Interleukin-17 governs hypoxic adaptation of injured epithelium

Skin-resident RORγt⁺ cells direct wound re-epithelialization

To define the composition of first responder lymphocytes in an unbiased manner, we performed cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) (11). CD45⁺CD90⁺ lymphocytes were purified and profiled from unwounded skin and at day 3 (D3) and D5 after full-thickness wounding (fig. S1, A and B). Principal component analysis (PCA) and uniform manifold approximation and projection (UMAP) plots with unsupervised analysis of 4923 cells revealed 12 clusters with a clear segregation of effector and regulatory T (T_reg) lymphocytes (fig. S1C). Because lymphocytes are notorious for low-abundance transcripts, we used surface protein epitopes alongside transcripts for definitive cluster annotations (fig. S2, A to C). T helper and cytotoxic T (T_H/T_C) cell clusters 1 to 3, γδ T cell and mucosal-associated invariant T (γδ T/MAIT) cell cluster 1, the natural killer and natural killer T (NK2/NKT) cell cluster, and T_reg cell clusters 2 and 3 were enriched in wounded skin compared with unwounded controls (Fig. 1A and fig. S2D). Thus, contrary to expectations, a dominant repair-associated lymphocyte population does not emerge. Rather, many different immune cell populations are substantially remodeled in acute injury.

We next examined the expression of cytokines that typify type 1, type 2, type 17, and T_reg archetypes (12). Of these, the type 17 cytokines Il17a and Il17f and the type 1 cytokine Ifng were enriched in wounded samples (Fig. 1B). Type 17 immunity encompasses different cell types controlled by a master transcription factor RORγt, which has also been implicated in repair (13, 14). Indeed, RORγt⁺ lymphocytes were enriched in both the human and murine epithelial wound front (Fig. 1, C and D, and fig. S3A). Rorgt-EGFP mice have enhanced green fluorescent protein (EGFP) knocked into the Rorgt locus, rendering it nonfunctional (15). Using homozygous Rorgt-EGFP mice (here-after called GFP-KI), we measured gross wound closure in the absence of RORγt. Both genetic loss of RORγt in GFP-KI mice and topical administration of an RORγt inhibitor, digoxin, resulted in a 40 to 50% reduction in wound closure rate compared with respective controls (fig. S3, B and C) (16).

Human skin heals exclusively by re-epithelialization, a process reliant on epithelial proliferation and subsequent migration into the wound bed. By contrast, mice heal by re-epithelialization and myofibroblast-mediated dermal contraction (17). We therefore secured murine wounds with silicone splints to eliminate dermal contraction. Microscopic visualization revealed a specific impairment of the keratin 14⁺ (K14⁺) integrin α5⁺ migrating epithelial tongue, but not epithelial proliferation, in GFP-KI splinted wounds (Fig. 1E and fig. S3D). Similarly, epithelial migration was markedly compromised in unsplinted GFP-KI wounds relative to controls (Fig. 1E).

Given the rapid enrichment of RORγt lymphocytes in wounds, we postulated that resident RORγt lymphocytes locally expand to fuel repair. Indeed, wound RORγt⁺ cells had heightened expression of cell-cycle genes (Mki67, Cdk1, and Top2a) and 5-ethynyl-2′-deoxyuridine (EdU) incorporation relative to unwounded skin and to their systemic lymph node counterparts (Fig. 1, F and G, and fig. S3E). To determine whether skin RORγt⁺ lymphocytes were sufficient to instigate repair, we treated mice with FTY720, a sphingosine-1 phosphate receptor agonist that inhibits lymphocyte egress into blood by “trapping” them within lymph nodes (fig. S3F) (18). FTY720 blockade did not alter the frequency of wound RORγt cells or epithelial migration (Fig. 1H and fig. S3G). Thus, the expansion of preexisting skin RORγt⁺ cells is sufficient to drive re-epithelialization.

