Ageing promotes early T follicular helper cell differentiation by modulating expression of RBPJ

Ageing profoundly changes our immune system and is thought to be a driving factor in the morbidity and mortality associated with infectious disease in older people. We have previously shown that the impaired immunity to vaccination that occurs in aged individuals is partly attributed to the effect of age on T follicular helper (Tfh) cell formation. In this study, we examined how age intrinsically affects Tfh cell formation in both mice and humans. We show increased formation of Tfh precursors (pre‐Tfh) but no associated increase in germinal centre (GC)‐Tfh cells in aged mice, suggesting age‐driven promotion of only early Tfh cell differentiation. Mechanistically, we show that ageing alters TCR signalling which drives expression of the Notch‐associated transcription factor, RBPJ. Genetic or chemical modulation of RBPJ or Notch rescues this age‐associated early Tfh cell differentiation, and increased intrinsic Notch activity recapitulates this phenomenon in younger mice. Our data offer mechanistic insight into the age‐induced changes in T‐cell activation that affects the differentiation and ultimately the function of effector T cells.

The formation, maintenance and function of the GC depend on T follicular helper (Tfh) cells, a distinct CD4 + helper T-cell subset that provides essential help to B cells. Though critical for generating appropriate B-cell responses to infection and vaccination, dysregulated Tfh cell responses can occur and are associated with a variety of autoimmune and inflammatory disorders, many of which develop with increasing age. This suggests that age can impact Tfh cell development and/or function (Blanco et al., 2016;Deng et al., 2019;Edner et al., 2020;Qin et al., 2018;Vinuesa et al., 2016). In mice, age-dependent changes in CD4 + T cells have been shown to contribute to the defective GC and antibody responses (Eaton et al., 2004;Lefebvre et al., 2016;Maue & Haynes, 2009;Yang et al., 1996).
However, the aged microenvironment also affects GC responses and can inhibit generation of Tfh cells (Lefebvre et al., 2012(Lefebvre et al., , 2016Stebegg et al., 2020). In previous work, we showed that older people have reduced numbers of Tfh cells in peripheral blood following influenza vaccination. This was associated with a 5-fold reduction in production of vaccine-specific antibodies (Stebegg et al., 2020). This was also seen following immunisation of older mice where reductions in GC size, titres of antigen-specific antibodies and numbers of GC-Tfh cells were evident (Stebegg et al., 2020). This suggested an altered GC-Tfh response in older individuals and that similar age-associated defects in the Tfh cell response are common between mice and humans. It remains unclear how age affects Tfh cell formation and little is known about the molecular mechanism/s that underpin the altered Tfh cell biology seen in older individuals.
Generation of GC-Tfh cells occurs by a multi-step process: T-cell priming, differentiation to Tfh precursor (pre-Tfh) cells and then further differentiation to GC-Tfh cells (Crotty, 2019;Webb & Linterman, 2017). Typically, priming occurs by a dendritic cell delivering three key signals to a naïve T cell: ligation of the TCR by peptide-MHCII, co-stimulation by CD80/86 and cytokine stimulation. These signals converge to direct differentiation of the T cell. In human T cells, early Tfh cell polarisation can be replicated in vitro when naïve CD4 + T cells are stimulated via CD3/CD28 in the presence of TGFβ or Activin A with IL-12 (Locci et al., 2016;Ma et al., 2009;Schmitt et al., 2009Schmitt et al., , 2014. Cells generated in this system have increased expression of CXCR5, PD-1, ICOS, MAF and BCL6, a phenotype akin to the pre-Tfh cells formed in vivo when later (B-cell-dependent) stages of GC-Tfh cell generation are blocked (Choi et al., 2011;Kerfoot et al., 2011;Kitano et al., 2011).
In this study, we show that ageing results in an accumulation of pre-Tfh cells following immunisation of aged mice. We use the in vitro system described above to show that this also occurs in human T cells from older donors. We dissect the mechanisms that drive this early differentiation and established that ageing alters early signalling events resulting in increased expression of RBPJ, a transcription factor essential for the Notch pathway. We show that RBPJ and Notch work together to promote pre-Tfh cell differentiation. However, whilst Notch activity is required, it is the age-driven increase in RBPJ expression that drives early Tfh cell differentiation.

