Thymic stromal lymphopoietin induces IL-4/IL-13 from T cells to promote sebum secretion and adipose loss

Abstract

Background:

The cytokine thymic stromal lymphopoietin (TSLP) promotes type 2 immune responses and can induce adipose loss by stimulating lipid loss from the skin through sebum secretion by sebaceous glands, which enhances the skin barrier. However, the mechanism by which TSLP upregulates sebaceous gland function is unknown.

Objective:

Here, we investigated the mechanism by which TSLP stimulates sebum secretion and adipose loss.

Methods:

RNA sequencing analysis was performed on sebaceous glands isolated by laser capture microdissection and single cell RNA sequencing analysis was performed on sorted skin T cells. Sebocyte function was analyzed by histological analysis and sebum secretion in vivo and by measuring lipogenesis and proliferation in vitro.

Results:

We found that TSLP sequentially stimulated the expression of lipogenesis genes followed by cell death genes in sebaceous glands to induce holocrine secretion of sebum. TSLP did not affect sebaceous gland activity directly. Rather, single cell RNA sequencing revealed that TSLP recruited distinct T cell clusters that produce IL-4 and IL-13, which were necessary for TSLP-induced adipose loss and sebum secretion. Moreover, IL-13 was sufficient to cause sebum secretion and adipose loss in vivo and to induce lipogenesis and proliferation of a human sebocyte cell line in vitro.

Conclusion:

We propose that TSLP stimulates T cells to deliver IL-4 and IL-13 to sebaceous glands, which enhances sebaceous gland function, turnover, and subsequent adipose loss.

Graphical Abstract

Capsule summary:

This study shows that TSLP-induced IL-4 and IL-13, which are pathogenic in atopic skin disease, could also promote sebum secretion, which is an important component of the skin barrier.

Introduction

The skin is a critical barrier surface that needs to sense and protect the host from the environment (1). One way the skin protects against environmental threats is through the production and secretion of a lipid-rich substance called sebum. Sebum is produced by sebaceous glands that lie within the mid-dermis and make up part of the pilosebaceous unit alongside hair follicles (2). Sebaceous gland lipids help repel water and prevent water loss, act as a delivery vehicle for secreted antimicrobial peptides (3, 4), provide nutrients and differentiation signals for skin resident immune cells (5, 6), regulate the skin microbiota composition (7, 8), and have antibacterial and antifungal properties (6, 9, 10). Thus, the regulation of sebum secretion is a central component of skin barrier function.

Several factors are known to control sebum production. Androgens are the most well studied, as testosterone and 5α-dihydrotestosterone stimulate sebocyte proliferation, sebaceous gland development, and sebocyte lipogenesis through the upregulation of sterol regulatory element binding proteins (SREBPs) (11, 12). Growth hormones such as insulin-like growth factor-1 (IGF-1) promote sebocyte cell growth and lipogenesis (13, 14), while retinoids suppress lipogenesis and trigger sebocyte apoptosis (15–18). The neuropeptides corticotrophin-releasing hormone and α-melanocyte-stimulating hormone both stimulate sebocyte lipid synthesis (19, 20). The active form of Vitamin D, 1,25(OH)(2)D(3), promotes sebocyte proliferation, wax ester production, and lipid accumulation (21). Finally, like adipocytes, sebocytes signal through peroxisome proliferator-activated receptors, which upregulate cellular lipid processing machinery to promote intracellular lipid droplet formation (17), and through Liver X receptors, which increase lipogenesis through transcriptional upregulation of fatty acid synthetases and SREBP-1 (22, 23).

However, little is known about if and how the immune system might regulate sebum production. One study reported that skin innate lymphoid cells (ILCs), which are maintained in the skin by interleukin 7 (IL-7) and TSLP, produce tumor necrosis factor alpha (TNF-α) and lymphotoxins, which limit sebocyte growth via Notch signaling (7). RAG/IL2Rγ double knockout mice, which lack ILCs, had decreased production of palmitoleic acid, a sebum fatty acid that inhibits growth of Gram-positive aerobic cocci, such as S. aureus and S. xylosus (7). The authors concluded that ILCs can modulate the microbial landscape of the skin by regulating sebaceous gland function (7).

Recently, our group found that the cytokine TSLP, which is known for its role in activating type 2 immune responses and mediating allergic/atopic diseases, stimulates recruitment of previously activated conventional CD4⁺ and CD8⁺ T cells to the sebaceous glands and induces holocrine secretion of sebum (24). One striking byproduct of this TSLP-immune-sebum axis was the metabolic effect - mice in which TSLP was overexpressed showed decreased visceral white adipose tissue, and were protected from obesity, insulin resistance, and fatty liver disease (24). These mice also had a striking “greasy fur” appearance from the sebum hypersecretion (24). We found that these effects of TSLP were dependent on conventional T cells, but independent of most other immune cell types including eosinophils, ILCs, regulatory T cells, dendritic cells, and B cells (24). Given that TSLP is a barrier cytokine, and that sebum is an important component of cutaneous barrier defense, we reasoned that this TSLP-T cell-sebum axis likely evolved to promote skin barrier function, and at high supraphysiological TSLP levels, causes adipose loss as the body undergoes lipolysis to replenish lipid loss at the skin. The goal of this study was to further elucidate the mechanisms by which TSLP-stimulated T cells might promote sebum production.

Methods

Mice.

