Abstract Hyperosmotic stress occurs in several diseases, but its long-term effects are largely unknown. We used sorbitol-treated human fibroblasts in 3D culture to study the consequences of hyperosmotic stress in the skin. Sorbitol regulated many genes, which help cells cope with the stress condition. The most robustly regulated gene encodes serine protease 35 (PRSS35). Its regulation by hyperosmotic stress was dependent on the kinases p38 and JNK and the transcription factors NFAT5 and ATF2. We identified different collagens and collagen-associated proteins as putative PRSS35 binding partners. This is functionally important because PRSS35 affected the extracellular matrix proteome, which limited cell proliferation. The in vivo relevance of these findings is reflected by the coexpression of PRSS35 and its binding partners in human skin wounds, where hyperosmotic stress occurs as a consequence of excessive water loss. These results identify PRSS35 as a key regulator of the matrisome under hyperosmotic stress conditions. __________________________________________________________________ Hyperosmotic stress induces functional alterations of the fibroblast matrisome through induction of the protease PRSS35. INTRODUCTION Local osmolarity changes occur in several human diseases ([46]1), such as inflammatory bowel disease ([47]2, [48]3), dry eye syndrome (DES) ([49]4), irritant contact dermatitis (ICD) ([50]5), and atopic dermatitis (AD) ([51]6–[52]8), and contribute to their pathological features. In ICD and AD, a defect in the epidermal barrier leads to an increase in transepidermal water loss (TEWL) ([53]5, [54]9), which augments the osmotic pressure in the skin. This results in a self-aggravating loop of hyperosmotic stress and dryness. The importance of increased TEWL and the associated skin dryness in AD pathology is reflected by the reduced occurrence of AD in geographic regions with high environmental humidity ([55]10) and the frequent worsening of AD symptoms during the dry winter season ([56]11). In functional in vivo studies, exposure to high humidity rescued inflammation and keratinocyte hyperproliferation in mice that exhibit AD-like features ([57]6). The beneficial effect of high humidity may result from a reduction in proinflammatory cytokine expression, which was induced in dry skin as a consequence of hyperosmolarity ([58]12). Some of these cytokines also affect the underlying fibroblasts and promote fibrosis ([59]12). In addition, fibroblasts are most likely directly affected by hyperosmotic stress, e.g., in response to a major defect in the epidermal barrier as seen after skin wounding ([60]8, [61]13). Consistent with this assumption, moist dressings have long been known to promote wound healing and reduce scarring ([62]14, [63]15). However, the mechanisms underlying the response of dermal fibroblasts to hyperosmotic stress and the functional consequences are still poorly understood. Here, we used a physiologically relevant three-dimensional (3D) culture model and next-generation sequencing to identify genes that are regulated by hyperosmolarity in primary human dermal fibroblasts (HDFs). The most strongly up-regulated gene encodes the poorly studied serine protease 35 (PRSS35), which we identified as an important regulator of the matrisome under hyperosmotic stress conditions. These results demonstrate that fibroblasts respond to hyperosmotic stress with functionally relevant changes of the matrisome and identify a key regulator of the hyperosmotic stress response. RESULTS RNA sequencing identifies genes that are regulated by hyperosmotic stress in HDFs To identify osmoresponsive genes in HDFs, we cultured them in gelatin-based [gelatin methacryloyl (GelMA)] 3D hydrogels to mimic the conditions in the skin, challenged them with 200 mM sorbitol (∼530 mOsm/kg), and analyzed their transcriptome by RNA sequencing (RNA-seq; [64]Fig. 1A). Treatment with 200 mM sorbitol was previously shown to induce robust osmoregulated gene expression changes in fibroblasts ([65]16–[66]18). This sorbitol concentration activated the osmoregulated kinases p38 and c-Jun N-terminal kinase (JNK), as well as the mammalian target of rapamycin targets p70-S6 kinase and eukaryotic translation initiation factor 4E-binding protein 1 (fig. S1A), but it did not have major effects on cell viability, as shown