Th1 responses in vivo require cell-specific provision of OX40L dictated by environmental cues

The OX40-OX40L pathway provides crucial co-stimulatory signals for CD4 T cell responses, however the precise cellular interactions critical for OX40L provision in vivo and when these occur, remains unclear. Here, we demonstrate that provision of OX40L by dendritic cells (DCs), but not T cells, B cells nor group 3 innate lymphoid cells (ILC3s), is critical specifically for the effector Th1 response to an acute systemic infection with Listeria monocytogenes (Lm). OX40L expression by DCs is regulated by cross-talk with NK cells, with IFNγ signalling to the DC to enhance OX40L in a mechanism conserved in both mouse and human DCs. Strikingly, DC expression of OX40L is redundant in a chronic intestinal Th1 response and expression by ILC3s is necessary. Collectively these data reveal tissue specific compartmentalisation of the cellular provision of OX40L and define a mechanism controlling DC expression of OX40L in vivo. The OX40-OX40L axis is a crucial component of the costimulatory requirement of CD4 T cell responses. Here, the authors show context and cell type specific expression of OX40L for driving Th1 cell generation during acute and chronic models of infection.

P roductive T cell responses require interactions between costimulatory molecules in addition to signals conveyed through T cell receptor (TCR) engagement of peptide:major histocompatibility complex (MHC) complexes. Members of the tumour necrosis factor (TNF) receptor super family (TNFRSF) play critical roles in shaping T cell responses through sustaining proliferation, enhancing survival and directing cellular differentiation [1][2][3][4][5] . CD4 T cell responses are particularly reliant on OX40 6 , with OX40 −/− and OX40L −/− mice, alongside OX40 −/− TCR transgenic T cells, all implicating OX40 signals in effector memory responses [7][8][9] . Furthermore, a human patient lacking functional OX40 expression was unable to make rapid effector cytokines upon challenge 10 , entirely consistent with the CD4 T cell defects observed in OX40 −/− mice. Thus OX40 signals are thought to be critical in both generating effector CD4 T cell responses and the subsequent establishment of CD4 T cell memory. Since OX40 shares signalling pathways with its fellow TNFRSF member CD30, which is also induced upon T cell activation, there is also potential redundancy in the roles of these two molecules 2,[11][12][13] .
Despite some understanding of the importance of OX40: OX40L at the molecular level, the cellular interactions underpinning provision of this costimulatory signal remain unclear. T helper type 1 (Th1) and Th2 responses are both thought to be OX40 dependent [14][15][16][17][18] , while the generation of Th17 cells may be inhibited by OX40 signals 19 . Expression of OX40L has been described on different professional antigen (Ag)-presenting cells 20,21 alongside T cells 22 , innate lymphoid cells 23 and also non-haematopoietic populations 24,25 ; thus many different cellular interactions may underpin the phenotypes observed in OX40and OX40L-deficient mice.
The activation of naive T cells within secondary lymphoid tissue is initiated through T cell interactions with Ag-bearing dendritic cells (DC) within the T zone [26][27][28] . Thus DCs are obvious potential providers of costimulatory signals to activated T cells. However, after initial interactions with DCs, activated CD4 T cells migrate to the border between B and T zones and the interfollicular spaces, suggesting that some costimulatory ligands maybe provided by further cellular interactions that occur in distinct tissue microenvironments 29 . Cells isolated from secondary lymphoid tissue with the highest expression of OX40L were found to be lymphoid tissue-inducer cells (now considered a type of group 3 innate lymphoid cell (ILC3)) 23 and were located specifically within the interfollicular spaces of lymph nodes (LNs) and the functionally analogous bridging channels of the spleen 30,31 . Given that these ILC3s were also found to express high levels of CD30L 23,32 , they might be key providers of costimulatory signals to activated T cells moving through the interfollicular spaces. Such interactions would be consistent with evidence that ILC3s support splenic CD4 T cell responses 33 . Alternatively or in addition, activated T cells may encounter OX40L expression by B cells 14 or other T cells 22 . The question of whether specific cellular interactions are required after initial priming by DC remains an important mechanistic detail and answering this would facilitate efforts to manipulate CD4 T cell responses in vivo.
