TGF-β promotes the postselection thymic development and peripheral function of interferon-γ- producing iNKT cells

Abstract

Interferon-γ-producing invariant natural killer T (iNKT1) cells are lipid reactive innate-like lymphocytes that are resident in the thymus and peripheral tissues where they protect against pathogenic infection. The thymic functions of iNKT1 cells are not fully elucidated but subsets of thymic iNKT cells modulate CD8 T cell, dendritic cell, B cell and thymic epithelial cell numbers or function. Here we show that a subset of murine thymic iNKT1 cells required transforming growth factor (TGF)-β induced signals for their postselection development, to maintain hallmark TGF-β induce genes, and for expression of the adhesion receptors CD49a and CD103. However, the residency-associated receptor CD69 was not TGF-β-signaling dependent. Recently described CD244⁺ c2 thymic iNKT1 cells, which produce IFN-γ without exogenous stimulation and have NK-like characteristics, reside in this TGFβ-responsive population. Liver and spleen iNKT1 cells do not share this TGF-β gene signature but nonetheless TGF-β impacts liver iNKT1 cell phenotype and function. Our findings provide insight into the heterogeneity of mechanisms guiding iNKT1 cell development in different tissues and suggest a close association between a subset of iNKT1 cells and TGF-β producing cells in the thymus that support their development.

Introduction

Invariant natural killer T (iNKT) cells are a subset of innate like T lymphocytes that express an invariant T cell receptor that recognizes glycolipid antigens in the context of the non-classical major histocompatibility complex protein CD1d (1). These cells comprise a large portion of T lymphocytes in peripheral tissues. As much as 40% and 5% of liver T cells are iNKT cells in mice and humans respectively (2, 3). Most liver iNKT cells are tissue resident, interferon (IFN)-γ producing iNKT1 cells (3, 4). These cells patrol the sinusoidal space to protect against invading pathogens such as the spirochete Borrelia burgdorferi, the causative agent of Lyme’s Disease, and viral infection such as Hepatitis C Virus. (3, 5, 6). Chronic activation of iNKT cells is associated with disease severity in nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH) (2, 3, 7). iNKT cells have also been implicated in protection from pathogens in the lung and other tissues but may contribute to inflammatory disease such as asthma and atherosclerosis.

iNKT cell development initiates in the thymus from CD4⁺CD8⁺ thymocytes after rearrangement of the iNKT cell receptor (8, 9). During selection, iNKT cells are recruited to the thymic medulla by CCL21 and a subset of these immature postselection iNKT cells emigrate to seed peripheral tissues where they continue their maturation (10-12). The cells remaining in the thymus differentiate into multiple effector subsets but the majority of these cells in C57BL/6 mice are thymic resident IFN-γ-producing iNKT1 cells (13, 14). Cytokines produced by thymic iNKT cells, in particular IL-4, have the potential to impact the function of multiple cell types including CD8 T cells and dendritic cells (15-17). While our understanding of iNKT cell selection is well developed, the mechanisms that promote postselection iNKT cell differentiation in different tissues is not well understood.

Here, using a novel model for Cre-mediated recombination in postselection iNKT1 cells, we demonstrate that thymic iNKT1 cells differ from liver and spleen iNKT1 cells by their expression of genes associated with transforming growth factor (TGF)-β signaling and their lower expression of genes associated with IL6-JAK-STAT3 signaling, mTORC1 signaling and fatty acid metabolism. While TGF-β is known to regulate iNKT cell selection (18), we demonstrated that TGF-β signaling in postselection thymic iNKT1 cells enforced a classic TGF-β-associated gene signature, including repression of migratory genes and a cytokine signaling gene signature and induction of multiple residency-associated adhesion receptors. Postselection iNKT1 cells that cannot respond to TGF-β due to deletion of Tgfbr2 failed to differentiate into cells that expressed the residency-associated adhesion receptors CD49a and CD103 and the NK cell receptor CD244, which is associated with high baseline IFN-γ production (19). In contrast to studies in which Tgfbr2 was deleted in other IFN-γ producing cell types, including preselection iNKT cells (18), NK cells (20), or CD8 T cells that were differentiated into tissue-resident memory cells (Trm) (21-23), we found no role for TGF-β signaling in the control of the transcription factor T-bet or T-bet target genes. Despite the reduced TGF-β gene signature in liver iNKT1 cells, TGF-β signaling was required for their optimal expression of CD49a and supported their production of IFN-γ and IL-4. Our data reveal a selective requirement for TGF-β signaling in the postselection generation of thymic CD49a⁺CD103⁺ and CD244⁺ iNKT1 cells and optimal liver iNKT1 function.

