Arf1 and Arf6 Synergistically Maintain Survival of T Cells during Activation

The ADP-ribosylation factor (Arf) proteins are small GTPases belonging to the Ras superfamily and orchestrate intracellular protein trafficking under the control of their activators, guanine nucleotide exchange factors (GEFs), and inhibitors, GTPase-activating proteins (GAPs) (1, 2). Mice have six Arf isoforms (Arf1–Arf6), whereas Arf2 has been lost in humans (2). Among Arf family proteins, Arf1 and Arf6 are best characterized (3). Initial studies demonstrate that Arf1 regulates formation of coated vesicles mainly at the Golgi apparatus, as do Arf2-5, whereas Arf6, the most divergent isoform from Arf1, functions at the plasma membrane (4–7). Several reports have demonstrated, however, that both Arf1 and Arf6 facilitate migration of cancer cells by regulating integrin localization (8–10), indicating that Arf1 and Arf6 play overlapping roles at least in some circumstances. It has also been shown that Arf1 as well as Arf6 is involved in the activation of mTOR via PLD-mediated phosphatidic acid generation (11–13). Moreover, by using tissue-specific gene knockout strategy, Arf1 and Arf6 are revealed to play a critical role in the nervous system through Schwann cell myelination (14, 15). It is of interest to note that Arf6-GAP SMAP1 and Arf1-GAP SMAP2 synergistically regulate early embryonic development (16), also suggesting the genetic linkage between Arf1 and Arf6 during initial steps in ontogeny. In contrast, physiological roles of Arf family proteins in the immune system remain obscure.

Given that the intracellular protein trafficking system controls a diverse array of cellular responses, including proliferation, differentiation, and cell migration, one can assume that this is also the case with the immune responses. In fact, Rab family of small GTPases modulates TCR signal transduction through TCR recycling as well as trafficking various components to the immunological synapse (17, 18). In addition, Rab27a regulates the secretion of the lytic granules from cytotoxic T lymphocytes (19), and mutations in RAB27A gene are known to be associated with Griscelli syndrome type II in human (20). It has also been shown that CRACR2A, another member of Rab family, is involved in T cell activation by regulating vesicle trafficking to the immunological synapse (21). As is the case with Rab family, it seems likely that the Arf pathway contributes to T cell function as well. Accordingly, brefeldin A, a potent inhibitor of ArfGEFs, blocks the cytokine secretion from activated T cells, raising the possibility that Arf family proteins play important roles during T cell activation (22, 23).

To unmask the functional role of the Arf pathway in T cell, we established T lineage–specific Arf-deficient mice and found that the lack of both Arf1 and Arf6 renders T cells susceptible to apoptosis during activation. In contrast, contrary to our expectation, Arf deficiency had no impact on cytokine secretion. We also show that the Arf-deficient mice are resistant to Th17-mediated diseases while Ab responses are intact.

All mouse strains used in this study were backcrossed to C57BL/6 for at least seven generations. The floxed Arf1 (Arf1^fl/fl) mouse line (accession no. CDB1027K; http://www2.clst.riken.jp/arg/micelist.html) was established (detailed characterization was described in Supplemental Fig. 1) and crossed with Lck-Cre (24) along with Arf6^fl/fl mice (14). Age-matched Arf1^+/+: Arf6^+/+ and Lck-Cre mice or cohoused Arf1^fl/+: Arf6^fl/+: Lck-Cre and Arf1^fl/fl: Arf6^fl/fl littermates were used as a control. Rag2^−/− mice (stock number RAGN12) were obtained from Taconic. Mice systemically expressing tdRFP (referred to in this article as RFP mice) were derived from Rosa26-tdRFP reporter mice (kindly provided by H. J. Fehling, University of Ulm, Ulm, Germany) (25). We used male and female mice aged between 7 and 12 wk in all experiments unless otherwise stated. All mice were maintained in a specific pathogen-free facility and used according to our institutional guidelines.

