The metabolic hormone leptin promotes the function of TFH cells and supports vaccine responses

Follicular helper T (TFH) cells control antibody responses by supporting antibody affinity maturation and memory formation. Inadequate TFH function has been found in individuals with ineffective responses to vaccines, but the mechanism underlying TFH regulation in vaccination is not understood. Here, we report that lower serum levels of the metabolic hormone leptin associate with reduced vaccine responses to influenza or hepatitis B virus vaccines in healthy populations. Leptin promotes mouse and human TFH differentiation and IL-21 production via STAT3 and mTOR pathways. Leptin receptor deficiency impairs TFH generation and antibody responses in immunisation and infection. Similarly, leptin deficiency induced by fasting reduces influenza vaccination-mediated protection for the subsequent infection challenge, which is mostly rescued by leptin replacement. Our results identify leptin as a regulator of TFH cell differentiation and function and indicate low levels of leptin as a risk factor for vaccine failure.

F ollicular helper T (T FH ) cells constitute a specialised CD4 + T-cell subset that is essential for supporting humoral immunity 1,2 . The importance of T FH cells in health is evidenced by the identification of genetic mutations in genes encoding proteins critical to mounting T FH -cell-dependent humoral responses: inducible co-stimulatory receptor ICOS, SLAM-associated signal adaptor protein (SAP), cytokine receptors for IL-12 and IL-21, and signal transducer and activator of transcription 3 (STAT3). Loss-of-function mutations in these genes impair the generation of functional T FH cells and lead to defective humoral immunity 1,3,4 . On the other hand, excessive T FH generation and function promote the production of autoantibodies and the development of autoimmune diseases 1,4,5 . Studies suggest that aberrant T FH generation in autoimmune disease is likely driven by continuous (auto)antigen stimulation and the upregulation of T FH -inducing cytokines, such as IL-6 [5][6][7][8] . Between these two extreme ends of the spectrum of T FH cell function, we know very little about how the function of T FH cells is regulated in the general population.
As one of modern medicine's greatest successes, vaccinations against childhood and adult infections have saved millions of lives. However, it is not always effective in all individuals. Even in a healthy general population, vaccination is often limited by poor responses in a significant portion of people 9,10 . For example, the clinical efficacy of a standard dose of non-adjuvanted influenza vaccine was 70-90% in young adults but dropped sharply to only 17-53% in the elderly (aged ≥ 65 years) 11 . One key factor limiting vaccine responses is the activity of T FH cells. We and others have reported that T FH cells are induced early following vaccination 7,[12][13][14] , and the frequency of functional T FH cells correlates with the strength of B-cell responses and protective antibody production 12,13,[15][16][17] . Therefore, it is conceivable that variable vaccine responses among the healthy population might be regulated by individual activities of T FH cells. What determines the T FH activity among the general population?
Studies have highlighted the major influence that the metabolic state can exert on the immune system [18][19][20] . Employing a systemic approach, the multiscale, multifactorial response network, Li et al. revealed that metabolic phenotypes are tightly coupled with vaccine immunity in healthy individuals 21 . As an adipokine that is abundantly secreted by adipose tissue, the metabolic hormone leptin broadly regulates many immune cell types of both innate and adaptive immune systems 19,22,23 . Notably, leptin levels vary drastically (range over 10-fold) in healthy individuals 15 , but the function of leptin in regulating T FH cells and vaccine responses remains unclear. We, therefore, investigated whether leptin is a natural regulator of T FH function in healthy individuals and whether it plays a role in determining vaccine response.
In this study, we show that low levels of serum leptin in both young and elderly healthy groups are associated with reduced antibody responses to influenza and hepatitis B virus (HBV) vaccines. Leptin promotes the differentiation and function of both human and mouse T FH cells in culture and is required to support T FH function and effective humoral immunity to infection, immunisation and vaccination in mice. The mechanism of action of leptin is mediated partially through the activation of Stat3 and the mechanistic target of rapamycin (mTOR) pathways. Our results suggest leptin is a physiological regulator of T FH function and insufficient leptin might serve as a biomarker to identify the risk for low vaccine efficacy.