Here we use the reporting of effector cytokines by endogenous Ag-specific CD4 T cells to carefully dissect the provision of OX40L in vivo. Our data reveal that OX40:OX40L interactions are critical for the generation of functional Th1 effector cells in the response to acute systemic Listeria monocytogenes (Lm) infection, and using conditional OX40L knockout mice, we show that the expression of OX40L by DCs but not by T cells, B cells or ILC3s was required. Expression of OX40L by DCs is dependent upon crosstalk with natural killer (NK) cells that results in early interferon-γ (IFNγ) production directly signalling to the DCs to upregulate OX40L expression. In contrast, within a chronic intestinal Th1 response, DC OX40L expression is redundant and ILC3 expression of OX40L is required. Together, these studies define distinct cellular providers of OX40L in vivo, revealing that compartmentalisation of these signals, dependent on the nature of the response and the microenvironments, occurs to elicit optimal CD4 effector T cell responses.

Results
IFNγ-producing Th1 effector T cells are OX40 dependent. Studies in humans 10 and mice 8,34,35 indicate that effector CD4 T cell responses are highly OX40 dependent. To better understand the costimulatory requirements for CD4 effector cell function in vivo, mice deficient in both CD30 and OX40 (CD30 −/− × OX40 −/− ) were crossed with Great × Smart17A dual IFNγ and interleukin (IL)-17A reporter (GS) mice 36 . This approach enables reporting of cytokine expression without ex vivo manipulation. Furthermore, when combined with assessment of endogenous CD4 T cell responses using MHCII tetramers allows for the careful analysis of CD4 T cell function versus impacts on Agspecific CD4 T cell numbers. Infection with an attenuated Lm expressing the 2W1S peptide (Lm-2W1S) strain was used to model an acute intracellular bacterial infection that could be tracked with 2W1S-specific MHCII tetramers 37,38 . This wellcharacterised model provides a robust means to carefully dissect a robust Th1 response in vivo. To assess the impact of CD30 and OX40 deficiency on the effector CD4 T cell response, GS and GS CD30 −/− × OX40 −/− mice were assessed at 7 days post infection (dpi) with Lm-2W1S, revealing a substantial CD44 hi 2W1Sspecifc CD4 T cell population in both mouse strains, with the absence of CD30 and OX40 resulting in an approximate 2-fold reduction in total Ag-specific CD4 T cells (Fig. 1a, b). Expression of IFNγ, revealed by enhanced yellow fluorescent protein (eYFP) expression, was restricted to CXCR5 − 2W1S-specific T cells 38 , and notably, eYFP expression was almost completely absent in CXCR5 − 2W1S-specific T cells isolated from GS CD30 −/− × OX40 −/− mice (Fig. 1c, d). In contrast, the number of CXCR5 + follicular 2W1S-specific cells was not significantly changed ( Fig. 1e) consistent with previous data that the follicular T cell response to this attenuated Lm strain was not OX40 dependent 34 . Expression of the Th1 associated transcription factor T-bet was used to identify effector cells that might be unable to produce effector cytokines (Fig. 1f). Numbers of T-bet + CXCR5 − 2W1Sspecific T cells were reduced approximately 2-3-fold in the absence of CD30 and OX40 (Fig. 1g, h). This is compared to the tenfold loss in eYFP + CXCR5 − 2W1S-specific T cells indicating that the most substantial defect in the response is the ability of the cells to make effector cytokines. To determine whether the lack of IFNγ-producing Th1 cells at 7 dpi (the peak of the response) reflected an inability of these cells to survive after initial priming, the response was assessed at 4 dpi. Again, eYFP + 2W1S-specific CD4 T cells, but not CXCR5 + 2W1S-specific T cells, were heavily reduced in the absence of CD30 and OX40 (Fig. 1i-l) as observed at the peak of the response, suggesting that the effector Th1 cells were not being generated. Combined, these data reveal that, in the primary CD4 T cell response to Lm-2W1S infection, the number of Th1 effector CD4 T cells was reduced in the absence of CD30 and OX40, reflecting the early loss of cytokine-producing Th1 effector cells.
