Regulatory T cells control the dynamic and site-specific T helper bias following Salmonella infection

FoxP3+ regulatory T cells (Tregs) control inflammation and maintain mucosal homeostasis, but their functions during infection are less well understood. Th1, Th2 and Th17 cells can be identified by their master transcription factors (TFs) T-bet, GATA3 and RORγT; Tregs also express these TFs. While T-bet+ Tregs can selectively supress Th1 cells, it is unclear whether this is true of Tregs expressing other TFs, or whether such selective suppression can alter the balance of the Th cell response. To address these questions, we used Salmonella enterica serotype Typhimurium (STM) to induce non-lethal colitis. Following infection, we observed an early colonic Th17 response, followed by a Th1-dominated response. The early Th17 response parallels an increase in T-bet+ Tregs. Later, Th1 cells and RORγT+ Tregs dominate. This reciprocal dynamic between Th cells and Tregs expressing the same TF indicates that Tregs may selectively suppress Th subsets. To test this, Treg depletion experiments were performed. These demonstrated that Tregs enable both the early colonic Th17 response and the later Th1 response. Thus, Tregs can shape the tissue CD4 T cell response in a fine-tuned manner. This highlights the potential for subpopulations of Tregs to be used in targeted therapeutic approaches.


INTRODUCTION
Salmonella enterica serovar Typhimurium (STM) infects a wide range of animals and is a common non-typhoidal Salmonella (NTS) serotype in humans. To study intestinal STM infection a streptomycin pre-treatment model has been developed, creating a niche for STM to colonise 1,2 . This facilitates consistent intestinal infection and colitis 3,4 . Because the use of virulent STM strains in susceptible mice causes high mortality, we here combine streptomycin pre-treatment with an attenuated STM strain that induces consistent non-lethal colitis. This enabled analysis of the CD4 T cell response in the colonic lamina propria and the draining lymph nodes for at least nine weeks after infection.

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 STM infection and protection from re-infection 5,6 . In particular, Th1 cells are required for STM 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, where Th17 cells are important for limiting epithelial damage and systemic dissemination [10][11][12] . Thus, an effective immune response to STM involves both Th17 and Th1 cells. It is unclear, however, how the balance between these responses is controlled.

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 to the SI (Fig. 1c, Figure S2), led to the hypothesis that the MLN CD4 T cell response was focused in the colon and caecum draining MLNs (cMLNs). This hypothesis is consistent with previous work showing that different intestinal sites drain to specific MLNs 26,27 ( Figure S3a). With the aim of improving sensitivity of detecting STM responses in the MLNs, a comparison of CD4 T cells in the cMLN and sMLN was conducted following STM infection. 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 ( Figure S3b, c). Together, these data show that the CD4 T cell response to STM in the MLN is restricted to the colon and caecum-draining lymph nodes.