C57BL/6 mice were purchased from Charles River Laboratories (strain no. 556). Il4/13^fl/fl and Stat6^−/− mice were purchased from Jackson Laboratories (strain no. 031366 and 005977). Tslpr^−/− mice, as previously described (25), were a kind gift from W. J. Leonard (NIH) and were bred and maintained in our facility. Ebeta^−/−, Rag2^−/−, and Vav^Cre mice were also bred and maintained in our facility. For adipose mass measurement, mice were harvested 2 weeks after adeno-associated virus serotype 8 (AAV) injection and bilateral eWAT were excised and measured. Unless otherwise specified, all mice were 7-12 week-old males at the time of use, were housed in pathogen-free conditions, and were treated in strict compliance with the Institutional Animal Care and Use Committee regulations at the University of Pennsylvania.

AAV injections and cytokine levels.

Control-AAV (AAV8.TBG.PI.eGFP.WPRE.bGH), TSLP-AAV (AAV8.TBG.PI.mTSLP.IRES.eGFP.WPRE.bGH), and IL-13-AAV (AAV8.TBG.PI.mIL13.IRES.eGFP.WPRE.bGH) were generated by the Penn Vector Core. For experiments using TSLP-AAV, mice were injected intravenously with 1 × 10¹¹ genome copies of AAV. For experiments with IL-13-AAV, mice were injected intravenously with 2 × 10¹⁰ genome copies of AAV. Mice were euthanized, blood was collected and centrifuged, and serum IL-13 levels were measured using murine specific ELISAs (R&D, DY413-05) according to the manufacturer’s protocol.

Preparation of skin T cells for flow cytometry, cell sorting, and scRNA sequencing.

Dermal sheets of ear skin were separated, minced, and incubated in Hank’s balanced salt solution (ThermoFisher, 24020117) containing 0.25 mg/ml of Liberase TL (Roche, 5401020001), 0.1 mg/ml of DNase I (Roche, 10104159001), and 0.7 mg/ml of collagenase D (Roche, 11088882001) for 1 hour with shaking at 37°C. Contents were then strained though a 70 μm filter into a new tube containing 10 ml of PBS, centrifuged at 4°C for 5 min at 180×g, and resuspended in PBS. Cells were stained with live–dead stain and cell surface stains (Table E3) at 4°C for 15 min in PBS.

Flow cytometry and cell sorting.

Flow cytometry was performed with an LSR II or LSR Fortessa (BD Biosciences). Cell sorting was performed with a FACSAria cell sorter (BD Biosciences). Data were analyzed using FlowJo software (TreeStar). Staining antibodies used are listed in Table E3. Live/Dead Near-IR (ThermoFisher, L10119, 1:1000 dilution) stain was used to exclude dead cells.

qPCR of skin.

Dermal sheets of ear skin were separated and transferred into TRIzol (ThermoFisher, 15596026) and minced in tissue lyser tubes (MP Biomedicals, 116910050) using a homogenizer (Qiagen, 69980). RNA was isolated according to the TRIzol manufacturer’s instructions. Glycogen (Roche, 10 901 393 001) was used as a carrier. Isolated total RNA was quantified using a Nanodrop 1000. cDNA was synthesized using the Superscript Vilo kit (ThermoFisher, 11755050) according to the manufacturer’s protocol. qPCR was performed using the Taqman Fast gene expression assay (Taqman, 4444557) according to the manufacturer’s protocol, using the following primers from ThermoFisher: Il4 (Mm00445259_m1), Il13 (Mm00434206_g1) and Gapdh (Mm99999915_g1). IL-4 and IL-13 gene expression were normalized to housekeeping gene GAPDH expression. All qPCR reactions were performed on a ViiA7 Real-Time PCR instrument (ThermoFisher).

Adoptive transfers.

Il4/13^fl/fl and Vav^CreIl4/13^fl/fl splenic T cells were isolated using a T cell negative selection kit (STEMCELL Technologies, 19851) and then 2×10⁶ isolated cells were transferred intravenously into Rag2^−/− mice. Four weeks later, all adoptively transferred mice were injected with control or TSLP-AAV.

Lipid extraction from hair shafts and thin layer chromatography (TLC).

A 2.5 cm × 2.5 cm area of hair was shaved from back skin and immersed in 4 ml of chloroform×methanol (Sigma, 288306 and Sigma, 322415) (2:1 v/v) followed by 4 ml of acetone. Lipid extracts were pooled, syringe filtered, dried down overnight under a stream of N₂ gas, and resuspended in equal volumes of chloroform—methanol (4:1, v/v) for loading onto TLC plates (Sigma, 100390). The TLC plates were developed three times using the following: (1) hexane (Sigma, 296090):isopropyl diether (Sigma, 673803):acetic acid (80:20:1) up to 50% of the plate height, (2) hexane:benzene (Sigma, 401765) (1:1) up to 80% of the plate height, and (3) hexane up to 90% of the plate height. Plates were allowed to dry between each developing solution. Plates were sprayed uniformly with 10% cupric sulfate (Sigma, 451657)/8% phosphoric acid (Sigma, P6560) solution and then baked at 120°C for 20 min to visualize lipid species. ImageJ (NIH) was used to quantify the intensity and area of the bands. The TLC non-polar lipid mixture A (Matreya, 1129) was used as a standard to identify lipid classes.

Histology.