Materials and Methods

Mice

B6.129X-Gt(ROSA)26Sor^tm(EYFP)Cos/J (Rosa26-Stop-floxed-YFP), B6.CBA-Tg(Tbx21-cre)11Dlc/J (Tbx21^Cre) and B6;129-Tgfbr2^tm1Karl/J (Tgfbr2^F/F) mice were purchased from Jackson Labs (24-26). All mice were housed at the University of Chicago in concordance with the guidelines of the University of Chicago Institutional Animal Care and Use Committee.

Flow Cytometry Analysis

Mice were sacrificed using C0₂ and thymus and liver were isolated, dissociated and passed through a mesh filter to remove debris. Liver lymphocytes were isolated using Lympholyte M (CedarLane). Cells washed twice and stained on ice for 30 minutes at a concentration of 2 x 10⁷ cells/100 ul, washed two times and analyzed on a Fortessa flow cytometer (BD Biosciences). BV staining buffer was used when multiple BV dyes were used together (BD Biosciences). The Foxp3 staining kit was used for intracellular transcription factor staining and the Cytoperm/Cytofix Plus staining kit was used for cytokine staining (BD Biosciences). Data was analyzed using FlowJo (TreeStar). CD1d-PBS57 tetramers were titrated before use along with CD1d tetramers lacking glycolipid. Antibodies used include TCRβ, CD49a, CD103, CD69, ICAM-1, LFA-1, CD44, CD244 NK1.1, T-BET, PLZF, RORγt, GZMB, IL-4 and IFN-γ. Fluorochromes are indicated in the flow plots.

Cytokine Production Assays

For in vitro cytokine assays, 2 x 10⁶ cells thymocytes were transferred to 12 well culture plates and incubated with PMA (20 ng/ml), ionomycin (1 mg/ml) and Golgi Plug (1:1000) for 5 hours. Following incubation, cells were washed with FACS Buffer and stained with CD1d-tetramers and surface receptor before intracellular staining for IL-4 and IFN-γ using the Foxp3 staining kit.

In vivo blocking of S1P Lyase

Mice were given DOP (30 mg/ml) in drinking water containing glucose (10 g/L) for 10 days prior to sacrifice and flow cytometry analysis (27). Ctrl mice were given glucose containing drinking water.

RNA-sequencing and data processing

RNA was isolated from sorted iNKT1 cells using the RNAeasy MicroKit (Qiagen). RNA libraries were constructed using Nugen’s Ovation Ultralow Library systems followed by 76 cycles of NextSeq500 sequencing (for WT thymus, liver and spleen analysis) or 100 cycles of sequencing on NovaSeq (Ctrl versus cKO analysis). Raw sequence reads were trimmed using Trimmomatic v 0.33 (TRAILING:30 MINLEN:20) and then aligned to the mouse mm10 genome with STAR v2.5.2 (28, 29). Reads were assigned to genes using the htseq-count tool from HTSeq v 0.6.1 and gene annotations were from Ensembl release 78 (30, 31). The R package EdgeR was used to normalize the gene counts and to calculate differential expression statistics for each gene for each pairwise comparison of sample groups (32). Gene set enrichment analysis was performed using gene sets from the Hallmark Pathways of MSigDB (33). Genes were considered differentially expressed if the fold change was > or = 2 and false discover rate <0.05. RNA-sequencing data can be accessed in the Gene Expression Omnibus https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi (GSE215128 and GSE215338).

Statistics

The GraphPad Prizm software package was used for statistical analysis. A Student’s t-test was used to establish the level of significance between two groups. Groups of 3 or more were assessed using ANOVA with multiple comparisons with the Geisser-Greenhouse correction. *P<0.05, **P<0.01, ***P<0.005.

Results

Thymic iNKT1 cells express genes associated with TGF-β signaling

To trace postselection iNKT1 cells in different tissues we crossed a Tbx21^Cre bacterial artificial chromosome transgenic mouse to Rosa26-Stop-floxed-YFP mice (24, 25). Tbx21 encodes T-BET, the signature transcription factor for iNKT1 cells, and therefore we anticipated that iNKT1 cells and their effector fate restricted progenitors would be YFP⁺. Indeed, we found that CD1d-tet⁺TCRβ⁺ iNKT cells that were YFP⁺ in the thymus, liver and spleen were primarily CD44⁺NK1.1⁺, a phenotype associated with iNKT1 cells (Fig. 1). We used this model to isolate iNKT1 cells from the thymus, liver, and spleen and compare their gene programs by RNA-sequencing. A prior study demonstrated that iNKT1 cells isolated from different tissues, using a different selection strategy, had essentially identical gene programs (34). Consistent with this, we found that the transcriptome of liver and spleen iNKT1 cells were nearly identical with < 20 differentially expressed genes (DEGs) (Fig. 2A). The transcriptome of thymus and liver iNKT1 cells differed by 124 genes (adj.p < 0.01) and thymic and splenic cells differed by 107 genes (Fig. 2A). While these are small differences, there were specific genes of interest among the DEGs. Using Gene Set Enrichment Analysis (GSEA), we observed an enrichment of multiple Hallmark pathways in the liver as compared to the thymus including IL6-JAK-STAT3 signaling, mTORC1 signaling and fatty acid metabolism (Fig. 2B, S1). As expected, similar GSEA pathways were enriched in the liver and the spleen when compared to the thymus (Fig. S1).