IL-17RC–mediated γδ T cell–epithelial cross-talk controls re-epithelization

We mapped the dynamics of skin RORγt subsets over the course of repair using multi-parameter flow cytometry. Innate-like γδ T cells, MAIT (MR1 tetramer⁺), and invariant NKT (iNKT) cells (CD1d tetramer⁺) expanded first at D1 after wounding (fig. S4, A to C). Adaptive T_H17, T_C17, and RORγt⁺ T_reg cells subsequently increased between D3 and D5. Consistent with our CITE-seq analysis, the proportions of RORγt⁺ type 3 innate lymphoid cells (ILC3s) were unchanged by injury, suggesting that in the presence of adaptive immunity, ILC3s may not play a dominant role in repair (figs. S2D and S4, A to C).

To pinpoint which RORγt subset(s) was positioned at the re-epithelializing wound front, we turned to spatial transcriptomics (ST), a technology that enables spatially resolved transcriptional profiling of tissues at 50-μm resolution (8) (fig. S5, A to C, and table S4). PCA and UMAP projections distinguished seven clusters (tissue regions) in unwounded skin and six clusters in wounded skin (Fig. 2, A and B). The identified clusters were consistent with histological regions and genes that typified tissue architecture (Fig. 2, A and B, and fig. S5, D and E). Wound-edge epithelium was discernable by Krt14 and Krt17 expression (fig. S6A). Nevertheless, to confirm the identity of epidermal wound edge clusters and validate our ST data, we performed multimodal intersectional analysis (MIA) (8), a method that can infer cell-type enrichment in tissue regions. Integrating an independent scRNA-seq dataset (19), which defined wounded epithelium as growth arrest cells and wound distal epithelium as Col17a1⁺ cells, with our ST data revealed a significant degree of overlap between Col17a1⁺ cells and wound distal epithelial cluster 3 (P = 2.3 × 10⁻¹⁶) and growth arrest cells and wound-edge epithelial cluster 4 (P = 5.4 × 10⁻³⁵) (fig. S6, B and C).

We next leveraged MIA to resolve the location of wound lymphocyte populations identified by our CITE-seq data (Fig. 1A). MIA of the top 300 up-regulated genes (≥0.25 log FC) from each effector lymphocyte CITE-seq cluster and each ST cluster simultaneously located all of our lymphocyte populations in wounded skin and revealed a specific enrichment of γδ T/MAIT cluster 1 (P = 7.9 × 10⁻⁶) and cluster 2 (P = 1.9 × 10⁻⁵) within wound-edge epithelial cluster 4 (Fig. 2C and fig. S6D).

Both γδ T cells and MAIT cells have been implicated in wound repair (20). However, systematic comparisons between these and other RORγt⁺ cell subsets in the context of re-epithelialization are lacking. We therefore analyzed mice specifically lacking RORγt⁺ γδ T cells (Tcrd^CreER;Rorc^fl/fl), MAIT cells (Mr1^−/−), and T_H17/T_C17/MAIT/iNKT cells (Cd4^Cre;Rorc^fl/fl) (fig. S6E). After tamoxifen-induced depletion, Tcrd^CreER;Rorc^fl/fl mice exhibited a shorter migrating tongue than control animals, indicating a nonredundant role for RORγt⁺ γδ T cells in wound re-epithelialization (Fig. 2D). Wounds from both Mr1^−/− and Cd4^Cre; Rorc^fl/fl mice had migrating tongues comparable to their respective controls, underscoring the unicity of RORγt⁺ γδ T cells in directing re-epithelialization (Fig. 2, E and F).

RORγt⁺ γδ T cells exert potent effector function by producing the cytokines IL-17A and IL-17F, which provide instructive signals to the surrounding parenchyma (14, 21). Of all the effector lymphocytes analyzed by CITE-seq, γδ T/MAIT cell clusters expressed the highest level of these cytokines at D3 after wounding, hinting at their role in the RORγt⁺ γδ T cell–epithelial cross-talk at the wound’s edge (Fig. 2G). Accordingly, intradermal recombinant murine IL-17A (rmIL-17A) administration reversed the re-epithelization defect of GFP-KI mice (Fig. 2H).