Tissues were fixed in 10% formalin at 4°C overnight and embedded in paraffin before H&E or IHC staining. Skin samples were processed and stained (H&E and Ki67) by the University of Pennsylvania’s Cutaneous Phenomics and Transcriptomics Core. For quantification of sebocyte size, 8-10 sections at 20× magnification were captured per animal. ImageJ was used to draw circumscribing ellipses around sebocytes to quantify sebocyte area. For quantification of proliferating cells in various sebaceous gland zones, basal layer cells were identified as cells on the periphery of the sebaceous glands, while differentiated sebocytes were identified as larger cells in the interior of the sebaceous glands. % and number of Ki67⁺ basal cells were then quantified. Immunohistochemistry antibodies used are listed in Table E4. A mouse and rabbit specific HRP IHC kit (Abcam, ab236466) was used to detect staining.

SEB-1 sebocyte culture and treatment conditions.

SEB-1 cells were cultured in standard SEB-1 media consisting of: 5.5 mM low glucose DMEM (Invitrogen, 11886-092), 3:1 Ham’s F12 (Invitrogen, 11765-062), 2.5% fetal bovine serum (R&D, S11150), 0.0452 μg/mL hydrocortisone (Sigma, H-0396), 24 μg/mL adenine (Sigma, A-9795), 10 ng/mL insulin (Sigma, I-1882), 3 ng/mL epidermal growth factor, 1.2 × 10⁻¹⁰ M cholera toxin (Sigma, C-8052), and antibiotics (Invitrogen, 15240-062). For lipogenesis stimulation assays, cells were seeded at 120,000 cells per well in 6 well plates (Fisher, 3516) for TLC or 10,000 cells per well in 96 well plates (Fisher, 3904) for Nile Red fluorescence, and media was changed every other day until the cells were confluent, usually 6-7 days after initial seeding. Media was then changed into SEB1 stimulation media, which consisted of 5.5 mM low glucose DMEM and antibiotics, with or without 10 ng/mL IL-4 (Peprotech, 200-04) and 10 ng/mL IL-13 (Peprotech, 200-13).

LipidTOX staining and TLC.

For LipidTOX staining, cells were stained with 200 μL HCS LipidTOX^™ Green neutral lipid stain (Invitrogen, H34475) diluted to 1:150 in PBS for 25 min at 37°C. Staining solution was removed, and cells were trypsinized (Gibco, 25200056) for 5 minutes. Cells were then washed twice with PBS and flow cytometry was performed. For TLC, cells were collected into 15 mL conical tubes, spun down at 2000 rpm for 7 min at 4°C, and resuspended in 1 mL of 5.5 mM low glucose DMEM. 2mL of diethyl ether (Sigma, 32203) was added to each sample, samples were vortexed, and then spun down at 1000 rpm for 7 min at 4°C. Conical tubes were immediately frozen at −80°C for 7 min, and then the top liquid layer of each conical was poured into glass vials (Fisher, 14-955-327). 2 mL of diethyl ether was again added to each sample, samples were vortexed, spun down at 1000 rpm for 7 min at 4°C, frozen at −80°C for 7 min, and then the top layer of each sample tube was again poured into the glass vials, resulting in approximately a total of 4 mL of collected cell lipids dissolved in ethyl ether. Samples were evaporated under nitrogen gas to dry each sample, and dried samples were reconstituted in 20 μL of 4:1 v:v chloroform:methanol. TLC was then performed as described above.

Crystal violet cell viability assay.

SEB-1 cells were seeded at 7,000 cells per well in SEB-1 media in 96 well TC plates overnight to allow for adhesion. The next day, media was aspirated and replaced with fresh SEB-1 media with or without 10 ng/mL IL-4 and 13. Each day, from days 0-4, cells were washed once with PBS and stained with 50 μL of 0.5% crystal violet (Sigma-Aldrich, C0775) staining solution containing 80% dH20 and 20% methanol for 20 min at RT on a rocker. Cells were then washed gently with PBS 4 times and air-dried for 2 hours at room temperature (RT). To redissolve the crystal violet, 100 μL of methanol was added to each well and samples were incubated for 20 min at RT on a rocker. Plates were read at 570 nm OD, and baseline values of crystal violet staining in wells with media alone (no cells) were subtracted from all sample values.

scRNA sequencing methods.

Skin T cells were isolated and prepared as described above for flow cytometry and cell sorting. Single cell isolation and library preparation of FACS-sorted live, CD45.2⁺ CD90.2⁺ lymphocytes were performed using the Chromium platform (10× Genomics, v1.1 5’ chemistry with V(D)J sequencing) according to the manufacturer’s instructions. Sequencing was performed at the Center for Applied Genomics core at the Children’s Hospital of Philadelphia on an Illumina NovaSeq 6000 sequencer using an SP 100 cycles flow cell. FastQ files were then processed using the Cell Ranger count (v.2.1.0) analytical pipeline, and resulting count matrices were processed in R (v 4.2.2) using the Seurat (v 4.3.0) (26) and scRepertoire (v 1.7.0) (27) packages. Cell type annotations for Fig E1, A were provided by scType (28), and cell type annotations for Fig 1, C were manually annotated using reference genes from a published dataset (29). Gene set enrichment analysis (GSEA) was performed using the fgsea (v 1.22.0) (30) and msigdbr (v 7.5.1) (31–34) packages, while GO Enrichment analysis (35–37) was performed through the geneontology.org website.

