Regulatory T cells control the dynamic and site-specific polarization of total CD4 T cells following Salmonella infection

FoxP3+ regulatory T cells (Tregs) control inflammation and maintain mucosal homeostasis, but their functions during infection are poorly understood. Th1, Th2, and Th17 cells can be identified by master transcription factors (TFs) T-bet, GATA3, and RORγT; Tregs also express these TFs. While T-bet+ Tregs can selectively suppress Th1 cells, it is unclear whether distinct Treg populations can alter Th bias. To address this, we used Salmonella enterica serotype Typhimurium to induce nonlethal colitis. Following infection, we observed an early colonic Th17 response within total CD4 T cells, followed by a Th1 bias. The early Th17 response, which contains both Salmonella-specific and non-Salmonella-specific cells, parallels an increase in T-bet+ Tregs. Later, Th1 cells and RORγT+ Tregs dominate. This reciprocal dynamic may indicate that Tregs selectively suppress Th cells, shaping the immune response. Treg depletion 1–2 days post-infection shifted the early Th17 response to a Th1 bias; however, Treg depletion 6–7 days post-infection abrogated the Th1 bias. Thus, Tregs are necessary for the early Th17 response, and for a maximal Th1 response later. These data show that Tregs shape the overall tissue CD4 T cell response and highlight the potential for subpopulations of Tregs to be used in targeted therapeutic approaches.


INTRODUCTION
Salmonella enterica serovar Typhimurium (S. Tm) infects a wide range of animals and is a common non-typhoidal Salmonella serotype in humans. To study intestinal S. Tm infection, a streptomycin pretreatment model has been developed, creating a niche for S. Tm to colonize. 1,2 This facilitates consistent intestinal infection and colitis. 3,4 Because the use of virulent S. Tm strains in susceptible mice causes high mortality, we here combine streptomycin pretreatment with an attenuated S. Tm strain that induces consistent nonlethal colitis. This enabled analysis of the CD4 T cell response in the colonic lamina propria and the draining lymph nodes.
CD4 T helper (Th) cells drive intestinal immune responses and can express master transcription factors (TFs). T-bet, GATA3, and RORγT are used to identify Th1, Th2, and Th17 cells, respectively. CD4 T cells are critical for both clearance of S. Tm infection and protection from reinfection. 5,6 In particular, Th1 cells are required for S. Tm clearance and protective immunity in systemic infection. 5,[7][8][9] In addition to Th1 cells, an early Th17 response has been identified in the intestines, which is important for limiting epithelial damage and systemic dissemination. [10][11][12] Thus, an effective immune response to S. Tm involves both Th17 and Th1 cells. It is unclear however, how the balance between these responses is controlled.
FoxP3 + regulatory T cells (Tregs) regulate conventional T cells (Tconvs) and control the CD4 T cell response during infection. [13][14][15] Tregs can differentiate in the thymus (tTregs) or peripheral tissue (pTregs). While both tTregs and pTregs are found in the intestines, [16][17][18] their contribution to immune responses to infection is poorly characterized. In addition to FoxP3, Tregs can express Th master TFs T-bet, RORγT, or GATA3. [19][20][21][22] The role of these Treg TFs is the subject of ongoing research and accumulating evidence suggests that Tregs expressing Th TFs may selectively suppress their respective Th subset. 21,23,24 T-bet + Tregs can selectively target Th1 cells, but it is unclear whether Tregs expressing other TFs have similarly selective functions. 21,24 It is also unclear whether selective suppression can alter the balance of Th cell responses during infection.
To investigate the role of Tregs in controlling the CD4 T cell response, we employed a model of S. Tm colitis with an attenuated strain allowing identification of an S. Tm-specific T cell clonotype. Characterization of Tconvs and Tregs highlighted a dynamic colonic T cell response within total T cells at the infection site. This is dominated initially by Th17 cells, followed by a sustained Th1 response. Interestingly, we identified a reciprocal dynamic between Tconvs and Tregs expressing the same TFs. This reciprocity indicated that subpopulations of Tregs may selectively suppress Th subsets. Treg depletion experiments were therefore carried out using FoxP3 DTR mice. 25 Results show that Tregs are necessary for the early Th17 response and contribute to the later Th1 bias. Thus, Tregs not only inhibit CD4 T cells, but shape the T cell response in a fine-tuned, tissue-specific manner.

