The Lag3 and PD-1 pathways govern TCR signal duration and migration of CD4+ T cells

Anti-Lag3 and anti-PD-1 combination immunotherapy for melanoma has received recent regulatory approval, yet its mechanism of action is unclear, particularly for CD4+ T cells. Here, we determined the roles of the Lag3 and PD-1/PD-L1 pathways during CD4+ T cell activation in vivo. During primary immune responses, Lag3 played a redundant role with the PD1/PD-L1 axis dominant in regulating early CD4+ T cell activation. By exploiting an adaptive tolerance model, we reveal that Lag3 and PD-L1 co-blockade (CB) drove major changes in CD4+ T cells, resulting in a transcriptional profile dominated by a hybrid T follicular helper (Tfh)/Th17 cell phenotype. Mechanistically, CB enhanced T cell receptor (TCR) signal duration, thereby inducing an NFAT-biased transcriptional motif, previously linked to positive clinical outcomes for melanoma patients. Additionally, CB synergistically upregulated CCR6, leading to enhanced CCL20-mediated CD4+ T cell migration. Analysis of melanoma patients on anti-PD-1 pathway immunotherapies revealed that tumour CCR6 expression stratifies clinical outcomes and effector memory CD4+ T cells upregulate CCR6 in patients on anti-PD-1 and anti-Lag3 combination immunotherapy. Our data reveal that Lag3 and PD-1 pathways have context dependent roles in regulating the activation and migration of CD4+ T cells and highlights CCR6 as a biomarker for immunotherapy responses.


Introduction
Canonical T cell activation is initiated by the binding of the T cell receptor (TCR) to peptide-loaded major histocompatibility complexes (MHC).The integration of signals from co-stimulatory and co-inhibitory molecules fine-tunes this process (1) and can alter the threshold for T cell activation (2).Understanding how T cell activation thresholds are modulated has fundamental implications for the treatment of T cellmediated diseases and the exploitation of T cell-modulating immunotherapies.
An emerging paradigm is that the targeting of more than one immune checkpoint can lead to superior clinical outcomes (21).This has been observed first-hand in the clinic: combination therapy of Ipilimumab (anti-CTLA-4) and Nivolumab (anti-PD1) offers better patient survival compared to monotherapy in melanoma (7).Similarly, a recent study demonstrated superior progression-free survival (PFS) in patients on a combination therapy of Relatlimab (anti-Lag3) with Nivolumab (8).Although cross-trial comparisons are challenging, there is evidence that grade 3/4 adverse effects may be lower with Relatlimab plus Nivolumab combination therapy (21.1% combination therapy vs 11.1% in Nivolumab monotherapy (22)) compared to those observed in Ipilimumab plus Nivolumab (59% combination therapy vs 22% in Nivolumab monotherapy (23)).
Nivolumab and Relatlimab were approved for use as a combination therapy for unresectable advanced melanoma in 2022 by the FDA and EMA, and in 2024 by NICE in the UK.Given that significant proportions of patients fail to respond to monotherapy (8,23), a combination therapy with enhanced efficacy and more tolerable side effects is highly desirable.However, our understanding of how immune checkpoints fundamentally interact to regulate different CD4 + T cell subsets is unclear.
We recently developed an accelerated adaptive tolerance model that can evaluate the effects of immune checkpoints on CD4 + T cell activation in vivo (2,24).Using this model, we showed that PD-1 acts as a rheostat to control the threshold for T cell activation.Its blockade leads to an enhanced TCR signal strength and the upregulation of co-stimulatory molecules, such as ICOS and OX40 (2).Importantly, we found that the co-inhibitory receptor Lag3 was strongly up-regulated in these tolerogenic CD4 T cells, and so we questioned what effect this pathway would have on T cell function in the context of PD-1 pathway blockade.
To investigate how Lag3 may regulate CD4 + T cells in the context of PD-1 pathway blockade, we exploited the Tg4 TCR transgenic model, which is specific for myelin basic protein (MBP) peptide.MBP contains a lysine (K) at position 4, which exhibits unstable MHC binding to the I-A U class II molecule (25,26).Substituting for a tyrosine (Y) generates an affinity for I-A U which leads to highly stable peptide/MHC II complexes (25,27) which is potently recognised by the Tg4 transgenic TCR.Given that Lag3 has a preferential function for targeting stable peptide/MHC II complexes (28,29) we were interested to further understand its role in the context of PD-1/PD-L1 pathway blockade during longer periods of immune activation with [4Y]-MBP.
In this study, we demonstrate that Lag3 and PD-L1 cumulatively govern CD4 + T cell activation by increasing the duration of TCR signalling.This leads to an enhanced state of T cell activation, bolstered migration, and a T follicular helper (Tfh)/Th17 cellbiased phenotype.Finally, we show that these same signatures of extended TCR signalling duration and CCR6-associated migration can be found in melanoma patients receiving PD-1 incorporating immunotherapies, thus revealing new insights into fundamental CD4 T cell biology, while also unveiling new biomarkers for immunotherapy.

