Innate T cells, including CD1d-restricted invariant natural killer T (iNKT) cells, are characterized by their rapid activation in response to non-peptide antigens, such as lipids. While the transcriptional profiles of naive, effector, and memory adaptive T cells have been well studied, less is known about the transcriptional regulation of different iNKT cell activation states. Here, using single-cell RNA-sequencing, we performed longitudinal profiling of activated murine iNKT cells, generating a transcriptomic atlas of iNKT cell activation states. We found that transcriptional signatures of activation are highly conserved among heterogeneous iNKT cell populations, including NKT1, NKT2, and NKT17 subsets, and human iNKT cells. Strikingly, we found that regulatory iNKT cells, such as adipose iNKT cells, undergo blunted activation and display constitutive enrichment of memory-like cMAF+ and KLRG1+ populations. Moreover, we identify a conserved cMAF-associated transcriptional network among NKT10 cells, providing novel insights into the biology of regulatory and antigen-experienced iNKT cells.
iNKT cellsactivationtranscriptional remodelingadipose tissuescRNA-Seqcellular metabolismResearch organismHumanMousehttp://dx.doi.org/10.13039/100000041American Diabetes Association1-16-JDF-061LynchLydiahttp://dx.doi.org/10.13039/100000002National Institutes of HealthR01AI134861LynchLydiahttp://dx.doi.org/10.13039/501100000781European Research Council679173LynchLydiahttp://dx.doi.org/10.13039/501100001602Science Foundation IrelandKaneHarryLynchLydiahttp://dx.doi.org/10.13039/100000002National Institutes of HealthAI113046BrennerMichael BThe funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.Author impact statementInvariant natural killer T (iNKT) cells have some shared yet some distinct metabolic and transcriptional programs for their in vivo effector functions, including a novel population of memory-like regulatory iNKT cells.Introduction
Activation of T cells following recognition of cognate antigen is essential for mounting effective immune responses against pathogens and tumors (Kumar et al., 2018). Typically, in the case of MHC-restricted adaptive CD4+ and CD8+ T cells, this requires extensive transcriptional remodeling over several days to facilitate proliferation and differentiation of naive T cells into clonal effector populations that traffic to sites of infection or tissue damage (Chen et al., 2018a). Transcriptional and metabolic remodeling is also needed to generate memory T cells that can be rapidly reactivated following secondary antigen encounter during reinfection (Chen et al., 2018a). Innate T cells, including CD1d-restricted invariant natural killer T (iNKT) cells, contrast and complement this paradigm by exiting thymic development as poised ‘effector-memory-like’ cells already capable of mounting potent cytokine responses within minutes of activation. This allows iNKT cells to rapidly transactivate other immune populations and orchestrate immune responses (Smyth et al., 2005; Reilly et al., 2012). Activation also induces iNKT cell proliferation, generating an expanded pool of effector cells within 72 hr, most of which subsequently undergo apoptosis as the expanded iNKT cell pool contracts within 7 days (Cameron and Godfrey, 2018; Wilson et al., 2003; Parekh et al., 2005). However, some iNKT cells persist after the immune response subsides (Wilson et al., 2003; Parekh et al., 2005; Shimizu et al., 2014), and there is evidence that antigen challenge induces long-term changes in the iNKT cell repertoire analogous to memory T cell differentiation. For example, several studies have demonstrated that activation of iNKT cells with α-galalctosylceramide (αGalCer), a potent glycolipid antigen, induces the emergence of novel KLRG1+ and Follicular Helper iNKT (NKTFH) cell populations that are greatly enriched after 3–7 days, and still detectable >30 days after αGalCer challenge (Shimizu et al., 2014; Chang et al., 2011; Murray et al., 2021; Chen et al., 2018b). However, our knowledge of the transcriptional programs underpinning iNKT cell activation remains limited, and there are also relatively few transcriptional resources available for studying activated iNKT cells, especially compared to adaptive T cells (Andreatta et al., 2021).
Analysis and interpretation of iNKT cell biology is also challenging because iNKT cells exhibit heterogeneity, including NKT1, NKT2, and NKT17 subsets that broadly mirror CD4+ Th1, Th2, and Th17 cells (Engel et al., 2016). Past studies of NKT1, NKT2, and NKT17 subsets largely focused on iNKT cell thymic development or steady-state phenotype in the absence of activation (Engel et al., 2016; Harsha Krovi et al., 2020; Lee et al., 2015; Baranek et al., 2020), and less is known about iNKT cell subsets after activation. Using parabiosis models, we and others have also shown that iNKT cells are predominantly tissue resident (Lynch et al., 2015; Thomas et al., 2011), and that this can strongly influence their biology (LaMarche et al., 2020). For example, iNKT cells resident in adipose tissue exhibit an unusual regulatory phenotype characterized by increased KLRG1 expression, reduced expression of the transcription factor promyelocytic leukemia zinc-finger (PLZF), and increased production of IL-10 through an IRE1a-XBP1s-E4BP4 axis, enabling these cells to suppress inflammation and promote metabolic homeostasis (Lynch et al., 2015; LaMarche et al., 2020). Interestingly, Sag et al., 2014 demonstrated that IL-10+ iNKT (NKT10) cells emerge in other organs such as the spleen after repeated antigen challenge (Sag et al., 2014), indicating that TCR stimulation can induce a regulatory phenotype, and that NKT10 cells can potentially be considered a memory-like population. However, the relationship between NKT10 cells and other memory-like populations, such as KLRG1+ and NKTFH cells, remains unclear. Furthermore, it is unknown whether similar factors regulate NKT10 cells present in adipose tissue versus those induced after antigen challenge.
To characterize transcriptional remodeling in activated iNKT cells while also considering subset and tissue-associated heterogeneity, we performed single-cell RNA-sequencing (scRNA-seq) of 48,813 murine iNKT cells from spleen and adipose tissue at steady state and 4 hr, 72 hr, and 4 weeks after in vivo stimulation with αGalCer, as well as after repeated αGalCer challenge. We also reanalyzed published human and murine data to generate a transcriptomic atlas of iNKT cell activation states. We found that activation induces rapid and extensive transcriptional remodeling in iNKT cells, and that a common transcriptional framework underpins the activation of diverse iNKT cell populations. However, regulatory iNKT cell populations demonstrate largely blunted activation in response to αGalCer and display enrichment of memory-like KLRG1+ and cMAF+ iNKT cell subsets expressing a T regulatory type 1 (Tr1) cell gene signature. We also show that cMAF+ iNKT cells are enriched for NKT10 cells and express a gene signature similar to NKTFH cells. Overall, this study provides novel insights into longitudinal transcriptional remodeling in activated iNKT cells and the phenotype of regulatory iNKT cells, while also generating a novel transcriptomic resource for interrogation of iNKT cell biology.
ResultsiNKT cells undergo rapid and extensive transcriptional remodeling in response to αGalCer
To investigate transcriptional remodeling in activated iNKT cells, we performed 10× scRNA-seq of whole murine adipose and splenic iNKT cells 4 hr, 72 hr, and 4 weeks after in vivo stimulation with αGalCer and reanalyzed our published scRNA-seq of steady-state murine adipose and splenic iNKT cells (LaMarche et al., 2020; GSE142845, Figure 1A). We first analyzed our steady-state, 4 hr and 72 hr splenic iNKT cell data. After quality control measures, we obtained 16,701 splenic iNKT cells, including >4000 cells per activation state. After performing uniform manifold approximation and projection (UMAP), we observed minimal overlap between iNKT cells from different activation states (Figure 1B), indicating that iNKT cells undergo rapid and extensive transcriptional remodeling during early activation. Using gene expression analysis (Supplementary file 1), we found that steady-state iNKT cells displayed enrichment of NKT1 and NKT17 cell markers such as Il2rb, Klrb1c, Rorc, and Il7r (Figure 1C; Engel et al., 2016), but following activation iNKT cells rapidly downregulated these genes within 4 hr and upregulated expression of T cell activation markers and cytokines, including Il2ra, Irf4, Nr4a1, Pdcd1, Ifng, Il4, and Il17a (Figure 1C). This was accompanied by increased expression of Zbtb16 (PLZF) and the PLZF regulon genes Icos and Cd40lg (Figure 1C), consistent with published data demonstrating that PLZF is required for the innate response of iNKT cells to antigen and is upregulated after activation (Kovalovsky et al., 2008; Oleinika et al., 2018). Activated iNKT cells also downregulated expression of the transcription factor Id2 (Figure 1C), which plays an essential role in normal iNKT cell activation (Stradner et al., 2016), and upregulated expression of genes regulating T cell metabolic activation, including Myc, Hif1a, and Tfrc (Marchingo et al., 2020; Finlay et al., 2012; Wang et al., 2018c), suggesting that activated iNKT cells undergo metabolic remodeling.
Invariant natural killer T (iNKT) cells undergo rapid and extensive transcriptional remodeling in response to α-galalctosylceramide (αGalCer).
(A) Cartoon illustrating the experimental design for the generation of all scRNA-seq data. (B) Uniform manifold approximation and projection (UMAP) of murine splenic iNKT cells with cell cycle regression. (C) Heatmap of scaled averaged gene expression with hierarchical clustering in the data from (B). (D) Histograms showing expression of metabolic gene module scores in the data from (B). (E) UMAP of human peripheral blood mononuclear cell (PBMC) iNKT cells reanalyzed from GSE128243. (F) Histograms showing expression of functional and metabolic gene module scores in the data from (E).
Single cell gene expression of splenic iNKT cells at steady state, and 4 and 72 hours post activation.
(A) Uniform manifold approximation and projection (UMAP) of murine splenic invariant natural killer T (iNKT) cells with cell cycle regression applied. (B) Density plots of cell cycle and proliferation marker gene expression in the data from (A).
Expression plots of proliferation gene scores and select genes in splenic iNKT cells at 72 hours post activation.
(A) Expression plots of S phase and G2/M phase proliferation gene scores in murine splenic invariant natural killer T (iNKT) cells at 72 hr post-α-galalctosylceramide (post-αGalCer). (B) Density plots of gene expression in murine splenic iNKT cells 72 hr post-αGalCer.
Enrichment of cMAF in activated iNKT cells.
(A) Dot plot of Klrg1 and Maf expression in murine splenic invariant natural killer T (iNKT) cells at steady state, 4 hr, and 72 hr post-α-galalctosylceramide (post-αGalCer). (B) Representative histogram of cMAF expression in murine splenic iNKT cells at steady state or 72 hr post-αGalCer (⍺GC). iNKT cells were defined as live, single CD19- CD8- CD3low CD1d-PBS57 tetramer+ cells. Histograms were normalized to the mode. (C) Box plot quantifying cMAF median fluorescence intensity (MFI) in the data from (B). N = 3 biological replicates from one experiment. Experiment performed twice. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. The central box plot horizontal line denotes the median value. (D) Box plot quantifying the percentage data from (B), as performed for MFI data in (C). (E) Representative pseudocolor plots showing expression of IFNγ versus IL-10 in murine splenic iNKT cells at steady state or 72 hr post-αGalCer after restimulation for 4 hr with 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin (ex vivo). iNKT cells were defined as live, single CD19- CD8- CD3low CD1d-PBS57 tetramer+ cells. (F) Box plot quantifying the percentage data in (E). N = 3 biological replicates from one experiment. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. The central box plot horizontal line denotes the median value.
By 72 hr, however, expression of activation and cytokine genes was greatly reduced, and we identified enrichment of genes associated with proliferation, stem-like T cells, and NKT2 or stage 2 iNKT cells, including Mki67, Slamf6, Tcf7, and Ccr7 (Figure 1C, Figure 1—figure supplements 1 and 2; Utzschneider et al., 2016; Cohen et al., 2013). We observed that some 72 hr cells displayed enrichment of TFH and NKTFH markers, including Cxcr5, Il21, and Maf (Chang et al., 2011; Chen et al., 2018b; Andreatta et al., 2021; Andris et al., 2017), and memory-like iNKT cell markers, such as Itga4 and Klrg1 (Shimizu et al., 2014; Figure 1C, Figure 1—figure supplements 2 and 3), corresponding with previous studies documenting the appearance of NKTFH and KLRG1+ iNKT cells after αGalCer challenge (Shimizu et al., 2014; Chang et al., 2011; Murray et al., 2021; Rampuria and Lang, 2015). We also found increased expression of genes associated with the KLF2 regulon, including Klf2 and S1pr1 (Figure 1C). KLF2 is known to induce T cell thymic egress and trafficking through secondary lymphoid organs (Carlson et al., 2006), and has been found to play an important role in iNKT cell migration and thymic emigration (Harsha Krovi et al., 2020; Baranek et al., 2020; Wang et al., 2022; Wang and Hogquist, 2018a). While iNKT cells are generally tissue resident under steady-state conditions (Lynch et al., 2015; Thomas et al., 2011), increased expression of Klf2 and S1pr1 at 72 hr post-αGalCer could suggest that activated iNKT cells may traffic to other sites. However, previous work has also shown that hepatic iNKT cells arrest after becoming activated (Geissmann et al., 2005; Liew et al., 2017). Therefore, further analysis of migration in iNKT cells from different organs and at distinct stages of activation will be required to elucidate if and when activated iNKT cells undergo migration.
Having identified enrichment of Myc, Hif1a and Tfrc 4 hr post-αGalCer, we wondered what type of metabolic remodeling activated iNKT cells undergo in vivo. To map metabolic gene changes during iNKT cell activation, we generated gene module scores using the KEGG pathway (Kanehisa et al., 2021) and Gene Ontology Consortium (Carbon, 2021) databases (Supplementary file 2) and scored our data. We found 4 hr activated cells upregulated glycolysis, amino acid metabolism, polyamine synthesis, and fatty acid synthesis signatures, whereas oxidative signatures were downregulated compared to steady-state iNKT cells (Figure 1D). This suggests that, despite being poised at steady state for cytokine production, activated iNKT cells, like adaptive T cells, may switch on aerobic glycolysis and upregulate biosynthethic pathways to fuel cytokine production, growth, and proliferation (Marchingo et al., 2020; Angiari et al., 2020; Wu et al., 2020). Our data is also consistent with recent work identifying glucose as an important fuel for iNKT cell effector function (Fu et al., 2019; Kumar et al., 2019). Interestingly, we found that 72 hr activated cells engage oxidative signatures while maintaining elevated expression of glycolytic genes (Figure 1D), suggesting that the metabolic requirements of iNKT cells change across different activation states. We also observed reduced expression of polyamine synthesis and amino acid metabolism signatures 72 hr post-αGalCer, indicating that those pathways may be coupled to early iNKT cell activation and cytokine production, while oxidative metabolism is associated with proliferation when the iNKT cell pool expands four- to tenfold in vivo by 72 hr ; (Wilson et al., 2003; Parekh et al., 2005).
Having profiled transcriptional remodeling in activated murine iNKT cells, we wondered whether similar remodeling occurs in human iNKT cells. To investigate human iNKT cell activation, we reanalyzed published scRNA-seq data of human iNKT cells isolated from peripheral blood mononuclear cells (PBMCs) and stimulated ex vivo with phorbol 12-myristate 13-acetate (PMA) and ionomycin (GSE128243; Zhou et al., 2020). Following quality control measures, we obtained 13,957 cells and found that human iNKT cells also undergo rapid and extensive transcriptional remodeling after activation (Figure 1E and F). Furthermore, activated human iNKT cells recapitulated the metabolic gene reprogramming observed in activated murine iNKT cells, displaying upregulated glycolytic, amino acid metabolism and polyamine synthesis signatures, and reduced expression of oxidative signatures (Figure 1F). Thus, transcriptional signatures of iNKT cell activation are conserved across species.
Oxidative phosphorylation differentiates functional responses to αGalCer in NKT2 and NKT17 cells versus NKT1 cells
We next asked whether iNKT cell subsets expressed different transcriptional signatures after activation. We performed subclustering of murine splenic iNKT cells at steady state and 4 hr post-αGalCer, and identified clusters corresponding to NKT1, NKT2, and NKT17 cells (Figure 2A, Figure 2—figure supplements 1 and 2; see ‘Methods’) using the published transcription factors Tbx21, Zbtb16, and Rorc, and the flagship cytokines Ifng, Il4, Il13, Il17a, and Il17f (Cameron and Godfrey, 2018; Engel et al., 2016; Venken et al., 2019; Figure 2B, Figure 2—figure supplements 1 and 2). Notably, we found that all subsets upregulated Zbtb16 after activation (Figure 2B), suggesting that PLZF may play a subset-independent role during activation. We found that Tbx21 expression was nonspecifically increased across all iNKT cell subsets after activation (Figure 2B), and therefore, we did not use Tbx21 to demarcate activated iNKT cells. When we performed gene expression analysis, we found that all subsets demonstrated upregulation of activation, cytokine, glycolysis, amino acid metabolism, polyamine synthesis, and fatty acid synthesis signatures after αGalCer (Figure 2C, Figure 2—figure supplement 3), indicating that a common transcriptional framework underpins the activation of functionally diverse iNKT cell subsets. We also identified genes specifically enriched in one or more subsets, such as Gzmb and Ccl4 in NKT1 cells (Figure 2C; Supplementary file 3). Strikingly, we found that NKT2 and NKT17 cells, but not NKT1 cells, shared expression of many genes, including Lif, Lta, Cd274, and Ncoa7 (Figure 2C). Activated NKT2 and NKT17 cells also demonstrated increased whole-transcriptome correlation compared to activated NKT1 cells (Figure 2D), indicating that activated NKT2 and NKT17 cells are transcriptionally similar compared to NKT1 cells.
Oxidative phosphorylation differentiates functional responses to α-galalctosylceramide (αGalCer) in NKT2 and NKT17 cells versus NKT1 cells.
(A) Uniform manifold approximation and projection (UMAP) of murine splenic invariant natural killer T (iNKT) cells. (B) Box plots showing gene expression in the data from (A). ⍺GC denotes αGalCer. The central box plot horizontal line denotes the median value. (C) Heatmap of scaled averaged gene expression with hierarchical clustering in NKT1 (1), NKT2 (2), and NKT17 cells (17) from the data in (A). (D) Correlation plot of total normalized RNA counts for all genes across steady state and activated NKT1, NKT2, and NKT17 cell subsets. (E) Gene set enrichment analysis (GSEA) plot of KEGG oxidative phosphorylation comparing activated NKT2 cells and NKT17 cells versus activated NKT1 cells. NES, normalized enrichment score; Padj, adjusted p-value for the enrichment. (F) Histograms showing staining of MitoTracker Green FM and TMRM in CD44+ NKT1, NKT2, and NKT17 cells from BALB/c mouse thymus. iNKT cells were defined as live, single CD19- CD8- CD3low CD1d-PBS57 tetramer+ cells. Histograms were normalized to the mode. (G) Bar plots quantifying the mean fluorescence intensity (MFI) data from (F). N = 4 biological replicates from one experiment. Experiment performed at least twice. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. Source data provided in Figure 2—source data 1. (H) Line plots showing production of flagship cytokines by CD44+ BALB/c mouse thymic iNKT cells: NKT1 (IFNγ), NKT2 (IL-4 and IL-13), and NKT17 cells (IL-17A) without stimulation and after 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin in the absence or presence of 40 nM oligomycin (all 4 hr ex vivo). N = 5 biological replicates from one experiment. Experiment performed at least twice. One-way ANOVA and Tukey’s post hoc test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. Source data provided in Figure 2—source data 2.
MFI expression levels of mitotracker and TMRM in iNKT cell subsets by flow cytometry.
Amount of cytokines produced in each iNKT cell subset with or without metabolic inhibitors.
Distinct gene expression differences in iNKT cell subsets.
(A) Principal component analysis (PCA) plot of murine splenic invariant natural killer T (iNKT) cells at steady state. (B) Violin plots of iNKT cell subset marker gene expression in the data from (A). (C) Violin plots of NKT1-A versus NKT1-B marker gene expression in the data from (A). (D) PCA plots and uniform manifold approximation and projection (UMAP) plot of murine splenic iNKT cells at 4 hr post-α-galalctosylceramide (post-αGalCer). (E) Feature plots of cytokine gene expression in the data from (D).
UMAP plots of NKT subsets, resting and post-activation.
(A) Uniform manifold approximation and projection (UMAP) plots of murine splenic NKT1, NKT2, and NKT17 cells at steady state or 4 hr post-α-galalctosylceramide (post-αGalCer). (B) UMAP plots of reclustered murine splenic NKT1, NKT2, and NKT17 cells into ‘true’ steady state or activated (4 hr post-αGalCer) clusters. (C) Feature plots of expression of invariant natural killer T (iNKT cell activation gene module scoring in the data from A, B). (D) Feature plots of expression of the flagship iNKT cell subset cytokines Ifng, Il4, and Il17a in the data from (A, B).
Module score analysis of scRNA-Seq data reveals changes in metabolic pathway gene expression following iNKT cell activation.
(A) Box plots of metabolic pathway gene module score expression in murine steady state and activated NKT1, NKT2 and NKT17 cells.