RESULTS
Salmonella infection with BRD509-2W1S causes nonlethal colitis with transient weight loss and increased numbers of total and 2W1S-specific CD4 T cells To investigate the CD4 T cell response to S. Tm in the intestines, attenuated strain BRD509-2W1S was orally administered to C57BL/ 6 mice 24 h following streptomycin treatment. Infection caused nonlethal colitis and allowed identification of S. Tm-specific T cells. Representative images of large intestines, including the cecum and colon, are shown from infected (top) and uninfected (bottom) animals 6 days post-infection (p.i.) (Fig. 1a). Infected tissues show colonic shortening, edema, empty and shriveled caeca, and an enlarged cecal patch (Fig. 1a, inset). Despite physical signs of colitis, S. Tm-infected mice lost <5% of initial weight, which recovered to levels of mock (PBS)-infected controls by 14 days p.i. (Fig. 1b). To assess bacterial dissemination, intestinal and lymphoid tissues were processed and plated. At day 6 p.i. S. Tm CFUs were recovered from the colon, mesenteric lymph nodes (MLN), Peyer's patches (PP), and small intestine (SI); significantly more CFUs were recovered from the colon and MLN than the SI (Fig. 1c).
Next, changes in total cell numbers, CD4 T cells and FoxP3 + Tregs were enumerated in the colonic lamina propria (LP) (Fig. 1d) and MLN (Fig. 1e) at days 6, 30, and 60 p.i. There was a significant increase in the number of total cells, CD4 T cells and Tregs at day 6 p.i., and total cell numbers remained elevated until 60 days p.i. (Fig. 1d). The number of total MLN cells also increased 6 days p.i., but this increase was resolved by day 30 p.i. Unlike in the colon, there was no significant increase in the number of MLN CD4 T cells or Tregs (Fig. 1e).
Infection with BRD509-2W1S allows identification of S. Tmspecific T cells with 2W1S:I-A b tetramers ( Supplementary Fig. S1). In the colon and cecum, 2W1S-specific cells comprise 5-10% of total CD4 T cells (Fig. 1f, Supplementary Fig. S2), but are less frequent in lymphoid tissues including MLNs (Fig. 1g, Supplementary Fig. S2). Tracking 2W1S-specific CD4 T cells over time showed these cells were detectable until at least 90 days p.i. (Fig. 1f, g). The large number of Tconvs specific for 2W1S reflects the immunodominance of the peptide, as previously reported. [25][26][27][28] The higher number of 2W1S-specific T cells in the intestinal LP than the MLNs is consistent with the hypothesis that these cells are accumulating at barrier sites to protect against invasion from the intestinal lumen, where the highest bacterial burden is present (Supplementary Fig. S4a).
The MLN CD4 T cell response to Salmonella is constrained to colon-draining MLNs The higher bacterial burden and number of 2W1S-specific T cells observed in the large intestine compared with the SI (Fig. 1c, Supplementary Fig. S2b), led to the hypothesis that the MLN CD4 T cell response was focused in the colon and cecum draining MLNs (cMLNs). This hypothesis is consistent with previous work showing that different intestinal sites drain to specific MLNs 29,30 ( Supplementary Fig. S3a). With the aim of improving sensitivity of detecting S. Tm responses in the MLNs, a comparison of CD4 T cells in the cMLNs and sMLNs was conducted. Results show that cMLNs but not sMLNs contain an increased number of total cells, CD4 T cells, CXCR3 + CD4 T cells, and 2W1S-specific T cells following S. Tm infection ( Supplementary Fig. S3b, c). Together, these data show that the CD4 T cell response to S. Tm in the MLN is restricted to the colon and cecum-draining lymph nodes.