PD-1 pathway is dominant over Lag3 in controlling initial CD4 + T cell activation
To investigate the functions of the PD-1 and Lag3 pathways during a primary response we employed two different models for CD4 + T cell activation.In all the following studies we used anti-PD-L1 to modulate the PD-1 pathway as we were interested in identifying T cell sub-populations expressing this marker.Firstly, Tg4 Nur77-Tempo mice were immunised with [4Y]-MBP peptide and we examined the ability of PD-L1, Lag3 or combination blockade (CB) to sustain TCR signalling during initial activation in vivo (Figure 1A-E).Nur77-Tempo mice report distal TCR signalling through expression of Fluorescent Timer(30) protein (FT), which has an initial Blue fluorescent form with a half-life of 4 hours before maturation to an irreversible Red fluorescent form.T cells that sustain TCR signalling during the first few days will appear as Blue + Red + cells at the analysis time point (Figure 1B).Our experiments showed that only anti-PD-L1 treated or CB-treated mice showed significantly increased populations of CD4 + T cell with persistent TCR signalling, with no noticeable effect of anti-Lag3 treatment alone (Figure 1C).In addition, we examined expression of ICOS, a marker linked to strong TCR signalling in response to anti-PD-1 treatment of re-activated CD4 + T cells (2).In this setting we saw a significant upregulation of ICOS in the CB-treated mice compared to the other conditions (Figure 1D).
Next, to understand the roles of PD-L1 and Lag3 during a polyclonal immune response we utilised an immunisation approach with the model antigen NP-Ova (Figure 1E).Mice were immunised with NP-Ova precipitated in alum.Twenty-four hours later mice received either IgG, anti-PD-L1, anti-Lag3 or combination and 7 days later (day 8 post-immunisation) the popliteal lymph nodes were isolated and CD4 + T cells analysed (Figure 1F-H).Anti-PD-L1 and CB significantly increased the frequency of Tfh cells compared to isotype and anti-Lag3 groups, but there was no additive effect of CB compared to anti-PD-L1 alone (Figure 1G).Analysis of ICOS levels on CD4 + T cells (Figure 1H) revealed that, as with the previous Tg4 Nur77-Tempo model, CB imparted a significant increase in ICOS levels on CD4 + T cells.In summary during early CD4 + T cell activation, whilst the PD-L1/PD-1 pathway is dominant over Lag3 in regulating the process CB led to more consistent CD4 + T cell activation in vivo.

Lag3 and PD-L1 pathways exert additive and synergistic effects on CD4 + T cell activation during secondary stimulation
To further investigate the effects of co-blockade (CB) of Lag3 and PD-L1 in vivo, we therefore utilised our accelerated adaptive tolerance model (2,24) to test CB potency effects following T cell reactivation (Figure 2A).In this model, a single high dose of [4Y]-MBP induces transient TCR signalling in vivo and upregulation of multiple immune checkpoints, including PD-1 and Lag3, thereby allowing an assessment of their functions (2).A secondary immunisation subsequently allows a dissection of the roles that Lag3 and PD-L1 play in the activation thresholds for CD4 + T cells.CB significantly enhanced Nr4a3-Red + Nr4a3-Blue + (i.e. cells that have persistent activity of the NFAT-Nr4a3 pathway (31) and henceforth referred to as "responder" cells; mean = 33.3%)compared to isotype control (mean=7.9%)and single agent anti-PD-L1 (mean=21.7%)or anti-Lag3 (mean=18.5%)treatment (Figure 2B and 2C).We have previously identified ICOS and OX40 as markers of PD-1 response in CD4 + T cells(2), so we were interested to further understand their regulation in the context of CB.Analysis of ICOS and OX40 expression within responder cells (to normalise for differences in the frequency of T cell activation in vivo, Supplementary Figure 1) revealed a strikingly large increase in CD4 + T cell surface expression of ICOS and OX40 in CB-recipient mice compared to IgG or single agent-treated mice (Figure 2D).We generated a surface protein metric called surface TCR.strong based on the geometric mean of OX40 and ICOS MFI (Supplementary Figure 1), the transcripts of which we have previously reported to be enriched in melanoma patient responders to Nivolumab (2).Analysis of surface TCR.strong expression suggested that CB was exerting synergistic effects on the levels of OX40 and ICOS compared to individual treatments (Figure 2E).To further explore the effects of Lag3 and PD-L1 CB on responder and surface TCR.strong metrics, we performed statistical testing to determine whether any interaction (i.e.potential synergy) existed between the treatments (Figure 2F).The effect of combination therapy was additive for the responder population (parallel lines, p value for interaction=0.865,ns); however, for the surface TCR.strong metric a significant interaction was detected, highlighting a synergistic effect on ICOS and OX40 expression.As the analysis was performed on only responder populations, this effect is not driven by differential activation states in response to the different treatment regimes.These findings were also detected at a later time point (18hrs) post-secondary stimulation (Figure 2G).In summary, these data reveal that CB of Lag3 and PD-L1 drives additive increases in the proportion of CD4+ T cells that re-activate as well as synergistic increases in markers linked to strong TCR signalling in vivo.