Fig 1.

scRNAseq of skin T cells in mice given Ctrl-AAV vs. TSLP-AAV (A) Schematic of scRNAseq of skin T cells. Skin T cells were isolated and pooled from mice treated with Ctrl-AAV (N = 30) or TSLP-AAV (N = 10 mice). (B) TCR clonotype abundance and distribution from TCR sequencing of skin T cells. (C) UMAP of α or β TCR -expressing single cells with annotations. (D) UMAP plot comparison of α or β TCR -expressing single cells from mice treated with control vs. TSLP-AAV. (E) Heatmap analysis of the top 10 most differentially expressed genes per cluster. (F) GSEA analysis on the DEGs between activated CD4+ T cells (0, 1, 2, 3, 8, 9, and 13) in the TSLP condition and the Ctrl condition. (G) Targeted analysis and dot plot of Th2 genes.

Laser capture microdissection.

Laser capture microdissection (LCM) was performed using the LMD 7000 system (Leica Microsystems Inc., Chicago, IL). Formalin-fixed, paraffin-embedded mouse skin was processed onto a polyethylene naphthalate slide designed for LCM processing (Leica Catalog #11505158). For adequate material for RNA sequencing of sebaceous glands, at least 1,000 sebaceous glands or 1,000,000 μm² of tissue was isolated.

RNA extraction.

RNA extraction from sebaceous glands was performed using a Qiagen All Prep DNA/RNA FFPE Kit (Qiagen Catalog #80234) according to the manufacturer’s instructions. RNA concentration was measured by Qubit fluorometric quantification. Samples were assessed for RNA quality via BioAnalyzer.

Library preparation and sequencing.

cDNA library preparation was performed using Illumina Stranded Total RNA Prep with Ribo-Zero Plus Kit (Illumina Catalog # 20040529) with IDT for Illumina RNA UD Indexes, Set A (Illumina Catalog # 20040553). Libraries were assessed for cDNA quantity and library quality using Qubit and BioAnalyzer, respectively. For any samples with excess primer dimers present in the sample post-library preparation, extra bead washing was performed to purify these samples further. Samples were then pooled and sequenced on a Nextseq 550 using a NextSeq 500/550 High Output Kit v2.5 (150 Cycles) (Illumina Catalog # 20024907).

RNA sequencing analysis.

Bioinformatic analysis of the sebaceous gland RNAseq samples was performed in the R statistical computing environment version 4.2 and RStudio version 2022.02.1 using a published pipeline adopted from an open-source toolkit (38). Pseudo-aligning reads to the GRCm39 mouse cDNA reference genome was performed using Kallisto and reads were annotated using the Ensembl database (Kallisto). Data was filtered based on genes with less than 1 count per million in the number of samples equal to n of the smallest group. The Trimmed Mean of M-values method of normalization was used from the EdgeR package. Post-filtration, post-normalization, data was then variance stabilized via the Limma package using the VOOM function. Limma was also used for differentially expressed gene (DEG) analysis with multiple testing correction via the Benjamini-Hochberg method. The gprofiler2 R package was used to perform gene ontology pathway analysis. The msigdbr, clusterprofiler, and enrichplot R packages were used to perform gene set enrichment analyses using the C2 curated database from MSigDB, Reactome, and KEGG databases.

Statistical analysis and figure construction.

Data are reported as mean ± SEM. All measurements were made from distinct biological samples. Sample sizes were determined based on prior lab experiences on the number of mice needed for statistical significance, as well as the availability of mice. Samples were randomly allocated to experimental treatment groups. Statistical analyses were performed with Prism 9. Data figures were created with Prism 9.

Results

Single cell RNA sequencing of skin lymphocytes

We started our investigation with an unbiased approach to examine the factors produced or pathways present in skin T cells that might be important in mediating TSLP-driven adipose loss and sebum secretion. To this end, we overexpressed TSLP by injecting wildtype (Wt) mice with a TSLP-expressing adeno-associated virus serotype 8 (AAV) and harvested the ear skin 10 days post-injection. We then isolated individual cells from the ears, pooled these cells together from each group, and sorted live, CD45.2⁺CD90.2⁺ lymphocytes for sequencing. Paired 10x single cell 5’ RNA sequencing (scRNAseq) and T cell receptor (TCR) sequencing was then performed (experiment schematic shown in Fig 1, A). Uniform manifold approximation and projection (UMAP) analysis with annotation by scType showed that T cells represented the vast majority of cells that were sequenced (Fig E1, A and B). Interestingly, with TSLP treatment, there was a relative increase in the proportion of skin effector T cells and a relative decrease in the proportion of skin naive B and T cells, basophils, and natural killer cells (Fig E1, A and B).

TCR sequencing and clonotype abundance analysis showed that the diversity of the TCRs in T cells from both the control and TSLP groups were relatively similar, with T cells from the TSLP group possessing a decrease in expanded clones and therefore higher TCR diversity compared to control mice (Fig 1, B). This aligns with our prior findings that an activated T cell state, but not antigen specificity, is required for the effects of TSLP (24). We also previously showed that FTY720 treatment, which inhibits T cell migration, renders mice insensitive to TSLP, indicating that T cell migration is required for the effect of TSLP on sebum secretion (24). Thus, TSLP causes recruitment of migratory T cells in an antigen-independent manner that likely results in increased TCR diversity in the skin.

