of lymphocyte responses was protective against viral shedding. Oxidative stress and antiviral-related proteins showed lower expression in 1DS versus MDS participants (p<0.05). Proteins identified on D3 (p<0.05) also included lymphocyte function associated IL-2, IFN-γ, SLAMF1 and IL-5 (Fig. 6B), with NS participants still showing higher expression compared to 1DS and MDS (p<0.05) participants. IL-2, SLAMF1 and IL-5 concentrations were also higher in 1DS versus MDS participants (p<0.05). Cytostim induces non-specific activation of T cells by binding the TCR and crosslinking it to a major histocompatibility complex (MHC) antigen-presenting cell (APCs). Expression of lymphocyte markers (LAG3, CD8A, IL-2, IFN-γ) were not statistically different between the shedding groups in culture supernatants of PBMCs collected at D1 and D3 post-challenge and treated with Cytostim ex vivo (fig. S11), showing that the reduced expression in virus-stimulated PBMCs from shedding participants was not due to inherent defects in lymphocyte responsiveness.

Fig. 6: Differential PBMC ex vivo responsiveness in multimodal shedding and vaccination groups.

PBMCs collected from participants at D1 and D3 post-challenge were stimulated ex vivo with infectious A/Bethesda/MM2/H1N1 virus (n=44 participants). Supernatant proteins were quantified by Olink proteomics. (A) Plots show expression of proteins from PBMCs collected D1 post-challenge associated with T cell activation and recruitment, oxidative stress, or antiviral responses. Values were statistically significant between the NS, 1DS or MDS groups using one-way ANOVA (p<0.05). *p<0.05. (B) Plots showing expression of proteins from PBMCs collected D3 post-challenge. Values were statistically significant between the NS, 1DS or MDS groups using one-way ANOVA (p<0.05). *p<0.05. (C) Plots showing expression of proteins collected D1 post-challenge that showed significant differences between vaccinated (vax) and unvaccinated (unvax) MDS shedders by standard t-test (p<0.05). Data are presented as box plots, which show the minimum, the first quartile (Q1), the median, the third quartile (Q3), and the maximum. Scale of Olink NPX is log₍₂₎.

Comparison of PBMC responsiveness following ex vivo exposure to A/H1N1 virus in vaccinated versus unvaccinated MDS participants identified 8 and 10 proteins in supernatants from PBMCs isolated D-1 and D1 post-influenza challenge, respectively (p<0.05). For PBMCs collected at D-1, most proteins (7 of 8) showed higher values in unvaccinated shedders (p<0.05) (fig. S12). These included the anti-inflammatory proteins IL-10 and TGF-β, in addition to pro-inflammatory chemokine CX3CL1, which is chemotactic for monocytes and lymphocytes. IL-2RB and ADA, which are involved in T cell-mediated responses, and IL-24, which plays a key role in immune cell proliferation and survival, were also elevated (p<0.05) in unvaccinated versus vaccinated shedders (1DS and MDS). FAM3B and CLEC6A showed higher expression in vaccinated shedders in PBMCs collected at D-1 (p<0.05). In contrast, responses in PBMCs collected D1 post-challenge showed 7 of 10 of proteins had higher values in vaccinated versus unvaccinated participants (p<0.05) (Fig. 6C), including pro-inflammatory cytokine CCL23, highly chemotactic for resting T cells, and CXCL10, secreted in response to IFN-γ and VEGF-A. AREG, expressed by activated Th2 cells, is critical for regulatory T cell function and GALNT3 plays a role in regulation of NF-κB during IAV infection (15). Proteins with higher expression in unvaccinated shedding participants (p<0.05) included IL-17A, a proinflammatory cytokine produced by activated T cells that links T cell activation to neutrophil recruitment and activation (16), SLAMF1, also produced by activated lymphocytes, and FGF-23 (Fig. 6C).

Discussion

Understanding mucosal innate and adaptive immunity in humans, particularly in populations with pre-existing immunity, is critical for next-generation vaccine development, yet much remains unknown of the relationships between these responses that determine infection outcome (17). Although vaccines have been developed inducing long-lived immunity from systemically replicating viruses (e.g., measles, mumps, and rubella), vaccines producing durable immunity against respiratory mucosal RNA viruses remain elusive (18). The ability of respiratory RNA viruses to evolve necessitates constant updating of seasonal vaccines as new variants arise and spread (19). This is largely due to vaccine strategies designed to elicit serum antibody responses to single antigens (e.g., HA for influenza and spike protein for coronaviruses) but not mucosal immune or memory T cell responses. T cell responses are critical for cross-protective immunity and vaccine efficacy (18, 20) and numerous studies have shown essential roles of CD8 T cell-mediated immunity in reducing morbidity and mortality during respiratory viral infection (18). Another challenge to next-generation vaccine development is that infection with respiratory RNA viruses themselves do not result in long-term immunity (18), as reinfection is common, even with genetically identical viruses (21).