Transcriptional analysis reveals Th17/Tfh cell bias and activation of an NFATbiased gene signature in response to Lag3 and PD-L1 blockade
To gain an unbiased overview of phenotypic changes in CD4 + T cells in response to CB, we repeated the experiments performed in Figure 2 but this time used FACS to isolate responder cell populations before performing 3' mRNA sequencing (Figure 3A and 3B).Isolating the responders alone (i.e.Timer+ cells) allowed us to normalise for different levels of activation between groups and to home-in on the qualitative changes induced by these treatments.Principal component analysis (PCA) showed that the treatments were largely separated by PC1, with CB-recipient mice the most distant from isotype and anti-Lag3 and anti-PD-L1-treated mice clustering close together (Figure 3C).Analysis of DEGs revealed that Lag3 and PD-L1 co-regulated genes associated with strong TCR signalling (Icos, Tnfrsf4, Il21, Tnfrsf9, Maf, Irf8( 2)) but the DEG showing the highest log fold-change was the chemokine receptor Ccr6 (Supplementary Table 1, Figure 3D).In addition, there were notable upregulated genes involved in glycolysis/metabolism (Aldoa, Hk2, Hif1a), NF-kB regulator gene Bcl3, and genes involved in ribosome biogenesis (Gnl3, Wdr43).Downregulated genes included the Th2-associated Gata3, as well as integrins Itga4, Itga6 and also genes involved in lymphoid tissue retention (Sell, S1pr1) (Figure 3E).As we have observed previously (2) in strongly signalling CD4 + T cells, genes encoding molecules involved in TCR signalling were also downregulated (Cd4, Rasgrp2).Having noticed a reduction in Gata3 (Th2 master transcription factor), we curated an analysis of genes associated with different T helper subsets (Figure 3F).Visualisation revealed a clear footprint for Tfh cells (notable differences for Il21, Icos, and Maf) and Th17-associated markers (although Il17a and Il17f transcripts were not detectable), including Ccr6 which was also elevated in anti-PD-L1 treated mice (Figure 3F).Th1-, Th2-and Treg-associated transcripts were less pronounced after CB treatment, albeit with clear variation in expression, reflecting low transcript abundance for many of these genes.
Given that Nr4a3 is a direct NFAT target (31), we wanted to further understand whether the DEGs identified showed any relationship to NFAT signalling pathways.A recent study determined that 1089 genes in T cells are directly regulated by NFAT and ERK pathways during the first 30 hrs(32) -a similar time frame to our in vivo studies.Through incorporation of inhibitors of NFAT or ERK pathways, Wither et al.
determined the relative dependence of genes on these signalling axes.We intersected the 189 genes that were upregulated in CD4 + T cell responders between CB-treated and isotype-treated mice with the 1089 identified in Wither et al.Thirtynine genes were found to overlap between the two data sets (Figure 3G).The expression of these genes was then visualised across the four treatment groups from RNA-seq data (Figure 3H), with additional annotation based on the dependence of ERK and/or NFAT pathways as defined in Wither et al (32).The vast majority of the genes (33 out of 39) were associated with co-operative ERK/NFAT signalling, but with an NFAT-biased signature, implying a stronger contribution to their regulation from the NFAT pathway compared to ERK.Included in these groups were Tnfrsf4 (OX40), Icos and Irf8 which exhibited co-operative regulation (with NFAT bias), highlighted by red arrows.We confirmed the binding of NFAT1 to the genes encoding Nr4a3, ICOS, IRF8 and OX40 (Supplementary Figure 2) in publicly available datasets from murine T cells (33).All four genes displayed sharp NFAT1 binding peaks that were evident 0-10kb upstream from the transcription start sites of each gene, highlighting a link to TCR-NFAT signalling in driving their sustained expression.Taken together, CB imparts a novel transcriptional signature on CD4 + T cells undergoing reactivation in vivo, exhibiting hallmarks of NFAT-biased TCR signalling, metabolic priming and a T follicular helper cell bias.

