Impaired HA-specific T follicular helper cell and antibody responses to influenza vaccination are linked to inflammation in humans.

Antibody production following vaccination can provide protective immunity to subsequent infection from pathogens such as influenza. However, circumstances where antibody formation is impaired after vaccination, such as in older people, require us to better understand the cellular and molecular mechanisms that underpin successful vaccination in order to improve vaccine design for at risk groups. Here, by studying the breadth of anti-hemagglutinin (HA) IgG, serum cytokines, and B and T cell responses by flow cytometry before and after influenza vaccination, we show that formation of circulating T follicular helper cells (cTfh) cells are the best predictor of high titre antibody responses. Using MHC class II tetramers we demonstrate that HA-specific cTfh cells can derived from pre-existing memory CD4+ T cells and have a diverse TCR repertoire. In older people, the differentiation of HA-specific cells into cTfh cells was impaired. This age-dependent defect in cTfh cell formation was not due to a contraction of the TCR repertoire, but rather was linked with an increased inflammatory gene signature in cTfh cells. Together this suggests that strategies that temporarily dampen inflammation at the time of vaccination may be a viable strategy to boost optimal antibody generation upon immunisation of older people.

efficacy varies significantly between different vaccines. The seasonal influenza vaccine needs 23 to be administered each year in order to provide protection against the most prevalent 24 circulating influenza strains, but its efficacy typically ranges from 40-80% even when the 25 vaccine is antigenically matched to circulating viruses. This inefficacy contributes to the 26 millions of severe influenza cases and hundreds of thousands of deaths globally (1), which 27 could be potentially prevented by a more effective vaccine . 28 The reasons that the seasonal influenza vaccine provides protection in some 29 individuals, but not others, have yet to be fully established. Antibodies against the influenza 30 surface glycoprotein haemagglutinin (HA) are capable of limiting infection, and anti-HA 31 antibody titres and inhibitory activity are the most commonly used correlate of protection (2). 32 The human antibody response to influenza vaccination is highly variable, but what causes 33 this inter-individual variation is not well understood. Twin studies estimate that genetics can 34 account for less than 20% of the variation in antibody responses to influenza vaccination, 1 implicating non-heritable factors as key contributing influences (3). Age, sex, chronic viral 2 infections, and non-communicable diseases have all been reported to influence antibody titre 3 following vaccination (4)(5)(6)(7)(8)(9), but how these various factors impact immune responses to 4 vaccination have yet to be fully unravelled. 5 The generation of protective humoral immunity is supported by CD4 + helper T cells 6 (10), which, like neutralising antibodies, are correlates of protection for influenza infection 7 (11). The majority of work on human T cell responses to influenza vaccination has focussed 8 on T helper type 1 cells, largely because of the relative ease of detecting antigen-specific 9 cytokine-secreting cells upon ex vivo peptide restimulation. However, this approach fails to 10 identify T cell types, such as Tfh cells, that do not readily secrete cytokines (12,13), and 11 therefore our understanding of how the human CD4 T cell response is linked with high titre 12 antibody responses upon vaccination is limited. Here, we use MHC class-II tetramers (14,15) 13 to directly assess helper T cell responses to the seasonal influenza vaccine. We find that 14 differentiation of antigen-specific circulating T follicular helper (cTfh) cells, but not the total 15 number of HA-specific CD4 + T cells, is correlated with high titre antibody production upon 16 vaccination. HA-specific cTfh cells are clonally expanded from memory cells present pre-17 vaccination, and share a transcriptional profile with human lymph node Tfh cells. Further, we 18 find that in older people there is a specific defect in the formation of cTfh cells upon 19 vaccination. Interestingly, this was not explained by limited T cell receptor diversity of the 20 responding T cells, as is commonly proposed as the cause of poor T cell responses in older 21 people (16,17). Rather, poor cTfh and antibody responses correlated with an enhanced 22 inflammatory gene signature. Together, this implicates cTfh cells as key mediators of 23 antigen-specific immunity and suggests that vaccine strategies that limit, rather than enhance, 24 the inflammation associated with ageing may be more successful in older individuals. 25 26

Results: 27