Results
Low leptin levels are associated with reduced vaccine responses. We first examined the relationship between serum leptin levels and the immune response in healthy adults (aged 18-60 years, n = 76) immunised with a standard dose of non-adjuvanted trivalent influenza vaccine 14,24 . We classified individuals into responder group (HI titer fold-change to any one of the three strains at baseline and after vaccination ≥2, n = 71) and non-responder group (HI titer to all three strains <20 and HI titer fold-change to all three strains <2, n = 5) (Supplementary table 1). We found the serum leptin levels of the non-responder group showed about 2.5-fold reduction compared to those of the responder group: median, 4.2 vs 10.5 ng/mL (P = 0.03, Student's t-test; Fig. 1a). The lower leptin levels in non-responders than responders suggested leptin insufficiency as a risk factor for vaccine response. Therefore, individuals were divided into the group with relatively low levels of leptin (lower quartile, n = 19) and sufficient levels of leptin (middle and upper quartiles, n = 57). More strikingly, we observed that non-responders were highly enriched in the low leptin group: the non-response rate of 21.1% in the low leptin group vs 1.8% in the sufficient leptin group (P = 0.003, Chi-Square test; Fig. 1b).
The vaccine efficacy significantly drops in the elderly group 11 . This promoted us to conduct a similar analysis in a group of healthy elderly adults (aged ≥ 65 years, n = 83) immunised with a standard dose of non-adjuvanted trivalent influenza vaccine. We observed that low leptin levels were again associated with poor antibody responses after vaccination in healthy elderly adults. The serum leptin levels of the non-responder group were significantly lower than those of the responder group: median, 8.0 vs 15.5 ng/ mL, P < 0.001, Student's t-test; Supplementary table 2, Fig. 1c). The individuals with relatively low levels of leptin (the lower quartile, n = 21) were highly enriched with non-responders: the non-response rate was 52.4% in the low leptin group vs 6.5% in the sufficient leptin group (P < 0.001, Chi-Square test; Fig. 1d).
To test whether the association between leptin levels and vaccination responses was a common phenomenon for a different vaccine, we next examined a group of young adults (aged 18-25 years, n = 22) with low serological anti-HBs (<100 mIU/mL) who were re-immunised with the vaccine for HBV (Supplementary  table 3). The non-responder group (serum antibody titre <200 mIU/mL after primary immunisation, n = 8) showed about 5.0fold lower leptin levels than the responder group (serum antibody titter fold changes ≥2, n = 14): median 3.2 vs 16.0 ng/mL (P = 0.004, Student's t-test; Fig. 1e). The non-response rate was 80.0% in the low leptin group (the lower quartile, n = 5) and 23.5% in the sufficient leptin group (n = 17) (P = 0.021, Chi-Square test; Fig. 1f). These results demonstrated that in different vaccination and different age groups, non-responders showed lower leptin levels than responders, suggesting that low levels of leptin could be a limiting factor for effective vaccine responses or indicate a relationship between vaccine responses and systemic metabolic state, since serum leptin concentrations correlate with the percentage of body fat 15,25 .
Leptin potentiates the differentiation and function of human T FH cells. Previous studies including ours revealed that the T FH activity is critical for the effectiveness of human vaccination [12][13][14]16 . We next asked whether leptin levels affected the T FH generation after vaccination. Among the three cohorts above, peripheral blood mononuclear cells (PBMCs) were collected and stored for the middle-age cohort (18-60 years old, n = 76) immunised with influenza vaccines (Fig. 1a, b). When the T FH generation in PBMCs was profiled, we found that the vaccine-induced T FH differentiation correlated with the production of protective antibodies (Supplementary Fig. 1), confirming previous reports that the T FH activity controlled antibody production after vaccination 12,14,16,17 . Importantly, those with high T FH activities (T FH fold-change >1) had significantly higher serum leptin than those with low T FH activities: median, 11.1 vs 6.3 ng/mL (P = 0.046, Student's t-test; Supplementary table 4, Fig. 2a). The rate of low T FH activity was 36.8% in the low leptin group, compared to 14.3% in the sufficient leptin group (P = 0.031, Chi-Square test; Fig. 2b). These data suggested that leptin might be required for the optimal generation of T FH cells.