Combination blockade biases towards Tfh cell development and synergistically upregulates CCR6
Our transcriptional analysis showed that CB imparts a Tfh-like phenotype in CD4 + T cells.Tfh cells are known to possess a signature of genes activated downstream of the TCR (including metabolic, cell cycle and cytokine genes (34)) and are sustained by AP-1-independent NFAT signals( 35) (i.e.NFAT-biased signalling).Therefore, we sought to validate that PD-L1 and Lag3 enhanced Tfh cell differentiation in vivo during CD4 + T cell reactivation (Figure 4A).We employed our accelerated adaptive tolerance model and analysed CD4 + T cells for markers of Tfh cell development: PD-1, CXCR5 and ICOS (Figure 4B).As can be seen, CB drove enhanced differentiation of Tfh cells (Figure 4C) compared to all other treatment conditions, and these cells expressed significantly elevated levels of ICOS at both the bulk CD4 + T cell level (Figure 4D) as well as within the Tfh subset (Figure 4E) compared to all other treatment groups.
Since transcriptional analysis by RNA-seq provides only a 'snapshot-in-time' view, we wanted to analyse protein levels of the other notable markers identified previously.We performed an identical experiment as in Figure 2A and then evaluated protein intracellular levels of IRF8 (a DEG and part of the TCR.strong metric(2)) and IRF4 (previously linked in this model to strong TCR signalling(2)) (Figure 4F and 4G).Analysis revealed that CB increased levels of IRF8 compared to all three treatment groups, whilst for IRF4 CB increased the levels compared to control and anti-PD-L1 treated mice.In addition, we validated whether the upregulation of Ccr6 also occurred at the protein level.Analysis of surface CCR6 revealed a sharp, synergistic upregulation (more than 30% of cells now positive for this marker) in response to CB (Figure 4H and 4I).In sum, protein validation of the RNA-seq revealed that CB enhanced Tfh-like cell development and increased surface CCR6 protein levels.

PD-L1 and Lag3 co-operatively regulate TCR signal duration in CD4 + T cells
The maturation of Timer Blue proteins into Timer Red can give an estimation of the signal duration within an actively signalling T cell population (Figure 5A).This is because TCR signalling triggers new formation of Timer Blue proteins, which have a half-life of four hours, before maturation into the longer lived Red proteins from (halflife of 3-5 days).A short TCR signal will thus lead to an increase and then fall in Red proteins, whilst sustained signalling will lead to accumulation of mature Timer Red proteins.We hypothesised that the enhanced NFAT-biased signature observed is a result of CB sustaining TCR signalling for a longer period following reactivation in vivo and we therefore re-analysed the levels of Timer Red protein within "responder" cells from Figure 2F and 2G across the four treatment groups.Analysis at 12 hrs post-reactivation in the presence of CB showed a small but significant accumulation of Timer Red proteins compared to the three other groups, however a small increase in Timer Red was also observed in the anti-PD-L1 treated mice (Figure 5B).Analysis at the later 18 hr timepoint revealed that only CB exhibited a significant increase in Timer Red, which demonstrates that these cells have signalled for a longer period (Figure 5C).
We next wanted to test whether key markers identified in Figure 3 5E).ICOS expression showed a delayed response but rose prominently from 4 hrs to 12 hrs with a gradual increase up to 24 hrs (Figure 5E).IRF8 showed a similar pattern but its induction required longer periods of TCR signalling (minimum of 12 hrs), and gradually increased thereafter (Figure 5E).OX40 showed a time dependent increase from the initiation of TCR signalling but plateaued from 12 to 24 hrs (Figure 5E).Thus, CB results in an increased duration of TCR signalling, which accounts for the time-dependent and synergistic upregulation of ICOS, IRF8 and OX40.

Elevated tumour CCR6 expression associates with enhanced survival in melanoma patients on PD-1 pathway targeting immunotherapies
We have previously shown that ICOS, IRF8 and OX40 gene expression is linked to successful outcomes to PD-1 pathway immunotherapy in melanoma patients (2).
Given the stark increase in CCR6 in anti-PD-L1+anti-Lag3-treated mice we investigated whether CCR6 expression is clinically relevant in the context of PD-1 pathway immunotherapy through analysis of RNA-seq datasets of patient tumour biopsies taken before and after anti-PD-1.Analysis of the Riaz et al. (36) cohort revealed that pre-therapy, tumour CCR6 expression showed no significant association with clinical outcomes (Figure 6A).However, on-therapy tumour biopsies showed significantly higher CCR6 expression in patients that showed treatment benefit (Figure 6B).Furthermore stratifying the Riaz cohort by previous Ipilimumab status revealed that those who were ipilimumab naïve (i.e.only received anti-PD-1 with no prior anti-CTLA-4 treatment) had significantly increased overall survival when patients were stratified by CCR6 expression levels within their ontherapy tumour biopsies (Figure 6C-D).In addition CCR6 transcripts showed a significant positive correlation with CCL20, the ligand for CCR6, within the ontherapy tumour biopsies (Figure 6E).The findings regarding tumour CCR6 expression was replicated in a separate patient cohort of early during treatment biopsies of melanoma patients on PD-1/PD-L1 immunotherapy from the Gide et al. (37) cohort (Figure 6F).Combining the datasets from both the Gide and Riaz studies (where both datasets were classified as high or low based on each cohorts respective median expression and then pooled) revealed that high tumour CCR6 expression in on-therapy tumour biopsies significantly associated with increased overall survival (Figure 6G).