Naive (CD44^loCD62L^hiCD25⁻) or effector (CD44^hiCD62L^loCD25⁻) CD4⁺ T cells were obtained from splenocytes by using a MojoSort Mouse CD4 T Cell Isolation Kit (BioLegend), followed by cell sorting on a FACSAria III (BD Biosciences). RPMI 1640 (Fujifilm) supplemented with 10% FCS (HyClone), 55 μM 2-ME, 100 U/ml penicillin, 100 μg/ml streptomycin, nonessential amino acids, 1 mM sodium pyruvate, and 10 mM HEPES was used as a culture medium. Cytokines used in this study were purchased from BioLegend. Naive CD4⁺ T cells were stimulated either with immobilized anti-CD3ε mAb (5 μg/ml) and 1 μg/ml soluble anti-CD28 mAb or 10 ng/ml IL-7 in 24-well plates. In some experiments, T cells were activated in the presence of 20 ng/ml IL-2, 20 ng/ml IL-4, 20 ng/ml IL-7, 50 ng/ml IL-21, or 50 μM Z-VAD-FMK (MBL). In vitro Th differentiation was induced with 10 ng/ml IL-12 (Th1), 10 ng/ml IL-4 (Th2), 10 ng/ml IL-2 and 20 ng/ml TGF-β (regulatory T cell [Treg]), 30 ng/ml IL-6 and 3 ng/ml TGF-β (nonpathogenic Th17), or 30 ng/ml IL-6, 50 ng/ml IL-23, 30 ng/ml IL-1β, and 1 μg/ml anti–IL-2 Ab (pathogenic Th17), followed by stimulation with Dynabeads Mouse T-Activator CD3/CD28 (Thermo Fisher Scientific). To isolate lamina propria (LP) lymphocytes, the colon was digested with HBSS containing 1.3 mM CaCl₂, 0.5 mM MgCl₂, Liberase TM (200 μg/ml; Roche Diagnostics), and DNase I (10 μg/ml; Roche Diagnostics) after removing intraepithelial lymphocytes with HBSS containing 5 mM EDTA and 1 mM DTT. LP lymphocytes were obtained from the 40/80% Percoll (GE Healthcare) interface after centrifugation of the digested colon at 920 × g for 20 min at 25°C.

Total RNA was extracted using RNeasy Micro Kit (QIAGEN), followed by PrimeScript RT Master Mix (Takara) reaction. THUNDERBIRD SYBR qPCR Mix (Toyobo) was used to evaluate gene expression on a Rotor-Gene Q (QIAGEN). Primers used were as follows: Arf1, 5′-ACAGAGAGCGTGTGAACGAG-3′ and 5′-TGGCCTGAATGTACCAGTTC-3′; Arf6, 5′-TCCTAATGAGCGTCCTCCAC-3′ and 5′-TCCTAGGAATGGGTTTTGGA-3′; and Cyclophilin A, 5′-ATGGCACTGGCGGCAGGTCC-3′ and 5′-TTGCCATTCCTGGACCCAAA-3′.

Concentrations of IL-2, IFN-γ, IL-4, and IL-17A in culture supernatants (1 × 10⁵ cells/ml) were detected with ELISA Max Deluxe (BioLegend). To evaluate OVA-specific Ab response, mice were primed i.p. with 100 μg of OVA (Sigma-Aldrich) along with Imject Alum adjuvant (Thermo Fisher Scientific) and boosted 14 d later. Alternatively, mice were immunized i.p. with 25 μg of OVA along with Sigma Adjuvant System (SAS; Sigma-Aldrich). On day 28, sera were subjected to ELISA in 96-well plate coated with 10 μg/ml OVA along with HRP-linked anti-mouse IgG F(ab′)₂ Ab (GE Healthcare). Fecal extract was obtained with 20% w/v PBS containing 2 mM PMSF and 0.2 mg/ml benzamidine, and IgA was quantified by ELISA using SBA Clonotyping System-C57BL/6-HRP (SouthernBiotech).

Abs used in this study were as follows: Mcl-1 (no. 600-401-394) from Rockland Immunochemicals; Erk2 (D2) from Santa Cruz Biotechnology; CD4 (GK1.5), CD8 (53-6.7), and CD62L (MEL-14) from Tonbo Biosciences; CD44 (IM7), CD25 (PC61), IL-17A (TC11-18H10.1), CD98 (RL388), CD71 (RI7217), CD45.1 (A20), CCR7 (4B12), and annexin V (no. 640912) from BioLegend; Foxp3 (MF23) and CXCR5 (2G8) from BD Biosciences; PD-1 (RMPI-30) from eBioscience; HSP90 (no. 4874), phospho-S6 ribosomal protein (Ser^235/236) (D57.2.2E), Bim (no. 2819), Bcl-2 (D17C4), and Bcl-x_L (54H6) from Cell Signaling Technology.

Isolated cells were stained with the appropriate Abs along with 7-aminoactinomycin D (7-AAD; Sigma-Aldrich). To evaluate cell proliferation, cells (2.5 × 10⁶ cells/ml in PBS) were labeled with 3 μM Cell Proliferation Dye eFluor 450 (eF450; eBioscience) at 37°C for 5 min, followed by washing with the culture medium. For intracellular staining of cytokines and transcription factor, cells were stimulated with PMA and ionomycin in the presence of brefeldin A (eBioscience) for 4 h, followed by treatment with Fixation and Permeabilization Buffer (eBioscience) or Cytofix/Cytoperm (BD Biosciences). Glucose uptake was assessed by using 2-NBDG (Thermo Fisher Scientific) (26). Reactive oxygen species (ROS) were measured by incubation for 30 min at 37°C with 5 μM CellROX Deep Red (Thermo Fisher Scientific) after staining of surface makers. To assess the impact of ROS on cell survival, CD4⁺ T cells were stimulated with or without 10 mM N-acetyl-l-cysteine (NAC) for 3 d, followed by evaluation with annexin V and 7-AAD staining according to the manufacturer’s instructions (BioLegend). For the analysis of the sub-G1 population, cells were fixed and permeabilized with 75% ethanol and stained with 10 μg/ml 7-AAD. Data were acquired on an FACSCanto II (BD Biosciences) and analyzed with FlowJo Software (Tree Star).

