Reciprocal Regulation of Glycolysis-Driven Th17 Pathogenicity and Regulatory T Cell Stability by Cdc42

Upon Ag recognition, naive CD4⁺ T cells differentiate to several types of effector cells, including IFN-γ–producing T helper 1 (Th1) and IL-17–producing Th17 cells. Th17 cells, induced by the transcriptional factor RORγT, are functionally diverse. On the one hand, Th17 cells play a critical role in the clearance of extracellular bacterial and fungal infections. On the other hand, these cells can be pathogenic and induce tissue inflammation and autoimmune diseases (1). Protective, nonpathogenic Th17 cells can be generated in vitro from naive T cells by using TGF-β and IL-6, whereas pathogenic Th17 cells can be generated by using TGF-β + IL-6 + IL-23 or IL-1β + IL-6 + IL-23 (2). It has recently been found that pathogenic Th17 cells upregulate 16 proinflammatory genes (e.g., T-bet, IL-23R, and IL-22), whereas they downregulate seven regulatory genes (e.g., IL-10) (3). Th17 cells may become plastic as they can transdifferentiate to a cell subset coexpressing Th17 and Th1 cytokines and to Th1 cells (4). Th17 plasticity has been implicated in pathogenesis of inflammatory bowel diseases (4). In line with the fact that pathogenic Th17 cells express Th1 signature transcriptional factor T-bet, it is conceivable that Th17 pathogenicity is closely associated with their plasticity. Nonetheless, the molecular mechanisms that control Th17 pathogenicity/plasticity remain poorly defined, identification of which could allow selective suppression of pathogenic Th17 cells while sparing nonpathogenic Th17 cells.

Regulatory T cells (Tregs) are the cells that maintain immune tolerance by inhibition of T cell proliferation and effector T cell function (5). Tregs are identified by the expression of transcriptional factor Foxp3 and can be classified as natural Tregs (nTregs) and induced Tregs (iTregs). nTregs are developed from the thymus whereas iTregs are derived from peripheral naive T cells (6). TGF-β has been shown to maintain nTregs in the periphery and drives iTreg differentiation (6).

The balance between Th17 cells and Tregs is important for host immunity and immune tolerance. Loss of this balance can lead to excessive Th17-mediated immune responses and undesired autoimmunity (7, 8). It is generally considered that loss of Th17/Treg balance is attributed to altered differentiation and/or impaired Treg suppressive function (5). Nonetheless, most recent studies suggest that Treg lineage instability also contributes to imbalance between effector T cells (e.g., Th17) and Tregs (9–16). Treg lineage instability is a phenomenon where Tregs lose Foxp3 and acquire effector T cell functions and during which an intermediate cell population coexpressing Foxp3 and effector T cell signature genes may emerge (4, 9–13). The mechanisms underlying Treg instability are largely undetermined.

Cdc42 of the Rho family small GTPases is an intracellular signal transducer that cycles between GTP-bound active and GDP-bound inactive states (17, 18). In T cells, overexpression of constitutively active or dominant negative mutant of Cdc42 suggests that Cdc42 plays a role in T cell actin and tubulin cytoskeleton polarization, migration, and development (19–22). By genetic deletion of Cdc42 in T cells, we have recently reported that Cdc42 positively regulates naive T cell homeostasis but negatively regulates T cell activation and Th1 cell differentiation (23, 24). In this study, we further report that Cdc42 is required for Th17/Treg balancing by negative regulation of Th17 differentiation/pathogenicity and positive regulation of Treg differentiation/suppressive function/stability through integrating transcriptional and metabolic signaling.

Cdc42^fl/fl mice were generated as described previously (25). LCK^Cre mice and Foxp3^YFP-Cre were purchased from The Jackson Laboratory. Cdc42^fl/fl mice were bred with LCK^Cre mice or Foxp3^YFP-Cre mice in our animal facility to generate Cdc42^fl/flLCK^Cre mice or Cdc42^fl/flFoxp3^YFP-Cre mice. RAG1^−/− mice and BoyJ mice were obtained from the Cincinnati Children’s Hospital Research Foundation Comprehensive Mouse and Cancer Core. All mice were housed under specific pathogen-free conditions in the animal facility at the Cincinnati Children’s Hospital Research Foundation in compliance with the Cincinnati Children’s Hospital Medical Center Animal Care and Use Committee protocols.

Abs for flow cytometry were obtained from BD Biosciences, eBioscience, or BioLegend. Anti-CD3 (clone-145-2c11) (Cat. no. 553057) and anti-CD28 (clone-37.1) (Cat. no. 553294) were purchased from BD Biosciences. Anti-pSmad2/3 (Cat. no. 8828) and anti-Smad2/3 (Cat. no. 5678) were purchased from Cell Signaling Technology. Anti–IFN-γ and anti–IL-4 and recombinant cytokines IL-6 and IL-2 were purchased from R&D Systems. TGF-β was purchased from Peprotech. Dextran sodium sulfate (DSS) (Cat. no. 0216011050) was purchased from MP Biomedicals. 2-deoxy-D-glucose (2-DG) (Cat. no. D6134), collagenase (Cat. no. C5138), PMA (Cat. no. P8139), and ionomycin (Cat. no. I0634) were purchased from Sigma (St. Louis, MO). IL-17A and IFN-γ ELISA sets were purchased from BD Biosciences. ELISA kits for anti-dsDNA and antinuclear Ab were purchased from Alpha Diagnostics.

Total splenocytes were subjected to CD4⁺ T cell isolation using CD4⁺ T cell isolation kit (Cat. no. 130-104-454) (Miltenyi Biotec), according to the manufacturer’s protocol. The MACS-sorted CD4⁺ T cells were stained for CD4, CD62L, and CD44. Naive CD4⁺CD62L⁺CD44⁻ cells were FACS-sorted and stimulated with plate-bound anti-CD3 (2 μg/ml) and soluble anti-CD28 (2 μg/ml) along with either TGF-β (1 ng/ml), IL-6 (30 ng/ml), anti–IFN-γ (10 μg/ml), and anti–IL-4 (10 μg/ml) for Th17 differentiation for 4 d or TGF-β (5 ng/ml), IL-2 (10 ng/ml), anti–IFN-γ (10 μg/ml), and anti–IL-4 (10 μg/ml) for iTreg differentiation for 4 d. Cells were restimulated with PMA (5 ng/ml) and ionomycin (50 ng/ml) along with GolgiPlug (BD Biosciences) for 4 h and then harvested and prepared either for intracellular staining or for real-time RT-PCR.

