NKT cells represent a small subset of glycolipid-recognizing T cells that are heavily implicated in human allergic, autoimmune, and malignant diseases. In the thymus, precursor cells recognize self-glycolipids by virtue of their semi-invariant TCR, which triggers NKT cell lineage commitment and maturation. During their development, NKT cells are polarized into the NKT1, NKT2, and NKT17 subsets, defined through their cytokine-secretion patterns and the expression of key transcription factors. However, we have largely ignored how the differentiation into the NKT cell subsets is regulated. In this article, we describe the mRNA-binding Roquin-1 and -2 proteins as central regulators of murine NKT cell fate decisions. In the thymus, T cell–specific ablation of the Roquin paralogs leads to a dramatic expansion of NKT17 cells, whereas peripheral mature NKT cells are essentially absent. Roquin-1/2–deficient NKT17 cells show exaggerated lineage-specific expression of nearly all NKT17-defining proteins tested. We show through mixed bone marrow chimera experiments that NKT17 polarization is mediated through cell-intrinsic mechanisms early during NKT cell development. In contrast, the loss of peripheral NKT cells is due to cell-extrinsic factors. Surprisingly, Roquin paralog–deficient NKT cells are, in striking contrast to conventional T cells, compromised in their ability to secrete cytokines. Altogether, we show that Roquin paralogs regulate the development and function of NKT cell subsets in the thymus and periphery.
Natural killer T cells express an evolutionarily conserved semi-invariant TCR and are characterized by an activated phenotype and rapid secretion of effector cytokines in response to innate and antigenic stimulation. In the mouse, NKT cells represent 0.2–0.5% of lymphocytes in thymus, spleen, and bone marrow and ∼30% in the liver, whereas in humans the fractions are smaller (<0.1 and 1%, respectively) (1). Despite their rarity, NKT cell responses can drive inflammation or tolerance, thereby impacting a wide range of immune cells, such as dendritic cells, NK cells, and B and T cells (2). NKT cells protect their host organisms from certain strains of bacteria, sustain antiviral responses, and contribute to the suppression of certain types of cancer and immune diseases (1, 3–6). In contrast, NKT cells are involved in the pathophysiology of allergic responses, ulcerative colitis, and liver cancer (7–9). Therefore, NKT cell activation can have dramatically different outcomes, depending on diverse environmental factors.
Rare Vα14-Jα18 (Vα14i) TCRα-chain rearrangements in CD4 CD8 double-positive (DP) thymocytes lead, in conjunction with a limited number of TCRβ-chains, to the expression of a glycolipid-recognizing TCR. The overwhelming majority of NKT cells express this Vα14i-containing semi-invariant TCR. Therefore, they are termed invariant NKT or Vα14i-NKT cells, but for simplicity we will refer to them as NKT cells. Recognition of glycolipid Ags presented by CD1d on DP thymocytes through the semi-invariant TCR expressed by precursor cells (stage 0: CD24high, CD44low, NK1.1−) triggers NKT cell development (10, 11). These TCR signals induce massive proliferative expansion, downregulation of CD24 (stage 1), and expression of the key transcriptional regulator of NKT cell maturation, promyelocytic leukemia zinc finger (PLZF) (12–15). Thymic NKT cell maturation is indicated by subsequent upregulation of the memory marker CD44 (stage 2) and expression of NK cell markers, most prominently NK1.1 (stage 3) (16, 17). While passing through stages 1–3, NKT cells also differentiate into functional subsets termed NKT1 (PLZFlow, T-bet+), NKT2 (PLZFhigh, GATA3+), and NKT17 (PLZFint, RORγt+) cells, which are polarized toward the preferential production of cytokines reminiscent of the respective Th cell lineages (18, 19). The signals and events that drive the differentiation of these NKT cell subsets are incompletely understood.
Roquin-1 and its mammalian paralog Roquin-2, encoded by rc3h1 and rc3h2 respectively, are mRNA-binding proteins that repress immune reactions by redundant and unique posttranscriptional regulation of genes, including costimulatory molecules and proinflammatory cytokines (20–24). Thereby, these proteins act as safeguards against autoimmunity by preventing the spontaneous generation of follicular Th (TFH) cells, IL-17–producing Th17 cells, and IFN-γ–producing short-lived effector CD8 T cells (20, 22, 25). However, their role in the development of agonist-selected T cell subsets, such as NKT cells, is unknown.
