Skin Tregs are highly activated on Postnatal day 6
The accumulation of Tregs in skin peaks on postnatal day 13 (P13) of life, likely derived from lymphocytic seeding from lymphoid organs (6). At this time-point Tregs display a highly proliferative and activated profile. We sought to perform immune profiling at earlier timepoints, prior to P13, upon first entry of Tregs in skin. We first characterized Treg abundance and phenotypic marker expression in steady-state C57BL/6 neonatal skin and skin-draining lymph node (SDLNs) on P6, P9, P12, and later P28 (juvenile) and P49 (adult) timepoints.
Flow cytometric profiling revealed the presence of Tregs at detectable but low numbers on P3 (Fig. 1A-B; full gating strategy is shown in Supplementary Fig. 1). The abundance and proportion of CD4+Foxp3+ Tregs steadily increased and peaked during the second week after birth, which is consistent with previous studies (Fig. 1A-B) (6). This wave of Treg accumulation in P12 skin was unique to Tregs as our global analysis of other major skin-resident T cell subsets during this time frame did not display a similar pattern, namely Foxp3− T effector cells (Teffs) and CD8+ T cells (Supplementary Fig. 2A-F). Of note, Treg percentages on P8 and P12 were comparable to that found in adult skin on P49 (Fig. 1A), corroborating a recent study reporting similar findings (8). By contrast, the percentage of FoxP3+ Tregs in SDLNs was less variable and remained steady throughout the analysis period. The proliferative index of Teffs, CD8 + T cells, and Tregs was the highest in P6 skin but similar amongst all T cell subsets in the three time-points assessed, as evidenced by equivalent levels of Ki67 expression. However, in P9 SDLNs, Tregs were more proliferative relative to other T cell subsets (Supplementary Fig. 2C, 2F).
Given the rapid influx of Tregs in skin during the early neonatal period of P6-P12, we assessed the expression of the activation-associated markers CD25, CD27, CTLA4, and ICOS. Between P6 to P9, the median flourescence intensity (MFI) values of all markers, except CD27, were significantly higher in P9 skin Tregs relative to their SDLN counterparts (Fig. 1C). In addition, the MFI of all markers were upregulated by at least 2-fold between P6 and P9 skin Tregs, but not in SDLN Tregs (Fig. 1C). The proportions of Treg activation marker expression largely resembled the fluctuations observed in MFI values for both skin and SDLN compartments (Fig. 1D). Interestingly, later skin seeding Tregs (LTregs) in P12 skin expressed lower levels of CD25, CD27, and CTLA4 relative to P9. Taken together, these data suggest P3-P12 is an important interval where Tregs accumulate in skin and identifies P6-P9 as a specific time window where early skin seeding Tregs (ETregs) become highly, but transiently, activated in the tissue.
Neonatal Tregs suppress inflammation and skin pigmentation in later life
Given that Treg activation and proliferation is highly dynamic in P6-P12 skin, we hypothesized that Tregs in this time window may play an important role in regulating either inflammatory responses or postnatal skin development, or both. To test this, we utilized Foxp3-DTR transgenic mice where the diphtheria toxin receptor (DTR) is expressed ahead of the Foxp3 promoter (9). Following administration of DT, these mice permit highly robust Treg depletion in both SDLNs and skin (5, 10, 11). To ablate Tregs from P6-P12, we administered DT on P6, P8, P10 and P12 (hereafter referred to as the “ΔΤreg” group). We chose P28 as the timepoint for analysis as the major hallmarks of stem cell (SC) mediated skin development are manifested at this age, namely hair follicle SC (HFSC) activity and melanocyte SC (MeSC) mediated skin pigmentation.
While Treg sufficient controls on P28 developed normal appearing dorsal skin with black pigmentation, this was markedly impacted in ΔTreg animals, where the skin failed to pigment (Supplementary Fig. 3A-B). This phenotype was confirmed histologically using Fontana & Masson (F&M) staining that detected melanin granules produced by MeSCs in Treg sufficient controls but not in ΔΤreg skin (Supplementary Fig. 3C). To ascertain whether stem cell regulation is impacted upon Treg depletion, we profiled CD34+ Itga6+ HFSCs and CD117+ MeSCs by flow cytometry (Supplementary Fig. 3D-E). Both the proliferation and abundance of HFSCs were unaffected in ΔTreg skin relative to controls on P28. However, MeSC proliferation was markedly reduced in ΔTreg skin. These findings suggest the melanogenic function of MeSCs in the HFs is under Treg control during the P6-P12 time frame, whereas HFSCs are not impacted.