The CD4 T cell response to STM comprises an early colonic Th17 response followed by a sustained Th1 response
To characterise the CD4 T cell response we analysed activated conventional T cells (Tconvs) expressing master transcription factors (TFs). Tconvs were identified as CD4 T cells (gated as described in Fig. 1d) that are CD44 hi and FoxP3 - (Fig. 2a, left). Tconvs were characterised by expression of T-bet, RORγT and GATA3 to identify Th1, Th17 and Th2 cells respectively 28 . We initially examined the 2W1S-specific T cell response, and observed that 2W1S-specific T cells homogenously display a Th1 phenotype at all time points following infection ( Figure S4), consistent with previous reports 29, 30 . In contrast, the total CD4 T cell response following STM infection comprises heterogenous and dynamic populations of Th1 cells, Th17 cells, and Tregs. To assess changes in Th subsets and Tregs following infection, the bulk CD4 T cell pool was therefore characterised in the subsequent experiments.
The number and proportion of cMLN (Fig. 2b) and colonic LP (Fig. 2c) CD44 hi Tconvs are plotted from mock (PBS)-infected mice and STM-infected mice harvested at 6, 30 or 60 days p.i. In the cMLN, there are increased numbers (Fig. 2b, left) and proportions (Fig. 2b,  there is an increased number of Tconvs expressing T-bet (Th1), RORγT (Th17) or T-bet and RORγT 6 days p.i (Fig. 2c, left). While Th1, Th17, and T-bet + RORγT + cells increase in number, the increase in Th17 cells is greater, leading to an increased proportion of Th17 cells and a decreased proportion of Th1 and Th2 cells at d6 (Fig. 2c, right). This Th17 bias at 6 days p.i. is transient however, and by day 30 p.i. a strong Th1 bias is observed. Cytokine staining of PMA Ionomycin-stimulated cells shows that T-bet + Tconvs produce predominantly IFNγ, and RORγT + Tconvs produce IL17A, which supports the characterisation of these cells as Th1 and Th17 ( Figure S5). In summary, there is an initial colonic Th17 response which shifts to a long-lasting Th1 response. This dynamic CD4 T cell response is observed in the colonic mucosa but not in the cMLN. ns, not significant; *p<.05; **p<.01; ***p<.001; ****p<.0001.

Reciprocal dynamics between colonic Tconvs and Tregs expressing the same TFs
Following characterisation of the Tconv response, 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 STM-infected mice harvested at 6, 30 or 60 days p.i. (Fig. 3b). There was an increased number of Tregs that were T-bet + and Tbet + RORγT + at 6 days p.i., but no increase of RORγT + or GATA3 + Tregs. The changes in absolute number correspond with an increased proportion of Tregs that were T-bet + 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). To further investigate this reciprocity, a more detailed time-course was carried out during the first 30 days p.i. In Fig. 3c-d, the proportions of Tregs that are T-bet + or RORγT + (black lines) are overlaid with the proportion of Tconvs (grey lines) that express the same TFs. These plots highlight a reciprocal dynamic between Tconvs and Tregs expressing T-bet or RORγT in the colon, but not the cMLN. Furthermore, these data show that day 6 p.i. is the peak of the colonic Th17 response and a Th1 bias is re-established by day 11 p.i (Fig. 3c). In summary, the early Th17 response in the colon coincides with an increased proportion of T-bet + Tregs, and the later Th1 bias coincides with an increased proportion of RORγT + Tregs. This dynamic is consistent with the hypothesis that Th1 cells are selectively suppressed by T-bet + Tregs and Th17 cells are selectively targeted by RORγT + Tregs.

T-bet expression by Colon CD4 T cells T-bet expression by cMLN CD4 T cells RORγT expression by cMLN CD4 T cells
RORγT expression by colon CD4 T cells

Helios hi Tregs include T-bet + and GATA3 + cells but not RORγT + cells
The reciprocal dynamic between Tregs and Tconvs expressing the same TFs raises questions about the ontogeny and stability of Tregs that express T-bet or RORγT following STM infection. Helios expression has been proposed to differentiate thymic Tregs (tTregs) from peripherally induced Tregs (pTregs) 31 , but this is controversial as Helios expression has also been reported in pTregs [32][33][34] .
Recent research has shown that Helios hi and Helios lo populations have distinct TCR repertoires and represent distinct populations 35 . To assess whether Tregs expressing Th TFs co-express Helios following STM infection, t-SNE plots were generated from colon CD4 T cells 11 days after mock-or STM-infection (Fig. 4b, c). Parameters used include CD44, FoxP3. T-bet, RORγT, GATA3, Helios, CXCR3 and CCR6.
Comparing t-SNE plots of Th TFs with Helios, RORγT + Tregs are mainly confined to Helios lo/int cells.
In contrast, GATA3 + Tregs are primarily Helios hi cells, while Tbet + Tregs include cells with a range of Helios expression levels. STM infection clearly increases the proportion of Tbet + Tregs, and the majority of these cells are Helios hi . The proportion of Helios lo Tregs is reduced by infection but RORyT + Tregs remain Helios lo/int cells (Fig. 4b-c). Therefore, while STM infection alters the proportion of different Treg populations, the co-expression of Th TFs with Helios remains stable.
Helios is a marker of tTregs 31 and RORγT is a marker of pTregs 22,36