CCR6 upregulation in human effector memory CD4 + T cells on Nivolumab and Relatlimab combination therapy
A recent study provided insight into transcriptional changes occurring on therapy in peripheral blood T cells in patients receiving anti-PD-1 and anti-Lag3 combination therapy (specifically Relatlimab and Nivolumab (38)).We investigated whether transcriptional changes identified in our murine models were also reflected in this cohort of human patients on combination therapy.We selected DEGs in human CD4 Finally, given the link between CCR6 expression levels and immunotherapy outcomes, we investigated the migratory capacity of FACS-purified CD4 + T cells following in vivo CB using a migration assay (Figure 7D).In response to CCL20, CD4 + T cells from CB-treated mice showed significantly enhanced migration towards CCL20, highlighting that PD-1 incorporating immunotherapies can enhance CCR6dependent migration of CD4 + T cells.

Discussion
Our study demonstrates that the PD-L1 and Lag3 pathways exert non-redundant control on the duration of TCR signalling in CD4 + T cells.PD-1 itself can directly target the TCR (10) and CD28 (11) molecules, which allows PD-1 to act as a rheostat to control TCR signal strength (2).PD-1 contains tyrosine-based motifs (ITIM and ITSM (44)) that allow recruitment of SHP-2 phosphatases that mediate its inhibitory function (9).Lag3 on the other hand has proven more enigmatic in its function.Whilst its binding to MHC Class II is well established (28,29), Lag3 also has other reported ligands of relevance to T cell biology: galectin-3(45), fibrinogen-like protein 1 (FLP-1)( 46) and the TCR-CD3 complex (47).Whilst stable peptide MHC Class II appear more important than FLP-1 for T cell inhibition in vitro, the hierarchy of ligands in vivo remains unclear (48).Our findings show that Lag3 works co-operatively with the PD-1/PD-L1 pathway to regulate the T cell activation threshold of CD4 + T cells in vivo.
Mechanistically, co-blockade of PD-L1 and Lag3 results in an increased duration of TCR signalling, which drove synergistic increases in the TCR signal durationdependent receptors OX40, ICOS.
Interestingly, the effect of Lag3 blockade alone was generally weaker than the PD-L1 pathway in vivo, which Vignali and colleagues have also previously noted (47).In addition, Lag3 had minor effects on the primary CD4+ T cell response and its major effects were only exerted in the presence of PD-1 pathway blockade.These findings may be due to broader T cell PD-1 expression and the susceptibility of Lag3 to ADAM-mediated shedding (49).Indeed, ADAM-resistant forms of Lag3 confer resistance to anti-PD-1 immunotherapy in vivo (50), which is linked to reduced CD4 + T cell help for CD8 + T cells.It is possible that PD-1 blockade itself may augment the ability of Lag3 to interact with its potential ligands (e.g., MHC class II), which could result in an increased potency of Lag3 in our model, particularly as [4Y]-MBP peptide generates highly stable peptide-MHC II complexes.Given that in the immunised setting blockade of Lag3 resulted in much weaker effects than in the adaptive tolerance conditions, this may reflect the higher levels of Lag3 that are induced on the tolerised CD4 + Tg4 T cells.
The fact that PD-1 pathway appears more potent than Lag3 could also reflect hierarchical levels of T cell tolerance, whereby upon loss of the PD-1 pathway, Lag3 plays a much greater functional role.It is also likely that blockade of the PD-1 pathway may alter the dwell times of T cells on APCs in vivo.Tolerised islet antigenspecific T cells showed enhanced motility when scanning DCs in pancreatic draining lymph node (51).Blockade of PD-L1 or PD-1 led to reduced T cell motility and precipitation of autoimmune diabetes; however blockade of CTLA-4 had no effect on T cell motility(51) despite past reports that CTLA-4 may limit formation of stable APC:T cell conjugates (52).
A previous study demonstrated functional synergy of PD-1 and Lag3 in the control of anti-tumour immunity (53), and this was linked to increases in IFNγ-expressing cells amongst both CD4 + and CD8 + T cells (53).Unlike CD8 + T cells, CD4 + T cells have less clearly defined roles in tumour development, exerting anti-tumour effects through effector cytokines, cytotoxicity, and T cell help (reviewed in (54)).However, evidence for pro-tumour development comes from the association of CD4 + Foxp3 + Tregs with poor clinical outcomes (55).Nonetheless, expansion of Tfh cells in human tumours has been observed following anti-PD-1 immunotherapy (56).In solid tumours, Tfh cells likely have a beneficial role (57) and their tumour infiltration is positively correlated with clinical outcome in human breast cancer (58).
Mechanistically, Tfh cell-derived IL-21 was linked to enhancing CD8 + T cell function in mouse models of lung adenocarcinoma (59).