Since the effect of TSLP on sebum secretion is mediated by TCRαβ T cells (24), we restricted our gene expression analysis to cells expressing at least one α or β TCR chain to increase the specificity of cell selection while preserving as much of the dataset as possible. Manual annotation of the shared nearest neighbors clustering and UMAP plot of these cells (Fig 1, C and Fig E2) showed that indeed, a large proportion of the cells were of conventional TCRαβ lineage, with a small proportion being other T cell types including gamma-delta T cells (cluster 10) and mucosal-associated invariant T cells (MAITs, cluster 7). Of note, clusters 12 and 14 consisted of cells that did not fall under typical labels. Cells in both unclassified clusters appeared low for Cd3d, but those in cluster 12 still expressed Th2 cytokines with mixed low expression of Cd4 and Cd8b1 while those in cluster 14 seemed to be a mix of different Th types (Fig E2). TSLP drove the upregulation of the specific distinguishing clusters 0, 3, 8, and 9, which were not seen or were underrepresented in the control group (Fig 1, D). Heatmap analysis of the top differentially expressed genes per cluster (Fig 1, E) showed that the most distinguishing genes in cluster 0 were shared by cells in the other TSLP-specific clusters 3, 8, and 9. Among the TSLP-specific clusters, clusters 0 and 3 had more in common with each other, while clusters 8 and 9 were more similar to each other.

To delineate which pathways might be upregulated in the TSLP-specific T cell clusters, we looked at the top differentially expressed genes (DEGs) in the combined clusters of 0, 3, 8 and 9 when compared to the combination of all other clusters (Table E1). We then performed gene ontology (GO) enrichment analysis on these TSLP-specific cluster DEGs and found that the T helper 2 (Th2) differentiation pathway was one of the top three upregulated GO pathways (Table E2). As Th2 differentiation is primarily associated with CD4⁺ T cells, the TSLP-specific clusters appeared to be highly enriched in CD4⁺ T cells (Fig E3, A). GSEA of the DEGs between CD4⁺ T effector cells (clusters 0, 1, 2, 3, 8, 9, and 13) in the TSLP condition versus the control condition showed that CD4⁺ T cells from the TSLP group were more likely to upregulate genes that are upregulated by cells of the Th2 lineage when compared to naïve CD4⁺ T cells (Fig 1, F). This was confirmed by targeted analysis of Th2 genes, which showed that although cells in TSLP-specific clusters expressed similar levels of the TSLP receptor (TSLPR) Crlf2 when compared to other clusters, they expressed higher levels of the Th2 genes Gata3, Il4, Il13, and Il5, which was not seen in other clusters (Fig 1, G). Flow cytometric analysis of Gata3 expression in skin CD4 T cells also confirmed that there were increased Th2 cells in the skin of mice given TSLP (Fig E3, B and C).

IL-13 and/or IL-4 mediate TSLP-induced adipose loss and sebum hypersecretion

Based on findings from the transcriptional analysis of T cells, we next examined the role of IL-4 and 13 in driving TSLP-induced adipose loss and sebum hypersecretion. We previously reported that IL-4-deficient mice are susceptible to TSLP-mediated adipose loss (24). However, we reasoned that IL-4 might be dispensable, because IL-4 and 13 play redundant roles by signaling through STAT6. To test this possibility, we injected control or TSLP-AAV into Stat6^−/− mice and measured the epididymal white adipose tissue (eWAT, visceral fat) mass 2 weeks later. We also performed hair lipid extraction from the mice and analyzed the lipid class composition using thin layer chromatography (TLC). Wax esters are sebum-specific and thus can be used as a marker for sebum production in mice. Stat6^−/− mice did not lose adipose or hypersecrete wax esters when injected with TSLP-AAV (Fig 2, A and B), suggesting that STAT6-dependent signaling is important for this process.

Fig 2.

T cell derived IL-4 and 13 are necessary for TSLP-induced adipose loss and sebum secretion (A) eWAT masses and (B) TLC quantification of lipids extracted from hair of Stat6^−/− mice, 2 weeks post AAV (N = 4-5 mice/group, 1 representative of 3 experiments). CE = cholesterol esters, WE = wax esters, FFA = fatty acids, and FC = free cholesterol. (C) eWAT masses and (D) TLC quantification of lipids extracted from hair of Il4/13^fl/fl and Vav^CreIl4/13^fl/fl mice, 2 weeks post AAV (N = 3 mice/group, 1 representative of 3 experiments). (E) eWAT masses and (F) TLC quantification of hair lipids from Rag2^−/− mice transferred with T cells from either of Il4/13^fl/fl or Vav^CreIl4/13^fl/fl mice, 2 weeks post AAV (N = 3mice/group, 1 representative of 2 experiments). N.S. = not significant, *P<0.05, **P<0.01, ***P<0.001, ****P0.0001 by Student t-test. Data are shown as mean ± SEM.

To specifically test the role of IL-4 and 13, we crossed Il4/13^fl/fl to Vav^Cre mice to generate Vav^CreIl4/13^fl/fl mice, which lack IL-4 and IL-13 expression specifically in hematopoietic cells. Il4/13^fl/fl mice given TSLP-AAV showed an increase in IL-4 and IL-13 expression in the skin, which was abolished in Vav^CreIl4/13^fl/fl mice (Fig E4, A and B). Furthermore, while Il4/13^fl/fl control mice given TSLP-AAV displayed decreased eWAT mass and increased sebaceous gland wax ester secretion, TSLP did not cause eWAT loss or increased wax esters in Vav^CreIl4/13^fl/fl mice (Fig 2, C and D). This suggested that IL-4 and/or IL-13 produced by hematopoietic cells is critical for TSLP-induced adipose loss and sebum secretion. Importantly, the accumulation of TCRαβ T cells in the skin was not affected in Vav^CreIl4/13^fl/fl mice, as both TCRαβ T cell percentages and numbers in the skin were increased with TSLP-AAV, similarly to I14/13^fl/fl control mice (Fig E5, A–C).