by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay reflecting mitochondrial activity and the lack of obvious changes in cell morphology (fig. S1, B and C). However, higher doses of sorbitol induced cell death (fig. S1C). Fig. 1. Transcriptome changes in HDFs following exposure to hyperosmotic stress. [67]Fig. 1. [68]Open in a new tab (A) RNA-seq results [heatmap with top hits (log[2] FC ≥ 1, FDR < 0.05)] from HDFs embedded in gelatin-based (GelMA) hydrogels and treated with 200 mM sorbitol for 9 or 24 hours (n = 3). (B) Venn diagram showing the overlap of the early and late osmodependent gene response. (C) Top six osmoregulated pathways based on differentially expressed genes at the 9- or 24-hour time points (1 ≤ log[2] FC ≤ −1, P < 0.01) according to Ingenuity Pathway Analysis (IPA). (D) Summary of the osmodependent gene response in different model systems and species. The responses of nine candidate genes identified by RNA-seq to hyperosmotic (sorbitol, salt, and sucrose) or hypoosmotic (urea) stimuli were assessed by quantitative reverse transcription polymerase chain reaction (qRT-PCR) using RNA from cells grown in 2D. HDFs from three different donors and murine primary fibroblasts from two to three wild-type mice were used. Green, significant up-regulation in response to hyperosmotic treatment and down-regulation in response to hypoosmotic treatment; yellow, nonsignificant trend; red, no or opposite response. (E) qRT-PCR for PRSS35 relative to RPL27 using RNA from HDFs in 2D treated with different concentrations of sorbitol or salt for 9 hours (N = 3, n = 6) and kinetics of PRSS35 expression in HDFs treated with 200 mM sorbitol for 3 to 24 hours (N = 3, n = 6). (F) qRT-PCR for PRSS35 using RNA from the dermal equivalent of organotypic human skin cultures treated with 200 mM sorbitol for 9 or 24 hours (N = 3, n = 7 to 10). (G) qRT-PCR using RNA from HDFs seeded on tissue culture polystyrene (TCPS) plates or cultured in 3D GelMA or 3D polyethylene glycol (PEG) hydrogels, treated with 50 mM sorbitol for 9 hours (N = 2, n = 6). Bar graphs indicate means with SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, one-way analysis of variance (ANOVA) with Bonferroni post-test [(E) and (F)] or unpaired t test (G). Red and rosé bars show sorbitol-treated cells, and blue bars show control cells. For the identification of hyperosmolarity-induced transcriptome changes, a 9-hour treatment with 200 mM sorbitol was chosen to capture the early and most likely direct osmodependent gene response, while a 24-hour time point was used to identify the secondary osmotic stress response ([69]Fig. 1A). In the presence of sorbitol, expression of 667 genes was significantly up-regulated, and expression of 486 genes was down-regulated [1 ≤ log[2] fold change (FC) ≤ −1, false discovery rate (FDR) < 0.05; [70]Fig. 1B]. The up-regulated genes include known osmoregulated genes, e.g., genes encoding solute carrier family members (SLCO4A1, SLC5A3, and SLC6A12) ([71]19, [72]20), aldose reductase (AKR1B1), and the matricellular protein osteopontin (SPP1; [73]Fig. 1A) ([74]21). High positive enrichment scores [normalized enrichment scores (NES)] in up-regulated genes at both time points with five “osmo-gene sets” derived from published studies of salt-, sucrose-, and sorbitol-treated mouse embryonic fibroblasts, rat kidney cells, and mouse trophoblast stem cells (fig. S1D) ([75]20, [76]22–[77]25) validated our experimental setting. Leading edge analysis of the 24-hour sorbitol dataset with these five published osmo-gene sets identified, for example, PRSS35, guanine nucleotide -binding protein–coupled estrogen receptor 1 (GPER1) and arginase 2 (ARG2) as conserved osmoregulated genes (fig. S1E). Ingenuity Pathway Analysis (IPA; P < 0.01; [78]Fig. 1C) revealed a remarkable enrichment of the 24-hour osmostress genes in the cholesterol biosynthesis pathway. The genes that were regulated after 9 hours could not clearly be attributed to a specific pathway, but some of them point to a proinflammatory effect of acute sorbitol treatment ([79]Fig. 1C). We examined the nine most strongly up-regulated genes from the screen for their response to various osmotic stimuli in 2D monolayers using quantitative reverse transcription polymerase chain reaction (qRT-PCR; [80]Fig. 1D and fig. S2, A to F). Primary fibroblasts originated from three