To formally test whether leptin potentiates human T FH differentiation, human naïve CD4 + T cells were stimulated with anti-CD3/CD28, in the presence or absence of leptin. The addition of leptin enhanced the mRNA of multiple markers for T FH cell differentiation (Fig. 2c). As a control, leptin treatment reduced IL4 and increased IL17 transcripts in activated human CD4 + T cells ( Supplementary Fig. 2a), in agreement with its reported role in suppressing T H 2 and promoting T H 17 cells [26][27][28] . Flow cytometric analysis indicated that leptin potently promoted the expression of CXCR5 (Fig. 2d) and the production of IL-21 ( Fig. 2e) from cultured human CD4 + T cells in a dose-dependent manner. TGF-β plus IL-12 or IL-23 were shown to promote CXCR5 expression and T FH differentiation 29 . The addition of leptin could also synergise with each of these reported T FH -inducing cytokines to enhance the expression of CXCR5 and IL-21 ( Fig. 2f, g). The direct effect of leptin in promoting human T FH differentiation supports the notion that low levels of leptin might impair the activities of T FH cells and confers the risk of poor vaccine responses.

Defective leptin signalling impairs antibody responses in mice.
Can leptin regulate antibody responses in vivo? We first detected leptin by immunofluorescent staining and found it was abundantly produced in white pulps including T-cell zones and B-cell follicles ( Supplementary Fig. 3a, b), suggesting that CD4 + T cells and B cells could access leptin throughout antibody responses. The expression of LepR was higher on T FH cells than on naïve CD4 + T cells ( Supplementary Fig. 3c) and was further induced by antigen, co-stimulatory signals and leptin (Supplementary Fig. 3d- (Fig. 3a, Supplementary Fig. 4a). Despite the increased viral antigens, db/db mice produced significantly less virusspecific antibodies in sera and bronchoalveolar lavage fluid (BALF) (Fig. 3b, c). Draining lymph node is important for memory T FH formation and vaccination efficacy 30 , flow cytometric analysis revealed a significant reduction of T FH cells (Fig. 3d), accompanied by the decrease in germinal centre (GC) B cells and antibody-secreting cells (ASCs) (Fig. 3e, f) in the mediastinal lymph nodes (mLNs). An Enzyme-Linked Immuno-Spot (ELISPOT) assay also showed a two-fold reduction of ASCs in spleens of db/db mice, compared to WT mice (Fig. 3g). We previously showed that inflammatory cytokines induced by acute malaria infection could suppress T FH differentiation 31 . To rule out the possibility that the weakened T FH generation and humoral response in db/db mice was caused by the persistence of high virus titres and enhanced inflammatory cytokines, we immunised WT and db/db mice with 2,4,6,-trinitrophenyl (TNP)-keyhole limpet hemocyanin (KLH) in complete Freund's adjuvant (CFA) and we again observed defective T FH and GC B-cell differentiation and reduced antibody production, compared to WT mice (Supplementary Fig. 4b-d). IL-21 is the key cytokine produced by T FH cells that underpins the B-helper function [32][33][34][35] . Compared to WT cells, db/db CD4 + T cells produced less IL-21 (Supplementary Fig. 4e). These data indicate that LepR deficiency impairs T FH and B-cell responses following infection and protein immunisation.