FIGURE 1:

Validation of the Tbx21^Cre mouse and analysis of RNA-seq data. Flow cytometry analysis of NKT cells from the thymus, liver, and spleen of Tbx21^Cre;Rosa26-stop-flox-YFP mice. iNKT cells were identified as CD1d-Tet⁺ and TCRβ⁺. YFP⁺ iNKT cells were then examined for expression of NK1.1 and CD44. Mature iNKT1 cells express NK1.1 and progenitors that are T-BET⁺ can be found among CD44⁺NK1.1⁻ cells. This mouse was 6 weeks old, but similar results were observed in mice between the ages of 6 and 16 weeks of age.

FIGURE 2:

Thymic iNKT1 cells are enriched for genes associated with TGF-β signaling compared to liver and spleen iNKT1 cells. RNA was sequenced from iNKT1 cells isolated from the thymus, liver, and spleen of 10-week-old WT mice and normalized read counts were compared. (A) Number of differentially expressed genes (DEG) between WT thymus, liver, and spleen iNKT1 cells. (B) Hallmark Pathways identified by GSEA as enriched in liver compared to thymus iNKT1 cells. (C) Reads for 3 replicate RNA-seq samples from WT thymus (T, blue), liver (L, red), and spleen (S, green) for Itgb3, Dhfr, Cpt1a and Decr, which are representative genes from the pathways identified in (B). (D) Hallmark Pathways identified as enriched in thymus as compared to liver iNKT1 cells. (E) Normalized reads from replicate RNA-seq samples showing genes representative of the pathways identified in (D). (F) RNA-seq reads for Tgfbr1 and Tgfbr2. *P<0.05, **<P0.01, ***P<0.005 by ANOVA with multiple comparisons. iNKT1 cells for RNA-seq were isolated from mice that were 6 or 7 weeks of age.

In the IL6-JAK-STAT3 pathway, Itgb3 was more abundant in liver and spleen compared to the thymus iNKT1 cells (Fig. 2C). Socs3 mRNA, encoding a suppressor of cytokine signaling that binds STAT3 (35), was also more highly expressed in liver and spleen iNKT1 cells although this difference did not quite reach significance (Fig. S1). The mTORC1 pathway was selectively enriched in the liver to thymus comparison and Dhfr mRNA was highest in the liver (Fig. 1C, S1). Interestingly, mRNA encoding enzymes in the fatty acid metabolism pathway, including Cpt1a and Decr1 mRNAs, encoding carnitine palmityoyltransferase 1A and 2,4 dienoyl-CoA reductase respectively, were higher in both the liver and spleen compared to the thymus even though GSEA identified this pathway as selectively enriched in the liver (Fig. 2C, S1C). These data suggest that liver and spleen iNKT1 cells have active IL6-JAK-STAT3 signaling and may utilize fatty acid metabolism but that these pathways are less active in thymic iNKT1 cells.

Thymic iNKT1 cells were significantly enriched for only the hallmark MYC-Targets pathway but only one gene in this pathway, Rabepk, encoding a mannose 6-phosphage receptor transport protein, reach statistical significance when tested by ANOVA (Fig. 2D, E). In contrast, enrichment of the TGF-β signaling pathway did not reach significance by GSEA; however, multiple genes in this pathway were differentially expressed (Fig. 1D, E). These genes include Cdh1, encoding the adhesion protein Cadherin, Itgae, encoding the alpha chain of the integrin CD103, Itga1, encoding the alpha chain of CD49a, and Vim, encoding the type III intermediate filament protein vimentin (Fig. 2E). Tgfbr1 and Tgfbr2, encoding components of the TGF-β receptor, were expressed similarly in iNKT1 cells from all tissues, suggesting that differences in TGF-β signaling-associated genes were not a consequence of differential TGF-β receptor expression (Fig. 2F). Tbx21, encoding the transcription factor T-BET, which is required for development of iNKT1 cells and CD8 Trm cells but is dampened by TGF-β signaling in many cell types, was also not a differentially expressed gene (DEG) (Fig. S1) (21, 22, 36-38). These data indicate that the transcriptome of iNKT1 cells is very similar in the thymus, liver, and spleen, but there are DEGs associated with cytokine signaling, mTORC1, fatty acid metabolism, MYC and TGF-β signaling.

