In wild-type (WT) naïve CD4⁺ T cells, Cars2 expression is downregulated after TCR stimulation in vitro, suggesting that CARS2/CPERS may play a role in T cell function ( Supplementary Figure S1 ). To address the question of whether CARS2/CPERS-dependent supersulfide metabolism is functional in CD4⁺ T lymphocytes, we first examined thymic T cell development in WT versus Cars2 ^+/- mice, the latter of which we previously reported that the same metabolic pathway is significantly reduced in (11). To do so we compared double negative (DN; CD4^- CD8^-), double positive (DP; CD4⁺ CD8⁺), CD4 and CD8 single positive (SP; CD4⁺ CD8^- and D4^- CD8⁺, respectively), and Foxp3⁺ thymocytes between WT and Cars2 ^+/- mice at the age of 2-3 months (young) and 12 months (old). WT and Cars2 ^+/- mice exhibited the unaltered numbers of these thymocyte subsets ( Supplementary Figures S2A, B ), suggesting that CARS2/CPERS-dependent supersulfide metabolism does not significantly affect T cell development in the thymus.

To examine if CARS2/CPERS affects peripheral CD4⁺ T cell homeostasis, we analyzed the same cells in the spleen, mesenteric lymph nodes (mLNs), and the colon obtained from young and old WT versus Cars2 ^+/- mice. While CD4⁺ T cells were equally present in the spleen of WT and Cars2 ^+/- mice, those and especially their CD44^hi CD62L^lo effector/memory subset in mLNs and the colon significantly increased in number in old but not young Cars2 ^+/- mice ( Figures 1A–C ). In consistent, CD4⁺ T lymphocytes histologically accumulated in the colon of old Cars2 ^+/- mice, which led to hyperplastic mucosa with increased inflammatory infiltrates in the same tissues ( Figures 1D, E ). Among other immune cells infiltrating in the colon, CD44^lo CD62L^hi naïve CD4⁺ T cells accumulated slightly more in the Cars2 ^+/- than WT old mice, whereas we could not detect any difference in other cell types such as Foxp3⁺ Tregs, CD8⁺ T cells and neutrophils ( Supplementary Figures S3A–C ). Furthermore, the weight of each organ, colon length and body weight were not altered between WT and Cars2 ^+/- old mice ( Supplementary Figures S4A–C ), and we could not detect any difference in Cars2 expression in CD4⁺ T cells isolated from spleen and mLN between young and old WT mice ( Supplementary Figure S5 ). Together these data indicate that CARS2/CPERS tonically inhibits accumulation of effector/memory CD4⁺ T cells in the colon and gut-associated lymphoid tissues at homeostasis.

Homeostasis of naïve and effector/memory CD4⁺ T lymphocytes is governed by a proliferative response called “homeostatic proliferation” (16). Experimentally, this response can be best examined in lymphopenic settings. Thus, when transferred into lymphopenic animals such as gene-manipulated or sublethally irradiated mice, some naïve CD4⁺ T cells rapidly proliferate to give rise to a subpopulation with an effector/memory phenotype in lymphoid as well as non-lymphoid tissues (17–19). To determine whether the increased number of effector/memory CD4⁺ T lymphocytes in mLNs and the colon of aged Cars2 ^+/- mice ( Figures 1B, C ) is attributed to enhanced homeostatic proliferation, we next sought to examine the same proliferative response of naïve CD4⁺ T cells derived from WT versus Cars2 ^+/- animals. To do so we isolated naïve cells from these two types of mice, labeled with carboxyfluorescein diacetate succinimydyl ester (CFSE), transferred into congenic recipients rendered acutely lymphopenic by sublethal irradiation, and analyzed the donor cells in the spleen, mLNs, and the colon 9 days later (the experimental design shown in Figure 2A ). While the total donor cell number was unaltered between WT and Cars2 ^+/- groups ( Figure 2B ), Cars2 ^+/- as compared to WT CD4⁺ T cells more rapidly proliferated in the colon as reflected by the heightened CFSE^- fractions (CD44^hi CD62L^lo), with the slowly proliferating CFSE⁺ cells (CD44^lo CD62L^hi) largely unchanged ( Figure 2C , Supplementary Figure S6 ). Thus, CARS2/CPERS suppresses fast but not slow homeostatic proliferation of naïve CD4⁺ T lymphocytes in a cell-intrinsic manner.