The spleen, mesenteric lymph node (MLN), and Peyer patches (PPs) were fixed with 1% paraformaldehyde. After fixation, tissues were equilibrated gradually with 10, 20, and 30% sucrose in PBS at 4°C, embedded in O.C.T. compound (Sakura Finetek), and frozen at −80°C. Frozen sections (10 μm) were made using a cryostat (Leica Biosystems) and postfixed with cold acetone for 3 min. Sections were stained with PE-conjugated anti-CD3ε (145-2C11), Alexa Fluor 488–conjugated anti-B220 (RA3-6B2) (eBioscience), and anti-desmin (Abcam), followed by Alexa Fluor 633–conjugated anti-rabbit IgG (Molecular Probes). The specimens were examined using an FV1200 confocal microscope (Olympus). Digital images were prepared using FV10-ASW (Olympus) and Adobe Photoshop CS6 (Adobe Systems).

Age- and sex-matched mice (7–9 wk old) were immunized s.c. in the flanks with an emulsion containing the myelin oligodendrocyte glycoprotein (MOG) peptide MOG_35–55 (200 μg/mouse) and Mycobacterium tuberculosis H37Ra extract (5 mg/ml in CFA, 200 μl/mouse). Pertussis toxin (200 ng/mouse) was administered i.p. on days 0 and 2. Mice were assigned scores daily on a scale of 0–5 in a double-blinded manner as described (27).

Unless otherwise indicated, statistical analysis was performed using unpaired Student t test. A p value <0.05 was considered statistically significant.

According to cDNA microarray analysis data presented on RefDIC (http://refdic.rcai.riken.jp/profile.cgi; RMPSTC027003 and RMPSTC028001), Arf1 and Arf6 are the top two Arf family members predominantly expressed in splenic CD4⁺ and CD8⁺ T cells. We thus focused on Arf1 as well as Arf6 to investigate the role of the Arf pathway in T cells. Given that loss of Arf1 or Arf6 results in embryonic lethality (28, 29), we established T lineage–specific Arf1-deficient (Arf1-KO), T lineage–specific Arf6-deficient (Arf6-KO), and T lineage–specific Arf1/Arf6 doubly deficient (Arf1/6-KO) mice by crossing conditional knockout mice (Supplemental Fig. 1A) with Lck-Cre transgenic mice. Both Arf1-KO mice and Arf6-KO mice had normal numbers of thymocytes and peripheral T cells (Supplemental Fig. 1B, 1C), whereas PCR analysis of genomic DNA obtained from purified T cells revealed that Arf genes were completely deleted by Lck-Cre (data not shown). By using quantitative PCR analysis, we also confirmed that expression levels of Arf mRNAs were virtually absent (Fig. 1A). Essentially, the same result was obtained with Arf1/6-KO mice, but the detailed analysis revealed that Arf1/6-KO mice alone exhibited decreased numbers of CD4⁺ and CD8⁺ T cells in the spleen (Fig. 1B), whereas numbers of CD4 single-positive (SP) and CD8SP thymocytes were increased (Supplemental Fig. 1D, 1E). These observations suggest that Arf1 and Arf6 play redundant roles in the T cells.

Increased numbers of CD4SP and CD8SP cells in the thymus can be explained by either enhanced efficiency in positive selection or suppression of T cell egress from the thymus. To examine these two possibilities, we first focused on the double-positive (DP) stage when positive selection takes place. We found that proportions of TCRβ^hiCD69^hi DP cells, which correspond to postselected DP cells, were comparable between control and Arf1/6-KO mice (Supplemental Fig. 1F). We further examined the expression profiles of chemokine receptor CCR7 in combination with CD69 to track developing thymocytes but found little or no difference between control and Arf1/6-KO mice (Supplemental Fig. 1G). Taking these results into account, it seems unlikely that positive selection is affected by the lack of Arf1 and Arf6. We next examined expression profiles of CD62L and CD69 because CD4SP as well as CD8SP cells can be subdivided into immature (CD62L^loCD69^hi) and mature (CD62L^hiCD69^lo) SP cells, the latter of which are known as egress-competent cells (30). Arf1/6-KO mice contained a higher proportion of mature SP cells compared with control mice (Supplemental Fig. 1H), suggesting that CD4SP and CD8SP cells have a defect in thymic egress in Arf1/6-KO mice. Consistently, mixed bone marrow chimera between control (CD45.1⁺) and Arf1/6-KO (CD45.1⁻) mice demonstrated that the proportions of splenic CD4⁺ cells relative to thymic CD4SP cells (referred to in this study as the egress index, Supplemental Fig. 1I) were substantially decreased in Arf1/6-KO mice–derived CD45.1⁻ cells. Decreased numbers of splenic CD4⁺ and CD8⁺ T cells in Arf1/6-KO mice can also be explained by this partial defect in thymic egress.