To determine if loss of neonatal Tregs results in systemic inflammation in later life we firstly monitored body weight gain. Depletion of Tregs during this six-day window resulted in weight reduction by P28. This was despite of a repopulating Treg presence that was significantly higher than Treg-sufficient controls (Supplementary Fig. 3F). In addition, immune profiling of all major skin-resident T cell subsets revealed an increased abundance and activation status of Teffs and CD8+ T cells on P13 – only 1 day after the Treg depletion regimen (Supplementary Fig. 3G-H). This inflammatory phenotype persisted even at the later timepoint of P28 (Supplementary Fig. 3I-K). In particular, CD8+ T cells were highly proliferative and outnumbered Tregs, as evidenced by an increased CD8:Treg ratio in ΔTreg mice relative to controls. These data suggest there is an immediate inflammatory response following the six-day Treg depletion regime that may persist until later life. Indeed, the same Treg depletion regimen in adult mice does not cause overt skin inflammation (10), supporting the idea that neonatal and adult Tregs are functionally distinct (7). Overall, the loss of neonatal Tregs from P6-P12 leads to defective MeSC mediated melanogenesis and a consequent defect in skin pigmentation, that may be associated with a systemic inflammatory response.
Early skin seeding Tregs (ETregs) are required for skin pigmentation.
In our initial experiments, we chose a 4-dose DT administration to deplete Tregs from P6-P12 (Supplementary Fig. 3). While this resulted in a defective skin pigmentation phenotype, it was also accompanied by reduced weight gain and an immediate inflammatory response that persisted until P28 (Supplementary Fig. 3F-K). Thus, we set out to determine whether defective MeSC function was a result of prolonged systemic inflammation and to also define a precise ‘window’ of time for Treg requirement. Given ETregs acquire a highly activated profile between P6 to P9, and the less activated profile of LTregs in P12 skin (Fig. 1C-D), we sought to address how transient loss of Tregs at these two timepoints would impact both skin development and local inflammation. To do so, we implemented a 2-dose DT regimen by administering DT on P6 and P8 to deplete ETregs (hereafter referred to as the “ΔETreg” group), and on P10 and P12 to deplete LTregs (hereafter referred to as the “ΔLTreg” group). Efficient Treg depletion in the skin was confirmed one day after the last DT injection on P9 (for the ETreg group) and P13 (for LTreg group) (Supplementary Fig. 4A-B). Notably, transient depletion of Tregs in the ETreg group recapitulated the attenuation of dorsal skin pigmentation on P28 observed in the 4-dose DT-treated ΔTreg group (Fig. 2A-E). In contrast, the skin of ΔLTreg animals was not affected relative to controls (Fig. 2B-E). Histologic F&M examination showed a significant reduction in melanin production in ΔETreg, but not in ΔLTreg animals, as quantified using a skin pigmentation index (Fig. 2D-E). Importantly, Treg ablation during the P6 to P8 window in ΔETreg animals had no significant effect on weight gain relative to Treg sufficient control groups (Supplementary Fig. 4C). These results indicate an early 2-dose depletion regimen targeting ETregs, but not LTregs, impairs skin pigmentation without compromising animal fitness in later life.
Next, we performed flow cytometry to test the hypothesis that defective skin pigmentation is linked to a disruption of SC maintenance and/or activation. Profiling of MeSCs and HFSCs showed that both the abundance and proportion of Ki67 expression remains comparable between control and ΔETreg groups on P28 (Fig. 2F-G). Therefore, SC maintenance is unaffected in the absence of ETregs.