| Increased pre-Tfh cells following immunisation of older mice
In order to look at the effect of age on early Tfh cell differentiation in vivo, we first sought to determine the characteristics of pre-Tfh cells to enable their identification following immunisation. Though often described in the literature, identification/characterisation of pre-Tfh cells is less clear (Krishnaswamy et al., 2018;Ma et al., 2012;Read et al., 2016;Song & Craft, 2019). The expression of the Tfh cell markers CXCR5 and PD-1 varies depending upon the location of the T cell; Tfh cells within the GC express the highest levels of CXCR5 and PD-1 whilst the pre-Tfh cells (found outside the follicle) express intermediate levels (Shulman et al., 2013). Bcl6 is upregulated early in all activated T cells, and its continued expression is required for the maintenance of both the pre-Tfh and GC-Tfh cells, whilst SAP is essential for maintaining the cognate T-cell-B-cell interactions required only for full GC-Tfh cell differentiation (Choi et al., 2011;Kerfoot et al., 2011;Kitano et al., 2011). Consistent with this, in mice with Bcl6-deficient T cells, neither CXCR5 int PD-1 low/int pre-Tfh nor CXCR5 high PD-1 high GC-Tfh cells form after influenza A infection (Figure 1a-c). In contrast, CXCR5 int PD-1 low/int pre-Tfh cells are present in SAP-deficient F I G U R E 1 Age promotes pre-Tfh cell differentiation in mice. Representative flow plots showing CXCR5 and PD-1 expression on CD4 + cells isolated from draining lymph nodes of Bcl6 fl/fl Cd4 +/+ and Bcl6 fl/fl Cd4 cre/+ mice on day 14 postinfluenza A virus infection (a) and graphs showing percentage of pre-Tfh (b; CXCR5 + PD-1 low/int ) and GC-Tfh cells (c; CXCR5 hi PD-1 + ). Representative flow plots showing CXCR5 and PD-1 expression on CD4 + cells isolated from draining lymph nodes of Wt and Sh2d1a −/− mice on day 14 postinfluenza A virus infection (d) and graphs showing percentage of pre-Tfh (e) and GC-Tfh (f) cells. Representative flow plots showing CXCR5 and PD-1 expression on CD4 + T cells from inguinal lymph nodes in unimmunised 8-to 12-week-old and >85-week-old mice (g). Flow plots showing the pre-Tfh and GC-Tfh cells present in draining lymph nodes of 8-to 12-week-old and >85-week-old C57Bl/6 mice on day 6 postinfluenza A virus infection (h). Percentages of pre-Tfh cells (i) and GC-Tfh cells (j), numbers of pre-Tfh (k), GC-Tfh cells (l) and ratio of pre-Tfh to GC-Tfh cell number (m) in draining lymph nodes of 8-to 12-week-old and >85-week-old mice on day 6 postinfection. Percentages of naïve/effector (n), pre-Tfh (o) and GC-Tfh (P) CD4 + T cells expressing Bcl6 of draining lymph nodes from 8-to 12-week-old and >85-week-old mice on day 6 postinfection. Each symbol is representative of an independent biological replicate, and the height of the bar represents the mean. Statistics were calculated using Mann-Whitney U test *p < 0.05, **p < 0.005, ***p < 0.0005 mice after infection, but CXCR5 high PD-1 high GC-Tfh cells do not form (Figure 1d-f). We then used these criteria to identify pre-Tfh cells and GC-Tfh cells in younger adult (8-12 weeks) and aged (>85 weeks) mice. In unimmunised adult and aged mice, CD4 + T cells do not express CXCR5 or PD-1 (Figure 1g)  (Figure 1p). Using a different immunisation strategy that enabled identification of antigen-specific T cells responding directly to challenge, we were able to confirm that ageing promotes antigen-specific pre-Tfh cell differentiation.
Mice were immunised with 1W1K-NP in alum and the differentiation of total and 1W1K-IAb binding T cells assessed. GCs were established by day 10 postimmunisation but few were seen on day 7 when T cells were analysed, ensuring that the T cells have not yet entered the GC ( Figure S1A). In aged mice, the percentage of both the total (Figure S1B, C) and antigen-specific pre-Tfh cells were increased ( Figure S1E, F), but there was no corresponding increase in the percentage of antigen-specific GC-Tfh cells ( Figure S1E, G). This demonstrates that pre-Tfh cells are newly generated in response to the immunisation, and their differentiation is increased in aged mice. However, there is no such increase in the percentage of GC-Tfh cells, and combining this with the lymphopenia seen in aged mice, there will be very few GC-Tfh in aged mice that can provide help to GC B cells. Since many of the cellular players involved in pre-Tfh cell differentiation are susceptible to age-driven changes, we turned to an in vitro system where pre-Tfh cell differentiation can be assessed in the complete absence of other accessory cells.