healthy human donors and from two to three mice. Hyperosmolarity was induced by the addition of sorbitol, salt, or sucrose to the medium, and hypoosmolarity was induced by the addition of urea. The most robust and consistent regulation was observed for the PRSS35/Prss35 gene ([81]Fig. 1D and fig. S2, A to F). Its increased expression was already observed in 2D at 100 mM sorbitol or 50 mM salt and peaked at 200 mM sorbitol and 100 mM salt when incubated for 9 to 18 hours ([82]Fig. 1E). In summary, the PRSS35/Prss35 gene responded to all tested stimuli in fibroblasts from both species at early and late time points ([83]Fig. 1D). The differential expression of the other eight tested genes under osmotic stress conditions was also verified in at least four settings ([84]Fig. 1D). Other robustly regulated genes encode SPP1, GPER1, and ARG2, which produces urea. The accumulation of this metabolite might help the cells establish isotonicity. Sorbitol-dependent PRSS35 regulation also occurred in the dermal layer of scaffold-free 3D organotypic skin cultures ([85]Fig. 1F) ([86]26). Culture in a 3D (hydrogel) versus 2D environment increased the responsiveness of fibroblasts to a lower dose (50 mM) of sorbitol ([87]Fig. 1G), demonstrating that they are able to respond to small changes in osmolarity in a physiologically relevant 3D environment. The PRSS35 gene is regulated by p38/JNK–ATF2 and JNK-NFAT5 in response to hyperosmotic stress To study the mechanisms underlying the osmodependent gene regulation in HDFs, we focused on the gene encoding poorly studied PRSS35 and studied the involvement of p38 and JNK in its regulation. Simultaneous inhibition of these stress-activated kinases significantly reduced the sorbitol-induced PRSS35 up-regulation from 60-fold to 7.3-fold, while inhibition of either p38 or JNK alone had no or only a very mild effect ([88]Fig. 2A). Fig. 2. Hyperosmotic stress regulates PRSS35 expression via the p38/JNK-NFAT5/ATF2 axis. [89]Fig. 2. [90]Open in a new tab (A) qRT-PCR for PRSS35 using RNA from HDFs pretreated with 10 μM SB202190 (p38 inhibitor) for 1 hour and/or with 70 μM SP600125 (JNK inhibitor) for 2 hours before the addition of 200 mM sorbitol for 9 hours (N = 3, n = 6). The highest value of each experiment was set to 100%. (B) Transcription factor binding site prediction in the promoter region of PRSS35 [PRSS35Pro; P < 0.001, Find Individual Motif Occurrences (FIMO)], defined as the 1000 bp upstream of the PRSS35 transcription start site (TSS). (C) Western blot of lysates from HDFs treated for 30 min with 200 mM sorbitol for total and phosphorylated ATF2, JNK, p38, and tubulin (TUB; loading control). (D) qRT-PCR for NFAT5 using RNA from HDFs treated with 200 mM sorbitol for 3 to 24 hours (N = 3, n = 6). (E) qRT-PCR for nuclear factor of activated T cells (NFAT5) using RNA from HDFs pretreated with p38 and/or JNK inhibitors and incubated with 200 mM sorbitol ± inhibitors for 9 hours (N = 3, n = 6). (F to H) qRT-PCR for PRSS35 using RNA from HDFs, which had been transfected with NFAT5 and/or ATF2 (N6 and N7/A7 and A8) or scrambled (Scr) short interference RNAs (siRNAs) for 39 hours and treated with 200 mM sorbitol for 9 hours (N = 3, n = 6 to 9). (I) Lentiviral construct containing PRSS35Pro in front of a firefly luciferase (FFLuc) reporter and the GFP gene fused to a puromycin resistance cassette (PuroR) under control of a constitutively active promoter. (J) Luciferase activity in HDFs stably transduced with the construct depicted in (I) and treated with 150 to 250 mM sorbitol or 75 to 125 mM salt for 9 hours (N = 3, n = 7 to 13). RLU, relative light units. Bar graphs indicate means with SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, two-way ANOVA with Bonferroni post-test [(A) and (E) to (H)] or one-way ANOVA with Dunnett post-test [(D) and (J)]. Red bars, sorbitol-treated cells; blue bars, control cells. Analysis of the PRSS35 core promoter (PRSS35Pro), here defined as up to 1000 base pairs (bp) upstream of the transcriptional start site, revealed potential binding sites for activating transcription factor 2 (ATF2) and nuclear factor of activated T cells (NFAT5) ([91]Fig. 2B). Both transcription factors are activated by hyperosmotic stress in different cell types ([92]6, [93]20, [94]22, [95]27, [96]28). Activation of ATF2 by hyperosmolarity was abrogated upon combined