The study’s cohorts have been described in detail previously (15). Briefly, as part of an ongoing study by the University of Georgia, Athens (UGA), a total of 690 participants were recruited during five seasons between 2016 and 2020 (UGA1–5). The study procedures, informed consent, and data collection documents were reviewed and approved by the Institutional Review Board of the University of Georgia (IRB #3773). Participants received the standard dose (15 μg/antigen) split-virion (IIV) version of licensed Fluzone by Sanofi Pasteur (Swiftwater, PA, USA). Vaccine formulation was based on WHO recommendations for the Northern Hemisphere influenza seasons beginning in the fall. For the 2019–2020 season (UGA4), the strains were A/Brisbane/2018 (H1N1), A/Kansas/2017 (H3N2), B/Phuket/2013 (Yamagata lineage), and B/Colorado/2017 (Victoria lineage). Participants provided blood samples on the day of vaccination (“Day 0”) prior to vaccine administration and 7/28 days post‐vaccination (“Day 7” and “Day 28”). HAI assays were performed on the Day 0 and Day 28 blood samples against each of the vaccine strains as well as other strains. For this study, a subset of 100 participants enrolled in UGA4, who had been enrolled and vaccinated as part of UGA3, were chosen. These participants were selected so that we could explore the impact of pre-vaccination immunity, as well as provide a wide distribution of participant ages.

Plasma were assayed for cytokines and chemokines involved in stimulating neutrophils, antigen-presenting cells (APCs) (i.e., CXCL1, CXCL2, CXCL5, CXCL8, CCL2, CCL5, CCL20, IL-1a, IL-1b, IL-17, IFN-γ, and IFN I), T cells (GM-CSF, IFN-γ, IL-2, IL-4, IL-5, IL-13, and IL-17, among others), and B cells (IFN-γ, IL-12, IL-2, IL-4, IL-13, IL-10, and TGF-β) via 38-plex Luminex multibead arrays (Millipore Sigma, Burlington, Massachusetts) acquired on a Luminex 200 System (Luminex). Standard curves per analyte were used to define pg/mL concentrations for each sample. Plasma analytes were quantified on blood samples from Day 0 and Day 7, but not Day 28.

To maximize the resolution for detecting changes in targeted cytokines, we utilized the ultrasensitive Single Molecule Array Technology (Simoa) immunoassay to quantify IL-1α, IL-1β, and IL-15 (Quanterix, MA, USA). The limits of detection for IL-1α, IL-1β, and IL-15 were 0.062, 0.011, and 0.003 pg/mL, respectively, by Simoa, compared with 4.8, 1.6, and 3 pg/mL, respectively, by Luminex. Standard curves per analyte were used to define pg/mL concentrations for each sample. Plasma analytes were quantified on blood samples from Day 0 and Day 7, but not Day 28.

Influenza strain-specific HAI titers were analyzed pre-vaccination (“Day 0”) and post-vaccination (“Day 28”). Additionally, log2 fold-change in HAI titers from pre- to post-vaccination was analyzed. Analyses of post-vaccination HAI titers or changes in titers included pre-vaccination HAI titer as a covariate. Analytes with >25% of samples undetectable for an individual assay were excluded from downstream analyses. Using Luminex, 16/38 analytes were included. Using Simoa, three analytes were included. For Day 0 and Day 7 analyses, chemokine/cytokine concentrations in pg/mL were used. Additionally, log2 fold-change in concentrations from Day 0 to Day 7 was analyzed. Univariate analysis of plasma analytes with respect to HAI titers against individual strains was conducted by linear regression, adjusting for age, gender, and body mass index (BMI); see the Results section for the detailed justification of covariates. Models were considered significant if p-value ≤ 0.05.

HAI titers against all four strains of influenza were incorporated in the UGA4 vaccine construct post-vaccination ( Table 1 ). The greatest increase was observed against A/Kansas/2017, which also had the lowest pre-vaccination titer. To understand which demographic factors were associated with HAI titers and serological response to influenza vaccination, we analyzed the association between these variables and strain-specific HAI titers at Day 0 and Day 28. Age was significantly associated with pre-vaccination HAI titers against all four strains and post-vaccination titers against three out of four strains ( Table 2 , Supplementary Figure S1 ). Female participants trended toward lower pre-vaccination HAI titers against influenza B strains, significantly for B/Colorado/2017 ( Table 2 , Supplementary Figure S2 ), but there was no significant association between gender and post-vaccination titers. BMI was not associated with pre- or post-vaccination titers against three out of four strains ( Table 3 , Supplementary Figure S3 ), but there was a positive association between BMI and titers against B/Colorado/2017, significantly so post-vaccination.

Of the 39 plasma chemokines and cytokines measured in this study, 16 (41%; eotaxin, FGF-2, GRO, IFN-γ, IL-10, IL-15, IL-17A, IL-1α, IL-1Rα, IL-8, IP-10, MCP-1, MDC, MIP-1β, sCD40L, and TNFα) were above the limit of detection in >75% of participants across all study timepoints ( Table 3 ). Of these analytes, seven were significantly associated with age. The association was positive for six out of seven, and only IL-1a levels were significantly negatively associated with age. Additionally, five analytes were significantly associated with gender and BMI. When median levels of all analytes were clustered together ( Supplementary Figure S4 ), there was some evidence of older and female participants clustering together with higher levels of multiple analytes. Due to the observed association of both HAI titers and plasma analyte levels with age, gender, and BMI, and previously reported impact of these demographic factors on response to influenza vaccination, all three were included as covariates in our analyses.