Characterization of 2W1S-specific T cells To investigate the CD4 T cell response to S. Tm infection we initially examined the 2W1S-specific T cell response, and observed that 2W1S-specific T cells homogenously display a Th1 phenotype ( Supplementary Fig. S4), consistent with previous reports. 28,31,32 As such, 2W1S-specific cells expressed T-bet but not RORγT or FoxP3, up to and including 60 days p.i. (Supplementary Fig. S4a, b). Cytokine profiling of 2W1S-specific cells confirms a Th1 phenotype, with cells expressing IFNγ but not IL17A ( Supplementary  Fig. S4c, d). It has been previously shown that S. Tm-specific CD4 T cells comprise cells that recognize a wide range of peptides, and both S. Tm-specific and non-S. Tm-specific "bystander" cells play an important role in Salmonella clearance and protection from reinfection. 31,[33][34][35] Because of the phenotypic homogeneity of 2W1S-specific cells and the importance of a heterogenous CD4 T cell response to S. Tm, we next characterized changes in the total CD4 T cell pool.
S. Tm infection induces an early colonic Th17 response followed by a sustained Th1 response In contrast to 2W1S-specific cells, total CD4 T cells comprise heterogenous and dynamic populations of Th1 cells, Th17 cells, and Tregs following S. Tm infection. To characterize bulk CD4 T cells, that includes a mixture of CD4 T cell responding to the bacteria, potential bystander cells, and cells not involved in the response, we analyzed activated Tconvs by transcription factor (TF) and cytokine expression. Tconvs were identified as CD4 T cells (gated as described in Fig. 1d) that are CD44 hi and FoxP3 - (Fig. 2a, left). Tconvs were characterized by expression of T-bet, RORγT, and GATA3 to identify putative Th1, Th17, and Th2 cells respectively. 36 The number and proportion of cMLN ( Fig. 2b) and colonic LP ( Fig. 2c) CD44 hi Tconvs are plotted from mock (PBS)-infected mice and S. Tm-infected mice examined at 6, 30, or 60 days p.i. In the cMLN, there are increased numbers and proportions of Tconvs that are T-bet + , RORγT + , and T-bet + RORγT + cells at 6 days p.i. (Fig. 2b). Numbers of Tconv subsets return to baseline levels by day 30 p.i., except T-bet + cells, which remain elevated until 60 days p.i. In the colonic LP there is an increased number of Tconvs expressing T-bet, RORγT, or T-bet and RORγT 6 days p.i (Fig. 2c, left). While T-bet + , RORγT + , and T-bet + RORγT + cells increase in number, the increase in RORγT + cells is greater, leading to an increased proportion of RORγT + cells, and a decreased proportion of T-bet + and GATA3 + cells at day 6 ( Fig. 2c, right). This increase in RORγT + Tconvs at day 6 p.i. is transient however, and by day 30 a strong bias toward T-bet + Tconvs is observed. In summary, there is an initial colonic RORγT + response which shifts to a long-lasting T-bet + response. This dynamic CD4 T cell response is observed in the colonic mucosa but not in the cMLN.
Reciprocal dynamics between colonic Tconvs and Tregs expressing the same TFs Following characterization of Tconvs by TF, FoxP3 + Tregs were assessed for expression of Th master TFs T-bet, RORγT, and GATA3 ( Fig. 3a, b). The number and proportion of colon Tregs expressing these TFs were plotted from mock (PBS)-infected mice or S. Tminfected mice examined at 6, 30, or 60 days p.i. (Fig. 3b). There was an increased number of T-bet + and T-bet + RORγT + Tregs at 6 days p.i., but no increase of RORγT + or GATA3 + Tregs. Changes in absolute numbers correspond with an increased proportion of Tbet + Tregs and a reduced proportion of RORγT + Tregs at day 6 p.i. While these proportional changes are partially resolved by day 60 p.i., the percentage of Tregs that are T-bet + remain elevated while the proportion of RORγT + Tregs remain lower than mock-infected controls (Fig. 3b, right).
Comparing the proportion of colonic Tconvs and Tregs expressing T-bet or RORγT reveals a reciprocal dynamic between populations expressing the same TF. The early decrease in the proportion of T-bet + Tconvs (Fig. 2c, right) occurs concurrently with an increased proportion of T-bet + Tregs (Fig. 3b, right). Conversely, the transient increase in the proportion of RORγT + Tconvs corresponds with transient decrease in the proportion of RORγT + Tregs.