| Naïve CD4 + T cells from older donors initiate cytokine-independent pre-Tfh cell differentiation
To determine whether age also affected human pre-Tfh cell formation, we used an established model of pre-Tfh cell differentiation (Locci et al., 2016;Ma et al., 2009;Schmitt et al., 2009Schmitt et al., , 2014. This in vitro system enabled us to look at the effects of ageing on human T cells in the absence of other cell types, thereby allowing the T-cellintrinsic effects of ageing to be directly assessed. Human naïve T cells were flow-sorted from peripheral blood using CD45RA and CD27 as markers of naivety. Inclusion of CD27 enabled exclusion of T effector memory (TEMRA) cells which increase with age and re-express CD45RA ( Figure S2A). Consistent with their naivety, we found ageassociated decreases in the frequency of CD45RA + CD27 + CD4 + T cells ( Figure S2B). There was no difference in the expression of CD28 ( Figure S2C), which has been previously noted on CD45RA + CD4 + T cells, possibly reflecting decreased expression by CD45RA + CD27 − TEMRA cells (Weng et al., 2009). Naïve CD4 + T cells were purified from younger (17-39 years) and older (60-76 years) donors and stimulated via CD3/CD28 in the presence or absence of the Tfhpolarising cytokines, IL-12 and TGFβ. The expression of the Tfh cell surface receptors CXCR5 and PD-1 was assessed daily and was evident on day 3 of activation ( Figure S2D). Surprisingly, naïve CD4 + T cells from older donors generated CXCR5 + PD-1 + pre-Tfh cells in the absence of Tfh-polarising cytokines (Figure 2a, b, Figure S2D Figure S2D). When ICOS was used as a marker for pre-Tfh cells instead of PD-1, a similar effect of age on early Tfh cell differentiation was seen ( Figure S2E). The pre-Tfh status of the cells generated (either cytokine-or age-induced) was confirmed by the lack of downregulation of CCR7 and lack of induction of CD57 ( Figure S2F), features of fully differentiated human Tfh cells found within GCs (Wong et al., 2015). The activation of cells from older donors in the presence of neutralising antibodies to Activin A, IL-12 and TGFβ had no effect on pre-Tfh cell differentiation (Figure 2d), confirming that this age-associated increase in pre-Tfh cell differentiation occurs independently of the "classic" Tfh-polarising cytokines. Since the T-cell cytokines, IL-21 and IL-2, can affect Tfh cell differentiation, we F I G U R E 2 Age promotes pre-Tfh cell differentiation in humans. Flow plots showing the frequency of CXCR5 + PD-1 + cells amongst CD4 + following 4 days in vitro activation of naïve CD4 + T cells taken from younger (17-39 yrs) and older (>60 yrs) donors in the presence/absence IL-12 and TGFβ (a). Percentage of CXCR5 + PD-1 + following activation of naïve CD4 + T cells from younger and older donors in the absence (b) or presence (c) of IL-12 and TGFβ. Percentage of CXCR5 + PD1 + cells following activation of naïve CD4 + T cells from older donors with or without neutralising antibodies to IL-12, Activin A and TGFβ (d). Bar graphs showing percentage of cells expressing pSTAT3 (e, f), and pSTAT5 (g, h) on day 3 of culture in presence/absence of IL-12 and TGFβ in donors of the indicated ages. RT-PCR determination MAF (i), BCL6 (j), FOXP3 (k) and CXCL13 (l) following 4 days activation of naïve CD4 + T cells from younger and older donors in the absence of polarising cytokines. Flow plots (m-o) and bar graphs (n-o) showing percentage of IFNγ and IL-21 expressing cells following PMA and ionomycin mediated restimulation after 4 days in vitro culture of naïve CD4 + T cells from younger and older donors in the absence of IL-12 and TGFβ (mo). Bar graph showing levels of IgG produced by B cells following co-culture with CD4+ T cells from day four cultures from young and older donors, in the presence/absence of Tfh-polarising cytokines (p). Each symbol is representative of an independent biological replicate, and the height of the bar represents the mean. Statistics were calculated using Mann-Whitney U test *p < 0.05, **p < 0.005, ***p < 0.0005 also examined expression of pSTAT3 and pSTAT5 during culture. As

