The BNT162b2 mRNA vaccine demonstrates reduced age-associated T_H1 support in vitro and in vivo

Summary

mRNA vaccines demonstrate impaired immunogenicity and durability in vulnerable older populations. We hypothesized that human in vitro modeling and proteomics could elucidate age-specific mRNA vaccine actions. BNT162b2-stimulation changed the plasma proteome of blood samples from young (18-50Y) and older adult (≥60Y) participants, assessed by mass spectrometry, proximity extension assay, and multiplex. Young adult up-regulation (e.g., PSMC6, CPN1) contrasted reduced induction in older adults (e.g., TPM4, APOF, APOC2, CPN1, PI16). 30–85% lower T_H1-polarizing cytokines and chemokines were induced in elderly blood (e.g., IFNγ, CXCL10). Analytes lower in older adult samples included human in vivo mRNA immunogenicity biomarkers (e.g., IFNγ, CXCL10, CCL4, IL-1RA). BNT162b2 also demonstrated reduced CD4⁺ T_H1 responses in aged vs. young adult mice. Our study demonstrates the utility of human in vitro platforms modeling age-specific mRNA vaccine immunogenicity, highlights impaired support of T_H1 polarization in older adults, and provides a rationale for precision mRNA vaccine adjuvantation to induce greater immunogenicity.

Graphical abstract

Highlights

• •mRNA vaccine innate immunogenicity was modeled in human whole blood in vitro
• •Older adults had distinct, primarily lower T_H1, mRNA vaccine immune activation
• •Induction of CXCL10, IL-1RA, IFNγ, and CCL4 was reduced with age
• •In vivo mRNA vaccination in aged mice demonstrated impaired T_H1 immunity

Health sciences; Geriatrics; Immunology ; Immunity; Proteomics

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) mRNA vaccines were rapidly developed, authorized, approved, and implemented to address the public health threat of coronavirus disease 2019 (COVID-19). Initial reports suggested a high vaccine efficacy (VE) of ∼90–95% for BNT162b2 (Comirnaty, Pfizer-BioNTech) and mRNA-1273 (Spikevax, Moderna) in reducing severe COVID,^1,2 (compared to other respiratory viruses, e.g., yearly influenza VE of 10–80%^3,4). While mRNA vaccines remain key to combating COVID-19 morbidity and mortality, it is increasingly evident that VE varies by target population. While SARS-CoV-2 mRNA vaccines have demonstrated VE across a range of ages, they have been less effective at preventing hospitalization and symptomatic infection 9.5% reduction in VE in those >65 years (Y) versus 18-65Y adults,¹ and 20% less efficacy in >80Y compared to 60-69Y.⁵ A meta-analysis identified a 9.3% decrease of VE preventing infection in older populations compared to the general populace,⁶ and another review identified consistently lower VE in those ≥65 than <65, with up to 15% less VE in elders.⁷ Generally, age-associated infection vulnerability has been attributed to increased disease severity and reduced vaccine-induced protection.^5,8,9,10 Vulnerability, linked with immunosenescence in older adults, has been observed both with respect to SARS-CoV-2 infection-induced immunogenicity¹¹ and reduced immunogenicity of mRNA vaccines.^{12,13,14,15,16,17,18} These observations mirror impaired immunogenicity in elder individuals across a range of vaccines targeting diverse microbial pathogens,^{19,20,21,22,23} impacting both humoral and cell-mediated immunity, culminating in diminished VE in older adults.^13,24,25,26

Reduced vaccine immunogenicity in older adults likely reflects immunosenescence. Advancing age has been associated with impaired immunity including reduced neutralization, phagocytosis, chemotaxis, and co-stimulatory molecule expression.^14,20,27 This impairs B cell class switching, with distinct T follicular helper CD4⁺ T cell (T_FH) activity, and increased T regulatory cell (T_reg) frequency restraining responses.^{14,19,20,21,22,23} Age-associated changes in T cell immunity could contribute to reduced cellular and antibody (Ab) functionality following the mRNA vaccination of older adults.^12,13,16 Older populations (>60Y, >65Y, and >80Y) demonstrate lower cell mediated immunity (CMI), with impaired CD4⁺ and CD8⁺ activation following BNT162b2 immunization, compared to middle aged adults.^14,15,28 Impaired CD4⁺ and CD8⁺ T cell responses were also observed in older adults post-SARS-CoV-2 infection,²⁹ suggesting distinct immunity. Additional booster immunizations with mRNA vaccines encoding Wuhan1 or bivalent Wuhan1 with BA.4/BA.5 mRNA encoding spike protein have been applied to overcome elder immunosenescence.^30,31,32 The third immunization with Wuhan1 mRNA transiently amplified immunogenicity against Wuhan1, and the variants Delta (B.1.617.2), and Omicron (B.1.1.529),³⁰ but failed to induce durable immunity as >65Y individuals had more rapid waning of immunity compared to <65Y.³¹ The CDC Advisory Committee on Immunization Practices (ACIP) recommendation for those ≥65 Y to receive 2 mRNA immunizations per year was driven by expected and exacerbated waning immunity in elder populations.³³ A better understanding of age-specific immunity may provide insights that can inform efforts to optimize mRNA vaccines to enhance VE.

An emerging approach to characterize vaccine action is human in vitro modeling,³⁴ employing primary human leukocytes and autologous plasma, which is a rich source of age-specific soluble factors.^{9,34,35,36,37} Indeed, the recent passage into law of the FDA Modernization Act 2.0 provides for the use of human in vitro systems to support drug and vaccine development.³⁷ Such in vitro systems enable the characterization of vaccine action in a species-specific manner wherein the same study participant can serve as the control (vehicle) and test condition. Such assays are amenable to the downstream measurement of a range of analytes via systems biology enabling the discovery of biomarkers that may correlate with vaccine safety (e.g., reactogenicity) and immunogenicity in vivo. Characterizing mRNA vaccine-induced immune activation in vitro may provide insight into human- and age-specific immunogenicity to inform future enhancement and optimization of mRNA vaccines.

We hypothesized that comparing responses of young and older adults to mRNA vaccines via human in vitro modeling coupled with proteomics would demonstrate distinct age-specific responses to BNT162b2 stimulation, providing an understanding of age-specific mRNA vaccine immune activation. To this end, we studied BNT162b2 immune stimulation in adult and elder human whole blood assay (WBA) in vitro and characterized the supernatant proteomes using liquid chromatography mass spectrometry (LC/MS), Ab-based proximity extension assay (PEA), as well as cytokine and chemokine multiplexing. We observed impaired induction of a range of proteins including T_H1-polarizing cytokines and chemokines in elders. This age-dependent mRNA vaccine-induced impaired T_H1 immunogenicity was confirmed in young and aged mice, in vivo. Impaired T_H1 polarization with age may contribute to the reduced mRNA vaccine-induced immunogenicity that is observed in older adults. These observations provide insight into age-dependent mRNA vaccine action and can inform the discovery and development of next generation vaccines optimized for enhanced immunogenicity and protection in vulnerable older adults with distinct immunity.

Results

Cohort description

Human research study participants donated peripheral blood which was evaluated in vitro for immune activation following stimulation with the BNT162b2 lipid nanoparticle (LNP) encapsulated mRNA vaccine (Pfizer/BioNTech). Participants were grouped by age, with younger and middle-aged adults of 18-50Y, and older adults, elders, ≥ 60Y. These populations had a mixed vaccination and infection history, detailed in Tables S1 and S2. In brief, 41% of adults evaluated had a self-reported exposure to SARS-CoV-2 Spike antigen via infection or vaccination, while 86% of elder participants had a self-reported exposure.

BNT162b2-stimulation impacts the whole blood proteome

Human blood was stimulated with an mRNA LNP in a whole blood assay (WBA), broadly measuring up-/down-regulation and age-associated differences of immune factors by supernatant LC/MS proteomics (n = 12–14, Table S1). LC/MS has been employed to identify disease severity-associated responses after SARS-CoV-2 infection in humans,³⁸ but has not yet been employed to evaluate in vitro WBA with BNT162b2-stimulated supernatant. Protein expression was evaluated by generalized estimating equations generalized linear model (GEEGLM) analysis,^39,40,41,42 assessing mRNA vaccine stimulant concentration-dependent impact on analyte fold change (FC) of stimulated over matching vehicle control for baseline-normalization. Adult participants had 20 upregulated and 4 downregulated proteins (Figure 1A), while elder participants demonstrated 4 upregulated and 12 downregulated proteins (Figure 1B). Heatmap visualization displayed age-dependent patterns of the regulation (Figures S1A and S1B). Notable BNT162b2-induced proteins in adults included a proteasome regulatory unit protein (PSMC6), hemoglobin subunit epsilon (HBE1), carboxypeptidase N catalytic chain (CPN1), and bisphosphoglycerate mutase (BPGM). Downregulated proteins included peptidyl-prolyl cis-trans isomerase A (PPIA) and neutrophil defensin 3 (DEFA3). Elder sample protein upregulation included serum amyloid A-1 protein (SAA1) and the fibrinogens-γ and -β (FGG and FGB), while downregulated proteins included DEFA3, tropomyosin alpha-4 chain (TPM4), apolipoprotein F (APOF), apolipoprotein C-II (APOC2), CPN1, and peptidase inhibitor 16 (PI16). The supernatant was further evaluated with 4 Target 96 kits (inflammation, inflam; oncology III, onco; neurology, neuro; and cardiometabolic, cardio) of PEA-based proteomics, quantifying 368 proteins (from n = 4–5 participants, Table S2). PEA of supernatants identified the upregulation of predominantly inflammatory markers in adult BNT162b2-stimulated samples compared to vehicle (e.g., chemokine (C-C motif) ligand 2 (CCL2), CCL3, CCL4, CCL7, CCL8, CCL11, chemokine (C-X-C motif) ligand 8 (CXCL8), IL-1β, and IL-6),^{38,43,44,45,46,47,48,49} with 60 upregulated and 5 downregulated differentially expressed proteins (DEPs) with nominal p < 0.05 (Figure 1C, 21 upregulated and 1 downregulated with Benjamini-Hochberg false discovery rate (FDR) < 0.05). In stark contrast, the only PEA-quantified inflammatory markers induced in BNT162b2-stimulated elder participant samples, compared to vehicle control, were CCL8 and CXCL10 (Figure 1D). In general, elder participant blood samples were less responsive to BNT162b2 stimulation, totaling 4 upregulated and 16 downregulated DEPs with p < 0.05 (0 DEP with FDR <0.05). Unsupervised heatmap evaluation of the top 30 DEPs resulted in 4/5 adult samples clustering by treatment versus no clustering in the elder study participants (Figures S1C and S1D).