Flow cytometry validation of iNKT cell subset identity in the BALB/c CD43-HG/ICOS model using expression of flagship NKT2, NKT17 and NKT1 cell transcription factors.
(A) Representative histograms of flagship invariant natural killer T (iNKT) cell subset transcription factor staining in CD44+ NKT1, NKT2, and NKT17 cells from BALB/c mouse thymus. iNKT cells were defined as live, single CD45+ CD19- CD8- CD3low CD1d-PBS57 tetramer+ cells. NKT1 cells were defined as CD43-HG- ICOS- iNKT cells, NKT2 cells were defined as CD43-HG- ICOS+ iNKT cells, and NKT17 cells were defined as CD43-HG+ iNKT cells. Histograms were normalized to the mode. Experiment performed once with N = 4 biological replicates.
Correlation of mitochondrial marker expression with NKT subsets.
(A) Representative contour plots of MitoTracker Green FM and TMRM staining versus NK1.1 expression in murine steady-state splenic invariant natural killer T (iNKT) cells. iNKT cells were defined as live, single CD45+ CD19- CD3low CD1d-PBS57 tetramer+ cells. (B) Quantification of the data in the left panel (MitoTracker Green FM) of (A). Percentages were calculated by taking the absolute number of cells in the NK1.1+ or NK1.1- MitoTracker GreenHigh gate as percentage of the total number of NK1.1+ or NK1.1- iNKT cells, respectively. N = 5 biological replicates from one experiment. Experiment performed at least three times. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. The central box plot horizontal line denotes the median value. (C) Quantification of the data in the right panel (TMRM) of (A), as performed for MitoTracker Green FM data in (B). (D) Representative pseudocolor plots showing expression of IFNγ versus IL-4 in murine splenic iNKT cells at steady state after stimulation for 4 hr ex vivo with 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin in the absence (left panel) or presence (right panel) of 20 nM oligomycin. iNKT cells were defined as live, single CD45+ CD8- CD19- F4/80- CD3low CD1d-PBS57 tetramer+ cells. (E) Box plot quantifying the percentage data in (D). N = 5 biological replicates from one experiment. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. NS denotes not significant (p>0.05). The central box plot horizontal line denotes the median value.
To investigate the shared transcriptional signatures of NKT2 and NKT17 cells, we performed gene set enrichment analysis (GSEA) (Korotkevich et al., 2016; Subramanian et al., 2005) comparing activated NKT2 and NKT17 cells versus activated NKT1 cells using the KEGG pathway database (Kanehisa et al., 2021). We identified enrichment of oxidative phosphorylation (Figure 2E), suggesting that NKT2 and NKT17 cells use oxidative metabolism more than NKT1 cells. To validate this result, we first measured mitochondrial mass and membrane potential in thymic CD44+ NKT1, NKT2, and NKT17 cells from BALB/c mice (Figure 2—figure supplement 4; Cameron and Godfrey, 2018), and we found that NKT2 and NKT17 cells had significantly increased mitochondrial mass and membrane potential compared to NKT1 cells (Figure 2F and G). We also found that the expression of NK1.1, a surface marker that is known to segregate NKT1 versus NKT2 and NKT17 cells (Engel et al., 2016; Lee et al., 2013), was able to distinguish distinct mitochondrial phenotypes among splenic iNKT cells from C57BL/6 mice (Figure 2—figure supplement 5). We next investigated whether NKT2 and NKT17 cells were more functionally dependent on oxidative metabolism than NKT1 cells by stimulating BALB/c thymic iNKT cells ex vivo with PMA and ionomycin for 4 hr in the presence or absence of oligomycin, to inhibit oxidative phosphorylation (Lopes et al., 2021). Treatment with oligomycin globally reduced cytokine production across all iNKT cell subsets; however, we found that production of IL-4, IL-13, and IL-17A was almost completely ablated compared to production of IFNγ (Figure 2H). We also found that treatment with oligomycin resulted in significantly reduced production of IL-4 but not IFNγ by splenic iNKT cells from C57BL/6 mice (Figure 2—figure supplement 5). Collectively, our data demonstrate that oxidative metabolism is essential for the production of NKT2 and NKT17 cytokines but less so for NKT1 cytokines.
Adipose iNKT cells display blunted and delayed activation after αGalCer, enrichment of Tr1 cell markers, and hallmarks of chronic endogenous activation
We and others have shown that adipose iNKT cells display an unusual regulatory phenotype characterized by E4BP4 (Nfil3) expression and enrichment of NKT10 cells (Lynch et al., 2015; LaMarche et al., 2020; Sag et al., 2014). Having characterized activated splenic iNKT cells, we next investigated activated adipose iNKT cells. Combined analysis of adipose and splenic iNKT cells at steady state, 4 hr post-αGalCer, and 72 hr post-αGalCer returned 28,561 cells, including 11,860 adipose iNKT cells, and >2900 cells per activation state. When we performed UMAP, we found that adipose and splenic iNKT cells displayed minimal overlap (Figure 3A), indicative of constitutive transcriptional differences between adipose and splenic iNKT cells. Cross-dataset differential gene expression analysis (‘Methods’) identified 971 genes enriched among all adipose iNKT cells regardless of activation status or subset, versus only 65 genes enriched among all splenic iNKT cells (Supplementary file 4), indicating that adipose but not splenic iNKT cells retain expression of many conserved genes during activation. When we profiled these conserved genes using over-representation analysis against the KEGG pathway database, we identified Ribosome as the sole enriched pathway among splenic iNKT cells, whereas adipose iNKT cells displayed enrichment of pathways related to uptake from the extracellular environment (endocytosis, regulation of actin cytoskeleton, FcγR-mediated phagocytosis), adhesion (focal adhesion, leukocyte transendothelial migration), cytotoxicity and cell death (NK cell-mediated cytotoxicity, apoptosis), cellular senescence, chemokine signaling, and TCR signaling (Figure 3B).
Adipose invariant natural killer T (iNKT) cells display blunted and delayed activation after α-galalctosylceramide (αGalCer), enrichment of Tr1 Cell markers, and hallmarks of chronic endogenous activation.
(A) Uniform manifold approximation and projection (UMAP) of murine adipose and splenic iNKT cells with cell cycle regression. (B) Lollipop plot showing enrichment of nondisease KEGG pathways based on conserved enrichment 971 genes in murine adipose iNKT cells and 65 genes in murine splenic iNKT cells. Source data provided in Figure 3—source data 1. (C) Heatmap of scaled averaged gene expression with hierarchical clustering in the data from (A). (D) Line plots showing median expression of functional and metabolic gene module scores in the data from (A). (E) Box plots showing gene expression in murine steady-state splenic or adipose iNKT cells. The central box plot horizontal line denotes the median value.
Comparison of splenic and adipose iNKT cells at 72hr post-activation.
Comparison of splenic and adipose iNKT cells at 72hr post-activation.
(A) Uniform manifold approximation and projection (UMAP) plot of murine splenic and adipose invariant natural killer T (iNKT) cells at 72 hr post-α-galalctosylceramide (post-αGalCer). (B) Feature plots of Nr4a1 and cytokine gene expression in the data from (A).
We next compared gene expression among individual iNKT cell activation states. Intriguingly, we found that there was greatly reduced expression of activation markers such as Il2ra and Nr4a1 (Nur77) in adipose versus splenic iNKT cells at 4 hr post-αGalCer (Figure 3C). We previously found that Nur77 is enriched in adipose iNKT cells compared to splenic iNKT cells at steady state (Lynch et al., 2015; LaMarche et al., 2020), but these data suggest that adipose iNKT cells demonstrate reduced upregulation of Nur77 compared to splenic iNKT cells upon activation. Using module scoring, we found that adipose iNKT cells showed a blunted initial response to αGalCer characterized by reduced upregulation of activation and cytokine signatures, and reduced metabolic remodeling (Figure 3D). At 72 hr post-αGalCer, when proliferative gene signatures are upregulated, adipose and splenic iNKT cells showed similar enrichment of proliferation markers (Figure 3D), but adipose iNKT cells had reduced expression of stem-like markers such as Tcf7 and Slamf6, and reduced expression of TFH markers such as Cxcr5 and Tox2 (Figure 3C). Adipose iNKT cells showed increased expression of Il10, Lag3, Ctla4, Il21, Ccr5, Hif1a, Maf, and Gzmb (Figure 3C), which are markers typically associated with Tr1 cells, a heterogeneous population of regulatory T cells that do not express FOXP3 (Yan et al., 2017; Gruarin et al., 2019; Chihara et al., 2016; Alfen et al., 2018; Mascanfroni et al., 2015). We previously found that adipose iNKT cells do not express FOXP3 and instead express the transcription factor E4BP4, which regulates IL-10 production (Lynch et al., 2015; LaMarche et al., 2020). Other genes enriched among adipose iNKT cells included Il21r, the adenosine receptor Adora2a, and the exhaustion marker Tox (Figure 3C). In summary, activation with αGalCer induces differential transcriptional remodeling in adipose versus splenic iNKT cells, and the peak of the regulatory response in adipose iNKT cells is delayed compared to the rapid cytokine burst in splenic iNKT cells.
Notably, although adipose iNKT cells displayed blunted activation and metabolic remodeling compared to splenic iNKT cells, adipose iNKT cells were already enriched for activation, glycolysis, and amino acid metabolism gene signatures at steady state (Figure 3D), suggesting that adipose cells are activated at baseline. At steady state, adipose iNKT cells had increased expression of key activation markers such as Il2, Cd69, Myc, and Nr4a genes, as well as increased Nfat gene expression (Figure 3E). Nr4a and Nfat genes are commonly upregulated in chronically activated T cells, often in combination with Tox (Liu et al., 2019; Seo et al., 2019; Kumar et al., 2020), suggesting that adipose iNKT cells experience chronic activation. Furthermore, we found that steady-state adipose iNKT cells demonstrated increased expression of markers of antigen experience, Klrg1 and Itga4 (Figure 3E; Shimizu et al., 2014), and we have previously shown that KLRG1+ iNKT cells are greatly enriched in adipose tissue (Lynch et al., 2015). We also found that adipose iNKT cells, but not splenic iNKT cells, continued to express cytokine transcripts at 72 hr post-αGalCer (Figure 3—figure supplement 1), indicative of a ‘smoldering’ activation phenotype among adipose iNKT cells. Overall, our data suggest that adipose iNKT cells experience chronic endogenous activation, which may dampen their ability to undergo further rapid activation in response to αGalCer, and could explain the nature of their regulatory phenotype.
scRNA-seq identifies transcriptional signatures of adipose iNKT cell subset activation and adipose NKT10 cells
We have recently shown that adipose iNKT cells are more heterogeneous than originally anticipated. At steady state, these cells are comprised of NK1.1+ NKT1 cells, NK1.1- NKT1 cells, and NKT17 cells (LaMarche et al., 2020). NK1.1+ and NK1.1- adipose iNKT cells are functionally distinct as NK1.1+ cells produce IFNγ, and NK1.1- iNKT cells produce IL-4 and IL-10. However, it is unknown whether these populations engage distinct molecular programs in response to αGalCer, a proposed therapy for type 2 diabetes and obesity. We also wondered whether the reduced responsiveness to αGalCer in adipose iNKT cells was linked to any particular adipose iNKT cell population. To answer these questions, we performed analysis of adipose iNKT cells at steady state and 4 hr post-αGalCer. Unlike splenic iNKT cells, where activation accounted for most of the variance (Figure 2A), transcriptional differences between adipose NKT1 cells and NKT17 cells accounted for most of the variance among adipose iNKT cells (Figure 4A). We found that adipose NKT1 and NKT17 cells both upregulated expression of activation, cytokine, and metabolic gene signatures after αGalCer (Figure 4B). Comparison of adipose and splenic iNKT cell subsets showed that all adipose iNKT cell subsets had reduced activation and cytokine gene signature expression, and reduced metabolic remodeling at 4 hr post-αGalCer (Figure 4C). This indicates that all adipose iNKT cell subsets respond to αGalCer but no single subset (e.g., NKT1 or NKT17) was uniquely hyporesponsive vis-a-vis the spleen. Interestingly, we identified increased oxidative gene expression in adipose NKT17 cells versus adipose NKT1 cells (Figure 4C), similar to our finding in splenic iNKT cells (Figure 2). We have previously shown that γδ17 cells also display enrichment of oxidative metabolism (Lopes et al., 2021), suggesting that this is a conserved feature of innate T cells that produce IL-17.
scRNA-seq identifies transcriptional signatures of adipose invariant natural killer T (iNKT) cell subset activation.
(A) Uniform manifold approximation and projection (UMAP) of murine adipose iNKT cells. (B) Heatmap of scaled averaged gene expression with hierarchical clustering in the data from (A). (C) Line plots showing median expression of functional and metabolic gene module scores in murine adipose and splenic NKT1, NKT2, and NKT17 cell subsets. (D) Density plots of gene expression in murine adipose cytokinepos NKT1 cells at 4 hr post-αGalCer (cluster 6, A). (E) Radar chart showing Log2(Fold Change) values of genes enriched in Il10pos adipose iNKT cells versus Il10neg adipose iNKT cells at 4 hr and 72 hr post-α-galalctosylceramide (post-αGalCer).
Adipose iNKT cell heterogeneity at 72hr post aGalCer activation.
(A) Density plots of cytokine gene expression in murine adipose cytokinepos invariant natural killer T (iNKT) cells 72 hr post-α-galalctosylceramide (post-αGalCer) with cell cycle regression applied. (B) Density plots of cytokine and secreted factor gene expression in murine adipose Il10pos iNKT cells at 4 hr and 72 hr post-αGalCer with cell cycle regression applied.
Analysis of cytokine gene expression among adipose iNKT cells revealed that Il10 was only expressed by NKT1 cells (Figure 4B). Since cytokinepos adipose NKT1 cells (cluster 6) lacked or had downregulated expression of Klrb1c (NK1.1) by 4 hr post-αGalCer (Figure 4B), we could not stratify cytokine production in adipose NKT1 cells using Klrb1c. Therefore, we performed unbiased fine clustering of cytokinepos adipose NKT1 cells. We identified one population of cells co-expressing Ifng, Il4, and Il2, and a second population enriched for Il10 (Figure 4D), suggesting that adipose NKT10 cells are a distinct population. Analysis of cytokinepos adipose iNKT cells at 72 hr post-αGalCer, when expression of Il10 was highest among adipose iNKT cells (Figure 3), identified some NKT10 cells co-expressing Il10 with Ifng, Il4, and Il21, and other NKT10 cells that co-expressed Il10, Ifng, and Gzmb, suggesting that expanded NKT10 cells are heterogeneous (Figure 4—figure supplement 1). To investigate the transcriptional profile of adipose NKT10 cells, we performed global analysis of Il10pos versus Il10neg adipose iNKT cells at 4 hr and 72 hr post-αGalCer. We identified 207 genes enriched in adipose NKT10 cells (Supplementary file 5), including the Tr1 cell markers Lag3, Ctla4, Pdcd1, Gzmb, Il21, Hif1a, Maf, and Ccr5 (Figure 4E; Chihara et al., 2016; Mascanfroni et al., 2015; Grossman et al., 2004; Apetoh et al., 2010), suggesting that adipose NKT10 cells may be functionally similar to Tr1 cells. In summary, our analysis suggests that adipose iNKT cells primarily segregate by function after αGalCer, and that regulatory adipose NKT10 cells are a transcriptionally distinct population similar to Tr1 cells.
Chronic activation of splenic iNKT cells induces an adipose-like phenotype and the emergence of Tr1 iNKT cells
Since adipose iNKT cells displayed blunted and delayed activation after αGalCer, and enrichment of Tr1 cell markers, we wondered whether these were conserved features of regulatory iNKT cell biology. To answer this question, we repeatedly activated splenic iNKT cells, which induces IL-10 production (Sag et al., 2014). We sequenced 5433 αGalCer activated splenic iNKT cells, including 2117 cells isolated 4 weeks after mice received one dose of αGalCer (resting) and 3316 cells isolated 4 hr after reactivation with a second dose of αGalCer (reactivated; Figure 5A). Comparison of iNKT cells at steady state (no αGalCer) and resting (4 weeks post-αGalCer) revealed that resting iNKT cells displayed a reduced response to restimulation with αGalCer, similar to the adipose iNKT cell response to one dose of αGalCer (Figure 5B). This indicates that blunted activation after αGalCer is a conserved feature of regulatory iNKT cells. Furthermore, gene expression analysis revealed that resting iNKT cells were transcriptionally similar to adipose iNKT cells, displaying reduced Ifng, Il4, and Il2 expression, and increased expression of Tr1 cells markers such as Gzmb, Il10, Maf, Il21, Ctla4, Hif1a, Ccr5, and Lag3, especially after αGalCer rechallenge (Figure 5C), suggesting that regulatory iNKT cells are similar to Tr1 cells. We also reanalyzed previously published microarray data of control and αGalCer-pretreated splenic iNKT cells (GSE47959; Sag et al., 2014; Figure 5—figure supplement 1) and identified a similar activation phenotype in αGalCer-pretreated splenic iNKT cells versus control iNKT cells after short-term αGalCer stimulation. Interestingly, although resting iNKT cells expressed Tox (Figure 5C), we did not identify enrichment of other chronic activation markers, such as Nr4a1 (Nur77) (Figure 5—figure supplement 2, Supplementary file 6), indicating that prior exposure of splenic iNKT cells to antigen does not completely reproduce the phenotype of adipose iNKT cells, which may be exposed to chronic endogenous activation in situ.
Chronic activation of splenic invariant natural killer T (iNKT) cells induces an adipose-like phenotype and the appearance of populations expressing Tr1 cell markers.
(A) Uniform manifold approximation and projection (UMAP) of murine splenic iNKT cells. (B) Line plots showing median expression of functional and metabolic gene module scores in the data from (A). (C) Heatmap of scaled averaged gene expression with hierarchical clustering in the data from (A). (D) UMAP of subclustered murine splenic reactivated iNKT cells from (A). (E) Density plots of functional and metabolic gene module score and gene expression in the reactivated iNKT cells from the data in (D). (F) Representative pseudocolor plots of Granzyme A versus KLRG1 expression in steady state and resting splenic iNKT cells at 4 weeks post-α-galalctosylceramide (post-αGalCer) and after restimulation with 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin for 4 hr ex vivo (left), and box plot quantification of the pseudocolor plot data (right). iNKT cells were defined as live, single CD45+ CD8- CD19- F4/80- CD3low CD1d-PBS57 tetramer+ cells. N = 5 biological replicates from one experiment. Experiment performed at least twice. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. Source data provided in Figure 5—source data 1. The central box plot horizontal line denotes the median value. (G) Representative pseudocolor plots of cMAF versus IL-10 expression in steady state and resting splenic iNKT cells at 4 weeks post-αGalCer and after restimulation with 50 ng PMA and 1 µg Ionomycin for 4 hr ex vivo (left), and box plot quantification of the resting iNKT cell data from the pseudocolor plot data (right). iNKT cells were defined as live, single CD45+ CD8- CD19- F4/80- CD3low CD1d-PBS57 tetramer+ cells. N = 5 biological replicates from one experiment. Experiment performed at least twice. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. Source data provided in Figure 5—source data 2. The central box plot horizontal line denotes the median value.
% IL-10 production by flow cytometry in cmaf postive or negative iNKT cells.
% KLRG1 Granzyme A expression by flow cytometry by iNKT without or without treatment.
Reanalysis from GSE47959 showing gene expression differences between splenic iNKT with and outout a previous aGalCer activation.
(A) Robust Multichip Average (RMA)-normalized microarray gene expression in control and α-galalctosylceramide (αGalCer)-pretreated murine splenic invariant natural killer T (iNKT) cells after activation with 1 μg αGalCer for 90 min in vivo prior to isolation. Reanalyzed from GSE47959. Values below 0.25 units of expression were set to 0.25 for plotting. (B) RMA-normalized microarray gene expression in control and αGalCer-pretreated murine splenic iNKT cells at rest. Reanalyzed from GSE47959. Values below 0.25 units of expression were set to 0.25 for plotting.
Activation markers in iNKT cells at different activation states.
(A) Uniform manifold approximation and projection (UMAP) of murine splenic invariant natural killer T (iNKT) cells at steady state (Steady State), activated iNKT cells at 4 hr post-α-galalctosylceramide (post-αGalCer) (1 μg; Activated), resting iNKT cells at 4 weeks post-αGalCer (4 μg; Resting), and reactivated iNKT cells at 4 weeks post-αGalCer (4 μg) and 4 hr post-reactivation with αGalCer (1 μg; Reactivated). (B) Violin plots showing gene expression of iNKT cell activation markers in the data from (A).
Functional and metabolic gene module scores in different iNKT cell clusters.