Differential expression of adhesion receptors on thymic and liver iNKT1 cells

The enrichment of some TGF-β signaling induced transcripts was of interest given that iNKT1 cells are tissue-resident cells with hybrid NK cell, ILC1, and CD4 and CD8 Trm cell characteristics (39). Response to TGF-β depends on Runx3, which is expressed in CD8 Trm but not CD4 Trm cells, resulting in a differential response to this cytokine (40). Our RNA-seq data revealed a modest increase in Runx1 mRNA in liver and spleen as compared to thymus iNKT1 cells and broad expression of Runx3 mRNA in all iNKT1 cells, consistent with the ability of thymic iNKT1 cells to respond to TGF-β (Fig. S1). Itgae and Itga1, components of CD103 and CD49a respectively, are induced on subsets of CD8 Trm cells by TGF-β produced from epithelial cells (40-43). By flow cytometry we confirmed that CD103 was expressed on a subset of iNKT1 cells in the thymus but not in the liver (Fig. 3A, B). By contrast, CD69, a protein expressed on tissue-resident cells that functions as an inhibitor of the S1P receptor S1P1 (44, 45), was highly expressed on both thymic and liver iNKT1 cells with a subtle increase in frequency on liver iNKT1 cells (Fig. 3A, B). CD49a was expressed on both thymic and liver iNKT1 cells although at a lower frequency and lower mean fluorescence intensity (MFI) on liver iNKT1 cells (Fig. 3A, C). A substantial population of thymic iNKT1 cells expressed CD49a along with CD103 and CD69 (Fig. 3A). These receptors are associated with tissue residency of CD8 T cells; however, liver iNKT1 cells are known to require LFA-1 and ICAM-1 to maintain tissue residency (6). Notably, the MFI for ICAM-1 and LFA-1 on liver iNKT1 cells was higher than on thymic iNKT1 cells (Fig. 3D, E). CD44, another adhesion receptor associated with mature iNKT cells (46), was also expressed at slightly higher levels on liver as compared to thymic iNKT1 cells (Fig. 3F). These data reveal striking differences in the tissue adhesion programs of thymic and liver iNKT1 cells and are consistent with a TGF-β induced program in a subset of thymic cells.

FIGURE 3:

Thymic iNKT1 cells uniquely express TGF-β-associated adhesion proteins. (A) Flow cytometry for CD49a versus CD103 (top) or CD69 (bottom) on iNKT1 cells from the thymus (left) or liver (right). n = 4 (B) Summary of multiple flow cytometry experiments showing the percent of iNKT1 cells that are positive for CD103 or CD69 or (D) CD49a in the thymus (grey) or liver (black). MFI for CD49a is shown in the right panel. Each data point is an independent mouse. Representative flow cytometry histograms (top) and summary of multiple experiments (bottom) for (D) ICAM-1, (E) LFA-1 or (F) CD44 on thymus (shaded) or liver (open) iNKT1 cells. Light shaded histogram is isotype control. Data are representative of 3 experiments. *P<0.05, **<P0.01, ***P<0.005 by Student’s t-test. Mice were between 6 and 10 weeks of age.

TGF-β signaling is required for development of CD49a⁺CD103⁺ thymic iNKT1 cells

Previous studies demonstrated a requirement for TGF-β during selection of iNKT cells and a possible role for TGF-β in promoting iNKT17 differentiation while limiting iNKT1 differentiation (18, 47, 48). In those studies, Tgfbr2 was inactivated in all T cells prior to T cell receptor-mediated selection making it difficult to dissociate effects on selection from effects on differentiation. To address the requirement for TGFβ signals in iNKT1 cell development postselection, we crossed Tbx21^Cre Rosa26-StopFlox-YFP mice to mice with floxed alleles of Tgfbr2, encoding TGFβRII (24), to create Tgfbr2^F/F Tbx21^Cre Rosa26-StopFlox-YFP (cKO) mice. In these mice, the total number of CD1d-Tet⁺TCRβ⁺ YFP⁺ iNKT cells (iNKT1) in the thymus was decreased by approximately 30% compared to Tbx21^Cre Rosa26-StopFlox-YFP (Ctrl) mice (Fig. 4A. B). Using antibodies to PLZF and RORγt, the frequency of PLZF⁻RORγt⁻ (iNKT1) cells among CD1d-Tet⁺TCRβ⁺ cells decreased while the frequency of PLZF⁺RORγt⁻ (iNKT2) and PLZF^loRORγt⁺ (iNKT17) iNKT cells increased (Fig. 4C). Quantification of CD1d-Tet⁺TCRβ⁺PLZF⁻RORγt⁻ iNKT1 cells also indicated an approximate 30% decrease in cKO mice whereas CD1d-Tet⁺TCRβ⁺PLZF⁺RORγt⁻ iNKT2s and CD1d-Tet⁺TCRβ⁺PLZF^loRORγt⁺ iNKT17s were not different from Ctrl (Fig. 4D, E). In CD8 Trm cells, CD4 Th1 cells and NK cells, TGF-β signaling dampens the expression of T-BET, despite that T-BET is required for the development of these cells (20-22, 37). T-BET is also a central regulator of iNKT1 cell development (38, 49) and therefore we examined its expression in iNKT cells from Ctrl and cKO mice. As expected, the frequency of T-BET⁺ cells among CD1d-Tet⁺TCRβ⁺ cells was lower in cKO mice compared to Ctrl mirroring the decreased frequency of iNKT1 cells; however, the MFI for T-BET was similar (Fig. 4F, G). To address the functionality of cKO thymic iNKT1 cells we stimulated these cells in vitro with PMA + ionomycin for 5 hours prior to measuring IFN-γ and GzmB. There were fewer thymic CD1d-Tet⁺TCRβ⁺ cells from cKO mice that produced IFN-γ, and the total number of IFN-γ producing iNKT cells was nearly half of that from Ctrl mice (Fig. 4H, I). However, the frequency of GZMB⁺ iNKT cells was slightly higher in cKO mice than in Ctrl (Fig. 4I). Taken together, these data demonstrate that Tgfbr2 is required for the optimal development of a subset of postselection iNKT1 cells.