| Increased expression of CXCR5 following activation of naïve CD4 + cells from older donors
The

| Age-induced CXCR5 expression is driven by the transcription factor RBPJ
To further investigate the activation-induced molecular changes in aged T cells, we also analysed gene expression by RNA sequencing.
Naïve CD4 + T cells from younger and older donors were stimulated via CD3/CD28 in the presence/absence of the polarising cytokines and gene expression determined 3 days later. In T cells from younger donors, the expression of 239 genes was changed by the addition of Tfh-polarising cytokines, whilst 121 transcripts were differentially expressed between younger and older donors when T cells were activated in the absence of these cytokines (Figure 5a). The 34 genes common between these two lists were potential candidates for supporting cytokine-independent pre-Tfh cell formation seen in older donors ( Figure 5a). The transcription factor RBPJ was an exemplar of this gene expression pattern, associated with cytokine-dependent pre-Tfh cell formation in younger, and with cytokine-independent Tfh formation in older donors (Figure 5b). Since CXCR5 expression was measured by flow on cultures used to generate RNA-Seq libraries, we were able to determine whether the expression of CXCR5 protein correlated with RBPJ expression. We found a strong correlation for both CXCR5 mRNA and protein with RBPJ (Figure 5c, d).
Furthermore, the age-induced increase in RBPJ expression occurred by day 2, a timepoint where there is little CXCR5 expression is seen, establishing that RBPJ expression precedes that of CXCR5 ( Figure   S4A). To determine whether RBPJ is required for CXCR5 expression, and older (red circles) donors. Expression of CXCR5 (f, g) and PD-1 (h, i) relative to CTV dilution 72 hr after activation of naïve CD4 + T cells from younger (clear circles) and older (red circles) donors. Each symbol is representative of individual values from independent donors, the height of the bar represents the mean, and statistics were calculated using Mann-Whitney U or Kruskal-Wallis test *p < 0.05, **p < 0.005 an IRF4-dependent manner (Figure 5i). This demonstrates that Rbpj expression can be regulated by IRF4 in mouse T cells, and suggests that increased IRF4 may contribute to increased RBPJ expression.

