Abstract
In the small intestine, a cellular niche of diverse accessory cell types supports the rapid generation of mature epithelial cell types through self-renewal, proliferation, and differentiation of intestinal stem cells (ISCs). However, not much is known about interactions between immune cells and ISCs, and it is unclear if and how immune cell dynamics affect eventual ISC fate or the balance between self-renewal and differentiation. Here, we used single-cell RNA-seq (scRNA-Seq) of intestinal epithelial cells (IECs) to identify new mechanisms for ISC–immune cell interactions. Surprisingly, MHC class II (MHCII) is enriched in two distinct subsets of Lgr5+ crypt base columnar ISCs, which are also distinguished by higher proliferation rates. Using co-culture of T cells with intestinal organoids, cytokine stimulations, and in vivo mouse models, we confirm that CD4+ T helper (Th) cells communicate with ISCs and affect their differentiation, in a manner specific to the Th subtypes and their signature cytokines and dependent on MHCII expression by ISCs. Specific inducible knockout of MHCII in intestinal epithelial cells in mice in vivo results in expansion of the ISC pool. Mice lacking T cells have expanded ISC pools, whereas specific depletion of Treg cells in vivo results in substantial reduction of ISC numbers. Our findings show that interactions between Th cells and ISCs mediated via MHCII expressed in intestinal epithelial stem cells help orchestrate tissue-wide responses to external signals.
Introduction
The intestinal mucosa maintains a functional equilibrium with the complex luminal milieu, which is dominated by a spectrum of gut microbial species and their products. The functional balance between the epithelium and the lumen plays a central role in maintaining the normal mucosa and in the pathophysiology of many gastrointestinal disorders [1]. To maintain barrier integrity and tissue homeostasis in response to immune signals and luminal contents [1], the gut epithelium constantly regenerates by rapid proliferation and differentiation [2]. This process is initiated by intestinal stem cells (ISCs), which give rise to committed progenitors that in turn differentiate to specific IEC types [3, 4].
ISC differentiation depends on external signals from an ecosystem of non-epithelial cells in the gut niche. In particular, canonical signal transduction pathways, such as Wnt and Notch [5, 6], are essential to ISC maintenance and differentiation, and rely on signals from stromal cells [7, 8]. The intestinal tract is also densely populated by innate and adaptive immune cells, which maintain the balance between immune activation and tolerance [1, 9]. However, it is unknown if and how immune cells and the adjacent ISCs interact.
Several studies suggest an important role for immune cells in tissue homeostasis. Tissue-resident innate immune cells, such as macrophages and type 3 innate lymphoid cells (ILC3s), can play a role in regeneration of the gut [7, 10] and other tissues [11, 12]. Among adaptive immune cells, recent studies have implicated T regulatory cells (Tregs) in regeneration within muscles, lungs, and the central nervous system [13-15]. Skin-resident Tregs were very recently shown to be involved in maintaining hair follicle stem cell (HFSC) renewal through Jagged1-mediated Notch signaling [16]. In the gut, mouse models of intestinal infection, T cell depletion, and inflammatory bowel disease (IBD) all display aberrant epithelial cell composition, such as goblet cell hypoplasia or tuft cell expansion [17-20]. These phenotypes have been primarily interpreted as reflecting intestinal epithelial cell dysfunction and changes in gut microbial populations [9, 17, 21, 22].
Here, we set out to identify and characterize novel mechanisms for interaction between immune cells and ISCs. Using scRNA-seq, we identified a putative molecular mechanism for CD4+ T cell interaction with specific subsets of Lgr5+ ISCs with enriched expression of MHC class II (MHCII) molecules and higher proliferation rates. We characterized this putative interaction using scRNA-Seq and in situ analysis of canonical in vivo infection models, organoid assays, and T cell-depleted, Treg-depleted, and inducible epithelial-specific MHCII-KO mouse models. We found that CD4+ T helper cells influence ISC renewal and epithelial differentiation via MHCII interaction. Our study underscores the important anatomic positioning of CD4+ T cell–ISC interactions in the context of ISC renewal or contraction, gut inflammation, and tumorigenesis.