FIGURE 4:

TGFβRII is required for the generation of thymic iNKT1 cells. (A) Flow cytometry analysis for thymic CD1d-tet⁺TCRβ⁺ (top) and YFP⁺ (bottom) iNKT1 cells in Ctrl and cKO mice. (B) Summary of thymic iNKT1 cell numbers in Ctrl (blue) and cKO (black) mice based on the staining strategy shown in (A). (C) Expression of PLZF and RORγt in thymic iNKT cells to identify iNKT1, iNKT2 and iNKT17 as indicated. (D) Summary of thymic iNKT1 cell numbers in Ctrl and cKO mice as determined in (C). (E) Summary of number of thymic iNKT2 and iNKT17 in Ctrl and cKO mice as determined in (C). (F) Expression of PLZF and T-BET identifying iNKT1 and iNKT2 cells among total iNKT cells in the thymus of Ctrl (blue) and cKO (black) mice. (G) Summary of the MFI for T-BET thymic iNKT1 cells identified as in (F). (H) GZMB and IFN-γ in thymic iNKT cells 5 hours after stimulation with PMA + ionomycin in vitro. (I) Summary of the number of thymic IFN-γ⁺ iNKT cells (left) or the % GZMB⁺ in Ctrl and cKO mice. *P<0.05, **<P0.01, ***P<0.005. ****P<0.001 by Student’s t-test. Mice were between 6 and 10 weeks of age.

The iNKT1 cells remaining in the cKO thymus lacked expression of CD103 but the frequency of CD49a⁺CD103⁻ cells was not increased suggesting that the development of CD49a⁺CD103⁺ iNKT1 cells, rather than just their expression of CD103, depended on TGFβ signaling (Fig. 5A, B). The MFI of CD49a on CD49a⁺CD103⁻ iNKT1 cells was reduced, and CD49a⁻CD103⁻ iNKT1 cells increased, indicating that TGFβRII may be required for proper expression of CD49a (Fig. 4A, B). Interestingly, CD69 was expressed on a higher frequency of cKO iNKT1 cells resulting in a similar total number of CD69⁺ iNKT1 cells in Ctrl and cKO mice (Fig. 5C, D). A recent study identified a small population of thymic CD244⁺ iNKT cells as constitutive IFNγ producing cells with migratory properties akin to NK cells (19). We found that CD244⁺ cells are CD103⁺ in Ctrl mice and are absent from the thymus of cKO mice (Fig. 5E, F). These data demonstrate that TGFβRII is required for development of CD49a⁺CD103⁺ and CD244⁺ thymic iNKT1 cells.

FIGURE 5:

TGFβRII is required for the generation of thymic CD49a⁺CD103⁺ iNKT1 cells. (A) Expression of CD49a versus CD103 (top) on Ctrl (left) or cKO (right) thymic iNKT1 cells. (B) Summary of the number of thymic CD49a⁺CD103⁺ or CD49a⁺CD103⁻ iNKT1 cells (top), or CD49 MFI and number of CD49a⁻CD69⁻ iNKT1 cells (bottom) in the thymus of Ctrl (grey) and cKO (black) mice. (C) Expression of CD49a versus CD69 on Ctrl or cKO thymic iNKT1 cells. (D) Summary of the number of CD69⁺ thymic iNKT1 cells in Ctrl and cKO mice. (E) Expression of CD244 versus CD103 on thymic iNKT1 cells from Ctrl (left) and cKO (right) mice. (F) Summary of the number of CD244⁺ thymic iNKT1 cells in Ctrl and cKO mice. **<P0.01, ***P<0.005 by Student’s t-test. Mice were between 7 and 9 weeks of age.