Hemagglutinin-specific CD4 + T cells expand and differentiate in response to seasonal 28 influenza vaccination . 29 In order to track influenza HA-specific CD4 T cells directly ex vivo we recruited two cohorts  Table S1, Figure S1-3). IgG levels increased against all measured HA 2 proteins from the vaccine influenza strains at d7 and d42 (Fig 1B). We analysed post-3 vaccination time-points (d7 and d42) compared to baseline to identify which immunological 4 parameters were altered by vaccination (Fig. 1C). From our panel of 32 HA proteins, IgG 5 titres to 31 (96%) were altered at day 42 relative to baseline (d0), with the greatest fold 6 changes were observed for HA strains contained in the TIV. These data indicate that vaccine-7 induced IgG responses were able to partially cross-react across multiple influenza strains 8 (Fig. 1C). The increase in anti-HA antibody titre was coupled with an increase in 9 hemagglutination inhibitory antibodies to A.Cali09, the one influenza A strain contained in 10 the TIV that was shared across the two cohorts (Fig. 1C). Our analysis of 8 cytokines by 11 Luminex identified that 4 cytokines were upregulated (CXCL13, BCMA, APRIL and 12 Osteopontin) and 2 were downregulated (BAFF and TWEAK) at d7 post-vaccination ( Fig.  13 1C, S4). We did not detect alterations in the frequency of any B cell subsets either cohort by 14 flow cytometry, however the frequency of cTfh cells (CD45RA -CXCR5 + PD1 ++ ) was 15 increased on d7, a population that we and others have previously shown share transcriptional 16 and clonal similarity with germinal centre Tfh cells (18)(19)(20)(21). Furthermore, the expression of 17 ICOS on cTfh cells was increased on d7, confirming the cTfh population is an activated 18 effector population that forms in response to vaccination (Fig. 1C, S4). 19 20 To study HA-specific CD4 + T cell responses, we focussed our analysis on the A.Cali09 strain 21 as this was included in both seasons' vaccine formulations, and used MHC Class II tetramers 22 of HLADR*0701 and HLADR*1101 loaded with A.Cali09 HA peptides to identify antigen-23 specific T cells (15). Tetramer binding antigen-experienced CD4 + CD45RA -(Tet + ) T cells 24 showed the largest fold-change increase after vaccination of any parameter measured (Fig.  25 1C). Tet + cells were detected in all individuals before vaccination and expanded a median of 26 5-11-fold between d0 and d7 in both cohorts (Fig. 1D, E). These antigen-specific T cells had 27 upregulated ICOS after immunisation, indicating that they have been activated by vaccination 28 (Fig. 1F, G). In addition, one third of HA-specific T cells upregulated the Tfh markers 29 CXCR5 and PD1 seven days after immunisation (Fig. 1H, I), the majority of which expressed 30 CXCR3 (Fig. S5), consistent with the "Th1" skew in the total CD4 + T cell (15) and cTfh cell 31 response to influenza vaccination (22). Therefore, seasonal influenza vaccination increases 32 anti-HA IgG titres, induces a cTfh response, and promotes the expansion and differentiation 33 of HA-specific CD4 + T cells. 34 1 Circulating HA-specific Tfh cells correlate with vaccine IgG response. 2 The majority of successful vaccines provide protection against re-infection through the 3 production of antibodies. The development of antibody secreting plasma cells requires a 4 concerted effort of multiple cell types of the immune system, which have been investigated in 5 detail in mice, but not well in humans. Therefore we sought to determine which immune 6 parameters were linked with A.Cali09 IgG titre six weeks post-vaccination ( Fig. 2A-D). Pre-7 existing antibody titres have been linked with diminished responses to subsequent 8 vaccination (23-25), but while we observed a slight negative correlation at d0 in support of 9 this, the relationship was not statistically significant for any of the Flu HA Luminex or 10 A.Cali09 HAI titres (Fig. 2B). In contrast, the IgG responses to a range of HA proteins from 11 different influenza strains correlated strongly with A.Cali09 IgG (d42-d0), indicating that 12 those individuals with large vaccine-induced IgG responses also developed cross-reactive 13 antibodies against multiple strains (Fig. 2B). Changes between d0 and d7 in serum BCMA 14 was positively correlated, and BAFF negatively correlated with A.Cali09 IgG (d42-d0) 15 responses, in both cohorts (Fig. 2C, E). The frequency of B cell subsets, including antibody 16 secreting cells at d7 was not associated with day 42 A.Cali09 IgG (d42-d0), whereas cTfh 17 cell frequency correlated with antibodies in both cohorts (Fig. 2D, E). A reproducible 18 positive correlation was observed between A.Cali09 IgG (d42-d0) and HA-specific cTfh cells 19 ( Fig. 2D, E), but not for total Tet + CD4 + T cells (Fig. 2D). This indicates that the 20 differentiation of antigen-specific Tfh cells is more relevant for antibody responses than the 21 overall frequency of antigen-specific helper T cells. 22

clonotypes. 25