Results
High expression of MHCII genes by ISC subsets
To identify potential mechanisms for ISC–immune cell interactions, we searched for genes that are specifically expressed by ISCs compared to other gut epithelial cells and that encode cell surface or secreted proteins capable of interacting with cognate molecules on immune cells. We collected full-length, high-coverage scRNA-seq (scRNA-seq) data of 1,522 EpCAM+ intestinal epithelial cells (IECs) [18] from crypt-enriched small intestine of WT and Lgr5-GFP mouse models [23] (Methods). Using unsupervised clustering (k-nearest neighbor (k-NN) graph-based clustering, Methods) of the 1,522 cells (Table S1) we identified 637 Lgr5-high (Lgr5High) stem cells (Figure S1A,B), as well as clusters corresponding to mature enterocytes, Paneth, goblet, tuft, and enteroendocrine cells [18]. Clustering of only the 637 ISCs (Methods) further partitioned the ISCs into three distinct subsets (ISC-I, -II and -III, Figure 1A,B), all of which express known stem cell markers [3] including Lgr5 (Figure S1C). This was consistent with recent scRNA-seq reports [24, 25]. We confirmed that all three subsets comprise Lgr5+ ISCs using the Lgr5-GFP mouse model [26]: the three stem cell populations were strongly enriched for GFPhigh cells (Figure S1D), over 90% of the GFPhigh cells were allocated to one of the three stem cell subsets (Figure S1E), and the three subsets are present in similar proportions in the duodenum, jejunum, and ileum (Figure S1F and Methods). Lastly, we identified differentially expressed genes between the three subsets, as well as between all ISCs and the other IECs that are annotated as receptors or ligands for cell-cell interactions [27] (Figure S1G and Table S2).
We found that CD74, the invariant chain of the MHCII complex, was highly expressed and enriched in ISCs (Figure 1C-E and Figure S1G). Moreover, other MHCII genes were among the most strongly expressed by two out of three ISC subsets (ISC-II and -III) (Figure 1C,D) compared to other IECs (Figure S2A). These included many canonical components of the MHCII machinery, including H2-Ab1, H2-DMb1, H2-DMa, H2-Aa, Cd74, and the recently discovered co-stimulatory molecules Sectm1a and Sectm1b [28] (Figure 1D), but not the canonical co-stimulatory molecules CD80 and CD86 (data not shown). Although MHCII expression has been previously reported in intestinal epithelial cells [29-32], it was not shown to be enriched in ISCs. We found that MHCII expression in the ISCMHCII+ (ISC-II and -III) groups was the highest among all IECs at both the mRNA and protein levels (Figure S2A,B). We confirmed MHCII protein expression by ISCs using an immunofluorescence assay (IFA) and immunohistochemistry (IHC) with anti-MHCII antibodies in wild type mice, and its absence in a constitutive MHCII knockout (KO) [33] (Figure 1E and Figure S2C).
MHCII-expressing ISCs are more proliferative
The three ISC subsets vary not only in their expression of the MHCII system, but also in their expression of signatures of the cell-cycle [34, 35] (Figure 2A,B). The subset with highest MHCII expression consisted primarily of cells in G1/S. The second subset, with lower but significant MHCII expression, had cells spanning several phases of the cell-cycle including G2/M. We concluded that cells in both of these subsets are likely in highly proliferative states and termed these ISC-II and ISC-III, respectively. In contrast, the cells in the subset with low or no detectable expression of MHCII, termed ISC-I, also had low G1/S and G2/M scores (129 of 209 cells) and likely represented cells in G0. The low-cycling state of ISC-I was further supported by the higher expression of the histone methylase Kdm5b, which is highly expressed in post-mitotic differentiated cells of the small intestine (Figure S2D,E) and in low-cycling or quiescent cells in other systems [34, 36-39]. Such heterogeneity in the proliferative state of ISCs has been recently reported, including a quiescent ISC subset, which is enriched for Mex3a and correlates well with our low cycling ISC-I subset (data not shown) [24, 25]. Importantly, while the cell-cycle status aligns with the partitioning of ISC-I, II and III subsets, the ISC subsets are discernable even when we exclude canonical cell-cycle genes (Figure 2C, Figure S3A, Table S3 and Methods) and even when analyzing only the 183 ISCs scoring at G0 (Figure S3B).