To determine the individual contributions of CD30 and OX40 to the effector T cell response, the 2W1S response at 7 dpi in OX40 −/− and CD30 −/− × OX40 −/− mice was compared (Fig. 2a). Enumeration of total 2W1S-specific CD4 T cells (Fig. 2b), 2W1Sspecific Th1 effectors (Fig. 2c, d) and 2W1S-specific CXCR5 + follicular cells (Fig. 2e) revealed that loss of OX40 signals accounted for the defects observed in CD30 −/− × OX40 −/− mice. Thus, for numbers of Ag-specific T cells at the peak of the response, OX40 expression was key. To determine whether CD30 played a role in the functional capacity of the Th1 effector population, GS mice with only single copies of either CD30 or OX40 (GS OX40 −/− CD30 −/+ and GS OX40 −/+ CD30 −/− ) were also generated. A single copy of OX40 in the absence of CD30 was sufficient for normal IFNγ (eYFP) production (Fig. 2f, g) and formation of a T-bet + effector CD4 T cell population (Fig. 2h, i), indicating that these aspects of the response were not CD30 dependent. Combined, these experiments demonstrate that signals through OX40, rather than through CD30, were critical for the primary Th1 response induced by Lm-2W1S infection. While OX40L is the only reported ligand for OX40, we confirmed that observations made in OX40 −/− mice were consistent with mice lacking OX40L by crossing OX40L f/f mice 39 with PGK cre mice (which express cre recombinase ubiquitously) to generate a total OX40L-deficient mouse model. As anticipated, b Enumeration of 2W1S-specific CD4 T cells. c Enumeration of eYFP + 2W1S-specific CD4 T cells. d Percentage of 2W1S-specific CD4 T cells expressing eYFP. e Enumeration of CXCR5 + 2W1S-specific CD4 T cells. f Expression of T-bet versus CXCR5 by 2W1S-specific CD4 T cells. g Enumeration of T-bet + 2W1S-specific CD4 T cells. h Percentage of 2W1S-specific CD4 T cells expressing T-bet. Data were pooled from 3 independent experiments (n = 11 mice per group). i Enumeration of 2W1S-specific CD4 T cells. j Enumeration of eYFP + 2W1S-specific CD4 T cells. k Percentage of 2W1S-specific CD4 T cells expressing eYFP. l Enumeration of CXCR5 + 2W1S-specific CD4 T cells. Data were pooled from 2 independent experiments (n = 7 mice per group). Values on flow cytometric plots represent percentages; bars on scatter plots represents the median. Statistical significance was tested by using an unpaired, nonparametric, Mann-Whitney two-tailed T test: *p ≤ 0.05, ***p ≤ 0.001, ****p ≤ 0.0001. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17293-3 ARTICLE NATURE COMMUNICATIONS | (2020) 11:3421 | https://doi.org/10.1038/s41467-020-17293-3 | www.nature.com/naturecommunications total OX40 −/− or PGK cre × OX40L f/f mice showed a comparable disruption of the effector Th1 response (Fig. 2j-l).
DC expression of OX40L is required in systemic Th1 response. The specific cellular providers of OX40L signals within secondary lymphoid tissue remain to be elucidated, with professional Agpresenting cells, T cells and ILC populations all possible sources in vivo. We hypothesised that in the response to Lm-2W1S there was a critical cellular provider of OX40L signals that underpinned the CD4 effector T cell response. To test this, OX40L f/f mice were crossed with multiple cell-specific cre strains to target DCs, ILC3s, T cells and B cells, since all of these have been reported to express OX40L 5 . Specific deletion of OX40L expression on the appropriate populations was confirmed through analysis of splenocytes ex vivo alongside fate-mapping of cre expression ( Supplementary Fig. 1). Given their expression of OX40L 23,32 , positioning in lymphoid tissue 31 and support for splenic CD4 T cell responses 33 , we first hypothesised that ILC3 expression of OX40L was required for functional Th1 responses. To test this, Rorc cre versus Rorc cre × OX40L f/f mice were infected with Lm-2W1S and assessed at 7 dpi, alongside OX40 −/− controls. Strikingly, no defect was observed in the 2W1S-specific response when ILC3 lacked OX40L (Fig. 3a-c), indicating that ILC3 provision of OX40L was not required for the generation of effector Th1 responses in the spleen. Given that the defect in the Th1 effector response was established by 4 dpi, we hypothesised that DCs were the critical source of OX40L signal in vivo. To first confirm that the 2W1S-specific response to Lm-2W1S was dependent upon interactions with DCs, MHCII expression by DCs was deleted using CD11c cre × H2Ab-1 f/f mice. As expected, in the absence of MHCII expression on CD11c + cells, the 2W1S-specific response was completely abrogated (Fig. 3d, e). To then test whether DC provision of OX40L was necessary for normal Th1 effector T cell responses, CD11c cre × OX40L f/f mice were infected with Lm-2W1S, alongside CD11c cre and CD30 −/− × OX40 −/− controls. Compared with CD11c cre controls, CD11c cre × OX40L f/f mice had significantly reduced numbers of 2W1S-specific CD4 T cells at 7 dpi (Fig. 3f, g) due to substantial loss of the T-bet + CXCR5 − subset ( Fig. 3h-j). Strikingly, the defects in CD11c cre × OX40L f/f mice were comparable to those observed in CD30 −/− × OX40 −/− controls, demonstrating that the expression of OX40L by DCs was absolutely required for the Th1 effector response and could account for the impaired Th1 effector response observed in the total absence of CD30 and OX40 signals. While these data clearly establish the importance of OX40L expression by DCs, we sought to further confirm that there was no absolute requirement for provision of OX40L by lymphocytes during the primary response, perhaps acting subsequent to interactions with DCs. Therefore, B cell-specific (Mb1 cre ) and T cell-specific (CD4 cre ) conditional OX40L-deficient mice were infected with Lm-2W1S and the 2W1S-specific CD4 T cell response was analysed. No significant differences were observed versus cre-only controls in either strain confirming that there was no requirement for B nor T cell provision of OX40L in vivo for the normal response to Lm-2W1S ( Supplementary Fig. 2). Together, these data reveal that, in the primary response to Lm-2W1S, provision of OX40L by a single cellular source, the CD11c + DCs, is required for the Th1 effector T cell response.