Cells were incubated with anti-CD16/32 (2.4G2) (BD Biosciences) to block FcγR II/III, and then stained with various conjugated Abs as indicated. BD Cytofix/Cytoperm kit (BD Biosciences) was used for intracellular staining. Stained cells were analyzed by BD LSRII, FACSCanto, and LSRFortessa flow cytometers. Data were analyzed with BD FACSDiva and FlowJo software.

For 5-bromo-2′-deoxyuridine (BrdU) incorporation assay, mice were injected i.p. with 500 μg BrdU. Two hours after injection, splenocytes were isolated and immunolabeled with anti-CD4 Ab and BrdU incorporation was analyzed by a BrdU Flow kit as per the manufacturer’s protocol (BD Biosciences).

For cell apoptosis assay, freshly isolated splenocytes were incubated with anti-CD4 Ab for 20 min. The cells were washed, incubated with Annexin V (BD Biosciences) for 20 min, and then analyzed by flow cytometry.

FACS-sorted naive CD4⁺CD62L⁺CD44⁻ T cells were activated with plate-bound anti-CD3 and anti-CD28 along with anti–IFN-γ, anti–IL-4, and recombinant IL-2 for 24 h. Retroviral mock vector (PBABE-neo) or Smad2 (pBABE RFP1-Smad2 neo) (Addgene) was added to the activated T cells followed by centrifugation at 1000 × g for 90 min at 32°C. After 24 h, another round of spin infection was performed. Cells were then cultured under iTreg polarizing condition for 3 d. After 3 d, cells were restimulated with PMA/ionomycin along with GolgiStop followed by intracellular staining and flow cytometry analysis of Foxp3⁺ cells.

Total RNA was extracted with RNeasy mini kit from Qiagen. Isolated RNA was converted to cDNA by using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time RT-PCR was performed with Platinum SYBR Green qPCR SuperMix-UDG with ROX and measured on StepOnePlus Real-Time PCR System (Applied Biosystems). Data were normalized to 18S rRNA.

Primer sequences were as follows: RORγT: forward (Fwd) 5′-TTTGGAACTGGCTTTCCATC-3′, reverse (Rev) 5′-AAGATCTGCAGCTTTTCCACA-3′; DUSP2: Fwd 5′-CGGTTTCAAAAGCTTCCAGA-3′, Rev 5′-CAGGTAGGGCAAGATTTCCA-3′; IL-23R: Fwd 5′-TTCAGATGGGCATGAATGTTTCT-3′, Rev 5′-CCAAATCCGAGCTGTTGTTCTAT-3′; T-bet: Fwd 5′-CCTCTTCTATCCAACCAGTATC-3′, Rev 5′-CTCCGCTTCATAACTGTGT-3′; IL-22: Fwd 5′-GTGAGAAGCTAACGTCCATC-3′, Rev 5′-GTCTACCTCTGGTCTCATGG-3′; Hif1α: Fwd 5′-AGCTTCTGTTATGAGGCTCACC-3′, Rev 5′-TGACTTGATGTTCATCGTCCTC-3′; PDK1: Fwd 5′-GGACTTCGGGTCAGTGAATGC-3′, Rev 5′-TCCTGAGAAGATTGTCGGGGA-3′; PGM1: Fwd 5′-CAGAACCCTTTAACCTCTGAGTC-3′, Rev 5′-CGAGAAATCCCTGCTCCCATAG-3′; Smad2: Fwd 5′-CTCCAGTCTTAGTGCCTCGG-3′, Rev 5′-AACACCAGAATGCAGGTTCC-3′; Smad3: Fwd 5′-TTCACTGACCCCTCCAACTC-3′, Rev 5′-CTCCGATGTAGTAGAGCCGC-3′; HMGCR: Fwd 5′-TGGTCCTAGAGCTTTCTCGTGAA-3′, Rev 5′-GGACCAAGCCTAAAGACATAATCATC-3′; FASN: Fwd 5′-TGGGTTCTAGCCAGCAGAGT-3′, Rev 5′-ACCACCAGAGACCGTTATGC-3′; HK2: Fwd 5′-TGATCGCCTGCTTATTCACGG-3′, Rev 5′-AACCGCCTAGAAATCTCCAGA-3′; Cdc42: Fwd 5′-TGCAGGGCAAGAGGATTATG-3′, Rev 5′-GATGGAGAGACCACTGAGAAA-3′; 18S: Fwd 5′-GTAACCCGTTGAACCCCATT-3′, Rev 5′-CCATCCAATCGGTAGTAGCG -3′; AldoC: Fwd 5′-AATTGGGGTGGAGAACACTG-3′, Rev 5′-TGTCAACCTTGATGCCTACG-3′; Slc2a: Fwd 5′-CAGTTCGGCTATAACACTGGTG-3′, Rev 5′-GCCCCCGACAGAGAAGATG-3′; Foxp3: Fwd 5′-GGTACACCCAGGAAAGACAG-3′, Rev 5′-ATCCAGGAGATGATCTGCTTG-3′; HMGCS: Fwd 5′-CAGGGTCTGATCCCCTTTG-3′, Rev 5′-CAGAGAACTGTGGTCTCCAGGT-3′; SQLE: Fwd 5′-GCCTCTCAGAATGGTCGTCT-3′, Rev 5′-CGCATCTCCCAGAATAAGGA-3′; ACACA: Fwd 5′-GGCCAGTGCTATGCTGAGAT-3′, Rev 5′-CCAGGTCGTTTGACATAATGG-3′; IL-4: Fwd 5′-AGATCATCGGCATTTTGAACG-3′, Rev 5′-TTTGGCACATCCATCTCCG-3′; IFN-γ: Fwd 5′-GATGCATTCATGAGTATTGCCAAGT-3′, Rev 5′GTGGACCACTCGGATGAGCTC-3′; GATA3: Fwd 5′-AGAACCGGCCCCTTATCAA-3′, Rev 5′-AGTTCGCGCAGGATGTCC-3′; DNMT3a: Fwd 5′-ACTTGGAGAAGCGGAGTGAA-3′, Rev 5′-GGATTCGATGTTGGTCTGCT-3′; and TET1: Fwd 5′-ATCATTCCAGACCGCAAGAC-3′, Rev 5′ AATCCATGCAACAGGTGACA-3′.