IL-17A/F signaling is mediated by an IL-17 receptor A (IL-17RA)–IL-17RC heterodimer, whereas IL-17F can also bind an IL-17RC/RC homodimeric complex (14, 22). ST plots showed increased Il17ra and Il17rc transcripts at the wound’s edge, with the most prominent expression being Il17rc (Fig. 2I). We therefore generated and wounded epithelial-specific IL-17RC–deficient mice (Krt14^Cre;Il17rc^fl/fl labeled Il17rc^EKO) (fig. S6F). Compared with IL-17RC–sufficient mice, epithelial-specific loss of Il17rc mirrored the re-epithelialization defect of GFP-KI and Tcrd^CreER;Rorc^fl/fl mice (Fig. 2J). Thus, IL-17A/F supplied by RORγt⁺ γδ T cells directly engages epithelial IL-17RC to induce migration and re-epithelialization.

HIF1α is a re-epithelialization factor controlled by RORγt⁺ cells

PCA and UMAP plots integrating wild-type (WT) and GFP-KI tissue sections from D3 wounds discerned 15 clusters with high reproducibility between biological replicates (Fig. 3, A and B, and fig. S7, A and B). We focused on wound edge clusters 7 and 8 on the basis of histology and expression of the re-epithelialization factors Gjb2, Lamc2, Mmp9, Mmp13, and Stfa1 (fig. S7A). We further delineated cluster 7 as wound-edge basal epithelium (Krt5, Krt6, Krt14, and Krt17) and cluster 8 as wound-edge differentiated epithelium (Krt10, Lce1a1, and Lor). Pathway analysis of WT and GFP-KI cluster 7 differentially expressed genes revealed enrichment of IL-17 signaling, skin/epidermal development, epidermal differentiation, epithelial cell/tissue migration, and T cell activation pathways in WT (Fig. 3C and fig. S8A). HIF1 signaling, which is typically associated with tissue hypoxia, also emerged as a differentially expressed pathway (Fig. 3C) (10), with Hif1a transcripts higher in WT compared with GFP-KI wound edges (Fig. 3D).

To validate this finding, we analyzed the transcriptomes of integrin α5 and integrin α6 double-positive wound-edge epithelium (migrating tongue) and integrin α6⁺-only wound distal epithelium, from WT and GFP-KI wounds (fig. S8B). We observed very modest differences (62 differentially expressed genes, adjusted P < 0.1) between WT and GFP-KI wound distal epithelium (fig. S8C). By contrast, WT and GFP-KI migrating tongue cells had 1548 differentially expressed transcripts (adjusted P < 0.1) (fig. S8D). HIF1α signaling and associated glycolysis pathways were significantly enriched in WT samples (fig. S8E).

Consistent with the enrichment of Hif1a transcripts, WT migrating epithelial tongue robustly expressed nuclear HIF1α, which was absent from GFP-KI wound fronts (Fig. 3E and fig. S9, A and B). GFP-KI mice did not compensate for the loss of HIF1α by up-regulating HIF2α (fig. S9D). Furthermore, Il17rc^EKO mice phenocopied the HIF1α defect observed in GFP-KI wounds (Fig. 3F). Nuclear HIF1α was also robustly expressed in the migrating epithelial front of human wounds, suggesting a highly conserved role for this transcription factor in re-epithelialization (fig. S9C).

HIFs ubiquitously mediate cellular adaptation to low-oxygen environments (23). Therefore, we measured hypoxia in WT and GFP-KI wounds using both a fluorescent hypoxia marker, pimonidazole (PIM), and a fiber probe to directly measure tissue O₂. WT and GFP-KI wounds had equivalent PIM signals and O₂ levels (~10 to 12 mmHg or 2% O₂) (Fig. 3, G and H). Moreover, WT and GFP-KI epithelia expressed comparable levels of prolyl hydroxylases (Egln1, Egln2, and Egln3) and Vhl ubiquitin ligase (fig. S9D). Thus, wound-edge epithelium failed to sustain HIF1α without RORγt⁺ cell–derived signals even in the presence of tissue hypoxia and while having the necessary regulatory machinery.