| Increased RBPJ expression enables Notch-mediated CXCR5 expression in cells from older donors
RBPJ is a transcription factor that works in collaboration with Notch (Amsen et al., 2009;Jarriault et al., 1995;Kopan & Ilagan, 2009 (Bray, 2016). RBPJ is an essential component of the Notch pathway, and there is little evidence to suggest Notch-independent functions of RBPJ in mammalian cells (Bray, 2016). Previous studies have shown that deletion of Notch in T cells prevents Tfh cell differentiation in vivo (Auderset et al., 2013;Dell'Aringa & Reinhardt, 2018). It is widely assumed that dur- ing T-cell priming Notch is activated by ligands expressed on APCs (Amsen et al., 2009;Dell'Aringa & Reinhardt, 2018). Since there are no accessory cells in the in vitro system, we first determined whether Notch was activated during culture. We found that T-cell activation generated NICD and induced expression of Notch1 and Notch2 ( Figure 6a-d). However, there were no age-related changes in the expression of Notch or Notch cleavage (Figure 6a-d), suggesting that it was the age-induced increase in RBPJ expression that was driving pre-Tfh cell differentiation. To determine whether Notch was required for RBPJ-driven CXCR5 expression, we used a low dose of the gamma-secretase inhibitor, Ly411-575, to partially inhibit Notch activation (Figure 6e). This dose was chosen as higher doses prevented cell proliferation. As a control, we used the gamma-secretase inhibitor, JLK6, which lacks the ability to inhibit Notch cleavage (Petit et al., 2003). We found that whilst Ly411-575 could inhibit age-induced CXCR5 expression and pre-Tfh cell differentiation, JLK6 had no effect (Figure 6f-i). We went on to examine the expression of the Tfh-associated transcription factor BCL6, MAF, the Th1associated transcription factor, TBX21 and RBPJ. We found that only Tfh-associated transcription factors were reduced by inhibition of Notch cleavage (Figure 6j-m). This suggests that inhibition of Notch is required for RBPJ-driven pre-Tfh cell differentiation even though its expression and activation are unaffected by age.
Seven days after immunisation, mice treated with Ly411-575 had a reduced frequency of pre-Tfh but not GC-Tfh cells ( Figure S4B-D).
To ascertain that this was CD4 + T cell-intrinsic and to attempt to recapitulate the ageing phenotype in vivo, we looked at the effect of inducing activated Notch in T cells during immunisation. For this, we generated mice in which Notch signalling and linked GFP expression can be induced upon tamoxifen administration in ovalbumin-specific transgenic T cells (Murtaugh et al., 2003). In this OT-II ERT2- that are associated with GC-Tfh cells (Figure 6n, p). Indeed, active Notch increased the proportion of cells that became pre-Tfh but did not promote full differentiation into GC-Tfh cells (Figure 6q, r). This demonstrates that by enhancing the Notch pathway specifically in T cells, the age-associated accumulation of pre-Tfh cells can be recapitulated in younger adult animals.