To further investigate this reciprocity, a more detailed timecourse was carried out during the first 30 days p.i. In Fig. 3c . In all charts, mean ± standard error of the mean (SEM) are plotted. Statistical significance between weights at specific time points are calculated by multiple Student's t tests. Statistical differences between all other groups are calculated by one-way ANOVA with Tukey's test. ns not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.
proportions of Tregs that are T-bet + or RORγT + (black lines) are overlaid with the proportion of Tconvs (gray lines) that express the same TFs. These plots highlight a reciprocal dynamic in the colon but not the cMLN. Furthermore, these data show that day 6 p.i. is the peak of the colonic RORγT + Tconv response and that the Tbet + Tconv bias is re-established by day 11 p.i (Fig. 3c). In summary, the early RORγT + Tconv response in the colon coincides with an increased proportion of T-bet + Tregs, and the later T-bet Tconv bias coincides with an increased proportion of RORγT + Tregs.

Effect of bacterial burden on T cell responses following S. Tm infection
To understand how the presence of S. Tm affects changes in the T cell response, we treated mice with antibiotics to clear the infection. S. Tm is known to cause persistent infection in mice and bacterial clearance requires prolonged treatment. 28, 39 We employed a previously described enrofloxacin treatment from day 5 to day 29 p.i. 33 In untreated mice, S. Tm CFUs could be recovered in luminal contents and the spleen at day 30 p.i.; following treatment, CFUs could not be recovered from any sites or tissue ( Supplementary Fig. S5a). Treatment reduced the proportion and number of tetramer + T cells at day 30, although the number of 2W1S-specific cells in treated animals remained above levels observed at day 6 ( Supplementary Fig. S5b, c). Enrofloxacin treatment did not change the T-bet + phenotype of 2W1S-specific cells ( Supplementary Fig. S5d), but reduced the proportion of Tbet + Tconv and RORgT + Tregs at day 30 p.i. This indicates that S. Tm persistence contributes to the Th1 response, but that the reciprocal dynamic between Tconvs and Tregs is maintained even when bacteria are cleared ( Supplementary  Fig. S5e, f).
Helios hi Tregs can co-express T-bet + and GATA3 + but not RORγT The reciprocal dynamic between Tregs and Tconvs raises questions about the ontogeny and stability of Tregs that express T-bet or RORγT. Helios expression has been proposed to differentiate tTregs from pTregs, 40 but this is controversial as Helios expression has also been reported in pTregs. 41
Finally, FoxP3 + Tregs were gated as Helios hi (dark purple), Helios int (pink), and Helios lo (lilac) populations (Fig. 5a, right). Overlaying these populations reveals that Helios hi Tregs overlap one lobe of the Treg cluster, while Helios lo/int Tregs overlay another (Fig. 5b, c, right). RORγT + Tregs are confined to Helios lo/int cells, GATA3 + Tregs are primarily Helios hi cells, and Tbet + Tregs include cells with a range of Helios expression levels. The proportion of Helios lo Tregs is reduced by infection but RORyT + Tregs remain Helios lo/int cells (Fig. 5b, c). Therefore, while S. Tm infection alters the proportion of different Treg populations, the co-expression of Th TFs with Helios remains stable.
The nonoverlapping expression of Helios and RORγT is consistent with previous work suggesting Helios is a marker of tTregs 40 and RORγT is a marker of pTregs. 22,44 Defining cell function or ontogeny by marker expression is fraught with  Treg depletion alters the balance of Th1 and Th17 responses following S. Tm infection The reciprocal dynamic between Tregs and Tconvs during S. Tm infection (Fig. 3) is consistent with the hypothesis that Tregs expressing Th master TFs selectively target Th subsets and skew the Th response. To determine whether Tregs shape the dynamic Th response to S. Tm, Treg depletion experiments were conducted using FoxP3 DTR "DEpletion of REGulatory T cell" (DEREG) mice, which allow ablation of FoxP3 + Tregs following treatment with diphtheria toxin (DT). 46 Tregs were efficiently depleted from the peripheral blood and colon after 2 DT treatments on consecutive days. Five days post-treatment there were~67% fewer Tregs in the blood, and~90% fewer Tregs in the colon than in DT-treated littermate controls ( Supplementary Fig. S6a, b). Similar to C57BL/6 mice, Treg-replete DEREG mice exhibit an early Th17 response and later Th1 bias following S. Tm infection, and T-bet + and RORgT + Tregs show the same reciprocal dynamic with Tconvs in both strains ( Supplementary Fig. S6d).