Cells were extracted using RIPA lysis buffer (1× PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM PMSF, and protease inhibitors). Lysates were resolved by SDS-PAGE, electrophoretically transferred onto PVDF membrane (Bio-Rad), and probed with the indicated Abs. The bands were visualized with the ECL system (Thermo Scientific).

Colitis was induced in 8–10-wk-old mice by giving the mice 2.2% DSS (m.w. 36,000–50,000) in drinking water for 5 d followed by normal water for another 5 d. Mouse body weight was measured daily. The mice were sacrificed after 10 d and the spleen, mesenteric lymph node (mLN), and colon were harvested for intracellular cytokine and Foxp3 staining.

Colitis was induced by injecting i.p. 0.6 × 10⁶ of FACS-sorted naive CD4⁺CD62L⁺CD44⁻ cells into age- and sex-matched RAG1^−/− mice. Mice were weighed daily and sacrificed upon losing 20–25% of initial body weight. The spleen, mLN, and colon were harvested for intracellular cytokine and Foxp3 staining.

The colon was harvested from the colitis mouse models into harvesting media (HBSS) with addition of glucose and 2% FBS. The colon was then opened longitudinally and cut into small pieces and incubated with constant stirring in harvesting media with addition of DTT and EDTA at 37°C for 30 min. The colon pieces were then washed twice with harvesting media with addition of EDTA and digested with collagenase in HBSS with addition of glucose and 5% FBS at 37°C for 30 min. The digested colon was then filtered with 70-μm cell strainer and centrifuged at 1500 rpm at 4°C for 5 min. The pellets were dissolved in 8 ml of 40% Percoll and slowly layered over 5 ml of 80% Percoll in a 15 ml falcon tube. The falcon tube was then centrifuged at 2000 rpm at 4°C for 20 min with acceleration of 3 and deceleration of 2. The lamina propria lymphocytes were carefully removed from the interface between 40% Percoll and 80% Percoll. The isolated lamina propria lymphocytes were washed twice with complete RPMI 1640 T cell medium, cultured with anti-CD3 (4 μg/ml) and anti-CD28 (8 μg/ml) for 5 h in the presence of Golgi plug for the last 4, and then prepared for intracellular staining.

Tissues were sectioned, fixed in 4% formaldehyde solution, embedded in paraffin, and stained with H&E. The sections were analyzed by light microscopy from Fisher Scientific Moticam at 20× magnification at room temperature. The images were acquired by using the software Motic Images Plus 2.0.

Th17 cells or iTregs were resuspended in assay medium (pH 7.4) (Sigma) containing 2 mM l-glutamine, plated onto XF24 cell culture plates precoated with Cell Tak (BD Biosciences), and incubated without CO₂ at 37°C. Extracellular acidification rate (ECAR), an indicator of aerobic glycolysis, was measured using the Seahorse XF24 Analyzer (Seahorse Bioscience) in the presence of the glycolysis substrate glucose (10 mM).

CD25^hiIL-7R^lo iTregs were sorted by FACS. A total of 3 × 10⁵ iTregs were mixed with 6 × 10⁵ naive CD4⁺CD62L⁺CD44⁻ cells from congenic BoyJ mice and transferred i.p. into age- and sex-matched RAG1^−/− mice. The recipient mice were weighed every other day for the first week and every day thereafter. The mice were sacrificed and the colon was isolated for histopathological analysis and intracellular cytokine staining.

A total of 200 ng genomic DNA was subjected to sodium bisulfite treatment and purified using the EZ DNA Methylation-Gold Kit (Zymo research, Irvine, CA) according to the manufacturer’s specifications. Two rounds of standard PCR amplification reaction were performed to amplify targeted gene fragment at an annealing temperature of 50°C before being subjected to pyrosequencing. The generated pyrograms were automatically analyzed using PyroMark analysis software (Qiagen, Valencia, CA). The pyrosequencing assay was validated using SssI-treated human genomic DNA as a 100% methylation control and human genomic DNA amplified by GenomePlex Complete Whole Genome Amplification kit (Sigma) as 0% methylation control. Primers used for bisulfite pyrosequencing are as follows: mFoxp3_assay1_NF: 5′-TTTGTGTTTTTGAGATTTTAAAATT-3′; mFoxp3_assay1_NR: /5Biosg/5′-AAAAATAACTAATCTATCCTATAACC-3′; mFoxp3_assay1_LF: 5′-TATTTTTTTGGGTTTTGGGATATTA-3′; mFoxp3_assay1_LR: 5′-ACAAATAATCTACCCCACAAATTTC-3′; mFoxp3_assay1_S2: 5′-GGGTTTTTTTGGTATTTAAGAAAG-3′; mFoxp3_assay1_S3: 5′-GGGTTTTGTATGGTAGTTAGATGG-3′; mFoxp3_assay1_S4: 5′-AGTATTTATATTATTTTATTTGGG-3′; mFoxp3_assay2_NF: 5′-GGTTATAGGATAGATTAGTTATTTTT-3′; mFoxp3_assay2_NR: /5Biosg/5′-CCAACTTCCTACACTATCTATTAAAAC-3′; mFoxp3_assay2_LF: 5′-TTTATATTATTTTATTTGGGTTTATT-3′; mFoxp3_assay2_LR: 5′-ATAACTATATAATACATCAATACATTCTCA-3′; mFoxp3_assay2_S2: 5′-GTTTTTTTTTTTTTTTTTTTGTTG-3′; mFoxp3_assay2_S3: 5′-GGTTGTGATAATAGGGTTTAGATGTAG-3′; and mFoxp3_assay2_S4: 5′-GTTTTTAAGAAATAGTTAAATAGG-3′.