TGFβRII promotes a gene program involving multiple adhesion receptors in thymic iNKT1 cells

To gain a global view of the gene program promoted by TGF-β signaling in thymic iNKT1 cells we performed RNA-sequencing using Ctrl and Tgfbr2-deficient thymic iNKT1 cells. We found 110 DEGs (adj. p < 0.01), 68 that were more highly expressed in Ctrl and 42 that were more highly expressed in cKO iNKT1 cells (Fig 6A). GSEA revealed that Ctrl cells were enriched for only one pathway, the Hallmark TGF-β signaling pathway, whereas cKO cells were enriched for multiple pathways including the IL-6-JAK-STAT3 pathway (Fig. 6B, S2). Genes associated with the IL6-JAK-STAT3 pathway, including Socs3 and Itgb3, were both significantly higher in cKO as compared to Ctrl cells (Fig. 6C). Tgfbr2 mRNA was confirmed to be decreased in the cKO cells as was Smad7 mRNA, a TGF-β -induced inhibitor of TGF-β signaling (Fig. 6D). Classic targets of TGF-β signaling such as Cdh1, Itgae, Ski, Vim, and Itga1 were reduced in cKO iNKT1 cells (Fig. 6E). Inpp4b, encoding inositol polyphospate-4-phospatase type II, which was recently implicated in TGF-β receptor endocytosis (50), was also decreased in cKO cells (Fig. S2). Multiple genes associated with migration were increased in cKO cells including Mmp9, S1pr5, Sell and Hif1a (Fig. 6F, S2). Interestingly, the gene encoding HOBIT, Zfp683, a transcription factor that functions redundantly with BLIMP1 to enforce tissue residency gene programs, was also decreased (Fig. 6F) (51). However, Prdm1 mRNA, encoding BLIMP1, was not decreased in the absence of Tgfbr2. Tbx21 mRNA was not impacted, consistent with our observation that T-BET is expressed similarly in Tgfbr2 cKO and Ctrl iNKT1 cells (Fig. S2). Eomes mRNA, encoding a T box binding transcription factor related to T-BET, was very low in both Ctrl and cKO thymic iNKT1 cells (Fig. S2). Interestingly, Zeb2, a known T-BET target gene whose protein product regulates S1pr5 (52, 53), was also not a DEG despite that S1pr5 mRNA increased in cKO iNKT1 cells (Fig. 6F, S2). These data implicate TGFβR2 in the regulation of classic TGF-β target genes involved in adhesion and migration in iNKT1 cells and in the repression of IL6-JAK-STAT3 signaling but not in the regulation of critical iNKT1 transcription factors.

FIGURE 6:

TGFβRII promotes a classic TGF-β gene signature and represses an IL6-STAT3 signaling signature in thymic iNKT1 cells. (A) Heat map for differentially expressed genes between Ctrl and cKO thymic iNKT1 by RNA-seq. (B) Top enriched Hallmark Pathways by GSEA in Ctrl versus cKO thymic iNKT1 cells. (C) Normalized reads for replicate RNA-seq samples for representative IL6-JAK-STAT3 signaling genes, (D) Tgfbr2 and Smad7, (E) classic TGF-β signaling targets, and (F) genes associated with migration. (G) Graph of Log2FC and adj.p of genes in Ctrl and cKO thymic iNKT1 cells selected for those that are TGF-β regulated in skin CD8 Trm cells (54). Colors show genes that are up-regulated (blue) or down-regulated (red) or not changed (black) in our dataset. A subset of genes are labeled for clarity. *P<0.05, **<P0.01, ***P<0.005 by Student’s t-test. The iNKT1 cells for the RNA-seq were isolated from mice at 6.5, 5, and 15 weeks of age.

The gene program regulated by TGF-β in iNKT1 cells shares significant overlap with that of skin and salivary gland CD8 Trm cells (54). However, there were also some notable differences including Runx3, S1pr1, Junb, Cxcr4, and Lef1, which were identified as TGF-β regulated in skin CD8 Trm (54) but were not DEGs in our dataset (Fig. 6G). We also examined the expression of genes associated with CD244⁺ iNKT cells and found that only a subset of these genes decreased in cKO thymic iNKT1 cells (Fig. S2). For example, Cdh1, Itga1, Cd244 and Cish were identified as associated with CD244⁺ cells and these genes decreased in cKO iNKT1 cells. In contrast, Ccl5, Itga5, Trf, and Serpinb6b were also associated with CD244⁺ cells but these genes were increased in cKO thymic iNKT1 cells. Therefore, a subset of Trm associated genes and genes associated with CD244⁺ iNKT1 are decreased in cKO thymic iNKT1 cells but these gene programs do not completely overlap.