Given the important role of T cells in disorders of skin pigmentation (12), we assessed whether T cell numbers are impacted in ΔETreg skin on P28. Indeed, the skin T cell pool expands, but maintain homeostatic levels of proliferation, Teff:Treg ratios, and CD8:Treg cell ratios (Fig. 2H-K). We also assessed for the hallmark indicator of global tissue inflammation by quantifying epidermal hyperproliferation of the skin. Histological quantification of epidermal thickness showed no difference between ΔΕTreg and control skin on P28 (Fig. 2J). Additionally, Teff numbers decreased in ΔETreg skin on P9, whilst the abundance and proliferation of other T cell subsets were unchanged (Fig. 2L-M). Nonetheless, we observed an increased IFNγ production by CD8+ T cells in the skin on P9 following ETreg depletion (Supplementary Fig. 5A-E), and elevated levels of IFNγ and TNFα in SDLNs (Supplementary Fig. 5F-I). To determine if CD8+ T cells play a functional role in ETreg-mediated skin pigmentation, we co-depleted Tregs with CD8+ T cells to determine if MeSC function could be rescued in the absence of Tregs. Under these conditions, skin pigmentation was not restored, suggesting that suppression of CD8-mediated inflammation is not a major mechanism by which ETregs promote MeSC function (Supplementary Fig. 6A). Taken together, we are unable to detect any overt T cell-mediated inflammation in ΔETreg mice, either immediately on P9 or in later life on P28, that contributes to the defective skin pigmentation phenotype.
Melanocytes and PPARγ pathway are under the control of ETregs
We next set out to elucidate the cellular and molecular mechanisms responsible for ETreg- mediated control over MeSC function. We were intrigued by the requirement of Tregs during a very short 3-day window from P6-8, and the lack of fulminant inflammation in ΔETreg mice. Supporting our findings are previous studies showing that adult tissue-resident Tregs facilitate tissue homeostasis largely independently of suppressing conventional inflammatory responses; namely Jagged-1 expressing skin Tregs that promote hair regeneration and Areg-producing Tregs that support lung repair (10, 13). We therefore hypothesized that ETregs regulate either i) the recruitment of specific immune cell subsets we have not assessed in our immune profiling, or ii) specific cellular pathways in non-immune cells, thereby enabling MeSCs to effectively synthesize melanin and maintain postnatal skin homeostasis.
To test this hypothesis, we took a discovery-based approach and reasoned that defining the immediate transcriptomic changes under ETreg control would be key to elucidating the pathways contributing to defective skin pigmentation. We therefore performed bulk RNA-sequencing of whole skin taken from Treg-sufficient controls, ΔETreg, ΔLTreg, and ΔTreg groups 24 hours after the last DT treatment – P9 for ΔETreg or P13 for ΔLTreg and ΔTreg groups (Fig. 3A). We conducted pathway enrichment analysis using differentially expressed genes (p < 0.05) between P9 and P13 skin with ShinyGO (14). Interestingly, terms such as melanin biosynthesis, melanin metabolic processes, and melanocyte differentiation showed greater than 15-fold enrichment in P9 controls relative to ΔETreg skin. Additionally, transcriptomic defects in interferon-β responses were evident in the ΔETreg group (Fig. 3B). In comparison, ΔLTreg skin displayed changes primarily associated with cell development and morphogenesis (Fig. 3C). These observations suggest that melanocytes are the major cell lineage impacted upon depletion of ETregs, but not LTregs. This notion is further substantiated by the downregulation of key melanocyte transcripts (Tyr, Tyrp1, Dct, Slc45a2, Oca2, Slc24a5) specifically in the ΔETreg and ΔTreg groups, but not in the ΔLTreg group (Fig. 3D). In situ quantification of HF MeSCs using the specific lineage marker CD117 revealed no change in the overall abundance in P9 control or ΔETreg skin, suggesting the transcriptomic changes were not attributed to a quantitative reduction in MeSC numbers (Fig. 3E-F). Rather, it suggests MeSC function is impacted in ΔETreg animals likely due to downregulation of transcripts associated with melanin synthesis, including Tyr and Oca2. Overall, we found that melanocytes are key targets of ETregs and that the absence of ETregs lead to disruption of the melanocyte transcriptome. This conclusion is further supported by our histological quantification of MeSC-Treg distance, Teff-Treg and CD8-Treg distances. During early life, Tregs reside in close proximity to MeSCs in the HF region of skin, but not their canonical targets Teffs and CD8+ T cells (Supplementary Fig. 7A-C).