We next generated and wounded epithelial-specific HIF1α-deficient mice (Krt14^Cre; Hif1a^fl/fl labeled Hif1a^EKO) (fig. S9E). Loss of epithelial HIF1α impaired the development of migrating tongue (Fig. 3I). Additionally, unlike GFP-KI mice, rmIL-17A failed to rescue the re-epithelialization defect of Hif1a^EKO mice (Fig. 2H and fig. S9F). Thus, epithelial HIF1α is necessary to respond to IL-17A signals and activate a program of migration.

IL-17RC signaling induces HIF1α through extracellular signal–regulated kinase 1/2 (ERK 1/2)-AKT–mammalian target of rapamycin (mTOR)

To test our hypothesis that γδ T cell–derived IL-17 serves as a second signal to activate HIF1α, we intradermally injected rmIL-17A, rmIL-17F, or rmIL-22 into unwounded skin. Administration of rmIL-17A and, to a lesser degree, rmIL-17F was sufficient to induce epidermal HIF1α activation and hyperplasia (Fig. 4A and fig. S10, A and B). IL-22 marginally increased epidermal thickness but did not up-regulate HIF1α (fig. S10, A and B). Skin epithelial organoids cultured for 5 days in the presence of rmIL-17A were significantly larger than controls and robustly expressed HIF1α (Fig. 4B and fig. S10, C to F). Other damage-associated cytokines [rmIL-1, rmIL-6, and recombinant murine tumor necrosis factor α (rmTNFα)] did not activate HIF1α (fig. S10G). Epidermal growth factor also failed to augment HIF1α to the levels observed after rmIL-17A treatment, underscoring the specificity of IL-17 in amplifying this pathway (fig. S10H).

IL-17RA and IL-7RC were robustly expressed in our epithelial organoids and wound-edge integrin α5⁺ cells (fig. S10I) (14, 22). Epithelial-specific loss of Il17rc abrogated the ability of rmIL-17A to activate HIF1α both in vivo and in vitro, excluding a role for accessory cells in this cross-talk (Fig. 4, C and D, and fig. S10, J and K). To delineate how epithelial IL-17RC signaling resulted in HIF1α activation, we returned to our comparative GFP-KI and WT wound transcriptomics (Fig. 3, A to C). mTOR signaling, which is implicated in HIF1α regulation, was salient as a key differentially expressed pathway (24, 25). We therefore evaluated ribosomal protein S6^S240/244 phosphorylation, a major downstream substrate of mTOR complex 1. rmIL-17A robustly increased epithelial pS6^S240/244 in WT but not Il17rc^EKO mice and organoids (Fig. 4, E and F, and fig. S10L). rmIL-17A–treated organoids also showed enhanced activation of the upstream kinase phosphorylated S6 kinase^T389 (p-S6K^T389) and phosphorylated mTOR^S2448 (p-mTOR^S2448) (Fig. 5, A and B). Inhibition of mTOR with rapamycin significantly reduced HIF1α protein downstream of IL-17A (Fig. 5, A to C, and fig. S11, A and B). Additionally, Hif1a^EKO and control organoids had comparable sizes and mTOR activation (p-S6^S240/244) after rmIL-17A stimulation, confirming that mTOR is upstream of HIF1α (fig. S11, C and D).

In addition to augmenting HIF1α protein, rmIL-17A stimulation also induced Hif1a gene expression in an mTOR-dependent manner (Fig. 5D). Increased Hif1a transcripts in rmIL-17A–treated organoids were not caused by enhanced mRNA stability, and HIF1α protein was more rapidly degraded in the presence of rmIL-17A (fig. S11, E and F). Thus, mTOR controls HIF1α transcription and protein levels, but not stability, downstream of IL-17A.