Considering that brefeldin A, a well-known inhibitor for ArfGEFs, which functions upstream of Arf1, is widely used to block cytokine secretion from activated T cells, it seems reasonable to speculate that Arf deficiency abrogates cytokine production. Contrary to our expectation, however, naive CD4⁺ T cells obtained from Arf1/6-KO mice produced IL-2 during activation to a level comparable to those from control mice (Fig. 1C). We also found that brefeldin A successfully blocked IL-2 secretion in both control and Arf1/6-deficient CD4⁺ T cells (Fig. 1D), excluding the possibility that Arf deficiency is compensated by a brefeldin A–independent secretory machinery. Consistent with this observation, Arf1/6-deficient T cells normally proliferated upon TCR stimulation (Fig. 1E). We further found that Arf1/6-deficient CD4⁺ T cells retained the potential to produce other cytokines, including IFN-γ, IL-4, and IL-17A as well (Fig. 1F). Although IL-4 production was decreased, Arf1/6-deficient CD4⁺ T cells secreted more IL-17A and IFN-γ than control cells. Collectively, we conclude that Arf1 and Arf6 are dispensable for cytokine secretion, including IL-2, IFNγ, and IL-17A in activated T cells.

One of the most important functions of CD4⁺ T cells in immune system is to provide help for B cells to produce Ab against “nonself” in germinal centers (GCs). To examine whether Arf deficiency affects Ab response in vivo, we evaluated Ab production against OVA along with Th1-polarizing adjuvant SAS or Th2-polarizing adjuvant Alum (31, 32). Contrary to the in vitro observations (Fig. 1F), Arf1/6-KO mice produced normal level of OVA-specific Ab under Th2-polarizing conditions, although Th1-driven response was attenuated (Fig. 2A). Histological analysis revealed that Arf1/6-deficient mice contain seemingly normal lymphoid tissues, including spleen and MLN (Fig. 2B, 2C). Although the proportion of CD4⁺ T cells in the MLN were decreased in Arf1/6-KO mice, this may reflect the reduced numbers of CD4⁺ T cells in the spleen (Fig. 2D). We also found that there was no difference in the amount of fecal IgA, which is induced against intestinal commensal bacteria in the GALTs, including the PPs between control and Arf1/6-KO mice (Fig. 2E), albeit with slightly reduced proportions of Tfh cells (CD4⁺CXCR5⁺PD-1^hi) in the PPs (Fig. 2F) in Arf1/6-KO mice. These data indicate that T cell–dependent Ab response remains intact in the gut of Arf1/6-KO mice.

Given that Arf1 contributes to the amino acid–induced mTORC1 activity in Drosophila S2 cells (33) and that Arf6 regulates tumor cell proliferation via the PLD–mTORC1 pathway (11), we examined whether Arf deficiency affects mTORC1 signal in T cells. Arf1/6-deficient CD4⁺ T cells exhibited slightly lower levels of phosphorylated ribosomal protein S6 (pS6), a measure of active mTORC1 signal, when compared with control CD4⁺ T cells (Fig. 3A). Consistent with the previous reports demonstrating that mTORC1 regulates glycolytic metabolism in activated T cells, glucose uptake upon TCR stimulation was moderately impaired in Arf1/6-deficient CD4⁺ T cells (Fig. 3B). In contrast, we found little or no difference between control and Arf1/6-deficient CD4⁺ T cells in surface expression levels of the amino acid transporter CD98, a well-known downstream target of the mTORC1 pathway (Fig. 3C). We also evaluated expression levels of another target of the mTORC1 pathway, transferrin receptor CD71, and found that CD71 induction was slightly attenuated at 24 h after TCR stimulation in Arf1/6-deficient CD4⁺ T cells (Fig. 3C). It should be noted, however, that there was no significant difference in CD71 levels between control and Arf-deficient CD4⁺ T cells at 72 h after stimulation (data not shown). These results suggest that mTORC1 signal in Arf1/6-deficient CD4⁺ T cells, albeit slightly reduced, sufficiently supports metabolic reprogramming during T cell activation.