We next examined whether TSLP-induced IL-4 and 13 expression specifically in T cells drives adipose loss and sebum secretion. To this end, we adoptively transferred T cells from Il4/13^fl/fl mice or Vav^CreIl4/13^fl/fl mice into T cell-deficient Rag2^−/− mice. 4 weeks after reconstitution, we injected the mice with either Control or TSLP-AAV. Rag2^−/− mice transferred with Il4/13^fl/fl T cells displayed TSLP-induced eWAT loss and increased wax esters secretion, whereas Rag2^−/− mice transferred with Vav^CreIl4/13^fl/fl T cells did not lose eWAT or hypersecrete wax esters (Fig 2, E and F). Although T cells may not be the sole source of IL-4/13, this signifies that the production of IL-4 and 13 in T cells downstream of TSLP is necessary for adipose loss and sebum secretion.

We next investigated whether IL-13 alone is sufficient to induce adipose loss and sebum secretion. Overexpression of IL-13 was achieved by injection of IL-13-AAV, which resulted in serum IL-13 levels of ~130 pg/mL (Fig E6, A). Similar to TSLP-AAV-injected mice (24), mice given IL-13-AAV displayed fat loss, wax ester hypersecretion, smaller sebaceous glands on histology, and exhibited increased Ki67 staining of sebaceous gland basal stem cells (Fig 3, A–E and Fig E6, B). This suggests that IL-13 promotes turnover of sebocyte progenitors to replenish the terminally differentiated sebocytes that undergo holocrine rupture and thus sebum secretion. The effect of IL-13 was independent of TSLP signaling, as Tslpr^−/− mice were still susceptible to IL-13-driven adipose loss and wax ester hypersecretion (Fig 3, F and G), suggesting that IL-13 lies downstream of TSLP-mediated T cell activation. Consistent with this notion, IL-13 did not cause an increase in skin T cell numbers (Fig E6, C and D). Concordantly, Ebeta^−/− mice, which lack TCRαβ T cells and are resistant to TSLP-induced adipose loss and sebum secretion, were still susceptible to IL-13 and displayed eWAT loss and wax ester hypersecretion when given IL-13-AAV (Fig 3, H and I). In contrast to Ebeta^−/− mice, Stat6^−/− mice were resistant to IL-13-AAV and did not lose any adipose mass or hypersecrete wax esters (Fig 3, J–K). Together, these results suggest that TSLP acts on T cells to stimulate IL-4 and 13 production, which then mediates adipose loss and sebum secretion through STAT6 signaling.

Fig 3.

IL-13 is sufficient to drive adipose loss and sebum hypersecretion. (A) eWAT masses and (B) TLC quantification of hair lipids from wildtype (WT) mice, 2 weeks post AAV (N = 3 mice/group). (C) Quantification of sebaceous gland (SG) area in WT mice, 2 weeks post AAV (n = 105 Ctrl and 95 TSLP SGs from 3 mice/group). (D) Ki67 staining (red arrows indicate Ki67+ cells) and (E) quantification of SGs in WT mice, 2 weeks post AAV (N = 3 mice/group). (F) eWAT masses and (G) TLC quantification of hair lipids from Tslpr^−/− mice, 2 weeks post AAV (N = 3 mice/group). (H) eWAT masses and (I) TLC quantification of hair lipids from Ebeta^−/− mice, 2 weeks post AAV (N = 3 mice/group). t-test. (J) eWAT masses and (K) TLC quantification of hair lipids from Stat6^−/− mice, 2 weeks post AAV (N = 4 mice/group). For each experiment, 1 representative of 2 experiments is shown. N.S. = not significant, *P<0.05, **P<0.01, ****P<0.0001 by Student t-test. Data are shown as mean ± SEM.

IL-4 and IL-13 stimulate sebocyte lipogenesis and proliferation

To understand how IL-4 and 13 affect sebocytes, we utilized the SEB-1 cell line, a human sebocyte cell line known to express characteristic sebocyte proteins and to accumulate lipid droplets in the cytoplasm (39). To first assess the effect of IL-4 and 13 on SEB-1 lipogenesis, we grew SEB-1 cells until confluency, and then stimulated the cells with lipogenesis media either with or without IL-4 and 13. We found that SEB-1 cells treated with IL-4 and 13 displayed increased lipid droplet formation comparable to that seen with insulin treatment, as measured by LipidTOX staining and flow cytometry (Fig 4, A and Fig E7). LipidTOX stains neutral lipids such as triglycerides, which are the major component of sebum in humans, as opposed to wax esters in mice. This increase in triglycerides with IL-4 and 13 treatment was also confirmed by TLC on lipid extracts from treated SEB-1 cells (Fig 4, B). We further found that IL-4 and 13 also increased SEB-1 sebocyte cell numbers over time (Fig 4, C), which is consistent with the increased Ki67 staining of basal sebocytes seen with both TSLP (24) and IL-13-AAV (Fig. 3, E).