In the first depletion experiments, DEREG mice were DT-treated day 1 and 2 p.i., with an endpoint at day 6 p.i. (Fig. 6a). There was no significant difference in weight change between DEREG mice and DTR -(Treg replete) littermates ( Supplementary Fig. S7a) nor in S. Tm CFUs recovered 6 days p.i. (Supplementary Fig. S7c). The number of total colon cells remained unchanged (Fig. 6b) but both the number of CD4 T cells (Supplementary Fig. S7d) and CD44 hi Tconvs (Fig. 6c) increased following Treg depletion. Although the number of tetramer + cells trended higher, differences were insignificant (Supplementary Fig. S7e). The increase in CD4 T cells was confined to T-bet + Th1 cells and Tbet + RORγT + cells, while the number of RORγT + Th17 cells and/or IL17A by Tconvs that are T-bet + , RORγT + , or T-bet + RORγT + are depicted in donut charts. Means ± SEM are plotted. Data points represent individual animals (n = 3-4) from a representative example of two independent experiments. Statistical significance calculated by one-way ANOVA with Holm-Šídák test (b); regression p values calculated by Pearson's correlation coefficient (c). ns not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. remained unchanged (Fig. 6d). The increased number of Th1 cells underlies a decreased proportion of Th17 cells and an increased Th1 bias (Fig. 6d, Supplementary Fig. S7b). The decreased proportion of Th17 cells and increased proportion of Th1 cells following early Treg depletion is robust, also being observed when data from three independent experiments are pooled (Supplementary Fig. S7b). These pooled data show a colonic Th17 bias in Treg-replete DTRmice, while there is a Th1 bias in Treg-depleted DEREG littermates ( Supplementary Fig. S7b, right). These data demonstrate that the Th17 bias at 6 days p.i. (Fig. 2b) is dependent on Tregs, which have an increased proportion of T-bet + cells at this timepoint (Fig. 3c, Supplementary Fig. S6d).
Next, Treg depletion experiments were conducted to determine if Tregs were also required to drive the Th1 bias that becomes prominent by day 11 p.i. (Fig. 2b). DT was administered day 6-7 p. i. and the endpoint was day 11 p.i. (Fig. 6e). As described following earlier DT treatment, later Treg depletion did not cause a significant difference in weight change ( Supplementary Fig. S7f), the number of S. Tm CFU (Supplementary Fig. S7h) nor the number of total colon cells (Fig. 6f) between DEREG mice and DTRlittermates. There was an increase in the number of CD4 T cells ( Supplementary Fig. S7i), CD44 hi Tconvs (Fig. 6g) and tetramer + 2W1S-specific T cells (Supplementary Fig. S7j). There was also an increased number of both colonic Th17 and Th1 cells (Fig. 6h, left) but the increase in Th17 cells was greater, reflecting an increased proportion of Th17 cells at day 11 p.i. (Fig. 6h, right). The increased proportion of Th17 cells following Treg depletion is also observed when data from three independent experiments are pooled ( Supplementary Fig. S7g). These pooled data show the Th1 bias in Treg-replete mice at day 11 p.i. is abrogated following Treg depletion ( Supplementary Fig. S7b, right).
To confirm that these changes in Th bias reflect dysregulation of the immune response to Salmonella, and not general dysregulation following Treg depletion, we compared the effect of Treg depletion in uninfected DEREG mice (Supplementary Fig. S6). Four days after DT treatment there was no significant change in Th bias between Treg-depleted DEREG mice and DTR − littermates ( Supplementary Fig. S6e,f). This contrasts with robust changes in Th bias of Treg-depleted mice following S. Tm infection (Fig. 6d, h;  Supplementary Fig. S7b, g). This indicates Treg depletion does not shift Th bias in the absence of S. Tm infection.