T-cell-specific LepR is required for optimal T FH and antibody responses. We previously reported that leptin promotes B-cell homeostasis by inhibiting apoptosis and inducing cell-cycle entry 36 . To investigate the role of cell-autonomous leptin signalling in T and B-cell differentiation, we transferred combined T cells and B cells from either WT or db/db mice into Rag1 -/recipient mice. After 1 week, mice were immunised with 4hydroxy-3-nitrophenyl (NP)-ovalbumin (OVA) in CFA and analysed after 9 days ( Supplementary Fig. 5a). The lack of leptin signalling in either T or B cells was sufficient to impair the generation of GC and ASCs ( Supplementary Fig. 5b). This observation was also supported by testing the capability between WT and db/db T FH cells in supporting GC B cells to produce antibodies in the in vitro co-culture ( Supplementary Fig. 5c). Isolated db/db T FH cells were less capable than their WT counterparts in this assay while co-culturing db/db T FH and db/db GC  Fig. 5d). To test cell-autonomous leptin signalling in a physiological manner, we reconstituted irradiated Rag1 -/mice with 50:50 mixed bone marrow (BM) cells from db/db or WT (both CD45.2) mice and congenically marked WT (CD45.1) mice. After reconstitution, BM chimeric mice were immunised with NP-OVA in CFA and analysed 9 days after immunisation. An analysis of CD4 + T cells showed that db/db and WT-derived Foxp3 + T REG and total Foxp3 -CD44 high effector subsets were comparable. However, the CXCR5 + Bcl6 + T FH subset was reduced by 2-fold in db/db cells, compared to WT cells ( Supplementary Fig. 6a), suggesting an intrinsic defect of db/db T FH cells. We observed that db/dbderived NP + IgG1 + antigen-specific B cells and CD138 high ASCs were reduced compared to counterpart WT cells ( Supplementary  Fig. 6b). The frequencies of the memory and GC B-cell subsets in NP + IgG1 + antigen-specific B cells were comparable between db/ db and WT cells ( Supplementary Fig. 6b). These sets of experiments demonstrated that leptin signalling in both T and B cells are required for optimal humoral responses, especially for T FH differentiation and the generation of ASCs.
To evaluate the contribution of T-cell specific leptin signalling to humoral immunity in vivo, we crossed Lepr floxed mice with Cd4-Cre transgenic mice to specifically delete LepR signalling in T cells. Mutant (Cd4-Cre:LepR fl/fl ) and WT (Cd4-Cre:LepR +/+ ) mice showed comparable B cells, Foxp3 + T REG cells and CD44 + effector CD4 + T cells 37 . Both groups of mice were infected with the H3N2 influenza virus and analysed after 9 days. There was no significant difference between the two groups for the percentages of total CD44 + effector CD4 + cells or Foxp3 + T REG cells (Fig. 4a). However, the CXCR5 + Bcl6 + T FH percentages in effector cells and their numbers in the draining mediastinal LN were almost halved in mutant mice, compared to WT controls (Fig. 4b). We also observed a significantly reduced production of IL-21 in mutant CD44 + effector CD4 + cells, compared to WT cells (Fig. 4c). On the contrary, the lack of LepR signalling showed negligible effect on the production of major effector cytokine IFN-γ in this model, neither IL-4 nor IL-17 ( Supplementary Fig. 7a-c). We also quantified follicular regulatory T (T FR ) cells, a subset of T REG cells that localise in B-cell follicles to suppress T FH cell-mediated antibody responses 38,39 . There was a trend for the reduction of T FR cells in the absence of LepR signalling but it didn't reach significance ( Supplementary Fig. 7d). Finally, we examined the key question of whether the LepR deficiency in T cells perturbed GC formation and antibody production. The T-cell-specific deletion of LepR led to about 2-fold reduction of GL-7 + Fas + GC B cells (Fig. 4d) and almost a 3-fold reduction of CD138 + TACI + ASCs (Fig. 4e), which was accompanied by the significant decrease of influenza virus-specific IgG1 and IgG2 antibodies (Fig. 4f). Therefore, defective leptin signalling autologously restrained T FH numbers and function, which constituted a limiting factor for antibody responses.
Leptin promotes T FH function via Stat3 and mTOR signalling pathways. Compared to WT cells, effector CD4 + T cells in db/ db or Cd4-Cre:LepR fl/fl mice produced less IL-21, the key effector cytokine for T FH function, than their WT counterparts ( Supplementary Fig. 4e and Fig. 4c), suggesting a functional impairment of T FH cells in the absence of leptin signalling. To test this, we added IL-21 in the T-B co-culture and found the defective B-helper function of db/db T FH cells was improved to the levels similar to that of WT T FH cells (Fig. 5a). We also confirmed that leptin directly promoted the production of IL-21 from stimulated mouse CD4 + T cells (Fig. 5b), similar to the effects of promoting IL-21 secretion by cultured human CD4 + T cells (Fig. 2e).