In this article, we report that conditional ablation of both Roquin paralogs in T cells essentially prevents the generation of mature NKT cells in the thymus, with the exception of NKT17 cells, whose production is dramatically increased by cell-intrinsic mechanisms. Roquin-1/2–deficient NKT cells express high amounts of NKT17-associated markers but display a hyporesponsive phenotype, as shown by impaired cytokine production. In addition, peripheral Roquin-1/2–deficient NKT cells are essentially absent, mainly due to cell–extrinsic mechanisms.
Materials and Methods
CD4Cre (26), Rc3h1F/F (24), Rc3h2F/F (22), Vα14iStopF (27), IL6Rα−/−, and IL6RαF/F (28) mice were kept on a C57BL/6 genetic background. Mice were housed in specific pathogen–free animal facilities at the Max-Planck Institute of Biochemistry and the Technische Universität München. Unless indicated otherwise, age-matched animals were analyzed at 6–12 wk of age. CD4Cre Rc3h1-2F/F mice were analyzed before any visible signs of pathology. All experiments were performed in accordance with German Federal Animal Protection Laws and approved by the Regierung of Oberbayern. In most experiments, Rc3h1F/F Rc3h2F/F, Rc3h1F/F Rc3h2F/+, and/or CD4Cre mice were used as controls. Controls were pooled, because one copy of the CD4Cre transgene does not significantly affect thymocyte numbers or splenic NKT cell numbers (29).
Single-cell suspensions were stained using the following commercial Abs and kits: Bcl-6 (K112-91), B220 (RA3-6B2), CCR-6 (29-2L17), CD4 (GK1-5, RM4-5), CD8a (53-6.7), CD11b (M1/70), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD62L (MEL-14), CD90.2 (53-2.1), CD127 (A7R34), CD138 (281-2), c-MAF (sym-0F1), CXCR5 (2G8), Egr2 (erongr2), Foxp3 (FJK-16s), GATA3 (TWAJ), Gr-1 (RB6-8C5), IL-17 (eBio17B7), IL-4, (11B11), IL-6Rα (D7715A7), INF-γ (XMG1.2), Ly-6C (HK1.4), Neuropilin-1 (JE12), NK1.1 (PK136), PD-1 (J43), PLZF (9E12), RORγt (AFKJS-9, Q31-387), TCRβ (H57-597), T-bet (4B10), and ThPOK (2POK). For intracellular transcription factor staining, cells were fixed and permeabilized with a Foxp3 Transcription Factor Staining Buffer Set (eBioscience). For intracellular cytokine staining, cells were fixed and permeabilized with a Cytofix/Cytoperm kit (BD) or a Foxp3 Transcription Factor Staining Buffer Set (eBioscience). For analysis of intracellular phosphorylated proteins, single-cell suspensions were prepared using serum-free buffer. Cells were processed using a Transcription Factor Phospho Buffer Set or fixed with Phosflow Lyse/Fix Buffer and permeabilized with Perm Buffer III (all from BD). Cells were stained overnight using commercially available mAbs against p-STAT3 (pS727), p-mTOR (MRRBY), p-4EBP1 (236B4) and p-AKT (S473) (M89-61), p-ZAP70 (n3kobu5), and phosphorylated phospholipase-Cγ (PLCγ) (K86-689.37) (all from BD). For detection of apoptotic cells, cells were stained with Annexin V and 7-aminoactinomycin D (7-AAD) using an Annexin V detection kit (eBioscience). mCD1d PBS-57 tetramers were provided by the National Institutes of Health Tetramer Core Facility. For enrichment of thymic NKT cells, CD8+ thymocytes were depleted using CD8a MicroBeads (Miltenyi Biotec). For intracellular cytokine staining, cells were stimulated for 4 h with 100 ng/ml PMA (Sigma) and 1000 ng/ml ionomycin (Calbiochem) together with 2 μM monensin (eBioscience). Samples were acquired on a FACSCanto II or LSRFortessa (BD) and analyzed with FlowJo software. For gating on NKT cells, the following gating strategy was used: singlets → LIVE/DEAD stain, gated on living cells → lymphocyte gate → TCRβint mCD1d-PBS57+ cells. NKT cell stages and subsets were gated as indicated in the FACS plots. To analyze the relative expression of extracellular and intracellular proteins, median fluorescent intensities (MFIs) were calculated, and the mean MFI of mCD1d-PBS57 tetramer− TCRβ+ T cells of the respective control mice was set to 1.