Early IFNγ regulates OX40L expression by DCs. Defining the cellular interaction required for provision of OX40L enabled the investigation into how expression of this key costimulatory pathway was regulated in vivo. The kinetics of DC expression of OX40L was assessed by infecting mice with Lm-2W1S and then analysing splenic DC expression of OX40L at different time points post infection. Although a basal level of OX40L expression by the splenic DCs was evident, Lm-2W1S infection resulted in the rapid upregulation of OX40L expression by 24 h, which then declined to basal levels once more by 72 h post infection ( Fig. 4a, b). The dynamics of OX40 expression relative to OX40L expression in the same spleen tissue was then assessed to better understand the relationship between maximal DC expression of the ligand and Ag-specific T cell expression of the receptor ( Supplementary Fig. 3). OX40 expression was evident on 2W1S-specific CD4 T cells at 48 h, where approximately 50% of the cells expressed OX40 and many also co-expressed CD25, described to be expressed by early Th1 effector T cells 38 . The proportion of OX40-expressing 2W1S-specific CD4 T cells peaked at 2 dpi and then rapidly decline such that by 4 dpi the vast majority of the responding CD4 T cells lacked any detectable expression of OX40 and DC expression of OX40L had returned to baseline. Thus the expression of both OX40 and OX40L expression was focused very early in the response and OX40 expression was limited to only a proportion of the responding 2W1S-specific CD4 T cells with a phenotype consistent with Th1 effector cells.
While OX40L expression can be upregulated in vitro by stimulation through CD40 21 , the rapid upregulation of OX40L expression observed after Lm-2W1S infection suggested that early signals from innate immune cells would precede CD40L expression by activated T cells. Activation of NK cells contributes to robust Th1 responses [40][41][42] , and more recently, IFNγ release by activated NK cells was revealed as the mechanism by which an adjuvanted vaccine response enhanced Th1 responses 43 . Furthermore, crosstalk between NK cells and DCs has been previously identified as a key mechanism through which both populations are fully activated 44 . To investigate early innate cell production of IFNγ in vivo, GS mice were infected with Lm-2W1S and assessed 24 h later. Gating on NK cells revealed substantial IFNγ expression compared to NK cells from uninfected GS control mice (Fig. 4c, d). Moreover, NK cells were the dominant source (approximately 70%) of IFNγ expression at this time, with T cells contributing most of the remaining signal (Fig. 4e, f). To ask whether this early IFNγ signal was required for upregulation of OX40L on DCs, wildtype (WT) and IFNγ −/− mice were infected with Lm-2W1S and DC expression of OX40L assessed 24 h later. Notably, the upregulation of OX40L expression by DCs isolated from IFNγ −/− mice was impeded compared to WT controls but was higher than the basal level observed in uninfected mice (Fig. 4g). These data indicate that early IFNγ was resulting in the increase in OX40L expression by DCs, but other signals might also contribute. In vitro, Th1 responses to heat-killed Lm required IL-12 45,46 , and neutralisation of IL-12 in vivo results in reduced resistance to Lm, which could be reversed by recombinant IFNγ 47 . To assess the requirement for IL-12, IL-12p35 −/− mice were infected with Lm-2W1S and assessed 24 h later for NK cell production of IFNγ and OX40L expression by splenic DCs. The To confirm that mice deficient in OX40L phenocopied mice deficient in OX40, expression of T-bet versus CXCR5 by 2W1S-specific CD44 hi CD4 T cells in PGK cre , PGK cre × OX40L f/f versus OX40 −/− mice was assessed. j Enumeration of 2W1S-specific CD4 T cells. k Enumeration of T-bet + 2W1S-specific CD4 T cells. l Percentage of 2W1Sspecific CD4 T cells expressing T-bet. Data were pooled from 2 independent experiments (n = 6 PGK Cre mice, n = 6 PGK Cre × OX40L f/f mice, n = 9 OX40 −/− mice). Values on flow cytometric plots represent percentages; bars on scatter plots represents the median. Statistical significance was tested by using an unpaired, non-parametric, Kruskal-Wallis one-way ANOVA with post hoc Dunn's test: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17293-3 ARTICLE NATURE COMMUNICATIONS | (2020) 11:3421 | https://doi.org/10.1038/s41467-020-17293-3 | www.nature.com/naturecommunications proportion of NK cells producing IFNγ (Fig. 4h, i) and the amount of IFNγ produced (Fig. 4j) were significantly reduced in IL-12p35 −/− mice and DC upregulation of OX40L was impaired (Fig. 4k, l), indicating that IL-12-mediated NK cell expression of IFNγ was linked to DC OX40L expression. We further confirmed that expression of IL-12 mRNA was equally upregulated by DCs from CD11c cre × OX40L f/f and OX40L f/f littermate controls after infection with Lm-2W1S ( Supplementary Fig. 4), indicating that IL-12 production by OX40L-deficient DCs was not impaired and the loss of Th1 effector cells not due to impaired polarisation during priming.
Since the activation of NK cells can occur subsequent to the rapid activation of invariant NK T (iNKT) cells and mediated through very early iNKT expression of IFNγ 48 , we assessed the requirement for iNKT cell activation using CD1d −/− mice. WT, CD1d −/− and IFNγ −/− mice were infected with Lm-2W1S, and NK cell IFNγ expression was assessed 24 h later. NK cells from CD1d −/− mice showed comparable IFNγ expression to NK cells  from WT controls (Fig. 4m, n). Furthermore, DC expression of OX40L was also not impaired in CD1d −/− mice (Fig. 4o). Combined, these data reveal that iNKT cells are not required for IFNγ production by NK cells in response to Lm-2W1S and highlight the importance of IFNγ in enhancing DC expression of OX40L.
IFNγ produced by NK cells signals directly to DCs. To investigate whether IFNγ could directly upregulate OX40L expression by DCs, splenocytes were cultured in vitro with recombinant IFNγ alongside anti-CD40 antibodies (Abs) as a positive control 21 . Culture with IFNγ induced robust upregulation of OX40L expression by DCs (Fig. 5a, b), suggesting that IFNγ might  Values on flow cytometric plots represent percentages; bars on scatter plots represents the median. Statistical significance was tested in b, d by using Kruskal-Wallis one-way ANOVA with post hoc Dunn's test and in e, g by using an unpaired, non-parametric, Mann-Whitney two-tailed T test: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001. directly stimulate OX40L expression via signals through the IFNγR. To test this in vivo, CD11c cre × IFNγR f/f mice were generated and infected with Lm-2W1S alongside CD11c cre and IFNγ −/− controls. DCs isolated from either CD11c cre × IFNγR f/f or total IFNγ −/− mice expressed significantly less OX40L than controls (Fig. 5c, d). Flow cytometric analysis of IFNγR expression by CD11c + DCs confirmed efficient deletion of this receptor in CD11c cre × IFNγR f/f mice (Fig. 5e). Finally, DCs from CD11c cre × IFNγR f/f mice cultured with recombinant IFNγ failed to upregulate OX40L to the level observed for WT controls (Fig. 5f, g). Combined, these data indicate a mechanism where DC-expressed IL-12 instructs NK production of IFNγ, which then directly signals back to the DC to enhance OX40L expression and orchestrate the generation of a robust Th1 effector T cell response.