A total of 2 × 10⁶ donor bone marrow cells were mixed with recipient bone marrow cells at 1:1 ratio and injected into the tail veins of lethally irradiated BoyJ mice. Eight weeks after transplantation, the chimeric mice were sacrificed, and CD45.2⁺ thymocytes and splenic T cells derived from donor bone marrow cells were analyzed.

All the statistics were performed with the two-tailed Student t test. Data were expressed as mean ± SD. p ≤ 0.05 was considered significant.

To study the role of Cdc42 in Th17 cell differentiation, Cdc42-deficient and wild-type (WT) CD4⁺ T cells were isolated from mice bearing Cdc42 deletion in T cells (Cdc42^fl/flLck^Cre) and WT mice (Cdc42^+/+Lck^Cre), respectively, and cultured under nonpathogenic Th17 differentiation condition (TGF-β + IL-6). We found that Cdc42 deficiency significantly increased IL-17 production upon 2 d culture (Supplemental Fig. 1A, 1B) or PMA/ionomycin restimulation at day 5 (Fig. 1A), suggesting that Cdc42 deficiency causes aberrant Th17 cell differentiation. Consistently, Cdc42-deficient Th17 cells expressed higher levels of RORγT (Fig. 1B), a key transcriptional factor for Th17 induction (1). Noting that RORγT expression is controlled by Stat3 (1), Stat3 activity was markedly enhanced in Cdc42-deficient Th17 cells with or without IL-6 restimulation (Fig. 1C). Conversely, DUSP2, a phosphatase inactivating Stat3 (26), was downregulated in Th17 cells lacking Cdc42 (Fig. 1D). Interestingly, under nonpathogenic Th17 differentiation condition, Cdc42 deficiency resulted in emergence of IL-17⁺IFN-γ⁺ cells and IFN-γ⁺ Th1 cells and increased secretion of IFN-γ (Fig. 1A, Supplemental Fig. 1C), suggesting that ablation of Cdc42 converts nonpathogenic Th17 cells to pathogenic Th17 cells. In support, we found that IL-23R, T-bet, and IL-22, three important signature genes of pathogenic Th17 cells (3), were substantially elevated in Cdc42-deficient Th17 cells (Fig. 1E). These data suggest that Cdc42 restrains aberrant Th17 differentiation and Th17 pathogenicity.

Glycolysis has recently been shown to be essential for Th17 cell differentiation (27). We found that Cdc42 deletion led to an increased glycolysis in Th17 cells (Fig. 2A). In agreement with this, Hif1α, PDK1, PGM1, HK2, AldoC, and Slc2a that indirectly (e.g., Hif1α, PDK1, PGM1, Slc2a) or directly (e.g., HK2, AldoC) regulate glycolysis were upregulated in Cdc42-deficient Th17 cells (Fig. 2B). Importantly, inhibition of glycolysis by 2-DG reversed IL-17 expression in pathogenic Cdc42-deficient Th17 cells (Fig. 2C). Of note, 2-DG also suppressed IL-17 expression in nonpathogenic WT Th17 cells (Fig. 2C), supporting the essential role of glycolysis in Th17 cell differentiation. Interestingly, upon 2-DG treatment, IL-17 in Cdc42-deficient Th17 cells was decreased to comparable levels to that in WT Th17 cells but not to that in 2-DG–treated WT Th17 cells (Fig. 2C). Furthermore, 2-DG treatment reduced the transdifferentiation of Cdc42-deficient Th17 cells to IL-17⁺IFN-γ⁺ cells and IFN-γ⁺ cells (Fig. 2C). These data suggest that inhibition of glycolysis diminished aberrant differentiation and pathogenicity of Cdc42-deficient Th17 cells. Lipid metabolism may play a role in Th17 differentiation (28). Consistent with the increased Th17 differentiation, we found that Cdc42 deficiency upregulated lipid metabolism regulators HMGCS, FASN, and SQLE in Th17 cells (Supplemental Fig. 1D).

Because Cdc42-deficient Th17 cells became pathogenic, we next examined whether they had functional impact in vivo on colitis development in which pathogenic Th17 cells play an important role (4, 29). As expected, Cdc42-deficient mice lost more weight compared with WT mice, upon treatment with DSS (Fig. 3A), a well-known regimen to model human colitis in mice (30). This is associated with more IL-17⁺ cells in the colon (Fig. 3B). Furthermore, in support of a recent report that Th17 cells give rise to Th1 cells in the pathogenesis of colitis (29), we found increased IFN-γ⁺ cells in DSS-treated Cdc42-deficient mice (Fig. 3B). To substantiate these results, we adopted a naive T cell transfer-induced mouse model of colitis in which naive CD4⁺ T cells are transferred into RAG1^−/− mice to induce colitis (30). As shown in Fig. 3C, both WT and Cdc42 knockout naive T cells induced weight loss in RAG1^−/− mice starting at about two and half weeks after cell transfer. Nonetheless, RAG1^−/− mice receiving Cdc42 knockout naive T cells suffered from significantly more weight loss 5 wk later. The severe wasting disease in the mice receiving Cdc42 knockout naive T cells is associated with increased IL-17A⁺, IFN-γ⁺, and IL-17A⁺IFN-γ⁺ cells in the spleen and/or mLNs (Fig. 3D). Surprisingly, IL-17⁺, IFN-γ⁺ and IL-17A⁺IFN-γ⁺ cells in lamina propria were decreased or tended to decrease (Fig. 3D), which might reflect a negative feedback effect. Taken together, these results suggest that Cdc42 prevents colitis development through suppression of Th17 differentiation and pathogenicity.

We next examined the effect of Cdc42 deficiency on iTreg differentiation. As shown in Fig. 4A, deletion of Cdc42 dampened iTreg differentiation as Foxp3⁺ cells were considerably reduced when Cdc42-deficient CD4⁺ T cells were differentiated with TGF-β and IL-2. Interestingly, under iTreg differentiation condition, there was a significant increase in Foxp3⁺IL-17⁺, IL-17⁺, IL-17⁺IFN-γ⁺, and IFN-γ⁺ cells in the absence of Cdc42 (Fig. 4A, 4B), suggesting that Cdc42 deficiency caused iTreg instability. In addition, Cdc42 deficiency-induced defect in iTreg differentiation was mirrored in vivo. Thus, DSS-treated Cdc42-deficient mice and RAG1^−/− mice transferred with naive CD4⁺ Cdc42-deficient T cells showed fewer Foxp3⁺ cells in either their lamina propria, spleen, or mLN (Fig. 4C, 4D).