Our results demonstrated that HA-specific cTfh cells are correlated with the antibody 26 response to seasonal influenza vaccination, and so we next investigated how vaccination 27 influences the T cell receptor (TCR) repertoire and transcriptional signatures of HA-specific 28 cTfh cells and their precursors. We sort-purified and RNA sequenced tetramer-binding CD4 + 29 T cells (Tet + cells) from d0 and Tet + cTfh cells from d7 (Fig. 3A, S3, S6), and retrieved a 30 total of 1405 and 2085 TCRb clonotypes at d0 and d7, respectively. Expanded clones were 31 observed on d7 (Fig. 3B), resulting in a decrease in the diversity of the TCR repertoire at d7 32 relative to d0 (Fig. 3C), and an increase in the number of co-dominant TCRb clonotypes 33 acquire the immune trajectory of younger individuals. 23 We defined the correlates of influenza vaccine antibody responses in 18-36 year olds to be 24 serum concentrations of BAFF and BCMA, and the frequency of both polyclonal and HA-25 specific cTfh cells. Next we wanted to investigate how the immune response to seasonal 26 influenza vaccination was impacted in older individuals, a group where the influenza vaccine 27 is less efficacious (32, 33). We compared our cohorts of 18-36 year olds with that of 28 individuals over 65 years old in the same vaccination years (Fig. 5A, cohort 1 median 69 29 years old, cohort 2 median 73.43 years old), and observed that older individuals showed 30 lower HA-specific IgG responses compared to [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36] year olds for all the vaccine strains 31 measured (Fig. 5B). To investigate whether ageing broadly influenced immune status pre-32 and post-vaccination, we applied a diffusion pseudotime algorithm to 23 antibody, cytokine 33 or immunophenotyping variables measured at d0 and d7 for both age-groups and cohorts 34 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint (Table S1). The immune profiles formed a single continuous spectrum, with the pseudotime  1   'vaccination trajectory' beginning with d0 samples and ending with d7 samples from  2 18-36 year old samples (Fig. 5C). There was no pre-vaccination difference in trajectory 3 values between the age-groups, indicating that ageing did not impact the baseline status for 4 the immune parameters involved in the vaccination response (Fig. 5D). In contrast, over 65 5 year olds failed to progress along the vaccination trajectory to the same extent as their 18-36 6 year olds counterparts (Fig. 5D), indicating a failure to appropriately respond to vaccination. 7 11 immune parameters correlated with the vaccination trajectory values, including antibody 8 responses to the vaccination strains, cTfh cells, HA-specific T cells and diminished serum 9 BAFF (Fig. 5E). These results indicate that in over 65 year olds there is a failure to fully 10 engage the adaptive immune system in response to vaccination compared to younger 11 individuals. 12

13
In order to identify which immune parameters may explain the age-related difference in 14 vaccination trajectory, we compared the 53 immune parameters between age-groups before 15 and after vaccination. Older individuals had higher titres of IgG to several HA proteins from 16 different influenza A strains at d0, however no difference was seen in IgG or HAI titre levels 17 for the influenza strains in the vaccines administered to our cohorts ( Fig. 5F) (Fig. 5F). For serum cytokines, we observed consistently lower baseline April and 23 TWEAK concentrations, and higher d42 SCF levels in over 65 year olds across both cohorts 24 ( Fig. 5G, S7), but there was no age-dependent difference in BAFF or BCMA. No age-related 25 differences in circulating B cell populations were observed pre-or post-vaccination, and no 26 consistent differences were seen for CD4 + T cell subsets pre-vaccination, including for the 27 frequency of d0 HA-specific CD4 + cells (Fig. 5H, S7). At d7, the frequency of polyclonal 28 cTfh cells and HA-specific Tet + cTfh cells were the only CD4 + T cell subsets consistently 29 reduced in older individuals across both cohorts, with the strongest effect on antigen-specific 30 cTfh cells (Fig. 5H-J). There was no consistent difference in the total d7 Tet + HA-specific T 31 cell population with age for both cohorts (Fig. 5H, S7), which suggests that the poor vaccine 32 antibody responses in older individuals is impacted by impaired cTfh cell differentiation (Fig. 33 5J) rather than size of the vaccine-specific T cell pool. 34 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint 1 TCR repertoire diversity in HA-specific cTfh cells is comparable in younger and older 2

individuals. 3