Although NK cells were the largest population of IFNγexpressing cells, CD3 + IFNγ + cells were evident 24 h after infection with Lm-2W1S raising the possibility that multiple sources of IFNγ expression contributed to DC activation. To better understand the spatial distribution of IFNγ-producing cells in the spleen after Lm infection, IFNγ expression was localised in tissue sections 24 h post infection. Expression of IFNγ was detected within some white pulp areas of the spleen of infected mice ( Supplementary Fig. 5). Both IFNγ + NK cells and IFNγ + CD8 T cells were detected in approximately equal frequency in the proximity of DCs in the splenic white pulp (Fig. 6a, b). To specifically test in vivo whether NK cells were a critical source of IFNγ, mice were treated with either anti-NK1.1 Abs alone or in combination with anti-CD8 Abs, then infected with Lm-2W1S and analysed 24 h later. Depletion of the targeted cells and a loss IFNγ expression among this population was confirmed by flow cytometry (Supplementary Fig. 6), and the expected reduction in IFNγ expression was observed when splenocytes were analysed after ex vivo culture in the presence of brefeldin A (Fig. 6c-e). Importantly, depletion of NK cells resulted in the impaired upregulation of OX40L expression by DCs, comparable to that observed in IFNγ −/− controls (Fig. 6f), indicating that NK cells were a critical in vivo source of the cytokine. Given the proximity of IFNγ + CD8 T cells to DCs in the splenic white pulp of Lm-2W1S-infected mice, mice were also treated with a combination of anti-NK1.1 and anti-CD8 Abs; however, this did not significantly reduce the OX40L expression by DCs further.
While our data indicated an important role for IFNγ in regulating OX40L expression, the OX40L expression by DC when IFNγ signalling was disrupted remained higher than in uninfected mice, indicating that other signals likely contributed to the upregulation of this ligand. To investigate this further, we employed in vitro cultures of splenocytes to screen other cytokines that could boost the expression of OX40L by DCs. Splenocytes from WT and IFNγ −/− mice were compared in parallel to help determine potentially indirect effects via stimulating IFNγ expression in the cultures. Notably, basal expression of OX40L in IFNγ −/− splenocyte cultures was lower than in WT controls. Recombinant IFNγ robustly upregulated OX40L expression as observed previously, but both IL-18 and TNFα were identified as further cytokines that could enhance OX40L expression by DCs in this assay (Supplementary Fig. 7).
NK cell-derived IFNγ upregulates OX40L on human DCs. Having identified how DC expression of OX40L in response to Lm-2W1S infection was controlled in vivo, we sought to confirm that this pathway was conserved in human DCs. Human cDC2 were isolated to high purity by fluorescence-activated cell sorting (FACS) and then cultured alone or with autologous FACSisolated NK cells (Supplementary Fig. 8). Under these conditions, recombinant human IFNγ was sufficient to significantly enhance OX40L expression consistent with direct effects of this cytokine on human cDC2 (Fig. 7a-c). Activation of NK cells in co-culture with cDC2 through addition of IL-12 and IL-15 was also able to enhance OX40L expression to similar levels as the recombinant IFNγ. Thus the mechanism of direct IFNγ signalling to DCs resulting from crosstalk with NK cells would also appear to operate for human DCs.
DC expression of OX40L redundant in intestinal Th1 response. Our data have revealed that OX40L expression by DCs, licenced by early crosstalk with NK cells, is sufficient for Th1 cell expansion and function in a model of acute systemic bacterial infection. To ask whether this mechanism was specific to the splenic response observed with Lm-2W1S infection, we sought to test the requirement for DC OX40L under quite different experimental conditions. To this end, we established an oral Salmonella infection model using an attenuated Salmonella-2W1S strain 49 , thus enabling tracking of the same endogenous 2W1S-specific CD4 T cell response within the draining secondary lymphoid tissue (mesenteric LNs (mLNs)) and the non-lymphoid effector site (colon). Having established that OX40L expression by ILC3s was not required in the response to Lm-2W1S, the Salmonella infection model provided an opportunity to further test the potential role of ILC3s in supporting CD4 T cells at mucosal sites via their expression of OX40L 50 . Thus CD11c cre × OX40L f/f and Rorc cre × OX40L f/f mice, alongside littermate OX40L f/f only and OX40-deficient controls, were infected, and the 2W1S-specific response was assessed 7 days later in the colon (Fig. 8a). To aid assessment of cytokine production, the different OX40L conditional mice had been further crossed with Great × Smart17A dual IFNγ and IL-17A reporters. Enumeration of the total CD4 T cell response confirmed a significant reduction in the intestinal CD4 T cell compartment in mice lacking OX40 (Fig. 8b) 51 , which was not detected in the conditional OX40Ldeficient mice. Strikingly, the number of 2W1S-specific CD4 T cells, alongside the proportion and number of IFNγ + (eYFP + ), was specifically reduced in Rorc cre × OX40L f/f , but not in CD11c cre × OX40L f/f mice (Fig. 8c-e), indicating that, for the effector Th1 cells in the colon, OX40L expression by ILC3s rather than by DCs was required. Furthermore, assessment of the response in the mLN (Fig. 8f) again revealed that the 2W1S Th1 response was specifically reduced in Rorc cre × OX40L f/f mice ( Fig. 8g-i), suggesting that ILC3s might provide OX40L within the draining lymphoid tissue, either in addition or instead of provision within the colon. Regardless, the data clearly showed that, in the primary response to Salmonella-2W1S, the expression of OX40L by DCs was not required to sustain the effector Th1 response.