(A) Uniform manifold approximation and projection (UMAP) plot of subclustered murine splenic reactivated invariant natural killer T (iNKT) cells. (B) Histograms showing expression of functional and metabolic gene module scores in the data from (A). (C) Density plots of cytokine and secreted factor gene expression in subclustered cluster B cells from the data in (A). (D) Density plots of surface marker, transcription factor, and secreted factor gene expression in subclustered cluster A cells from the data in (A).
Distinct expression patterns of MAF, IL10, Granzyme and KLRG1 associated with different iNKT cell states.
(A) Representative pseudocolor plots of cMAF expression versus CD1d-PBS57 tetramer staining in murine splenic invariant natural killer T (iNKT) cells at steady state (left panel) versus resting murine splenic iNKT cells at 4 weeks (WK) post-α-galalctosylceramide (post-αGalCer) (⍺GC; right panel). iNKT cells were defined as live, single CD45+ F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (B) Box plot quantifying the percentage data in (A). N = 5 biological replicates from one experiment. Experiment performed at least three times. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. The central box plot horizontal line denotes the median value. (C) Representative pseudocolor plots of Granzyme A versus IL-10 expression in murine splenic (left) or adipose (right) iNKT cells at 4 weeks post-αGalCer and after restimulation for 4 hr ex vivo with 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin. iNKT cells were defined as live, single CD45+ CD8- F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (D) Box plot quantifying the percentage data in(C). N = 5 biological replicates from one experiment. Experiment performed once. The central box plot horizontal line denotes the median value. (E) Representative pseudocolor plots of Granzyme A versus KLRG1 expression in murine adipose iNKT cells at 4 weeks post-αGalCer and after restimulation for 4 hr ex vivo with 50 ng PMA and 1 µg ionomycin. iNKT cells were defined as live, single CD45+ CD8- F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (F) Box plot quantifying the percentage data in (E). N = 5 biological replicates from one experiment. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. The central box plot horizontal line denotes the median value.
Since we had identified a distinct population of NKT10 cells in adipose tissue after αGalCer, we wondered whether we could also identify an NKT10 population among reactivated splenic iNKT cells. Graph-based clustering of reactivated splenic iNKT cells revealed two major populations, clusters A and B (Figure 5D), which were differentiated by their response to reactivation. Only cluster B cells (~50% of cells) demonstrated significant activation, cytokine transcript expression, and metabolic gene reprogramming after the second dose of αGalCer (Figure 5E, Figure 5—figure supplement 3). Functional stratification of reactivated splenic iNKT cells revealed a population of NKT1 cells co-expressing Ifng and Il4, a population of NKT17 cells expressing Il17a, a population of NKT10 cells co-expressing Il10 and Maf, as well as intermediate Ifng and Il4, and a population of cells co-expressing Gzma and the memory-like iNKT cell marker Klrg1 (Figure 5E, Figure 5—figure supplement 3). These data suggest that prior immunization with αGalCer is associated with the appearance of two new functional iNKT cell populations expressing Klrg1 or Maf. We confirmed these findings using flow cytometry, identifying enrichment of Granzyme A+ KLRG1+ iNKT cells in the spleen at 4 weeks post-αGalCer versus steady state (Figure 5F), increased expression of cMAF at 4 weeks post-αGalCer (Figure 5—figure supplement 4), and increased IL-10 production among restimulated cMAF+ iNKT cells compared to cMAF- iNKT cells at 4 weeks post-αGalCer (Figure 5G). We also found that expression of Granzyme A and IL-10 was mutually exclusive at the protein level (Figure 5—figure supplement 4), matching our scRNA-seq analysis. Therefore, these data suggest that repeated antigen exposure induces a regulatory phenotype in splenic iNKT cells, which is associated with the appearance of two new functionally distinct iNKT cell populations expressing cMAF and KLRG1.
Memory-like cMAF<sup>+</sup> and KLRG1<sup>+</sup> iNKT cells are induced in the spleen following αGalCer challenge, and similar populations are constitutively present in adipose tissue
Having identified enrichment of iNKT cells expressing Maf and Klrg1 among reactivated splenic iNKT cells, we wondered whether we could identify similar populations among resting iNKT cells. We also sought to characterize the transcriptional profile of these populations. Analysis of steady-state and resting iNKT cells revealed distinct NKT1, NKT2, NKT17, and cycling cell populations on the basis of graph-based clustering, spatial separation, and Tbx21, Zbtb16, Rorc, and Mki67 expression (Figure 6A and BFigure 2—figure supplement 1, Figure 6—figure supplement 1). We additionally found that NKT1 cells separated into two distinct clusters, NKT1-A and NKT1-B cells, with NKT1-B cells displaying increased expression of genes associated with the KLF2 regulon (Figure 6A and B, Figure 1—figure supplement 3, Figure 6—figure supplement 1). This demonstrated that iNKT cell subset diversity is preserved after antigen challenge. Among resting iNKT cells, however, two new clusters emerged, expressing Maf or Klrg1 (Figure 6A and B), matching our reactivated data. Independent of our clustering, we also detected mutually exclusive gene-level expression of Maf and Klrg1 among resting iNKT cells (Figure 6C), consistent with two distinct iNKT cell populations expressing Maf or Klrg1. We confirmed this finding using flow cytometry, identifying significant enrichment of mutually exclusive cMAF+ and KLRG1+ iNKT cell populations among resting versus steady-state splenic iNKT cells (Figure 6D and E). This demonstrates that cMAF+ and KLRG1+ iNKT cells are induced after antigen challenge. Interestingly, we found that the frequency of cMAF+ iNKT cells was positively correlated with antigen load (Figure 6—figure supplement 2), suggesting that TCR signal strength may regulate cMAF expression in iNKT cells, a signaling axis which has previously been reported in γδ T cells (Zuberbuehler et al., 2019).
Memory-like cMAF<sup>+</sup> and KLRG1<sup>+</sup> invariant natural killer T (iNKT) cells are induced in the spleen following α-galalctosylceramide (αGalCer) challenge, and similar populations are constitutively present in adipose tissue.
(A) Uniform manifold approximation and projection (UMAP) of murine splenic iNKT cells. (B) Heatmap of scaled averaged gene expression with hierarchical clustering in the data from (A). Both datasets were merged and normalized together, and cycling cells were excluded from the analysis. (C) Density plots showing mutually expression of Maf and Klrg1 in the data from (A) (bottom UMAP). (D) Representative pseudocolor plots of RORγT- (spleen, adipose tissue, lung, and inguinal lymph nodes) or total (liver) murine splenic iNKT cells at steady-state or resting iNKT cells at 4 weeks post-αGalCer. iNKT cells were defined as live, single CD19- CD8- F4/80- CD3low CD1d-PBS57 tetramer+ cells. iLN, inguinal lymph node. (E) Box plots quantifying the KLRG1 percentage data from (D). N = 4 biological replicates for hepatic iNKT cells at 4 weeks post-αGalCer and N = 5 biological replicates for all other data from one experiment. Experiment performed at least twice in the spleen and once for other tissues. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. Source data provided in Figure 6—source data 1. The central box plot horizontal line denotes the median value. (F) Box plots quantifying the cMAF percentage data from (D). N = 4 biological replicates for hepatic iNKT cells at 4 weeks post-αGalCer and N = 5 biological replicates for all other data from one experiment. Experiment performed at least twice in the spleen and once for other tissues. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. Source data provided in Figure 6—source data 1. The central box plot horizontal line denotes the median value.
Gene expression patterns in resting iNKT, 4 weeks post stimulation.
Gene expression of spenic iNKT cells resting 4 weeks post activation.
(A) Principal component analysis (PCA) plots and uniform manifold approximation and projection (UMAP) plot of murine resting splenic invariant natural killer T (iNKT) cells at 4 weeks post-α-galalctosylceramide (post-αGalCer). (B) Violin plots of marker gene expression in the clusters from (A). (C) Correlation plot of total normalized RNA counts for all genes for the clusters from (A) (cycling cells were excluded).
Strength of TCR signal is associated with strength of MAF expression.
(A) Representative pseudocolor plots of cMAF versus CD3 expression in murine splenic invariant natural killer T (iNKT) cells at steady state (panel 1) or 4 weeks (WK) post-activation with 1 μg α-galalctosylceramide (αGalCer) (panel 2), 2.5 μg αGalCer (panel 3), or 5 μg αGalCer (panel 4). iNKT cells were defined as live, single CD45+ CD8- CD11b- CD19- CD3low CD1d-PBS57 tetramer+ cells. (B) Box plot quantifying the percentage data in (A). N = 3 biological replicates from one experiment. Experiment performed once. One-way ANOVA and Tukey’s post hoc test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. The central box plot horizontal line denotes the median value. (C) Representative pseudocolor plots of cMAF versus KLRG1 expression in RORγT- murine splenic iNKT cells at 4 weeks post-activation with 4 μg αGalCer (left) or 10 μg αGalCer (right). iNKT cells were defined as live, single CD45+ CD8- F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (D) Box plot quantifying the percentage data in (C). N = 5 biological replicates. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, *p<0.05; **p<0.01; ***p<0.001. The central box plot horizontal line denotes the median value.
Comparison of splenic and adipose iNKT cells expression of functional markers.
(A) Representative pseudocolor plots of Granzyme A versus IFNγ expression in murine splenic (left) or adipose (right) invariant natural killer T (iNKT) cells at 4 weeks post-α-galalctosylceramide (post-αGalCer) and after restimulation for 4 hr ex vivo with 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin. iNKT cells were defined as live, single CD45+ CD8- F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (B) Box plot quantifying the percentage data in (A). N = 5 biological replicates from one experiment. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. The central box plot horizontal line denotes the median value. Padj denotes the adjusted p-value. (C) Representative pseudocolor plots of IL-10 versus IFNγ expression in murine splenic (left) or adipose (right) iNKT cells at 4 weeks (wk) post-αGalCer (⍺GC) and after restimulation for 4 hr ex vivo with 50 ng PMA and 1 µg ionomycin. iNKT cells were defined as live, single CD45+ CD8- F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (D) Box plot quantifying the percentage data in (C). N = 5 biological replicates from one experiment. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. The central box plot horizontal line denotes the median value.
IL17 and IL10 expression is mutually exclusive on iNKT cells.
(A) Representative pseudocolor plot of IL-10 versus IL-17A expression in murine splenic invariant natural killer T (iNKT) cells at 4 weeks post-α-galalctosylceramide (post-αGalCer) and after restimulation for 4 hr ex vivo with 50 ng phorbol 12-myristate 13-acetate (PMA) and 1 µg ionomycin. iNKT cells were defined as live, single CD45+ CD8- F4/80- CD19- CD3low CD1d-PBS57 tetramer+ cells. (B) Box plot quantifying the percentage data in (A). N = 5 biological replicates from one experiment. Experiment performed once. The central box plot horizontal line denotes the median value.
Resting adipose iNKT cells 4 weeks post aGalCer activation.
(A) Uniform manifold approximation and projection (UMAP) of murine resting adipose invariant natural killer T (iNKT) cells at 4 weeks post-α-galalctosylceramide (post-αGalCer). (B) Heatmap of scaled averaged gene expression with hierarchical clustering in the data in the data from (A). (C) UMAP of subclustered murine steady-state adipose NK1.1- NKT1 cells. (D) Heatmap of scaled averaged gene expression with hierarchical clustering in the data in the data from (C). (E) Box plots of gene expression in the data from (C). The central box plot horizontal line denotes the median value.
We next performed gene expression analysis to analyze the transcriptional profile of cMAF+ and KLRG1+ iNKT cells. cMAF+ cells displayed enrichment of memory and exhaustion markers associated with CD4+ memory T cells, stem-like memory T cells, Tregs, and T precursor exhausted cell (TPEX) populations, including Cd4, Slamf6, Tcf7, Ctla4, Lag3, Cxcr3, and Tox (Figure 6B; Andreatta et al., 2021; Utzschneider et al., 2016; Jadhav et al., 2019). We also detected enrichment of Zbtb16, Izumo1r (FR4), and the Tr1 cell markers Il27ra, Maf, and Hif1a (Figure 6B; Chihara et al., 2016; Mascanfroni et al., 2015; Apetoh et al., 2010). By contrast, KLRG1+ cells showed enrichment of cytotoxic T cell, effector memory CD8+ T cell, and NK cell markers such as Gzmb, Gzma, Klrg1, Ncr1, Klrd1, S1pr5, and Klre1 (Figure 6B). KLRG1+ cells also expressed Spry2 and Cx3cr1 (Figure 6B), which were previously identified in KLRG1+ iNKT cells by Murray et al., 2021, and the transcription factor Zeb2. ZEB2 is a key transcription factor regulating terminal differentiation of KLRG1+ CD8+ effector cells (Omilusik et al., 2015), suggesting that ZEB2 may be a candidate regulator of KLGR1+ iNKT cells. cMAF+ and KLRG1+ cells both demonstrated enrichment of the antigen experience marker Itga4 (Shimizu et al., 2014; Grau et al., 2018), suggesting that these two populations are memory-like or ‘trained’. Interestingly, whole-transcriptome correlation analysis revealed that cMAF+ cells and KLRG1+ cells were more similar to NKT1 cells than to NKT2 or NKT17 cells (Figure 6—figure supplement 1), and we noted that production of Granzyme A and IL-10 by iNKT cells at 4 weeks post-αGalCer, which we previously associated with KLRG1+ and cMAF+ iNKT cells, respectively (Figure 5), was also associated with co-production of IFNγ (Figure 6—figure supplement 3). Conversely, we did not detect co-production of IL-17A and IL-10 (Figure 6—figure supplement 4). This suggests that cMAF+ and KLRG1+ cells are NKT1-like populations and/or that cMAF+ and KLRG1+ cells may differentiate from NKT1 cells.
Having identified memory-like splenic cMAF+ and KLRG1+ iNKT cell populations that appear after αGalCer challenge, we wondered whether analogous populations also appeared in other organs after αGalCer challenge. Using flow cytometry, we identified a significant enrichment of KLRG1+ iNKT cells at 4 weeks post-αGalCer in the lung, liver, adipose tissue, and inguinal lymph nodes (Figure 6D and E), while cMAF+ iNKT cells were enriched in the liver, adipose tissue, and inguinal lymph nodes (Figure 6D and E). scRNA-seq also revealed distinct cMAF+ and KLRG1+ iNKT cell populations among 3014 adipose iNKT cells at 4 weeks post-αGalCer (Figure 6—figure supplement 5), indicating conserved enrichment of cMAF+ and KLRG1+ iNKT cells after αGalCer challenge across different tissues. Interestingly, we noted that minor populations of cMAF+ and KLRG1+ iNKT cells were present in the adipose tissue at steady state (Figure 6D and E), which correlated with our previous finding that adipose iNKT cells constitutively express markers of antigen experience and chronic activation (Figure 3). We have previously shown that adipose NK1.1- iNKT cells express Klrg1, Maf, and Itga4 at steady state (LaMarche et al., 2020), and subclustering of steady-state adipose NK1.1- iNKT cells revealed distinct cMAF+-like or KLRG1+-like iNKT cell subpopulations at the transcriptional level (Figure 6—figure supplement 5). Overall, these data indicate that cMAF+ and KLRG1+ iNKT cell populations are induced across multiple organs in response to antigenic stimulation and are constitutively present in adipose tissue.
Identification of a conserved cMAF-associated and NKT<sub>FH</sub>-like transcriptional state in NKT10 cells
Having identified transcriptional signatures of regulatory iNKT cells in adipose tissue and after serial antigen activation, we next sought to describe shared transcriptional features of these different NKT10 cell populations. Gene expression analysis identified 110 genes enriched among splenic NKT10 cells (Supplementary file 7), including Ctla4, Pdcd1, Lag3, Il21, Maf, Hif1a, and Ccr5 (Figure 7A), all of which were already identified in adipose NKT10 cells. We identified 39 genes conserved across NKT10 cells from both tissues (Figure 7B), including Tr1 cell markers, Tgfb1, and the tolerogenic factors Slfn2 and Vsir (Berger et al., 2010; ElTanbouly et al., 2020; Figure 7B). We also found that splenic NKT10 cells expressed the adipose iNKT cell marker Nfil3 (Figure 7A), which we previously linked to IL-10 production by regulatory adipose NKT10 cells (Lynch et al., 2015; LaMarche et al., 2020; Motomura et al., 2011). However, gene regulatory network analysis using GENIE3 (Huynh-Thu et al., 2010) of Il10 versus Nfil3 and other transcription factors identified in NKT10 cells revealed that Maf demonstrated the greatest correlation with Il10 in NKT10 cells (Figure 7C), matching our previous transcriptional and functional analysis correlating IL-10 with cMAF (Figure 5). Since cMAF is known to regulate IL-10 in other immune populations, such as Tr1 cells, B cells, and macrophages57,66,67, our data suggest that Maf is a major candidate regulator of NKT10 cells. Our analysis also indicates that cMAF+ iNKT cells are a memory-like population of NKT10 cells or that NKT10 cells are significantly enriched among cMAF+ iNKT cells.
Identification of a conserved cMAF-associated and NKT<sub>FH</sub>-like transcriptional state in NKT10 cells.
(A) Radar chart showing Log2(Fold Change) values of genes enriched in murine splenic activated Il10pos invariant natural killer T (iNKT) cells versus activated Il10neg iNKT cells at 4 hr post-α-galalctosylceramide (post-αGalCer). (B) Venn diagram showing overlap of genes enriched in murine splenic (left) and adipose (right) Il10pos iNKT cells after αGalCer (activated 4 hr post-αGalCer splenic iNKT cells, and 4 hr and 72 hr post-αGalCer for adipose iNKT cells). (C) Lollipop plot showing correlation (Link Weight) values of different transcription factors versus Il10 from GENIE3 analysis of total splenic and adipose NKT10 cells. Source data provided in Figure 7—source data 1. (D) Violin plots showing expression and enrichment of NKTFH signature gene module scoring in resting splenic iNKT cells, with NKT1-A, NKT1-B, NKT2, NKT17 and cycling cell clusters pooled together (all other iNKT cells) and compared versus cMAF+ and KLRG1+ iNKT cell clusters. The central violin plot horizontal line denotes the median value. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. (E) Bar plots showing bulk RNA-seq gene expression values in untreated total murine splenic iNKT cells (steady state) and murine splenic NKTFH or non-NKTFH cells (6 days post-αGalCer). Reanalyzed from GSE161492. Asterisks indicate significantly increased expression versus all other populations or versus steady state alone (Itga4 only). * Padj<0.05; ** Padj<0.01; *** Padj<0.001. (F) Representative pseudocolor plots of CXCR5 versus PD-1 expression in murine total steady-state splenic iNKT cells (1), total resting splenic iNKT cells at 4 weeks post-αGalCer (2) or splenic RORyT- cMAF+ iNKT cells at 4 weeks post-αGalCer (3). iNKT cells were defined as live, single CD19- CD8- F4/80- CD3low CD1d-PBS57 tetramer+ cells. (G) Box plots quantifying the data from (F). N = 5 biological replicates from one experiment. Experiment performed once. Source data provided in Figure 7—source data 2. The central box plot horizontal line denotes the median value.
NKTFH signature gene expression in control and αGalCer-pretreated splenic iNKT cells at rest.
(A) Robust Multichip Average (RMA)-normalized microarray NKTFH signature gene expression in control and α-galalctosylceramide (αGalCer)-pretreated murine splenic invariant natural killer T (iNKT) cells at rest. Reanalyzed from GSE47959. Values below 0.25 units of expression were set to 0.25 for plotting. (B) Representative histogram of BCL6 expression in total murine splenic iNKT cells at 4 weeks post-αGalCer, RORγT- cMAF+ murine splenic iNKT cells and murine CXCR5+ PD-1+ (NKTFH) cells. iNKT cells were defined as live, single CD19- CD8- CD3low CD1d-PBS57 tetramer+ cells. Histograms were normalized to the mode. (C) Box plot quantifying BCL6 median fluorescence intensity (MFI) in the populations from (B). N = 5 biological replicates from one experiment. Experiment performed once. Student’s unpaired t-test. Asterisks denote significance, * Padj<0.05; ** Padj<0.01; *** Padj<0.001. The central box plot horizontal line denotes the median value. (D) Violin plot of Bcl6 expression in murine resting splenic iNKT cell clusters at 4 weeks post-αGalCer.