Mechanistically, Cdc42 deficiency led to an attenuated activation and mRNA/protein expression of transcriptional factors Smad2 and Smad3 that are known to be induced by TGF-β and are important for Foxp3 expression in iTregs (5) (Fig. 5A, 5B). Reconstitution of Smad2 into Cdc42-deficient iTregs partially restored their differentiation (Fig. 5C, Supplemental Fig. 1E, 1F), suggesting that the decreased Smad2 expression in Cdc42-deficient iTregs contributes to the impaired differentiation and that Smad2 is necessary for Cdc42 to regulate iTreg differentiation. Noting that overexpression of Smad2 in WT iTregs did not enhance iTreg differentiation (Fig. 5C, Supplemental Fig. 1E, 1F), it appears that Smad2 is not sufficient to promote iTreg differentiation. In contrast to its inhibitory effect on Smad2/3 expression/activation, Cdc42 deficiency in iTregs had no effect on IL-2–induced Stat5 activation (Fig. 5D). Lipid metabolism and glycolysis have been indicated in Treg differentiation and/or stability (11, 12, 31–34). We found that although Cdc42 deficiency in iTregs had no effect on the expression of lipid metabolism enzymes HMGCS and ACACA, it inhibited HMGCR and FASN expression (Fig. 5E), suggesting a defective lipid metabolism in Cdc42-deficient iTregs. Cdc42 deficiency also dampened glycolysis and the expression of glycolysis mediators Hif1α and HK2 in iTregs (Fig. 5F, 5G). Importantly, Cdc42 deficiency-induced increase in Foxp3⁺IL-17⁺ cells was abolished by addition of glycolysis-derived pyruvate into culture medium of Cdc42-deficient iTregs (Fig. 5H), an approach mimicking restoration of glycolysis through bypassing the intermediate steps of glycolysis (35). In contrast, Cdc42 deficiency-induced decrease in Foxp3⁺ cells was not reversed by addition of pyruvate (Fig. 5H). These data suggest that Cdc42-mediated glycolysis is important for iTreg stability but not differentiation. To further demonstrate the importance of glycolysis in maintaining iTreg stability, we treated WT iTregs with 2-DG and found that 2-DG caused iTreg instability as evidenced by reduced Foxp3⁺ cells and concomitantly increased IL-17⁺ cells (Fig. 5I). Collectively, our data suggest that Cdc42 is required for iTreg differentiation and stability through coordinating transcriptional and metabolic signaling.

We then evaluated functional capacity of Cdc42-deficient iTregs. To this end, WT or Cdc42-deficient iTregs were transferred together with congenic naive CD4⁺ T cells from BoyJ mice into RAG1^−/− mice. We found that although WT iTregs abolished naive T cell transfer-induced weight loss, Cdc42-deficient iTregs failed to suppress the wasting disease (Fig. 6A). The loss of function of Cdc42-deficient iTregs in control of colitis was associated with their inability to suppress infiltration of inflammatory cells (Fig. 6B), particularly IL-17– and IFN-γ–producing cells (Fig. 6C), into the colon.

The observation that Cdc42 deficiency caused an enhanced Th17 cell differentiation and pathogenicity in vitro and in induced colitis models prompted us to examine whether Th17 cells were spontaneously developed in steady-state Cdc42^fl/flLck^Cre mice. Indeed, Cdc42^fl/flLck^Cre mice had more IL-17–producing cells compared with control mice (Supplemental Fig. 1G). And these cells appeared to be pathogenic as they inclined to develop to IFN-γ–producing cells (Supplemental Fig. 1G). Brief restimulation of splenocytes from Cdc42^fl/flLck^Cre mice in vitro also induced more IL-17⁺ cells than that from WT mice (Supplemental Fig. 1H). However, Cdc42^fl/flLck^Cre mice did not appear to have spontaneous autoimmune diseases (data not shown) and apparent pathological abnormalities in the spleen, kidney, liver, and colon (Supplemental Fig. 1I). This is likely due to a compensatory increase in the frequency but not the absolute numbers of nTregs (Supplemental Fig. 1J, 1K) (24).

The compensatory increase in nTregs in Cdc42^fl/flLck^Cre mice precludes us from studying nTreg autonomous role of Cdc42 in these mice. To circumvent this hurdle, we achieved Treg-specific deletion of Cdc42 by crossing Cdc42^fl/fl mice with Foxp3^YFP-Cre mice. The resultant Cdc42^fl/flFoxp3^YFP-Cre mice showed effective Cdc42 deletion in their nTregs (Fig. 7A, Supplemental Fig. 2A for nTreg flow-sorting strategy). Strikingly, Cdc42^fl/flFoxp3^YFP-Cre mice were small in size, lacked mobility, and developed a hunched posture, crusting of the ears, eyelids and tail, and skin ulceration, before they became moribund within ∼3–5 wk after birth (Fig. 7B, data not shown). Moreover, the mice showed splenomegaly and lymphadenopathy (Fig. 7C), massive leukocyte infiltration and/or distorted architecture in the colon, kidney, lung, and skin (Fig. 7D), and increased autoantibodies (e.g., anti-dsDNA, antinuclear Ab) (Fig. 7E). Such severe phenotypes are reminiscent of that observed in mice with the scurfy mutation of Foxp3 gene and mice depleted of Tregs (36–38). The systemic inflammatory diseases in Cdc42^fl/flFoxp3^YFP-Cre mice are associated with altered T cell homeostasis. As such, the mice showed more activated T cells (CD62L^loCD44^hi) in their CD4⁺ and CD8⁺ compartments (Fig. 7F) and more effector T cells (e.g., IL-17–producing Th17, IFN-γ–producing Th1, IL-4–producing Th2) (Fig. 7G). Consistently, these mice had increased cells expressing Th1 and Th2 signature transcriptional factor T-bet and GATA3, respectively (Supplemental Fig. 2B). On the other hand, Cdc42^fl/flFoxp3^YFP-Cre mice had fewer Foxp3⁺ cells (Fig. 7H), suggesting that the immune activation in Cdc42^fl/flFoxp3^YFP-Cre mice is likely due to the decreased nTregs.