Figure 1.

mRNA vaccine induced greater in vitro inflammatory protein release in adult vs. elder whole blood assay

Adult and elder whole blood stimulated with vehicle control (RPMI) or BNT162b2 had supernatants characterized by LC/MS (A-B) and PEA (C-D). GEEGLM evaluation testing concentration-dependent down- (purple) and up-regulated (orange) protein expression is presented against log₂FC of BNT162b2-stimulated samples over matching control, from (A) adult and (B) elder participant samples. Log₂FC of 2μg/ml BNT162b2’s mRNA over matching vehicle control display PEA-quantified (C) adult and (D) elder responses. Horizontal dotted lines represent -log10(0.05). For (A-B) n = 10 to 14; for (C-D) n = 4 to 5. Statistical significance was determined by (A-B) GEEGLM, and (C-D) paired moderated T-test.

The inflammatory proteome was lower in elders’ than adults’ BNT162b2-stimulated whole blood

BNT162b2-induced LC/MS proteomic protein profiles in the WBA differed by age, with adults and elders expressing 21 and 13 unique proteins, respectively (Figure 2A). Just 3 significant DEPs overlapped across age groups, with only DEFA3 downregulated in response to BNT162b2 stimulation in both age groups. The other 2 overlapping proteins had different directionality (Alpha-1 microglobulin, AMBP, was downregulated in adults but upregulated in elders, and vice versa for CPN1). BNT162b2 generally induced greater responses in adults vs. elders (Figures 2B, S2A and S2B). Principal component analysis (PCA) clustering of PEA proteomics displayed distinct adult and elder patterns with separation only in adult BNT162b2-stimulated vs. vehicle control (RPMI, Figure S2C). PEA comparison of BNT162b2-stimulated adult vs. BNT162b2-stimulated elder samples also demonstrated a generally greater upregulation of inflammatory markers in adults (Figure 2C). Advancing age had a significant correlation with lower BNT162b2-induced CCL4 (BNT162b2 stimulation slope −0.13, p = 0.04, Figure 2D), with trends toward lower CXCL8 (BNT162b2 stimulation slope −0.1, p = 0.09), and CCL2 expression (BNT162b2 stimulation slope −0.07, p = 0.11) (Figure S2D). Network representation of DEP pathway analyses indicated some similar pathways induced in adult and elder participant samples (Figures S2E and S2F, E g., “signaling by interleukins”). Elder sample profiles had fewer proteins contributing to each pathway node and an additional predominantly downregulated “immunoregulatory interactions between a lymphoid and a non-lymphoid cell” node that was not observed in adult study participants. Additionally, the “IL-4 and IL-13 signaling” that was enriched in adult samples was not observed in elder samples. Overall, BNT162b2 mRNA vaccine stimulation in the WBA resulted in dampened proteome responses in elder participant samples across two proteomic platforms.

Figure 2.

Lower BNT162b2-induced inflammatory response in elder vs. adult whole blood assayed in vitro

(A) DEPs from BNT162b2-stimulated samples against paired vehicle controls (RPMI) were predominantly nonoverlapping between age groups. Comparing BNT162b2-stimulated adult and elder samples identified upregulation in adult participants with analyte quantification by (B) LC/MS-proteomics or (C) PEA-proteomics. (D) Advancing age (years, Y) negatively correlated with normalized protein expression (NPX) in BNT162b2 (BNT)-induced CCL4 (Spearman’s correlation analysis p = 0.04), with 95% confidence interval graphed in gray. (B-C) Horizontal dotted lines represent -log10(0.05). For (A-B) n = 10 to 14; for (C-D) n = 4 to 5. Statistical significance was determined by (B) GEEGLM, (C) paired moderated T-test, and (D) Spearman’s correlation.

Cytokine and chemokine induction by BNT162b2 was verified by bead-based multiplex, with lower T_H1 support from aged participants

An additional evaluation by a targeted multiplex bead-based assay identified titratable production of interleukin-6 (IL-6), CXCL8, tumor necrosis factor (TNF), and interferon gamma (IFNγ) in adult WBA samples (Figure 3A). Other cytokines measured, such as IL-17A, were not induced. Adult and elder responses were FC-normalized (stimulated divided by paired vehicle control) (Figure 3B), and multiple analytes were induced in both age groups, including CXCL10, IL-1RA, and IFNγ. Nevertheless, across multiple stimulation doses, elder samples had 30-59% lower IFNγ, 42–85% lower CXCL10, and 54–85% lower IL-1RA FC induction, compared to adults. Importantly, CXCL10, IFNγ, IL-1RA, and CCL4 have been associated with high responsivity in young adults following human mRNA vaccine immunization,⁵⁰ that was also higher in adults than in elders (Figure 3C). Multiplex-quantified analytes were grouped by function (per Table S3) as T_H1, T_H2, T_H17, or T regulatory (T_reg) polarizing, chemokine, hematopoiesis-supporting, or those associated with trained immunity. A linear modeling analysis, GEEGLM, evaluated if age interacted with each function. T_H1 support was significantly impaired (p = 0.027) in elders compared to adults, with an average of 7.2% less in each analyte involved (Figures 4A–4C). The other functions evaluated were not significantly different (Figure S3) indicating a predominant impairment in inducing T_H1-polarizing analytes.

Figure 3.

BNT162b2 induced concentration- and age-dependent cytokine and chemokine production in in vitro human whole blood assay

Multiplex quantification of secreted analytes identified BNT162b2-induced responses compared to vehicle control (RPMI).

(A) mRNA vaccine concentration-dependent induction of IL-6, CXCL8, TNF, and IFNγ was noted in adults.

(B) Fold Change (FC) baseline standardization of stimulated over matching vehicle control demonstrated greater production of certain analytes such as CXCL10, IL-1RA, and IFNγ in adult (solid red line) vs. elder (dashed blue line) blood.

(C) Volcano plot of analytes with greater BNT162b2 fold-induced stimulation in adults than in elders, with circles representing 0.2 μg/mL, squares 0.67 μg/mL, and diamonds 2.0 μg/mL of mRNA encapsulated in BNT162b2. Non-filled, crossed points represent markers associated with high vaccine responsiveness. The dotted line represents significance, with points annotated above 1.3 -log10(p-value). For (A-C) n = 12 to 14. Boxplots display the median, interquartile range (IQR), with the identification of the furthest values from the median not exceeding 1.5 × IQR. Statistical significance was determined by Shapiro-Wilk then (A) paired Wilcoxon rank-sum test, (B, C) 1-sided unpaired T tests on log-transformed fold-change, with p-values annotated as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Figure 4.

Impaired BNT162b2-induced T_H1-polarizing cytokine production in human older vs. young adults’ blood

Radar plots displaying log₁₀-transformed, FC averages of multiplex-quantified analytes per spoke, separating adult (orange-red) and elder (blue-teal) participants. Stimulation with BNT162b2 (BNT) encapsulated mRNA weights of (A) 0.2 μg/mL, (B) 0.67 μg/mL, and (C) 2 μg/mL had T_H1-polarizing analytes (per Table S3) significantly induced (one-sided T-tests hypothesizing induction compared to vehicle control, color-coded orange adult and teal elder asterisks presented above each analyte). GEEGLM analyses evaluating the interaction of age and induction of T_H1 polarizing analytes demonstrated 7.2% less T_H1-polarizing cytokine production in elder participant samples compared to adult samples (p = 0.027). For (A-C), n = 12 to 14. Significance displays one-sided unpaired T-tests compared to vehicle control, with p-values annotated as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Reduced mRNA vaccine-induced T_H1 cellular immunity in aged compared to young adult mice

To assess whether the age-dependent differences observed in vitro may also be reflected in vivo, we assessed murine intramuscular BNT162b2 vaccination (Figure S4A). As observed in humans,^12,13 aged mice (>10 months) sera displayed significantly lower total immunoglobulin G (IgG), IgG2a, and IgG1 Ab immunogenicity, with a lower anti-receptor binding domain (RBD) Ab titers than adult mice (Figure 5A). Aged murine immunity was rescued with higher antigen doses, with non-significant differences between 0.5 μg-immunized adult and 1.0 μg-immunized elder, or 1.0 μg and 5.0 μg immunized adult and elder animals, respectively. Adult and aged mice displayed waning immunity between Days (D) 42 and 210 post-prime immunization, at various immunization doses (Figures S4B–S4G). Mirroring human elder observations,³¹ greater waning of immunity was observed in 1 μg-immunized aged mice, with 63–75% more waning immunity across IgG, IgG2a, and IgG1 based on the median fold change of D210 over D42 between age groups (Figure S4H). A trend of 30–83% faster waning was observed at other immunization doses. Ab isotypes IgG2a and IgG1, respective markers of T_H1 and T_H2 polarized immunity,⁵¹ were induced over nonvaccinated controls (Figure 5A). The IgG2a/IgG1 relative ratio inferring T_H1 (>1) or T_H2 (<1) polarization identified an impairment of T_H1 associated responses in aged mice on D28 post-prime (dual immunized), but not D14 post-prime (singly immunized) (Figures S5A and S5B). Ab function was inferred via sera inhibition of RBD binding to recombinant human angiotensin-converting enzyme 2 (hACE2) in a surrogate virus neutralization assay (sVNT), as a correlate of protection.^52,53 Aged mice had lower sVNT than adult mice at multiple immunization doses (Figure S5C). A Spearman’s rank correlation test was performed in both age groups, measuring whether an independent variable, anti-spike protein IgG Ab response, statistically significantly correlates with a dependent variable, sVNT. Spearman’s rho correlation coefficients, measuring the strength (+1 or −1 rho representing perfect associations, and 0, no association) and direction (a positive rho indicating that increasing one variable would increase the other) of the correlation between anti-spike IgG and sVNT, were r = 0.87 and r = 0.75 for adult and aged mouse groups, respectively (Figure S5D). D28 post-prime murine Ab neutralization of live Washington-1 (WA-1) SARS-CoV-2 in vitro demonstrated an impaired aged response compared to adult mice (Figure S5E). Spike peptide splenocyte stimulation induced CD4⁺ T cell IFNγ, IL-2, TNF, and dual stained IL-4 and -5 positivity, alongside CD8⁺ TNF (Figure S6, key resources table). Baseline population differences in CD4⁺ T cell populations were accounted for by dividing mouse BNT162b2-immunized responses by the average of age-matched vehicle control immunized mice. Aged mice had significantly less fold-induction of CD4⁺ T cell IFNγ and TNF cell positivity compared to adult mice (59% and 43% lower median fold induction, respectively, Figure 5B). IL-2 was unchanged, while IL-4/5 demonstrated a lower trend in aged mice (54% lower median aged FC induction) that was not statistically significant. Similarly, CD8⁺ TNF⁺ T cell fold induction was significantly impaired in aged vs. adult mice (45% less median elder FC, Figure 5C). Overall, in vivo murine evaluation mirrored human results with age-associated impaired Ab production, Ab function, class switching, and CD4⁺ and CD8⁺ CMI.