Historically, human influenza challenge and vaccine efficacy studies have enrolled and studied responses in participants with little pre-existing immunity to the challenge virus strain (11, 12, 22). Although these studies have provided important insights into the natural history of infection, these populations do not reflect real-world immune complexity and dynamics to influenza infection. This is especially true for mucosal immunity, which represents the first line of defense against infection. Human IAV immunity is complex and arises from a lifetime of exposure and vaccination to continually evolving circulating influenza strains. Here, nasal mucosal and systemic immune transcriptomic, proteomic, and B and T cell VDJ repertoire dynamics following A/H1N1pdm09 challenge were characterized in participants with complex pre-existing immunity, where half the cohort had recently been vaccinated. Responses correlated with differential clinical outcomes ranging from “fully protected” participants with no viral shedding or symptoms through “partial protection” with reduced shedding and symptoms to “unprotected participants” with multiday shedding and increased symptoms.

A model of nasal mucosal and systemic responses was created correlating immune protection and multimodal shedding outcomes (fig. S13). In NS participants, protection from active viral replication and clinical illness correlated with increased baseline HA-mediated antibody immunity, in serum HAI and nasal mucosal IgA, and expansion of low frequency B and T cell clones post-challenge. There was no evidence of vaccine-expanded B cell clones re-expanding post-challenge; however, this may be due to the difference in timing of B cell activation and sampling in this study. Different populations of low frequency B cell clones, including broadly neutralizing VDJ clones, expanded D7 post-challenge, likely reflecting activation of memory B cell responses from prior IAV infections. In contrast, some T cell VDJ clones that expanded following vaccination also expanded D7 post-challenge. IAV HA, as the major vaccine antigen in split and subunit IAV vaccines, does not contain the major known IAV T cell epitopes (23), as these vaccines are designed to primarily induce serum antibodies against HA. Protection from viral infection in NS was also correlated with increased baseline nasal mucosal immune cell pathway-related gene expression and serum protein expression that was dominated by proteins associated with NK cell activation. This suggests that prophylactic stimulation of nasal mucosal and systemic cellular immune responses could be used in a non-specific manner to prevent or limit infection.

Vaccinated shedders were 18-times more likely to be 1DS than MDS if they had TRBV VDJ clones expanding post-vaccination and D7 post-challenge. Also, 1DS participants were three times more likely to be female than male, which may be linked to increased protective immune responses contributed to differences in production of cytokines/chemokines, numbers of immune cells, and how pattern recognition receptors detect IAV infection (24, 25). In 1DS and MDS, neither nasal mucosal SIgA nor serum HAI titers correlated with D1 viral load, further supporting the importance of T cell immunity. MDS participants lacked sufficient nasal mucosal humoral and cellular immune responses to control infection. Protective efficacy may be improved by nasal mucosal vaccine strategies to induce better SIgA and IgG titers against IAV surface glycoproteins, and T cell responses against IAV internal protein antigens. Since mucosal antibody titers wane rapidly following vaccination, strategies to induce mucosal T cell responses (e.g., an increase in mucosal TRM cells) would be beneficial, especially with the rapid antigenic drift of IAV HA and NA glycoproteins (26, 27).

Limitations of the study include the small sample size and young age of participants (average 34 years) lacking underlying co-morbidities. The elderly and those with pre-existing risk factors (e.g., cardiopulmonary disease, obesity, immunosuppression, pregnancy, etc.) may demonstrate different responses, and thus these data may not be predictive of those at risk for severe IAV infections. Other limitations that were due to clinical study design and blood volume limitations include the inability to perform detailed immune cell phenotyping by flow cytometry that have would allowed us to distinguish different subsets of B and T cells, including effector and memory cells, antigen purification methods for influenza-specific VDJ repertoire analysis, and the lack of nasal mucosal sampling for scRNA-seq. Analysis of nasal mucosa-resident memory B and T cells collected from germinal centers would also add valuable data. These issues will be addressed in future studies.