Collectively these data demonstrate a critical role for OX40L in supporting the effector Th1 response and that in vivo there are specific cellular providers of these signals dependent on the microenvironment and the response.

Discussion
Many cells are able to present Ag to T cells and potentially provide the costimulatory signals critical for optimal responses 52 . This suggests that CD4 T cells may interact with distinct Agpresenting cell populations during the initial stages of the immune response as they move through different microenvironments within secondary lymphoid tissue 29 . However, despite its relevance for understanding how CD4 T cell responses are regulated, such a model has not been tested in vivo. Here we have provided a detailed in vivo dissection of the cellular provision of OX40L for Th1 effector CD4 T cells contrasting an acute systemic response with a chronic intestinal infection. Our data reveal that, in an acute Th1 infection model, OX40L provision by DCs is critical with OX40L expression by T cells, B cells or ILC3s redundant for the Th1 effector T cell response. In contrast, DC OX40L expression was not needed for the Th1 effector cells in response to oral Salmonella infection and here OX40L expression by ILC3s, or potentially T cells, was absolutely required. Thus, through conditionally deleting OX40L on different immune cell populations, our data reveal that, in vivo, there are specific cellular interactions that are critical for normal OX40L provision and this is tissue and response specific.
Having identified that DC expression of OX40L was required for generating a robust Th1 response, we were then able to dissect the pathway through which expression of OX40L by the DC was regulated, thereby identifying that NK cell:DC crosstalk ensures optimal OX40L expression through early IFNγ production signalling back to the DC. That this mechanism was also evident for regulating human DC expression of OX40L suggests this is a broadly relevant pathway controlling the expression of this costimulatory molecule. However, it is evident that other proinflammatory cytokines can also enhance DC expression of OX40L and in vitro culture with recombinant IL-18 or TNFα enhanced OX40L expression by DCs and both of these cytokines are required for optimal clearance of virulent Lm infection 53,54 . That DC expression of OX40L was critical in the response to Lm-2W1S likely reflects the very acute nature of this attenuated infection, resulting in only a short window of OX40 expression by activated T cells and thus presumably limited time for different cellular interactions providing OX40L signals. Where high levels of Ag are maintained for longer, such as in more chronic infections, T cell expression of OX40 is prolonged and here other cellular interactions after initial DC priming become relevant.
In the response to Salmonella-2W1S, deletion of OX40L using the Rorc cre resulted in the loss of IFNγ-producing Th1 CD4 T cells in both the draining lymphoid tissue and the effector tissue (colon). While T cell provision of OX40L cannot be excluded, these data are consistent with ILC3 OX40L expression critical for intestinal T cell responses, significantly developing previous studies using immunodeficient mice 50,55 where ILC frequencies are grossly perturbed 31 . The lack of a role for ILC3s in the splenic response to Lm-2W1S may reflect the acute response or simply perhaps the very low frequency of ILC3 in the spleen and the lack of clear clusters of ILC3, as seen in the intestine and their draining LNs 30 . Within the intestinal tract, it is evident that ILC3s restrict responses to commensal bacteria not only at the level of effector T cell responses 56,57 but also through T follicular helper cells and IgA production 58 . Thus ILC3s promote some T cell responses, while limiting others, indicating a critical regulatory role in intestinal immunity, potentially presenting an attractive cellular target for therapeutic manipulation 59 . Exactly where ILC3s interact with activated T cells remains unanswered; however, the impaired 2W1S-specific response observed in both the mLN and colon suggests a role in the draining lymphoid tissue, and the interfollicular spaces of the mLN may be the critical microenvironment fostering such crosstalk 31 . The 2W1S-specific CD4 T cell response to Lm-2W1S has been characterised in detail 38 and contains both T-bet-dependent Th1 effector cells and Bcl-6-dependent follicular T cells in roughly equal proportions. How the individual cells within this polyclonal response are pushed down one differentiation pathway or another is poorly understood but at least partially reflects differences in the strength of TCR signalling 60 . Our data suggest that only about half the activated CD4 T cells responding to Lm-2W1S ever express OX40 and differentiate into Th1 effector cells through upregulation of Blimp-1 and T-bet, alongside repression of Bcl-6 35,61 . Within the chronic Salmonella model, deletion of OX40L on ILC3 and T cells again resulted in a significant loss in the number Th1 effector cells; however, those cells that survived still expressed IFNγ. Thus signals through OX40 support Th1 effector cell expansion and survival rather than CD4 T cell differentiation 9 . The central finding of this manuscript is that, depending on the nature of the response, provision of OX40L signals is achieved through quite distinct interactions, with critical contributions by DCs and ILC3 demonstrated in vivo.