Fig 4.

IL-4 and 13 induce lipogenesis and proliferation of SEB-1 cells. (A) LipidTOX flow staining of SEB-1 cells treated with media alone, media with insulin, or media with IL-4 and 13 (N = 3 samples/group, 1 representative of 2 experiments). MFI = mean fluorescence intensity. (B) TLC triglyceride (TG) levels of SEB-1 cells treated with or without IL-4 and 13 (N = 6 samples/group, 1 representative of 3 experiments). (C) Crystal violet optical density (OD) time course over 5 days of treatment with or without IL-4 and 13 (N = 5 samples/group, 1 representative of 3 experiments). N.S. = not significant, *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 by Student t-test. Data are shown as mean ± SEM.

In vivo analysis of sebaceous gland gene expression

To further understand the effects of TSLP on sebocytes in vivo, we performed laser capture microdissection (LCM) to isolate sebaceous glands from formalin-fixed, paraffin-embedded skin sections from mice given control or TSLP-AAV at 7 days and 14 days post-injection. We then extracted RNA and performed RNA sequencing to determine sebaceous gland gene expression. Sebaceous glands occupy only a small proportion of the skin area, so this technique allowed us to specifically examine gene expression in sebaceous glands. At Day 7 post-injection, we found a clear separation of control vs. TSLP-treated samples by principal component analysis (PCA) with 202 upregulated DEGs and 97 downregulated DEGs (Fig 5, A and Fig E8, A). Many of the Day 7 DEGs were associated with cell death and lipid metabolism pathways (Fig 5, B). Accordingly, GO term analysis showed that there was an upregulation of lipid metabolism and cell death pathways in TSLP-treated sebaceous glands (Fig 5, C). GO term analyses also showed downregulation of some lipid metabolism pathways in the TSLP samples. However more GO lipid metabolism pathways were upregulated than downregulated, and the upregulated pathways contained more DEGs known to be involved in sebocyte lipid metabolism (such as Fads3, Agpat1, Dgat216). GSEA showed that there was downregulation of gene pathways known to be downregulated during adipocyte differentiation in the TSLP samples (Fig 5, D), suggesting that TSLP may promote pathways associated with adipocyte differentiation in sebocytes.

Fig 5.

TSLP promotes lipogenesis at day 7 and cell death at day 14 in sebaceous glands in vivo. LCM of isolated skin sebaceous glands and RNAseq analysis at 7 (Fig 5, A–D, n = 3 samples/group) and 14 (Fig 5, E–J, n = 5 samples/group) days after control vs. TSLP-AAV injection. (A) Principle component analysis (PCA) of control (blue) vs. TSLP (red) samples. (B) Heat map of DEGs, zoomed in on cell death and lipid metabolism genes. (C) Upregulated GO terms induced by TSLP. (D) GSEA plot representing a downregulated pathway induced by TSLP, identified via the C2 curated MSigDB database. (E) PCA of control (blue) vs. TSLP (red) samples. (F) Heat map of DEGs, zoomed in on cell death and lipid metabolism genes. (G) Upregulated GO terms induced by TSLP. (H) Downregulated GO terms induced by TSLP. (I) GSEA representing enrichment of cell death genes, identified via the Reactome database. (J) GSEA representing downregulation of fatty acid metabolism genes with TSLP, identified via the KEGG database. All GO terms were identified from the GO knowledgebase; FDR adj-p-val < 0.05; number listed beside each bar corresponds to number of DEGs overlapping with genes in term dataset.

At 14 days post-AAV, RNA-seq analysis also showed clear separation of control vs. TSLP treated sebaceous glands by PCA (Fig 5, E). There were 576 upregulated DEGs and 156 downregulated DEGs (Fig E8, B), many of which were again associated cell death and lipid metabolism (Fig 5, F). GO pathway analysis at day 14 showed that there was strong upregulation of cell death pathways but significant downregulation of lipid metabolism pathways (Fig 5, G–H). This was also seen with GSEA analysis, which showed enrichment of programmed cell death pathways (Fig 5, I) and decreased expression of fatty acid metabolism genes at day 14 (Fig 5, J). Overall, these data suggest that at day 7 of TSLP treatment, lipid metabolism and cell death is enhanced as TSLP induces lipogenesis and holocrine secretion in sebocytes. By day 14, lipid metabolism has been downregulated and cell death is further upregulated, as sebocytes undergo holocrine secretion and die. These findings are consistent with our previous data showing that TSLP-AAV injection induces an increased proportion of mature sebocytes that lack nuclei (24), as they die and release intracellular lipid contents into the hair follicle.

Discussion

The results presented here support a model in which TSLP stimulates enrichment of Th2 T cells in the skin that produce IL-4 and 13 to promote sebaceous gland lipogenesis at early timepoints (7 days post-AAV) and then subsequent holocrine secretion at later timepoints (14 days post-AAV). IL-13 alone phenocopies both the metabolic and sebaceous gland effects of TSLP in vivo. Treatment of SEB-1 sebocytes with IL-4 and IL-13 leads to increased lipogenesis and proliferation while in vivo analysis of sebaceous glands from TSLP-treated mice shows increased lipogenesis and then subsequent cell death of mature sebocytes, with concomitant proliferation of sebocyte stem cells. The difference between these in vitro and in vivo findings is likely because cultured SEB-1 cells better represent sebaceous gland stem cells that may not fully recapitulate the in vivo sebocyte differentiation profile (39), whereas in vivo sebaceous glands are comprised of a mix of basal, mature, and differentiating sebocytes.