Poor T cell responses to vaccination have been previously attributed to a contraction of the 4 naïve CD4 + T cell population and restriction of the TCR repertoire with age (16,17), 5 although most human studies have examined the whole repertoire, rather than focusing on 6 vaccine-specific T cells. Our ability to track vaccine specific CD4 + T cells enables direct 7 assessment of the repertoire of responding cells, and therefore we sought to determine 8 whether there were ageing-related differences in the clonal diversity of pre or post 9 vaccination of HA-specific CD4 + T cells (Fig. 6A). The pre-vaccination diversity is 10 important as our data from young donors demonstrates that a large contribution to the d7 11 response originates from memory CD4 + T cells present prior to vaccination. In pre-12 vaccination HA-specific Tet + cells there was a slight reduction in TCR diversity in over 65 13 year olds compared to younger people, although this was only statistically significant in one 14 cohort (Fig. 6B). However, after vaccination no difference was seen in the TCR diversity of 15 the resulting d7 Tet + cTfh population between age groups (Fig. 6B), despite the diminished 16 frequency of this population in older individuals. Likewise, no age-related difference in TCR 17 diversity was observed for the total pool of vaccine-specific Tet + population at day 7 ( Fig.  18 6C). As we had observed in younger donors, TCRb clones were present in both the d0 Tet + 19 and d7 Tet + cTfh populations. There was a trend towards a lower absolute number of recalled 20 clones in older individuals (Fig. 6D), in line with the lower numbers of d7 sequenced cells 21 due to the reduced cTfh population frequency (Fig. S6). However, the proportion of the d7 22 population that consisted of recalled clones was not impacted by age (Fig. 6E), indicative of 23 equivalent contribution from pre-existing memory cells into the resulting cTfh population 24 between younger and older donors. The affinity of TCR for antigen shapes Tfh differentiation 25 and Tfh vs. Th1 cell fate after influenza vaccination (34)(35)(36), and therefore we sought to 26 determine if the age-related decline in Tet + cTfh differentiation could be explained by an age-27 dependent skew in the TCR repertoire away from Tfh differentiation. We examined the 28 TCRb clones that were present in both the d7 Tet + and d7 Tet + cTfh populations, and 29 observed strong correlations between clone frequency between the two populations 30 irrespective of age-groups, suggesting that there was no restriction in the TCR repertoire able 31 to give rise to Tfh cells to the HA peptides studied here (Fig. 6F). Together, our data indicate 32 that TCR diversity is not a limiting factor for the Tfh cell response to seasonal influenza 33 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint vaccination in older individuals, and suggests that the age-associated reduction in cTfh cell 1 frequency is due to other cell-intrinsic or environmental effects on Tfh cell differentiation or 2 survival. 3 4 HA-specific cTfh cells from older individuals fail to induce Tfh transcriptional 5 signatures and display aberrant inflammatory signatures. 6 We investigated the transcriptome of HA-specific cTfh cells from older individuals to 7 determine if we could identify pathways that could explain the poor Tfh cell differentiation in 8 ageing. With supervised principal component analysis using the 684 genes that are 9 differentially expressed in 18-36 year olds d7 cTfh cells compared to d0, we observed that d7 10 cTfh cells from over 65 year olds were clustered between d0 and d7 samples from younger 11 donors for PC1 in both cohorts (Fig. 6G). Indeed, of the 425 genes that were consistently DE 12 between HA-specific cells from d0 and d7 in older donors from both cohorts, only 170 (40%) 13 were part of the younger d7 signature (Data file S5). Importantly, we did not observe any 14 consistent age-related gene expression differences in pre-vaccination d0 cells, indicating that 15 ageing is not associated with transcriptional changes in resting HA-specific memory cells. 16 These data indicate that cTfh cells from older individuals failed to acquire the full gene 17 signature seen in Tfh cells from younger people. 18 19 To further resolve these age-related transcriptional differences, we performed enrichment 20 analysis with the Hallmarks genesets and observed 7 gene sets that were consistently 21 positively enriched in d7 HA-specific Tet + cTfh cells from over 65 year old compared to 18-22 36 year old individuals (Fig. 6H) indicating that the enrichment for numerous inflammatory pathways in Tfh cells from older 32 people occurred as part of the response to vaccination, rather than being a generalisable 33 feature of CD4 + biology in ageing. To further investigate the upregulation of inflammation 34 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. and cytokine signalling in cTfh cells from older donors, we used the leading-edge output 1 from the gene set enrichment results to curate 3 non-overlapping gene lists of 'TNF', 'IL2' 2 and 'Inflammation' (Table S2). We then calculated 3 gene list scores by generating gene z-3 scores for each cohort, and then summing the z-scores for the genes in each of the gene lists 4 