It is possible that the Tfh cell correlation with clinical outcomes in cancer may reflect neoantigen release which may promote their differentiation and execution of antitumour immune functions.However, in the context of immunotherapy, Tfh cell differentiation has been shown to be notably altered (57).A study assessing cancerirrelevant immune responses to seasonal flu vaccination describe an increased Tfh cell response in patients on anti-PD-1 therapy (60).Tfh cell signatures are positively correlated with immune related adverse events (irAEs) in humans (61).Interestingly, similar findings have been observed in murine studies using aged mice (18-24 months), which showed the development of IgG-mediated disease that was linked to an IL-21-producing Tfh-like subset (62).In the present study, combination treatment robustly drove enhanced Tfh cell development, but its most notable effect was the induction of higher ICOS expression.Given that ICOS is an essential costimulatory molecule for Tfh cell responses in vivo (63) it will be important to understand its wider potential effects on humoral immunity.
The potential manipulation of Tfh cell biology by checkpoint blockade comes with inherent risk.Genetic analysis of SNPs have linked a variant (rs17388568) associated with increased risk of ulcerative colitis(64) and type 1 diabetes(65) with an enhanced response to anti-PD-1 in metastatic melanoma patients (66).The risk variant maps to a genomic region containing IL2 and IL21, which are critical cytokines involved in regulation of CD4 + and Tfh cell responses.Future work should address the extent to which Tfh cell development is altered in patients on immune therapies targeting PD-1 and/ or Lag3 pathways and how these relate to irAEs.
Upregulation of CCR6 was a clear feature of CD4 + T cells restimulated in the presence of antibodies to PD-L1 and Lag3 -a finding also seen in analysis of singlecell RNAseq data of human CD4 + effector memory subsets in peripheral blood (38).
CCR6 has been identified as a marker of autoreactive but tolerogenic CD4 memory T cells in vivo (67).Interestingly, IL-21 has been shown to be the major factor that leads to CCR6 upregulation on T cells in culture (68).The CCL20/CCR6 axis has been heavily implicated in the metastasis of cancer cells(69) but CCR6 is also being exploited as a tumour-homing chemokine in the Chimeric Antigen Receptor (CAR) T cell field (70).Our findings that PD-1 and Lag3 pathway blockade can alter T cell migratory properties is therefore important and warrants further investigation.
In addition to CCR6, we also observed changes to the integrins Itga4 and Itga6 and downregulation of L-selection (encoded by Sell).Therefore, CB of Lag3 and PD-L1 promotes an effector/ migratory CD4+ T cell phenotype.Whilst CCR6 is associated with Th17 cells (71), it is also associated with antigen-experienced memory T cells in humans (72), and subsets of so-called "circulating" Tfh cells in humans (73).The CCR6:CCL20 axis in mice has been shown to be pivotal to Th17 cell-mediated autoimmune arthritis in mice (74).This axis has also been linked to the recruitment of T cells into inflammatory tissues (75).Therefore, upregulation of CCR6 may enhance the recruitment of T cells to inflamed sites.However, recent work addressing correlates of toxicity to immunotherapy has highlighted that pre-existing abundance of activated CD4 + memory T cells and TCR diversity are significantly associated with irAEs in melanoma (76).Recent data from the RELATIVITY-047 phase 2/3 clinical study, which demonstrated that Relatlimab (human IgG4-Lag3 blocking antibody) in combination with Nivolumab showed superior progression free survival (8).
Nonetheless, combination therapy was associated with an increased level of grade 3/4 treatment related adverse events (21.1 vs 11.1%) (22).It will be imperative to further explore potential mediators of these irAEs in combination therapy-treated patients in the future.Identification of potential biomarkers for treatment monitoring in this study may aid in further understanding the correlates of treatment efficacy as well as their relationship to irAEs.
A recent study suggested that the ligand-dependent suppression by Lag3 depended on its cytoplasmic RRFSALE motif (77).Intriguingly, this study also showed that blocking Lag3 can enhance T cell activation even in the absence of ligands for Lag3 in vitro (77).Given that PD-1 can exhibit tonic signalling which drives a basal level of suppression( 78), it will be interesting to determine whether the mere expression of Lag3 is also in itself sufficient to mediate some of its inhibitory functions in CD4 + T cells.
In summary, our study provides new insight into how the PD-1 and Lag3 pathways intersect to govern multiple aspects of CD4 + T cell activation in vivo and identifies CCR6 as a clinically meaningful readout for PD-1 incorporating immunotherapies.