The human respiratory mucosal immune system exists in semi-organized and specific regional areas, including nasopharyngeal associated lymphoid tissue, bronchial associated lymphoid tissue, and separate pulmonary compartments. Future studies will explore identification and tracking of influenza-specific B and T cells, either from laboratory-based methods (e.g., flow cytometry of cell populations responding to influenza-specific antigens) or advances that identify signatures of influenza exposures in silico. Future clinical studies should include sampling in additional regions of the respiratory tract, both for evidence of viral replication and inflammatory/immune responses, and HLA and SNP analyses, as well as a more comprehensive assessment of pre-existing immunity to past IAV exposures in challenge cohorts, not only for antibodies against challenge strain HA and neuraminidase, but also for the role of cross-protective T and B cell memory responses (28, 29). Previous exposure history has been shown to influence vaccine responses (30, 31), and future vaccine-challenge studies comparing intramuscular and mucosal-delivered vaccination (18) will be crucial for next-generation IAV vaccine development.

Materials and Methods

Study design.

The objective of this study was to characterize viral shedding-correlated nasal mucosal and systemic responses in samples from an unblinded, non-randomized A/H1N1 influenza challenge in people (n=74) with complex pre-existing immunity, including half who were recently vaccinated (mean 64 days; (11)). Participants were enrolled independent of pre-existing immunity against H1N1pdm09 challenge virus (see the supplementary materials file for detailed power analysis and participant inclusion/exclusion criteria). Measurements were performed on all available samples collected from 74 participants. Analyses of samples included nasal wash viral shedding by BioFire FilmArray (n=669) and qRT-PCR (n=148); PBMC B and T cell VDJ repertoire by immunosequencing (n=444 each); nasal mucosal cell (n=297) and PBMC (n=339) RNA expression by microarray; and protein expression in nasal mucosa (n=340), serum (n=338), and supernatants of ex vivo H1N1 virus-stimulated PBMCs (n=405) by Olink assay. Technical replicate measurements ranging from singlet to triplicate were determined by assay type as described below.

Clinical study and sample collection.

The challenge study (clinicaltrials.gov NCT01971255) was performed at the NIH Clinical Center between April and October 2019 (11). It was approved by the NIAID Institutional Review Board (19-I-0058) and conducted in accordance with the provisions of the Declaration of Helsinki and Good Clinical Practice guidelines. Briefly, 40 participants were enrolled in the vaccinated cohort and were vaccinated intramuscularly with QIV (Flucelvax), referred to as phase 1 of the study; another 40 participants were not vaccinated. Although the participant cohort was not randomized, similar distributions of age, race, gender, and ethnicity were enrolled in vaccinated and unvaccinated cohorts. All 80 participants were brought into the NIH Clinical Center as mixed cohorts and challenged with 10⁷ TCID₅₀ of influenza A/Bethesda/MM2/H1N1 virus (median 64 days post-vaccination) and assessed daily for a minimum of 9 days, referred to as phase 2. Of 80 participants, six were excluded due to study criteria. Nasal mucosal and systemic samples were collected pre-challenge (day prior to inoculation) and D1, D3, D5, and D7 post-challenge, with PBMC samples also collected D28 and D56 post-challenge (fig. S1).

Nasal mucosal cell samples were collected D-1, D1, D5, and D7 from the middle turbinate using a nasal speculum and pathology brush, placed in RNAlater and stored at −80°C prior to RNA extraction. Nasal mucosal secretions were collected using a synthetic absorbent material (SAM) strip, frozen at −80°C and then thawed on ice for elution in 300 μl 1% bovine serum albumin (BSA) in phosphate-buffered saline (PBS). The eluate was aliquoted and stored at −80°C. PBMCs for DNA extraction and ex vivo stimulation with infectious A/Bethesda/MM2/H1N1 virus were isolated using SepMate PBMC isolation tubes (STEMCELL Technologies). Viral stimulation was performed on all participants (n=44) for which a minimum of 5 million viable cells/ml cells were available. Controls included a negative (PBS) and positive (Cytostim) per sample.

Measurement of proteins using Olink targeted proteomics platform and prioritization.