for each sample, before and after vaccination. For all three gene signatures, while there was 5 no pre-vaccination difference between the age-groups, d7 cTfh cells from older donors failed 6 to downregulate the expression of these genes to the same extent as cTfh cells from 18-7 36 year olds (Fig. 6I). In summary, our transcriptional analysis indicated that in older 8 individuals, cTfh cells display evidence of inflammatory cytokine signalling that is not 9 typical of Tfh cells in younger people. Our findings suggest that the presence of pro-inflammatory cytokine signalling in cTfh cells 18 induced by vaccination negatively impacts optimal vaccine responses. In order to determine 19 whether the observed negative correlation between the three inflammatory gene signatures 20 and IgG response to seasonal influenza vaccination could be replicated in other cohorts, we 21 analysed PBMC microarray data and A.Cali09 anti-HA antibodies from an independent 22 cohort of 50 adults aged 18-86 year olds from the U.S.A (8). Consistent with our cohorts, we 23 did not detect any age-related difference in gene signatures present prior to vaccination, but 24 the gene signatures induced by vaccination d7 (relative to d0) in PBMCs were negatively 25 correlated with IgG production at d28 (Fig. 7C). This demonstrated that expression of genes 26 associated with inflammation, TNF, and IL-2 signalling in blood is negatively associated 27 with impaired IgG responses to seasonal influenza vaccination, irrespective of age. These 28 data suggest that generation of a high-titre antibody response to vaccination requires the 29 cytokine milieu in secondary lymphoid tissues to be carefully controlled to limit persistent 30 pro-inflammatory cytokine signalling during Tfh differentiation. 31 32 Discussion: 33 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The formation of virus neutralising antibodies after influenza vaccination provides 1 protections against subsequent infection, yet the cellular and molecular pathways that support 2 a high-titre antibody response in humans remain incompletely defined. Here a systems 3 immunology approach was used to determine which immune parameters are associated with 4 antibody formation upon vaccination. We used MHC class II tetramers to track the formation 5 and differentiation of haemagglutinin-specific CD4 + T cells after vaccination, ensuring the 6 specificity of the CD4 + T cell response can be accurately matched to the A.Cali09 antibody 7 response. HA-specific cTfh cell frequency was strongly correlated with anti-A.Cali09 8 antibodies six weeks after vaccination, and there was not a reproducible relationship between 9 total HA-specific CD4 + T cells and antibody titre, highlighting the importance of 10 antigen-specific Tfh cell differentiation to support humoral immunity. TCR repertoire 11 analysis showed that HA-specific cTfh cells formed from pre-existing memory cells in both 12 younger and older adults, but their differentiation was reduced in older people. Interestingly, 13 there was no difference in TCR repertoire diversity in cTfh cells in ageing, indicating that a 14 contraction of the TCR repertoire with ageing is unlikely to be the cause of poor Tfh cell 15 differentiation in older people. The defective cTfh response in ageing was, however, 16 associated with an enhanced pro-inflammatory gene expression signature, suggesting that 17 excess inflammation can limit the response to vaccination. We were able to confirm that 18 these enhanced inflammatory signatures associated with poor antibody titre in an independent 19 cohort of influenza vaccinees. Together, this suggests that formation of antigen-specific Tfh 20 cells is essential for high titre antibody responses, and that excessive T cell-extrinsic 21 inflammatory factors contribute to poor cTfh cell and antibody responses to vaccination. 22 23 In order to understand what type of immune response supports protective antibody 24 production upon vaccination we comprehensively analysed the immune response to seasonal 25 influenza vaccination by measuring IgG responses to 32 HA proteins, 8 cytokines and 26 chemokines, B cells, CD4 + T cells and HA-specific CD4 + T cells. This combined approach 27 enabled us to identify which parameters were correlated with the antibody response after 28 vaccination. In contrast to previous studies (22,37), we observed that the frequencies of 29 circulating B cell populations including plasmablasts did not correlate with the long-term 30 antibody response. This suggests that the circulating plasmablasts observed seven days after 31 vaccination in people are likely biomarkers of the early extrafollicular antibody response that 32 provides a short initial burst of antibodies, rather than of long-lived plasma cells (38). 33 However, the down-regulation of BAFF and up-regulation of one of its receptors, BCMA, in 34 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021.