To identify the signaling pathways responsible for pigmentation downstream of neonatal Tregs, we employed specific inclusion criteria based on the observation that pigmentation fails to develop in the ΔETreg and ΔTreg groups. We selected genes that met the following criteria: 1) Not differentially expressed between the ΔETreg and ΔTreg groups, 2) Differentially expressed between the ΔLTreg and ΔTreg groups, and 3) Differentially expressed between the control and ΔTreg groups (Fig. 3G). We identified 195 genes that fulfilled these criteria. Pathway enrichment analysis revealed dysregulation of transcripts under the control of the peroxisome proliferator-activated receptor (PPAR) pathway (Fig. 3H). Amongst the three PPAR isoforms – PPARα, PPARβ, and PPARγ – only PPARγ transitions from high expression to low expression from neonatal to adult skin (15). We therefore focused on validation of PPARγ target genes. To validate the immediate decrease in PPARγ target genes, we performed RNAscope analysis of neonatal P9 control and ΔETreg skin to quantify the in situ expression of the PPARγ target genes, Fabp7 and Adipoq. The expression of both transcripts specifically in the HFs were significantly attenuated in ΔETreg skin relative to Treg sufficient controls. (Fig. 3I-L). Overall, our findings convincingly demonstrate that MeSC function in skin is under the tight control of ETregs during neonatal development. We identify a transient window for establishing a Treg-MeSC axis, during which loss of ETregs leads to significant disruption of the PPARγ signalling pathway.
PPARγ pathway is necessary and sufficient for pigmentation
We next sought to determine if the signalling pathways identified in our transcriptomic analysis are responsible for the pigmentation defect downstream of Tregs. In ΔETreg skin, the two main pathways impacted were associated with Type-I interferon response and PPAR signalling (Figs. 3B). These candidate pathways have previously been associated with adult melanocyte function in vitro (16, 17), but whether they play a role during postnatal skin development in vivo has not been assessed. To determine if regulation of the interferon pathway is a major mechanism by which ETregs facilitate MeSC function, we neutralized interferon alpha-receptor (IFNAR) function in ETreg depleted mice using an anti-IFNAR monoclonal antibody. Neutralization of IFNAR was unable to reinstate skin pigmentation in ΔETreg mice, ruling out the interferon-β response as a functional candidate in this process. (Supplementary Fig. 6B).
The next set of experiments addressed if modulation of the PPAR pathway plays a functional role during steady state neonatal skin development. Amongst three isoforms of PPAR (α, β and γ), PPARγ is expressed at higher levels in neonatal skin relative to adult skin (15). Therefore, we utilised small molecule modulators targeting the γ isoform. Administration of the highly specific antagonist Τ0070907 to C57BL/6J mice on P6 and P8 significantly impaired skin pigmentation on P28 (Fig. 4B-E). Defective pigmentation was accompanied by a reduction in the number, but not proliferation, of MeSCs (Fig. 4F-G). Importantly, these results indicate global PPARγ signalling during neonatal life is required for MeSC-mediated skin pigmentation.
Next, tο validate whether PPARγ functions downstream of Tregs, we administered a highly specific PPARγ agonist, GW1929, to ΔETreg animals on P6 and P8 (Fig. 4H). PPARγ agonization restored skin pigmentation in ETreg-depleted mice to baseline levels (Fig. 4I-L). However, MeSC numbers and proliferation were unaffected by PPARγ agonization (Fig. 4M-N), suggesting that either qualitative changes in MeSC function, or the presence of an intermediate PPARγ responsive non-immune cell type, may play a role in promoting melanogenesis. Lastly, the observed PPARγ-dependent changes in pigmentation were not associated with significant alterations in skin-resident T cell abundance or proliferative capacity (Supplementary Fig. 8A-H). Therefore, it seems likely that Treg-PPARγ axis functions independently of T cells. Taken together, our findings suggest that a dominant function of ETregs during postnatal development is to regulate the PPARγ signalling axis to establish MeSC-mediated skin pigmentation in later life.