Figure 5.

Reduced humoral and T_H1 cellular immunogenicity of BNT162b2 vaccine in aged mice

BNT162b2-immunized adult and aged mice had humoral immunity evaluated on Day 42 post-prime immunization for receptor binding domain (RBD) responsivity. (A) Total anti-RBD IgG was significantly induced, but with significantly lower Ab titers in aged mice. With FC normalization, aged mice additionally had (B) lower CD4⁺ T cell IFNγ⁺ and TNF⁺ (T_H1) positivity, and (C) lower CD8⁺ T cell TNF⁺ positivity. For (A-C), n = 5 to 10. Boxplots display the median, interquartile range (IQR), with the identification of the furthest values from the median not exceeding 1.5 × IQR. Statistical significance was determined by Shapiro-Wilk, then (A) Kruskal-Wallis and one-sided Wilcoxon rank-sum hypothesizing vaccine-associated induction compared to vehicle control, and two-sided Wilcoxon rank-sum test comparing age groups, (B) two-sided Wilcoxon rank-sum, (C) two-sided T-test, with significance annotated as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Discussion

Herein, we have characterized BNT162b2-induced immunity with the proteomic assessment of age-specific human whole blood stimulation in vitro. We demonstrate that (a) human in vitro modeling of proteomic responses to mRNA vaccines is feasible, (b) such modeling demonstrates marked age-dependent differences in mRNA vaccine-induced analytes including those supportive of T_H1 immunity; (c) impairment in mRNA vaccine-induced T_H1 polarized responses was validated in mice, and (d) analytes identified in vitro (e.g., IFN-γ, CXCL10, IL.1RA, CCL4) corresponded with those that correlated with higher immunogenicity in humans, in vivo, following mRNA vaccination. Vaccines have been crucial in combatting the SARS-CoV-2 pandemic and mRNA vaccines are being assessed for utility against other infectious and non-infectious diseases, yet much remains to be learned regarding their age-specific immunogenicity.^{13,24,25,26,54} As elders have higher rates of severe COVID-19⁵⁵ and reduced vaccine immunogenicity,^{12,13,14,15,28,56} understanding the contributing factors to weaker immunogenicity is an urgent unmet need.

While the mechanisms of mRNA vaccine activation are under active investigation, a possible contribution from self-adjuvantation from ionizable lipids and mRNA can enhance immunogenicity.^57,58,59,60 Self-adjuvantation has been associated with high efficacy in live-attenuated vaccines, arising from pathogen-associated molecular patterns activating and enhancing innate immune responses.^61,62,63 Immune activation following BNT162b2 stimulation may be mediated by pattern recognition receptors (PRRs) such as Toll-like receptor (TLR)-2, −3, −4, −7, and/or −8, as well as retinoic acid-inducible gene I (RIG-I), and melanoma differentiation-associated protein 5 (MDA-5) recognizing multiple vaccine components.^64,65,66,67 Among other PRRs, SARS-CoV-2 can also activate MDA-5 signaling.⁶⁸ Development of vaccine formulations that trigger similar innate immune activation as natural infection may enhance immunogenicity against microbial pathogens.^62,69 Consequently, DEPs from BNT162b2 stimulation were compared to SARS-CoV-2 infection-associated responses. We employed two complementary proteomic approaches for the in vitro evaluation of BNT162b2-induced WBA responses. LC/MS proteomics identified DEPs of BNT162b2-stimulated WBA adult samples compared to vehicle control (Figure 1). Adult samples, but not those from older adults, demonstrated BNT162b2-induced increases in the ATPase PSMC6, hemoglobin HBE1, and the metalloprotease CPN1, each implicated in the host response to SARS-CoV-2, COVID-19 severity, and/or have anti-viral activity.^{70,71,72,73,74,75} Additionally, stimulated adult, but not elder, samples had reduced peptidylprolyl isomerase A (PPIA), lower plasma concentrations of which are associated with better COVID-19 prognosis.⁷⁶

Elder participants’ proteomic responses were markedly distinct from adults. There was only a single overlapping downregulated protein between the age groups assessed by LC/MS, DEFA3 (Figure 1), which has been associated with lipid envelopes.⁷⁷ Adult participants had 19 up- and 2 down-regulated proteins, while elder participants had a starkly contrasting 3 up- and 10 down-regulated proteins, highlighting divergent immune responses (Figure 2A). Unique proteins included SAA1, FGG, and FGB which were induced in elder BNT162b2-stimulated samples, but not adults, and have been associated with SARS-CoV-2 infection and/or COVID severity.^78,79,80,81 Downregulated DEPs in BNT162b2-stimulated elder samples, compared to vehicle control stimulations, included TPM4, APOF, APOC2, CPN1, and PI16, which were also downregulated by exposure to SARS-CoV-2 virions in humans in vitro and in vivo, associating with poor prognosis.^75,82,83,84 Overlap of impaired elder BNT162b2 responses with factors that have been associated with disease susceptibility may reflect important common signaling pathways shaped by immunosenescence that may contribute to both COVID susceptibility and impaired vaccine responses. A secondary guided PEA-based proteomic assay validated results (Figures 1C and 1D), observing similar age-dependent patterns of BNT162b2-induced adult up and elder downregulation as the LC/MS proteomics, but with distinct analytes. The striking differences in proteomic responses between adult and elder participants may contribute to age-dependent differences in BNT162b2 immunogenicity.

Directly comparing the proteome derived from BNT162b2-stimulated adult and elder WBA supernatants via LC/MS (Figure 2B) and PEA (Figure 2C) demonstrated marked age-dependent differences. Antibody and bead-based fluorescent multiplex analysis revealed that elders had significantly (p = 0.04) impaired BNT162b2-stimulated chemokine CCL4 production (Figure 2D), though the functional categorization of multiplex-quantified chemokines did not identify broad differences in chemokine induction (Figure S3). CCL4 induction has been negatively correlated with age,²⁰ potentially impacting monocyte and antigen-presenting cell (APC) chemotaxis to the injection site and lymph nodes, respectively,^85,86,87,88 both of which would impact adaptive immunity. Network analysis of DEPs (Figures S2E and S2F) further delineated lower mRNA vaccine-induced activation in elderly blood. In contrast to young adults, elder participant samples lacked BNT162b2-induced activation of the ‘IL-4 and IL-13 signaling’ network which supports T_H2, B cell differentiation, and class switching.⁸⁹ Additionally, while adults had an unaltered pathway, elders demonstrated BNT162b2-induced down-regulation of the “immunoregulatory interactions between a lymphoid and non-lymphoid cell” pathway. An associated analyte within this pathway, Cytotoxic and regulatory T cell molecule (CRTAM), supports CD4⁺ and CD8⁺ T cell differentiation,⁹⁰ such that downregulation in elders may contribute to reduced CMI.

BNT162b2 stimulation in a whole blood assay (WBA) resulted in dose-dependent induction of multiple analytes, including IL-6, CXCL8, TNF, and IFNγ (Figure 3A). The WBA induced similar analytes as those from adult human in vitro stimulations with LNP encapsulated mRNA (encoding non-SARS-CoV-2 antigens and with a distinct cationic lipid, SM-102),⁹¹ IL-1β, IL-1RA, IL-6, TNF, CCL2, CCL3, and CCL4 (Figures 4 and S3), indicating similar immune activation of peripheral lymphocytes in a WBA. Myocarditis, a serious mRNA vaccine-associated adverse event, has been correlated with vaccine-induced plasma IL-1β, IL-1RA, IL-15, CCL4, CXCL1, and CXCL10,⁹² each of which was also significantly induced in vitro with WBA BNT162b2 stimulation (Figures 4 and S3), demonstrating the potential utility of the WBA platform for studying vaccine safety. Lower mRNA vaccine-induced reactogenicity in older adults has been associated with lower immunogenicity,⁹³ therefore mRNA vaccine activation of the innate immune system may concurrently contribute to both mRNA vaccine self-adjuvantation and reactogenicity.