Isolation of lamina propria and lung lymphocytes
For isolation of lamina propria lymphocytes, Peyer’s patches and fat tissue were removed, and intestines were flushed with ice-cold PBS. Intestines were opened longitudinally and cut into 1–1.5-cm pieces. After vigorously vortexing in PBS, samples were incubated two times for 15–20 min at 37°C in HBSS buffer containing 5% FCS and 5 mM EDTA, 1 mM DTT, and 10 mM HEPES. Then, cells were washed in PBS, and intestines were digested for 45 min at 37°C in PBS+Ca/+Mg containing 5% FCS, 1 mg/ml Collagenase, Type 2 (CellSystems), and 0.1 mg/ml DNase I (Roche). Lymphocytes were then purified by a 40/80% Percoll (Biochrom) gradient.
To prepare single-cell suspensions, lungs were digested using RPMI 1640 containing 0.02 mg/ml Liberase TM and 10 U DNase (both from Roche Diagnostics).
Bone marrow chimeras
B6.SJL-Ptprca Pepcb/BoyJ (B6.SJL-congenic); C57BL/6 heterozygous recipient mice (CD45.1/2) were lethally irradiated with 2 × 5.5 Gy (4 h apart) and injected i.v. with 4–5 × 106 bone marrow cells in a B6.SJL (CD45.1)/C57BL/6 (CD45.2) ratio of 1:1. Before transplantation, bone marrow was depleted of T cells using CD90.2 MicroBeads (Miltenyi Biotec). Two to three days before until 2 wk after transplantation recipient mice were treated with Borgal 24% (Virbac Animal Health) in drinking water at a dose of 0.1 ml/kg body weight/d. Mice were analyzed 7–9 wk after transplantation.
Statistical analysis was performed by GraphPad prism version 6. The p values were calculated as indicated in the figure legends.
Ablation of Roquin proteins leads to a reduction in thymic NKT cells and a loss of peripheral NKT cells
To assess the role of Roquin proteins in NKT cells, we analyzed mice with T cell–specific ablation of one or both of these proteins. We used CD4Cre for this purpose, because it efficiently recombines conditional alleles, including Rc3h1F/F (24) and Rc3h2F/F (22), in DP thymocytes. CD4Cre-mediated protein ablation is complete in NKT cell precursors (16, 17, 30, 31), which represent older DP thymocytes because of the late timing of the Vα14-Jα18 rearrangement (16, 17, 31, 32). Knockout of Roquin-1 and coablation of Roquin-1/2 led to a 2-fold reduction in NKT cell proportions and numbers in the thymus (Fig. 1A, 1B). Analysis of thymic developmental stages 1–3 revealed normal numbers at stage 1, gain of numbers at stage 2, and significant loss of numbers at stage 3 in NKT cells in CD4Cre Rc3h1F/F mice compared with control mice; this was strongly exacerbated when Roquin-2 was coablated (Fig. 1A, 1C). Accordingly, the absolute numbers of stage 3 NKT cells were reduced from an average of 32 × 104 cells in controls to 16 × 104 in CD4Cre Rc3h1F/F mice and 1 × 104 in CD4Cre Rc3h1-2F/F mice (Fig. 1C).
Moreover, loss of Roquin proteins also affected NKT cells in peripheral lymphoid organs. In the spleens of CD4Cre Rc3h1F/F mice, NKT cell proportions and absolute numbers were reduced by >50%, primarily due to a reduction in stage 3 NKT cells (Fig. 1D, 1E, Supplemental Fig. 1A). Strikingly, NKT cells were virtually absent in the spleens of CD4Cre Rc3h1-2F/F mice (Fig. 1D, 1E). A similar picture was observed in the liver, peripheral lymph nodes (pLNs), mesenteric lymph nodes (mLNs), and lung (Fig. 1F, Supplemental Fig. 1B). Immune cells were dramatically expanded in liver and most prominently in pLNs, but not in mLNs, of CD4Cre Rc3h1-2F/F mice compared with control mice. Therefore, absolute NKT cell numbers were reduced in liver and mLNs but not pLNs (Fig. 1F). In contrast to Roquin-1, individual loss of Roquin-2 has no discernable effects on NKT cell development and numbers (Supplemental Fig. 1C, 1D).
Taken together, ablation of Roquin proteins leads to a dose-dependent decrease in thymic NKT cell generation, with a clear maturation block between stages 2 and 3. In the periphery, NKT cells were only barely detectable in CD4Cre Rc3h1-2F/F mice.