Tgfbr2 cKO iNKT1 cells are not lost from the thymus due to emigration

Thymic iNKT1 cells are tissue-resident cells and CD103 is associated with the Trm phenotype (55). However, a previous study demonstrated that CD103 is not required for thymic retention of iNKT cells (10). Nonetheless, we considered the possibility that multiple TGF-β regulated genes could contribute to thymic iNKT1 residency, and that their loss could promote emigration of the CD103⁺ population. To test this, we analyzed iNKT1 cells after 10 days of continuous feeding of mice with the S1p lyase inhibitor 4-deoxypyridoxine (DOP) (27). S1P lyase is essential for maintaining the gradient of S1P that promotes the emigration of thymocytes (27). Disrupting this gradient results in retention of cells that are emigrating including mature CD4 and CD8 T cells. As expected, treatment of Ctrl and cKO mice with DOP for 10 days was sufficient to promote an accumulation of thymic TCRβ⁺ T cells (Fig. S3). As shown previously (11), thymic iNKT1 cell numbers increased in Ctrl mice after inhibition of emigration (Fig. S3). In contrast, thymic iNKT1 numbers increased only subtly in cKO mice treated with DOP (Fig. S3B, D). Importantly, the frequency of CD49a⁺CD103⁺ iNKT1 cells did not change in either Ctrl or cKO mice after treatment with DOP (Fig. S3), indicating that CD49a⁺CD103⁺ iNKT1 cells in Tgfbr2 cKO mice are not reduced in number due to S1p-mediated emigration.

Tgfbr2 is required for optimal expression of CD49a and liver iNKT1 cell function

Liver iNKT1 cells expressed some TGF-β-associated genes, such as Inpp4a and Vim, and they expressed surface CD49a (Fig. 2E). Therefore, we examined the consequence of deleting Tgfbr2 with Tbx21^Cre on liver iNKT1 cells. In contrast to the thymus, the frequency of iNKT1 cells among liver lymphocytes was not affected in Tgfbr2 cKO mice (Fig. 7A, B); however, the frequency of CD49a⁺ cells and the MFI of CD49a was decreased (Fig. 7C, D). CD69 continued to be expressed and the frequency of CD49a⁻CD69⁻ iNKT1 cells did not increase significantly (Fig. 7C, E), suggesting that CD49a is downregulated but the cells continue to express CD69. The observation that TGFβRII-deficiency impacted CD49a on liver iNKT1 cells prompted us to test the functional capacity of these cells. TGF-β impairs the functionality of CD8 Trm cells in the skin and salivary gland (54). By contrast, we found that liver iNKT1 cells from Tgfbr2 cKO mice were somewhat less capable of producing both IFN-γ and IL-4 in response to an in vivo injection of αGalCer than liver iNKT1 cells from Ctrl mice (Fig. 7E, F). Liver iNKT1 cells with or without TGFβRII failed to produce GZMB. Taken together, these data demonstrate that despite the absence of a strong TGF-β gene signature in liver iNKT1 cells, TGF-β modulates gene expression in these cells and optimizes their ability to make cytokines in response to antigen stimulation.

FIGURE 7:

TGFβRII regulates CD49a and cytokine production in liver iNKT1 cells. (A) Flow cytometry analysis for liver CD1d-tet⁺TCRβ⁺YFP⁺ iNKT1 cells in Ctrl (left) and cKO (right) mice. (B) Summary of the percent of iNKT1 cells among liver lymphoid cells in Ctrl (red) and cKO (black) mice. (C) Expression of CD9a versus CD103 (top) or CD69 (bottom) on Ctrl (red) or cKO (black) liver iNKT1 cells. (D) Summary of the percent of iNKT1 cells that are CD49a⁺, the MFI of CD49, and the percent that are CD49a⁻CD69⁻. (E) Intracellular staining for IL-4 and IFN-γ (top) or GZMB and IFN-γ (bottom) and (F) summary of the percent of liver iNKT1 cells that are IFN-γ⁺ or IL-4⁺ two hours after injection of αGalCer. *P<0.05, **<P0.01, ***P<0.005 using Student’s t-test. Mice were between 6 and 12 weeks of age.