Although dysregulated IL-4 and 13 play a pathogenic role in atopic diseases, these cytokines are the hallmark cytokines of Th2 immunity and have been shown to play important roles in skin, lung, and gut barrier site immunity, defending the host against helminths and parasites (40). These cytokines mediate their effects by inducing immunoglobulin E (IgE) secretion, stimulating mucous production, triggering the release of effector molecules such as antimicrobial peptides (AMPs), recruiting immune cells to sites of damage, and promoting smooth muscle contractility (41). Other host-protective responses of Th2 immunity include inducing itch and scratching to remove parasites and irritants, and sensitizing neurons to histamine and IL-31 (42). Thus, it makes sense that IL-4 and 13 also control sebum secretion, as sebum plays a key role by contributing lipids, antioxidants, vitamin E and, antimicrobial peptides to promote skin and immune barrier defense (4, 43–45). Previous studies in humans have also shown that sebaceous gland rich areas of the skin have higher TSLP expression, increased AMP production, distinct microbiomes, and a tolerogenic, homeostatic surveillance immune profile (46, 47). Our findings suggest that IL-4 and 13 may act downstream of TSLP to stimulate sebaceous gland lipid synthesis and subsequent sebocyte turnover to maintain these unique characteristics of sebaceous gland rich skin.

Type 2 immune cells are also primed to be sensors and regulators of the body’s nutritional and metabolic status, so can thus coordinate maintenance of metabolic homeostasis. For example, in adipose tissue, type 2 cytokines regulate adipocyte proliferation and differentiation, promote beige adipocyte function, and induce thermogenic caloric expenditure. IL-4, IL-5, and IL-13 secreted by type 2 immune cells suppress adipose tissue inflammation, control glucose and lipid metabolism, promote insulin sensitivity, and stimulate adipose browning (48–51). In the liver, IL-13 has also been shown to signal through STAT3 to directly inhibit hepatic gluconeogenesis and lipogenesis and increase energy expenditure (52).

Although sebaceous glands have not traditionally been thought of as a metabolic organ, the sebaceous gland is remarkably similar to adipose tissue and produces many of the same lipids that adipose tissue produces, with the exception of wax esters and squalene, which are specific to sebum (53). Sebaceous glands are found throughout the entire human body except on the palms and soles. Sebaceous glands undergo a unique maturation process of holocrine secretion whereby peripheral stem-like cells detach from the basal layer and differentiate into mature sebocytes by accumulating lipids, and when these maturing sebocytes reach the center of the gland, they disintegrate and release their intracellular lipid contents into the follicle through the process of holocrine secretion. Although the adipocytes do not undergo holocrine secretion, the biology of both cell types closely parallels each other. Both cell types undergo lipogenesis and express many of the same enzymes such as liver x receptor, stearoyl-CoA desaturase 1, and diacylglycerol acyltransferase. Sebocytes also uptake lipids from the circulating lipid pool through the low density lipoprotein (LDL) receptor and fatty acid transporter 4 (FATP4) proteins, so there is likely close cross communication between the adipose tissue and skin sebaceous glands (43). Thus, it is not completely unexpected that type 2 cytokines such as IL-4 and 13 which play integral roles in adipose tissue function might also play major roles in sebocyte metabolism, lipid production, and holocrine secretion.

Not much is known about the roles of IL-4 and 13 in sebocyte function, but one recent study showed that IL-4 and 13 drive skin lipid abnormalities through the expression of the enzyme 3β-hydroxysteroid dehydrogenase. The authors found that IL-4 and 13 enhance androgen synthesis and thereby affect sebaceous gland production of triglycerides, diglycerides, and sphingomyelin (54). Although we did not find a role for androgens in our model, we found that TSLP-driven T cell derived IL-4 and 13 promote sebaceous gland maturation, lipogenesis, turnover, and ultimately, holocrine secretion. Although non-T cell sources of IL-4 and IL-13 could be present in the skin following TSLP overexpression, our data suggest that these sources alone are not sufficient for TSLP-induced sebum secretion and subsequent adipose loss. One explanation could be that given the migratory nature of T cells, T cells may function as the carriers and delivery vehicles of IL-4 and 13 to specific microenvironments such as sebaceous glands, while preventing the overt systemic inflammation that can be driven by these cytokines when localization is not controlled. There likely remain complex and unexplored roles in the interplay between T cells, type 2 immunity, skin barrier function, and sebocyte biology that help promote healthy skin and metabolic homeostasis.

Clinical Implication:

Thus, beyond causing skin inflammation, IL-4 and IL-13 can also promote sebum secretion, which is an important component of the skin barrier.

Abbreviations:

AAV

Adeno-associated virus serotype 8

DEG

Differentially expressed genes

eWAT

epididymal white adipose tissue

GO

gene ontology

GSEA

Gene set enrichment analysis

ILC

innate lymphoid cell

LCM

Laser capture microdissection

PCA

principal component analysis

scRNAseq

single cell 5’ RNA sequencing

SREBP

Sterol regulatory element binding proteins

TC

T cell receptor

Th2

T helper 2

TLC

Thin layer chromatography

TSLP

Thymic stromal lymphopoietin

UMAP

Uniform manifold approximation and projection

Wt

wildtype

Data and materials availability:

All data are available in the main text or the supplementary materials.