To further clarify the relationship between the Arf pathway and mTORC1 signal, we next focused on T cell differentiation program. The mTORC1 pathway is known to control the balance of Th17 versus Treg, and the blockade of mTORC1 signal results in skewing of naive CD4⁺ T cells toward Treg differentiation while attenuating Th17 differentiation (34, 35). We therefore examined whether Arf deficiency affects differentiation program of naive CD4⁺ T cells in vitro and found that Arf1/6-deficient CD4⁺ T cells normally differentiated to Th17 or Treg under appropriate conditions (Fig. 3D). We thus conclude that the Arf pathway is dispensable for mTORC1-regulated T cell function.

Homeostatic proliferation generates two distinct populations: slow-dividing cells are induced in the presence of environmental cytokines like IL-7, whereas fast-dividing cells are thought to respond to gut microbiota (36). Interestingly, when compared with control CD4⁺ T cells, fast-dividing Arf1/6-deficient CD4⁺ T cells were markedly diminished, whereas slow-dividing ones appeared to be intact during homeostatic proliferation in lymphopenic Rag2^−/− mice (Fig. 4A), raising the possibility that Arf deficiency causes a defect in immune reaction in the gut. Actually, CD4⁺ T cells in the colonic LP were significantly decreased in Arf1/6-KO mice (data not shown). One can argue, however, that decreased proportion of Arf1/6-KO CD4⁺ T cells in the LP just reflects decreased number of splenic T cells (Fig. 1B). We thus analyzed mixed bone marrow chimeric mice as in Supplemental Fig. 1I and found that Arf1/6-deficient CD4⁺ T cells were defeated by control CD4⁺ T cells in the colonic LP, but not in the MLNs (Supplemental Fig. 2A), confirming that Arf deficiency leads to the reduction of CD4⁺ T cell number specifically in the gut. We therefore speculated the following two possibilities: Arf1/6-deficient CD4⁺ T cells have a defect in either migration into or survival in the colon.

Given that the CCR9 and the integrin α4β7 play important roles in T cell homing to the gut under homeostatic conditions (37, 38), we first examined the surface expression levels of CCR9 in colonic CD4⁺ T cells. However, we failed to detect CCR9 even in the CD4⁺ T cells obtained from the colonic LP of control mice, presumably because of its internalization upon ligand binding, as is the case with other chemokine receptors, including CCR7 (39). Instead, we found that CD4⁺ T cells derived from Arf1/6-KO mice expressed CCR9 on the surface of thymocytes to a similar level to those from control mice, suggesting that the Arf pathway is dispensable for CCR9 expression (data not shown). We also found that lack of both Arf1 and Arf6 rather augmented α4β7 expression levels in CD4⁺ T cells, albeit only slightly (Supplemental Fig. 2B). It should be noted, however, that the affinity of integrin for its ligands is tightly regulated by intracellular signaling, so-called inside-out signal, which plays a pivotal role in cell adhesion, spreading, and migration (40, 41). To further investigate whether integrin normally functions in Arf1/6-deficient CD4⁺ T cells, we estimated the size of adhesion area formed by interaction between LFA-1 integrin and its ligand ICAM-1 upon CCL21 stimulation, a measure of inside-out signal, and found that CCL21-stimulated, Arf1/6-deficient CD4⁺ T cells efficiently spread on the ICAM-1–coated plate (data not shown). Actually, by transferring a mixture of control (CD45.1⁺) and Arf1/6-deficient CD4⁺ T cells (CD45.1⁻) into Rag2^−/− mice, we confirmed that control and Arf1/6-deficient CD4⁺ T cells equally migrated into the colon (Supplemental Fig. 2C), excluding the possibility that decreased CD4⁺ T cell number in the colon is caused by impaired cell migration.

We next focused on the hypothesis that Arf deficiency in T cells leads to a defect in survival in the colon. Given that LP CD4⁺ T cells are in a state of activation because of continuous exposure to microbial Ags, we examined whether TCR-induced activation affects viability of CD4⁺ T cells lacking Arf1 and Arf6. Stimulation with anti-CD3ε and anti-CD28 mAbs resulted in marked decrease of Arf1/6-deficient CD4⁺ T cells in a signal intensity–dependent manner, although there was little or no difference in cell proliferation between control and Arf1/6-deficient CD4⁺ T cells, suggesting that Arf1/6-deficient CD4⁺ T cells are susceptible to cell death during activation (Fig. 4B). In contrast, CD4⁺ T cells lacking either Arf1 or Arf6 alone had no such defect in survival during TCR stimulation (Supplemental Fig. 3A, 3B), suggesting that Arf1 and Arf6 play a redundant role in cell survival. Essentially the same results were obtained with Arf1/6-deficient CD8⁺ T cells (Supplemental Fig. 3C, 3D). We further found that upon TCR stimulation, Arf1/6-deficient CD4⁺ T cells contained a higher proportion of sub-G1 cells (Supplemental Fig. 3E), which reflects apoptosis-related DNA fragmentation. Accordingly, treatment with Z-VAD-FMK, a pan-caspase inhibitor, partially blocked loss of Arf1/6-deficient CD4⁺ T cells induced by TCR stimulation (Supplemental Fig. 3F). It seems thus likely that the Arf pathway protects activated T cells from a caspase-mediated apoptotic program. Interestingly, the lack of Arf1 and Arf6 rendered only naive, but not effector, CD4⁺ T cells susceptible to apoptosis upon TCR stimulation (Fig. 4C). In contrast, we found no significant defect in IL-7–mediated survival between control and Arf1/6-deficient naive CD4⁺ T cells (Fig. 4C), consistent with the observation that Arf deficiency had no effect on slow-dividing population during homeostatic proliferation (Fig. 4A). These results taken together clearly demonstrated that Arf1 and Arf6 play a protective role exclusively in naive T cells during TCR-induced activation.