We next set to identify the molecular mechanism by which leptin stimulates IL-21 production. As a key downstream signal pathway that mediates the function of leptin 22,40 , Stat3 signalling is critical for the generation of both human and mouse T FH cells and IL-21 production by these cells 1,4 . Leptin induced Stat3 phosphorylation in mouse CD4 + T cells (Fig. 5c, raw gel blots in Source Data). Using the Chromatin Immunoprecipitation (ChIP) assay, we found that leptin increased the binding of Stat3 to the Il21 promoter in mouse CD4 + T cells (Fig. 5d, raw gel blots in Source Data), suggesting that the leptin-Stat3 axis enhances the Il21 transcription. To directly examine it, the Il21 promoter was cloned into a luciferase reporter construct (Fig. 5e), which was then transduced into 293 T cells. Leptin treatment was able to enhance the transcriptional activity of the Il21 promoter, measured by the luciferase activities (Fig. 5f). In line with this, Stat3-deficient CD4 + T cells failed to respond to leptinstimulated IL-21 production (Fig. 5g). Together, these data suggest that Stat3 mediates leptin-stimulated IL-21 production.
In addition to Il21, adding leptin in cultured CD4 + T cells also increased the mRNA of other major markers for mouse T FH cell differentiation, including Cxcr5, Bcl6, Icos and Cd40l, but not Il4 ( Fig. 6a and Supplementary Fig. 2b), suggesting that leptin alone induces the overall T FH differentiation program, consistently with the promoting function of leptin for human T FH differentiation (Fig. 2d). Indeed, not only did leptin treatment enhance the expression of CXCR5 on activated CD4 + T cells (Fig. 6b), it also induced the expression of the key T FH transcription factor Bcl6 (Fig. 6c, raw gel blots in Source Data). Both mTOR and Stat3 pathways are required for mouse T FH cell differentiation and function 1,41,42 . Besides the Stat3 pathway (Fig. 5c), lepR signalling also activates the mTOR pathway in cultured CD4 + T cells, as measured by the phosphorylation of p70 S6K and S6, one major pathway downstream of mTOR (Fig. 6d, raw gel blots in Source Data). We found that leptin-induced Bcl6 expression was blunted by the treatment of the inhibitors for the Stat3 pathway (S3I-201) or the mTOR pathway (Rapamycin) (Fig. 6e, f, raw gel blots in Source Data). It has been shown that mTOR kinase complexes 1 and 2 (mTORC1 and mTORC2) are both required for T FH responses, with mTORC2 more specifically regulating T FH differentiation and mTORC1 mainly promoting CD4 + T-cell proliferation [41][42][43] . Correspondingly, CD4 + T cells from Cd4-Cre: Stat3 fl/fl or Cd4-Cre:Rictor fl/fl mice were unable to upregulate Bcl6 expression upon the leptin treatment (Fig. 6g, h). Collectively, leptin induces the activation of Stat3 and mTOR pathways to promote T FH differentiation and function.

Leptin insufficiency impairs vaccination-mediated protection.
After establishing that the genetic ablation of leptin signalling impaired humoral immunity, we next asked whether the physiological restriction of leptin production attenuates antibody responses in vivo. We first employed an acute fasting model that had been shown to reduce leptin production 25 . OT-II CD4 + T cells that specifically recognise the OVA peptide were adoptively transferred into congenic WT mice that were subsequently immunised with NP-OVA in CFA. Mice were fed with a normal diet (Fed) or acutely fasted (Fast) on day 7 and 8 when T FH and GC responses were at peak 7 . Fasted mice were divided into two groups, which were treated with either leptin to compensate for fasting-induced leptin insufficiency or PBS as control (Supplementary Fig. 8a), to delineate the effect of leptin insufficiency on the antibody response. The expansion of transferred CD45.1 + OT-II cells in recipient mice with acute fasting was significantly inhibited (Supplementary Fig. 8a, b) while the differentiation of CXCR5 + Bcl6 + T FH cells was further repressed for more than 2fold ( Supplementary Fig. 8a,b). Leptin supplementation during acute fasting partially improved the expansion of OT-II cells and largely rescued the T FH differentiation ( Supplementary Fig. 8a, b). Importantly, leptin treatment completely restored serum IL-21 levels, which was reduced by over 2-fold by acute fasting for 48 h ( Supplementary Fig. 8c). These results demonstrate that the physiological restriction of leptin by acute fasting diminished both the generation and function of T FH cells. We also measured the generation of NP-specific Fas + GL-7 + GC B cells and CD138 +  Fig. 8d). Although leptin supplementation was unable to rectify the generation of NP-specific B cells weakened by acute fasting (Supplementary Fig. 8e), it fully recovered GC and ASC differentiation ( Supplementary Fig. 8d,e). Therefore, leptin insufficiency caused by acute fasting plays a central in suppressing T FH function and humoral immunity.