In vitro restimulation of antigen adapted T cells (hybrid model)
For the hybrid tolerance model Tg4 Nr4a3-Tocky Tiger (Il10-GFP) mice were immunized through subcutaneous injection of 4 mg/kg [4Y] MBP peptide.24 h later mice were euthanised and spleens were removed and digested as described above.
Digested cells were washed once and 5x10 5

RNA-seq library preparation and analysis
RNA was extracted from lysates using the PureLink RNA Micro Scale Kit (Invitrogen) according to the manufacturer's instructions.5 ng of RNA was used for generation of sequencing libraries using the Quantseq 3' mRNA-seq Library Preparation kit (Lexogen).Briefly, library generation was commenced with oligodT priming containing the Illumina-specific Read 2 linker sequence.After first strand synthesis, RNA was degraded.Second strand synthesis was initiated by random priming and a DNA polymerase.Random primers contained the illumina-specific Read 1 linker sequence.Double stranded DNA was purified from the reaction using magnetic beads and libraries amplified (18 cycles) and sequences required for cluster generation and sample indexes were introduced.Libraries were normalized and pooled at a concentration of 4 nM for sequencing.Libraries were sequenced using the NextSeq 500 using a High 75 cycle flow cell.Cluster generation and sequencing was then performed and FASTQ files generated.FASTQ files were then processed by Lexogen as follows: Trimming was performed using Cutadapt version 1.18, followed by mapping using STAR version 2.6.1a to the GRCm38 genome.Reads counting as performed using FeatureCounts version 1.6.4.Uniquely mapped read counts conts in the .txtformat were used for further analysis using DESeq2 in R version 4.0 (83).A DESeq dataset was created from a matrix of raw read count data.
Data were filtered to remove genes with fewer than 30 reads across all samples.
Log2 fold change estimates were generated using the DESeq algorithm and shrinkage using the ashr algorithm (84) to estimate log2 fold changes (lfc).A lfc threshold of 0.2 was set and genes with an adjusted p value<0.05 and an s value<0.005were used to filter out genes with low expression or negligible gene expression changes.Normalized read counts were transformed using the regularised log (rlog) transformation.Heatmap analysis was performed on the rlog transformed data using the R package gplots.For KEGG pathway analysis clusterProfiler (85), DOSE (86) and biomaRt(87) packages were used.Venn diagrams were generated using the CRAN package VennDiagram.

Analysis of published ChIP-Seq data
Data was analysed as previously described (31).Briefly, processed bigwig data files deposited in GEO: GSE64409 (were downloaded and hosted in CyVerse Discovery Environment (https://de.cyverse.org/de)and then mapped against the mm9 genome using the UCSC genome browser.

Analysis of Riaz and Gide cohorts
FPKM for CCR6 or CCL20 were extracted for melanoma patients from supplementary files appended to GEO: GSE 91061 (36), or generated as described previously (2).FPKM were converted to TPM (88) through dividing each gene level FPKM by the sum of all FPKM in annotated genes within that sample.This figure was then multiplied by 1 e 6 then offset by 0.01 to avoid 0 values.Patient responses were characterised as: complete remission (CR), partial remission (PR), stable disease (SD) or progressive disease (PD) as per Riaz et al. (36) and Gide et al. (37).
For combined survival analysis, patients within each cohort were classified as High or low based on their own cohorts median CCR6 TPM expression and then pooled for Kaplan-Meier visualisation.

TCR signal duration termination by PP2 administration
Nr4a3-Tocky Tg4 Tiger mice were subcutaneously injected with 4 mg/kg [4Y] MBP peptide.Mice were culled 24 h later and spleens were dissociated and single cell suspensions of splenocytes were made using the digest method as described above.
Cells were washed once and cultured at 5x10 5

Immunization for Tfh cell responses
C57BL/6J mice were immunized with alum-precipitated (9% aluminum potassium sulfate (Sigma-Aldrich)) NP-OVA conjugate (NP-conjugated ovalbumin), alum NP-OVA.These were respectively prepared by mixing endotoxin-free NP-conjugated ova (kind gift from Prof Kai Toellner, University of Birmingham) with a 9% alum solution.Per mouse a total of 20 μg NP-OVA was mixed with same volume of 9% alum and pH was adjusted using NaOH and HCL.After washing, 20μg NP-ova/Alum was resuspended in a total volume of 20μl PBS and injected into the left footpad.24 h later mice were administrated via the intraperitoneal (i.p.) route with 0.5 mg anti-PD-L1 (clone MIH5), anti-Lag3 (clone C9B7W), combination therapy of their respective isotype control (pooled at 1:1 ratios).On day 8, mice were euthanised and the popliteal lymph node was harvested.LNs were forced through 70 µM filters, before cells were stained for flow cytometric analysis.

Analysis of human combination therapy datasets
Differentially expressed genes compared to month 3 or month 1 compared to baseline in CD4 Naïve, CD4 effector memory, and CD4 Treg clusters from human patients on anti-PD-1 and anti-Lag3 therapy were extracted from supplementary files from (38) and intersected with DEGs identified from Isotype and CB mice from Figure 3 (Supplementary Table 1).