Roquin proteins restrict NKT17 cell differentiation
Given that most NKT17 cells are phenotypically stage 2, we determined the subset composition (33). We found increased proportions of NKT17 cells in CD4Cre Rc3h1F/F mice, whereas most of the NKT cells in CD4Cre Rc3h1-2F/F mice were NKT17 (Fig. 2A). Loss of Roquin-2 alone had no effect on NKT subset composition (Supplemental Fig. 1E). Importantly, the absolute numbers of NKT17 cells were increased >10-fold, from an average of 0.5 × 104 in controls to 6.5 × 104 cells in CD4Cre Rc3h1-2F/F mice. Therefore, we conclude that the increased proportions are due to strongly increased NKT17 cell production in the absence of Roquin-mediated regulation and not merely a consequence of a decreased NKT1 population (Fig. 2B). At this point, because we observed more pronounced changes in the absence of both Roquin proteins, we focused on the analysis of thymic NKT cells from CD4Cre Rc3h1-2F/F mice in further experiments.
ICOS is the first described target of Roquin’s posttranscriptional activities, but ICOS is also differentially expressed among NKT cell subsets. Therefore, we compared ICOS MFIs on NKT1, NKT2, and NKT17 cells. In controls, the highest ICOS expression was found on NKT17 and NKT2 cells, whereas NKT1 cells expressed similar amounts as TCRβ+ tetramer− thymocytes (Fig. 2C). Loss of both Roquin paralogs enhanced ICOS expression to a similar extent in all subsets (Fig. 2C), indicating that the differential ICOS expression in NKT cell subsets occurs largely independently of Roquin-mediated regulation.
NKT17 cells are characterized by enhanced expression of the chemokine receptor CCR-6, the integrin αEβ7 (CD103), syndecan-1 (CD138), and neuropilin-1 (Nrp-1) (Fig. 2D) (34–36). Unlike controls, Roquin-1/2–deficient NKT17 cells were uniformly positive for all of these markers (Fig. 2D). Interestingly, a fraction of Roquin-1/2–deficient NKT1 and NKT2 cells expressed CD138, the surface protein most faithfully representing NKT17 lineage commitment. This could indicate a bias toward NKT17 differentiation in these cells.
In summary, we demonstrate that Roquin proteins prevent polarization into the NKT17 lineage.
Expression of a knock-in Vα14i TCR facilitates development and peripheral accumulation of all NKT cell subsets in CD4Cre Vα14iStopF Rc3h1-2F/F mice
Transgenic expression of a Vα14i TCR has been used frequently to rescue defects in early NKT cell development. The presence of normal numbers of stage 1 NKT cell precursors in CD4Cre Rc3h1-2F/F mice essentially excludes defects in TCRα rearrangements or thymocyte survival, which are commonly overcome in such experiments. We recently generated the Vα14iStopF allele, a novel type of Vα14i transgene that expresses a Vα14i TCR from within the endogenous TCRα gene locus upon Cre-mediated recombination. In combination with CD4Cre, this leads to massive NKT cell overproduction (Fig. 3A) (27). In CD4Cre Vα14iStopF Rc3h1-2F/F mice, expression of Cre leads to recombination of all conditional alleles, ultimately causing the inactivation of the Roquin-1/2 alleles and induction of the Vα14i TCRα-chain. Vα14i TCR expression occurs within 6 h after Cre expression (S. Bortoluzzi, C. Drees, and M. Schmidt-Supprian, unpublished observations), whereas it takes considerably longer before the cells are functionally depleted of Roquin-1/2 proteins (S. Bortoluzzi, C. Drees, and M. Schmidt-Supprian, unpublished observations). Furthermore, it is very likely that a cell has activated the Vα14iStopF allele before it has inactivated all four Roquin alleles. Therefore, in contrast to the situation in CD4Cre Rc3h1-2F/F mice, in CD4Cre Vα14iStopF Rc3h1-2F/F mice a newly expressed Vα14i TCR initially signals to a cell that still contains some Roquin proteins, which are then lost during NKT cell development. Vα14i TCR expression restored the development of NKT cells of all stages in CD4Cre Vα14iStopF Rc3h1-2F/F mice, although the overall numbers were still significantly reduced compared with controls (Fig. 3A, 3B). In Vα14i knock-in mice, NKT cell development is strongly biased toward cells early in their development (Fig. 3B), which is also reflected in the expression profiles of RORγt and PLZF (Supplemental Fig. 2A). Within the mCD1d-PBS57 