Based on previous findings demonstrating a correlation between the degree of homeostatic proliferation of CD4⁺ T cells and severity of colitis (20, 21), we hypothesized that enhanced homeostatic proliferation of Cars2 ^+/- naive CD4⁺ T cells can exacerbate colitis in certain circumstances. To seek this possibility, we used a mouse model of colitis where naïve CD4⁺ T cells are transferred into lymphodeficient mice (21–23). In such chronically lymphopenic environment naive CD4⁺ T cells robustly proliferate to trigger colitis through fast homeostatic proliferation (16). Using this approach, we adoptively transferred WT or Cars2 ^+/- naïve CD4⁺ T cells into Rag2 ^-/- mice and analyzed the donor cells as well as the host animals at different time points. Rag2 ^-/- hosts that received Cars2 ^+/- versus WT CD4⁺ T cells exhibited severer body weight loss and histological colitis ( Figures 3A–C ). In consistent, the number of Cars2 ^+/- as compared to WT donor cells was significantly higher in the gut but not in lymphoid organs 4 weeks after transfer ( Figure 3D and Supplementary Figures S7A, D ). Furthermore, the frequency of effector cytokine-producing cells (IFN-γ⁺, IL-17A⁺, and IFN-γ⁺ IL-17A⁺) among the total donor population was comparable between the two groups in all of the organs examined, and as a consequence, the total number of cytokine-secreting donor cells accumulating in the gut 4 weeks after transfer was higher in the Cars2 ^+/- CD4⁺ T cell-transferred group ( Figure 3E , Supplementary Figures S7B, E ). Because the frequency of dead cells was equivalent between the two groups ( Figure 3F , Supplementary Figures S7C, F ) and because the in vitro assay demonstrated that CARS2 deficiency had no effect on T cell activation and differentiation ( Supplementary Figures S8A, B ), these results suggest that augmented accumulation of cytokine-secreting Cars2 ^+/- CD4⁺ T cells in the colon may reflect an enhanced rate of lymphopenia-induced homeostatic proliferation.

To test this hypothesis, we performed cell cycle analysis of the donor cells during the course of the experiments described above. One week after naïve CD4⁺ T cell transfer, the frequency of cells in G0 phase was significantly lower in Cars2 ^+/- versus WT donor cells while that of cells in G1 phase higher in the former donor cell population, with cells in S and G2/M phases largely unaffected ( Figure 4A ). These differences were exclusively detected in donor cells accumulating in the gut but not in lymphoid organs ( Figure 4A ), and were not observed at a later time point ( Figure 4B ). In consistent, we detected downregulation of a cell cycle inhibitor Trp53 in Cars2 ^+/- CD4⁺ T cells at one week post transfer ( Figure 4C ). Furthermore, when naive CD4⁺ T cells were cultured in vitro in the presence of CD3/CD28 antibodies, Cars2 ^+/- CD4⁺ T cells exhibited a higher frequency of cell cycle entry ( Figure 4D ). Together these data demonstrate that CARS2/CPERS suppresses cell cycle entry of pathogenic CD4⁺ T cells in a cell-intrinsic manner at an early stage of intestinal inflammation.

Naïve CD4⁺ T cells generate a few Foxp3⁺ regulatory T cells (Tregs) when transferred into lymphopenic mice (24). To examine whether CARS2/CPERS affects differentiation and/or suppressive function of the latter cells, we analyzed Tregs that differentiate from naïve CD4⁺ T cells in the above colitis model. To do so we transferred naïve CD4⁺ T lymphocytes into Rag2 ^-/- animals and analyzed the donor cells at different time points. As shown in Supplementary Figure S9A , there was no difference in the number of generated Tregs, making the possibility unlikely that CARS2/CPERS is essential for differentiation of Tregs from naïve precursors. To further determine whether CARS2/CPERS affects suppressive function of Tregs, we co-transferred Cars2 ^+/+ Foxp3^- naïve CD4⁺ T cells together with Foxp3⁺ cells of Cars2 ^+/+ or Cars2 ^+/- origins into Rag2 ^-/- mice. The two types of Tregs equally suppressed naive cell-induced colitis ( Supplementary Figure S9B ). Hence, CARS2/CPERS appears to be dispensable for Treg differentiation and function.