We validated the association between the ISC subsets, MHC-II, and cell-cycle status by co-staining in situ (Figure 2D), by in vivo EdU labeling followed by single-nucleus RNA-seq (Div-Seq) [40] (Figure 2E), and by determining the proportion of EdU+ cells in subsets of GFPhigh cells with different levels of MHCII expression (Figure 2F). We also sorted single MHCIIhigh and MHCIIlow ISCs from Lgr5-GFP mice and collected 503 full-length scRNA-seq profiles. MHCIIhigh ISCs had a higher proportion of cells with high scores for ISC-II and -III state signatures, whereas MHCIIlow ISCs had a higher proportion of cells scoring highly for the ISC-I state, consistent with our in silico analysis (Figure 2G and Figure S3C). Taken together, these results support an association between MHCII expression and proliferative state within ISCs in vivo. Thus, the heterogeneity of Lgr5+ stem cells and MHC II expression is correlated with proliferation rates.
T helper subsets and their signature cytokines regulate ISC renewal in organoids
We hypothesized that ISCs may interact with CD4+ T helper (Th) cells via MHCII recognition and, as a consequence, CD4+ Th cells may affect ISC fate via cytokine–receptor interaction. Importantly, IECs, including ISCs, express receptors for Th cytokines interferon gamma (IFNγ), interleukin-10 (IL-10), IL-13, IL-4 and IL-17A (data not shown). Furthermore, intra-vital imaging showed that CD4+ Th cells can be in very close proximity to stem cells in small intestinal crypts (Figure S4A and movie S1). Moreover, scRNA-seq of IECs following infection of mice with Salmonella enterica (Salmonella) or Heligmosomoides polygyrus (H. polygyrus) [18], which induce Th1 (Figure S4B-D and Table S4) and Th2 responses, respectively, shows not only distinct shifts in the proportions of post-mitotic cells, such as tuft (in H. polygyrus) and Paneth (in Salmonella) cells [18] (Figure S4E-F), but also a reduction in ISC and stemness programs, and especially in ISC-I cells (Figure S4G-J). The observed elevation in ISCMHCII+ programs is consistent with our hypothesis, but in principle many other indirect cellular and molecular mechanisms in the complex cellular ecosystem of the crypt may also be involved.
To dissect the potential interactions between T helper cells and ISCs independently of other contributions to the niche, we therefore next used the intestinal organoid system [26] in which immune cells are natively absent but can be added in a controlled manner [41]. We introduced either specific CD4+ T helper subsets (Figure S5A) or their corresponding signature cytokines to organoid cultures, and used scRNA-seq to identify changes in the proportions or expression programs of ISCs. In one set of experiments, we co-cultured organoids with CD4+ T cells that were polarized ex vivo towards Th1, Th2, Th17, and iTreg cells [42] (Figure S5B). In a parallel set, we stimulated organoids derived from C57BL/6J WT mice with key cytokines produced by each of the four T helper subsets: IFNγ (Th1), IL-13 (Th2), IL-17a (Th17), and IL-10 (inducible Treg, iTreg). In each experiment, we collected droplet-based scRNA-seq profiles (Methods). For co-cultures, we computationally distinguished (post-hoc) T cells from epithelial cells by their profiles (Methods) and confirmed the Th cell state by mRNA expression of signature cytokines and transcription factors (Figure S5C). Although ex vivo polarized T helper cells share many hallmarks with their in vivo counterparts, they do not perfectly recapitulate them. In particular, Th2 differentiation yielded only 16.5% IL-4 and IL-13 expressing cells, while other T helper subsets had higher differentiation rates (Figure S5B). There are also several differences between organoids and in vivo IECs (Figure S5D-G): Organoids are enriched for stem cells [26, 43] (Figure S5D), the goblet and Paneth lineages do not fully diverge (Figure S5E), also in independently-generated organoids [44] (Figure S5F), and MHCII expression was not detected in our organoid culture (Figure S5G); thus, any impact of Th cell co-cultures is likely mediated through cytokine secretion from the polarized Th cells.