serum over the first 7 days after vaccination clearly demonstrate that dynamic regulation of B 1 cell responses through BAFF and its receptors are tightly intertwined with the magnitude of 2 the antibody output to vaccination. Consistent with this, the up-regulation of TNFRSF17, the 3 gene for BCMA, has been well reported to correlate with vaccine titres in PBMC 4 transcriptomics (8,(39)(40)(41). Importantly, we show that expansion of Tfh cells, both polyclonal 5 and HA-specific, were consistently correlated with the magnitude of antibody responses 6 across two cohorts. Through sequencing the HA-specific cells we were able to show that 7 cTfh cells are recalled from the resting memory CD4 + T cell compartment, that public TCR 8 clonotypes are readily detectable in antigen-specific cTfh cells, and that antigen-specific cTfh 9 cells share a transcriptional program with lymph node Tfh that includes the downregulation 10 of TNF and IL-2 signalling. As age is one of the key factors that influences antibody 11 responses to vaccines (42)(43)(44), by clearly defining ideal immune responses in younger 12 individuals we were able to gain novel insights into how ageing negatively impacts the 13 immune response resulting in low titre antibody responses following seasonal influenza 14 vaccination. 15 16 Impaired T cell responses to vaccines in ageing has been proposed to be caused by 17 contraction of the TCR repertoire and the accumulation of terminally differentiated effector 18 cells. However, in our study, we observed impaired antigen-specific Tfh differentiation in 19 older people, despite no defect in the overall antigen-specific CD4 + T cell response. Through 20 analysing the clonal relatedness in the TCR repertoire of HA peptide-specific T cells from 21 before and after vaccination, we observed that similar frequencies of d7 cTfh cell clonotypes 22 were recalled by vaccination from resting memory CD4 + T cell compartment in both age 23 groups. This suggests that the ability of memory cells to be recalled by vaccines in ageing is 24 not compromised. Furthermore, we observed no difference in the diversity of the TCR 25 repertoire of the cTfh cells that formed after vaccination between age-groups. Therefore, this 26 suggests that a loss of diversity in the T cell repertoire with age does not explain the impaired 27 Tfh differentiation we observed in our study. Impaired Tfh responses to vaccines have been 28 observed in humans and mice (9,45), however, interestingly the formation of pre-Tfh cells 29 remains intact but there is a failure to generate bona fide GC Tfh cells (45,46). Together with 30 our data, this suggests that instead of intrinsic defects in CD4 + T cells pre-vaccination in 31 older people, the reduced cTfh cell frequency post vaccination may be explained by extrinsic 32 factors, such as inflammation, that lead to a failure to appropriately acquire the GC Tfh cell 33 gene signature. 34 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint 1 Ageing is often associated with a state of low grade inflammation, termed 2 'inflammaging' (47)(48)(49). However, in this study we observed no age-related difference in 3 inflammatory gene signatures in HA-specific CD4 + T cells at day 0, which suggests the cTfh 4 cell response was negatively impacted by inflammation post vaccination, rather than pre-5 existing inflammation having affected resting memory T cell function prior to vaccination. 6 Because older people are more at risk of severe health outcomes after infection, significant 7 effort has been made to alter vaccine formulations to enable them to be effective in this age 8 group. Modifications to vaccines to increase the antibody response in older individuals now 9 include increasing the antigen dose and using more potent adjuvants (50, 51), which enhance 10 the inflammatory response. We have previously demonstrated in older people and aged mice 11 that type 1 interferon and TLR7 signalling is important for conventional dendritic cells type 2 12 to support Tfh differentiation after vaccination (9). These studies, together with the data 13 presented here, prompt the hypothesis that while some types inflammation are 'good' for 14 promoting long-lived antibody responses in ageing, cytokines such as TNF and IL-2 are 'bad ' 15 and negatively impact a Tfh-supported response to vaccination. Consistent with this, 16 inhibition of inflammatory monocyte recruitment into the skin of older people enhanced the 17 local CD4 T cell response to varicella zoster virus antigen challenge (52). Therefore, 18 vaccination strategies that support humoral immunity through limiting 'bad' pro-19 inflammatory signalling may be the key to improving vaccine efficacy in older people. 20

21