CARS2/CPERS regulates sulfur metabolism through the production of supersulfide (11). To ask whether CARS2/CPERS exerts its suppressive activity on colitogenic T cells via supersulfide, we administered a supersulfide donor glutathione trisulfide (GSSSG) to Rag2 ^-/- mice that had received naïve CD4⁺ T lymphocytes. Treatment with GSSSG significantly ameliorated body weight reduction and histological colitis induced by Cars2 ^+/- naïve CD4⁺ T cell transfer ( Figures 5A–C ). On the other hand, the same treatment did not rescue severity of colitis driven by WT CD4⁺ T cells ( Supplementary Figures S10A–C ). Further analysis of Cars2 ^+/- donor cells revealed that GSSSG inoculation significantly repressed the absolute number of donor cells accumulating in the gut ( Figure 5D ). While the frequency of cytokine-producing cells among gut-infiltrating CD4⁺ T cells was unchanged regardless of GSSSG supplementation, their absolute numbers were lowered in the presence of GSSSG ( Figure 5E ). The same treatment had no influence on cell death of Cars2 ^+/- CD4⁺ T cells ( Figure 5F ). These data argue that augmented severity of colitis in Rag2 ^-/- mice that have received Cars2 ^+/- CD4⁺ T cells is attributed to insufficient amounts of supersulfide in the same cells.

We next examined cell cycle status of Cars2 ^+/- donor cells in the above experiments. As shown in Figure 5G , GSSSG treatment significantly suppressed cell cycle entry of CD4⁺ T cells at 1 but not 4 week(s) post transfer. Moreover, Trp53 expression in CD4⁺ T cells was upregulated by GSSSG administration ( Figure 5H ). To further confirm whether GSSSG can directly act on CD4⁺ T cells, we stimulated Cars2 ^+/- naïve CD4⁺ T cells in vitro and analyzed their cell cycle status in the presence or absence of GSSSG. Addition of GSSSG diminished the frequency of cycling cells while increased the quiescent fraction ( Figure 5I ). Together these findings suggest that supersulfide can inhibit cell cycle entry of CD4⁺ T cells.

The forementioned results show that reduced levels of CARS2/CPERS-dependent supersulfide metabolism lead to excess proliferation of CD4⁺ T lymphocytes, thereby contributing to intestinal inflammation in mice. To gain insight into the question of whether lowered levels of CARS2 in CD4⁺ T cells are associated with pathogenesis of IBD in humans, we re-analyzed publicly available datasets of single cell RNA sequencing of intestinal lamina propria mononuclear cells obtained from CD patients versus controls (25, 26). After combining samples from 10 CD patients and 7 controls, we annotated T cell clusters based on the unique signature gene CD3E ( Supplementary Figures S11A, B ). We next performed re-clustering of the CD3E-positive T cell fraction and identified 22 subclusters ( Figure 6A , Supplementary Figure S11C ). Among them, clusters 0, 2, 3, 5, 7 – 9, 12, 13, 15, 21 expressed both CD3E and CD4, and defined them as CD4⁺ T cells ( Figures 6A, B , Supplementary Figure S11C ), while cluster 1, 4, 6, 10, 11, 15, 16, 19 – 21 were CD8⁺ T cells expressing CD8A. Intriguingly, expression levels of CARS2 were lower in CD4⁺ T lymphocytes in CD patients as compared to controls ( Figure 6C ). Additionally, the frequency of CARS2-expressing cells was lower in CD4⁺ T cells in the former group ( Figure 6D ). These findings suggest that CARS2/CPERS may be functional in human CD4⁺ T cells in the development and/or augmentation of IBD.