| DISCUSS ION
This study shows that ageing affects the early stages of Tfh cell

F I G U R E 4
Ageing changes the response to activation of human and murine naïve CD4 + T cells. Representative flow plot (a) and quantitation (b) of calcium flux after CD3/CD28 cross-linking induced flux in human naïve T cells taken from younger (clear circles) and older (red circles) donors. Representative flow plots of pCD247 (c), pERK (d) and pS6 (e) after CD3/CD28 cross-linking induced stimulation of human naïve T cells taken from younger (grey line) and older (red line) donors at 10 min compared with unstimulated older donors (black line). Percentage of cells positive for pCD247 (f), pERK (g) and pS6 (h) following 10 min of CD3/CD28 activation in human naïve CD4 + T cells from younger (clear circles) and older (red circles) donors. Representative histograms and graphs showing levels of IRF4 expression in human naïve CD4 + T cells from younger and older donors 20 hr postactivation with CD3/CD28 (i-k). Representative histograms showing levels of IRF4 expression expressed 20 hr post-CD3/CD28 activation of mouse naïve CD4 + T cells purified from spleens of 8-to 12-week-old and >95-week-old mice 20 hr (l-n). Statistics were calculated using Mann-Whitney U test *p < 0.05 of pre-Tfh cells in aged mice prior to immunisation (Almanan et al., 2020;Lefebvre et al., 2016). We did not observe this and suggest that this difference most likely reflects different antigen load in different facilities. Using naive CD4 + cells from older human donors, we showed that ageing enables cytokine-independent pre-Tfh cell differentiation. These data resolve the discrepancy between the many studies that have reported increased (Herati et al., 2014;Sage et al., 2015;Zhou et al., 2014) and decreased Tfh differentiation in ageing (Lefebvre et al., 2012(Lefebvre et al., , 2016Stebegg et al., 2020). The accumulation of pre-Tfh cells in the absence of similarly increased proportions of GC-Tfh cells suggests that there is a further block in Tfh cell differentiation in aged mice as previously described (Lefebvre et al., 2016).
The aim of this study was to gain mechanistic insight into why pre-Tfh cell differentiation is affected by age. fects their ability to respond to antigenic stimulation. Inflammaging, a chronic low-grade inflammation, is associated with increased age and may alter T-cell responses (Fulop et al., 2017). Naïve T cells also depend upon IL-7 and tonic signalling for their long-term survival (Koenen et al., 2013). These signals may also change with advanced age (Nikolich-Zugich, 2018). We already know that the stromal cells (which are the main source of IL-7) show profound age-associated changes (Becklund et al., 2016;Denton et al., 2020).
The in vitro culture system enabled us to look at early T-cell activation and showed that CXCR5 was upregulated early and that TCR signalling was perturbed. This was confirmed by the increased expression of IRF4, a transcription factor sensitive to the intensity of TCR activation (Iwata et al., 2017;Krishnamoorthy et al., 2017;Man et al., 2013). Recent work has suggested that age is associated with a bias towards Th9 cell differentiation (Hu et al., 2019). Although our study did not extend to analysis of Th9 cells, the activation-induced increase in IRF4 expression associated with age corroborates our findings. Together, these studies suggest that ageing results in a loss of CD4 + T cell unbiased multipotency (Hu et al., 2019 Le Page et al., 2018;Li et al., 2012;Ye et al., 2018). Our data show that age-induced changes in T-cell activation have profound outcomes, affecting the differentiation and ultimately the function of effector T cells. Given that age is the biggest risk factor for poor health, this work highlights the need to understand the ageing immune system, especially if we want to develop novel interventions and therapeutics to ameliorate immuneand inflammation-associated diseases in older people-arguably the sector of the population that is most in need of such interventions.
F I G U R E 5 Role for transcription factor RBPJ in age-associated CXCR5 expression in human CD4 + T cells. Venn diagram summarising data from RNA-Seq analysis where age-induced genes are shown in red and cytokine-induced genes in white with the 34 genes present in both samples listed in the table (a). Bar graph showing individual values of normalised reads for RBPJ in RNA-Seq libraries in the different cell culture conditions indicated (b). Correlations between levels of RBPJ with expression of CXCR5 as determined by both RNA-Seq (c) and flow cytometry (d) on the same samples 72 hr postactivation. Graphs showing inhibition of RBPJ expression by shRNA as determined by RT-PCR (e). Representative flow cytometry plots showing effect of shRBPJ and control (shScr) on pre-Tfh cell differentiation at 72 hr postactivation (f). Graphs showing CXCR5 (g), PD-1 (h) protein expression, IL21 (i) and BCL6 (j) relative to inhibition by control lentivirus (shScr) for CD4 + cells from individual older donors. ATAC-Seq of naïve and activated (72 hr) Wt and Irf4 −/− mouse CD4 + T cells at Rbpj locus. IRF4 ChIP-Seq tracks showing site of IRF4 binding within the Rbpj locus of naïve murine CD4 + T cells (k). Each symbol is representative of individual values from independent donors. Statistics calculated using Mann-Whitney U test, paired test *p < 0.05, **p < 0.005, ***p < 0.0005 1W1K-conjugate or NP-OVA (4-hydroxy-3-nitrophenylacetyl-ovalbumin) in Alum. Imject Alum (#77161) was purchased from Thermo Fisher Scientific, and NP-OVA (#N-5051-100) was purchased from Biosearch Technologies. The 1W1 K conjugate was generated as previously described (Stebegg et al., 2020).
For adoptive transfer experiments, 5 × 10 4 Vα2 TCR transgenic TEα or 1 × 10 5 OT-II CD4 + T cells (isolated from spleen or LNs) were injected via the tail vein. For tamoxifen treatment, mice were orally gavaged with 200 mg/kg body weight of emulsified tamoxifen in sunflower oil. Committee) and processed as previously described (Hill et al., 2019).