We next turn to address the question of whether leptin insufficiency during vaccination affects vaccination-mediated protection. WT mice were fed with normal diet (Fed) or alternate-day fasted from day 5 to 15 (Fast) following influenza vaccination. Alternate-day fasting immediately reduced the leptin levels and body weight, which could be recovered by normal feeding (Supplementary Fig. 9). The choice of the fasting period from day 5 to 15 was designed not to affect T-cell priming (before day 5) and to selectively target T FH differentiation, which starts around day 5 7 . After a period of normal feeding from day 16 to 21 so that fast mice recovered body weight ( Supplementary Fig. 9), mice were challenged with the H1N1 influenza virus at a median lethal dose (LD50). Fasted mice were again divided into two groups, which were treated with either leptin or PBS as the control on alternate fasting days from day 5 to 15 (Fig. 7a). Compared to normal feeding, alternate-day fasting during the period of vaccination greatly reduced the survival rate of virus-infected mice from 72.2% to 38.9% (Fig. 7b). Noticeably, leptin replacement on alternate fasting days almost completely restored vaccinationmediated protection in fasted mice, with a survival rate of 63.2% (Fig. 7b). Influenza vaccine-rendered protection is considered predominantly mediated by humoral immunity 44,45 . We, therefore, examined virus-specific antibody production and revealed that the decreased antibody production in alternate-day fasted mice was rescued by leptin replacement (Fig. 7c). Correspondingly, flow cytometric analysis demonstrated that alternate-day fasting retarded the generation of CXCR5 + Bcl6 + T FH cells, as well as GL-7 + Fas + GC B cells and CD138 high ASCs but this was also compensated by leptin treatment (Fig. 7d-f). To further examine whether the observed effect of GC and T FH response by fastinginduced leptinemia and leptin replacement was dependent on leptin signalling in T cells, we conducted a similar experiment in Cd4-Cre:LepR fl/fl mice. Mice were fed with normal diet or alternate-day fasted from day 5 to 15 following influenza vaccination and then analysed the antibody response at day 21 when bodyweights had recovered (Supplementary Fig. 10a, b).  Whitney U-test (a, b), or two-way ANOVA (f). *P < 0.05, **P < 0.01. Results are representative of three independent experiments.
In contrast to wild-type mice, alternate-day fasting or leptin replacement didn't show noticeable effect on T FH cell differentiation or GC and ASC formation ( Supplementary Fig. 10c, d). Intriguingly, fasting or leptin treatment didn't change the production of vaccine-specific IgG2c and IgG2b antibodies but altered IgG1 antibodies to all three viral strains ( Supplementary  Fig. 10e). The results from the physiological restriction of leptin support the conclusion that leptin insufficiency is a significant risk factor for T FH function, antibody responses and vaccinationmediated protective humoral immunity.

Discussion
Metabolism plays a fundamental role in regulating the immune system. Not only are the energetic and biosynthetic pathways within immune cells essential in controlling immune cell development, activation, functional specification and fate determination [18][19][20] , but the state of systemic metabolism can also directly influence the immune system through signals conveyed by hormones and cytokines, in which the adipokine leptin has been extensively studied 22,23,40 . Leptin specifically regulates the differentiation of T-cell functional subsets. It enhances T H 1 but suppresses T H 2 cytokine production 26 . We and others have demonstrated that leptin activates the Stat3 pathway to promote the generation of T H 17 cells 27,28,37 , while it limits the proliferation of T REG cells through the activity of mTOR pathway [46][47][48] . Moreover, leptin has been shown to regulate the metabolic fitness of CD8 + T cells and control their effector function in the tumour microenvironment and the responses to anti-PD-1 therapies [49][50][51] . In this study, we revealed leptin in promoting T FH function and supporting antibody responses.