Discussion

In summary, our study revealed that iNKT1 cells in the thymus differ from those in the liver and spleen by only a few genes but that these genes function in processes important to iNKT1 development and function. In particular, thymic iNKT1 cells expressed genes induced by TGF-β signaling and had lower expression of genes associated with IL6-JAK-STAT3 signaling and fatty acid metabolism. We demonstrated a role for TGF-β in supporting a TGF-β gene signature as well as the proper development of cells that express CD49a, CD103 and CD244. Our data support the hypothesis that TGF-β functions to promote the development and maturation of thymic iNKT1 cells. Liver iNKT1 cells, while not enriched for the TGF-β gene signature, do express a subset of TGF-β target genes and require TGF-β signaling for optimal expression of CD49a and production of IFN-γ in response to αGalCer. These data demonstrate broad functions for TGF-β signaling in the phenotype and function of postselection iNKT1 cells.

Our data suggest that a subset of iNKT1 cells are in contact with TGF-β producing cells in the thymus, possibly cortical or medullary thymic epithelial cells (cTEC or mTEC). Indeed, it is known that LTβR-dependent CCL21⁺ CD104⁺ MHCII^low mTECs impact iNKT1 cells through provision of IL-15Rα-IL-15 complexes (56, 57). It is feasible that these mTECs cells also provide TGF-β to promote the maturation of iNKT1 cells. Interestingly, Aire⁺ mTECs are dependent on CD1d, suggesting a direct role for iNKT cells in mTEC regulation, likely through provision of RANKL on iNKT2 cells (57, 58). However, IFN-γ has effects on Aire expression in thymic mTEC raising the possibility that CD49a⁺CD103⁺ iNKT1 cells could also impact mTEC (59). This possibility requires further exploration. Indeed, we found that CD244⁺ iNKT1 cells, a subset of iNKT1 cells that produce IFN-γ without the need for exogenous stimulation, are within the TGF-β responsive CD49a⁺CD103⁺ population, suggesting that they could encounter TEC. Our data reveal a role for TGF- β signaling in guiding the generation of a subset of thymic iNKT1 cells, including iNKT1 cells that constitutively express IFN-γ.

We also found that TGF-β impacted expression of CD49a on liver iNKT1 cells and modestly impacted their ability to produce cytokines after activation. This effect of TGF-β on liver iNKT1 cells likely explains why the hallmark TGF-β signaling pathway was not significantly enriched in the differential comparison of liver and thymic iNKT1 cells despite the strong expression of TGF-β-dependent genes in the thymic iNKT1 cells. Our data are consistent with the possibility that TGF-β directly impacts liver iNKT1 cells. However, it is possible that this effect of TGF-β is due to its function on precursors in the thymus that subsequently migrate to the liver after being imprinted by TGF-β. In this regard, we note that while iNKT1 cells that leave the thymus were proposed to be upstream of effector fate differentiation some of these recent thymic emigrants expressed detectable T-BET and could be deleted of Tgfbr2 in our mice (10). Moreover, recent scRNA-seq studies revealed multiple iNKT cells subsets that may have the potential to emigrate from the thymus, some of which already express Tbx21 (12). Therefore, it is feasible that emigrating iNKT1 progenitors are imprinted by TGF-β signaling prior to leaving the thymus. However, our experiments in which the S1P gradient was inhibited with DOP revealed that thymic iNKT1 cell numbers are not reduced in Tgfbr2 cKO because of increased thymic emigration.

Our study revealed substantial heterogeneity in the thymic iNKT1 cell population through differential expression of multiple cell surface receptors. This observation raises the question of whether all the changes in gene expression that we observe in cKO iNKT1 cells are the consequence of gene regulation by TGF-β signaling or whether some of these changes are due to changes in the population composition. The loss of hallmark TGF-β target genes after deletion of Tgfbr2 seems likely to be the consequence of direct regulation by TGF-β. Moreover, the loss of CD49a⁺CD103⁺ iNKT1 cells in cKO mice correlates well with the decrease in iNKT1 cell numbers in these mice. We would anticipate that genes expressed in CD49a⁺CD103⁻ and CD49⁻CD103⁻ populations are enriched in our dataset, because of the relative increase in their numbers and are presumably represented among the DEGs that increase. Indeed, we anticipate that cells lacking CD103 and CD49a are not regulated by TGF-β signaling and remain intact in the cKO thymus. We also note that genes regulated by TGF-β in CD8 Trm and genes expressed in CD244+ iNKT1 cells were not uniformly altered in cKO iNKT1 cells, an observation that may suggest that changes in gene expression rather than population alterations also occur. The best way to distinguish between differential gene expression and loss of specific cell subsets would be through single-cell RNA-sequencing of Ctrl and cKO iNKT1 cells, the absence of which is a limitation of our study. While these questions remain to be investigated, our data demonstrate that TGF-β plays a critical role in regulating gene expression and homeostasis in thymic iNKT1 cells.

Key Points:

Signaling and metabolic pathways differ in thymic versus liver and spleen iNKT1 cells

TGF-β promotes the postselection maturation of thymic iNKT1 cells.

TGF-β controls TGF-β and cytokine signaling gene signatures in thymic iNKT1 cells.