To further characterize the profiles of CD4⁺ T cells, we compared naive and effector markers between CD patients and controls. CD4⁺ T cells from CD patients exhibited lower levels of a naïve marker SELL and higher levels of activation markers HLA-DRA and IFNG (27) ( Figure 6E ). Similarly, expression of the proliferation marker MYC was high in CD4⁺ T cells of CD origins (28) ( Figure 6F ). While the expression level of TP53 was not significantly different in CD4⁺ T cells, the frequency of CARS2-positive cells among CD4⁺ cells positively correlated with that of TP53-positive cells ( Figures 6F, G ). To further clarify the profiles of each gene in CD4⁺ T cells, we subclustered the same cells based on gene expression patterns ( Supplementary Figure S11C ). Within CD4⁺ T cells, cluster 0, 12, 21 showed the marker of naïve CD4⁺ T cells: SELL, LEF1 and CCR7 (29). Cluster 2, 3, 7, 8, 9, 15 corresponded to effector CD4⁺ T cells expressing TIMP1, LGALS1, STAT1, SOCS1, ODC1, and/or IL2 (29, 30). Cluster 5 and 13 were both Treg based on FOXP3 expression. Because approximately 90% of CD4⁺ T cells are composed of effector subset ( Figures 6H, I ), we focused on the effector subset in the later analysis. Similar to the total CD4⁺ T cells, CARS2 expression and frequency were both lower in effector CD4⁺ T cells in CD patients ( Figures 6J, K ). Furthermore, effector CD4⁺ T cells from CD patients exhibited higher levels of activation markers (HLA-DRA, IFN-G) and proliferation marker MYC while that of cell cycle inhibitor TP53 was lower ( Figures 6L, M ). Moreover, the frequency of CARS2-expressing cells among effector CD4⁺ cells positively correlated with that of TP53-positive cells ( Figure 6N ). These findings suggest that downregulation of CARS2/CPERS-dependent supersulfide metabolism in CD4⁺ T cells is associated with the pathogenesis of IBD in humans.

Last, we sought to formally test the relationship between CARS2-dependent supersulfide metabolism and cell cycle in human CD4⁺ T cells. To do so we isolated naïve CD4⁺ T cells from human healthy donors, stimulated with CD3/28 antibodies in the presence or absence of GSSSG, and analyzed the cultured cells 48 hours later. As shown in Figure 6O , addition of GSSSG significantly suppressed cell cycle entry of CD4⁺ T lymphocytes. Thus, CARS2/CPERS-dependent supersulfide metabolism inhibits CD4⁺ T cell proliferation by suppressing cell cycle entry in humans.

C57BL/6 CD45.2⁺ WT mice were purchased from Japan SLC (Hamamatsu, Japan). Rag2 ^-/- and CD45.1⁺ WT mice were obtained from breeding stock at Tohoku University Graduate School of Medicine. Foxp3-RFP reporter mice were obtained from Jackson Laboratory (Bar Harbor, ME) (49). Cars2 ^+/- mice are previously described (11). Since Cars2 ^-/- are embryonically lethal, we used Cars2 ^+/- mice in this study. Cars2 ^+/- Foxp3-RFP reporter mice were obtained by crossing Foxp3-RFP reporter with Cars2 ^+/- mice. All mice were maintained in specific pathogen-free environment in Tohoku University Graduate School of Medicine. The care and handling of the animals used in our study were in accordance with the animal study protocols approved by the Institutional Committee for the Use and Care of Laboratory Animals of Tohoku University (2019MdA-204-04).

GSSSG was synthesized according to our previous reports (11, 32, 50). In brief, 20 mM GSH was reacted with 20 mM NaHS in 20 mM Tris-HCl buffer (pH 7.4) containing 20 mM iodine at room temperature for 15 min. For purification of GSSSG, the reaction mixture was subjected to high-performance LC Prominence (Shimadzu Corporation, Kyoto, Japan) with a reversed-phase column YMC-Triart C18 column, 50×2.0 mm inner diameter (YMC, Kyoto, Japan), under the following elution conditions: mobile phase A (0.1% formic acid) with a linear gradient of mobile phase B (0.1% formic acid in methanol) from 5 to 90% for 15 min at a flow rate of 0.2 ml/min at 40 °C. Eluted GSSSG was dried in vacuo.