To determine whether and how IL-17RC proximal signaling links to mTOR, we next measured early (≤1 hour) mTOR activation after rmIL-17A stimulation and found enrichment of p-S6K^T389 and p-S6^S240/244 (fig. S11G). Working our way upstream, we detected activation of p-ERK1/2^T202/Y204, a canonical mediator of IL-17RC signaling, and protein kinase B (p-AKT ^S473), an upstream mTOR kinase, after IL-17A stimulation (Fig. 5E). Inhibition of either ERK1/2 (U0126) or AKT (MK-2206) after long- and short-term rmIL-17A treatment (5 days and 30 min, respectively), diminished p-S6K^T389, p-S6^S240/244, and HIF1α expression (Fig. 5E and fig. S11H). However, blocking ERK also resulted in a compensatory increase in p-AKT^S473. Dual inhibition of ERK and AKT ablated the mTOR-HIF1α response downstream of rmIL-17A stimulation (Fig. 5E and fig. S11H). Thus, IL-17RC–proximal AKT and ERK signaling directly activates mTOR and consequently induces HIF1α.

Given the prominence of hypoxia at the wound edge, we next examined the IL-17A–mTOR–HIF1α axis in organoids cultured under acute (5 hours) or chronic (5 days) hypoxic conditions (2% O₂). Acute hypoxia was sufficient to stabilize HIF1α, and IL-17A further augmented this response (Fig. 5F). By contrast, and in agreement with our in vivo wounding data, p-S6^S240/244 and HIF1α levels were markedly reduced by long-term hypoxic exposure and rescued by the presence of rmIL-17A (Fig. 5F). Rapamycin treatment abrogated the ability of IL-17A to augment HIF1α in chronic hypoxic cultures (Fig. 5G and fig. S11, I and J). Thus, IL-17A directly and potently induces mTOR to up-regulate HIF1α under both normoxic and hypoxic conditions (Fig. 5H).

IL-17A–dependent metabolic reprogramming is vital for re-epithelialization

Migration, a cardinal feature of re-epithelialization, requires cells to expend energy. However, exactly how migrating epithelial cells meet their energy demands in a hypoxic wound microenvironment is unclear (Fig. 3, G and H). Glycolysis and oxidative phosphorylation are the two major cellular energy–producing pathways (26). Of these, HIF1α is known to control a transcriptional program of glycolysis (10). Accordingly, the WT wound front expressed higher levels of glycolysis-associated genes [hexokinase 2 (hk2), phosphoglycerate kinase 1 (Pgk1), and glucose transporter protein 1 (Slc2a1)] than did that of GFP-KI (Fig. 6A).

rmIL-17A treatment significantly increased glucose transporter 1 (Slc2a1) and glycolytic enzyme expression in epithelial organoids (Fig. 6B and fig. S12, A and B). rmIL-17A–treated organoids had heightened glycolysis activity and glycolytic capacity and secreted higher levels of lactate, an end product of glycolysis, compared with controls (Fig. 6C and fig. S12, C and D). Consistent with a role for mTOR in regulating HIF1α, rapamycin-treated or Hif1a^EKO organoids expressed lower levels of glycolysis enzymes than did controls (Fig. 6, D and E, and fig. S12, E and F). GKP-KI, Il17rc^EKO, and Hif1a^EKO wound-edge epithelia expressed significantly lower glucose transporter 1 levels than their respective controls (Fig. 6, F and G, and fig. S12G). rmIL-17A administration rescued glucose transporter 1 levels in GKP-KI wounds (fig. S12H). Thus, the IL-17A–HIF1α axis induces a program of glycolytic metabolism in epithelial cells.

Our findings raised the possibility that the IL-17A–HIF1α–dependent program of glucose metabolism fuels epithelial migration. We topically treated mice with 2-deoxy-d-glucose (2-DG), a glucose analog that inhibits glycolysis. 2-DG–treated mice had significantly shorter integrin α5⁺ migrating tongues than their control counterparts (fig. S12I). Blocking glycolysis did not alter epithelial cell proliferation, indicating that glycolysis is dispensable for expanding the epithelial cell pool at the wound front (fig. S12J). 2-DG treatment also led to a reduction in wound-edge type 17 cells (fig. S12K). We therefore turned to an in vitro epithelial scratch assay to control for the immune effects of topical 2-DG treatment. 2-DG or HIF1α inhibitor (BAY872243) hindered the ability of human epithelial cells to migrate and seal the in vitro “wound” both in the presence and absence of recombinant human IL-17A (rhIL-17A) (Fig. 6H). Thus, IL-17A augments glycolysis, an essential bioenergetic pathway powering epithelial cell migration, to expedite healing.