Each of the Th co-cultures or corresponding cytokine treatments resulted in a distinct modulation of the organoid ISC compartment (Figure 3A,B and Figure S6A-C). Strikingly, co-cultures with iTregs and treatment of organoids with their associated cytokine IL-10 led to organoid ISC expansion (Figure 3A,B, Figure S6A-C, Methods and Table S5), while co-cultures with Th1, Th2 and Th17 cells or treatment with IL-13 or IL-17 all reduced the size of the ISC pool of the organoids. Consistent with their depleted stem cell pool, organoids co-cultured with Th1, Th2, or Th17 cells or treated with IL-17a all showed elevated numbers of TA cells (p<10-4, hypergeometric test, Figure 3A,B). Note, for IFNγ, we used a low concentration (0.5u/ml) to avoid organoid apoptosis [45], which did not elicit any effect, while Th1 co-culture resulted in elevation of MHCII signature in IECs (Figure 3A, top and Figure S5G). In addition, the treatments impacted cell differentiation: IL-13 treatment decreased the proportion of secretory ‘Paneth-goblet’ cells, and increased tuft cells (Figure 3A,C) [19, 46]; Th1 co-culture up-regulated Paneth cell-specific genes (Figure S6D-F), consistent with in vivo observations (Figure S4F); and Th2 cell co-cultures had the opposite effect (Figure S6D,E).
The effects of Th cell subsets and cytokines on ISC numbers suggest that they affect ISC renewal potential, which in turn should affect the ability of ISCs to form organoid cultures. To test this hypothesis, we assessed whether key cytokines affect ISC clonogenicity [47]. We reseeded equal numbers of cytokine-treated organoids in new cultures and quantified the number of organoids after three days (n=6 replicates per each group, Methods). Consistent with our hypothesis, there was a significant reduction in the clonogenicity of organoids treated with the ISC-reducing cytokine IL-13, whereas the ISC-expanding cytokine IL-10 induced higher clonogenicity (Figure 3D), confirming the ability of this Treg-generated cytokine to rejuvenate the stem cell pool.
Elevation in ISC pool under epithelial MHCII ablation in vivo
Since the MHCII system is not expressed in organoids, we next assessed its role in IECs in vivo by its conditional KO. We crossed H2-Ab1fl/fl [48] to Villin-Cre-ERT2 [49] mice, generating a mouse model of specific and inducible MHCII knockout in IECs (MHCIIΔgut). We profiled 1,559 IECs from the MHCIIΔgut mice (n=5) 10 days after Tamoxifen induction and 1,617 IECs from floxed control (MHCIIfl/fl) littermates (n=5 mice). We validated that MHCII is successfully knocked-out in EpCAM+ IECs (Figure 4A and Figure S7A), but not in CD11b+ dendritic cells in the mesenteric lymph node (Figure S7B).