Treg depletion alters the balance of the Th1 / Th17 response following Salmonella infection
The reciprocal dynamic between Tregs and Tconvs during STM 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 are required to control the dynamic Th response to STM, Treg depletion experiments were conducted using FoxP3 DTR DEREG mice, which allow ablation of FoxP3 + Tregs following treatment with diphtheria toxin (DT) 25 . 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 ( Figure S6a-b).
In the first set of depletion experiments, DEREG mice were DT-treated at day 1 and 2 p.i., before being culled at day 6 p.i. (Fig. 5a). There was no significant difference in weight change between DEREG mice and wild-type (WT) littermates ( Figure S6c). Treg depletion also had no significant impact on STM CFUs recovered in faeces 6 days p.i. (Figure S6d). The number of CD4 + T cells (Fig.   5b) and CD44 hi Tconvs (Fig. 5c) increased following Treg depletion, but this increase was confined to T-bet + Th1 cells and T-bet + RORγT + cells, while the number of RORγT + Th17 cells remained unchanged (Fig. 5d, left). The increased number of T-bet + Tconvs underlies a decreased proportion of Th17 cells and an increased Th1 bias (Fig. 5d, left). These data demonstrate that the Th17 bias at 6 days p.i. (Fig. 2b) is dependent on Tregs, which include a higher proportion of T-bet + cells at this time point (Fig. 3b).
Next, depletion experiments were conducted to determine if Tregs were also necessary to drive the Th1 bias that becomes prominent by day 11 p.i. (Fig. 2b). DT was administered day 6-7 p.i. and mice were culled at day 11 p.i. (Fig. 5e). DT treatment at this timepoint did not cause a significant difference in weight change between DEREG mice and WT littermates ( Figure S6f). There was also no significant difference between faecal STM CFUs ( Figure S6g), the number of total colon cells ( Figure S6h), CD4 T cells (Fig. 5f) or CD44 hi Tconvs (Fig. 5g). Following Treg depletion, there was an increased number of both colonic Th17 and Th1 cells (Fig. 5h, left). The increase we observed in Th17 cells was greater than that of Th1 cells, reflecting a decreased Th1 bias at day 11 p.i. (Fig. 5h, right). These results demonstrate that Tregs are required for both the early Th17 and also the later Th1-biased response. In Fig. 5i, a schematic diagram depicts the Th response following STM infection and the requirement for Tregs to shape the early Th17 response and to increase the later Th1 response. The dashed-arrow is used to represent the smaller effect of RORγT + Tregs in the later Th1 response, since the later Treg depletion only partially inhibits the Th1 response. Holm- Šídák test (d,h). ns, not significant; *p<.05; **p<.01; ***p<.001; ****p<.0001.