We next sought to compare the transcriptional signature of NKT10/cMAF+ cells against other memory-like iNKT cell populations. Interestingly, module scoring of our resting scRNA-seq data demonstrated that NKT10/cMAF+ cells but not KLRG1+ cells showed enrichment for NKTFH cell gene signatures (Figure 7D). Reanalysis of published bulk iNKT cell RNA-seq data 6 days post-αGalCer (GSE161492) (Murray et al., 2021) also revealed that NKTFH cells but not steady-state or non-NKTFH cells expressed NKT10/cMAF+ markers, including Maf, Cd4, Tox, Izumo1r, Slamf6, Hif1a, and Lag3 (Figure 7E, Supplementary file 7), and we found co-enrichment of NKT10/cMAF+ and NKTFH markers in published microarray data comparing αGalCer-pretreated versus steady-state splenic iNKT cells (GSE47959) (Figure 5—figure supplement 1, Figure 7—figure supplement 1, Supplementary file 8). These data indicated that NKT10/cMAF+ cells are transcriptionally similar to NKTFH cells, and suggested that these two memory-like iNKT cell populations might overlap. However, we found that most cMAF+ iNKT cells did not co-express the NKTFH markers CXCR5 and PD-1 (Figure 7F and G), and we only detected a minor population of CXCR5+ PD-1+ NKTFH cells among cMAF+ iNKT cells (~7% of cells, Figure 7F and G). We also did not detect enrichment of the flagship NKTFH transcription factor BCL6 in cMAF+ iNKT cells either by flow cytometry or by scRNA-seq (Figure 7—figure supplement 1). Therefore, our data indicate that while NKT10/cMAF+ iNKT cells are transcriptionally similar to NKTFH cells, most NKT10/cMAF+ iNKT cells are not bone fide NKTFH cells, and NKT10/cMAF+ cells represent a distinct lineage of memory-like iNKT cells induced following activation with αGalCer.
Discussion
iNKT cells and other innate T cells, such as γδ T cells and MAIT cells, rapidly activate following antigen or cytokine stimulus and produce potent cytokine responses. In this study, we characterized the transcriptional programs underpinning the iNKT cell response to antigen, revealing an initial phase of rapid cytokine production, Zbtb16 upregulation, and metabolic gene remodeling, which was followed by second phase of proliferation, further remodeling of metabolic gene programs, and acquisition of features associated with memory-like iNKT cells. This transcriptional framework was highly conserved across iNKT cells from different tissues and species, and between functionally distinct iNKT cell subsets, including NKT1, NKT2, and NKT17 cells. Interestingly, we identified many similarities with adaptive T cell activation, including early induction of biosynthesis and aerobic glycolysis gene signatures (Hartmann et al., 2021), although in keeping with the poised phenotype of iNKT cells this occurred over hours instead of days, in concert with innate-like cytokine production, and expression of metabolic activation and cytokine production gene signatures was largely diminished by 3 days post-activation.
We sequenced 48,813 iNKT cells, the largest number of iNKT cells analyzed by scRNA-seq to date, which allowed us to explore iNKT cell heterogeneity in unprecedented detail. We found that NKT2 and NKT17 cells, despite sharing many features of activation with NKT1 cells, were more enriched for genes associated with oxidative metabolism after activation, and production of NKT2 and NKT17 cell cytokines was more dependent on oxidative phosphorylation compared to NKT1 cell cytokines. Recent work by Raynor et al., 2020 has shown that NKT2 and NKT17 cells are more enriched for oxidative metabolism than NKT1 cells during thymic development, and downregulation of oxidative metabolism was required for establishment of an NKT1 cell phenotype (Raynor et al., 2020). Here, we show that that enrichment of oxidative metabolism persists in splenic NKT2 and NKT17 cells after thymic development and defines their function, and we also identified enrichment of oxidative gene signatures in adipose NKT17 cells. Interestingly, we have shown that γδ17 cells display enrichment of oxidative metabolic signatures (Lopes et al., 2021), suggesting that enrichment of oxidative metabolism may be a conserved feature of IL-17-producing innate T cells.
Our study found that regulatory iNKT cell populations exhibited a blunted and/or delayed response to αGalCer, coupled with increased expression of early exhaustion and regulatory Tr1 cell markers. This agrees with previous studies demonstrating long-term anergy in iNKT cells treated with αGalCer (Parekh et al., 2005; Sag et al., 2014) and enrichment of regulatory iNKT cells after prior αGalCer challenge (Sag et al., 2014). Interestingly, we found that adipose iNKT cells, which are constitutively enriched for NKT10 cells, showed evidence of chronic endogenous activation, and repeated αGalCer challenge induced an adipose-like phenotype in splenic iNKT cells. Thus, chronic activation can promote a type of iNKT cell anergy, characterized by enrichment of regulatory iNKT cells. Chronic activation is known to promote exhaustion and anergy in adaptive T cells (Liu et al., 2019; Seo et al., 2019), and a recent study by Vorkas et al., 2022 found that chronically activated human MAIT cells upregulate expression of FOXP3 and early exhaustion markers such as PD-1 and LAG-3 (Vorkas et al., 2022). Functionally, induction of regulatory and/or anergic innate T cells following chronic activation could serve as a protective mechanism to prevent overactivation of innate T cells and limit immunopathogenicity. This could be relevant in autoimmunity, and it is notable that chronic activation of iNKT cells with αGalCer has been shown to ameliorate experimental autoimmune encephalomyelitis (Jahng et al., 2001; Singh et al., 2001).
Characterization of regulatory iNKT cells revealed two distinct memory-like iNKT cell populations, cMAF+ cells and KLRG1+ cells, which were enriched across multiple organs after prior antigen exposure and constitutively present in adipose tissue regardless of activation status. KLRG1+ iNKT cells showed enrichment of gene signatures associated with cytotoxic effector memory CD8+ T cells and NK cells, and expressed the transcription factor Zeb2. By contrast, cMAF+ iNKT cells were transcriptionally similar to CD4+ memory T cells, stem-like memory T cells, precursor exhausted T cells, Tregs and NKTFH cells, and were enriched for IL-10-producing NKT10 cells. Although cMAF+ and NKTFH cells were similar at the transcriptional level, we found that most cMAF+ cells could not be classified as NKTFH cells, indicating that cMAF+ iNKT cells represent a distinct lineage of memory-like iNKT cells. However, given the transcriptional similarity between cMAF+ and NKTFH cells, it is possible these populations may follow a similar differentiation trajectory. Moreover, given the similarity between KLRG1+ cells and CD8+ effector/memory T cells, and CD4+ memory T cells, cMAF+ cells and NKTFH cells, we hypothesize that there could be a bifurcating differentiation branch during the formation of memory-like iNKT cell populations, whereby a CD8+ effector/memory-like KLRG1+ iNKT cell population differentiates along one trajectory, and CD4+ memory-like cMAF+ and/or NKTFH cell populations differentiate along another trajectory. Mechanistically, it is possible that some of the same transcription factors that control the differentiation of CD8+ vs. CD4+ memory could regulate the development of different memory-like iNKT cell lineages. For example, Zeb2 is known to be a key regulator of the terminal differentiation of effector memory T CD8+ cells (Omilusik et al., 2015; Evrard et al., 2022) and could play a similar role in KLRG1+iNKT cell differentiation. Further research will be required to elucidate the dynamics of memory-like iNKT cell differentiation in more detail.
We have shown that E4BP4 (Nfil3) rather than FOXP3 regulates production of IL-10 by adipose iNKT cells (Lynch et al., 2015; LaMarche et al., 2020), and we identified Nfil3 expression among adipose and splenic NKT10 cells. However, most adipose iNKT cells express E4BP4 and do not express IL-10, suggesting that other factors must regulate IL-10. Furthermore, E4BP4 has been shown to bind an intronic region of the IL-10 promoter (Motomura et al., 2011), suggesting that other promoter-associated transcription factors are likely required to effectively induce IL-10 production in iNKT cells. Here, we implicate the AP-1 family transcription factor cMAF as a candidate regulator of NKT10 cells, correlating with published literature identifying a role for cMAF in regulation of IL-10 production by macrophages, B cells, and adaptive T cells, including Tr1 cells (Chihara et al., 2016; Apetoh et al., 2010; Liu et al., 2018; Cao et al., 2005). cMAF is also expressed by NKT17 cells and is required for their development (Thapa et al., 2017), as well as for γδ17 cell development (Zuberbuehler et al., 2019). However, although NKT17 cells can produce IL-10 after in vitro expansion (Cameron and Godfrey, 2018), we and others did not identify co-expression of IL-10 and IL-17 among in vivo or ex vivo iNKT cells (Sag et al., 2014). Interestingly, RORγT has been demonstrated to suppress production of IL-10 in Th17 cells (Sun et al., 2019), suggesting that it could perform a similar role in iNKT cells. We found that NKT10 cells co-expressed Ifng, Il4, and Il21 with Il10, and lacked Rorc expression, indicating that NKT10 cells are more similar to NKT1 cells and NKTFH cells than NKT17 cells. Notably, Murray et al. showed that most NKTFH cells arise from T-bet+ NKT1 cells; given the transcriptional similarity of NKT10 and NKTFH cells, it is possible NKT10 cells also arise from an NKT1 cell lineage. We also noted that NKT10 cells were transcriptionally similar to Tr1 cells, which are regulatory but lack expression of FOXP3, and are known to express cMAF (Yan et al., 2017; Gruarin et al., 2019; Chihara et al., 2016; Alfen et al., 2018; Mascanfroni et al., 2015; Apetoh et al., 2010). Therefore, our data suggest that NKT10 cells may be similar to Tr1 cells (Yan et al., 2017; Chihara et al., 2016; Alfen et al., 2018). Tr1 cells are heterogeneous, and we found that expanded adipose NKT10 cells were heterogeneous, with some cells expressing Il4 and Il21, and other cells expressing Gzmb.
Understanding the factors controlling the phenotype and generation of regulatory iNKT cells has a several potential applications. For example, NKT10 cells have been shown to emerge in tumors, where they can promote tumor growth and regulatory T cell function (Wang et al., 2018b). However, our work indicates that NKT10 cells are capable of co-expressing IFNγ and IL-10. Skewing regulatory iNKT cells toward a less regulatory phenotype by enhancing IFNγ and inhibiting IL-10 might have a therapeutic benefit. We have shown that regulatory adipose iNKT cells can induce weight loss and suppress adipose tissue inflammation (Lynch et al., 2016), and are enriched in human adipose tissue (Lynch et al., 2009), suggesting that modulation of adipose iNKT cells might be beneficial during obesity and metabolic syndrome. Insights into the factors and transcriptional signatures governing iNKT cell activation are also potentially relevant to other innate T cell populations, as γδ T cells and MAIT cells respond similarly after activation (Godfrey et al., 2015; Vorkas et al., 2020), and the transcriptomic resource that we present here may assist other studies investigating these innate T cell responses.
MethodsIn vivo stimulations
αGalCer (Avanti Polar Lipids) was prepared by dissolving in sterile DMSO (Sigma) at a concentration of 1 mg/mL. αGalCer was prepared for injection by dilution in sterile phosphate-buffered saline (PBS) (Sigma), and the final concentration of DMSO was adjusted to less than 10% vol/vol. Mice were injected IP with 1 µg of αGalCer or PBS vehicle. Mice were sacrificed and tissues harvested after either 4 or 72 hr. For analysis of memory-like iNKT cells, mice were injected IP with 4 µg of αGalCer and sacrificed after 4 weeks, with some animals receiving an additional IP injection of 1 µg of αGalCer 4 hr before sacrifice. All injections had a final volume of 100 µL.
Tissue processing
Adipose tissue was excised, minced with a razor, and digested in 1 mg/mL Collagenase Type II (Worthington) in RPMI shaking for 25–30 min at 37°C. Digested cells were filtered through a 70 μM nylon mesh and centrifuged at 15,000 rpm for 5–7 min to pellet the stromovascular fraction (SVF). Spleens and thymi were disrupted through a 70 μM filter and pelleted. Red blood cells in the spleen and thymus were lysed with ACK Lysing Buffer (VWR) or RBC Lysis Buffer (BioLegend) prior to further analysis.
Ex vivo stimulations and mitochondrial staining
Where indicated, thymocytes, splenocytes, and the isolated SVF from adipose tissue were cultured for 4 hr in the presence of PMA and ionomycin (Cell Stimulation Cocktail, BioLegend) and Brefeldin A (BioLegend), and in the presence or absence of oligomycin (Sigma) or Monesin (BioLegend). Cultures were in complete RPMI media supplemented with L-glutamine, penicillin, streptomycin, and 10% FBS (Thermo Fisher Scientific). For mitochondrial staining, thymocytes and splenocytes were cultured for 30 min in complete RPMI media containing TMRM (Invitrogen) and/or MitoTracker Green FM (Thermo Fisher Scientific).
Flow cytometry and cell sorting
All antibody staining of live cells was performed in PBS (Gibco) with 1–2% FBS. Single-cell suspensions were incubated in Fc-receptor blocking antibody (Clone 93, BioLegend) during cell surface antigen staining. Dead cells were excluded with Fixable Viability dyes (UV, eFluor 780; Thermo Fisher Scientific) or Zombie Aqua (BioLegend). For intracellular antigen staining, cells were fixed with either True-Nuclear Transcription Factor Buffer Set (BioLegend) for 45 min at room temperature, Foxp3/Transcription Factor Fixation/Permeabilization kit (Thermo Fisher Scientific), or Cytofix/Cytoperm kit (BD Biosciences) for 30 min at room temperature. The following anti-mouse antibodies were obtained from BioLegend: anti-CD3 (17A2), anti-CD19 (6D5), anti-CD11b (M1/70), anti-CD45 (30-F11), anti-TCRβ (H57-597), anti-CD8a (53–6.7), anti-KLRG1 (2F1/KLRG1), anti-IL-17A (TC11-18H10.1), anti-IL-4 (11B11), anti-IFNγ (XMG1.2), anti-CD43 Activation Glycoform (1B11), anti-ICOS (C398.4A), and anti-CXCR5 (L138D7). The following anti-mouse antibodies were obtained from Thermo Fisher Scientific: anti-cMAF (sym0F1), anti-RORγT (B2D), anti-RORγT (AFKJS-9), anti-CD8a (53–6.7), anti-IL-10 (JES5-16E3), anti-IL-13 (eBio13A), anti-CD11b (M1/70), anti-F4/80 (BM8), anti-BCL6 (BCL-DWN), and anti-Granzyme A (GzA-3G8.5). The following anti-mouse antibodies were obtained from BD Biosciences: anti-IFNγ (XMG1.2). iNKT cells were identified as live, single lymphocytes binding to anti-TCRβ or anti-CD3 antibodies and αGalCer analog PBS57-loaded CD1d tetramer (NIH Tetramer Core Facility/Emory Vaccine Center). A ‘dump’ channel with antibodies against CD19, or CD19 and CD11b or F4/80, or CD19, CD8a, and CD11b or F4/80 was used to eliminate nonspecific staining. For staining of BALB/c thymic iNKT cell subsets, NKT1 cells were gated as CD43-HG- ICOS- CD3low cells, NKT2 cells were gated as CD43-HG- ICOS+ CD3high cells, and NKT17 cells were gated as CD43-HG+ cells. Samples were acquired using LSR Fortessa and FACS Canto II cytometers and sorted using a FACS Aria Fusion Cell Sorter. Flow cytometry analysis and plots were created using FlowJo version 10.0.7r2.
scRNA-seq sequencing and data preprocessing
scRNA-seq was performed on single-cell suspensions of sorted iNKT cells from the visceral adipose tissue and spleens of mice using the 10X Genomics platform. A total of 35 visceral adipose tissue deposits from 35 mice or 5 spleens from 5 mice were pooled for each scRNA-seq sample. Nine biological samples were sequenced in three batches (Supplementary file 9). For two of the batches, adipose and/or splenic iNKT cell samples were first tagged by TotalSeq-A Mouse hashtag antibodies (BioLegend) and then pooled for sequencing. Cell suspensions were loaded onto a 10x Chromium Controller to generate single-cell Gel Beads-in-emulsion (GEMS) and GEMs were processed to generate UMI-based libraries according to the 10X Genomics Chromium Single Cell 3′ protocol. Libraries were sequenced using a NextSeq 500 sequencer (Illumina). Raw BCL files were demultiplexed using Cell Ranger v3.0.2 mkfastq to generate fastq files with default parameter. Fastq files were aligned to the mm10 genome (v1.2.0) and feature reads were quantified simultaneously using Cell Ranger count for feature barcoding. The resulting filtered feature-barcode UMI count matrices containing quantification of gene expression and hashtag antibody binding were then utilized for downstream analysis.
Downstream scRNA-seq data analysis
A total of 48,813 cells murine iNKT cells expressing a minimum median of 1567 genes per cell and 4163 UMIs per cell were loaded from feature-barcode UMI count matrices using the Seurat v4.0.3 package (Hafemeister and Satija, 2019). scRNA-seq data for steady-state adipose iNKT cells and splenic iNKT cells at steady state and 4 hr post-αGalCer were previously uploaded to GSE142845 (LaMarche et al., 2020). Raw human iNKT cell scRNA-seq data was downloaded from GSE128243 (Zhou et al., 2020). Antibody hashtag data was demultiplexed using the Seurat HTODemux function with a positive quantile of 0.99 and centered log ratio transformation normalization. Cells positive for more than one antibody were removed from the analysis. Genes expressed in less than three cells were excluded from all analyses to prevent false-positive identification of transcripts. Cells expressing less than 1% minimum or more than 12.5% maximum of mitochondrial genes as a % of total gene counts were considered to represent empty droplets or apoptotic/dead cells and were removed from the analysis. Cells were also filtered based on total UMI counts and total gene counts on a per-sample basis to remove empty droplets, poor quality cells and doublets, with a minimum cutoff of at least 500 genes per cell across all samples. UMI counts were normalized using regularized negative binomial regression using sctransform v0.3.2 (Hafemeister and Satija, 2019). Where indicated, cycle regression was performed by first normalizing UMI counts using sctransform, then performing cell cycle scoring using the CellCycleScoring() function and cell cycle gene lists provided with the Seurat package, and then re-normalizing raw RNA count data with sctransform and regression of computed cell cycle scores applied.
Dimensionality reduction was performed using principal component analysis (PCA) with n = 100 dimensions and 2000 or 3000 variable features, and an elbow plot was used to determine the number of PCA dimensions used as input for UMAP (McInnes et al., 2018). For collective analysis of cells from different batches, the harmony v1.0 package (Korsunsky et al., 2019) was used with default settings to remove batch effects, and batch-corrected harmony embeddings were used for UMAP. Outlier cells expressing genes associated with macrophages (e.g., Adgre1, Cd14), B cells (e.g.,. Cd19), or CD8+ T cells (e.g., Cd8a) were also identified, and these cells were removed prior to the final analysis. UMAP was performed using a minimum distance of 0.3 and a spreading factor of 1. Shared nearest neighbor (SNN) graphs were calculated using k = 20 nearest neighbors. Graph-based clustering was performed using the Louvain algorithm. In some cases, overclustering was performed and clusters were manually collapsed, and/or the first two dimensions of the UMAP reduction were used as input for graph-based clustering instead of PCA or harmony embeddings. Steady-state splenic NKT1, NKT2, and NKT17 cell identification was performed using expression of Tbx21, Zbtb16, and Rorc after graph-based clustering. Cycling cells were identified using expression of Mki67. Activated NKT1, NKT2, and NKT17 cell identification was performed using expression of Ifng, Il4, Il13, Il17a, and Il17f after graph-based clustering. Identified NKT1, NKT2, and NKT17 cells from steady state and 4 hr post-αGalCer datasets were combined by subset, and low-level graph-based clustering was performed to stratify ‘true’ steady state and activated iNKT cells into separate steady state or activated clusters. Reclustered steady state and activated iNKT cells from different subsets were then recombined and renormalized for final downstream analysis.
Gene expression analysis was performed using the FindMarkers() or FindAllMarkers() Seurat functions and the Wilcoxon rank-sum test. All gene expression analyses were performed using log-normalized RNA counts. Gene set enrichment analysis (GSEA) (Subramanian et al., 2005) was performed using FGSEA v1.17.0 (Korotkevich et al., 2016) and clusterProfiler v3.99.2 (Yu et al., 2012) packages. KEGG (Kanehisa et al., 2021) pathway data was retrieved from the Molecular Signatures Database (MSigDB) (Liberzon et al., 2011) using the msigdbr v7.4.1 package. Over-representation analysis was performed using g:Profiler (Raudvere et al., 2019). For cross-dataset differential gene expression analysis of adipose versus splenic iNKT cells, individual gene expression analyses were first performed between adipose and splenic iNKT cells at steady state, 4 hr post-αGalCer and 72 hr post-αGalCer. Genes significantly enriched in either adipose or splenic iNKT cells across all three analyses were identified, which included 971 genes enriched among adipose iNKT cells and 65 genes enriched among splenic iNKT cells. Over-representation analysis was then performed on these enriched genes using gprofiler. Heatmaps were generated using the Complex Heatmap v2.7.13 and circlize v0.4.13 packages (Gu et al., 2016). Module scores were calculated using the AddModuleScore() Seurat function with n = 10 control features. Density plots were produced using the Nebulosa v1.1.1 package (Alquicira-Hernandez and Powell, 2021). Gene regulatory network analysis was performed using GENIE3 v1.14.0 (Huynh-Thu et al., 2010) with n = 10 iterations and link weight values were averaged between replicate analyses. Other plots were created using egg v0.4.5, GGally v2.1.2, ggiraphExtra v0.3.0, ggpubr v0.4.0, pals v1.7, patchwork v1.1.1, tidyverse v1.3.1, tidymodels v0.1.3, and viridis v0.6.1.