Strikingly, the fraction of Lgr5+ cells was 31.3% higher in MHCIIΔgut mice (p<0.05, likelihood-ratio test, Figure S8A), which we confirmed by Lgr5-smFISH (Figure 4B), and the proportion of ISCs as defined by unsupervised clustering (Figure S8B,C) which was 17.6% higher (FDR<0.05, likelihood-ratio test, Figure 4C). Consistently, stem cell markers are overrepresented (p<10-6, hypergeometric test, Figure S8D) among the genes up-regulated in the MHCIIΔgut (FDR<0.05, likelihood-ratio test), including canonical ISC markers (e.g., Lgr5, Olfm4, Smoc2, and Igfbp4, Figure S8D). Furthermore, we separately analyzed only MHCIIΔgut ISCs in which H2-Ab1 is confirmed to be silenced (defined as no detectable mRNA) or only MHCIIΔgut ISCs in which H2-Ab1 mRNA is still expressed (Figure S8E, bottom left vs. right). We find that in MHCIIΔgut ISCs in which H2-Ab1 is confirmed to be silenced, expression of stem cell markers [3] was significantly higher than in stem cells still expressing H2-Ab1 (p < 0.05, likelihood-ratio test). Finally, the ISC-III signature score was significantly lower in stem cells from MHCIIΔgut mice (Figure 4D), suggesting that the ISCs in the expanded pool are shifted toward the ISCMHCII- state. Taken together, these data suggest that MHCIIΔgut increased ISC numbers and the expression of stem cell markers.
T cells modulate ISC renewal and differentiation in vivo
Our organoid assays predicted that Th cell subsets have distinct effects on intestinal epithelial cell differentiation. To demonstrate the relevance of the T cell-ISC interaction in vivo, we first assessed two T cell-deficient mouse models. First, we profiled 2,967 individual IECs isolated from athymic B6 nude mice [50] (n=2), characterized by T cell depletion. Unsupervised clustering revealed a markedly higher fraction of stem cells (52.5% increase, FDR<10-3, likelihood-ratio test, Methods) compared to control mice (n=6, Figure S9A,B,E). Consistently, stem cell markers were enriched (56 of 1,804 genes, p<10-6, hypergeometric test, Figure S9A and Table S6) among genes overall up-regulated in cells of nude vs. controls (FDR < 0.05, likelihood-ratio test). Similar analysis of 9,488 individual IECs profiled from TCRβ-KO mice (n=2) [51], characterized by a lack of α/β T cells, also showed a significant expansion of the ISC pool (35.0% increase, FDR < 0.05, likelihood-ratio test; Figure S9C-E). We confirmed the increased ISC numbers in situ in both T cell depleted models using Lgr5 single-molecule FISH (smFISH, Figure S9F).
Treg cells are essential to maintain the ISC niche in vivo
Our organoid assays also predicted that Treg cells promote renewal of the ISC pool. To test for this effect in vivo, we used the Foxp3-DTR mouse [52], in which Treg cells are specifically depleted upon application of diphtheria toxin (DT). We profiled 3,387 IECs from both Foxp3-DTR (n=4) and matched control mice (n=5) treated with DT for 7 days and confirmed Treg ablation in the lamina propria (Figure S10A). At this time point, there were no signs of increased cell death in the small intestinal crypts or of major tissue inflammation in IECs of Foxp3-DTR vs. control mice (Figure S10B), suggesting that the longer term effects of Treg depletion are not yet apparent [52]. However, consistent with our hypothesis, there was a substantial reduction (66.3% decrease, FDR<0.005, likelihood-ratio test) in ISC numbers in the epithelia of the Foxp3-DTR mice, as assessed by unsupervised clustering (Figure 4E), the fraction of cells in which Lgr5 mRNA was detected (p<0.005, likelihood-ratio test, Figure S10C), and smFISH (Figure 4F). Consistently, stem cell marker genes were overrepresented among those down-regulated across all cells in Treg-depleted mice (p<0.005, hypergeometric test, Figure S10D). There was also a substantial depletion of mature enterocytes (5.8-fold decrease, FDR<10-5), and expansion of tuft cells (4.1-fold increase, FDR<0.005) (Figure S10E), which we confirmed by IFA staining (Figure S10F). We did not observe significant changes in the expression of Notch signaling pathway components (p=1), or Notch targets [53] (p=0.31, hypergeometric test), which a recent study implicated in regulation of hair follicle stem cells by Tregs [16] (data not shown).