Epithelial-intrinsic PPARγ signalling activity relies on the presence of ETregs
To begin to understand the diverse cell states in the skin microenvironment, and to further elucidate the cellular mechanisms underlying the ETreg-PPARγ axis, we performed single cell RNA-sequencing (scRNA-seq) of CD45 + and CD45- cells in the presence and absence of Tregs. We reasoned that since Tregs reside proximal to both HFs and the interfollicular epidermis (10, 18, 19), early compensatory responses of epithelial resident cell types to ETreg depletion likely precede effects on skin pigmentation in later life. Furthermore, because the expansion of activated self-reactive T cells is observed 3–4 days after Treg ablation (9) we sought to avoid these confounding variables by analyzing skin-resident cells 1 day after the last DT treatment, on P9. Also, given skin pigmentation requires the presence of P6-8 ETregs, but not P10-12 LTregs, P9 was a rational timepoint to analyze early transcriptional mechanisms that govern pigmentation downstream of ETregs.
Firstly, we quantified CD45+ immune cell subsets to assess the accumulation of the major inflammatory lymphoid and myeloid cell lineages in skin. In support of our flow cytometric profiling on P9 (Supp Fig. 5A-E), pronounced local immune cell activation and inflammation were not observed following ETreg depletion, even though NK cells were moderately increased (Fig. 5A-B). We also analyzed CD45− immune cells and identified melanocytes (Mel), epithelial cells (Epithelia), hair follicle cells (HF), and other structural cells such as fibroblasts (Fib) and vascular endothelial cells (Vasc). While the overall abundance of these cells was also minimally affected, the proportions of both Mel and Vasc cell populations appear to expand in ΔETreg skin. (Fig. 5C-D).
Next, we performed a pseudo-bulk differential gene expression analysis to ascertain the magnitude of transcriptomic changes induced across all cell populations in ΔETreg and control mice. The most pronounced changes were identified in the HF and Epithelia clusters encompassing all epidermal keratinocytes of skin. When combined, these two cell types account for almost 600 differentially expressed genes (DEGs) (Padj<0.05) in the dataset (Fig. 5E and Supplementary Table 2). Among the hematopoietic CD45+ fraction, the innate lymphoid cell (ILC) cluster was most impacted with 200 DEGs imparted by ETreg depletion (Fig. 5E and Supplementary Table 3). T cell numbers changed minimally, despite being considered the main targets of Treg-mediated suppression. Collectively, these results indicate that the major cellular targets under the control of ETregs in neonatal skin are epithelial and HF cells.
Whole skin bulk RNA-seq and functional in vivo skin pigmentation rescue experiments strongly implicate the PPARγ-pigmentation axis as a major mechanism under Treg control (Fig. 3B-L, 4I-K). Therefore, we next assessed changes in PPARγ activity across the captured cell types in the presence and absence of ETregs. We selected a set of experimentally validated PPARγ target genes (20) and constructed an activity score using the “AddModuleScore” function built into Seurat package. Increasing scores on this scale indicate increased expression of PPARγ target genes. Within the CD45+ fraction, neutrophils and macrophages displayed the highest activity levels which is in line with the known role for PPARγ in these cell types (21, 22). However, PPARγ scores in these cell types were largely unaffected by ETreg depletion. Instead, amongst the CD45− cells, the most pronounced downregulation of the PPARγ activity score was observed in HF keratinocytes (Fig. 5I). This result is in agreement with reduced expression of PPARγ target genes, such as Fabp7 and Adipoq in the HFs (Fig. 3I-L). This is perhaps an unsurprising result given the HF cluster demonstrated the highest levels of transcriptomic perturbation (Fig. 5E), and the spatial proximity of HF cells to MeSCs in skin. Further analysis of the HF subset showed that 49 of 363 DEGS (~ 14%) between the control and ETreg depleted groups were known transcriptional targets of PPARγ (p = 2.25×10− 9) (Fig. 5J). Further visualisation of the top 56 DEGs in HF cells highlight five direct PPARγ targets, two of which are Cxcr6 and F13a1 (Fig. 5K). These transcripts have previously been associated with asymmetric self-renewal capacity of stem cell populations and facilitation of skin wound healing responses, respectively (23, 24). These results indicate that in the absence of ETregs in murine skin, tissue remodelling processes acting via PPARγ signaling is significantly altered in HF cells.