tetramer+ NKT cells, we defined the earliest RORγt+ PLZF− CD44− CD24+ progenitor cells as Vα14i-DP, because they essentially represent DP thymocytes expressing a Vα14i-NKT cell TCR (Fig. 3D, 3E, Supplemental Fig. 2A). We termed the putative next RORγt− PLZF− NK1.1− developmental stage as pre-NKT, because it consists of various NKT cell precursor populations (Supplemental Fig. 2A). Proportions and absolute cell numbers indicated a slight block in the Vα14i-DP to pre-NKT transition upon ablation of Roquin-1/2 (Fig. 3C). The numbers of NKT1 (RORγt−, NK1.1+, PLZFint) and NKT17 (RORγt+, CD44+, CD24−) cells were similar in both genotypes, whereas loss of Roquin caused a decrease in NKT2 (RORγt−, PLZFhigh, CD24−) cells (Fig. 3C, Supplemental Fig. 2A). Further characterization of the early progenitor stages and of the functional subsets with T-bet and GATA3 confirmed these results (Fig. 3D–G). Mature Roquin-1/2–deficient NKT cells of all subsets and stages also were detected in the spleen, although the numbers were reduced compared with controls (Fig. 4). Most diminished were PLZFhigh (NKT2) cells, whereas NKT17 cell numbers were not significantly altered (Fig. 4).
In conclusion, we show that expression of a knock-in Vα14i NKT cell TCR, concomitant with ablation of Roquin-1/2, largely overcomes the block in thymic NKT cell development, permits the accumulation of Roquin-deficient mature peripheral NKT cells, and strongly reduces the bias toward NKT17 cell production in CD4Cre Vα14iStopF Rc3h1-2F/F mice.
Roquin paralogs cell intrinsically drive NKT17 differentiation, and the loss of peripheral NKT cells is also due to cell-extrinsic factors
The results derived from experiments with the Vα14i TCR indicated that Roquin proteins play decisive roles during the earliest stages of NKT cell development. DP thymocytes are critically important for positive selection of developing NKT cells by presenting Ag and through stimulation via homotypic interactions between signaling lymphocytic activation molecule (SLAM) family members (16, 37). However, the expression of CD1d, CD150 (SLAMF1), and Ly108 (SLAMF6), which are required for successful NKT cell development (37), were not significantly altered by the lack of Roquin paralogs in DP thymocytes (Fig. 5A), arguing against an indirect role for these cells. To directly distinguish between cell-intrinsic or cell-extrinsic mechanisms, we created mixed bone marrow chimeras. We adoptively transferred T cell–depleted CD45.1 B6.SJL-congenic bone marrow, mixed with an equal amount of CD45.2 control or CD45.2 CD4Cre Rc3h1-2F/F T cell–depleted bone marrow, into lethally irradiated CD45.1/2 DP mice (Fig. 5B, Supplemental Fig. 2B). Thymi and spleens of the resulting chimeras were analyzed 7–9 wk after transplantation. Even after lethal irradiation, 20–30% of thymic NKT cells in the resulting chimeras were derived from hosts, and similar proportions of radio-resistant CD45.1/2 host T and NKT cells were found in the spleens (Supplemental Fig. 2C). Further, to account for the engraftment variations that are typical of these experiments, we normalized the CD45.1/CD45.2 ratios to those of double-negative (DN) thymocytes and B cells, which are Roquin proficient. This analysis revealed that Roquin-1/2–deficient NKT cells develop in normal proportions from their precursors, similar to other thymocyte subsets and similar to CD45.1 competitor and CD45.2 control NKT cells. Strikingly, Roquin-1/2–deficient NKT cells could also be detected at nearly normal proportions in the spleen, in striking contrast to the situation in CD4Cre Rc3h1-2F/F mice (Figs. 1D, 1E, 5C, Supplemental Fig. 2C). The proportions of Roquin-1/2–deficient CD45.2 thymic NKT17 cells were strongly and significantly increased, in contrast to the CD45.1 competitor and CD45.1 and CD45.2 cells in the control chimeras (Fig. 5D). Roquin-1/2–deficient NKT cells of all subsets could be detected in the spleen, although NKT17 cells remained dominant (Fig. 5C, Supplemental Fig. 2C, 2D). Furthermore, we found that elevated ICOS expression and the spontaneous differentiation of memory/effector-like T cells and TFH cells of the Roquin-1/2–deficient T lineage are cell intrinsic (Supplemental Fig. 3), similar to the san/san T lineage (38).