Because common γ-chain (γc) cytokines like IL-2 promote the survival of activated T cells (42), we examined the survival effect of γc cytokines on naive T cells upon TCR stimulation. Unexpectedly, IL-2 or IL-7 had no impact on the survival of TCR-stimulated, Arf1/6-deficient CD4⁺ T cells (Fig. 4D). In marked contrast, IL-4 as well as IL-21 successfully restored the viability of Arf1/6-deficient CD4⁺ T cells to a level similar to Z-VAD-FMK treatment, suggesting that IL-4 and IL-21 attenuate apoptosis caused by loss of Arf1 and Arf6. Apoptosis in T cells is triggered by two different pathways: the intrinsic pathway, which is controlled by the balance between pro- and antiapoptotic Bcl-2 family members (43–46), and the extrinsic pathway, which is initiated by signals delivered from death receptors such as Fas (47). Because Fas expression was intact in Arf1/6-deficient CD4⁺ T cells (data not shown), we focused on Bcl-2 family members and found that expression level of proapoptotic Bim was substantially increased in Arf1/6-deficient CD4⁺ T cells upon TCR stimulation, which was nearly completely attenuated in the presence of IL-21 (Fig. 4E, Supplemental Fig. 3G). In contrast, neither Arf deficiency nor IL-21 treatment had marked impact on expression levels of Bcl-2 (Fig. 4E, Supplemental Fig. 3G). Among three major isoforms of Bim, BimEL and BimL play a fundamental role in T cells (48). By using Western blotting analysis, we further found that BimL was the most abundantly upregulated in TCR-stimulated Arf1/6-deficient T cells (Supplemental Fig. 3H). Interestingly enough, treatment of Arf1/6-deficient T cells with IL-21 markedly decreased exaggerated expression of BimL. We also found that Arf deficiency perturbed expression levels of other antiapoptotic Bcl-2 family members: Mcl-1 was downregulated, whereas Bcl-xL was rather upregulated, albeit to a slight extent (Fig. 4F). It should be noted, however, that IL-21 treatment attenuated Bcl-xL expression in both control and Arf1/6-deficient T cells, whereas Mcl-1 level was rescued to a level comparable to control cells. Considering the protective effect of IL-21 against apoptosis, it seems reasonable to assume that enhanced apoptosis in Arf1/6-deficient CD4⁺ T cells is attributed to upregulation of Bim along with downregulation of Mcl-1. The physiological relevance of Bcl-xL upregulation in Arf1/6-deficient T cells remained to be resolved. Because ROS enhance apoptosis of T cells during activation (49), we also assessed the impact of NAC, a widely used antioxidant, on the viability of Arf1/6-deficient CD4⁺ T cells. By treatment with NAC, the proportion of dead cells in Arf1/6-deficient CD4⁺ T cells was significantly reduced (Fig. 4G). We also found, however, that Arf-deficient CD4⁺ T cells exhibited a lower level of ROS (Supplemental Fig. 3I), raising the possibility that Arf1/6-deficient CD4⁺ T cells are more sensitive to ROS-induced apoptosis.