BNT162b2-induced WBA cytokine and chemokine induction were age-dependent with consistently observed lower elder participant responsivity across LC/MS proteomics, PEA proteomics, and multiplex platforms. Functionally, T_H1-polarized immunity is observed after BNT162b2 and mRNA-1273 immunization,^94,95,96 but is not equally induced across multiple age groups. Human elders have impaired induction of antibody isotypes IgG1 and IgG3,⁹⁷ associated markers of human T_H1 polarization,^98,99 and also have had direct impairment of CD4⁺ and T_H1 polarized T cell responses following BNT162b2-immunization.^15,100 Specific contributing mechanisms to age associated BNT162b2 responsivity have not yet been described. Post-BNT162b2 mRNA vaccination, systemic IL-1RA, and the T_H1-polarizing CXCL10, CCL4, and IFNγ^{101,102,103,104,105,106,107,108} production have been associated with higher SARS-CoV-2 specific immunogenicity in 28–59-year-old adults.⁵⁰ Of note, older adults demonstrated lower production of IL-1RA, CXCL10, CCL4, and IFNγ in vitro (Figures 3B and 3C), analytes associated with higher human immunogenicity in vivo.⁵⁰ These analytes may be pivotal as the T_H1 polarizing cytokine IFNγ can suppress T_H2-associated IL-4, and increase expression of IL-12p70 and its receptor.¹⁰⁷ Indeed, blocking IFNγ has been shown to impair BNT162b2 responsivity in adult mice.¹⁰⁹ These observations are consistent with the impaired production of IFNγ by monocyte derived Dendritic Cells (MoDCs) from older vs. young adults to other lipid nanoparticles.¹¹⁰ Reduced production of CXCL10 by older adults may constrain vaccine immunogenicity as exogenous CXCL10 incubation with naive T cells can support T_H1 and T_H17 cell differentiation,¹⁰⁴ DC-produced CXCL10 can promote lymph node DC-T-cell interactions during naive cell priming,¹¹¹ and knockout mice have impaired antigen-specific T cell responses.¹⁰⁸

Interpretation of polyfunctional cytokines was validated by functionally grouping analytes to measure broadly dysregulated T_H1, T_H2, T_H17, or T_reg polarizing, chemokine, hematopoiesis-supporting, or trained immunity-associated responses. Pairing individual analysis with functionally grouped analytes (Table S3) can broadly describe function-based differences to account for analyte redundancies.¹¹² Importantly, the functional assignment was not just based on being produced by a polarized cell but rather required evidence of supporting or being a polarizing molecule of naive T cells. A conservative GEEGLM analysis was employed, averaging the induction of functionally grouped analytes, including those that were not individually induced, thereby biasing toward no difference, so that only broad and significant differences would be observed. GEEGLM analysis identified a significant reduction (average 7.2% across analytes, p = 0.027) of T_H1 polarizing analyte induction in elder WBA responses (Figure 4). The other 6 functions evaluated were not significant, indicating no impairment, or sample size limitations. Age impacts DCs, monocytes, natural killer, and T cells,^113,114,115 and additional investigation is needed to identify which specific cell types have age-impaired mRNA vaccine responses. We observed decreased production of multiple analytes, particularly a decrease in those polarizing toward T_H1, in human elder samples, compared to adult samples. Immunophenotyping the cellular origin of mRNA vaccine-induced cytokine production is an important consideration and warrants further investigation.

Investigating BNT162b2-induced immune activation in vitro offers significant insights into species (i.e., human)- and age-specific responses, but may not completely reflect relevant vaccine responses in vivo. To assess for correlates of our in vitro observations in vivo, we employed an age-specific murine mRNA vaccination model.^116,117,118 Increased age is associated with impaired human humoral immunity following BNT162b2 or mRNA-1273 vaccination.^1,13,31,119 Similar to humans, aged mice demonstrated impaired Ab induction at all immunization doses (Figure 5A), and waning immunity was more rapid in aged mice (Figure S4).^31,120 Higher antigen doses resulted in enhanced immunogenicity in aged animals, restoring humoral titers and sVNT to levels observed in younger animals (Figures 5A and S5). Higher antigen doses for older adults have been recommended to be evaluated for mRNA SARS-CoV-2 vaccines,¹³ and which may be tolerated given generally lower elder reactogenicity.⁹³ T_H1/T_H2 polarization was indirectly inferred by a relative ratio of IgG2a and IgG1 Ab isotypes, respective T_H1 and T_H2 BALB/c markers.⁵¹ IgG2a/IgG1 after first immunization was moderately T_H2-shifted and not different between both age groups, while the post-booster was T_H1-shifted in adult, but not aged, mice (Figure S5), potentially indicating booster-specific impairments that may impact future vaccination campaigns. Immune polarization changes over time, with a shift toward IgG4 responses after repeated immunizations in humans, suggesting a shift toward the associated marker of anti-inflammatory and T_H2 polarization.^121,122 Sera neutralization capacity, an important correlate of protection,^{52,53,123,124} was impaired in both sVNT and live-virus WA-1 SARS-CoV-2 assays in aged mice (Figure S5). The murine setting enabled a controlled environment, minimizing confounders, and mirrored age-dependent human observations of neutralization.¹²⁵ By Spearman’s rank correlation, anti-spike IgG significantly correlated with sVNT immunity in both age groups (Figure S5). T cell support is essential to effectively develop humoral immunity and cytotoxic immunity against infected cells. T_H1-polarizing analytes and Ab isotypes infer polarization states, and direct T cell evaluation can provide additional insight. Stimulation of murine splenocytes with spike-specific peptide induced IFNγ and TNF in CD4⁺ T cells, indicating T_H1 polarization,¹²⁶ with 43–59% less median induction of cell positivity in aged vs. young adult mice (Figure 5B). Additionally, aged mice had significant impairments in BNT162b2-induced TNF⁺ CD8⁺ T cells (Figure 5C), an important cell subset for lysing infected cells.¹²⁶ Reduced BNT162b2 immunogenicity in aged mice included lower vaccine-induced Ab titers, impaired Ab neutralization capacity, diminished CD8⁺ T cell activity, and reduced T_H1 polarization of CD4⁺ T cells, coupled with more rapid waning immunity, indicating the utility of murine models to investigate age-associated changes in immunity.

The observed impaired BNT162b2-induced T_H1 immunogenicity in aged mice and older human adults reflects distinct immunity with age. Multiple approaches may amplify immunogenicity, including: (a) dose-escalation,¹³ (b) additional boosters to extend protection,^123,127 albeit temporarily, due to rapidly waning immunity in elders,³¹ and (c) use of T_H1-polarizing adjuvants targeted toward elder populations,⁹ including Alum:CpG, saponin, or MF59,^{62,128,129,130,131} and potentially the TLR7/8 agonist Alhydroxiquim-II,¹³² to enhance host defense against intracellular pathogens,^133,134 B cell class switching (human IgG1 and IgG3,^98,99 or murine IgG2a⁹⁸), and support T_FH-independent B cell responses.¹³⁵

Our study features multiple strengths, including (a) the use of a human WBA in vitro that is replete with age-specific cellular and soluble factors that preserve physiological states, and which may be predictive of vaccine responses in vivo,^35,136 (b) the use of three complementary proteomic approaches (mass spectrometry, PEA and multiplex assay) to gain a comprehensive view of the impact of BNT162b2 on the WBA proteome, and (c) validation of findings using aged vs. adult mice in vivo. The use of human in vitro assays enables human- and age-specific modeling with individuals serving as both control and test conditions, permitting paired analyses of new and established/licensed vaccine formulations, thereby accelerating and de-risking vaccine discovery and development.^34,35,36 Indeed, the U.S. FDA Modernization Act 2.0, signed into law in 2022, provides for the use of human in vitro systems coupled with bioinformatic biomarker analysis to advance drug and vaccine development.³⁷

In summary, supernatants from adult and elder WBA demonstrated distinct BNT162b2-induced immune activation patterns by LC/MS and PEA proteomics, with BNT162b2-induced adult upregulation and elder downregulation. LC/MS DEP profiles were markedly age-dependent, with only 1 overlapping significant protein downregulated in both adults and elders (DEFA3). Cytokine and chemokine multiplex demonstrated a vaccine concentration-dependent response in human adults in vitro, including IL-6, CXCL8, TNF, and IFNγ production. Functional categorization of analytes identified impaired T_H1-polarizing analyte induction in elder participants, potentially contributing to reduced immunogenicity. Murine in vivo experiments mirrored impaired humoral induction, T_H1 polarization, reduced IgG2a/IgG1 relative ratio, and a directly reduced CD4⁺ T cell IFNγ and TNF response to SARS-CoV-2 spike peptide. Our study has demonstrated the value of a human in vitro platform coupled with proteomic systems biology to model age-specific responses to the mRNA vaccine BNT162b2. As the U.S. FDA increasingly welcomes human in vitro data,³⁷ this approach may have broad applicability to advance mRNA vaccines that remain essential for combatting coronaviruses and hold great promise for protection against additional infectious diseases (e.g., influenza, RSV, and HIV), and in oncology.^57,137 Identifying age-specific responses to mRNA vaccines will inform the discovery and development of the next generation of precision mRNA vaccines to overcome immunosenescence. Given the potentially massive benefits of optimized vaccines,¹³⁸ further translational research is needed to enhance elder immune responses including expanded adjuvantation efforts to enhance T_H1 polarization, durable immunogenicity, and protection,⁹ including through the addition of mRNA encoding IL-12 to adjuvant mRNA vaccines.^139,140

Limitations of the study

As with any research effort, our study has multiple limitations, including (a) grouping into adult (18-50Y) and elder (≥60Y) categories (as in^{141,142,143,144,145}) precludes other sub-age groups (e.g., 50-60Y, >80Y,^13,144 or >100Y¹⁴⁶), (b) vaccine and infection history were self-reported and differential between groups, (c) the study is underpowered for small effects and therefore some observations of no significant difference may be Type II errors-i.e., failure to detect real differences, d) sample size precluded the assessment of confounders (e.g., obesity, corticosteroids, immunosuppression, cardiovascular disease, and smoking), and covariates (e.g., sex, gender, previous vaccination or infection, and so forth), requiring further investigation, (e) the in vitro WBA assay lacks fluid flow and tissue interaction (e.g., muscle), and did not identify cellular origin of T_H1 polarizing cytokine and chemokines, (potentially monocyte, macrophage, dendritic cells¹⁰⁹), (f) study participants were not representative of global populations, (g) immune proteins have redundancies¹¹² lost during immunosenescence,¹⁴⁷ and (h) age-specific investigation of mice may not be directly translatable to humans.