To test whether peripheral NKT cells are initially generated but lost over time in CD4Cre Rc3h1-2F/F mice, we analyzed NKT cell development in very young (16–20-d-old) mice. NKT cell numbers were reduced significantly in the thymi and were nearly absent in the spleen of young CD4Cre Rc3h1-2F/F mice compared with control mice (Fig. 6A, 6B). At this age, NKT2 cells represent the largest subset in thymi of control mice, whereas in the absence of the Roquin paralogs, NKT17 cells predominate, and very few NKT1/2 cells are generated (Fig. 6C), similar to the situation in older mice (Fig. 2B).
Thus, our results demonstrate that the massive NKT17 cell production in CD4Cre Rc3h1-2F/F mice is mediated through cell-intrinsic mechanisms, whereas the reduction in mature peripheral NKT cells is largely caused by cell-extrinsic mechanisms.
Roquin-1/2–deficient NKT cells are characterized by elevated c-Maf but reduced levels of Egr2, PLZF, and phosphorylated STAT3
Having demonstrated the cell-intrinsic nature of the NKT17 bias of Roquin-1/2–deficient NKT cells, we embarked on further molecular characterization. Compared with controls, NKT17 cells lacking Roquin paralogs express higher amounts of TCR on their surface, whereas NKT1 cells express lower levels (Fig. 7A). Conventional CD4 T cells upregulate Ly-6C upon deprivation of TCR-mediated self-recognition (39). Within the NKT lineage, NKT1 cells had the largest proportion of Ly-6C+ cells, whereas only a few NKT2 and essentially no NKT17 cells were putatively self-deprived, which inversely correlated with TCRβ expression. Loss of Roquin-1/2 decreased the percentage of Ly-6C+ cells in all NKT subsets, although the differences were not dramatic (Fig. 7B). Next, we assessed the expression of the TCR-induced transcription factor Egr2, which was also shown to induce PLZF by binding to its promoter in immature NKT cells (40). Loss of Roquin-1/2 led to a significant decrease in Egr2 levels in TCR+ thymocytes and all NKT subsets, including NKT17 cells, which had the highest TCR surface levels (Fig. 7C). In Roquin-1/2–deficient NKT17 cells, decreased Egr2 expression was accompanied by a 50% reduction in PLZF protein (Fig. 7C), but PLZF expression was unchanged in the other subsets. PLZF was shown to regulate the expression of c-Maf (41), a transcriptional regulator of cytokine production that is highly expressed in Th17 cells (42) and contributes to Th17 differentiation through the induction of RORγt (43). Interestingly, we found significantly increased c-Maf expression in Roquin-1/2–deficient NKT2 and NKT17 cells, independent of PLZF levels (Fig. 7C). To assess the direct contribution of TCR signals to these changes, we analyzed phosphorylation of the TCR proximal adapter molecule ZAP70 and the further downstream acting PLCγ (Fig. 7C). In Roquin-1/2–deficient NKT17 cells, there was a trend toward diminished intracellular TCR signals, but it was not statistically significant. Altogether, our analyses indicated that TCR signaling is not significantly impacted by the loss of Roquin-1/2 proteins. IL-6, together with TCR signals, strongly induces c-Maf expression via STAT3 (44). However, in accordance with previous data (45), we did not detect significant expression of the STAT3-activating IL-6R on bulk NKT cells. Complete or T cell–specific ablation of the IL-6Rα–chain did not affect NKT cell development and subset differentiation in the thymus, but it led to increased splenic NKT cell numbers (Supplemental Fig. 4). Nevertheless, STAT3 phosphorylation was increased in control NKT2 and NKT17 cells compared with NKT1 cells (Fig. 7C), indicating that another receptor might activate STAT3. However, this receptor or downstream signals do not appear to be regulated by Roquin proteins, because Roquin-1/2–deficient NKT1 and NKT17 cells showed slightly reduced STAT3 phosphorylation compared with controls (Fig. 7C). Fate decisions between NKT1 and NKT17 cells are controlled by the transcription factor Th poxviruses and zinc finger and Krüppel family (ThPOK), whose expression is strongly reduced in NKT17 cells (34, 46). ThPOK levels were elevated in the absence of Roquin-1/2 in all subsets, with the exception of NKT17 cells, in which they were reduced (Fig. 7D), possibly reflecting their strong polarization into this lineage.