DISCUSSION
Here we have developed an STM infection model to characterise a dynamic CD4 T cell response. In the colon we observe an early Th17 response and a later, long-lasting Th1 response. The Th1 response to STM has been shown to be important for bacterial clearance and protective immunity 5,8,9,37 . The early Th17 response described here is consistent with work identifying an early expansion and contraction of flagellin-specific Th17 cells in the intestine 38 .
A dynamic and site-specific CD4 T cell response may enable the immune system to respond to changing bacterial targets, but the factors that control this response are unclear. Increasing evidence shows that Tregs expressing Th 'master' TFs can selectively suppress Th subsets and shape CD4 T cell bias [21][22][23][24]39 . To analyse the role of Tregs in controlling the Tconv response to STM, we first characterised the Treg response following infection. Results highlight a reciprocal dynamic between Tconvs and Tregs expressing T-bet or RORγT in the colon but not the cMLN (Fig. 3). The early increased proportion of colonic Th17 cells coincides with a decreased proportion of RORγT + Tregs, and the later Th1 bias coincides with a decreased proportion of RORγT + Tregs (Fig. 3c, d). The reciprocal dynamic 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
reported. Tregs have been shown to upregulate T-bet in response to type 1 inflammation and T-bet + Tregs are required to control Th1-mediated 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 . Unlike previous research, here we have characterised a dynamic Th response that is reciprocal to a Treg response. This highlights the potential for Tregs to shape a multi-phase CD4 T cell response in a fine-tuned and orchestrated manner.
To demonstrate that Tregs are responsible for the changes in Th bias, we depleted Tregs at critical timepoints following infection. 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). In the first set of DEREG experiments, DT was administered at days 1 and 2 p.i. and mice were culled at day 6 p.i.
Treg depletion increased the number and activation of colonic CD4 T cells (Fig. 5.2F,G), consistent with previous reports [40][41][42][43] . Treg depletion did not significantly reduce the number of STM CFUs recovered from faeces however (Fig. 5.2D), in contrast to a previous report of reduced bacterial load in DEREG animals systemically-infected with STM 44 .
In the second set of depletion experiments, DT was administered days 6 and 7 p.i., and mice were Together, 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. On the other hand, later Treg ablation leads to a preferential increase in the number and proportion of Th17 cells (Fig. 5d, h). As such, Tregs containing a high proportion of T-bet + cells are necessary for the early Th17 response and Tregs containing an increased proportion of RORγT + cells are required for an optimum 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. 5i). 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 the later Th1 response.
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 Cultures were then adjusted to an OD600 of 1.00, centrifuged at 5,000G for 10 min, and resuspended in sterile phosphate buffered saline (PBS) at an estimated concentration of 1.0-1.5 x 10 9 CFU/ml.
Following infections, actual bacterial dosage was confirmed by plating serial dilutions of STM inocula onto LB agar plates.

In vivo infections
6-10 week old mice were pre-treated with 20 mg streptomycin (Sigma-Aldrich, USA) suspended in 10 µl sterile PBS by oral gavage 24 hrs before STM infection. Mice were infected by administration of 100 µl of STM inoculum by oral gavage.

Bacterial recovery
Single cell suspensions from tissues were pelleted by centrifuging at 400 G for 5 mins and resuspended in 0.1% Triton X-100 (Sigma-Aldrich) in PBS and incubated at room temperature (RT) for 10 mins. Samples were then washed and pelleted before being resuspended in PBS, serially diluted and plated on MacConkey agar No. 2 (ThermoFisher Scientific, UK) containing 5 µg/ml streptomycin (Sigma-Aldrich) and incubated O/N at 37°C before CFUs were calculated. Bacteria were recovered from faeces and caecal contents, which were collected, aliquoted into 100 µg samples, homogenised and serially diluted and plated on MacConkey agar plates as described above.

Tissue harvest and processing
External fat, Peyer' patches (PP) and caecal 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 mins, washed in EDTA buffer and the process was repeated twice. EDTA incubations and washes were repeated thrice before digestion.
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, Peyer's patches (PPs) and caecal patches (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 (ACK) lysing buffer (ThermoFisher) for 3-5 mins 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 (USA). Following infection with STM-2W1S strains, tissues were harvested and processed as described above. Samples were transferred into 96 well round bottom plates and resuspended in 25µl 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, USA). Plates were incubated for 2 hrs 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 ( Figure S7). Surface staining was performed at 4°C for 30 mins before washing and resuspension.
Transcription factor (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 hr 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 Figure 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 12-myristate 13-acetate (PMA), 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 hrs at RT. Following cytokine staining, cells were washed and resuspended in FACS buffer.

Flow cytometry
Samples were analysed using a BD LSR Fortessa analyser 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 hr interval, at a concentration equivalent to 30 ng/g mouse weight. DT was administered to DTR + DEREG mice and DTR ̄ littermates as WT controls.