Taken together, these results demonstrate that Tregs are required for both the early Th17 response and a maximal Th1bias at day 11 p.i. In Fig. 6i, a schematic diagram depicts the Th response following S. Tm infection and the effects of Tregs to shape the dynamic Th response.   test (d, h). ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

DISCUSSION
Here we have developed an S. Tm infection model to characterize a dynamic CD4 T cell response. We observe an early colonic Th17 response corresponding with a decreased proportion of RORγT + Tregs and a later, long-lasting Th1 response with a decreased proportion of T-bet + Tregs (Fig. 3b-d) within the total T cells at the infection site. The early Th17 response described here is consistent with work identifying an early expansion and contraction of flagellin-specific Th17 cells in the intestine. 33 The Th1 response to S. Tm has been shown to be important for bacterial clearance and protective immunity. 5,8,9,47 The dynamic and site-specific Th response may therefore be optimized to counter a multi-phase infection with changing bacterial targets. However, the factors that control this response are unclear.
The reciprocal dynamic we reveal between Tconvs and Tregs expressing the same TFs is consistent with the hypothesis that Th1 cells are being suppressed by T-bet + Tregs and Th17 cells are being targeted by RORγT + Tregs later after infection. That this dynamic occurs in the colon LP and not the cMLN suggests that reciprocity between Tconvs and Tregs is manifested in the colon itself. Selective suppression of Th1 and Th17 cells by Tregs expressing T-bet or RORγT has been previously reported. Tregs have been shown to upregulate T-bet in response to type 1 inflammation and T-bet + Tregs are required to control Th1mediated inflammation. 24 In the context of Listeria infection, it has been shown that T-bet + Tregs suppress Th1 cells and comprise a stable population that proliferates rapidly during reinfection. 21 It has also been shown that specific intestinal bacteria induce RORγT + Tregs, which limit Th17-mediated colitis, and ablation of Treg-specific STAT3 induces Th17 inflammation. 22,23 Microbiotaspecific CD4 T cells have also been shown to be multi-functional and highly plastic. 48 Unlike previous research, here we have characterized a dynamic Th response that is reciprocal to a Treg response. This highlights the potential for Tregs to shape a multiphase CD4 T cell response in an orchestrated and fine-tuned manner.
To assess the regulation of CD4 T cells, we first assessed changes in S. Tm-specific 2W1S:I-A b tetramer + T cells. 2W1Sspecific cells maintained a Th1 phenotype in all tissues and timepoints ( Supplementary Fig. S4a-d), in contrast to changes in the bulk CD4 T cell pool. Previous work shows an important role for Tconvs that recognize a wide range of S. Tm peptides, as well as non-S. Tm-specific non-cognate activated "bystander" cells, 31,33-35 which prompted our focus on the bulk CD4 T cells. As such, tetramer-negative cells are likely a heterogenous population including cells specific for multiple S. Tm and non-S. Tm antigens. Future work delineating the roles of S. Tm-specific and non-S. Tm specific cells could help elucidate both the reciprocal dynamics reported here and the functions and interactions of CD4 T cells with different antigen specificity. While the antigen specificities of tetramer-negative CD4 T cells described here are unknown, we show that S. Tm infection drives clear changes in these populations, including increased numbers of Tconvs and Tregs, and changes in their TF expression (Figs. 1d, 2c).
To investigate whether Tregs are responsible for changes in Th bias, Tregs were depleted at critical timepoints. Day 6 p.i. was identified as a peak of the early Th17 response and day 11 p.i. was identified as a timepoint where a Th1 bias was re-established (Fig. 3d). For the day 6 experiment, Tregs were ablated during the period when colonic Tregs are enriched for T-bet + cells; for the day 11 experiment, Tregs were ablated during the period when Tregs comprise a higher proportion of RORγT + cells (Fig. 3b, c,  Supplementary Fig. S6c).
At both timepoints Treg depletion increased the number of colonic Tconvs (Fig. 6c, g), consistent with previous reports. [49][50][51][52] Treg depletion did not significantly reduce the number of S. Tm CFUs recovered (Supplementary Fig. S7c, h), in contrast to a previous report of reduced bacterial load in DEREG animals systemically-infected with virulent S. Tm. 53 This discrepancy may result from our use of an oral infection model with an attenuated strain, instead of i.v.-infection with a virulent strain. Because the bacterial burden is low in our model ( Supplementary Fig. S5a), ablation of Tregs and shifts in Th bias may not have significant effects on clearance of the few remaining bacteria.