Bulk RNA-seq and microarray analysis
Raw RNA-seq count files were downloaded from GEO Repository GSE161492 (Murray et al., 2021). Microarray data was downloaded from GEO Repository GSE47959 (Sag et al., 2014). Raw CEL files were annotated against the Mouse430_2 Array (mouse4302.db) and Robust Multichip Average (RMA) normalized using the affy v1.7.0 R package. Discrete probes corresponded to the same gene were merged and values were averaged. Raw RNA-seq counts were transformed using the cpm() function with log = TRUE and trimmed mean of M-values (TMM) normalized with edgeR using edgeR v3.33.7 (Robinson et al., 2010). Genes with low read counts were filtered out using the edgeR filterByExpr() function. Testing for differential gene expression was performed with Limma-Voom using limma v3.47.16 (Smyth, 2005; Law et al., 2014) with standard settings and without a minimum fold change cutoff. Plots were generated using the same libraries as used for scRNA-seq data plotting.
Statistics
Sample size for adequate power was determined based on previous studies (Lynch et al., 2015; LaMarche et al., 2020). Significance was determined by Student’s two-tailed t-test with Holm–Bonferroni correction, or one-way ANOVA with Tukey’s post hoc, where indicated. Significance is presented as *p<0.05, **p<0.01, ***p<0.001, with p>0.05 considered nonsignificant. All statistical analyses were performed using rstatix v0.7.0, stats v4.1.2, and ggpubr v0.4.0 R packages or GraphPad Prism v9.4.1. All sequencing data analyses were performed using R 4.1.2 and RStudio Desktop v1.4.1712 on an Ubuntu 20.04 Linux GNU (64 bit) system.
Additional informationCompeting interests
No competing interests declared
No competing interests declared
Author contributions
Conceptualization, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing – review and editing
Formal analysis, Validation, Investigation, Writing – review and editing
Validation, Investigation
Data curation, Project administration
Data curation, Formal analysis, Methodology
Data curation, Formal analysis
Data curation
Data curation, Formal analysis
Resources, Funding acquisition, Writing – review and editing
Male C57BL/6 mice were purchased from Jackson Laboratory or Harlan Laboratory. BALB/c mice were purchased from Harlan Laboratory. Male and female BALB/c mice were bred under specific-pathogen-free facilities at Trinity College Dublin. Mice were used in experiments at 6–14 weeks of age. All animal work was approved and conducted in compliance with the Trinity College Dublin University Ethics Committee and the Health Products Regulatory Authority Ireland, and the Institutional Animal Care and Use Committee guidelines of The Dana Farber Cancer Institute and Harvard Medical School. C57BL/6 mice were used for all experiments unless otherwise specified. Balb/c mice were used for surface staining of markers delineating NKT subsets so that co-staining for mitochondrial markers was possible.
Additional files
Gene expression data associated with the scRNA-seq data in <xref ref-type="fig" rid="fig1">Figure 1B</xref>.
Lists of genes used for module scoring of scRNA-seq data, starting with <xref ref-type="fig" rid="fig1">Figure 1D</xref>.
Gene expression data associated with the scRNA-seq data in <xref ref-type="fig" rid="fig2">Figure 2A</xref>.
Lists of conserved gene expressed in adipose or splenic invariant natural killer T (iNKT) cells, associated with <xref ref-type="fig" rid="fig3">Figure 3B</xref>.
Lists of conserved gene expressed in adipose NKT10 cells, associated with <xref ref-type="fig" rid="fig4">Figure 4E</xref>.
Lists of conserved gene expressed in adipose NKT10 cells, associated with <xref ref-type="fig" rid="fig7">Figure 7A</xref>.
Gene expression data associated with reanalysis of bulk RNA-seq data from GSE161492 in <xref ref-type="fig" rid="fig7">Figure 7E</xref>.
Gene expression data associated with reanalysis of bulk microarray data from GSE47959, first referenced in <xref ref-type="fig" rid="fig5s1">Figure 5—figure supplement 1</xref>.
Information about scRNA-seq data batch processing and sample information, associated with the ‘Methods’ section.
Data availability
Sequencing data have been deposited in GEO under accession code GSE190201.
The following dataset was generated:
KaneHLaMarcheNMBrennerMBLynchL2021Single cell analysis of activated iNKT cells from murine epididymal adipose tissue and spleenNCBI Gene Expression OmnibusGSE190201
The following previously published datasets were used:
LaMarcheNMKaneHKohlgruberACLynchLBrennerMB2020Single cell analysis of iNKT cells from murine epididymal adipose tissue and spleenNCBI Gene Expression OmnibusGSE142845
WangJAdriantoIWuXZhouLMiQ2020Single-cell RNA-seq of human peripheral blood NKT cellsNCBI Gene Expression OmnibusGSE128243
WingenderGKronenbergM2014NKT-10 cells represent a novel invariant NKT cell subset with regulatory characteristicsNCBI Gene Expression OmnibusGSE47959
Acknowledgements
The authors thank the Brigham and Women’s Hospital Single Cell Genomics Core for assistance with sequencing and preprocessing of scRNA-seq data, the NIH Tetramer Core Facility for recombinant CD1d-PBS57 tetramers, and A.T. Chicoine for cell sorting. This work was supported by NIH grants R01 AI134861, American Diabetes Association 1-16-JDF-061 and ERC Starting grant 14283 (Project ID: 679173) (to LL), and R01 AI113046 (to MBB). Cartoons were created with BioRender.com.
ReferencesAlfenJSLarghiPFacciottiFGaglianiNBosottiRParoniMMaglieSGruarinPVascoCMRanzaniVFrusteriCIsepponAMoroMCrostiMCGattiSPaganiMCaprioliFAbrignaniSFlavellRAGeginatJ2018Intestinal IFN-γ-producing type 1 regulatory T cells coexpress CCR5 and programmed cell death protein 1 and downregulate IL-10 in the inflamed guts of patients with inflammatory bowel diseaseThe Journal of Allergy and Clinical Immunology1421537154710.1016/j.jaci.2017.12.98429369775Alquicira-HernandezJPowellJE2021Nebulosa recovers single cell gene expression signals by kernel density estimationBioinformatics372485248710.1093/bioinformatics/btab00333459785AndreattaMCorria-OsorioJMüllerSCubasRCoukosGCarmonaSJ2021Interpretation of T cell states from single-cell transcriptomics data using reference atlasesNature Communications12296510.1038/s41467-021-23324-434017005AndrisFDenanglaireSAnciauxMHercorMHusseinHLeoO2017The transcription factor c-Maf promotes the differentiation of follicular helper T cellsFrontiers in Immunology848010.3389/fimmu.2017.0048028496444AngiariSRuntschMCSuttonCEPalsson-McDermottEMKellyBRanaNKaneHPapadopoulouGPearceELMillsKHGO’NeillLAJ2020Pharmacological activation of pyruvate kinase M2 inhibits CD4+ T cell pathogenicity and suppresses autoimmunityCell Metabolism3139140510.1016/j.cmet.2019.10.01531761564ApetohLQuintanaFJPotCJollerNXiaoSKumarDBurnsEJSherrDHWeinerHLKuchrooVK2010The aryl hydrocarbon receptor interacts with c-Maf to promote the differentiation of type 1 regulatory T cells induced by IL-27Nature Immunology1185486110.1038/ni.191220676095BaranekTLebrigandKde Amat HerbozoCGonzalezLBogardGDietrichCMagnoneVBoisseauCJouanYTrotteinFSi-TaharMLeite-de-MoraesMMallevaeyTPagetC2020High dimensional single-cell analysis reveals inkt cell developmental trajectories and effector fate decisionCell Reports3210811610.1016/j.celrep.2020.10811632905761BergerMKrebsPCrozatKLiXCrokerBASiggsOMPopkinDDuXLawsonBRTheofilopoulosANXiaYKhovananthKMorescoEMYSatohTTakeuchiOAkiraSBeutlerB2010An SLFN2 mutation causes lymphoid and myeloid immunodeficiency due to loss of immune cell quiescenceNature Immunology1133534310.1038/ni.184720190759CameronGGodfreyDI2018Differential surface phenotype and context‐dependent reactivity of functionally diverse NKT cellsImmunology & Cell Biology9675977110.1111/imcb.1203429504657CaoSLiuJSongLMaX2005The protooncogene c-maf is an essential transcription factor for IL-10 gene expression in macrophagesJournal of Immunology1743484349210.4049/jimmunol.174.6.348415749884CarbonS2021The gene ontology resource: enriching a gold mineNucleic Acids Research49D325D33410.1093/nar/gkaa111333290552CarlsonCMEndrizziBTWuJDingXWeinreichMAWalshERWaniMALingrelJBHogquistKAJamesonSC2006Kruppel-Like factor 2 regulates thymocyte and T-cell migrationNature44229930210.1038/nature0488216855590ChangP-PBarralPFitchJPratamaAMaCSKalliesAHoganJJCerundoloVTangyeSGBittmanRNuttSLBrinkRGodfreyDIBatistaFDVinuesaCG2011Identification of bcl-6-dependent follicular helper NKT cells that provide cognate help for B cell responsesNature Immunology13354310.1038/ni.216622120117ChenYZanderRKhatunASchauderDMCuiW2018aTranscriptional and epigenetic regulation of effector and memory CD8 T cell differentiationFrontiers in Immunology9282610.3389/fimmu.2018.0282630581433ChenZZhuSWangLXieDZhangHLiXZhengXDuZLiJBaiL2018bMemory follicular helper invariant NKT cells recognize lipid antigens on memory B cells and elicit antibody recall responsesJournal of Immunology2003117312710.4049/jimmunol.170102629581354ChiharaNMadiAKarwaczKAwasthiAKuchrooVK2016Differentiation and characterization of tr1 cellsCurrent Protocols in Immunology113im0327s11310.1002/0471142735.im0327s11327038462CohenNRBrennanPJShayTWattsGFBriglMKangJBrennerMBImmGen Project Consortium2013Shared and distinct transcriptional programs underlie the hybrid nature of iNKT cellsNature Immunology14909910.1038/ni.249023202270ElTanboulyMAZhaoYNowakELiJSchaafsmaELe MercierICeerazSLinesJLPengCCarriereCHuangXDayMKoehnBLeeSWSilva MoralesMHogquistKAJamesonSCMuellerDRothsteinJBlazarBRChengCNoelleRJ2020Vista is a checkpoint regulator for naïve T cell quiescence and peripheral toleranceScience367052410.1126/science.aay052431949051EngelISeumoisGChavezLSamaniego-CastruitaDWhiteBChawlaAMockDVijayanandPKronenbergM2016Innate-Like functions of natural killer T cell subsets result from highly divergent gene programsNature Immunology1772873910.1038/ni.343727089380EvrardMWynne-JonesEPengCKatoYChristoSNFonsecaRParkSLBurnTNOsmanMDeviSChunJMuellerSNKannourakisGBerzinsSPPellicciDGHeathWRJamesonSCMackayLK2022Sphingosine 1-phosphate receptor 5 (S1PR5) regulates the peripheral retention of tissue-resident lymphocytesThe Journal of Experimental Medicine219e2021011610.1084/jem.2021011634677611FinlayDKRosenzweigESinclairLVFeijoo-CarneroCHukelmannJLRolfJPanteleyevAAOkkenhaugKCantrellDA2012Pdk1 regulation of mTOR and hypoxia-inducible factor 1 integrate metabolism and migration of CD8+ T cellsThe Journal of Experimental Medicine2092441245310.1084/jem.2011260723183047FuSZhuSTianCBaiSZhangJZhanCXieDWangLLiZLiJZhangHZhouRTianZXuTBaiL2019Immunometabolism regulates TCR recycling and iNKT cell functionsScience Signaling12eaau178810.1126/scisignal.aau178830808817GeissmannFCameronTOSidobreSManlongatNKronenbergMBriskinMJDustinMLLittmanDR2005Intravascular immune surveillance by CXCR6+ NKT cells patrolling liver sinusoidsPLOS Biology3e11310.1371/journal.pbio.003011315799695GodfreyDIUldrichAPMcCluskeyJRossjohnJMoodyDB2015The burgeoning family of unconventional T cellsNature Immunology161114112310.1038/ni.329826482978GrauMValsesiaSMafilleJDjebaliSTomkowiakMMathieuALLaubretonDde BernardSJouvePEVentreEBuffatLWalzerTLeverrierYMarvelJ2018Antigen-induced but not innate memory CD8 T cells express NKG2D and are recruited to the lung parenchyma upon viral infectionJournal of Immunology2003635364610.4049/jimmunol.170169829632146GrossmanWJVerbskyJWTollefsenBLKemperCAtkinsonJPLeyTJ2004Differential expression of granzymes A and B in human cytotoxic lymphocyte subsets and T regulatory cellsBlood1042840284810.1182/blood-2004-03-085915238416GruarinPMaglieSDe SimoneMHäringerBVascoCRanzaniVBosottiRNoddingsJSLarghiPFacciottiFSarnicolaMLMartinovicMCrostiMMoroMRossiRLBernardoMECaprioliFLocatelliFRossettiGAbrignaniSPaganiMGeginatJ2019Eomesodermin controls a unique differentiation program in human IL-10 and IFN-γ coproducing regulatory T cellsEuropean Journal of Immunology499611110.1002/eji.20184772230431161GuZEilsRSchlesnerM2016Complex heatmaps reveal patterns and correlations in multidimensional genomic dataBioinformatics322847284910.1093/bioinformatics/btw31327207943HafemeisterCSatijaR2019Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regressionGenome Biology2011510.1186/s13059-019-1874-1Harsha KroviSZhangJMichaels-FosterMJBrunettiTLohLScott-BrowneJGapinL2020Thymic inkt single cell analyses unmask the common developmental program of mouse innate T cellsNature Communications11623810.1038/s41467-020-20073-833288744HartmannFJMrdjenDMcCaffreyEGlassDRGreenwaldNFBharadwajAKhairZVerberkSGSBaranskiABaskarRGrafWVan ValenDVan den BosscheJAngeloMBendallSC2021Single-Cell metabolic profiling of human cytotoxic T cellsNature Biotechnology3918619710.1038/s41587-020-0651-832868913Huynh-ThuVAIrrthumAWehenkelLGeurtsPIsalanM2010Inferring regulatory networks from expression data using tree-based methodsPLOS ONE5e1277610.1371/journal.pone.001277620927193JadhavRRImSJHuBHashimotoMLiPLinJXLeonardWJGreenleafWJAhmedRGoronzyJJ2019Epigenetic signature of PD-1+ TCF1+ CD8 T cells that act as resource cells during chronic viral infection and respond to PD-1 blockadePNAS116141131411810.1073/pnas.190352011631227606JahngAWMaricicIPedersenBBurdinNNaidenkoOKronenbergMKoezukaYKumarV2001Activation of natural killer T cells potentiates or prevents experimental autoimmune encephalomyelitisThe Journal of Experimental Medicine1941789179910.1084/jem.194.12.178911748280KanehisaMFurumichiMSatoYIshiguro-WatanabeMTanabeMK2021Kegg: integrating viruses and cellular organismsNucleic Acids Research49D545D55110.1093/nar/gkaa97033125081KorotkevichGSukhovVBudinNShpakBArtyomovMNSergushichevA2016Fast Gene Set Enrichment AnalysisbioRxiv10.1101/060012KorsunskyIMillardNFanJSlowikowskiKZhangFWeiKBaglaenkoYBrennerMLohP-RRaychaudhuriS2019Fast, sensitive and accurate integration of single-cell data with harmonyNature Methods161289129610.1038/s41592-019-0619-031740819KovalovskyDUcheOUEladadSHobbsRMYiWAlonzoEChuaKEidsonMKimH-JImJSPandolfiPPSant’AngeloDB2008The BTB-zinc finger transcriptional regulator PLZF controls the development of invariant natural killer T cell effector functionsNature Immunology91055106410.1038/ni.164118660811KumarBVConnorsTJFarberDL2018Human T cell development, localization, and function throughout lifeImmunity4820221310.1016/j.immuni.2018.01.00729466753KumarAPyaramKYaroszELHongHLyssiotisCAGiriSChangCH2019Enhanced oxidative phosphorylation in NKT cells is essential for their survival and functionPNAS1167439744810.1073/pnas.190137611630910955KumarAHillTMGordyLESuryadevaraNWuLFlyakAIBezbradicaJSVan KaerLJoyceS2020Nur77 controls tolerance induction, terminal differentiation, and effector functions in semi-invariant natural killer T cellsPNAS117171561716510.1073/pnas.200166511732611812LaMarcheNMKaneHKohlgruberACDongHLynchLBrennerMB2020Distinct iNKT cell populations use IFNγ or ER stress-induced IL-10 to control adipose tissue homeostasisCell Metabolism3224325810.1016/j.cmet.2020.05.01732516575LawCWChenYShiWSmythGK2014Voom: precision weights unlock linear model analysis tools for RNA-seq read countsGenome Biology1511710.1186/gb-2014-15-2-r2924485249LeeYJHolzapfelKLZhuJJamesonSCHogquistKA2013Steady-State production of IL-4 modulates immunity in mouse strains and is determined by lineage diversity of iNKT cellsNature Immunology141146115410.1038/ni.273124097110LeeYJWangHStarrettGJPhuongVJamesonSCHogquistKA2015Tissue-Specific distribution of iNKT cells impacts their cytokine responseImmunity4356657810.1016/j.immuni.2015.06.02526362265LiberzonASubramanianAPinchbackRThorvaldsdóttirHTamayoPMesirovJP2011Molecular signatures database (msigdb) 3.0Bioinformatics271739174010.1093/bioinformatics/btr26021546393LiewPXLeeWYKubesP2017Inkt cells orchestrate a switch from inflammation to resolution of sterile liver injuryImmunity4775276510.1016/j.immuni.2017.09.01629045904LiuMZhaoXMaYZhouYDengMMaY2018Transcription factor c-Maf is essential for IL-10 gene expression in B cellsScandinavian Journal of Immunology88e1270110.1111/sji.1270129974486LiuXWangYLuHLiJYanXXiaoMHaoJAlekseevAKhongHChenTHuangRWuJZhaoQWuQXuSWangXJinWYuSWangYWeiLWangAZhongBNiLLiuXNurievaRYeLTianQBianX-WDongC2019Genome-Wide analysis identifies NR4A1 as a key mediator of T cell dysfunctionNature56752552910.1038/s41586-019-0979-830814730LopesNMcIntyreCMartinSRaverdeauMSumariaNKohlgruberACFialaGJAgudeloLZDyckLKaneHDouglasACunninghamSPrendevilleHLoftusRCarmodyCPierrePKellisMBrennerMArgüelloRJSilva-SantosBPenningtonDJLynchL2021Distinct metabolic programs established in the thymus control effector functions of γδ T cell subsets in tumor microenvironmentsNature Immunology2217919210.1038/s41590-020-00848-333462452LynchLO’SheaDWinterDCGeogheganJDohertyDGO’FarrellyC2009Invariant NKT cells and CD1d (+) cells amass in human omentum and are depleted in patients with cancer and obesityEuropean Journal of Immunology391893190110.1002/eji.20093934919585513LynchLMicheletXZhangSBrennanPJMosemanALesterCBesraGVomhof-DekreyEETigheMKoayH-FGodfreyDILeadbetterEASant’AngeloDBvon AndrianUBrennerMB2015Regulatory iNKT cells lack expression of the transcription factor PLZF and control the homeostasis of T (reg) cells and macrophages in adipose tissueNature Immunology16859510.1038/ni.304725436972LynchLHoganAEDuquetteDLesterCBanksALeClairKCohenDEGhoshALuBCorriganMStevanovicDMaratos-FlierEDruckerDJO’SheaDBrennerM2016Inkt cells induce FGF21 for thermogenesis and are required for maximal weight loss in GLP1 therapyCell Metabolism2451051910.1016/j.cmet.2016.08.00327593966MarchingoJMSinclairLVHowdenAJCantrellDA2020Quantitative analysis of how Myc controls T cell proteomes and metabolic pathways during T cell activationeLife9e5372510.7554/eLife.5372532022686MascanfroniIDTakenakaMCYesteAPatelBWuYKenisonJESiddiquiSBassoASOtterbeinLEPardollDMPanFPrielAClishCBRobsonSCQuintanaFJ2015Metabolic control of type 1 regulatory T cell differentiation by AhR and HIF1-αNature Medicine2163864610.1038/nm.386826005855McInnesLHealyJSaulNGroßbergerL2018UMAP: uniform manifold approximation and projectionJournal of Open Source Software386110.21105/joss.00861MotomuraYKitamuraHHijikataAMatsunagaYMatsumotoKInoueHAtarashiKHoriSWataraiHZhuJTaniguchiMKuboM2011The transcription factor E4BP4 regulates the production of IL-10 and IL-13 in CD4+ T cellsNature Immunology1245045910.1038/ni.202021460847MurrayMPEngelISeumoisGHerrera-De la MataSRosalesSLSethiALogandha Ramamoorthy PremlalASeoGYGreenbaumJVijayanandPScott-BrowneJPKronenbergM2021Transcriptome and chromatin landscape of inkt cells are shaped by subset differentiation and antigen exposureNature Communications12144610.1038/s41467-021-21574-w33664261OleinikaKRosserECMateiDENistalaKBosmaADrozdovIMauriC2018CD1d-dependent immune suppression mediated by regulatory B cells through modulations of inkt cellsNature Communications968410.1038/s41467-018-02911-y29449556OmilusikKDBestJAYuBGoossensSWeidemannANguyenJVSeuntjensEStryjewskaAZweierCRoychoudhuriRGattinoniLBirdLMHigashiYKondohHHuylebroeckDHaighJGoldrathAW2015Transcriptional repressor ZEB2 promotes terminal differentiation of CD8+ effector and memory T cell populations during infectionThe Journal of Experimental Medicine2122027203910.1084/jem.2015019426503445ParekhVVWilsonMTOlivares-VillagómezDSinghAKWuLWangC-RJoyceSVan KaerL2005Glycolipid antigen induces long-term natural killer T cell anergy in miceThe Journal of Clinical Investigation1152572258310.1172/JCI2476216138194RampuriaPLangML2015CD1d-dependent expansion of NKT follicular helper cells in vivo and in vitro is a product of cellular proliferation and differentiationInternational Immunology2725326310.1093/intimm/dxv00725710490RaudvereUKolbergLKuzminIArakTAdlerPPetersonHViloJ2019G: profiler: a web server for functional enrichment analysis and conversions of gene Lists (2019 update)Nucleic Acids