To sort for naïve CD4⁺ T lymphocytes, CD4⁺ cells were enriched from whole splenocytes using CD4 Microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany). Naïve cells (CD3⁺ CD4⁺ CD25^- CD44^lo CD62L^hi or CD3⁺ CD4⁺ Foxp3-RFP^- CD44^lo CD62L^hi) were then purified by FACS Aria II (BD Biosciences, San Jose, CA) ( Supplementary Figure S12A ). To obtain Tregs, CD3⁺ CD4⁺ Foxp3-RFP⁺ cells were sorted out. For examination of homeostatic proliferation, purified naïve CD4⁺ T cells were labeled with CFSE (Thermo Fisher Scientific, Waltham, MA) and injected intravenously into sublethally irradiated (5.5Gy) CD45.1⁺ WT recipient mice (1 x 10⁶ cells per recipient) as previously described (18). To induce colitis, WT or Cars2 ^+/- naïve CD4⁺ T cells were injected into Rag2 ^-/- mice (3 x 10⁵ cells per recipient). In Supplementary Figure S9 , Cars2 ^+/+ naïve CD4⁺ T cells were co-transferred with Tregs (1 x 10⁵ cells per mouse) derived from Cars2 ^+/+ or Cars2 ^+/- Foxp3-RFP reporter mice. For in vivo GSSSG treatment, Rag2 ^-/- recipient mice were daily injected intraperitoneally with GSSSG (400 nmol per mouse) starting on the same day of naïve CD4⁺ T cell transfer.

Three to four weeks after naive cell transfer, colons of Rag2 ^-/- hosts were fixed in 10% formalin. Paraffin-embedded samples were then cut into 5 μm sections and stained with hematoxylin and eosin (H&E) or CD4 (EPR19514) monoclonal antibody (mAb) (Abcam, Trumpington, UK). Images were acquired using BZ-X810 (Keyence, Osaka, Japan). The histological score was measured as previously described (51).

Single-cell suspensions from the thymus, spleen, and mLNs were prepared and red blood cells lyzed in ACK buffer. Colonic lamina propria cells were isolated using Lamina Propria Dissociation Kit (Miltenyi Biotec) according to the manufacturer’s instructions. Lymphocytes were then separated using Percoll (GE Healthcare, Chicago, IL).

For the measurement of Cars2 mRNA in cultured CD4⁺ T cells, naïve CD4⁺ T cells stimulated with plate-coated CD3 (1 µg/ml) (Biolegend, San Diego, CA) were collected. For the measurement of Trp53 mRNA in cultured CD4⁺ T cells, naïve CD4⁺ T cells stimulated with plate-coated CD3 (1 µg/ml) and soluble CD28 (1 μg/ml) (both from Biolegend) were collected. For detection of Trp53 mRNA in CD4⁺ T cells in vivo, colonic CD4⁺ T cells were sorted out. Total RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany) and reverse-transcribed with the High Capacity cDNA Reverses Transcriptase Kit (Thermo Fisher Scientific). Real-time PCR was performed using FastStart Universal SYBR qPCR Mix (TOYOBO, Osaka, Japan). qPCR analysis was carried out using the ABI 7500 Real-time PCR System (Thermo Fisher Scientific). Relative gene expression was calculated by the ΔCt method and normalized to the amount of Actin-beta (Actb). The following primer sets were used: Trp53: 5’-ACGCTTCTCCGAAGACTGG-3’ and 5’-AGGGAGCTCGAGGCTGATA-3’; Cars2: 5’-CAGGTGCATAACAGCCTCACT-3’ and 5’-CCACAGCTATACCAGGAGACTG-3’; Actb: 5’-GAAGATCAAGATCATTGCTCCT-3’ and 5’-TGGAAGGTGGACAGTGAG-3’.