To directly evaluate the fate of Arf1/6-deficient naive CD4⁺ T cells upon activation in the colon, we used a naive CD4⁺ T cell–induced colitis model, one of widely used mouse models of inflammatory bowel disease (IBD), in which naive CD4⁺ T cells transferred into Rag2^−/− mice are activated with microbial Ags and differentiate to pathogenic Th17 cells in the colon. Actually, we found that Rag2^−/− mice transferred with control CD4⁺ T cells exhibited marked weight loss, which reflects colitis development, around 3 wk after the transfer (Fig. 5A). In contrast, the transfer of Arf1/6-deficient naive CD4⁺ T cells had no impact on body weights. Both thickening of the colon wall and severe infiltration of mononuclear cells in the colon confirmed the onset of colitis in control CD4⁺ T cell–transferred Rag2^−/− mice, whereas such histological changes were not detected in the mice transferred with Arf1/6-deficient CD4⁺ T cells (Fig. 5B), suggesting that naive CD4⁺ T cells lacking Arf1 and Arf6 failed to induce colitis. Considering that Th17 cells play a pivotal role in the pathogenesis of naive CD4⁺ T cell–induced colitis, one can argue that suppression of colitis could reflect the impaired ability of Arf1/6-deficient naive CD4⁺ T cells to differentiate to pathogenic Th17 cells. However, Arf1/6-deficient CD4⁺ T cells normally differentiated to pathogenic Th17 cells in vitro (Fig. 5C). In addition, pathogenic Th17 cells generated from Arf1/6-deficient CD4⁺ T cells were maintained in vivo to a level comparable to those from control (Supplemental Fig. 4A). We also found that CD4⁺ T cell number was systemically decreased in the mice transferred with Arf1/6-deficient CD4⁺ T cells when compared with those transferred with control T cells, whereas the frequency of IL-17A–producing CD4⁺ T cells were rather augmented (Supplemental Fig. 4B, 4C). We further found that Arf1/6-deficient CD4⁺ T cells were activated to a level comparable to control CD4⁺ T cells in the colon, which is indicated by upregulation of an activation marker CD44 (Fig. 5D). Furthermore, when transferred with control T cells into Rag2^−/− mice, Arf1/6-deficient T cells were virtually outcompeted (Fig. 5E). These results taken together suggest that the reason why transfer of Arf1/6-deficient naive CD4⁺ T cells did not develop colitis is likely due to the failure in expansion of activated T cells, but not due to the impaired differentiation or survival of pathogenic Th17 cells. Similar results were observed in another type of Th17-mediated inflammatory disease of the CNS, experimental autoimmune encephalomyelitis (EAE): Arf1/6-deficient mice were completely resistant to induction of EAE after immunization with MOG_35–55 peptide, whereas control mice developed severe EAE (Fig. 5F). Collectively, these results suggest that the Arf pathway is required for the onset of Th17-mediated autoimmune diseases, which could be explained by the protective role of the Arf pathway against apoptosis during TCR stimulation.

Intracellular cytokine staining combined with flow cytometry is a powerful method to assess T cell responses, in which the fungal metabolite brefeldin A plays a key role to retain cytokines within the activated T cells. Although brefeldin A is thought to inhibit Arf1 activation through its GEFs, physiological role of Arf family proteins in T cells has been unclear. In this study, we demonstrated that Arf1 along with Arf6 plays a pivotal role in developing autoimmune disease like colitis and EAE, presumably through suppressing apoptosis in activated T cells. In contrast, to our surprise, cytokine secretion was seemingly normal in CD4⁺ T cells lacking both Arf1 and Arf6. It has been revealed that among Arf1-GEFs targeted by brefeldin A, GBF1, BIG1, and BIG2 are involved in the activation of Arf3 and/or Arf5 as well (50). Therefore, these other Arf isoforms might compensate for the absence of Arf1 as well as Arf6 in the activated CD4⁺ T cells. Actually, previous studies have demonstrated that Arf isoforms share some redundant functions: small interfering RNA–mediated single knockdown of Arf1, Arf3, Arf4, or Arf5 fails to cause any observable phenotype, whereas double knockdowns of any type of combination of them using pairwise small interfering RNA impair the particular steps in the vesicle trafficking pathway (51).