Proteins in serum or nasal mucosal samples from a subset of participants (n=66) for which an entire time-course (D-1, D1, D3, D5 and D7 post virus challenge) was available were quantitated using proximity extension assay (Olink Proteomics) as described with one sample per well (32). Ct values normalized against extension and inter-plate controls and adjusted with a correction factor according to the manufacturer’s instructions to calculate a normalized protein expression (NPX) value, which is log₍₂₎. Statistical analysis included standard t-test and ANOVA using Benjamini-Hochberg procedure to correct for the false discovery rate (FDR). For protein prioritization, proteins not detected in at least two samples were excluded and treated as non-detection events with assigned NPX value of zero. For each tissue and time point pair, the matrix of Olink values was analyzed using the Gene Prioritization approach of the KnowEnG Platform (33), calculating the result of the two-sided two-sample Student’s t-test for each group comparison. Nine participant group comparisons were defined based on D1, D3, and D5 shedders compared to that day’s non-shedders or the NS group and based on the pairwise comparison of three shedding classes (NS, 1DS, MDS). Three additional participant group comparisons were examined by splitting by vaccination status across 1) all participants, 2) D3 shedders, and 3) Multiday Shedders (MDS).

RNA isolation and microarray gene expression analysis.

Total RNA was isolated from nasal mucosal samples (n=74 participants). Nasal brushes in RNAlater were vortexed to dislodge cells from the brush. Four volumes of PBS were added, then centrifuged at 6,000 x g for 10 minutes. The cell pellet was washed with 500 μl PBS, centrifuged at 6,000 x g for 10 minutes, resuspended in 700 μl of Qiazol and the miRNeasy (Qiagen) according to the manufacturer’s instructions. Total RNA was isolated from PBMCs collected in Paxgene tubes (12). Gene expression profiling was performed using Agilent Human Whole Genome 44K microarrays with one sample per microarray (12). The data were uploaded into Genedata Analyst 9.0 (Genedata) and normalization was performed using central tendency, followed by relative normalization using each participant’s baseline (D-1) as the reference. Statistical analyses were performed using Genedata Analyst 9.0, GraphPad Prism and R statistical software (4.1.0/18 May 2021; Foundation for Statistical Computing). Differentially expressed genes and proteins were identified using standard t-test (at least two-fold difference in median expression between compared groups and p<0.05). Benjamini-Hochberg procedure (FDR; p<0.05) was used to correct for multiple comparisons. Ingenuity pathway analysis was used for gene ontology enrichment.

Quantitation of IAV M gene RNA in nasal mucosal samples.

IAV RNA expression in nasal mucosal samples was quantified using RT-qPCR. Reverse transcription of total RNA was performed using the Superscript III first-strand cDNA synthesis kit (Invitrogen) primed with an equal mix of random hexamers and the Uni-12 influenza primer: 5’-AGCRAAAGCAGG-3’. IAV M gene was quantified using the following primers and probe sequences: forward, 5’-GACCRATCCTGTCACCTCTGAC-3’; reverse, 5’-AGGGCATTYTGGACAAAKCGTCTA-3’; probe, 5’-TGCAGTCCTCGCTCACTGGGCACG-3’. qPCR was performed on a Bio-Rad CFX384 Touch Real-Time PCR Detection System with TaqMan 2X PCR Universal Master Mix using a 10 μl reaction volume in duplicate. Ct values were normalized to the calibrator gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (ThermoFisher Scientific #4333764 T) and final ΔCt values were inverted (40 – ΔCt).

Variable chain immunosequencing in T and B cells.

The CDR3 regions of human TRBV and IGHV chains were sequenced using Adaptive Immunosequencing (Adaptive Biotechnologies). Extracted genomic DNA from PBMCs was amplified in a bias-controlled multiplex PCR, followed by high-throughput sequencing. Sequences were collapsed and filtered to identify and quantitate absolute abundance of each unique CDR3 region for further analysis, as previously described (34, 35). The sum frequency of broadly neutralizing B cell motifs was calculated from CDR3 IgH amino acid sequences containing CxxxxxQxxVx(8-12)W and V01-18 V gene usage (36).

Analyses of TRBV and IGHV sequencing.

Two quantitative components of repertoire diversity were compared across samples in this study. First, Simpson clonality was calculated on productive rearrangements by: $\sqrt{{\sum }_{i=1}^R{p}_i^2}$; where $\mathrm{R}$ is the total number of rearrangements and $p_i$ is the productive frequency of rearrangement $i$. Values of Simpson clonality range from 0 to 1 and measure how evenly receptor sequences (rearrangements) are distributed. Clonality values approaching 0 indicate an even distribution of frequencies, whereas values approaching 1 indicate an increasingly asymmetric distribution in which one to a few clones are present at high frequencies. Second, sample richness was calculated as the number of unique productive rearrangements in a sample after computationally down sampling to a common number of cells to control for variation in sample depth or B and T cell fraction. Repertoires were randomly sampled without replacement five times and report the mean number of unique rearrangements. T cell fraction was calculated by taking the total number of T cell templates and dividing by the total number of nucleated cells, which is derived from a panel of reference genes. Clonal expansion was calculated according to a binomial distribution framework as described (37). A two-sided test of the null hypothesis that the probability of success in a Bernoulli experiment was computed for each clone. The Benjamini-Hochberg procedure was used to control FDR (p<0.01). Wilcoxon signed rank was performed to compare immune metric changes timepoint-to-timepoint. Statistical analyses were performed in R version 3.4.x.