To explore the mechanisms underlying the loss of nTregs in Cdc42^fl/flFoxp3^YFP-Cre mice, we measured nTreg proliferation and apoptosis. nTregs from Cdc42^fl/flFoxp3^YFP-Cre mice showed less BrdU incorporation (Fig. 8A) and more Annexin V staining (Fig. 8B, Supplemental Fig. 2C), compared with that from control Cdc42^+/+Foxp3^YFP-Cre mice, suggesting that Treg-specific deletion of Cdc42 dampens nTreg proliferation and survival, which likely contributes to the reduction in nTregs in Cdc42^fl/flFoxp3^YFP-Cre mice.

The impaired nTreg homeostasis in Cdc42^fl/flFoxp3^YFP-Cre mice could also have resulted from nTreg instability. Indeed, Cdc42^fl/flFoxp3^YFP-Cre nTregs became plastic and gained effector T cell programs as they produced more effector T cell cytokines IL-17, IFN-γ, and IL-4 and expressed more effector T cell transcriptional factors RORγT, T-bet, and GATA3 than by Cdc42^+/+Foxp3^YFP-Cre nTregs (Fig. 8C, Supplemental Fig. 3A–D). Moreover, Cdc42^fl/flFoxp3^YFP-Cre nTregs had lower expression of Foxp3, although they exhibited intact CD25 expression (Fig. 8D). And in vitro culture of purified Cdc42^fl/flFoxp3^YFP-Cre and Cdc42^+/+Foxp3^YFP-Cre nTregs with IL-2 (39) found that Cdc42 deficiency inhibited IL-2–induced expansion of nTregs and diminished Foxp3 expression (Fig. 8E). As reduced Foxp3 expression is known to convert Tregs to effector T cells (40), we next examined if Cdc42 deficiency caused conversion of Cdc42^fl/flFoxp3^YFP-Cre nTregs to effector T cells. To this end, we cultured purified Cdc42^fl/flFoxp3^YFP-Cre and Cdc42^+/+Foxp3^YFP-Cre nTregs in vitro with IL-6/IL-1 or IL-12 (10). We found that Cdc42^fl/flFoxp3^YFP-Cre nTregs were more susceptible than Cdc42^+/+Foxp3^YFP-Cre nTregs to convert to IL-17–producing effector T cells upon IL-6/IL-1 culture (Fig. 8F) or IFN-γ–producing effector T cells upon IL-12 culture (Fig. 8G). In support of the increased transdifferentiation capacity of Cdc42^fl/flFoxp3^YFP-Cre nTregs, we detected Cdc42 knockout allele in CD4⁺Foxp3⁻ non-Tregs from Cdc42^fl/flFoxp3^YFP-Cre mice (Fig. 8H). Because Foxp3^YFP-Cre presumably induces gene deletion specifically in Tregs, the appearance of Cdc42 deletion in Foxp3⁻ population suggests that these Cdc42-deficient non-Tregs are converted from Cdc42-deficient nTregs. Treg activation has recently been shown to lead to their instability (41). Consistently, Cdc42^fl/flFoxp3^YFP-Cre nTregs upregulated their activation/functional markers CD44, CD69, glucocorticoid-induced TNFR-related protein (GITR), programmed cell death protein 1 (PD-1), and cytotoxic T lymphocyte–associated protein 4 (CTLA-4) with concomitant downregulation of CD62L (Fig. 8I, 8J). Hypomethylation of the CNS2 region in the Foxp3 enhancer is important for maintenance of Foxp3 expression and thus Treg stability (10, 12). As reported (10, 12, 42), CD4⁺ naive and activated T cells from Cdc42^+/+Foxp3^YFP-Cre mice showed hypermethylation and nTregs from Cdc42^+/+Foxp3^YFP-Cre mice showed hypomethylation in all of the 14 methylation sites in the CNS2 region of the Foxp3 gene. In agreement with the loss of Foxp3 expression and instability, nTregs from Cdc42^fl/flFoxp3^YFP-Cre mice showed significantly more methylation in these methylation sites, compared with Cdc42^+/+Foxp3^YFP-Cre nTregs (Fig. 8K), which is associated with upregulated DNMT3A, a DNA methyltransferase, but not TET1, a DNA demethylase (Supplemental Fig. 3E). Lastly, Cdc42 deficiency in nTregs dampened signaling involved in lipid metabolism (Supplemental Fig. 3F) and glycolysis (Supplemental Fig. 3G) that have been implicated in regulating Treg stability (11). Importantly, although Cdc42 deficiency enhanced IL-12–induced transdifferentiation of nTregs to IFN-γ–producing effector T cells, this effect of Cdc42 deficiency was partially reversed by glycolysis-derived pyruvate (Supplemental Fig. 3H). Collectively, our data suggest that Cdc42 deficiency destabilizes nTregs, leading to their transdifferentiation to effector T cells.