Collectively, scRNA-seq analysis indicates ETreg depletion markedly impacts the abundancies of several non-immune cell subsets in the skin microenvironment while minimally impacting the adaptive lymphoid compartment. Most notably, our results reveal that both the transcriptome and PPARγ activity of HF cells are preferentially modulated by ETregs. As such, neonatal ETregs in skin likely facilitate MeSC function by sustaining PPARγ signalling in HF keratinocytes.
The PPARγ pathway is implicated in Vitiligo and during human skin development
Numerous disorders of skin pigmentation have been described in humans. The most prominent of which is the autoimmune skin disease, vitiligo, in which depigmented skin results from melanocyte destruction (25). Genome wide association studies have highlighted vitiligo susceptibility loci in genes that support Treg function, including IKZF4 and CTLA4 (26, 27), Additionally, Tregs are important for restraining skin depigmentation severity in lesional vitiligo skin, further suggesting a functional role for Tregs in disease pathogenesis. Given these associations, we sought to address whether PPARγ signalling is active in this setting. To test this, we analyzed scRNA-seq data of human skin samples from 5 healthy donors and 10 vitiligo patients (28). The final processed datasets yielded 8 main cell types. Consistent with our murine skin single cell data, we identified two major clusters of keratinocytes in healthy and vitiligo skin; ‘HF’ characterized by high expression of hair follicle–associated markers KRT14 and KRT15, and ‘Epithelia’ that expressed the interfollicular epidermal transcript KRT5 (Fig. 6D). We next applied the PPARγ scoring module to assess the level of activity across the captured cell populations (20). These analyses revealed that only the HF cell type from vitiligo skin exhibited a significant downregulation in PPARγ pathway activation, relative to heathy controls. (Fig. 6E). This result aligns closely to our murine single cell data where dysregulation of PPARγ signalling was identified in HF cells upon loss of ETregs in neonatal skin (Fig. 5I).
We next set out to address if the PPARγ pathway is involved in human neonatal melanocyte development. In utero, Tregs seed the skin early as gestational week (GW) 18 (15, 29). This timeframe aligns closely with murine skin development during the first week of life when ETregs begin to accumulate (Fig. 1). Importantly, this period of Treg infiltration coincides with human HF development, which involves the migration of melanoblasts to the HF to complete the melanocyte maturation process (30). Given these associations, we analyzed scRNA-Seq data of melanocytes isolated from fetal (9.5–18 GW), neonatal (0 years), and adult (24–81 years) tissue (28, 31). We identified 6 distinct clusters across the dataset. Chi-squared analysis revealed a marked enrichment of adult and neonatal melanocytes for cluster 0, while fetal melanocytes were enriched for cluster 2 (Fig. 6A-B). This observation suggests that melanocyte maturation involves the expansion of cluster 0. Indeed, cluster 0 expresses high levels of mature melanocyte markers such as DCT, PMEL, TYR. Cluster 2 expresses high level of proliferation-associated transcripts such as MKI67, TOP2A and CCNB1, therefore representing actively dividing melanocytes (Supplementary table 4). Further analysis identified elevated expression of multiple PPARγ target genes in cluster 0 (Fig. 6C) (32), including CAV1 (33), which regulates melanogenesis and SCARB1, which is involved in fatty acid uptake (34). Overall, an increased PPARγ activity is associated with melanocyte maturation. The transition frow low PPARγ activity to high PPARγ activity temporally coincides with Treg seeding. As such, our findings suggest the possibility that human Tregs may modulate PPARγ activity during human skin development.
Collectively, our analysis of human transcriptomic data suggests that the PPARγ pathway is active during melanocyte development and is preferentially disrupted in HF cells from the pigment-deficient disease state Vitiligo.