All cell types in the Foxp3-DTR mice, including ISCs, showed strongly elevated expression of MHCII genes (p<5×10-4, likelihood-ratio test, Figure 4G). Amongst stem cells, there was an increase in proliferation, as indicated by both the distribution of cell-cycle signature scores and mKi67 staining (p<0.005, likelihood-ratio test, Figure 4F and Figure S10B). Furthermore, and also consistent with our predictions, ISCs from Treg-ablated Foxp3-DTR mice had an increased proportion of MHCII positive, proliferative ISCs and a decrease in ISCMHCII- (ISC-I, Figure 4D).
Discussion
Previous studies of stem cell dynamics and differentiation processes [54, 55], focused on the role of the epithelial-intrinsic or stromal niche signals using lineage tracing. Here, we investigated the possibility of interactions between adaptive immune cells and ISCs. Combining scRNA-seq with homeostatic or perturbed conditions that manipulate either T helper cells, their cytokines, or MHCII expression by epithelial cells allowed us to assay comprehensive “snapshots” of ISC abundance and the fate of their progeny, followed by in silico inference of cell states and differentiation. In unperturbed mice, the expression of MHCII is high yet variable across ISCs, such that both ISC-II and III (ISCMHCII+) express high levels of the MHCII molecules, whereas ISC-I (ISCMHCII-) do not. Using controlled manipulation experiments in organoids and mice followed by scRNA-seq, we established a crosstalk between Th cells and ISCs.
In particular, our in vitro and in vivo results support a model in which Th cells interact with ISCs via MHCII molecules, impacting the ISC pool and resultant differentiation pathways through their key cytokines (Figure 4I). In this model, Treg cells, which are enriched in the small intestine, maintain the ISC niche. They may be elevated after a strong inflammatory response [56] to serve as a feedback effectors in order to replenish and maintain stem cell numbers. Conversely, Th1 and Th2 cells or their signature cytokines both reduce ISC numbers, and bias IEC differentiation toward specific epithelial cell-types, perhaps in order to respond to either bacterial (Th1 cells, Paneth cell increase) or parasitic (Th2 cells, tuft cell increase) insults. Th17 cells, which are highly enriched in the small intestine [57], reduce the number of ISCs, which may reflect a shift in the balance between stem cell renewal and differentiation. In this way, epithelial and immune response could be integrated to titrate responses to dense luminal flora, avoiding continuous inflammation, while reacting to pathogens: first, the intestinal stem cells utilize the equilibrium of pro-inflammatory and anti-inflammatory signals to balance between renewal and differentiation; second, distinct Th cell subsets can boost the desired immune response by affecting renewal and differentiation processes of the gut epithelia concordantly with signals arriving from the gut lumen. If this novel role for MHCII in T cell communication with stem cells also exists in other mucosal or non-mucosal compartments, it may open the possibility of a general mechanism in which adaptive immune cells regulate parenchymal stem cells in order to maintain tissue homeostasis under normal and pathological conditions.
Acknowledgements
We thank Leslie Gaffney for help with figure preparation, the Broad Flow Cytometry Facility: Patricia Rogers, Stephanie Saldi and Chelsea Otis and Tim Tickle for help with the Single Cell Portal. Work was supported by the Klarman Cell Observatory at the Broad Institute (AR and RJX), HHMI (AR), and a Broadnext10 award (AR an RJX). MB was supported by a postdoctoral fellowship from the Human Frontiers Science Program (HFSP). RJX funded by DK043351, DK097485 and Helmsley Charitable trust. ASK in a member of Searle Scholars Program, the Beckman Young Investigator Program and the recipient of the NIH New Innovator Award DP2 OD020839. ÖHY is funded by CA211184 and AG045144 NIH grants and is the recipient of the Sidney Kimmel Scholar Award, the Pew-Stewart Trust Scholar Award and a member in the American Federation of Aging Research. AR is a member of the SAB for ThermoFisher Scientific, Syros Pharmaceuticals, and Driver Group.