The involvement of leptin in antibody responses was hinted at by previous observations that, compared to WT mice, db/db mice produced less protective antibodies to a vaccine against Helicobacter pylori 52 , and fewer autoantibodies in the pristine-induced model for lupus 53 . However, the mechanism underlying these observations was unknown. By testing leptin treatment in cultured PBMCs from patients of rheumatoid arthritis, Wang et al. reported the leptin treatment could promote T FH function by enhancing IL-6 production in PBMCs although the study didn't address whether such regulation was T-cell intrinsic or not 54 . Using a set of experiments, our results identified T FH cells as the effector CD4 + T subset that is particularly affected by the lack of leptin signalling. We found that leptin signalling was not only required for optimal T FH cell differentiation but was also sufficient to promote T FH cell generation and directly induce IL-21 production in mice and humans. In culture CD4 + T cells, leptin stimulated the ICOS expression in human cells while both Icos and Cd40l in mouse cells, suggesting the function of leptin in broadly enhancing T FH effector function. It remains to be determined whether the regulation of ICOS and CD40L is directly targeted by leptin signalling or in an indirect manner such as through the upregulation of Bcl6. We have previously shown that leptin signalling is required for B-cell homeostasis 36 . In line with this, we also found a cell-autonomous role for leptin in the expansion of antigen-specific B cells and the differentiation of ASCs in the experiment of mixed transfer of T and B cells in RAG-1-deficient mice. Intriguingly, although the effects of alternate-day fasting or leptin replacement on GC and T FH cells were largely dependent on T-cell intrinsic leptin signalling, the production of IgG1 but not IgG2 antibodies was affected by T-cell extrinsic leptin signalling. Future studies are required to dissect leptin function intrinsic to B cells or other cell types in humoral immunity. One caveat of this study is the lack of investigation of how leptin regulates antibody affinity maturation. Reduced T FH and GC responses in the absence of leptin signal suggest leptin is required for optimal GC reaction underlying antibody affinity maturation, which should be formally evaluated in future. In conclusion, the reduction of leptin triggers multiple mechanisms of actions to restrain adaptive immunity, in this case, humoral immunity. We hold the view that leptinemia-associated immune incompetence constitutes a safeguard for energy-saving and sustaining the body's basic requirements during malnutrition.
Our results demonstrated that the non-responder groups for both influenza and HBV vaccinations were highly enriched in individuals with low serum leptin (Fig. 1). On the other hand, serum leptin levels didn't always positively correlated with absolute antibody titres post-vaccination or antibody titre changes in adults with influenza or HBV vaccines (Supplementary Fig. 11), in agreement with a previous publication 55 . Altogether, our data support the notion that leptin is a natural regulator of T FH cells in the general population. This should not be interpreted as 'higher leptin levels, stronger vaccine responses'. Instead, in our opinion, leptin represents a metabolism-mediated threshold factor that is required to mount normal vaccine responses. In other words, a low level of leptin is a profound risk factor for poor vaccine responses. It should also be noted that some individuals with low leptin levels could still generate relatively normal antibody responses, which implies that the negative impact caused by weak leptin signalling might have been compensated by other factors. For example, we previously reported that IL-6, rather than leptin, was highly elevated in patients with rheumatoid arthritis, which correlated with the generation of pathogenic T FH cells in this autoimmune disease 5 . It should be noted that females have higher leptin levels than males when matched by age, weight, or body fat 25 while females also typically develop higher antibody responses following vaccination than males 56 . Therefore, it will be important to further investigate the role of leptin in sexual dimorphism to humoral immunity and vaccine response.