To address the loss of NKT1 and NKT2 cells we monitored cell death, which revealed significantly increased apoptosis in Roquin-1/2–deficient and control NKT cells at stage 2 (Fig. 7E) harboring >90% of NKT17 cells. This indicates that Roquin-1/2 do not directly influence cell death but that their absence directs NKT cell differentiation into a proapoptotic compartment. NKT cells (47), and especially NKT17 cells, depend on IL-7 for their generation, homeostasis, and survival (48). Accordingly, control NKT17 cells expressed three times more IL-7R α-chain (CD127) than did NKT1 and NKT2 cells. Loss of Roquin-1/2 led to a selective increase in CD127 expression in NKT17 cells (Fig. 7F), suggesting that decreased IL-7 signaling does not underlie their enhanced propensity to die.
Together, our data indicate that Egr2 signals and STAT3-activating signals are diminished in Roquin-1/2–deficient NKT cells. Mechanisms, including undefined signals that enhance c-Maf expression, strongly drive them into the NKT17 lineage, where they are more prone to die.
Roquin paralog–deficient NKT cells are compromised in their cytokine-secretion abilities
The differentiation of NKT cell subsets and cytokine production also depend on signaling through the mechanistic target of rapamycin (mTOR) protein complexes (49), and ablation of the negative regulator tuberous sclerosis 1 (TSC1) leads to phenotypes that strikingly resemble the phenotypes caused by loss of Roquin-1/2 (50). However, we did not find evidence for increased mTOR signaling in Roquin-1/2–deficient NKT cells compared with controls. There was no difference in the phosphorylation of mTOR or downstream proteins, such as 4-EBP1 (mTORC1) and AKT (mTORC2) (Fig. 8A). Roquin deficiency controls the differentiation of proinflammatory T lymphocytes, as well as their effector functions (Fig. 8D) (25, 51). To our surprise, PMA/ionomycin stimulation induced significantly lower proportions of cytokine-producing cells in all Roquin-1/2–deficient NKT cell subsets compared with controls (Fig. 8B). Furthermore, cytokine production per cell appeared to be lower in Roquin-1/2–deficient NKT17 cells (Fig. 8C), whereas conventional effector T cells produced significantly more proinflammatory cytokines after Roquin ablation (Fig. 8D).
Therefore, in contrast to the increased expression of NKT17-specific markers, Roquin-1/2–deficient NKT cells showed a hyporesponsive phenotype that was characterized by overall defective cytokine production.
During their development in the thymus, NKT cells pass sequential stages and mature into functionally different subsets. In this article, we show that the RNA-binding Roquin-1 and -2 proteins prevent excessive differentiation into the NKT17 lineage. The numbers of NKT17 cells deficient for the Roquin paralogs exceeded their controls by >10-fold and showed strongly enhanced lineage commitment. Mixed bone marrow chimeras demonstrated that NKT17 polarization is cell intrinsic, and conditional expression of a Vα14i TCR, concomitantly with Roquin-1/2 ablation, suggested that signals early during NKT cell development play decisive roles.
It remains unclear when and how the NKT17 fate is initiated during NKT cell development. We detected selective PLZF downregulation in Roquin-1/2–deficient NKT17 cells and normal levels in the other subsets. This argues against a causal role for PLZF regulation, because enhanced PLZF levels, as seen in lethal-7 miRNA-deficiency, shift the balance from NKT1 to NKT2/17 (52). Ablation or loss-of-function mutations in ThPOK result in the production of CD8 and DN NKT cells that are characterized by low NK1.1 expression (34, 53, 54). This goes hand in hand with a dramatic polarization toward NKT17 differentiation, which is also observed, although to a much lesser extent, in haploinsufficiency (34, 46). The elevated ThPOK levels in Roquin-1/2–deficient thymocytes and NKT1/2 cells indicate that lower ThPOK levels in NKT17 cells are the consequence, rather than the cause, of exaggerated NKT17 differentiation in the absence of Roquin-1/2.
A significant reduction in ThPOK mRNA was also detected in NKT cells deficient for the mTOR negative-regulator TSC1 (50), which show similar phenotypes, including a dramatic NKT17 bias and a reduction in mature thymic and peripheral NKT cells. This phenotype was attributed and functionally linked to elevated mTORC1 activity, reduced expression of T-bet, and ICOS-dependent signals (50). In our studies, we did not detect significant differences in mTORC1 or mTORC2 pathway activation in Roquin-1/2–deficient NKT17 cells compared with control cells. Furthermore, at least in the CD8 T cell lineage, loss of Roquin activity strongly enhances the acquisition of an effector state that is characterized by high T-bet expression (24, 51). In TSC1-deficient NKT cells, reduced Tbet levels might facilitate the transcriptional acquisition of high ICOS expression, which contributes to NKT17 polarization (50). A similar outcome with respect to elevated ICOS and NKT17 differentiation was independently achieved through enhanced translation of ICOS mRNA in a Roquin-1/2–deficient situation.