In Figure 4 and Supplementary Figure S8 , naïve CD4⁺ T cells (5 x 10⁴ cells per well) isolated by Naïve CD4⁺ T cell isolation Kit (Miltenyi Biotec) were stimulated with antibodies against plate-coated CD3 (1 μg/ml) and soluble CD28 (1 μg/ml) (both from Biolegend) in RPMI complete medium supplemented with 50 μM β-mercaptoethanol for 2 to 3 days. For Th1 condition, recombinant mouse (rm) IL-12 (20 ng/ml; ThermoFisher) and anti-IL-4 antibody (11B11) (5 μg/ml; Biolegend) were added. For Th2 condition, rmIL-4 (10 ng/ml; Biolegend) and anti-IFN-γ antibody (XMG1.2) (10 μg/ml; Biolegend) were added. For Th17 condition, recombinant human (rh) IL-6 (30 ng/ml; ThermoFisher), rhIL-6R (66 ng/ml; ThermoFisher), anti-IFN-γ antibody (10 μg/ml; Biolegend) and anti-IL-4 antibody (5 μg/ml; Biolegend) were added. For Treg condition, rhTGF-β (0.5 ng/ml; ThermoFisher), anti-IFN-γ antibody (10 μg/ml; Biolegend) and anti-IL-4 antibody (5 μg/ml; Biolegend) were added. In Figures 5 , FACS-sorted Cars2 ^+/- naïve CD4⁺ T cells (5 x 10⁴ cells per well) were stimulated with antibodies against plate-coated CD3 (1 μg/ml) and soluble CD28 (1 μg/ml) (both from Biolegend) in RPMI complete media supplemented with 50 μM β-mercaptoethanol for 3 days in the presence or absence of GSSSG (1 μM).

Cells were incubated with CD16/32 mAb (for mice) or Fc receptor blocking solution (for humans; Biolegend) and stained with combinations of the following mAbs for 20 min on ice: CD3 (145-2C11), CD62L (MEL-14) (Thermo Fischer Scientific), CD4 (RM4-5), CD44 (IM7) (BD Biosciences), CD8α (53-6.7), CD25 (PC61), CD45.1 (A20), CD45.2 (104), CD69 (H1.2F3) (Biolegend) for mouse samples, and CD3 (SK7), CD4 (SK3), CD25 (M-A251), CD45RA (HI100), and CD45RO (UCHL1) (BD Biosciences) for human samples. Dead cells were removed using LIVE/DEAD fixable dead cell stain kit (Thermo Fisher Scientific). To detect intracellular antigens, cells were fixed and permeabilized using Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) for 30 min on ice after cell surface staining, followed by staining with mAbs against Foxp3 (FJK-16s), Ki67 (SolA15) (Thermo Fischer Scientific), IFN-γ (XMG1.2), IL-13 (eBio13A) and/or IL-17A (TC11-18H101) (Biolegend) for 30 min at room temperature. For cell cycle analysis, cells were incubated with 2 μg/ml Hoechst 33342 (Thermo Fischer Scientific) for 15 min after intracellular staining as previously described (52). Flow cytometry was performed using LSR Fortessa and the data analyzed with FlowJo software (both BD Biosciences). We counted the number of cells in each cell subset during flow cytometry analysis. The gating strategy is detailed in Supplementary Figure S12B .

Public transcriptomic datasets on colonic T lymphocytes obtained from CD patients and controls were downloaded from Gene Expression Omnibus website (http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE157477 (25) and Broad DUOS (https://duos.broadinstitute.org) under the accession number DUOS-000146 CD_Atlas_2021_GIDER; DUOS-000145 CD_Atlas_2021_PRISM (26) and re-analyzed using R with the package Seurat version 4.0 (http://satijalab.org/seurat/) (53). Analyzed samples were age matched (> 45 years old) and genes detected in less than five cells were excluded. Low-quality cells or empty droplets defined as those with expression of less than 3,000 genes were filtered out.

Whole blood samples were collected from healthy donors under the approval of Institutional Review Boards of Tohoku University Graduate School of Medicine (2021-1-1249). Peripheral blood mononuclear cells were prepared using Vacutainer (BD Biosciences) according to the manufacture’s protocols and purified for naïve CD4⁺ T cells (CD3⁺ CD4⁺ CD25^- CD45RA⁺ CD45RO^-) by using FACS Aria II (BD Biosciences) ( Supplementary Figure S12C ). Cells were then seeded onto a 96 well round bottom plate (5 x 10⁴ cells/well) and stimulated with Dynabeads™ Human T-Activator CD3/CD28 (Thermo Fisher Scientific) in RPMI complete media in the presence or absence of GSSSG (1 μM) for 48 hours.

For re-analysis of human single cell RNA sequencing data, we conducted a Wilcoxson Rank Sum test to establish statistical significance. In human T cell culture experiments, a paired Student’s t-test was performed. In all other instances, an unpaired t-test was applied. p values <0.05 were considered to be statistically significant.