Migration assays
To measure migration, FACS purified CD4 + T cells were cultured in a 5 μm polycarbonate 24-well transwell system (Corning).T cells were loaded into the top chamber at a density of 2.5 x 10 5 cells in 0.1 mL complete media, with the bottom chamber containing either 0 or 100 ng/mL CCL20 (BioLegend) in 0.6 mL complete media.After 4 hours, the contents of the bottom chamber were harvested and washed extensively with PBS (+ 2 mM EDTA and 2% FCS), followed by enumeration by flow cytometry.AccuCheck couning beads (Invitrogen) were used to calculate the absolute number of CD4 + T cells in each sample.Migration index was calculated by normalising cells in the lower chamber compared to the medium alone control from isotype-treated mice.

Statistical Analysis
For non-RNA-sequencing analysis, statistical analysis was performed on Prism 10 (GraphPad) software.Flow cytometry data were analyzed using FlowJo software were regulated by the duration of TCR signalling.Given the challenges of inhibiting TCR signalling in vivo, we established a new in vitro model to assess the relative contributions of the PD-L1 and Lag3 pathways controlling in vitro TCR signal strength in adaptively tolerised T cells (Supplementary Figure 3).Splenocytes from Tg4 Nr4a3-Tocky Il10-GFP mice were isolated 24 hrs after in vivo immunisation and cultured in the presence of IgG isotype control, anti-PD-L1, anti-Lag3 or combination in the presence of a dose titration of [4Y]-MBP (Supplementary Figure 3A).To demonstrate that this model can reveal manipulations into TCR signal strength, we evaluated ICOS and OX40 expression on CD4 + Nr4a3-Red + Nr4a3-Blue + T cells.Using 'surface' TCR.strong metric based on the geometric mean of surface ICOS and OX40 expression we could see that the model faithfully captured alterations in TCR signal strength (Supplementary Figure 3B) since CB of Lag3 and PD-L1 augmented TCR signal strength across a dynamic range of peptide concentrations.Next, to address how TCR signal duration influences expression of the key markers ICOS, OX40, IRF8 and Nr4a3 we cultured antigen-adapted Tg4 T cells in vitro in the presence of [4Y]-MBP and added the Src kinase inhibitor PP2 to terminate TCR signalling after 0, 4, 8, 12 or 24 hours (Figure 5D).Cells remained in culture for the full 24 hours before analysis of Nr4a3-Blue, ICOS, OX40 and IRF8 expression.As expected, Nr4a3-Blue expression continued to increase in parallel with the increase in TCR signal duration (Figure naive/central memory (CM), human CD4 effector memory (EM)/Th1 and human Treg and intersected these with DEGs identified between combination therapy treated murine CD4s and controls (Figure7A-C).Upregulated across all 4 conditions was the NF-kB regulator gene BCL3/Bcl3, which has also been linked to inhibiting Treg development and function(39).Interestingly, ICOS showed upregulation across human CD4 naïve/CM and EM/Th1 cells.In addition consistent with our observation in the Riaz(36) and Gide(37) cohorts, CCR6 was upregulated within human effector memory/Th1 subsets.In human Treg, the gene encoding OX40 (TNFRSF4) was also notably upregulated, along with RRS1 (ribosome biogenesis; a pathway noted in our murine RNAseq), ISG15 (involved in type I IFN signalling) and TENT5C (mRNA stabilisation and NFAT driven).Notable downregulated genes in all 4 subsets included GATA3 (Th2) and TRAF3IP3 (required for Treg maintenance and function(40)).Human CD4 Treg and human CD4 effector memory cells also showed downregulation of DGKA and SIT1 (involved in TCR signalling).Notable other genes downregulated in human CD4 T EM/Th1 group also included CORO1A (which has roles in regulating allograft tolerance ) CDC25B (cell cycle regulation) and ANXA6 (involved in IL-2 sensing(41)) and RAC2 (deficiency of which affects T cell activation(42)), and ITGA4 (linked in mice to Th1 cell homing)(43).These findings confirm that peripheral human CD4 + T cells accurately reflect the murine in vivo phenomena and highlight ICOS, CCR6 and OX40 as clinical relevant biomarkers for treatment monitoring.
cells per well on 96-well U-bottom plates (Corning) in presence of 0.1 μM [4Y]-MBP peptide in a final volume of 200 μl RPMI1640 + L-glutamine (GIBCO) containing 10% FCS and 1% penicillin/streptomycin (Life Technologies).Inhibitors were dissolved in DMSO.PP2 (Sigma, 10 μM) was added at different time points.Cells were incubated at 37°C and 5% CO2 and analysed at the indicated time points on flow cytometry.

Figure 2 :
Figure 2: Lag3 and PD-L1 pathways exert additive and synergistic effects on CD4 + T cell activation during secondary stimulation

Figure 3 :
Figure 3: Transcriptional analysis reveals Th17/Tfh cell bias and activation of