Research47W191W19810.1093/nar/gkz36931066453RaynorJLLiuCDhunganaYGuyCChapmanNMShiHNealeGSesakiHChiH2020Hippo/mst signaling coordinates cellular quiescence with terminal maturation in iNKT cell development and fate decisionsThe Journal of Experimental Medicine217e2019115710.1084/jem.2019115732289155ReillyECThompsonEAAspeslaghSWandsJRElewautDBrossayLSandbergJK2012Activated iNKT cells promote memory CD8+ T cell differentiation during viral infectionPLOS ONE7e3799110.1371/journal.pone.003799122649570RobinsonMDMcCarthyDJSmythGK2010EdgeR: a bioconductor package for differential expression analysis of digital gene expression dataBioinformatics2613914010.1093/bioinformatics/btp61619910308SagDKrausePHedrickCCKronenbergMWingenderG2014Il-10-Producing NKT10 cells are a distinct regulatory invariant NKT cell subsetThe Journal of Clinical Investigation1243725374010.1172/JCI7230825061873SeoHChenJGonzález-AvalosESamaniego-CastruitaDDasAWangYHLópez-MoyadoIFGeorgesROZhangWOnoderaAWuCJLuLFHoganPGBhandoolaARaoA2019Tox and TOX2 transcription factors cooperate with NR4A transcription factors to impose CD8+ T cell exhaustionPNAS116124101241510.1073/pnas.190567511631152140ShimizuKSatoYShingaJWatanabeTEndoTAsakuraMYamasakiSKawaharaKKinjoYKitamuraHWataraiHIshiiYTsujiMTaniguchiMOharaOFujiiS2014KLRG+ invariant natural killer T cells are long-lived effectorsPNAS111124741247910.1073/pnas.140624011125118276SinghAKWilsonMTHongSOlivares-VillagómezDDuCStanicAKJoyceSSriramSKoezukaYVan KaerL2001Natural killer T cell activation protects mice against experimental autoimmune encephalomyelitisThe Journal of Experimental Medicine1941801181110.1084/jem.194.12.180111748281SmythGK2005Limma:linear models for microarray dataGentlemanRCareyVJBioinforma. Comput. Biol. Solut. Using R BioconductorSpringer39742010.1007/0-387-29362-0_23SmythMJWallaceMENuttSLYagitaHGodfreyDIHayakawaY2005Sequential activation of NKT cells and NK cells provides effective innate immunotherapy of cancerThe Journal of Experimental Medicine2011973198510.1084/jem.2004228015967825StradnerMHCheungKPLasorellaAGoldrathAWD’CruzLM2016Id2 regulates hyporesponsive invariant natural killer T cellsImmunology and Cell Biology9464064510.1038/icb.2016.1926880074SubramanianATamayoPMoothaVKMukherjeeSEbertBLGilletteMAPaulovichAPomeroySLGolubTRLanderESMesirovJP2005Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profilesPNAS102155451555010.1073/pnas.050658010216199517SunMHeCChenLYangWWuWChenFCaoATYaoSDannSMDharTGMSalter-CidLZhaoQLiuZCongY2019Rorγt represses IL-10 production in th17 cells to maintain their pathogenicity in inducing intestinal inflammationJournal of Immunology202799210.4049/jimmunol.170169730478092ThapaPMansoBChungJYRomera ArochaSXueHHAngeloDBSShapiroVS2017The differentiation of ROR-γt expressing inkt17 cells is orchestrated by RUNX1Scientific Reports7701810.1038/s41598-017-07365-828765611ThomasSYScanlonSTGriewankKGConstantinidesMGSavageAKBarrKAMengFLusterADBendelacA2011Plzf induces an intravascular surveillance program mediated by long-lived LFA-1-ICAM-1 interactionsThe Journal of Experimental Medicine2081179118810.1084/jem.2010263021624939UtzschneiderDTCharmoyMChennupatiVPousseLFerreiraDPCalderon-CopeteSDaniloMAlfeiFHofmannMWielandDPradervandSThimmeRZehnDHeldW2016T cell factor 1-expressing memory-like CD8 (+) T cells sustain the immune response to chronic viral infectionsImmunity4541542710.1016/j.immuni.2016.07.02127533016VenkenKJacquesPMortierCLabadiaMEDecruyTCoudenysJHoytKWayneALHughesRTurnerMVan GassenSMartensLSmithDHarckenCWahleJWangC-TVerheugenESchryversNVarkasGCypersHWittoekRPietteYGyselbrechtLVan CalenberghSVan den BoschFSaeysYNaboznyGElewautD2019Rorγt inhibition selectively targets IL-17 producing iNKT and γδ-T cells enriched in spondyloarthritis patientsNature Communications10910.1038/s41467-018-07911-630602780VorkasCKKrishnaCLiKAubéJFitzgeraldDWMazutisLLeslieCSGlickmanMS2020Single Cell Transcriptional Profiling Reveals Helper, Effector, and Regulatory MAIT Cell Populations Enriched during Homeostasis and ActivationbioRxiv10.1101/2020.10.22.351262VorkasCKKrishnaCLiKAubéJFitzgeraldDWMazutisLLeslieCSGlickmanMS2022Single-cell transcriptional profiling reveals signatures of helper, effector, and regulatory MAIT cells during homeostasis and activationJournal of Immunology2081042105610.4049/jimmunol.210052235149530WangHHogquistKA2018aCcr7 defines a precursor for murine inkt cells in thymus and peripheryeLife7e3479310.7554/eLife.3479330102153WangYSedimbiSLöfbomLSinghAKPorcelliSACardellSL2018bUnique invariant natural killer T cells promote intestinal polyps by suppressing th1 immunity and promoting regulatory T cellsMucosal Immunology1113114310.1038/mi.2017.3428401935WangZYinWZhuLLiJYaoYChenFSunMZhangJShenNSongYChangX2018cIron drives T helper cell pathogenicity by promoting RNA-binding protein PCBP1-mediated proinflammatory cytokine productionImmunity49809210.1016/j.immuni.2018.05.00829958803WangJLovelessIAdriantoILiuTSubediKWuXHossainMMSebzdaEZhouLMiQ-S2022Single-Cell analysis reveals differences among iNKT cells colonizing peripheral organs and identifies KLF2 as a key gene for iNKT emigrationCell Discovery87510.1038/s41421-022-00432-z35915069WilsonMTJohanssonCOlivares-VillagómezDSinghAKStanicAKWangCRJoyceSWickMJVan KaerL2003The response of natural killer T cells to glycolipid antigens is characterized by surface receptor down-modulation and expansionPNAS100109131091810.1073/pnas.183316610012960397WuRChenXKangSWangTGnanaprakasamJRYaoYLiuLFanGBurnsMRWangR2020De novo synthesis and salvage pathway coordinately regulate polyamine homeostasis and determine T cell proliferation and functionScience Advances64275429110.1126/sciadv.abc4275YanHZhangPKongXHouXZhaoLLiTYuanXFuH2017Primary tr1 cells from metastatic melanoma eliminate tumor-promoting macrophages through granzyme B- and perforin-dependent mechanismsTumour Biology39101042831769755410.1177/101042831769755428378637YuGWangLGHanYHeQY2012ClusterProfiler: an R package for comparing biological themes among gene clustersOmics1628428710.1089/omi.2011.011822455463ZhouLAdriantoIWangJWuXDattaIMiQS2020Single-cell RNA-seq analysis uncovers distinct functional human NKT cell sub-populations in peripheral bloodFrontiers in Cell and Developmental Biology838410.3389/fcell.2020.0038432528956ZuberbuehlerMKParkerMEWheatonJDEspinosaJRSalzlerHRParkECiofaniM2019The transcription factor c-Maf is essential for the commitment of IL-17-producing γδ T cellsNature Immunology20738510.1038/s41590-018-0274-03053833610.7554/eLife.76586.sa0Editor's evaluationWangChyung-RuReviewing Editorhttps://ror.org/000e0be47Northwestern UniversityUnited States
This study presents a valuable finding on transcriptional profiles of various subsets of activated invariant natural killer T (iNKT) cells using longitudinal scRNA-seq analysis. The evidence supporting the conclusions is solid with rigorous and thorough bioinformatic analyses. The work will be of interest to scientists within the field of iNKT cells.
10.7554/eLife.76586.sa1Decision letterWangChyung-RuReviewing Editorhttps://ror.org/000e0be47Northwestern UniversityUnited States
Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.
Decision letter after peer review:
Thank you for submitting your article "Longitudinal analysis of invariant natural killer T cell activation reveals a cMAF-associated transcriptional state of NKT10 cells" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Tadatsugu Taniguchi as the Senior Editor. The reviewers have opted to remain anonymous.
The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.
Reviewer #1 (Recommendations for the authors):
1. Based on scRNA-seq data, the authors should address whether TCR repertoire of iNKT cells is altered at different stages of activation (e.g., in the steady, activation, and re-activation states).
2. In Figure 5D, the authors identified two major populations of reactivated splenic iNKT cells (cluster A and cluster B). What is the interrelationship of these two clusters of iNKT cells? Does the proportion of KLRG1+ and c-MAF+ iNKT cells change at various time points after ⍺-GalCer immunization (e.g., 3 days, 7 days, 30 days, and reactivation)? Does activation with ⍺-GalCer-pulsed DC also induce these two unique iNKT cell populations?
3. In Figure 6C, it is worthwhile to determine the frequency of KLRG1+ and c-MAF+ iNKT cells in other tissues (e.g., liver, lymph nodes) at steady and resting state by flow cytometry.
4. The authors should compare the function of activated and reactivated iNKT cells in the adipose tissues.
5.The authors suggested that "NKT10/cMAF+ cells are transcriptionally similar to NKTFH cells, and these two memory-like iNKT cell populations may phenotypically and functionally overlap". Does NKT10/cMAF+ cells also express CXCR5, PD-1 and Bcl6?
6. The authors should discuss how ⍺-GalCer immunization leads to the emergence of two distinct lineages of memory-like iNKT cells. Does TCR signaling strength affect the c-MAF expression in iNKT cells?
Reviewer #2 (Recommendations for the authors):
1. Please add reference from the Bendelac lab re: parabiosis and tissue residence. J Exp Med. 2011 Jun 6;208(6):1179-88.
2. In Figure 1C, the authors noted that some activated iNKT cells express genes of the KLF2 regulon and suggest that activated iNKT cells may traffic to other tissues. This should be contrasted to the work of the Kubes and Littman labs, which showed that activated iNKT cells stop patrolling blood vessels.
3. In Figure 1D, the analysis of metabolic pathway is interesting. However, I would consider toning down some of the conclusions at the data is limited to gene expression and pathway analyses. Functional metabolic studies are necessary to draw firm conclusions.
4. The experimental setting in Figure 2H should be detailed. Were subsets sorted prior to in vitro stimulation, or were total thymic cells stimulated? Were iNKT subsets discriminated by FACS using transcription factor staining on top of cytokine antibodies. A similar experiment should be done with cells from the spleen. This experiment used Balb/c mice. Was the in vivo activation performed earlier in the manuscript performed with Balb/c or C57Bl/6 mice? The background of mice used should be clarified for all experiments. In B6 mice, iNKT1 cells are known to produce IL-4 in addition to IFN-g. Is IL-4 production by iNKT1 cells also more affected by oligomycin than their production of IFN-g?
5. The authors state: "We previously found that Nur77 is enriched in adipose iNKT cells compared to splenic iNKT cells at steady state (17), but Nur77 expression does not increase upon activation in adipose iNKT cells, unlike splenic iNKT cells. Is this based on the transcriptomic data, or has this been confirmed by FACS using Nur77 antibodies, or Nur77 reporter mice?
6. The blunted initial response observed for adipose tissue iNKT cells could rather be indicative of a different kinetic of activation between spleen and adipose tissue, due to sub-anatomical location of iNKT cells in these tissues, and accessibility to aGalCer and CD1d (see work from the Hogquist lab). A fine kinetic study of adipose tissue vs. spleen iNKT cell response may be warranted to draw this conclusion. This is in line with the statement drawn from Figure 3D-E that adipose tissue iNKT cells are more activated at baseline. Maybe they respond faster following aGalCer administration than their spleen counterparts.
7. The suggestion (from supplementary Figure 3) that iNKT10 might be functionally heterogenous is interesting and the data seems convincing. However, this could be due to subtle differences in the activation kinetic in vivo. Is there a way to fate-map some of these genes following aGalCer activation in vivo?
8. On Figure 5A, it seems to me that steady-state and resting (4 weeks post-aGalCer) iNKTs cluster together while activated and reactivated iNKTs cluster together. The "greatly reduced response" and the similarity with activated adipose tissue iNKTs is no obvious from this UMAP data. The authors should clarify this.
9. In Figure 6A, please clearly describe in the text what markers were used to define iNKT subsets among steady-state and resting I NKT cells.
10. How do the clusters A-C described in Figure 6G compare with effector subsets described in Figure 4?
11. "Flow cytometry of splenic iNKT cells 4 weeks post-⍺GalCer confirmed that IL-10+ iNKT cells expressed cMAF, and cMAFneg cells produced little IL-10 (Figure 7F, Figure 7G)." These panels do not appear to show FACS data. Does this relate to panel D? It would be better to show IL-10 vs. cMAF from CD1d tetramer-positive cells on these plots rather than CD1d tetramer staining. The figure should also use PMA/Iono stimulation of steady-state iNKT cells from the same tissue as a control.
Reviewer #3 (Recommendations for the authors):
This paper is of interest to scientists within the field of iNKT cells. The authors conducted scRNA-seq to longitudinally profile activated iNKT cells and generated a transcriptomic atlas of iNKT cells at the activation states. The study suggests that transcriptional signatures of activation are highly conserved among heterogeneous iNKT cell populations and that the adipose iNKT cells undergo blunted activation and display constitutive enrichment of memory like population. In addition, the study also identifies a conserved cMAF- associated network in NKT10 cells. The data quality is relatively high.
1. Based on the violin plots shown in Figure 2B, in the 4 hours aGalCer stimulation group, it seems that Tbx21 and Zbtb16 are evenly spaced in all three clusters, where it is difficult to identify NKT1, NKT2 in this case. It would be clearer if the author could show NKT1/2/17 signatures expression pattern in umap plot.
2. The statement " Notably, we found that all subsets upregulated Zbtb16 and Tbx21 after activation (Figure 2B), suggesting that PLZF and T-bet play subset-independent roles during activation." This needs more evidence. The gene expression level (such as Tbx21, Zbtb16) increased post aGalCer activation could be due to the transcriptional bursting, author should show and compare the house keeping genes expression level in both steady state and aGalCer activation state.
3. In Figure 2D, using correlation heatmaps the authors mapped the gene profiles in different subsets of NKT cells before and after aGalCer stimulation, and claimed that "activated NKT2 and NKT17 cells are transcriptionally similar compared to NKT1 cells." The authors should consider that post aGalCer stimulation, proliferating genes/cells will be predominant. The similarity of NKT2 and NKT17 observed in the heatmap could be due to the overwhelming cell cycle genes, instead of the intrinsic genes of NKT2 and NKT17. The authors should remove cell cycle gene, and then do the mapping.
4. The study claimed that NKT10 and Tr1 share transcriptional features. It will make this study more interesting if the author could employ the experiments to check on their function potential.
5. Figure 2F only shows the representation of the experiment, statistics summary bar graphs should also be included.
6. Figure 3B is generated from bulk analysis of scRNA-Seq data, right? What about by Stages (4 and 72h) subclusters?
7. "Analysis of cytokine production among adipose iNKT cells revealed that IL-10 was only expressed by NKT1 cells (Figure 4B/D)." Better have flow data to support this claim. No scale stick in Y axis in Figure 4C.
8. Clusters showing in Figure 5F with NKT 10 cells are better validated in protein level by FACS.
9. Clusters showing in Figure 6G should be validated in protein level.
10. The author did not mention how many biological replicates for scRNA-seq experiment, since mice treated with activator could vary widely between mouse to mouse, at least 2 biological replicates are needed.
11. The author mentioned in the method that both C57BL6 mice and BALB/c mice were used in this study, but did not mention which experiments were performed using C57BL6 mice or BALB/c. NKT cell subsets are different between C57BL6 and BALB/C mice. Please discuss the rationale why these two strains were used and make them clearer.
10.7554/eLife.76586.sa2Author response
Reviewer #1 (Recommendations for the authors):
1. Based on scRNA-seq data, the authors should address whether TCR repertoire of iNKT cells is altered at different stages of activation (e.g., in the steady, activation, and re-activation states).
The idea of checking if and how the TCR repertoire of iNKT cells changes during activation is an excellent one, and we thank the reviewer for this recommendation. However, from a bioinformatic perspective, robust TCR repertoire analysis using scRNA-Seq typically requires that TCR sequencing be performed in combination with standard scRNA-Seq at the time of data generation. Since TCR repertoire analysis was not a specific aim of this study we did not perform TCR-Seq at the time of sequencing. Also, given the restricted nature of the iNKT cell TCR, we didn’t feel that it was a key factor when designing the experiment initially. Therefore, we respectfully suggest that scRNA-Seq-based analysis of iNKT cell TCR repertoires is not within the scope of this study.
2. In Figure 5D, the authors identified two major populations of reactivated splenic iNKT cells (cluster A and cluster B). What is the interrelationship of these two clusters of iNKT cells? Does the proportion of KLRG1+ and c-MAF+ iNKT cells change at various time points after ⍺-GalCer immunization (e.g., 3 days, 7 days, 30 days, and reactivation)? Does activation with ⍺-GalCer-pulsed DC also induce these two unique iNKT cell populations?
We thank the reviewer for these questions, which we also found to be interesting. To answer these questions, we performed new experiments by flow cytometry.
In Figure 5D of our original manuscript we identified two main clusters or populations of reactivated splenic cells, Cluster A and Cluster B. The main difference between these two clusters mainly was their differential expression of activation markers, cytokine genes, and transcripts associated with metabolic remodeling and T cell metabolic activation (Figure 5D5E, Figure 5—figure supplement 3). Cluster A iNKT cells (~50% of reactivated iNKT cells) demonstrated low expression of activation-associated genes compared to Cluster B iNKT cells (~50% of reactivated iNKT cells; Figure 5D-5E, Figure 5—figure supplement 3). The expression of activation-associated genes in Cluster A iNKT cells was comparable to that of resting iNKT cells (gray histogram) which were not reactivated with αGalCer (Figure 5figure supplement 3). We also performed subclustering analysis of the iNKT cells in Cluster A and Cluster B. Within Cluster B we identified distinct populations of iNKT cells expressing genes associated with NKT1 cells (IFNγ and Il4), NKT17 cells (Il17a), cMAF+ iNKT cells (Maf, Il10) and KLRG1+ iNKT cells (Klrg1, Gzma) (Figure 5E, Figure 5—figure supplement 3). This indicated that Cluster B is comprised of a heterogeneous mix of different iNKT cell populations. Interestingly, we also identified distinct iNKT cell populations expressing Klrg1 or Maf within Cluster A (Figure 5E, Figure 5—figure supplement 3), which suggested to us that prior activation of iNKT cells with αGalCer might induce the differentiation or development of memory-like or “trained” KLRG1+ and cMAF+ iNKT cell populations. In the absence of any evidence to suggest that Cluster A iNKT cells are significantly different to Cluster B iNKT cells in any respect other than in their expression of activation markers and cytokines, we hypothesize that the cells in Cluster A would eventually assume a more activated transcriptional profile, similar to the cells in Cluster B.