These experiments show that early Treg ablation leads to a selective increase in Th1 cells and prevents the Th17 bias at day 6 p.i. (Fig. 6d, Supplementary Fig. S7b). On the other hand, later Treg ablation leads to an increased number and proportion of Th17 cells and abrogation of the Th1 bias (Fig. 6h, Supplementary  Fig. S7g). As such, Tregs, which are enriched for T-bet + cells at early timepoints, are necessary for the early Th17 response. At later timepoints, when Tregs contain an increased proportion of RORγT + cells, Tregs are required for a maximal Th1 bias at day 11 p.i. This indicates that the reciprocal dynamics described here are not just correlative, but that Tregs actively shape the CD4 T cell bias (Fig. 6i). These results are consistent with the hypothesis that T-bet + Tregs selectively suppress the early Th1 response and shift the Tconv response towards a Th17 bias, while RORγT + Tregs inhibit Th17 cells, enhancing the later Th1 response. Future work assessing the effect of selectively depleting Treg subsets will help elucidate the extent to which the dynamic Th response is controlled by selective suppression by Tregs expressing T-bet or RORγT.
The potential for CD4 T cells to up-or down-regulate FoxP3 expression is an alternative mechanism that could underly reciprocity between populations of Tconvs and Tregs expressing the same TFs. Plasticity between regulatory and conventional CD4 T cells has been demonstrated in several contexts. For instance, RORγT + Tregs have been shown to downregulate FoxP3 in the absence of IL-15 and become "ex-Tregs" that drive Th17-mediated colitis. 54 It has also been shown that Th1 cells have the potential to upregulate FoxP3 to become T-bet + Tregs. 55,56 On the other hand, evidence from lineage tracking models, transfers and in vitro experiments suggest that T-bet + Tregs are highly stable and retain FoxP3 expression independent of environmental conditions. 21,57 Because of the stability of T-bet + Tregs, plasticity is a less convincing explanation of reciprocity at later phases of infection. Furthermore, the potential for co-expression of Helios with T-bet but not RORγT by Tregs suggests distinct ontogenies. As such, selective suppression is a possible explanation for both reciprocal dynamics and Treg-dependent Th dynamics, although plasticity between populations of Tconvs or between Tconvs and Tregs may also contribute. Future work using lineage-tracking models and TCR profiling would help delineate the contribution of plasticity between subpopulations.
This research reveals a dynamic Th response in the colon after S. Tm infection and a reciprocal dynamic between Tconvs and Tregs expressing the same TFs. Treg depletion experiments show that the early colonic Th17 response and subsequent Th1 bias are dependent on Tregs. In conclusion we highlight that Tregs not only inhibit the Tconv response but shape the CD4 T cell response in a highly nuanced, site-specific and time-dependent manner. These results not only have implications for understanding how the T cell response to S. Tm is controlled, but also for how intestinal T cell responses are balanced during infection or inflammation in general.  25,28 S. Tm cultures were streaked out on Luria-Bertani (LB) agar plates before culturing in LB broth for 5 h at 37°C in an incubator shaking at 180 rpm. Cultures were then back-diluted 1:10 before static culture at 37°C overnight (O/N). Cultures were then adjusted to an OD 600 of 1.00, centrifuged at 5000 G for 10 min, and resuspended in sterile phosphate buffered saline (PBS) at an estimated concentration of 1.0-1.5 × 10 9 CFU/ml. Following infections, actual bacterial dosage was confirmed by plating serial dilutions of S. Tm inocula onto LB agar plates. Bacterial recovery Single cell suspensions from tissues were pelleted by centrifuging at 400 G for 5 min and resuspended in 0.1% Triton X-100 (Sigma-Aldrich) in PBS and incubated at room temperature (RT) for 10 min. Samples were then washed and pelleted before being resuspended in PBS, serially diluted and plated on MacConkey agar No. 2 (ThermoFisher, UK) containing 5 µg/ml streptomycin (Sigma-Aldrich) and incubated O/N at 37°C before CFUs were calculated. Bacteria were recovered from feces and cecal contents, which were collected, aliquoted into 100 µg samples, homogenized and serially diluted and plated on MacConkey agar plates as described above.

MATERIALS AND METHODS
Enrofloxacin treatment Antibiotic treatment of S. Tm-infected mice was carried out by adding enrofloxacin (Bayer, Germany) to drinking water at 2 mg/ ml. Enrofloxacin treatment was provided from day 5 to 29 p.i. and water was replaced every 72 h.