Fisher’s exact test and Poisson regression model.

Fisher’s exact test examined the association between the two categorical variable pairs by constructing 2x2 contingency tables and comparing the observed data against distribution under no associations. For Poisson regression model, days of shedding is used as the outcome, which is modeled using the Poisson distribution, with age, gender, vaccination status, and D-1 nasal mucosal SIgA (or serum HAI antibody titers in a separate model) as predictors to calculate their estimated coefficients and corresponding p-values. Marginal screening was also performed to identify significantly differentially expressed (p<0.01) D-1 nasal mucosal and PBMC genes in the Poisson regression. For this analysis, the number of shedding days was modeled as a Poisson count outcome, adjusting for age, gender, and vaccination status as fixed predictors, along with gene expression and antibody titers, which were used to construct two separate gene lists. The two lists are constructed with the following models: (1) a model including single gene expression values, screening over the entire list of genes; or (2) a model that includes gene expression values and IgA titers for nasal data as well as a model that includes gene expression values and HAI titers for PBMC data. The top lists of significantly differentially expressed genes were then used in KEGG pathway enrichment analysis.

Test statistic distribution-based network clustering with Kolmogorov-Smirnov (KS) distance.

For each gene present in two sample groups, the log₍₂₎ fold-change in gene expression were calculated comparing post-challenge to baseline expression. A two-sided two-sample t-test, differentiating conditions such as vaccinated versus unvaccinated shedders, was then applied to the log₍₂₎ fold changes in gene expression. Corresponding t-statistics were converted into quantiles on the null distribution (t distribution with the proper degrees of freedom). Genes were then categorized based on KEGG biological pathways (38), resulting in 287 pathway groups. To ensure analytical robustness and stability, only pathways containing a threshold of at least 20 genes were included in further analysis. A similarity matrix was constructed on every unique pair of KEGG pathways using the KS test as described in (39). By treating the t-statistics for each gene within a pathway as observed values, the KS test was conducted to measure the distributional disparities of these t-statistics between pathways. Hence, it serves as a distance measure, leading to the construction of the similarity matrix. Hierarchical clustering was then applied to the similarity matrix to identify pathways with the highest proportion of significantly differentially expressed genes. To avoid violating the independence observation assumption for the KS test, a permutation test (40) was conducted to generate the null distribution of KS test statistics. Specifically, genes and their corresponding t-statistics from each pathway pair were represented by vectors $z_i,z_j$. The overlapping genes between these two pathways was denoted as $z_{ij}^{\ast }$, and the unique genes in each pathway were denoted as $z_i^{\ast },z_j^{\ast }$ respectively. To maintain the proportion of overlap between the two pathways, resampling from, the collective gene set ($z_i^{\ast },z_j^{\ast },z_{ij}^{\ast }$) was performed to create two new gene groups, ensuring that the number of shared genes and the number unique genes in each group are the same as the original data. By repeatedly applying these methods and running the KS test of each resample, the null distribution of the KS test statistics was approximated under the overlapping setting. This generated the p-values used in the similarity matrix.

Statistical analysis

Individual-level data are presented in data file S16. Significance testing performed is described in detail for each method above. Unless otherwise specified, one-sided Fisher’s exact test (p<0.05), Poisson regression (p<0.01), t-statistic (p<0.05), Wilcoxon signed rank (p<0.05), Student’s t-test (p<0.05) when comparing two groups, ANOVA (p<0.05) with Tukey’s HSD post-hoc analysis when comparing more than two groups, Spearman rank sum, Pearson correlation (r>0.6), or Kolmogorov-Smirnov test (p<0.01) were used. Post-hoc analysis was performed using Benjamini-Hochberg procedure (p<0.01) to correct for FDR in multiple comparisons. Ingenuity pathway analysis was used for gene ontology enrichment using Bonferroni correction. Suitable application of parametric testing was verified by confirming normal distribution of gene expression and Olink data (data file S16).

Supplementary Material

Supplementary materials and methods

Fig. S1 to S13

Data files S1 to S16