To improve our data presentation, and to comply with reviewer recommendations, we have reworked Figure 5 in our revised manuscript. We have replaced the previously separate density plots of gene expression in Cluster A and Cluster B cells (Figure 5F and 5G in the original manuscript) with density plots of gene expression and module scores for all reactivated iNKT cells in a single UMAP (Figure 5E). We believe that this updated plot provides a more direct and improved visualization of the data. The histograms and density plots from Figure 5E, 5F and 5G of the original manuscript are now presented in Figure 5figure supplement 3. We also present new flow cytometry data (Figure 5F, Figure 5G, Figure 5—figure supplement 4) to support our gene expression analysis, focusing on our identification of co-expression of Klrg1 and Gzma, and Maf and Il10 (Figure 5E, Figure 5—figure supplement 4). In our revised Figure 5F we identify a significant increase in the frequency of Granzyme A+ KLRG1+ iNKT cells among reactivated versus activated splenic iNKT cells (Figure 5F). Similarly, we show that there is significantly increased production of IL-10 in cMAF+ versus cMAF- resting splenic iNKT cells (Figure 5G) and significantly increased expression of cMAF among resting splenic iNKT cells (Figure 5—figure supplement 4).
The frequency of KLRG1+ iNKT cells after αGalCer stimulation has previously been assayed by Shimzu et al. (2014) and Murray et al. (2021), demonstrating that KLRG1 expression peaks between 5-7 days post-αGalCer and thereafter declines, although both our study and previous studies have demonstrated that KLRG1+ iNKT cells can persist for weeks to months after αGalCer stimulation1,2. Shimizu et al. (2014) identified KLRG1+ iNKT cells after immunization of mice with αGalCer-pulsed dendritic cells, indicating that memory-like or “trained” iNKT cell populations can be induced using this method of immunization. In our study we identified increased expression of Klrg1 among iNKT cells at 72 hours post-αGalCer compared to 4 hours post-αGalCer or steady state (Figure 1, Figure 1—figure supplement 3), consistent with previous data indicating that the frequency of KLRG1+ iNKT cells peaks several days post-αGalCer. We also found that expression of cMAF was significantly increased among iNKT cells at 72 hours post-αGalCer (Figure 1, Figure 1—figure supplement 3), suggesting that cMAF+ iNKT cells may also be induced at this time point. To validate this observation, we analyzed expression of cMAF among iNKT cells at 72 hours post-αGalCer using flow cytometry. Interestingly, we found that ~50% of iNKT cells expressed cMAF (Figure 1—figure supplement 3), a larger proportion of cMAF+ iNKT cells than we had identified at 4 weeks post-αGalCer (Figure 6D and Figure 6E). Interestingly, we also noted that this high percentage of cMAF+ iNKT cells at 72 hour post-αGalCer correlated with increased production of IL-10 after restimulation of iNKT cells at 72 hours post-αGalCer compared to primary stimulation of steady state iNKT cells (Figure 1—figure supplement 3). Therefore, our study demonstrates, in conjunction with published data, that the frequency of cMAF+ and KLRG1+ iNKT cells changes at different time points following αGalCer immunization, peaking several days post αGalCer, but remaining albeit at lower levels in some iNKT cells at 4 weeks post αGalCer.
3. In Figure 6C, it is worthwhile to determine the frequency of KLRG1+ and c-MAF+ iNKT cells in other tissues (e.g., liver, lymph nodes) at steady and resting state by flow cytometry.
We agree and have performed a new experiment to examine the frequency of KLRG1+ and cMAF+ iNKT cells at steady state and 4 weeks after αGalCer in different tissues using flow cytometry. We found that KLRG1+ and cMAF+ iNKT cells are present at low levels at steady state and enriched 4 weeks after αGalCer (Figure 6D, Figure 6E), in all organs tested including spleen, lung, liver, adipose tissue, and inguinal lymph nodes. KLRG1 and cMAF expression was mutually exclusive across the organs (Figure 6D-6F). Interestingly, we found that KLRG1+ iNKT cells, but not cMAF+ iNKT cells, were specifically enriched in the lung 4 weeks post-αGalCer (Figure 6D-6E), correlating with previous work by Shimzu et al. (2014) which demonstrated that KLRG1+ iNKT cells are strongly induced in the lung after immunization of mice with αGalCer-pulsed dendritic cells.
4. The authors should compare the function of activated and reactivated iNKT cells in the adipose tissues.
We thank the reviewer for this recommendation. We have previously profiled the function of activated adipose iNKT cells in detail (Figure 2 of LaMarche et al. (2020), DOI: https://doi.org/10.1016/j.cmet.2020.05.017)3 and here we add to that with new data on the function of reactivated adipose iNKT cells in response to the reviewer’s question. Restimulated adipose iNKT cells at 4 weeks post-αGalCer demonstrated mutually exclusive production of Granzyme A and IL-10 (Figure 5—figure supplement 4), and expression of Granzyme A was mainly associated with KLRG1+ iNKT cells (Figure 5—figure supplement 4). We also identified a similar expression pattern in splenic iNKT cells (Figure 5E). Granzyme A has previously been shown to be a flagship secreted factor produced by KLRG1+ iNKT cells1 and here we demonstrate that production of IL-10 is enriched among cMAF+ iNKT cells at 4 weeks post-αGalCer (Figure 5F). Therefore, our functional analysis of Granzyme A and IL-10 production supports our KLRG1 and cMAF expression data in the adipose tissue and the spleen (Figure 6D-6F).
Interestingly, we found that most Granzyme A+ iNKT cells and IL-10+ iNKT cells produced IFNγ (Figure 6—figure supplement 3). Similarly, we found that IL-17A and IL-10 expression was mutually exclusive in the spleen at 4 weeks post-αGalCer. We did not detect robust production of IL-17A in the adipose tissue at 4 weeks post-αGalCer, which matched our scRNA-Seq data demonstrating that there are very few RORyT+ NKT17 cells in the adipose tissue at 4 weeks post-αGalCer (Figure 6—figure supplement 5). This matches previous functional data on induced NKT10 cells published by Sag et al. (2014)4 and suggests that KLRG1+ iNKT cells and iNKT10 cells are NKT1-like populations (or were previously NKT1 cells). Surprisingly, we did not detect robust production of IL-21 among adipose or splenic iNKT cells at 4 weeks post-αGalCer, even though Il21 was one of the most differentially expressed genes among reactivated splenic iNKT cells and NKT10 cells by scRNA-Seq and in published microarray data from Sag et al. (2014)4 (Figure 5C, Figure 7A-7B, Figure 4—figure supplement 1, Figure 5—figure supplement 1, Figure 7—figure supplement 1). This could suggest that IL-21 is only expressed at the protein level under different activation circumstances. Further research will be required to elucidate if and when IL-21 might be expressed by reactivated iNKT cells or NKT10 cells.
5. The authors suggested that "NKT10/cMAF+ cells are transcriptionally similar to NKTFH cells, and these two memory-like iNKT cell populations may phenotypically and functionally overlap". Does NKT10/cMAF+ cells also express CXCR5, PD-1 and Bcl6?
We thank the reviewer for this important insight which we have now examined. Investigating the relationship between NKTFH cells and cMAF+ iNKT cells, we found a minority of splenic iNKT cells co-expressing the NKTFH cell markers CXCR5 and PD-1 at 4 weeks post-αGalCer (Figure 7F, Figure 7G; Resting; ~6% of iNKT cells on average). The average percentage of CXCR5+ PD-1+ NKTFH cells was much lower than the percentage of cMAF+ iNKT cells in the same gate (Figure 6D, Figure 6F, ~28% of iNKT cells on average), which indicated that most cMAF+ iNKT cells were unlikely to be NKTFH cells based on expression of CXCR5 and PD-1. Gating on cMAF+ iNKT cells, a minority (~7%) coexpressed CXCR5 and PD-1 (Figure 7F, Figure 7G), although this was a slight increase relative to total resting iNKT cells (Figure 7G, a ~1.2 fold increase in the frequency of NKTFH cells among cMAF+ iNKT cells versus total resting iNKT cells). We did not find a significant increase in BCL6 expression among cMAF+ iNKT cells versus total resting iNKT cells (Figure 7—figure supplement 1), matching our scRNA-Seq where Bcl6 was not significantly enriched among cMAF+ iNKT cells (Figure 7—figure supplement 1). This suggests that there is some overlap between NKTFH cells and cMAF+ iNKT cells, and that NKTFH cells are slightly enriched among cMAF+ iNKT cells, but most cMAF+ iNKT cells are not bone fide NKTFH cells. At the gene level, however, cMAF+ iNKT cells express many of the same genes as NKTFH cells (Figure 7D, Figure 7E), suggesting that these two populations are transcriptionally similar. This result and discussion are presented in the revised manuscript.
6. The authors should discuss how ⍺-GalCer immunization leads to the emergence of two distinct lineages of memory-like iNKT cells. Does TCR signaling strength affect the c-MAF expression in iNKT cells?
We thank the reviewer for their recommendation. We have added to the discussion to address how αGalCer leads to the emergence of two distinct lineages of memory-like iNKT cells. In brief, we note that KLRG1+ iNKT cells display enrichment of markers associated with NK cells, cytotoxic T cells and effector memory CD8+ T cells, including Klrg1, Gzma, Klf2, Gzmb, and Zeb25–7 (Figure 6B). This similarity of KLRG1+ iNKT cells to cytotoxic T cells and effector memory CD8+ T cells has previously been noted by Shimizu et al. (2014)1, and a similar comparison was also made between KLRG1-like nonNKTFH cells (termed NKTeff cells) and effector memory CD8+ T cells/cytotoxic T cells by Murray et al. (2021)2. We also note that the transcriptional signature/phenotype of KLRG1+ iNKT cells is similar to that of short lived effector CD8+ T cells (SLECs) associated with the SLEC/memory precursor effector cell (MPEC) model described in numerous studies6–8. By contrast, we found that cMAF+ iNKT cells and NKT10 cells display a gene signature more similar to CD4+ memory T cells, Tregs, precursor exhausted T (TPEX) cells and MPECs, including Maf, Cd4, Il10, Tox, Cxcr3, Slamf6 and Izumo1r7–10 (Figure 6B, Figure 7A, Figure 7B). Therefore, we propose that there may be a bifurcating differentiation branch during the formation of memory-like iNKT cell subsets following immunization with αGalCer, whereby a CD8+ effector memory-like KLRG1+ iNKT cell population differentiates along one trajectory and a CD4+ memory-like T cell cMAF+ iNKT cell population differentiates along another trajectory. Furthermore, given the transcriptional similarity of NKTFH cells and cMAF+ iNTK cells, it is possible that these two populations might share a similar differentiation trajectory. Mechanistically, it is possible that some of the same transcription factors that control the differentiation of CD8+ vs CD4+ memory may also regulate the development of different memory-like iNKT cell lineages. For example, in our study we identified enrichment of the Zeb2 transcription factor in KLRG1+ iNKT cells (Figure 6B), which is known to be a key regulator of the terminal differentiation of memory CD8+ effector T cells5,11. Further research will be required to elucidate the dynamics of memory-like iNKT cell differentiation in more detail.
The question of whether TCR signal strength controls cMAF expression in iNKT cells is a fascinating one. We present some new data suggesting that increased antigen load might regulate cMAF expression under some circumstances. We performed immunization of mice with 1μg, 2.5μg or 5μg of αGalCer, which showed a dose dependent trend whereby there was an increased frequency of cMAF+ iNKT cells in the spleen in response to increased antigen load (Figure 6—figure supplement 2). Moreover, immuniztion of mice with 10μg of αGalCer resulted in an increased frequency of cMAF+ iNKT cells in the spleen (~37%) versus 4μg of αGalCer (29%), which was the dose that we had used throughout the study (Figure 5—figure supplement 4). These data suggest that cMAF expression is increased in response to increasing TCR signaling strength. Notably, expression of cMAF has previously been linked to TCR signal strength in γδ T cells12.
Reviewer #2 (Recommendations for the authors):
1. Please add reference from the Bendelac lab re: parabiosis and tissue residence. J Exp Med. 2011 Jun 6;208(6):1179-88.
We thank the reviewer for this recommendation, and have added the reference to the manuscript.
2. In Figure 1C, the authors noted that some activated iNKT cells express genes of the KLF2 regulon and suggest that activated iNKT cells may traffic to other tissues. This should be contrasted to the work of the Kubes and Littman labs, which showed that activated iNKT cells stop patrolling blood vessels.
We thank the reviewer for highlighting the potential role of KLF2 in activated iNKT cells. We have added some discussion about the potential role of KLF2 in the Results section 1 (“iNKT cells undergo rapid and extensive transcriptional remodeling in response to αGalCer”) and contrasted our interpretation of increased KLF2 expression in activated iNKT cells versus previous work from the Kubes and Littman labs.
3. In Figure 1D, the analysis of metabolic pathway is interesting. However, I would consider toning down some of the conclusions at the data is limited to gene expression and pathway analyses. Functional metabolic studies are necessary to draw firm conclusions.
We have adjusted the language used in our manuscript and toned down our conclusions from these data, emphasizing that our analysis in Figure 1D is at the gene/pathway level.
4. The experimental setting in Figure 2H should be detailed. Were subsets sorted prior to in vitro stimulation, or were total thymic cells stimulated? Were iNKT subsets discriminated by FACS using transcription factor staining on top of cytokine antibodies. A similar experiment should be done with cells from the spleen. This experiment used Balb/c mice. Was the in vivo activation performed earlier in the manuscript performed with Balb/c or C57Bl/6 mice? The background of mice used should be clarified for all experiments. In B6 mice, iNKT1 cells are known to produce IL-4 in addition to IFN-g. Is IL-4 production by iNKT1 cells also more affected by oligomycin than their production of IFN-g?
We thank the reviewer for highlighting this so that we can further explain. The experimental system used for Figure 2H has previously been published and described by the Godfrey lab (please see Cameron et al. (2018), DOI: https://doi.org/10.1111/imcb.12034)15 as an effective model system for the study of NKT1, NKT2, and NKT17 cell subsets without the need to use transcription factors, so that later work can be performed on live cells. For example, mitochondrial staining with Mitotracker Green FM and TMRM is not possible in cells that have been fixed and permeabilized using transcription factor staining, but works well in live unfixed cells. NKT2 and NKT17 cells are also enriched in BALB/c thymus compared to many other organs in BALB/c or B6 mice17,18, which is also advantageous for the study of these typically rare iNKT subsets. For this reason, we employed the use of this model system to study NKT1, NKT2, and NKT17 cell function. However, we found that this staining method only works for BALB/c mice and is not clearly distinct enough in B6 mice. Briefly, functionally mature NKT1, NKT2, and NKT17 cells can be identified in the thymus of BALB/c mice using the surface markers ICOS and the activationassociated glycoform of CD43 (CD43-HG). NKT17 cells are CD43-HGhigh, NKT2 cells are ICOS+, and NKT1 cells are ICOS-. CD44 was additionally used as a marker to exclude functionally immature developing iNKT cells. We confirmed the identify of the NKT1, NKT2 and NKT17 cell subsets gated using this surface marker system using the transcription factor markers T-bet, PLZF and RORγT by flow cytometry (Figure 2—figure supplement 4). To clarify the background of mice used for all experiments in our study: BALB/c mice were only used to generate the data from Figures 2F, 2G and 2H, and Figure 2—figure supplement 4. All other experiments in our study, including scRNA-Seq experiments and in vivo stimulation experiments, were performed using C57BL/6 mice. BALB/c mice were only used for the purpose of studying NKT1, NKT2 and NKT17 cell subsets using the model system published by Cameron et al. (2018)15. To clarify the experimental procedure for the data from Figure 2H:
Total BALB/c thymocytes were stimulated with PMA (50ng/mL) and Ionomycin (1μg/mL) in complete RPMI media in the presence or absence of 40nM Oligomycin for 4 hours ex vivo before being fixed and stained for the presence of cytokines.
We have not had success using the Cameron et al. (2018) model15 with splenic iNKT cells, and therefore had to measure total cytokine production by iNKT cells without being able to gate on specific iNKT cell subsets when examining the effect of oligomycin on cytokine production by iNKT cells in the spleen. We found that incubation of total splenic iNKT cells from C57BL/6 mice with 20nM oligomycin reduced the frequency of IL-4pos IFNγneg iNKT cells and IL-4pos IFNγpos iNKT cells, but not IL-4neg IFNγpos iNKT cells after PMA and Ionomycin stimulation (Figure 2—figure supplement 5). We did not detect a sufficiently robust IL-17A production to determine if IL-17A expression was inhibited by oligomycin in the spleen. These data suggest that oxidative metabolism is critically important for the production of IL-4 by splenic iNKT cells, including NKT1 cells co-producing IL-4 and IFNγ, but not for production of IFNγ alone by NKT1 cells. Collectively, our functional data from the thymus and spleen suggest that oxidative metabolism may be critically important for production of Th2 and Th17 cell cytokines by iNKT cells, but less so for production of the NKT1 cell cytokine IFNγ. We also performed an experiment with Mitotracker Green FM and TMRM staining in the spleen. We identified an increased frequency of NK1.1- cells exhibiting higher Mitotracker Green FM and TMRM staining compared to NK1.1+ cells among splenic iNKT cells from C57BL/6 mice (Figure 5—figure supplement 4), indicative of increased mitochondrial mass and membrane potential in NK1.1- versus NK1.1+ splenic iNKT cells. Since expression of NK1.1 is greatly enriched among NKT1 cells versus NKT2 and NKT17 cells18,19, our data suggest, as with our thymic data from BALB/c mice, that NKT2 and NKT17 cells are more biased towards mitochondrial/oxidative metabolism than NKT1 cells.
5. The authors state: "We previously found that Nur77 is enriched in adipose iNKT cells compared to splenic iNKT cells at steady state (17), but Nur77 expression does not increase upon activation in adipose iNKT cells, unlike splenic iNKT cells. Is this based on the transcriptomic data, or has this been confirmed by FACS using Nur77 antibodies, or Nur77 reporter mice?
The statement “We previously found that Nur77 is enriched in adipose iNKT cells compared to splenic iNKT cells at steady state” is based on transcriptomic microarray analysis in Lynch et al. (2015)16, flow cytometry validation of Nur77 protein expression in our recent study LaMarche et al. (2020)3 and transcriptomic data from our manuscript. Please see Figures S2C and S2D from LaMarche et al. (2020), DOI: 10.1016/j.cmet.2020.05.017, which show increased expression of Nur77 in adipose versus splenic iNKT cells at steady state. To better illustrate this in our manuscript, we have added a citation of the LaMarche et al. (2021) study to the statement in question.
The statement “but Nur77 expression does not increase upon activation in adipose iNKT cells, unlike splenic iNKT cells” is actually slightly incorrect and is a mistake carried over from an earlier draft of the manuscript. We thank the reviewer for helping us spot this error so that it can be corrected. Nr4a1 expression is increased in a minority of adipose iNKT cells at 4 hours post-αGalCer. This is shown in Figure 4A and Figure 4B of the manuscript, where Nr4a1 expression is increased in a minority of adipose NKT1 and NKT17 cells at 4 hours post-αGalCer. However, it is correct to state that Nr4a1 expression is not increased upon activation in most adipose iNKT cells (at 4 hours post-αGalCer). We also show in Figure 3C that the magnitude of increase in Nr4a1 expression is significantly reduced in adipose iNKT cells versus splenic iNKT cells upon activation. Therefore, we have amended the statement highlighted by the reviewer, and the entire sentence now reads:
“We previously found that Nur77 is enriched in adipose iNKT cells compared to splenic iNKT cells at steady state17,19, but these data suggest that adipose iNKT cells demonstrate reduced upregulation of Nur77 compared to splenic iNKT cells upon activation.”.
6. The blunted initial response observed for adipose tissue iNKT cells could rather be indicative of a different kinetic of activation between spleen and adipose tissue, due to sub-anatomical location of iNKT cells in these tissues, and accessibility to aGalCer and CD1d (see work from the Hogquist lab). A fine kinetic study of adipose tissue vs. spleen iNKT cell response may be warranted to draw this conclusion. This is in line with the statement drawn from Figure 3D-E that adipose tissue iNKT cells are more activated at baseline. Maybe they respond faster following aGalCer administration than their spleen counterparts.