Understanding individual immune responses to vaccination and identifying related immune signatures or biomarkers are essential for vaccine development 57,58 . Emerging evidence suggests that T FH cells represent a major limiting factor for vaccine efficacy. For example, the induction of functional T FH subsets was shown to correlate with effective antibody responses to influenza vaccination 12,14,17 . In addition, in the most promising HIV vaccine trial to date (ALVAC+ AIDSVAX, RV144), the higher frequency of HIV-specific IL-21 + T FH cells was associated with protective antibody responses 16 . Therefore, we and others have proposed to target T FH function as a new approach to improve vaccine efficacy 59,60 . Through association studies using human samples and mechanistic studies using mouse models, we have provided solid evidence to link levels of the metabolic hormone leptin, T FH cell function and vaccine responses. The respective contribution of leptin from systemic production and autocrine secretion from immune cells in the reinforce of vaccine responses remains unclear and needs further investigation. Nevertheless, we suggest that individuals with low levels of leptin should be particularly targeted with strategies to promote their T FH function and enhance vaccine responses. For instance, high-dose and prolonged antigen stimulation favours T FH generation 6 . Indeed, the high-dose influenza vaccine has been shown to elicit greater T FH activation in older adults 61 , which can at least partially explain the increased efficacy of high-dose influenza vaccine in elderly individuals 62 with reduced leptin sensitivity 63 . Understanding the mechanisms underlying varying vaccine responses among healthy individuals will not only provide critical information on the regulation of T FH function in non-pathogenic conditions but is also required to guide the optimisation of immunisation strategies and the development of new vaccines 44,58,59 . Overall, the identification of leptin as a natural regulator of T FH cell function and vaccine responses has farreaching implications for vaccine development.  Influenza virus challenge. H1N1 strain A/Puerto Rico/8/1934 (PR8), or H3N2 strain A/X-31 influenza virus was obtained in the allantoic cavity of 10-dayold embryonated hens' eggs at 37°C with 65% humidity for 48 h 64 . The virus was concentrated from the allantoic fluid and purified in a 10-50% sucrose gradient by centrifugation at 25,000 × g, 4°C for 2 h. Virus samples were further inactivated with 0.25% formalin (v/v) at 4°C for 7 days. The 50% lethal dose (LD 50 ) of pdmH1N1 in WT female mice was determined after serial dilution of the viral stock, and LD 50 dose was used in viral challenge experiments. Female WT and db/ db mice (7-10 weeks of age), were anaesthetized with isoflurane before being intranasally challenged with 30 μL of pdmH1N1 diluted in PBS. Cd4-Cre and Cd4-Cre:Lepr flox mice were intranasally challenged with 10 4 PFU (plaque-forming unit, PFU) A/X-31(H3N2) diluted in 30 μL of PBS after isoflurane anaesthesia.

Methods
Bone marrow chimaeras. Rag1 −/− mice were irradiated and reconstituted with 50:50 mixed bone marrow (BM) cells from db/db or WT (both CD45.2) mice and congenically marked WT (CD45.1) mice. After 4-6 weeks of reconstitution, tail bleeding was analysed by FACS to confirm the success reconstitution of the hematopoietic system. Fasting and leptin administration. Vaccinated mice with human flu vaccine were divided into three groups on day 5: normal diet (Fed) and alternate-day fasting with i.p. injection of 200 μL of PBS (Fast + PBS) or leptin (Fast + Leptin) (1 μg/g per mouse initial body weight) twice daily at 9 am and 6 pm every other day from day 5 to 15. Other days all mice were switched to the normal diet.
Hemagglutination inhibition assays. Antibody titres in the sera were determined by hemagglutination inhibition assays with previously reported protocol 13 . Sera samples were pre-treated with a receptor-destroying enzyme (RDE) for 18 h at 37°C in a water bath and then heated for 1 h at 56°C. Sera were 2-fold diluted with saline solution (0.9%) in 96-well plates. Four agglutinating doses of A/H1N1, A/H3N2, or B strain were added to each well, mixed with sera samples and incubated at room temperature for 1 h. Fresh turkey red blood cells were diluted and added to each well, after 1-h incubation at room temperature, the presence of agglutination inhibition was evaluated. Antibody titres are expressed as the inverse sera dilution that inhibited viral agglutination. The fold changes of serum HI titres of H1N1, H3N2 and B strains on day 7, day 42 were calculated by dividing values at day 0 12,14 .