Our data demonstrates that IL-2 is one signalling pathway that is normally downregulated 22 during Tfh differentiation. IL-2 is a cytokine produced by T cells early after T cell activation 23 that promotes clonal expansion, and favours a Th1 cell fate at the expense of Tfh cell 24 differentiation (28,31), in line with the negative relationship observed here between cTfh cell 25 differentiation and the response to IL-2 in those cells. Tfh cells are reported to be IL-2 26 producers as a consequence of heightened T cell receptor signalling, and whilst they typically 27 do not respond to IL-2, they provide a source of this cytokine to support non-Tfh cells (29). 28 Transcript levels for IL2 were not increased in cTfh cells from older donors, however IL2RB 29 was increased suggesting that the failure to downregulate an IL-2 gene signature in ageing 30 could be due to increased response to IL-2 in the local environment after vaccination. 31 Alternatively, IL-2 signalling gene signature may reflect a failure to fully acquire the Tfh 32 transcriptional program, which may result from memory T cells receiving inadequate Tfh 33 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint lineage commitment signals from the local environment, such as dendritic cell priming which 1 is known to be impaired in aged mice (9). 2 3 TNF signalling in cTfh cells was also negatively correlated with Tfh differentiation and 4 antibody response upon vaccination. While this cytokine is necessary for the formation of 5 primary B cell follicles and follicular dendritic cell networks that underpin the formation of 6 germinal centers (53-56), it has also been implicated in the loss of germinal centres and 7 disorganisation of secondary lymphoid organ architecture during infections and 8 immunizations in mice (57, 58). The role for TNF in Tfh cell differentiation or survival 9 remains unclear, and it is noteworthy that this pleiotropic cytokine can regulate many aspects 10 of T cell biology, including NF-κB signalling, TCR signalling, and apoptosis pathways (59). 11 In our study, this TNF signalling signature did not include the TNF gene, nor were TNF 12 transcripts altered in cTfh cells with age. While in our study, serum concentrations for both 13 TNF and IL-2 were below the limit of detection in most samples, serum TNF levels and TNF 14 production from memory B cells have been reported to increase with age (60, 61), suggesting 15 our TNF signature in cTfh cells may result from paracrine sources within the lymph node. show that this is due to the second vaccine dose enhancing Tfh cells and the germinal center 3 response to a greater extent in aged mice than in younger adult animals (72). COVID19 4 mRNA vaccines induce high titre antibodies after two doses in all age groups (73-75), and 5 generate superior germinal center responses than protein subunit vaccines in young mice (76) 6 and humans (pre-printed in (77)). Together with the data presented here, this demonstrates 7 that effective vaccines are ones that promote the germinal centre response. Therefore 8 vaccination strategies that induce the optimal inflammatory environment to support Tfh 9 differentiation is key to generating enduring antibody mediated immunity. 10 11

Human Cohorts 13
Peripheral blood was collected from healthy UK adults recruited through the NIHR 14 Trust, as part of the routine operative procedure, as described previously (20,78).

Flow cytometry of B cells and FACS sorting of T lymphocytes 31
Cryopreserved mononuclear cells were thawed and rested for 1 hour at 37℃. Fc receptors 32 were blocked using anti-human CD32 antibody (clone 6C4, eBioscience). 4 million PBMCs 33 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint were stained with a panel of antibodies to measure B cells (Table S2), acquired on BD LSR 1 Fortessa5 cytometer, and gated according to the strategy outlined in Fig. S1. For T cell 2 staining and sorting, between 15-40 million PBMCs were first treated with 50nM dasatinib 3 (Sigma) for 10minutes at 37℃, and then stained with PE-conjugated tetramers for 2 hours at 4 room temperature with methods and reagents that have been previously reported (15)  100-bp single-end reads per sample. mRNA from lymph node CD4 + CD45RA − T cell 28 populations was isolated from 1000 cells sorted into lysis buffer from 6 individuals as 29 previously described (20). Sequencing reads were aligned to the reference human genome 30 GRCh38 using HISAT2 (81), and quantitated using Rsubread package (82). Samples were 31 excluded on the basis of poor cDNA quality prior to sequencing, or where drop-out genes 32 with zero counts represented more than 80% of total reads. Genes were filtered based on 33 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint having more than 10 reads in 20% of samples, yielding 19113 genes. Unwanted variation due 1 to batch effects and sex was determined using the RUVseq package (83) for each cohort 2 seperately using the model.matrix [~data$sex +data$group], where the 'group' factor 3 represents the combination of time-point, cell type and age-group. The RUVseq output was 4 incorporated into DESeq2 model matrix design using the following model [design = ~ W_1 + 5 W_2 + W_3 + sex + group] where W represents the three RUVseq variance factors identified 6 for each cohort. DESeq2 fold changes were adjusted using the lfcShrink normal method, and 7 variance stabilised normalisation was applied to the counts to give an expression value per 8 gene (DESeq2 package (84, 85)). Significantly differentially expressed genes had adjusted p- Raw. CEL files were downloaded from GEO with the corresponding annotation data. CEL 29 files were read into R via Affy package(89) and were normalised using VSN (84). The 30 dataset was then filtered to select only individuals with paired day 0 and day 7 samples. Gene 31 z-scores were then calculated and summed across gene signatures for each sample as 32 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint described above, with the summed gene signature scores for d0 subtracted from d7 scores for 1 each sample for correlation to antibody titres, which were made available upon request (8). 2