Resource availability

Lead contact

Request for further information, resources, and reagents can be directed to the lead contact, Dr. Ofer Levy (ofer.levy@childrens.harvard.edu).

Materials availability

This study did not generate novel reagents.

Data and code availability

• •Deidentified quality assured human data from this study is deposited in the repository ImmPort:SDY2630, as listed in the key resources table. Further inquiries could be directed to the corresponding author. Murine data will be made available upon requests submitted to the corresponding author.
• •This article does not report the original code.
• •For other items, please contact the corresponding author.

Acknowledgments

We thank the BCH Pharmacy and all our study participants, including members of the congregation of Eitz Chayim Synagogue (Cambridge, MA). We thank Dr. Benoit Fatou for assistance with proteomic sample processing and data analysis, Dr. Kinga Smolen for helpful discussions, as well as Drs. Sirano Dhe-Paganon and Hyuk-Soo Seo for the production of the recombinant spike and RBD antigens used in this study. We thank B. S. Graham (NIH Vaccine Research Center) for providing the plasmid for prefusion stabilized SARS-CoV-2 spike trimer. Katherine Chew, Maria DeLeon, Gandolina Melhem, and Yamile Lugo Rodriguez provided helpful technical support. We thank Dr. Emilie Clement, and Dr. Daniel Frederick of Olink for logistical and technical support. We thank the pharmacists of Boston Children’s Hospital for efforts to maximize the use of SARS-CoV-2 vaccines by saving leftover (overfill) of otherwise-to-be-discarded vaccine vials, and Meagan Karoly and Caitlin Syphurs of the Precision Vaccines Program Data Management & Analysis Core for data deposition support. We thank Dr. Asimenia Angelidou and Project Manager Kerry McEnaney for supporting phlebotomy efforts. E.N. is a JSPS Overseas Research Fellow and a joint Society for Pediatric Research and Japanese Pediatric Society Scholar. We thank Dr. Natalie Thornburg and the CDC for providing WA-1 SARS-CoV-2. D.J.D. thanks S. McHugh, G. Boyer, L. Conetta and the staff of Lucy’s Daycare, the staff the YMCA of Greater Boston, Bridging Independent Living Together (BILT) Inc., and the Boston Public Schools for childcare and educational support during the COVID-19 pandemic.

Author contributions

B.B. conceived, designed, performed, and analyzed the in vitro and in vivo experiments, and wrote the article. B.F. performed and analyzed LC/MS proteomics experiments. A.K.C. and J.A. analyzed PEA proteomics experiments. S.B. conceived, designed, performed, and analyzed flow cytometry experiments. C.S., A.N.B., and T.R.O. assisted with murine data acquisition. M.E.M., M.F. designed and performed true neutralization experiments. A.O., J.A., and H.S. advised analytical approaches. A.S., L.B., E.M., G.S.S., P.P., E.N., D.S., S.H., and H.S. contributed to experimental design discussions. All authors critically reviewed the article. D.J.D. and O.L. conceived the project, assisted with the design of the experiments, mentored B.B., and edited the article.

Declaration of interests

O.L. has served as a consultant to GlaxoSmithKline (GSK) and Hillevax. M.B.F. serves on the scientific advisory board of Aikido Pharma and has collaborative research agreements with Novavax, AstraZeneca, Regeneron, and Irazu Bio. B.B., E.N., T.R.O., D.S., S.H., O.L., and D.J.D. are named inventors on vaccine adjuvant patent(s). O.L., G.S.S., and D.J.D. are named inventors on patents related to human in vitro modeling of vaccine responses. O.L. and G.S.S. are recipients of a sponsored research agreement with GSK. D.J.D is on the scientific advisory board of EdJen BioTech and serves as a consultant with Merck Research Laboratories/Merck Sharp & Dohme Corp. (a subsidiary of Merck & Co., Inc.). O.L. and D.J.D. are co-founders of and advisors to Ovax, Inc. ACS and LRB are involved in HIV, COVID, and other vaccine clinical trials conducted in collaboration with the NIH, HIV Vaccine Trials Network (HVTN), COVID Vaccine Prevention Network (CoVPN), International AIDS Vaccine Initiative (IAVI), Crucell/Janssen, Moderna, and Sanofi. These commercial or financial relationships are unrelated to the current study.

The participating Precision Vaccines Program (PVP) laboratories were supported in part, by U.S. National Institutes of Health (NIH)/National Institutes of Allergy and Infectious Diseases (NIAID) awards, including Human Immunology Project Consortium award U19 AI118608, Adjuvant Discovery (HHSN272201400052C) and Development (HHSN272201800047C) Program Contracts to O.L.; Adjuvant Discovery Program contract (75N93019C00044) to O.L. and D.J.D as well as NIH grant (1R21AI137932-01A1) to D.J.D. O.L. is also funded by an award from the Coalition for Epidemic Preparedness Innovations (CEPI), via the International Network of Special Immunization Services (INSIS). The PVP is supported, in part, by the BCH Department of Pediatrics and philanthropy via the BCH Trust, including from the Barry Family and the Boston Investment Council. A.K.C. was supported by the Friedman Award for Scholars in Health, the University of British Columbia, and Mitacs Accelerate Canada.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Viability dye, Fluorochrome LIVE/DEAD Aqua, at 1:500	Invitrogen	L34966
Anti-CD3, Clone 17A2, Fluorochrome Brilliant Violet 785, at 1:40	BioLegend	Cat# 100232; RRID: AB_2562554
Anti-CD4, clone RM4-5, Fluorochrome APC/Fire 750, at 1:160	BioLegend	Cat# 100568; RRID: AB_2629699
Anti-CD8, clone 53-6.7, Fluorochrome Brilliant UltraViolet 395 (BUV395), at 1:80	BD Biosciences	Cat# 563786; RRID: AB_2732919
Anti-CD44, clone IM7, Fluorochrome PerCP-Cy5.5, at 1:160	BioLegend	Cat# 103032; RRID: AB_2076204
Anti-IFNγ, clone XMG1.2, Fluorochrome Alexa Fluor 488, at 1:160	BioLegend	Cat# 505813; RRID: AB_493312
Anti-IL-2, clone JES6-5H4, Fluorochrome PE, at 1:40	BioLegend	Cat# 503808; RRID: AB_315302
Anti-TNF, clone MP6-XT22, Fluorochrome PE Cy7, at 1:160	BioLegend	Cat#506324; RRID: AB_2256076
Anti-IL-4, clone 11B11, Fluorochrome BV421, at 1:40	BioLegend	Cat# 504119; RRID: AB_10896945
Anti-IL-5, clone TRFK5, Fluorochrome BV421, at 1:160	BioLegend	Cat# 504311; RRID: AB_2563161
Anti-mouse IgG	Southern Biotech	Cat# 1036-05; RRID: AB_2794348
Anti-mouse IgG2a	Southern Biotech	Cat# 1081-05; RRID: AB_2736843
Anti-mouse IgG1	Southern Biotech	Cat# 1071-05; RRID: AB_2794426
Anti-RBD Fc chimera	R & D	Cat# 10499-CV-100; RRID: N/A
Anti-Human IgG Fc-HRP	Southern Biotech	Cat# 2048-05; RRID: AB_2795688
Biological samples
Heparinized human whole blood (18–50 Y)	Precision Vaccines Program	NA
Heparinized human whole blood (>60 Y)	Precision Vaccines Program	NA
Heparinized human whole blood (>60 Y)	Brigham and Women’s Hospital	NA
Chemicals, peptides, and recombinant proteins
Heparin	American Pharmaceutical Partners Inc.	NDC71288-402-10
Urea	Sigma Aldrich	LC/MS grade
Ammonium bicarbonate	Sigma Aldrich	09830-1KG
Dithiothreitol	Sigma Aldrich	D9779-10G
Iodoacetamide	Sigma Aldrich	I1149-25G
Sera-Mag Speed Beads 65	Sigma-Aldrich	65152105050250
Sera-Mag Speed Beads 45	Sigma-Aldrich	45152105050250
HPLC-grade water	Sigma Aldrich	W5-4
Trypsin	Promega	V5117
Formic acid (LC/MS grade)	Thermo Scientific	A117-50
Recombinant RBD (R319-K529)	Nanishi et al.130	GenBank MN975262.1
Recombinant spike (M1-Q1208)	Nanishi et al.130	GenBank MN90894
Tetramethylbenzidine BD OptEIA	BD Biosciences	555214
Human ACE2	Sigma-Aldrich	SAE0064
Ammonium-Chloride-Potassium lysis buffer	Gibco	A10492-01
Acridine Orange/Propidium Iodide	Nexcelom	CS2-0106
PMA + ionomycin (at 1:500)	BioLegend	423301
Brefeldin (at 1:1000)	BioLegend	420601
Peptivator, wild type spike peptide pool	Miltenyi Biotec	130-126-700
Cytofix/Cytoperm	BD	554714
Critical commercial assays
Target 96 Inflammation panel	Olink	N/A
Target 96 Cardiometabolic panel	Olink	N/A
Target 96 Oncology III panel	Olink	N/A
Target 96 neurology panel	Olink	N/A
41-plex multiplex	Milliplex	HCYTOMAG-60K
Deposited data
Human LC/MS proteomics data	ImmPort	SDY2630
Human PEA proteomics data	ImmPort	SDY2630
Human bead-based multiplex cytokine data	ImmPort	SDY2630
Experimental models: Organisms/strains
BALB/c, female, >10 months	Envigo	BALB/cAnNHsd
BALB/c, female, 6–22 weeks	Envigo	BALB/cAnNHsd
SARS-CoV-2 (courtesy of Dr. Natalie Thornburg and CDC)	Nanishi et al.130	WA-1
Software and algorithms
Skyline	MacLean et al.148	V20.2.1.315
R	R Development Core Team149	Versions 3.3.2 and 4.1.1
R Studio	RStudio Team150	Versions 1.3.1093, 2022.02.3 + 492, 2022.07.1 + 554 2023.03.1 + 446, 2023.06.1 + 524
AnnotationDbi	Pagès et al.151	1.62.2
ReactomePA	Yu and He152	1.44.0
DOSE	Yu et al.153	3.26.1
Ggraph	Pedersen154	2.1.0.9000
Geepack	Søren Højsgaard et al.155	1.3.9
mixOmics	Rohart et al.156	6.16.3
Limma	Ritchie et al.157	3.60.0
Plyr	Wickham158	1.8.7
dplyr	Hadley Wickham et al.159	1.0.9
reshape2	Wickham160	1.4.4
ggplot2	Wickham161	3.3.6
ggpubr	Kassambara162	0.4.0
ggfortify	Tang et al.163 and Horikoshi et al.164	0.4.14
ggforce	Pedersen165	0.3.4
broom	Robinson et al.166	1.0.0
ggradar	Bion et al.167	0.2
tidyverse	Wickham et al.168	1.3.2
scales	Hadley Wickham169	1.2.0
FlowJo	BD	v.10.8.1
Biorender	Biorender.com	N/A
Other
BNT162b2 monovalent wildtype SARS-CoV-2 vaccine overfill	Boston Children’s Hospital pharmacy	EW0181
Nexera Mikros	Shimadzu	N/A
Macrospin C18 plate	The Nest Group Inc.	SNS SS18VL
Capillary C18 Column	Shimadzu	227-32100-02
LCMS-8060 triple quadrupole mass spectrometer	Shimadzu	N/A
Nonheparinized capillary tubes	Drummond	1-000-1000
High-binding 96-well plate	Corning	9018
SpectraMax iD3 microplate reader	Molecular Devices	N/A
Nexcelom Cellometer K2	Nexcelom	N/A
LSRFortessa	BD Biosciences	N/A, custom