As inflammatory disorders may impair Treg stability (43–45), the instability of Cdc42^fl/flFoxp3^YFP-Cre nTregs could be due to the severe inflammation in Cdc42^fl/flFoxp3^YFP-Cre mice. To test this, we analyzed Cdc42^fl/flFoxp3^YFP-Cre/+ female mice that are heterozygous for Foxp3^YFP-Cre and did not show overt inflammation (data not shown). Because of the X chromosome–linked nature of and random X chromosome inactivation by Foxp3^YFP-Cre knock-in transgene, Foxp3^YFP-Cre/+ female mice should maintain 50% of Foxp3⁺YFP⁺ nTregs and 50% of Foxp3⁺YFP⁻ nTregs (14). This was the case in Cdc42^+/+Foxp3^YFP-Cre/+ female mice in which Foxp3⁺YFP⁺:Foxp3⁺YFP⁻ cells were ∼1:1 (Fig. 9A, Supplemental Fig. 4A for gating strategy). However, the ratio of Foxp3⁺YFP⁺ versus Foxp3⁺YFP⁻ cells in Cdc42^fl/flFoxp3^YFP-Cre/+ female mice was significantly reduced (Fig. 9A). BrdU incorporation and active caspase 3 staining assays revealed decreased proliferation and increased apoptosis in Foxp3⁺YFP⁺ cells from Cdc42^fl/flFoxp3^YFP-Cre/+ female mice, compared with that in Foxp3⁺YFP⁺ cells from Cdc42^+/+Foxp3^YFP-Cre/+ female mice (Fig. 9B, 9C), which might contribute to the decreased ratio of Foxp3⁺YFP⁺ versus Foxp3⁺YFP⁻ cells in Cdc42^fl/flFoxp3^YFP-Cre/+ female mice. Because Foxp3⁺YFP⁺ nTregs from Cdc42^fl/flFoxp3^YFP-Cre/+ female mice presumably lack Cdc42 expression and Foxp3⁺YFP⁻ nTregs from the same mice express Cdc42, our data suggest that Cdc42-deficient Foxp3⁺YFP⁺ nTregs were outcompeted by Cdc42-sufficient Foxp3⁺YFP⁻ nTregs in Cdc42^fl/flFoxp3^YFP-Cre/+ female mice. The uncompetitiveness of Foxp3⁺YFP⁺ nTregs in Cdc42^fl/flFoxp3^YFP-Cre/+ female mice is associated with their instability as evidenced by increased expression of Th17, Th1, and Th2 cytokines IL-17, IFN-γ, and IL-4, respectively (Fig. 9D), decreased expression of Foxp3 (Fig. 9E), and increased expression of Treg activation/functional markers GITR, PD-1, CTLA-4, and ICOS (Fig. 9F), compared with Foxp3⁺YFP⁻ nTregs from the same mice. The instability of Cdc42-deficient nTregs in noninflammatory Cdc42^fl/flFoxp3^YFP-Cre/+ female mice suggests that the instability of Cdc42-deficient nTregs in Cdc42^fl/flFoxp3^YFP-Cre mice is not due to continuing inflammation but a cell-intrinsic effect. In further support of this, a competitive bone marrow transplantation assay found that the spleen of recipient mice of Cdc42^fl/flFoxp3^YFP-Cre bone marrow had fewer donor-derived nTregs but more Foxp3⁺IFN-γ⁺ and Foxp3⁺IL-4⁺ cells (Fig. 9G).

The reduction in nTregs in Cdc42^fl/flFoxp3^YFP-Cre mice could also have been an outcome of impaired nTreg differentiation in the thymus. To test this, we analyzed thymocyte development in Cdc42^fl/flFoxp3^YFP-Cre mice. We found that although Cdc42^fl/flFoxp3^YFP-Cre mice had very small thymus (Supplemental Fig. 4B) and defective thymocyte development (e.g., decreased frequency of CD4⁺CD8⁺ thymocytes; increased frequency of CD4⁻CD8⁻, CD4⁺, and CD8⁺ thymocytes) (Supplemental Fig. 4C), the frequency of nTregs was increased (Supplemental Fig. 4D). Nonetheless, Foxp3 expression was not altered (Fig. 10A). These data suggest that Cdc42 deficiency has no effect on nTreg differentiation and that the altered thymocyte development likely results from the severe inflammation and the increased frequency of nTregs might be a proportional increase resulting from the increased frequency of CD4⁺ thymocytes. In support, noninflammatory Cdc42^fl/flFoxp3^YFP-Cre/+ female mice had normal thymocyte development (data not shown) and normal ratio of Foxp3⁺YFP⁺ versus Foxp3⁺YFP⁻ cells in the thymus (Fig. 10B). Furthermore, a competitive bone marrow transplantation assay found that the thymus of recipient mice of Cdc42^fl/flFoxp3^YFP-Cre bone marrow had normal frequency of donor-derived Foxp3⁺ cells showing normal Foxp3 expression (Fig. 10C, 10D).

A delicate balance between effector T cells and Tregs is important for mounting protective immune responses to foreign Ags without losing immune tolerance to self-antigens. Understanding the molecular mechanisms by which this balance is regulated is thus of paramount importance. In the current study, we demonstrate that Cdc42 is pivotal for maintaining Th17/Treg balance. This conclusion is supported by several observations: 1) genetic deletion of Cdc42 in T cells enhances Th17 differentiation but diminishes iTreg differentiation in culture under Th17 differentiation condition and iTreg differentiation condition, respectively; 2) Under Th17 differentiation condition, Cdc42 deficiency impairs the generation of Foxp3⁺ cells (data not shown), whereas under iTreg differentiation condition, Cdc42 deficiency increases IL-17–producing cells; 3) RAG1^−/− mice transferred with Cdc42-deficient naive T cells show increased Th17 cells but decreased iTregs, compared with that transferred with WT naive T cells; and 4) Treg-specific Cdc42 deletion leads to decreased nTregs but increased Th17 cells. Along with our previous finding that Cdc42 deficiency enhances Th1 cell differentiation (24), this study also suggests that Cdc42 plays an important role in maintaining the balance between Tregs and Th1 or Th2 cells.

Cdc42-deficient Th17 cells become pathogenic, which is attributed to increased glycolysis, because inhibition of glycolysis by 2-DG in Cdc42-deficient Th17 cells suppresses aberrant IL-17 and IFN-γ production. Interestingly, 2-DG–treated Cdc42-deficient Th17 cells still show WT level of IL-17 production, whereas 2-DG–treated WT Th17 cells have decreased IL-17 production as previously reported (27). It thus appears that although glycolysis promotes Th17 differentiation, Cdc42 controls the magnitude of glycolysis to prevent aberrant/pathogenic Th17 cell differentiation. Noting that Cdc42-deficient Th17 cells upregulate Hif1α, deletion of which has been shown to inhibit Th17 glycolysis and differentiation (27), it is plausible that Cdc42 restrains glycolysis and its regulated Th17 differentiation through suppression of Hif1α expression. Cdc42-deficient iTregs become unstable. The instability, but not the defective differentiation, of Cdc42-deficient iTregs is attributed to decreased glycolysis, suggesting that Cdc42-promoted glycolysis is important for iTreg stability but not differentiation. As Cdc42-deficient iTregs show decreased Hif1α, we envision that Cdc42 promotes glycolysis and its regulated iTreg stability through inducing Hif1α expression.