Tissue harvest and processing External fat, PP and cecal patches (CP) were removed from intestinal samples and the remaining tissue was chopped and washed in HBSS with 2 mM EDTA (Gibco, UK). Samples were then incubated at 37°C shaking at 205 rpm for 10 min, washed in EDTA buffer and the process was repeated twice. EDTA incubations and washes were repeated thrice before digestion. Digest enzyme cocktails were prepared in complete RPMI media (RPMI 1640 with 100 µg/ml streptomycin, 100 U/m penicillin, 2 mM L-glutamine, and 50 µm 2-Mercaptoethanol) with 10% FCS (all Gibco). Colon and cecal tissue were digested in an enzyme cocktail of 0.45 mg/ ml collagenase V (Sigma-Aldrich), 0.65 mg/ml collagenase D (Roche, Switzerland), 1.0 mg/ml dispase (Gibco) and 30 µg/ml DNAse (Roche). Small intestines were digested with 0.5 mg/ml collagenase V (Sigma-Aldrich). Tissues were incubated at 37°C in an incubator shaking at 205 rpm for 15-20 min. Following digests, samples were filtered through 100 µm filters, washed twice with buffer (PBS with 2% FCS and 2 mm EDTA) and filtered through a 40 µm filter.
Lymph nodes, PPs and CPs were washed in HBSS and chopped into~1 mm pieces, passed through a 40 µm filter and suspended in buffer. Spleens were cleaned and fat was removed before passing through a 40 µm filter with FACS buffer. Samples were suspended in Ammonium-Chloride-Potassium lysing buffer (Ther-moFisher) for 3-5 min on ice, before washing and resuspending in FACS buffer.
Staining for flow cytometry 2W1S:I-A b tetramer was kindly provided by the NIH tetramer core facility (Atlanta, GA). Following infection with S. Tm-2W1S strains, tissues were processed as described above. Samples were transferred into 96 well round bottom plates and resuspended in 25 µl of tetramer mix containing complete RPMI with 10% FCS, 7 µg/ml 2W1S:IA b tetramer, 20 µl/ml mouse serum, and 10 µl/ml Fc block (CD16/32) (Biolegend, CA). Plates were incubated for 2 h at 37°C. Following tetramer staining, samples were stained with fixable viability dye at a 1 µl/ml in PBS. Next, antibodies for extracellular markers were prepared at 1:100 or 1:200 dilutions in FACS buffer, as specified in the list of antibodies used ( Supplementary Fig. S7). Surface staining was performed at 4°C for 30 min before washing and resuspension.
TF staining was carried out after cells were fixed using the eBioscience FoxP3 transcription factor staining kit (ThermoFisher) or BD Cytofix (BD Biosciences, Belgium) according to manufacturer's instructions. Cells were fixed for 1 h at room temperature (RT) in the dark before washing and resuspending in eBioscience FoxP3 permeabilization buffer (PB) (ThermoFisher). Following fixation, TF antibodies were diluted in PB at concentrations listed in Supplementary Fig. S7, added to samples and incubated O/N at 4°C.
For cytokine staining, cells were treated with eBioscience cell stimulation cocktail (ThermoFisher), containing phorbol 12myristate 13-acetate, ionomycin, brefeldin A, and monensin, according to manufacturer instruction. Samples were then surface stained and fixed with BD Cytofix (BD Biosciences) according to manufacturer's instructions and stained with cytokine antibodies diluted for 2 h at RT. Following cytokine staining, cells were washed and resuspended in FACS buffer.
Flow cytometry Samples were analyzed using a BD LSR Fortessa analyzer or sorted using a BD FACSAria IIu or III (BD Biosciences) at the Institute of Infection, Immunity and Inflammation Flow Core Facility (University of Glasgow).
Regulatory T cell depletion Diphtheria toxin (DT) (Sigma-Aldrich) was suspended in PBS at a concentration of 10 µg/ml. DT was administered intraperitoneally (IP) twice, at a 24 h interval, at a concentration equivalent to 30 ng/g mouse weight. DT was administered to DTR + DEREG mice and DTR̄littermates as Treg-replete controls.