ICOS, IL-6 in conjunction with TCR-induced Ca2+ signals (44), TGF-β (55), and IL-27 (56) are all implicated in the regulation of c-Maf expression. c-Maf is characteristically overexpressed in Th17 cells, TFH cells (42), and, as we show in this article, NKT17 cells. Normal TCR signal strength in NKT cells argues against a role for TCR signaling in enhancing c-Maf expression in the absence of Roquin-1/2. Because we and other investigators (45) showed that IL-6R signals are dispensable for NKT17 development, it seems more likely that cytokines, such as IL-27, trigger c-Maf expression in developing NKT17 cells. IL-27 also induces the Roquin target ICOS, which, in turn, amplifies expression of c-Maf (56). c-Maf induces an IL-17–producing cell fate, together with its binding partner Sox5t, by directly enhancing RORγt expression and by binding to additional gene loci implicated in Th17 differentiation (43), including the Roquin targets Irf4 and Nfkbiz (25). Further transcriptional activation of these and other target genes during initial NKT17 differentiation could then be potentiated by the absence of Roquin-1/2–mediated posttranscriptional control, resulting in the production of NKT17 cells at the expense of other subsets.
We cannot provide a conclusive explanation for the absence of Roquin-1/2–deficient peripheral NKT cells. The general inflammatory environment caused by the loss of Roquin paralogs in T cells could contribute to the loss of Roquin-1/2–deficient NKT cells in the periphery. The expression of a Vα14i TCR and the presence of a majority of wild-type T cells in our mixed bone marrow chimeras substantially reduce general inflammation. This might partially explain why mature Roquin-1/2–deficient NKT cells can be maintained in the periphery in both models.
Loss of Roquin functionality in conventional T cells leads to disease-causing effector T cell polarization and excessive cytokine production (20, 22, 25, 38). In striking contrast, after ablation of Roquin-1/2, all NKT cell subsets showed a decreased production of subset-specific cytokines, which, in the case of NKT17, correlates with reduced STAT3 phosphorylation. NKT cells acquire features of memory T cells and innate-like cytokine-production capabilities during their development, and it appears that they become desensitized toward autoantigen recognition during their development (27). We do not know whether loss of Roquin proteins dampens cytokine production by NKT cells through direct or indirect means. Upregulation of c-Maf is a consistent feature of Roquin-1/2–deficient thymic NKT cells. In addition to promoting the differentiation of IL-17–producing T cell subsets, c-Maf is highly expressed in exhausted CD4 and CD8 T cells. Overexpression of c-Maf was sufficient to inhibit CD8 T cell cytokine production and antitumor responses (57). Therefore, high c-Maf levels could contribute to the blunted cytokine responses of Roquin-1/2–deficient NKT cells.
NKT cells are a small TCR repertoire–restricted glycolipid-recognizing subset implicated in a large spectrum of human diseases, including infection, autoimmunity, and malignancy. During their development, they mature into functionally different subsets. In this article, we show that the posttranscriptional regulators Roquin-1 and -2 play key cell-intrinsic roles in restricting the excessive developmental generation of NKT17 cells.
We thank Julia Knogler and Martina Schmickl for technical assistance. Fluorophore-labeled mCD1d PBS-57 tetramers were provided by the National Institutes of Health Tetramer Core Facility.
This work was supported by the Deutsche Forschungsgemeinschaft through Grants SCHM2440-3 and SFB 1054 A02 (to M.S.-S.). J.C.V. and K.D.H. received Ph.D. stipends from the Ernst Schering Foundation and the Boehringer Ingelheim Fonds, respectively.
The online version of this article contains supplemental material.
Abbreviations used in this article:
median fluorescence intensity
mesenteric lymph node
mechanistic target of rapamycin
peripheral lymph node
promyelocytic leukemia zinc finger
signaling lymphocyte activation molecule
Th poxviruses and zinc finger and Krüppel family
tuberous sclerosis 1.
The authors have no financial conflicts of interest.