Collectively our data reveal that, despite the ability of many cell types to express costimulatory ligands for T cells, in vivo, there are specific cellular interactions that are needed for provision of these signals. Understanding these regulatory checkpoints and how costimulatory molecule expression is controlled in these situations will present further opportunities to better manipulate T cell responses for therapeutic benefit.
Lm-2W1S infection. Mice were infected with 10 7 actA-deficient Lm-expressing OVA-2W1S (Lm-2W1S, kind gift from Dr. M. Jenkins) through intravenous injection in the tail vein. Bacteria were pre-cultured in 10 ml of LB broth containing 20 μg/ml chloramphenicol from a single colony and grown overnight at 37°C at 200-250 rpm. Next day, 1 ml of Lm-2W1S was seeded into 200 ml of LB broth containing 20 μg/ml chloramphenicol and grown at 37°C at 200-250 rpm until OD600 of 0.6-0.7 was reached. The liquid bacterial cultures were then centrifuged for 20 min at 4°C, 4000 rpm. The supernatant was discarded and pellets were resuspended in LB broth with 15% glycerol. Glycerol stocks of Lm-2W1S were aliquoted into 1 ml cryovials and stored at −80°C. Three randomly chosen samples were used to assess the viability of bacterial stocks by plating serial dilution of bacteria on agar containing 20 μg/ml chloramphenicol. Prior administration, the 1 ml Lm-2W1S was centrifuged and washed with sterile phosphate-buffered saline (PBS) before resuspending in adequate volume of sterile PBS to obtain 10 7 bacteria per 200 μl. Glycerol stocks were prepared using the same method as described in previous section. Streptomycin was used as a selection antibiotic. The viability of the stocks was tested by randomly choosing 3 samples and plating the bacteria on MacConkey agar (Sigma-Aldrich) containing 100 μg/ml of streptomycin (Sigma-Aldrich).
In vivo Ab administration. The depletion of NK cells and CD8 T cells was achieved by intraperitoneal (i.p.) administration of a dose of 250 μg anti-NK1.1 and 400 μg anti-CD8 in PBS at 48 h prior infection with 10 7 Lm-2W1S. Abs were provided by AstraZeneca.   Statistical analysis. Data were analysed using GraphPad Prism (version 8). Nonparametric Mann-Whitney, two tailed or non-parametric Kruskal-Wallis one-way analysis of variance (ANOVA) with post hoc Dunn's tests or, following the successful normality tests, ordinary one-way ANOVA with post hoc Tukey's tests were c Enumeration of 2W1S-specific CD4 T cells in colon. d Enumeration of eYFP + 2W1S-specific CD4 T cells. e Percentage of 2W1S-specific CD4 T cells expressing eYFP. f Flow cytometric plots showing 2W1S-specific CD44 hi CD4 T cell response and the eYFP expression in the mesenteric LN of GS, GS × Rorc cre × OX40L f/f , GS × CD11c cre × OX40L f/f and GS × OX40 −/− × CD30 −/+ mice at D7 post infection. g Enumeration of 2W1S-specific CD4 T cells in the mesenteric LN. h Enumeration of eYFP + 2W1S-specific CD4 T cells. i Percentage of 2W1S-specific CD4 T cells expressing eYFP. Data were pooled from 3 independent experiments (n = 8 GS mice, n = 15 GS × Rorc cre × OX40L f/f mice, n = 11 GS × CD11c cre × OX40L f/f mice, n = 6 GS × OX40 −/− × CD30 −/+ mice). Values on flow cytometric plots represent percentages; bars on scatter plots represents the median. Statistical significance was tested by using Kruskal-Wallis one-way ANOVA with post hoc Dunn's test: *p ≤ 0.05, **p ≤ 0.01. used to determine significance, which was set at p ≤ 0.05. Median values were calculated and used in all analyses unless stated.