| Cell purification and sorting
Naïve CD4 + T cells were enriched using Magnisort Human Naive CD4 + T Cell Enrichment Kit and then stained with antibodies against CD3, CD4, CD27 and CD45RA prior to sorting on a BD Biosciences Influx cell sorter. Cells used were >99% pure as assessed by flow.
For mouse naïve CD4 + T-cell isolation, lymphocytes were isolated from lymph nodes of either 8-to 12-week-old or 95-to 102-week-old C57BL/6 mice. Cells were enriched using Magnisort Murine Naïve CD4 + T Cell Enrichment Kit and then stained with antibodies against CD3, CD4, CD44 and CD62L prior to sorting on a BD Biosciences Aria cell sorter. Cells used were >99% pure.
For T-cell-B-cell co-culture experiments, human naïve CD4+ T cells were purified and activated as described above. On day 3 of culture, cells were plated in duplicate 96 round-bottomed wells at 1 × 10 4 with 5 × 10 5 B cells from a young allogeneic donor.
Supernatants were harvested on day 11 and IgG levels determine by sandwich ELISA using anti-human IgG to coat plates and biotinylated anti-human IgG to detect IgG with a standard curve generated from purified human IgG.
For shRNA experiments, shRNA-encoding DNA oligonucleotides (custom made by Sigma) were cloned into the HpaI and XhoI site of the pLentilox3.7 green fluorescent protein (GFP) lentiviral vector (Addgene). The targeting F I G U R E 6 Increased RBPJ expression enables Notch to drive CXCR5 expression. Western blot (a) and quantification (b) of cleaved form of NICD after 48 hr of CD3/CD28 stimulation of human naïve CD4 + T cells from younger and older people. Levels of NICD relative to α-actin and are shown for younger and older donors (b). Surface expression of Notch 1 (c) and Notch 2 (d) as determined by flow cytometry following 48 hr CD3/CD28 activation of human naïve CD4 + T cells from younger (clear circles) and older (red circles) donors. Western blot of NICD levels after treatment of CD3/CD28-activated human naïve CD4 + T cells in the presence of a partially inhibiting dose of Ly411-575 (e). Effect of gamma-secretase inhibitors Ly411-575 (50 nM) and JLK6 (5uM; Notch inactive) on CXCR5 expression and pre-Tfh cell differentiation of human naïve T cells from donors >60 years of age following stimulation via CD3/CD28 (f-i). The expression of MAF (j), BCL6 (k), RBPJ (l), TBX21 (m), mRNA as determined by RT-PCR 72 hr postactivation in the presence of Ly411-575. Representative flow cytometry plots of active Notch (NICD-GFP) in transferred murine OT-II T cells taken from draining lymph nodes 6 days postimmunisation, and CXCR5 expression on GFP − CD4 + cells (top, grey) and GFP + CD4 + cells (lower, green; n). Graphs showing proportion of CXCR5 int (o) and CXCR5 hi (p), Pre-Tfh (q) and GC-Tfh (r) cells in either GFP + NICD + or GFP − NICD − murine OT-II T cells. Each symbol is representative of individual values from independent mice or donors. Statistics calculated using Mann-Whitney U test *p < 0.05, **p < 0.005, ***p < 0.0005, ****p < 0.0001 sequence for human shRBPJ has been previously described (Niessen et al., 2008), 5'-GCATGGCACTCCCAAGATTGA; shScrambled, 5′-GATTAGAACCCTCACGGTACG-3′.
To produce lentivirus, transfer vector, 4.2 pGagPol (packaging plasmid) and pEcoEnv (envelope plasmid) were transfected into HEK293 T cells using Lipofectamine 2000 (Thermo Fisher Scientific UK). Human naïve CD4 + T cells were activated for 24 hr with anti-CD3/CD28 activator beads and then spin infected with LentiX (Clontech) concentrated viral supernatants generated by transfected HEK293 T cells. T cells were assessed 2 days later.

| Flow cytometry analysis
Cells were harvested and stained with antibodies against cell surface antigens, CXCR5, PD-1, CD25, CD28, Notch1 and Notch2 (all antibodies used are shown in Table 1). To stain for 1W1 K-specific CD4 T cells,

| RNA isolation and RT-qPCR
Total RNA was extracted using RNeasy Micro kit (Qiagen), and firststrand cDNA was synthesised on 100 ng RNA using iScript RT-PCR overnight with primary antibodies (1:1000 rat anti-Notch1 and 1:10 000 mouse anti α-actin). Immunoreactive bands were visualised by enhanced chemiluminescence using Immobilon Western Reagent (Millipore Inc.) and detected with the G:Box System (Syngene).

| Statistical analysis
Prism software (GraphPad) was used for all statistical analysis.
Data were analysed with a two-sample unpaired (or paired, where appropriate) Mann-Whitney U test or Kruskal-Wallis test to compare variables across groups. p values were considered significant when <0.05.

ACK N OWLED G EM ENTS
We thank the National Health Service Blood and Transplant for provision of leucocyte cones. We are grateful to the staff of the