To evaluate the global relationship of participant demographics and plasma analyte levels with influenza HAI titers, we performed principal component analysis (PCA) on strain-specific HAI titers from Day 0 and Day 28 ( Figures 1A–C ). The first principal component (PC), which accounted for approximately 47% of the total variance, reflected increasing HAI titers against all strains ( Figure 1A ). Age was significantly negatively correlated with PC1 (Pearson’s r = −0.51, p = 0.6 × 10⁻⁷, Figure 1D ), reflecting decreasing HAI titers with age. Additionally, age was weakly negatively correlated with PC3 (r = −0.22, p = 0.03). The second PC, which accounted for ~17% of the total variance, showed some association with serological response to vaccination, with higher values on this PC for Day 28 titers compared with Day 0 ( Figures 1A, C ). In support of this, fold-change in HAI titer from Day 0 to Day 28 was significantly positively correlated with PC2 for all strains ( Figure 1E ). Gender was also significantly, but weakly, correlated with PC2 (r = 0.20, p = 0.04). The third PC, which accounted for ~14% of the total variance, appeared to differentiate participants based on strain-specific HAI titers ( Figures 1B, C ). The highest values were associated with increased HAI titers against A/Brisbane/2018, while lower values were associated with increased HAI titers against B/Colorado/2017.

We next determined the correlation between plasma analyte levels and each PC ( Figure 1F ) using analyte levels at Day 0 and Day 7 and the fold-change from Day 0 to Day 7 (“Day 7/Day 0”). No plasma analytes measured at Day 0 were significantly correlated with any PC. However, the inflammatory proteins IP-10 (CXCL10, r = −0.25, p = 0.01) and TNFα (r = −0.23, p = 0.02) measured at Day 7 were negatively correlated with PC1. Fold-change in TNFα from Day 0 to Day 7 was also negatively correlated with PC1 ( Figure 1G , r = −0.23, p = 0.03). We also observed a significant positive correlation between fold-change in IL-15 from Day 0 to Day 7 and PC2 ( Figure 1H , r = 0.25, p = 0.01).

We next determined the association between Day 0 pre-vaccination plasma analyte levels and Day 28 post-vaccination HAI titers against individual strains of influenza ( Figure 2 ). Overall, plasma analyte levels were more strongly associated with fold-change in HAI titer from Day 0 to Day 28, compared to Day 28 titers ( Figure 2A ). Eotaxin (CCL11) levels were negatively correlated with Day 28 titers against A/Kansas/2017 and B/Colorado/2017 ( Figure 2B ), as well as fold-change in titers for all four vaccine strains ( Figure 2C ), significantly so for B/Phuket/2013. Similarly, IL-10 levels were negatively correlated with Day 28 titers against A/Kansas/2017 and B/Colorado/2017 ( Figure 2D ), as well as fold-change in titers for all four vaccine strains ( Figure 2E ), significantly so for A/Brisbane/2018 and B/Phuket/2013. IFN-γ levels were also significantly negatively correlated with fold-change in titers against B/Phuket/2013 ( Figures 2F, G ). In contrast, MDC (CCL22) levels were positively correlated with fold-change in titers for all four vaccine strains ( Figures 2H, I ), significantly so for A/Brisbane/2018 and A/Kansas/2017.

We next asked whether the levels of plasma analytes at Day 7 post-vaccination were associated with post-vaccination HAI titers ( Figure 3 ). Strikingly, many of the associations between Day 0 plasma analytes and post-vaccination titers were maintained at Day 7 ( Figure 3A ), most notably the negative association with eotaxin ( Figures 3B, C ) and IL-10 ( Figures 3D, E ) and the positive association with MDC ( Figures 3H, I ). In addition, Day 7 MCP-1 (CCL2) levels were negatively associated with serological response to vaccination ( Figures 3F, G ), both Day 28 titers and fold-change in titers against A/Kansas/2017 and Day 28 titers only against B/Phuket/2013. The levels of sCD40L were negatively associated with fold-change in titers against A/Brisbane/2018, while IL-17A was weakly positively associated with Day 28 titers against A/Kansas/2017.

We next determined whether changes in any plasma analytes from Day 0 to Day 7 post-vaccination were associated with serological response to vaccination ( Figure 4 ). As described above, changes in plasma analyte levels were more strongly associated with fold-change in HAI titer from Day 0 to Day 28, compared to Day 28 titers ( Figure 4A ). Increasing levels of IL-17A, IFN-γ, and IL-15 were all associated with increasing post-vaccination titers against A/Kansas/2017 ( Figures 4B–G ), and they were also individually associated with increasing titers against A/Brisbane/2018 ( Figures 4D, E ), B/Colorado/2017 ( Figures 4F, G ), and B/Phuket/2013 ( Figures 4B, C ), respectively. In contrast to the negative association of Day 0 eotaxin with serological response to vaccination, increasing eotaxin at Day 7 was positively associated with post-vaccination titers against B/Colorado/2017 ( Figures 4H, I ). Additionally, as observed in the Day 7 analysis above, increasing sCD40L was negatively associated with fold-change in titers against A/Brisbane/2018.