Animals

The following mouse strains were purchased from The Jackson Laboratory: C57BL/6J, B6.129P2(Cg)-Rorc^tm2Litt/J (Rorgt-EGFP), B6N.Cg-Tg(KRT14-cre)1Amc/J, B6.Cg-Il17rc^tm1.1koll/J, B6.Cg-Tg(Cd4-cre)1Cwi/BfluJ, B6.129S-Tcrd^{tm1.1(cre/ERT2)Zhu}/J, B6(Cg)-Rorc^tm3Litt/J, B6.129-Hif1a^tm3Rsjo/J. Mr1^{tm1(KOMP)Vlcg} mice were generated from the KOMP ES cell line Mr1^{tm1(KOMP)Vlcg}, RRID:MMRRC_058774-UCD, obtained from the Mutant Mouse Resource and Research Center (MMRRC) at University of California, Davis. Tg(Rorc-EGFP)1Ebe mice were a gift from Dr. G. Eberl (Institut Pasteur). All animal studies were approved by the institutional animal care and use committee (IACUC). Mice were bred and maintained under specific-pathogen-free conditions at the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)–accredited facilities at the New York University Langone Health Center and housed in accordance with the procedures outlined in the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals. Experiments were performed with IACUC-approved protocols. Age- and sex-matched controls were used in all experiments presented.

Human wound samples

Normal and wounded human skin samples were collected from three healthy donors (a 48-year-old female, a 24-year-old female, and a 27-year-old male) at Karolinska University Hospital, Stockholm, Sweden. After local lidocaine injection, two full-thickness excisional wounds were created using a 4-mm biopsy punch at the upper buttock area of each donor. Central excised skin was used as unwounded control. Wound-edge skin was collected using a 6-mm biopsy punch 1 and 7 days later. All donors consented to the collection and use of clinical samples. The study was approved by the Stockholm Regional Ethics Committee and conducted according to the Declaration of Helsinki’s principles.

Skin wound-healing models

Skin tissue digest

Unwounded or 0.25 mm of skin around the wound edge was excised and digested in Liberase TL (Sigma-Aldrich) as previously described (9).

Flow cytometry and fluorescence-activated cell sorting (FACS)

Single-cell suspensions were preincubated with anti-CD16/32 before staining with surface fluorescent conjugated antibodies and/or oligo-tagged antibodies (table S1) at predetermined concentrations in a 100 μl of staining buffer [phosphate-buffered saline (PBS) containing 5% fetal bovine serum (FBS) and 1% HEPES] per 10⁷ cells. Stained cells were resuspended in 4′,6-diamidino-2-phenylindole (DAPI) in FACS buffer before purification or analysis. For intracellular staining, fixable Live/Dead dye was added before surface antibody incubation, which was performed according to the manufacturer’s instructions for the Foxp3/Transcription Factor Staining Buffer Set (BD Biosciences). Flow cytometry data were acquired on LSRII Analyzers (BD Biosciences) and then analyzed with FlowJo program. FACS was performed using FACSAria Cell Sorters (BD Biosciences).

Immunofluorescence and image analysis

Tissue was fixed, processed, and stained as previously described (9). EdU reactions were performed according to the manufacturer’s directions (Life Technologies). Migrating tongues were determined by measuring the length of K14⁺ integrin α5⁺ cells. The number of wound-edge proliferating epithelial cells was determined by counting EdU⁺ K14⁺ cells from the first hair follicle adjacent to the wound until the wound’s leading edge. To detect PIM, slides were pretreated with a mouse-on-mouse immunodetection kit (Vector Laboratories) and then stained with biotin–anti-PIM adduct antibodies per the manufacturer’s guidelines (Hydroxyprobe). ImageJ was used to quantify PIM and glucose transporter 1 integrated density, a well-established method of measuring fluorescence intensity that accounts for differences in the area of the signal in K14⁺ and integrin α5⁺ migrating tongue. Images are shown with the pseudocolor Fire from ImageJ after they were contrasted equally and background was uniformly removed.