Experimental model and study participant details

Human participant inclusion and exclusion criteria

Inclusion criteria included study participants >18 years of age who had responded to recruitment efforts, recruited between January-April, 2021, and who could give informed consent and who were willing and able to donate >75 mL blood. Participants were excluded if they had symptoms of an active infection (e.g., >38 C temperature), were recently immunized (14 days for non-live vaccines, 28 days for live attenuated), had donated >220 mL blood within the past 5 weeks, had taken anti-inflammatory medication that day, or taken immunosuppressants (e.g., corticosteroids, chemotherapy) within 3 months. Consenting participants had blood drawn, as summarized in the method details section, below.

Mice

BALB/c animals (BALB/cAnNHsd) were purchased from Envigo and housed at BCH. Female adult mice were used between 6 and 22 weeks of age, and aged female retired breeder mice were utilized within 43–59 weeks of age, a similar age group as used in other studies of immunosenescence,¹⁷⁰ communally housed with similarly aged animals. Adult mice were sex-matched to the availability of the aged female mice. Ear clipping enabled mouse tracking, and guided randomization balanced treatments across cages, to reduce variability. Mice were injected with 0.5, 1.0, or 5.0 μg of mRNA within monovalent wildtype BNT162b2, administered in 50 μL inoculum to the mouse’s right hindlimb via intramuscular (IM) injection in either conscious or isoflurane-anesthetized mice. A prime-boost schedule was followed, separated by 14 days. Methodology for the evaluation of post-vaccination mouse serum for anti-spike immunity, antibody isotype, surrogate virus neutralization, and true neutralization as well as splenocyte dissection, dissociation, and CD4⁺ and CD8⁺ T cell stimulation, staining, and flow cytometry are summarized in the method details section, below.

Method details

mRNA vaccine

In vitro (human) and in vivo (murine) studies employed residual overfill, after removal of injectant for human immunization, of the monovalent wildtype Pfizer/BioNTech BNT162b2 mRNA vaccine from the BCH Pharmacy (February 2020-August 2022), within 12 h of vial puncture. Only monovalent BNT162b2 (encoding wildtype SARS-CoV-2 spike protein) was used.

Human participant sample processing

Heparinized whole blood was collected from adult (18-50Y), and older adult (elders, ≥60Y) study participants. Demographics, self-reported SARS-CoV-2 infection, and vaccination history were summarized in Table S1 for participant samples evaluated by LC/MS proteomics and multiplex, and in Table S2 for PEA evaluation. Clinical data on co-morbidities (e.g., obesity, corticosteroids, immunosuppression, cardiovascular disease, smoking, and other respiratory impacts) were not collected. Blood was drawn into a final 20 units/mL of clinical grade, pyrogen-free heparin (American Pharmaceutical Partners Inc.). WBA stimulation was performed as in^171,172 with a few modifications. Specifically, 125 μL blood was mixed 1:1 with RPMI 1640 (Gibco 11875-119) plus stimuli in 96-well U-bottom tissue culture plates (Becton Dickinson) and stimulated for 24 h in a 37°C humidified, 5% CO₂ incubator. Per-well stimulations of 0.002, 0.02, 0.2, 0.67, and 2.0 μg/mL of mRNA contained within BNT162b2 (0.1 μg mRNA/μL), corresponds to 0.002, 0.02, 0.2, 0.67, and 2%, % v/v, respectively. Cell-free supernatant was collected post-centrifugation (500g, 10 min), and stored at −80°C.

Evaluation of human culture supernatant

The impact of in vitro stimulation with BNT162b2 in the WBA was quantified by targeted plasma proteomics (liquid chromatography, mass spectrometry, LC/MS), proximity extension assay (PEA, 4x Olink Target 96 platforms) proteomics, and bead-based multiplex quantifying inflammation and chemotaxis mediators. For proteomics, stimulated sample supernatants were randomized to avoid batch effects. Each assay, and methodology for analyte functional categorization is elaborated below.

Targeted plasma proteomics sample preparation

WBA samples were evaluated by LC/MS proteomics observing a dose-titration of BNT162b2-stimulated whole blood (WB). All chemicals and reagents were purchased at the highest purities available. Solvents used in this study were LC/MS grade and purchased from Fisher Chemicals (Thermo Fisher Scientific). Briefly, a volume of 10 μL of 10-fold diluted plasma was mixed with 60 μL of urea buffer (8M urea in 50 mM ammonium bicarbonate, Sigma Aldrich) and 15 μL of dithiothreitol buffer (DTT, 50 mM in urea buffer, Sigma Aldrich) before being incubated for 30 min on a thermomixer (800 rpm, room temperature, RT). The samples were alkylated with iodoacetamide buffer (375 mM in urea buffer, Sigma Aldrich) and incubated for 30 min (800 rpm, RT and dark). A volume of 10 μL of DTT buffer was added to quench the alkylation. The samples were transferred to the SP3 beads mixture (Sera-Mag SpeedBeads, 1:1 v/v, GE Healthcare) previously washed with high-performance liquid chromatography (HPLC)-grade water (Sigma Aldrich) at a 1:10 protein to bead ratio. A volume of 150 μL of absolute ethanol (Superlco) was added and incubated 15 min on a thermomixer (1,000 rpm at RT). The samples were placed on the magnetic rack and then the clear supernatant was removed. The beads were washed three cycles in 200 μL 80% ethanol. After the final washing step, the samples were trypsinized with 100 μL of trypsin buffer (Promega, 20 μg/mL in 50 mM ammonium bicarbonate) and placed on thermomixer (1,000 rpm, 2 h, 37°C). After digestion, samples were centrifuged to pull down the liquid and placed on magnetic rack to collect the supernatant and were acidified with 2% v/v formic acid in HPLC water. The C18 cleanup was performed using a 96-well MACROSPIN C18 plate (TARGA, The NestGroup Inc.) and the tryptic peptides were eluted off the C18 particles using 40% ACN/0.1% FA. The samples were then dried and stored at −20°C until LC/MS analysis. The samples were analyzed using an LC system (Nexera Mikros, Shimadzu) equipped with Capillary C18 column (0.2 × 100mm, 2.7 μm particle diameter, Shimadzu) coupled online to an LCMS-8060 triple quadrupole mass spectrometer instrument (Shimadzu). From each sample, 1 μg peptide quantity was separated using a non-linear gradient over 15-min run time operated at 10 μL/min (5% solvent B for 0.2 min; 5 to 40% B for 10.3 min; 85% B for 1.5 min and 5% for 3 min). The final scheduling method was performed using the following parameters: 1.2 s of maximum loop time with minimum dwell time of 2 msec and pause time of 1 msec, Q1 and Q3 resolution set at the ‘unit’ level.

Proximity extension assay (PEA)

To broaden the range of proteins measured, supernatants from human in vitro WBA assays were also evaluated by a Proximity Extension Assay (PEA) with Olink technology, as in,¹⁷³ following manufacturer recommendations. An n of 5 adult and 5 aged participants were evaluated, and an n 5 adult and 4 elder passed blinded quality control measures. Of note, due to quality control warnings, one elder research participant was removed from the original n 5, according to the manufacturer’s sample-blinded quality control recommendations. These participant samples were a subset of those investigated by LC/MS proteomics. Samples were selected while blinded to outcomes and were chosen to reduce the number of sample freeze thaw cycles. Proteins were labeled with a mixture of antibodies containing pairs of antibodies tagged with a DNA barcode that were able to recognize the same protein. Antibodies binding to the same target, in close proximity, have DNA tag hybridization, undergo DNA polymerase-dependent extension, subsequent PCR amplification, and next generation sequencing (NGS). The dual antibody binding and PCR amplification resulted in high specificity and sensitivity to evaluate normalized protein expression (NPX). Four Target 96 panels (Inflammation, Cardiometabolic, Oncology III, and Neurology) were assayed by Olink under a service agreement. PEA assay-quantified proteins were labeled with the platform name (Inflammation, ‘inflam’; Oncology, ‘onco’; Neurology, ‘neuro’; Cardiology III, ‘cardio’). A total of 368 proteins were evaluated while blinded to age group, and analyses were performed at BCH. For conditions comparing LNP-stimulations between age groups baseline-normalization via subtracting vehicle control from LNP-stimulated conditions to reduce inter-assay variability and were evaluated for differential (up/down regulation) normalized protein expression in samples stimulated with 2 μg/mL of encapsulated BNT162b2 mRNA against vehicle (RPMI) controls. PEA heatmap analysis was unsupervised to evaluate if patterns of LNP-induced proteins could differentiate stimulated from non-stimulated in adults and elders. Euclidean-clustering was applied to evaluate BNT162b2-stimulated adult clustering and elder non-clustering. The top 100 differentially expressed proteins were converted to Entrez IDs with the AnnotationDbi package and then enrichment analysis was performed with ReactomePA. Plotting of network interactions involved cnetplot and ggraph. Network analysis nodes were sized by the number of contributing proteins.