Upon TCR stimulation, Arf-deficient CD4⁺ T cells exhibited higher level of Bim, a predominant BH3-only protein functioning in the activated T cells, whereas expression of an antiapoptotic Bcl-2 family protein Mcl-1 was slightly impaired. Although we found that expression level of Bcl-xL, another Bcl-2 family member exerting antiapoptotic function, was substantially increased in Arf-deficient T cells as well, recent studies have revealed that Bcl-xL is dispensable for survival of peripheral T cells as well as development of effector T cells (52, 53), suggesting that enhanced apoptosis in Arf-deficient T cells during activation is presumably due to the perturbed expression of Bim and/or Mcl-1. It is well established that Bim expression is negatively regulated by the PI3K/Akt and the Ras/Erk pathways. The former suppresses Bim expression at a transcriptional level through inhibition of Foxo transcription factor (54), whereas the latter induces proteasomal degradation of especially BimEL via Erk-mediated phosphorylation at Ser⁶⁵ (48). However, considering that the PI3K/Akt pathway plays an essential role in Th17 differentiation, the fact that Arf deficiency had no impact on Th17 differentiation strongly suggests that the PI3K/Akt pathway is intact in CD4⁺ T cells lacking Arf1 and Arf6. Consistently, phosphorylation status of S6 protein mediated by mTORC1 signal, which is a well-known target of the PI3K/Akt pathway, seemed normal in Arf-deficient T cells. We also found that the lack of Arfs unaffected surface expression levels of CD62L and CD69 (data not shown), which are controlled downstream of the Ras/Erk pathway, excluding the possibility that Arf deficiency impairs activation of the Ras/Erk pathway either. So, what is the molecular mechanism linking Arf deficiency and enhanced apoptosis in the activated T cells? Our data suggest that ROS are involved, at least partly, in the apoptosis of Arf1/6-KO CD4⁺ T cells. Actually, it has been reported that the accumulation of ROS enhances the expression of Bim in T cells (55). Considering that ROS levels were decreased in Arf-deficient CD4⁺ T cells, however, we rather speculate that the lack of Arf1 and Arf6 renders T cells more sensitive to apoptosis-triggering cues including ROS. One intriguing possibility is autophagy. Several studies have demonstrated that both Arf1 and Arf6 play a critical role in autophagy through the formation of autophagosomes (56, 57). Because autophagy negatively regulates Bim expression as well (58), it stands to reason that the defect in autophagy process in Arf-deficient CD4⁺ T cells causes accumulation of Bim protein. Alternatively, as has recently been reported, Arf deficiency may induce ER stress response, rendering the activated T cells susceptible to cell death (59). Given that Arf1/6-KO effector CD4⁺ T cells were resistant to apoptosis upon TCR stimulation, the difference between naive and effector T cells, like metabolic states, would provide a clue for mechanism in augmented apoptosis in Arf-deficient naive T cells. One must await detailed analysis to reveal the molecular linkage between Arf deficiency and enhanced apoptosis in activated T cells.

Despite the fact that inflammatory responses including colitis and EAE were nearly completely abrogated, Ab responses especially under Th2-polarizing conditions seemed intact in Arf1/6-KO mice (Fig. 2A). Essentially the same results were obtained during Leishmania major infection, which is prevented by Th1 immune response: Ab against L. major was produced to a level comparable to control mice, whereas Arf1/6-KO mice failed to completely eliminate the parasite presumably because of the inefficient activation of Th1 cells (data not shown, and the details will be described elsewhere). We speculate this puzzling phenotype reflects the difference in environmental cues. We found that, among the γc cytokine family, IL-21 and IL-4 efficiently suppressed apoptosis in activated Arf1/6-KO T cells (Fig. 4D). GCs found in secondary lymphoid tissues are important sites for Ab production. In GCs, Tfh cells support B cell differentiation into Ab-secreting plasma cell by producing a substantial amount of IL-21. Arf-deficient T cells could survive in GCs, where plenty amount of IL-21 exists compared with the nonlymphoid tissues like the colon. Actually, in the mixed bone marrow chimeric mice, we found that Arf-deficient T cells were maintained to a level comparable to control T cells in the MLN, although proportions of Arf-deficient T cells were markedly decreased in the colonic LP (Supplemental Fig. 2A).

It is estimated that nearly 2.5 million people live with multiple sclerosis (MS) in the world, and 6.8 million patients are suffering from IBD primarily comprising of Crohn disease and ulcerative colitis globally in 2017 (60). Although most immune therapies for these patients are associated with immunosuppressive drugs, which typically target the adaptive immune system, these treatments may increase the risk of serious side effects like reduced Ab responses against bacteria and viruses. In marked contrast, loss of Arf function in T cells nearly completely suppressed the onset of naive CD4⁺ T cell–induced colitis as well as EAE without affecting OVA-specific Ab production, at least under Th2-polarizing conditions, demonstrating that the Arf pathway would be a good drug target for autoimmune disease like MS and/or IBD. Given that the ubiquitous role of the Arf pathways in vesicle trafficking, one can argue that the blockade of the Arf pathway would cause detrimental effects. To our surprise, however, the mice at 2 mo old were viable for 4 wk after Arf1 and Arf6 were systemically deleted by using the CreERT2 system (data not shown; the details will be described elsewhere), raising the possibility that the lack of Arf1 and Arf6 may have little or no fatal effect, at least within a short period of time. Interestingly, genome-wide association studies (https://www.ebi.ac.uk/gwas/) have linked common genetic variants at the locus of ASAP1, one of the ArfGAPs, to MS. In addition, another ArfGAP ASAP2 and an ArfGEF cytohesin 1 are suggested to associate with Crohn disease and ulcerative colitis, respectively. It should be noted that all of these molecules are known to regulate both Arf1 and Arf6 (2). Future studies will clarify the impact of therapeutic approach targeting upstream regulators of the Arf pathway on autoimmune diseases.

We thank Hans Joerg Fehling for Rosa26-tdRFP mice. We also thank Yoichi Maekawa, Yoshihiro Ueda, and Katsuhiko Yoshizawa for technical advice and Chikako Eguchi for technical assistance.

The authors have no financial conflicts of interest.