We thank the reviewer for this recommendation. We are aware of the interesting work from the Hogquist lab and this is an interesting point. We did not rule out in this study that the sub-anatomical location of adipose tissue might have an influence on the activation kinetic of adipose iNKT cells in response to αGalCer. However, we have previously shown that adipose iNKT cells also display reduced production of IFNγ, a key flagship iNKT cell cytokine, compared to splenic iNKT cells following in vitro activation with αGalCer; please see Figure 1D of Lynch et al. (2012). Therefore, we do not believe that this blunted or delayed activation phenotype that we observe in vivo is an artifact of differential activation due to the location of adipose tissue, but rather that is indicative of a different activation phenotype between adipose and splenic iNKT cells. To further support our argument, we show that adipose iNKT cells are still actively expressing cytokine and activation marker transcripts at 72 hours post-αGalCer, whereas splenic iNKT cells have mostly ceased expressing these transcripts by 72 hours post-αGalCer (Figure 3—figure supplement 1).
7. The suggestion (from supplementary Figure 3) that iNKT10 might be functionally heterogenous is interesting and the data seems convincing. However, this could be due to subtle differences in the activation kinetic in vivo. Is there a way to fate-map some of these genes following aGalCer activation in vivo?
We thank the reviewer for this recommendation. To our knowledge it would exceedingly difficult to fate-map most of the genes from Supplementary Figure 3B in the original manuscript (now Figure S in the revised manuscript) in vivo following αGalCer treatment using something like a reporter mouse model, as this would likely require the generation of several new bespoke mouse models and likely take years. We agree that this is a cool suggestion but is outside of the scope of the current study.
8. On Figure 5A, it seems to me that steady-state and resting (4 weeks post-aGalCer) iNKTs cluster together while activated and reactivated iNKTs cluster together. The "greatly reduced response" and the similarity with activated adipose tissue iNKTs is no obvious from this UMAP data. The authors should clarify this.
We thank the reviewer for this helpful comment that we now realize needs clarification. The phrase “greatly reduced response” is in reference to Figure 5B, which shows greatly reduced expression of activation and cytokine transcripts and reduced remodeling of metabolic genes among reactivated splenic iNKT cells compared to activated splenic iNKT cells. We agree that this reduced response is not as obvious from the UMAP in Figure 5A as it is with adipose iNKT cells from the UMAP in Figure 3A. Therefore, we have made two modifications to the manuscript. We have modified the UMAP in Figure 5A by splitting the data in steady state and activated (left panel) and resting and reactivated (right panel) so that the increased overlap between resting and reactivated iNKT cells is clearer. We believe that the reduced response is shown at the gene level in Figure 5B, but the difference between activated and reactivated splenic iNKT cells is not as obvious in the UMAP as it is with adipose, we have modified our language from “greatly reduced response” to “reduced response” to be more accurate.
9. In Figure 6A, please clearly describe in the text what markers were used to define iNKT subsets among steady-state and resting I NKT cells.
We thank the reviewer for this recommendation. We have reworked the Results text relating to Figure 6A to better describe key genes and markers used to demarcate the iNKT cell subsets identified among steady state and resting splenic iNKT cells in Figure 6A. We have also added three supplemental figures (Figure 2figure supplement 1, Figure 2—figure supplement 2, Figure 6—figure supplement 1) which help describe how we identified and defined iNKT cell subsets among steady state and resting iNKT cells. These new supplemental figures are referenced in the text describing the identification of iNKT cell subsets for Figure 6A.
10. How do the clusters A-C described in Figure 6G compare with effector subsets described in Figure 4?
The cells in Figure 6G of the original manuscript are NK1.1- adipose NKT1 cells at steady state. In Figure 4 we described a range of different clusters of adipose iNKT cells at steady state and at 4 hours post-αGalCer. The cells in Figure 6G of the original manuscript (now Figure 6—figure supplement 5 of the revised manuscript) are the same cells as the cells in Cluster 2 in Figure 4A (steady state adipose NK1.1- NKT1 cells).
11. "Flow cytometry of splenic iNKT cells 4 weeks post-⍺GalCer confirmed that IL-10+ iNKT cells expressed cMAF, and cMAFneg cells produced little IL-10 (Figure 7F, Figure 7G)." These panels do not appear to show FACS data. Does this relate to panel D? It would be better to show IL-10 vs. cMAF from CD1d tetramer-positive cells on these plots rather than CD1d tetramer staining. The figure should also use PMA/Iono stimulation of steady-state iNKT cells from the same tissue as a control.
We thank the reviewer for pointing out this typographical error – the line "Flow cytometry of splenic iNKT cells 4 weeks post-αGalCer confirmed that IL-10+ iNKT cells expressed cMAF, and cMAFneg cells produced little IL-10 (Figure 7F, Figure 7G)." should have instead referred to Figure 7D and Figure 7E from the original manuscript, which referenced the flow cytometry data. This is now corrected. We also thank the reviewer for the recommendation about how to display the flow data in Figure 7D of the original manuscript. For this analysis we pre-gated on cMAF+ and cMAF- iNKT cells before gating for expression of IL-10, in order to show that cMAF+ iNKT cells produce more IL-10 than cMAF- iNKT cells. We have, however, amended the panel to show the data in the format suggested by the reviewer. In addition, we have also added a panel of steady state iNKT cells from the same tissue (spleen). Please note – these revised flow cytometry data showing expression of IL-10 versus cMAF have been moved to from Figure 7 to Figure 5F of the revised manuscript, based on a request from reviewer #3.
Reviewer #3 (Recommendations for the authors):
This paper is of interest to scientists within the field of iNKT cells. The authors conducted scRNA-seq to longitudinally profile activated iNKT cells and generated a transcriptomic atlas of iNKT cells at the activation states. The study suggests that transcriptional signatures of activation are highly conserved among heterogeneous iNKT cell populations and that the adipose iNKT cells undergo blunted activation and display constitutive enrichment of memory like population. In addition, the study also identifies a conserved cMAF- associated network in NKT10 cells. The data quality is relatively high.
1. Based on the violin plots shown in Figure 2B, in the 4 hours aGalCer stimulation group, it seems that Tbx21 and Zbtb16 are evenly spaced in all three clusters, where it is difficult to identify NKT1, NKT2 in this case. It would be clearer if the author could show NKT1/2/17 signatures expression pattern in umap plot.
We thank the reviewer for this recommendation. This violin plot is correct, as we found that expression of Tbx21 (T-bet) and Zbtb16 (PLZF) did not effectively demarcate NKT1 and NKT2 cells at the RNA level after αGalCer, unlike at steady state, due to activation induced changes (see the left-hand side of Figure 2B). Therefore, we instead used flagship NKT1, NKT2, and NKT17 cell cytokines (IFNγ, Il4, Il13, Il17a and Il17f) to demarcate NKT1, NKT2, and NKT17 cell subsets. We have prepared an additional two supplemental figures (Figure 2—figure supplement 1 and Figure 2—figure supplement 2) and added some text to the Methods (see ”Downstream scRNA-Seq data analysis”) which should better clarify how we identified NKT1, NKT2 and NKT17 cells. These new supplemental figures also include UMAP plots of NKT1, NKT2 and NKT17 cell signature gene expression in accordance with the useful recommendation from the reviewer.
2. The statement " Notably, we found that all subsets upregulated Zbtb16 and Tbx21 after activation (Figure 2B), suggesting that PLZF and T-bet play subset-independent roles during activation." This needs more evidence. The gene expression level (such as Tbx21, Zbtb16) increased post aGalCer activation could be due to the transcriptional bursting, author should show and compare the house keeping genes expression level in both steady state and aGalCer activation state.
We thank the reviewer for this comment and recommendation. Increased PLZF expression after iNKT cell activation has previously been shown by Oleinka et al. (2018)20, (DOI:10.1038/s41467-018-02911-y), which is in line with what we found. However, increased expression of T-bet (Tbx21) has not previously been demonstrated among activated iNKT cells. We agree with the reviewer that more evidence is needed to show subset independent roles and so we have amended the referenced statement in our manuscript to “Notably, we found that all subsets upregulated Zbtb16 after activation (Figure 2B), suggesting that PLZF may play a subset-independent role during activation.”. For Tbx21, we clarify instead that Tbx21 expression is non-specifically increased across all iNKT cell subsets after activation at the RNA level, and for this reason we did not use Tbx21 as a marker gene for iNKT cell subset identification after activation.
3. In Figure 2D, using correlation heatmaps the authors mapped the gene profiles in different subsets of NKT cells before and after aGalCer stimulation, and claimed that "activated NKT2 and NKT17 cells are transcriptionally similar compared to NKT1 cells." The authors should consider that post aGalCer stimulation, proliferating genes/cells will be predominant. The similarity of NKT2 and NKT17 observed in the heatmap could be due to the overwhelming cell cycle genes, instead of the intrinsic genes of NKT2 and NKT17. The authors should remove cell cycle gene, and then do the mapping.
We thank the reviewer for highlighting this point. We demonstrate that cell cycle and proliferation genes are not strongly upregulated by 4 hours post-αGalCer (see Figure 1C, Figure 3D, and a new Figure 1—figure supplement 1). There is a strong upregulation of cell cycle and proliferation genes by 72 hours post-αGalCer (Figure 1, Figure 3), and as the reviewer points out this could dominate the any results including data from 72 hours post-αGalCer. For that reason we had performed cell cycle regression in all of our cross-analysis of scRNA-Seq data from 72 hours post-αGalCer with scRNA-Seq data from different time points, such as steady state or 4 hours post-αGalCer (see Figure 1 and Figure 3 for example). Therefore, we believe that it is unnecessary to perform cell cycle regression for the correlation analysis of iNKT cells at 4 hours post-αGalCer in Figure 2D, and we believed that doing so might lead to artifacts from an unnecessary gene regression being applied, similar to how performing batch correction on a scRNA-Seq dataset without a batch effect can lead to strange artifacts appearing in the data.
4. The study claimed that NKT10 and Tr1 share transcriptional features. It will make this study more interesting if the author could employ the experiments to check on their function potential.
We agree that this is a very interesting idea and could make for a cool follow-up study. However, in our study we only referenced Tr1 cells with respect to some overlap of gene signatures between what we found for adipose iNKT cells, splenic iNKT cells at 4 weeks post-αGalCer and what has been published in the literature for Tr1 cells. Therefore, we respectfully suggest that a detailed comparison of adipose iNKT cells and Tr1 cells on a functional level is outside of the scope of this study.
5. Figure 2F only shows the representation of the experiment, statistics summary bar graphs should also be included.
We agree, and we have added statistics summary bar graphs for this experiment in Figure 2G as requested, replacing the scatter plot present in the original manuscript.
6. Figure 3B is generated from bulk analysis of scRNA-Seq data, right? What about by Stages (4 and 72h) subclusters?
We apologize for any confusion with Figure 3B, we found this slightly more difficult to explain in a short space. To generate the analysis in Figure 3B we performed cross-dataset differential gene expression analysis of adipose and splenic iNKT cells. This was performed by first individually comparing adipose versus splenic iNKT cells at steady state, 4 hours post-αGalCer and 72 hours post-αGalCer using gene expression analysis, generating three different lists of differentially expressed genes between adipose and splenic iNKT cells, one list for each time point or comparison. We then identified genes which were significantly enriched in either adipose or splenic iNKT cells across all three comparisons. This identified a total of 971 genes enriched among adipose iNKT cells and a total of 65 genes enriched among splenic iNKT cells. We then performed overrepresentation analysis of these enriched genes using gprofiler21 to generate the pathway analysis shown in Figure 3B. To clarify this in the manuscript we have added additional explanation to the Methods (see “Downstream scRNA-Seq data analysis”) and referenced this in the text to aid the reader in understanding our analysis. The reason that we did our analysis in this way, rather than just performing a bulk comparison of all adipose and splenic iNKT cells at steady state, 4 hours post-αGalCer and 72 hours post-αGalCer together, is that our method allows us to better leverage the different temporal data that we have access to in our scRNA-Seq data, and enables the identification of genes which are always (no matter what activation state) enriched among adipose or splenic iNKT cells, providing insight into core genes associated with the phenotype of iNKT cells in the adipose tissue or spleen. We hope that this explanation provides more clarity for the reader.
7. "Analysis of cytokine production among adipose iNKT cells revealed that IL-10 was only expressed by NKT1 cells (Figure 4B/D)." Better have flow data to support this claim. No scale stick in Y axis in Figure 4C.
We note that the referenced statement in our manuscript reads “Il10 was only expressed by NKT1 cells”, which refers to Il10 gene expression being specific to NKT1 cells. We show this by heatmap in Figure 4B. However, we found that IL-10 and IFNγ were co-produced by splenic and adipose iNKT cells at 4 weeks post-αGalCer (Figure 6—figure supplement 3), and IL-10 and IFNγ were also co-produced by splenic iNKT cells at 72 hours post-αGalCer (Figure 1—figure supplement 3). As IFNγ is a flagship NKT1 cell cytokine, these data suggest that iNKT10 cells are an NKT1-like population (or were previously NKT1 cells). We have added scale sticks to the Y axes for Figure 4C.
8. Clusters showing in Figure 5F with NKT 10 cells are better validated in protein level by FACS.
We thank the reviewer for this recommendation and agree. We have reworked Figure 5, reducing the number of scRNA-Seq plots present (see Figure 5E), and we present new data showing co-expression of KLRG1 and Granzyme A (see Figure 5F) by flow cytometry, to better illustrate the identification of functional KLRG1+ iNKT cells among reactivated splenic iNKT cells. We have also reworked and moved flow cytometry data from Figure 7 of the original manuscript to now show co-expression of cMAF and IL-10 in the reworked Figure 5 (see Figure 5G), to better illustrate the identification of functional cMAF+ iNKT cells among reactivated splenic iNKT cells. We have also reworked the Results text associated with Figure 5 to reflect these new data.
9. Clusters showing in Figure 6G should be validated in protein level.
We thank the reviewer for this recommendation. The goal of Figure 6G in the original manuscript was to indicate that iNKT cell populations similar to the memory-like KLRG1+ and cMAF+ iNKT cells that we identified by scRNA-Seq and validated by flow cytometry in the spleen after prior αGalCer immunization (Figure 6C and 6D of the original manuscript) are always detected in the adipose tissue at steady state. In our analysis of adipose NK1.1- iNKT cells in Figure 6G of the original manuscript, we identified Cluster C, which might be representative of KLRG1+ iNKT cells present at steady state in the adipose tissue, and Cluster B, which might be representative of a cMAF+ iNKT cell population present in adipose tissue at steady state. We have now reworked Figure 6 and included new FACS data showing distinct KLRG1+ iNKT cell and cMAF+ iNKT cell populations present in the adipose tissue at steady state and after αGalCer treatment (see Figure 6D-6F). These new FACS data make the same point as the previous scRNA-Seq data did about KLRG1+ and cMAF+ iNKT cells being present in the adipose tissue at steady state while additionally showing the enrichment of KLRG1+ and cMAF+ iNKT cells across multiple organs after αGalCer, which was a request from reviewer #1. The scRNA-Seq data from Figure 6E, 6F, 6G and 6H of the original manuscript are now present in Figure 6—figure supplement 5. We have also reworked the Results text associated with Figure 6 to reflect these new data.
10. The author did not mention how many biological replicates for scRNA-seq experiment, since mice treated with activator could vary widely between mouse to mouse, at least 2 biological replicates are needed.
Thank you to the reviewer for highlighting this. In the “scRNASeq sequencing and data pre-processing” section of the Methods we state that “A total of thirty-five visceral adipose tissue deposits and five spleens were pooled for each sample”. To be more precise on this point, we have amended this statement to say “A total of thirty-five visceral adipose tissue deposits from 35 mice or five spleens from 5 mice were pooled for each scRNA-Seq sample”.
11. The author mentioned in the method that both C57BL6 mice and BALB/c mice were used in this study, but did not mention which experiments were performed using C57BL6 mice or BALB/c. NKT cell subsets are different between C57BL6 and BALB/C mice. Please discuss the rationale why these two strains were used and make them clearer.
We agree that it is confusing without further clarification, as reviewer 2 also mentioned, and we have added a clarification to the “Animals” section of the Methods stating that “C57BL/6 mice were used for all experiments unless otherwise specified” and have made it more clear in both the Figure Legends and Results text associated with Figure 2 and Figure 2—figure supplement 4, which are the only figures where BALB/c mice were used in some experiments. C57BL/6 mice were used for almost all experiments in our study, including the generation of all scRNA-Seq data. BALB/c mice were specifically used for the purpose of investigating the phenotype and function of NKT1, NKT2 and NKT17 cells using an excellent thymic model previously described in a publication by Cameron et al. (2018)15 from the Godfrey lab, in which functionally NKT1, NKT2 and NKT17 cells can be easily labeled using the surface markers ICOS and the activation-associated glycoform of CD43 (CD43-HG), so that the cells do not need to be fixed and permeabilized. This allows us to perform later assays on live cells, such as mitochondrial analysis, which is not possible if fixation and permeabilization buffer is used for transcription factor analysis. However, we found that this method only works in BALB/c mice and not well in C57BL/6 mice, hence for Figure 2 and Figure 2—figure supplement 4 only, BALB/c mice were used.
References:
1. Shimizu, K. et al. KLRG+ invariant natural killer T cells are long-lived effectors. Proc. Natl. Acad. Sci. U. S. A. 111, 12474–12479 (2014).
2. Murray, M. P. et al. Transcriptome and chromatin landscape of iNKT cells are shaped by subset differentiation and antigen exposure. Nat. Commun. 12, 1–14 (2021).
3. LaMarche, N. M. et al. Distinct iNKT Cell Populations Use IFNγ or ER Stress-Induced IL-10 to Control Adipose Tissue Homeostasis. Cell Metab. 32, 243-258.e6 (2020).
4. Sag, D., Krause, P., Hedrick, C. C., Kronenberg, M. and Wingender, G. IL-10–producing NKT10 cells are a distinct regulatory invariant NKT cell subset. J. Clin. Invest. 124, 3725–3740 (2014).
5. Omilusik, K. D. et al. Transcriptional repressor ZEB2 promotes terminal differentiation of CD8+ effector and memory T cell populations during infection. J. Exp. Med. 212, 2027– 2039 (2015).
6. Obar, J. J. et al. Pathogen-Induced Inflammatory Environment Controls Effector and Memory CD8 + T Cell Differentiation. J. Immunol. 187, 4967–4978 (2011).
7. Joshi, N. S. et al. Inflammation Directs Memory Precursor and Short-Lived Effector CD8(+) T Cell Fates via the Graded Expression of T-bet Transcription Factor. Immunity 27, 281 (2007).
8. Plumlee, C. R. et al. Early Effector CD8 T Cells Display Plasticity in Populating the Short-Lived Effector and Memory-Precursor Pools Following Bacterial or Viral Infection. Sci. Rep. 5, 1–13 (2015).
9. Kallies, A., Zehn, D. and Utzschneider, D. T. Precursor exhausted T cells: key to successful immunotherapy? Nat. Rev. Immunol. 2019 202 20, 128–136 (2019).
10. Ricardo Miragaia, A. J. et al. Single-Cell Transcriptomics of Regulatory T Cells Reveals Trajectories of Tissue Adaptation. (2019) doi:10.1016/j.immuni.2019.01.001.
11. Evrard, M. et al. Sphingosine 1-phosphate receptor 5 (S1PR5) regulates the peripheral retention of tissue-resident lymphocytes. J. Exp. Med. 219, (2021).
12. Zuberbuehler, M. K. et al. The transcription factor c-Maf is essential for the commitment of IL-17-producing γδ T cells. Nat. Immunol. 20, 73–85 (2019).
13. Fujino, M. et al. c-MAF deletion in adult C57BL/6J mice induces cataract formation and abnormaldifferentiation of lens fiber cells. Exp. Anim. 69, 242 (2020).
14. Gabryšová, L. et al. C-Maf controls immune responses by regulating disease-specific gene networks and repressing IL-2 in CD4+ T cells article. Nat. Immunol. 19, 497–507 (2018).
15. Cameron, G. and Godfrey, D. I. Differential surface phenotype and context-dependent reactivity of functionally diverse NKT cells. Immunol. Cell Biol. 96, 759–771 (2018).
16. Lynch, L. et al. Regulatory iNKT cells lack expression of the transcription factor PLZF and control the homeostasis of Treg cells and macrophages in adipose tissue. Nat. Immunol. 16, 85–95 (2015).
17. Lee, Y. J. et al. Tissue-Specific Distribution of iNKT Cells Impacts Their Cytokine Response. Immunity 43, 566–578 (2015).
18. Lee, Y. J., Holzapfel, K. L., Zhu, J., Jameson, S. C. and Hogquist, K. A. Steady-state production of IL-4 modulates immunity in mouse strains and is determined by lineage diversity of iNKT cells. Nat. Immunol. 14, 1146–1154 (2013).
19. Engel, I. et al. Innate-like functions of natural killer T cell subsets result from highly divergent gene programs. Nat. Immunol. 17, 728–39 (2016).
20. Oleinika, K. et al. CD1d-dependent immune suppression mediated by regulatory B cells through modulations of iNKT cells. Nat. Commun. 9, 1–17 (2018).
21. Raudvere, U. et al. G:Profiler: A web server for functional enrichment analysis and conversions of gene lists (2019 update). Nucleic Acids Res. 47, W191–W198 (2019).