Bead based multiplex

Human samples from WBA were also evaluated by a bead-based multiplex platform measuring 41 analytes (Milliplex HCYTOMAG-60K) following the manufacturer’s recommendations and excluding samples with insufficient bead counts (requiring ≥30 beads/analyte).

Functional categorization of analytes

Selection of individual significantly induced proteins increases the risk of interpretation bias due to analyte polyfunctionality and the potential of false positives. We augmented the classical approach of individual analyte interpretations from multiplex assays by additionally analyzing based on functional categorization to evaluate if age significantly interacted with each function. Immunosenescence could be driven by differential production of analytes capable of polarizing naive T cells toward CD4⁺ T helper cell (T_H) 1, T_H2, T_H17, and T_reg differentiation, and those supporting chemotaxis, hematopoiesis, and/or associated with secondary effects of vaccine (e.g., trained immunity, nonspecific effects). T_H1 polarized immune responses can trigger effective intracellular pathogen responses,¹³³ including CD8⁺ T cell-mediated immunity,¹³⁴ B cell class switching^98,99 and induction of T_FH-like activity for effective B cell responses in the absence of T_FH.¹³⁵ T_H2 responses can support Ab production but can bias toward IgE Ab class switching with potential age-dependent differences.^98,174 T_H17 has been associated with B cell differentiation and class switching to IgA,^175,176,177 with increased mucosal immunity.¹⁷⁸ Chemokine responses are critical for mounting an effective immune response,¹⁷⁹ through both initial recruitment of monocytes to the vaccination site, and subsequent chemotaxis of mature antigen-presenting cells (APC) to the draining lymph node.^85,86,87,88 T_reg can restrain germinal center reactions.^130,180 Hematopoiesis-associated factors could be important immunoregulators, as impaired hematopoiesis has been associated with reduced vaccine responsiveness in the aged.^181,182 mRNA vaccines may also induce trained immunity.^38,183 Impact of age on each function was evaluated through a targeted multiplex cytokine and chemokine assay measuring 41 predominantly polyfunctional analytes.

A literature review informed classification of each of the 41-plex measured analytes into the functional categories of T_H1, T_H2, T_H17, and T_reg polarizing, and chemokine, hematopoiesis, or vaccine associated trained immunity inducing functions. Particular attention to differentiate polarizing activity from analytes that were produced by polarized cells was performed. This review included various gene ontology (GO) terms, including ‘T cell differentiation’ (GO: 0030217, sub-divided to T_H1 or T_H2 or T_H17 polarizing), ‘T-helper 17 cell lineage commitment’ (GO: 0072540), ‘regulatory T cell number’ (GO: 0045066), ‘Chemokine’ (GO: 0032602), ‘Chemotaxis’ (GO 0006935), and ‘Hematopoiesis’ (GO: 0030097). Supplemental targeted searches of each function, and “polarizing” or “polarized,” in the case of T_H-polarizing activity, in the Google Scholar database between Aug-Oct 2021. Evidence from human sources was prioritized but supplemented with murine where human observations were not available. Categorization into the CD4⁺ T cell polarizing capacities required evidence of being required for polarization, or inducing polarization itself, rather than being induced by a polarized cell. The other evaluated functions included direct and indirect chemokine activity, hematopoiesis support or induction, and mediating secondary effects of vaccines.

Murine SARS-CoV-2 specific antibody evaluation

At 14, 28, 42, and 210 days post-prime immunization animals were anesthetized under 3% isoflurane and had 100-200 μL of blood collected by retroorbital bleed into non-heparinized glass capillary tubes (Drummond Cat. 1-000-1000). Prompt expelling of blood into microcentrifuge tubes was followed by allowing samples to clot. Blood was centrifuged within 2 h (1500g, 7.5 min), transferred to new microcentrifuge tubes, recentrifuged, and serum was aliquoted for storage at −80°C.

Anti-spike and anti-RBD titers were evaluated by ELISA as in.¹³⁰ In brief, flat-bottomed high-binding 96-well Corning plates (NY, catalog 9018) were coated with 25 ng per well of SARS-CoV-2 wildtype sequence of recombinant RBD (GenBank MN975262.1, amino acids R319-K529) or 50 ng per well of recombinant spike (GenBank MN90894, amino acids M1-Q1208) glycoprotein. These proteins were produced with constructs consisting of a TwinStrepTag, an HRV3C cleavage site, and an 8XHisTag C-terminal modification from Aaron G. Schmidt from the Ragone Institute, and Barney S. Graham from the NIH Vaccine Research Center, respectively. Overnight incubation at 4°C was followed by 0.05% Tween 20 in PBS-wash of plates, with subsequent 1% bovine serum albumin (BSA) blocking for 1 h at RT. Serum samples were initially diluted 1:100 then 4-fold serially diluted to a dilution factor of 1.05E8, followed by incubation in the pre-coated plate for 2 h at RT. Following 3 washes a 1 h RT incubation with horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG, IgG2a, or IgG1 (Southern Biotech respective cat. 1036-05, 1081-05, 1071-05) was performed. Following 5 × 0.05% Tween 20 in PBS washes, RT tetramethylbenzidine (TMB, BD OptEIA substrate solution from BD Biosciences) was added for 5 min, then stopped with sulfuric acid, 2N H₂SO₄. Optical density (OD) was determined at 450 nm in a SpectraMax iD3 microplate reader (Molecular Devices). Assignment of antibody titer was calculated from the final dilution where TMB was over 3x background. Any value below 3x background was assigned half the initial serum dilution of 100.

Murine surrogate virus neutralization titer (sVNT) evaluation

Murine sera were evaluated using a previously as in.¹³⁰ Specifically, flat-bottomed high-binding 96-well Corning plates (NY, catalog 9018) were incubated with 100 ng recombinant human angiotensin-converting enzyme 2 (hACE2, Sigma-Aldrich) in PBS, per well, overnight at 4°C. Following 3 x washes with 0.05% Tween 20 in PBS, plates were blocked for 1 h RT with 1% BSA. Sera were initially diluted 1:160, then incubated with 3 ng of RBD conjugated with an Fc fragment of IgG (RBD-Fc) for 1 h at RT. Sample mixtures were transferred to the hACE2 coated plates alongside the positive control (PC) of non-serum, RBD alone and the negative control (NC) of 1% BSA in PBS alone. After a 1 h RT incubation, 3 x washes with 0.05% Tween 20 in PBS were incubated with anti-human IgG Fc with HRP-conjugation (Southern Biotech). After an additional 5 x washes plates were TMB-developed, H₂SO₄ stopped, and read at 450 nm, as in the Ab evaluation. Percent inhibition was evaluated by calculating (1 – (Sample OD - NC OD)/(PC OD – NC OD)) x 100.

Live SARS-CoV-2 virus neutralization test

Murine samples were evaluated as in.^123,124 Specifically, processing and scoring of samples were performed randomized and blinded to animal treatment. Sera were heat-inactivated at 56°C for 30 min to deactivate complement. Once equilibrated to RT, samples were processed in duplicate to evaluate neutralization titer. Samples were initially diluted 1:20, followed by a 1:2 serial dilution resulting in a 12-dilution series with each well containing 60 μL. Dilutions employed Dulbecco’s Modified Eagle Medium (DMEM, Quality Biological) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (HI-FBS, Gibco), 1% penicillin/streptomycin (v/v, Gemini Bio-products) and 1% L-glutamine (v/v, 2mM final concentration, Gibco). Dilution plates were transported to biosafety level (BSL)-3 where 60 μL of diluted SARS-CoV-2 inoculum (WA-1 strain, courtesy of Dr. Natalie Thornburg, and the CDC) was added to each well with serum, resulting in a multiplicity of infection of 0.01, corresponding to 100 pfu/well. Each plate had a non-treated virus-only control and a mock-infection well to establish cytopathic effects. After 1 h incubation at 37°C with 5% CO₂, 100 μL of sample-virus complexes were transferred to a 96-well plate with confluent (∼1e4) Vero Transmembrane serine protease 2 (TMPRSS2) cells. Incubation of cells with virus permitted evaluation of cytopathic effect (CPE) after 72 h where the first dilution displaying CPE was set as the minimum sample dilution needed to neutralize >99% of the SARS-CoV-2 tested.