T cell receptor sequencing and clonotype analysis: 4
TCRβ clonotypes were called from adaptor-trimmed RNA sequencing fastq files using 5 MIXCR (version 2.1.9;(90)) run in RNA-Seqmode with rescuing of partial alignments and 6 set to collate TCRA or TCRB clonotypes at the amino acid level and requiring more than five 7 reads to identify a clonotype. Clonotype diversity was determined using vdjtools (version

Vaccination trajectory and pseudotime analysis 15
The trajectory analysis was assembled as previously described (7, 92). Briefly, the 16 frequencies for immune parameters were first normalized for abundance by subtracting the  (Table S1) 19 Principal components analysis was then performed, in which the samples from different time-20 points showed separated by PC1. Therefore, the immune parameters that had a correlation to 21 PC1 greater than 0.4 were included (n = 23 cell types). The diffusion maps algorithm was 22 then applied to the scaled frequencies using the destiny R package(93, 94), the resulting 23 diffusion pseudotime values (described as vaccination trajectory) were scaled to a range of 0 24 to 1 and compared between time-points and age-groups. Correlations between vaccination 25 trajectory and cell type frequencies were analysed by Spearman's correlation. 26 27

Statistics 28
All statistical tests for cell type frequencies assumed nonparametric data. Fold change for 29 immune parameters was calculated by dividing d7 or d42 frequency by d0, and post 30 vaccination increases relative to baseline were determined by subtracting d0 from d7 or d42 31 data. Two-group comparisons were made using either two-tailed Mann-Whitney tests or 32 Wilcoxon tests for paired data from the same individual at different time points. Multiple 33 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. group comparisons p-values were calculated using Dunn's post hoc test. Correlation analyses 1 used Spearman correlation, with the exception of log normalised TCRB clone frequency 2 analysis which used Pearson correlation. As necessary, FDR (Benjamini-Hochberg) 3 adjustment was used to adjust for multiple testing on two-group comparisons. Heatmaps of 4 manual gated cell subsets that were altered by vaccination or age group were generated using 5 Pheatmap package (95).       . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint All p-values were calculated using a Mann-Whitney U test.
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint a) Overview of age-groups and sample sizes for each cohort. b) IgG responses to HA proteins from vaccine influenza strains measured by Luminex for each age-group. c) Diffusion map dimensionality reduction of 122 samples from both cohorts combined using scaled values for 23 immune parameters and the diffusion-pseudotime algorithm (d0 18-36yo n= 27; d0 65+yo n= 32; d7 18-36yo n= 30; d7 65+yo n=33). Each dot represents a sample, shape represents time point (d0 = squares, d7 = circles), and colour either the pseudotime 'vaccination trajectory' output value or age-group category. Diffusion components (DC) 1 and 2 shown. d) Vaccination trajectory values for sample in each age category from both cohorts combined, with p-values calculated using Dunn's post hoc test. e) Spearman correlation coefficients for the 11 parameters that significantly correlated with the vaccination trajectory variable (padj<0.05). f-h) Heatmap of FDR-adjusted p-values from Mann-Whitney U test comparing immune parameters between age-groups for Cohort 1, Cohort 2 and both cohorts combined (All), at d0, d7 and d42, and at d7 and d42 after subtracting each individuals d0 baseline value (d7-d0, d42-d0) for f) Flu HA Luminex and HAI, g) Cytokine Luminex, h) B cells, CD4 + T cells and HA-specific CD4 + T cells. Colour corresponds to p-value and the direction of change. i) The percentage of CXCR5 + PD1 ++ cTfh cells and j) TET + cTfh cells for each age-group and each cohort, with p-values calculated by Mann-Whitney U test (Cohort 1 18-36yo n= 17, 65+yo n=17; Cohort 2 18-36yo n=20, 65+yo n=21).
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint a) Correlation between cTfh Tet + cells and Inflammation, TNF or IL2 gene signatures scores in d7 Tet + cTfh cells. b) Correlation between A.Cali09 IgG titre (d42-d0) and Inflammation, TNF or IL2 gene signatures scores in d7 Tet + cTfh cells. c) Correlation between A.Cali09 IgG titre (d28-d0) and Inflammation, TNF or IL2 gene signatures scores determined from microarray data of peripheral blood mononuclear cells on d0 or d7 after seasonal influenza vaccination from publicly available datasets (n= 50 total; 26-41yo n=18 , 66-86yo = 32) (8).
Correlation coefficients and p-values calculated using Spearman's Correlation coefficient.
Solid line represents linear regression fit. Color corresponds to age-group (Green = younger people; aqua = older people).
. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 13, 2021. ; https://doi.org/10.1101/2021.04.07.21255038 doi: medRxiv preprint