Epithelial organoids were released from basement membrane matrix (BME) with organoid harvesting solution (Cultrex), fixed in 4% paraformaldehyde for 2 hours at 4°C, and then incubated in acetone at −20°C. Organoids were stained with conditions similar to tissue sections and embedded in agarose for imaging. Data were analyzed using ImageJ. Organoid HIF1α intensity was determined using the Single Target Translocation Assay from HSC studio (Thermo Fisher). For the complete list of immunofluorescence antibodies and concentrations used, see table S2.

Microscopy and image processing

Images were acquired with a Zeiss Axioplan2 using a Plan-Apochromat 20×/0.8 air objective or Zeiss LSM 710 laser scanning confocal microscope using Plan-Apochromat 20×/0.75 air and Plan-Apochromat 63×/1.40 oil objectives. Tiled and stitched images of sagittal sections were collected using a Plan-Apochromat 20×/0.75 air objective, controlled by Zen software (Zeiss).

Organoids were imaged using Cell Insight CX7-LZR (Thermo Scientific), and images were processed with HCS Studio Cell Analysis software (Thermo Scientific). Maximal projection z-stacks are presented, and co-localization was interpreted only in single z-stacks. z-stacks were projected using ImageJ. Images were assembled in Adobe Illustrator CC2015.3.

Animal drug treatments

CITE-seq

FACS-purified live CD45.2⁺CD90.2⁺ TCRVg3⁻ cells from unwounded skin at D3 and D5 after wounding were prelabeled with surface-epitope-marking oligo-tagged antibodies and sample-specific oligo-tagged Totalseq-A antibodies (BioLegend; see table S1). Hashed samples were pooled at a 1:1.5:1 ratio of control: D3:D5 before library preparation (Chromium Single Cell 3′ Library, 10X Genomics) and sequenced on an Illumina HiSeq 4000 as 150-bp paired-end reads. Sequencing results were demultiplexed and converted to FASTQ format using Illumina bcl2fastq software. The Cell Ranger Single-Cell Software Suite was used to perform sample demultiplexing, barcode processing, and single-cell 3′ gene counting. The cDNA insert was aligned to the mm10/GRCm38 reference genome. Only confidently mapped, non–polymerase chain reaction (PCR) duplicates with valid barcodes and unique molecular identifiers (UMIs) were used to generate the gene-barcode matrix. Cell Ranger output was further analyzed in R using the Seurat package (32). Surface epitope oligo sequences were merged with cell transcriptome data by matching the cell barcode IDs.

Spatial transcriptomics

Organoid assays

Human primary keratinocyte scratch assay

Primary human keratinocytes were isolated from the foreskin and then cultured in serum-free Keratinocytes Growth Medium KGM-Gold (Lonza, Basel, Switzerland) supplemented with bovine pituitary extract, human endothelial growth factor, bovine insulin, hydrocortisone, gentamicin-amphotericin B (GA-1000), epinephrine, and transferrin.

For scratch migration assays, keratinocytes were allowed to reach confluency. Cells were pretreated with mitomycin C (8 μg/ml) 2 hours before and with 2-DG (40 mM) and BAY872243 (10 μM) 1 hour before scratch assay. Scratches were created by manual scraping of the cell monolayer with a 1-ml pipette tip. Cells were treated with rhIL-17A (100 ng/ml, Peprotech). The scratch area was imaged at 0 and 24 hours using EVOS (Thermo Fisher) and quantified using ImageJ.

Statistics

Data are presented as mean ± SEM. Group sizes were determined on the basis of the results of preliminary experiments. Mice were assigned at random to groups. Experiments were not performed in a blinded fashion. Statistical significance for each experiment was determined as shown in the figure legends, where n = the number of independent biological replicates (animals, unless noted as cells) per group, and N = the number of independent experimental replicates. Statistical significance of bulk RNA-seq data was calculated using an adjusted P value cutoff <0.1 in the DESeq2 R package. Statistical analyses were performed in Prism (GraphPad), DESeq2, or R. No data points were excluded.