Murine splenocyte evaluation

Assessment of cell mediated immunity in adult and older mice made use of CO₂-euthanizing mice between days 39 and 41 post-prime immunization with prompt aseptic collection of spleens at 4°C in 1 mL of RPMI 1640 (Gibco 11875-119) with 10% HI-FBS (HyClone, GE Healthcare) that was 0.22 μm-filtered. Mouse euthanasia was batched to reduce the amount of time (<10 min) that the spleen was within the mouse without active circulation. Downstream splenocyte processing was batched with no more than 3 mice at a time to reduce the amount of time that cells were without circulatory support and off ice. Aseptic dissection included care to dissect away pancreatic tissue, which otherwise can impact cell viability. Splenocytes were dissociated by gently pressing the spleen through a 70 μm cell strainer (Falcon cat. 352350) using the plastic portion of a 3 mL syringe’s plunger, aseptically removed from its wrapper. After twice rinsing the strainer and plunger with 1 mL cold RPMI (4°C), an additional 16 mL rinse of the strainer alone was performed. Following centrifugation (315g for 10 min) supernatant was decanted so that ≤200 μL of liquid remained, cells were resuspended in residual volume, and red blood cells (RBCs) lysed with 1 mL of RT Ammonium-Chloride-Potassium (ACK) lysis buffer (Gibco, Cat A10492-01, Waltham, MA) for exactly 2 min at RT. Osmotic lysis was neutralized immediately and cells washed with 25 mL cold RPMI, passed through a new 70 μm cell strainer, centrifuged, resuspended in RPMI +10% HI-FBS, and cells were counted by dual Acridine Orange/Propidium Iodide (AOPI) staining (Nexcelom Cellometer K2, CS2-0106). To restore basal activity levels, cells were plated at 2 x 10⁶ total cells/well in 200 μL in a 96 well U-bottom plate, then rested overnight (37°C, 5% CO₂) in T cell media consisting of RPMI 1640 (Gibco, Waltham, MA) supplemented with 10% HI-FBS (HyClone, Cytiva), 100 U/mL Penicillin and 100 mg/mL Streptomycin (Gibco, Waltham, MA), 55 mM 2-mercaptoethanol (Gibco, Waltham, MA), 60 mM non-essential Amino Acids (Gibco, Waltham, MA), 11 mM HEPES (Gibco, Waltham, MA), and 800 mM L-Glutamine (Gibco, Waltham, MA).

Flow cytometry of murine splenocytes

Following overnight rest, processed splenocytes were stimulated with SARS-CoV-2 wild type spike peptide pools (PepTivator, #130-126-700, Miltenyi Biotec) at 1 μg/mL in the presence of anti-mouse CD28/49days (1 μg/mL, BD) and brefeldin A (5 μg/mL, BioLegend). After 6h of stimulation, cells were washed twice with PBS and blocked with Mouse Fc Block (BD Biosciences) according to the manufacturer’s instructions. After blocking, cells were washed once with PBS and stained with Aqua Live/Dead stain (Life Technologies, Carlsbad, CA) for 15 min at RT. Following two additional PBS washes, cells were resuspended in 100 μL of FACS buffer (PBS supplemented with 0.2% BSA (Sigma-Aldrich)) containing mouse specific cell surface markers for flow cytometry. Markers included anti-mouse CD44 PerCP-Cy5.5, CD3 BV785, CD4 APC/Fire750 and CD8 BUV395. Clone and manufacturer in the customized nine color, 10 marker flow cytometry panel are documented in the Key Resources table, and as in.¹⁸⁴ Cells were incubated with surface markers for 30 min at 4°C. Cells were PBS-washed and fixed/permeabilized with a Cytofix/Cytoperm kit (BD, #554714), following manufacturer’s recommendations. Cells were washed in 1X perm/wash solution and subjected to intracellular staining (30 min at 4°C) using a cocktail of the following Ab: anti-mouse IFNγ Alexa Fluor 488, TNF PE Cy7, IL-2 PE, IL-4 BV421 and IL-5 BV421 in 1X perm/wash solution. Finally, cells were washed in PBS and fixed in PBS containing 1% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) for 20 min at 4°C. After two final washes in PBS, the cells were resuspended in PBS and stored at 4°C until acquisition. Samples were acquired on a BD LSRFortessa (BD Biosciences; San Jose, CA) configured with blue (488 nm), yellow/green (568 nm), red (640 nm), violet (407 nm), and ultraviolet (355 nm) lasers using standardized good clinical laboratory practice procedures to minimize variability of data generated. Analysis was performed using FlowJo software, v.10.8.1 according to the gating strategy outlined in Figure S6. Positive gates for each cytokine were determined using fluorescence minus one (FMO) controls for IFNγ, TNF, IL-2, and IL-4/5 where all antibodies were used except the targeted one. Population gating was performed blinded to treatment group. PBS group samples were stimulated with mitogen (BioLegend, #423301 at 1:500) for 6 hr as positive controls for Th1 and Th2 signatures. Baseline CD4⁺ T cell activation can be impacted by inflammaging,¹⁸⁵ therefore we performed baseline normalization by evaluating adult and aged mice for their fold induction of CD4⁺ T cell responses from immunized mice over the average age-matched vehicle control.

Quantification and statistical analysis

Proteomic, multiplex, and murine immunogenicity data were analyzed and graphed using R (versions 3.3.2 and 4.1.1). Raw data from LC/MS proteomics were exported into Skyline software (v20.2.1.315)¹⁴⁸ for peak area and retention time refinement. PEA was evaluated by Olink, blinded to treatment, with normalized protein expression (NPX) data sent to BCH. Missing data (e.g., below lower limit of quantification (LLOQ) or were NA (no data)) were replaced with the limit of quantification values as recommended by Olink. 39 of 368 PEA-measured analytes had >60% missing data, necessitating removal before analyses. LC/MS proteomic responses were expressed as fold change of stimulated samples divided by matched controls calculated for baseline adjustment. LC/MS titratable up- and down-regulated responses were evaluated by generalized estimating equations generalized linear model (GEEGLM) analysis,^39,40,41 leveraging multiple stimulation concentrations into 1 measure to determine if stimulation and age interacted and impacted analyte levels. Specifically, the 'geepack’ package in R was used to evaluate log fold change (logFC) of analyte induction in BNT162b2 stimulated over paired RPMI control (logFC) against (‘∼’) the treatment concentration (μg/mL of mRNA in BNT162b2), with an added evaluation (‘+’) of the interacting effect of participant age group (adult or elder) and (‘∗’) categorized analyte function (e.g., T_H1-polarizing or not). This assessment was repeated for each functional role evaluated. PEA assays did not have the same spread of stimulation doses as LC/MS and were evaluated by a moderated T-test between the WBA stimulated with 2 μg/mL of mRNA encapsulated in BNT162b2, versus vehicle (RPMI) control. Multilevel principal component analysis on NPX was performed using the PCA function in mixOmics 6.16.3 package. PEA quantified responses were evaluated with Spearman’s correlation analyses.

Bead-based multiplex samples were evaluated by linear modeling testing for dose dependency of each analyte in non-transformed pg/mL. Fold change (FC) of stimulated sample divided by a matched RPMI control was performed, then log₁₀ transformed. GEEGLM evaluated the interacting effect of age on various cytokine functions with non-interacting effect of stimulation by evaluating fixed effects of ‘BNT162b2 stimulation amount’ and ‘age group’ on LogFC of analyte induction, as above. Exponentiation of the point estimate for each fixed effect allowed for an interpretation of effects as percent increase/decrease, with confidence intervals determined by adding or subtracting 1.96 multiplied by the standard error prior to exponentiation. Data were presented by radar plot per-functional category, filtering for only the analytes associated with each function.

Evaluation of murine samples was based on Shapiro-Wilk test for normality then T-test for normally distributed, or Kruskal-Wallis and Wilcoxon rank-sum tests for non-normally distributed data. Locally estimated scatterplot smoothing (loess) was selected to model best-fit lines between correlations. Spearman correlation test of a monotonic relationship between murine IgG and SVNT was performed due to nonnormal distribution¹⁸⁶ of anti-spike IgG, enabling evaluation of whether an increase in anti-spike IgG would correspond to an increase in SVNT between age groups.

Statistical significance was denoted graphically by ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Sample size was selected based on sample and kit availability (collected in the midst of the SARS-CoV-2 public health emergency), triggering evaluation upon sufficient sample accumulation based on previous experience in modeling age-dependent differences by LC/MS and bead-based multiplex approaches (n ≥ 6 and 10 per age group),^35,187,188 and based on PEA kit availability (n 5 per age group) for exploratory investigation. Specifically, in Figure 1, sample sizes were (A-B) n = 10–14 and (C-D) n = 4–5, with significance by (A-B) GEEGLM analysis with nominal p-value <0.05, and (C-D) paired moderated T-test reporting nominal p-values <0.05. Horizontal dotted lines represent -log10(0.05), and non-significant (NS) findings were visualized as gray circles. In Figure 2, Sample sizes were (A-B) n = 10–14 and (C-F) n = 4–5. Evaluation of significance was by (B) GEEGLM with nominal p-values, (C) paired moderated T-test with adjusted p-values <0.05, and (D) Spearman’s correlation with nominal p-values. (B, C) Horizontal dotted lines represent -log10(0.05). In Figure 3, analyses were performed with (A) Shapiro-Wilk then Wilcoxon rank-sum tests evaluating paired analyses. (B, C) Age group comparisons were evaluated by 1-sided unpaired T-tests on log-transformed fold-change. Concentration-dependent induction was evaluated with linear modeling of log10-transformed analyte levels, with R² and significance annotated by age. Boxplots display median with interquartile range, with n = 12–14. In Figure 4, sample sizes were n = 12–14, with significance evaluated by 1-sided T-test hypothesizing induction. In Figure 5, sample sizes were n = 5–10. Significance was determined by Shapiro-Wilk, Kruskal-Wallis, then (A) one-sided Wilcoxon rank-sum hypothesizing vaccine-associated induction compared to vehicle control, and two-sided Wilcoxon rank-sum test comparing younger adult to aged mice, (B) two-sided Wilcoxon rank-sum test, and (C) two-sided T-test. The graphics for figures and the graphical abstract were created with BioRender.com.

Data management and deposition

Data quality control (QC) was performed for each platform by the endpoint laboratories following assay-specific outputs described above. Data quality assurance (QA) by the Precision Vaccines Program Data Management & Analysis Core (DMAC) entailed verifying application of QC criteria within a centralized cloud-based infrastructure. Deidentified quality assured human data from this manuscript is publicly deposited in the NIH/NIAID-supported repository Immport: SDY2630.

Ethics study approval statement

Experiments were performed under institutional and national guidelines. Volunteer study participants donated blood samples following informed consent, approved by Boston Children’s Hospital (BCH) Institutional Review Board (IRB, X07-05-0223, IRB-P00013867) and Biosafety (IBC-P00001416), and Brigham and Women’s Hospital IRB 2013P002473). Animal procedures were approved by the Institutional Animal Care and Use Committee (00001573), with supervision from the Department of Animal Resources at BCH.

Published: September 26, 2024