It remains unknown why the effects of Cdc42 deficiency are opposite between Th17 cells and iTregs. Of note, strong TCR signaling has been shown to promote Th17 cell differentiation but suppress iTreg differentiation (46–48). In this context, the enhanced Th17 differentiation and the diminished iTreg differentiation upon Cdc42 deletion may be attributed to heightened TCR signaling. Indeed, we have previously shown that Cdc42 deficiency leads to sustained ERK activation in activated T cells (23). Interestingly, Cdc42 deficiency also augments mTOR signaling in both Th17 and iTregs (data not shown). As mTOR is known to promote Th17 differentiation but suppress iTreg differentiation (49), the increased mTOR signaling may contribute to the agonistic effect of Cdc42 deficiency on Th17 differentiation and to the antagonistic effect of Cdc42 deficiency on iTreg differentiation. Moreover, given that mTOR is an essential metabolic sensor and is important for glycolysis in Th17 cells (27), the increased mTOR signaling may also contribute to the upregulated glycolytic signaling in Cdc42-deficient Th17 cells and to the downregulated glycolytic signaling in Cdc42-deficient iTregs. In addition to TCR signaling strength, the different skewing cytokines (IL-6 + TGF-β versus IL-2 + TGF-β) that drive Th17 versus iTreg differentiation may contribute to the opposite effects of Cdc42 deletion on Th17 cells versus iTregs. We speculate that on the one hand, IL-6 + TGF-β induce Th17 cell differentiation through upregulation of glycolysis; on the other hand, they activate Cdc42 to restrain glycolysis in a negative feedback manner to prevent aberrant/pathogenic Th17 cell differentiation. In iTregs, IL-2 + TGF-β activate Cdc42 to promote iTreg differentiation independent of glycolysis and to stabilize differentiated iTregs via upregulation of glycolysis.

Treg-specific deletion of Cdc42 by Foxp3^YFP-Cre causes early, fatal inflammatory disorders, suggesting that Cdc42 is a bona fide gatekeeper of peripheral tolerance. We have found that the inflammatory disorders in Cdc42^fl/flFoxp3^YFP-Cre mice are caused by increased effector T cells and decreased nTregs. The decrease in nTregs is due to impaired proliferation, survival, and stability but not differentiation. The instability of Cdc42^fl/flFoxp3^YFP-Cre nTregs is evidenced by downregulation of Foxp3 and upregulation of transcription factors and cytokines characteristic of effector T cells. As reduced Foxp3 expression is known to convert Tregs to effector T cells (40), at least a proportion of the increased effector T cells in Cdc42^fl/flFoxp3^YFP-Cre mice could have been derived from Cdc42^fl/flFoxp3^YFP-Cre nTregs. In support, we show that Cdc42^fl/flFoxp3^YFP-Cre nTregs are more susceptible to transdifferentiate to Th17 and Th1 cells in vitro, compared with Cdc42^+/+Foxp3^YFP-Cre nTregs. Meanwhile, the appearance of Cdc42^fl/flFoxp3^YFP-Cre nTregs expressing effector T cell markers might have exacerbated effector T cell functions and thus lymphoproliferative diseases in Cdc42^fl/flFoxp3^YFP-Cre mice, given that Tregs expressing effector T cell markers are implicated in autoimmune diseases such as inflammatory bowel diseases (4).

We provide evidences to indicate that the instability of Cdc42^fl/flFoxp3^YFP-Cre nTregs could have resulted from their activation, the Foxp3 enhancer hypermethylation, and/or attenuated glycolysis. However, although Cdc42^fl/flFoxp3^YFP-Cre nTregs upregulate their activation/functional markers in a cell-intrinsic manner, the activation phenotype of these cells might just reflect a compensatory effect because of the reduced frequency. On the other hand, by pyruvate rescuing, we show that the decreased glycolytic activity contributes to the instability of Cdc42^fl/flFoxp3^YFP-Cre nTregs, similar to that observed in Cdc42-deficient iTregs. Thus, our data depict that weakened glycolysis can lead to Treg instability. This is in contrast to a couple of recent studies that have associated heightened glycolysis to Treg instability (11, 12, 50). We reason that physiologic level of glycolysis is important for maintaining Treg stability. As deletion of Cdc42 links weakened glycolysis to Treg instability and deletion of PTEN or ATG7 links heightened glycolysis to Treg instability (11, 12, 50), we postulate that Cdc42 upregulates glycolysis to promote Treg stability, whereas PTEN/ATG7 controls magnitude of Cdc42-induced glycolysis to prevent hyperglycolysis-induced Treg destabilization.

In summary, we have identified Cdc42 as a master regulator of the balance between Th17 and iTregs or nTregs. Cdc42 controls Th17/iTreg balance by inhibiting aberrant differentiation/pathogenicity of Th17 cells and promoting differentiation/stability/suppressive function of iTregs. Cdc42 controls Th17/nTreg balance by maintaining nTreg homeostasis through regulating nTreg proliferation/survival/stability. Cdc42 may do so by integrating transcriptional and metabolic signaling in Th17 cells and Tregs. Consequently, T cell–specific deletion of Cdc42 leads to exacerbated wasting disease in mouse models of colitis and Treg-specific deletion of Cdc42 causes spontaneous, early fatal lymphoproliferative diseases. Therefore, interference of Cdc42 pathway in T cells might open a new avenue in autoimmune disease therapy and in cancer immunotherapy. Particularly, the observation that inhibition of glycolysis in Cdc42-deficient Th17 cells diminished their pathogenicity but not WT level of differentiation suggests that targeting Cdc42-regulated glycolysis selectively inhibits pathogenic but not protective Th17 cells and therefore could have important implications in the clinical therapy of Th17-mediated immune diseases.

We thank Veda Yadagiri and Hong Ji at the Cincinnati Children’s Hospital Research Foundation Pyrosequencing Core for analysis of the methylation of Foxp3 gene and for writing the related methods.

The authors have no financial conflicts of interest.