inhibition of JNK and p38 ([97]Fig. 2C), confirming the relevance of the JNK/p38-ATF2 axis in osmoregulation. The NFAT5 gene itself responded to hyperosmotic stress ([98]Fig. 2D), and its osmoregulated expression was dependent on JNK ([99]Fig. 2E). A 70 to 90% knockdown of NFAT5 by two different short interference RNAs (siRNA N6 and N7) reduced the sorbitol-induced PRSS35 up-regulation, while ATF2 knockdown (siRNA A7 and A8) had no effect ([100]Fig. 2, F and G, and fig. S3, A and B). Combined knockdown of NFAT5 and ATF2 further suppressed PRSS35 expression under hyperosmotic stress conditions ([101]Fig. 2H and fig. S3C). These findings demonstrate that NFAT5 is of key relevance for the osmoregulation of PRSS35 and that ATF2 exacerbates this effect. We next cloned the PRSS35 core promoter in front of a firefly luciferase (FFLuc) reporter ([102]Fig. 2I). This plasmid was used for stable transfection of HDFs. Incubation of these cells with sorbitol or salt caused a significant increase in FFLuc activity ([103]Fig. 2J), confirming the osmosensitivity of PRSS35Pro. PRSS35 and other osmoregulated genes are overexpressed in healing wounds Previous qRT-PCR studies showed that Prss35 expression increases in murine wounds and in wound-induced papillomas compared to normal skin ([104]29). We confirmed this using publicly available transcriptomics data and found that the increased expression in wounds is not exclusive for Prss35. Rather, gene set enrichment analysis (GSEA) using six wound healing datasets ([105]30–[106]33) showed a general positive enrichment of sorbitol-regulated genes in gene sets of whole wounds in the early (inflammatory) phase of healing and in wound fibroblasts derived from early-stage wounds ([107]Fig. 3A). Among the drivers for the positive enrichment score were SPP1, PRSS35, and ARG2 ([108]Fig. 3B). Published spatial transcriptomics data of 7-day murine splinted wounds ([109]34) identified six spatial clusters correlating with histologically distinct areas ([110]Fig. 3C). Wound tissue ([111]Fig. 3D, top), but not unwounded skin ([112]Fig. 3D, bottom), exhibited a high “osmo-signature” (defined by the nine validated sorbitol-regulated genes shown in [113]Fig. 1D). Enrichment of osmodependent genes was mainly detected in the granulation tissue ([114]Fig. 3E), and Prss35 transcripts were mainly localized to the granulation tissue at the edge of the wound at the borders of the eschar. Fig. 3. PRSS35 and other osmoregulated genes are highly expressed in skin wounds. [115]Fig. 3. [116]Open in a new tab (A) GSEA comparing own RNA-seq data (HDFs in GelMA hydrogels treated with 200 mM sorbitol for 9 and 24 hours, FDR < 0.05) to published transcriptomics data of whole skin wounds and wound fibroblasts. Data are presented as a pseudo-heatmap with an NES magnitude from −2 (blue) to 2 (red). Gene sets were obtained from the studies ([117]30–[118]33). (B) Leading edge analysis of skin wound datasets [published gene sets ([119]30–[120]33) versus 9-hour sorbitol gene set from this study] identifies the 29 top enriched common genes (red) and ranks them according to the enrichment score (ES) sum. Bolded are validated osmoresponsive genes in HDFs. (C) Spatial transcriptomics (ST) analysis of murine splinted wounds 7 days after wounding ([121]34) reveals six spatial clusters (clusters 0 to 5) correlating with characteristic areas of the wound tissue based on histological analyses. (D) The expression of the osmoresponsive gene scores in specific voxels is color-coded, ranging from −0.5 (blue, low osmo-score) to 0.5 (red, high osmo-score). (E) Nine validated osmoresponsive genes defining the osmo-signature are significantly enriched in the granulation tissue of wounds, with Prss35 being specifically up-regulated at the edge of the granulation tissue. (F) Representative immunohistochemistry stainings of sections from 15-day human excisional wounds using two different PRSS35-specific antibodies (#PA58641 and #HPA038788). Scale bars, 500 μm (top), 50 μm (bottom). D, dermis; E, epidermis; G, granulation tissue. To determine whether the increased expression of PRSS35 in skin wounds results mainly from hyperosmotic stress or from other challenges/conditions present in wound tissue, we treated cultured fibroblasts with agents that induce inflammation {lipopolysaccharide, polyinosinic:polycytidylic acid [poly(I:C)], or tumor necrosis factor α}, with fetal bovine serum that includes growth factors present in skin wounds, or with agents that induce reactive oxygen species [fig. S4, A to C (top graphs)]. None of these challenges/conditions induced PRSS35 expression in the absence of sorbitol, although they strongly regulated their known effector genes, including radical S-adenosyl methionine domain containing 2 (RSAD2), stress-associated endoplasmic reticulum (ER) protein 1 (SERP1), or NAD(P)H quinone dehydrogenase 1 [NQO1; fig. S4, A to C (bottom graphs)]. These data provide indirect evidence for a key role of hyperosmotic stress in the upregulation of PRSS35 expression in skin wounds, although the predominant increase in the expression of this gene at the wound edge points to an involvement of additional factors. Consistent with the spatial transcriptomics data, two different antibodies detected PRSS35 protein in the granulation tissue at the edge of acute human excisional wounds, particularly in the upper wound region close to the epidermal border ([122]Fig. 3F and fig. S4D for the secondary antibody–only control). PRSS35 was not or only weakly expressed in the dermis of unwounded skin (fig. S4E). Keratinocytes of normal and wounded skin also expressed PRSS35/Prss35 ([123]Fig. 3F and fig. S4, E and F). Furthermore, expression of PRSS35 increased in primary human keratinocytes in response to sorbitol treatment (fig. S4G), suggesting an as-yet unknown function of this protein in keratinocytes in response to hyperosmotic stress. Cytoplasmic/extracellular staining was obtained with an antibody detecting the N-terminal region, whereas an antibody against the C terminus stained mainly cell nuclei ([124]Fig. 3F). Similar findings were obtained with murine excisional wounds (fig. S4F). Cleavage of PRSS35 by furin generates fragments with different sub- and extracellular localizations PRSS35 is known as a secreted protein acting in the extracellular matrix (ECM) ([125]29, [126]35, [127]36) or in the secretome ([128]37), whereas a nuclear localization has not been described. A closer look at the PRSS35 protein sequence, however, revealed a nuclear localization signal (NLS; [129]Fig. 4A, orange) in the middle of the sequence. Because PRSS35 also has a signal peptide ([130]Fig. 4A, red) and therefore enters the secretory pathway, we determined its sub- and extracellular localizations in vitro in response to sorbitol. For this purpose, we used antibodies detecting an N-terminal ([131]Fig. 4A, blue) or a C-terminal epitope ([132]Fig. 4A, green) of PRSS35. Western blot analysis identified only low amounts of PRSS35 when fibroblasts were cultured in isotonic control medium, but a strong induction occurred upon sorbitol addition ([133]Fig. 4B). Full-length (FL; 47 kDa) PRSS35 peaked transiently at 9 to 16 hours after the addition of sorbitol, whereas its cleaved products, a C-terminal fragment (~30 kDa) and an N-terminal fragment (~15 kDa), peaked at 16 to 24 hours. The N-terminal fragment was present in the supernatant ([134]Fig. 4B), while the C-terminal fragment was enriched in the nucleus and was the predominant form after 18 hours of sorbitol treatment ([135]Fig. 4C). The FL protein was detectable in the total lysate ([136]Fig. 4, B and C) and in the cytoplasmic lysate ([137]Fig. 4C), which also includes proteins from organelles (except the nucleus). We confirmed the specific recognition of the PRSS35 protein and its cleavage products by both antibodies using fibroblasts with short hairpin RNA (shRNA)–mediated PRSS35 knockdown (Sh8 and Sh10) upon sorbitol treatment for 9 hours (fig. S5A). After 48 hours of conditioning, the N-terminal fragment had accumulated in the supernatant even in the absence of sorbitol, and the amounts increased further in the presence of sorbitol (fig. S5B). By contrast, neither FL PRSS35 nor the C-terminal fragment were detectable in the supernatant (fig. S5B, left). Fig. 4. Processing and localization of PRSS35 in HDFs in 2D culture. [138]Fig. 4. [139]Open in a new tab (A) PRSS35 protein sequence, containing a signal peptide (red), an NLS (orange), a putative furin cleavage site (//), and a T instead of an S in the previously predicted catalytic center (T bold). Epitopes recognized by two different PRSS35-specific antibodies are highlighted in green and blue. (B) Western blot analysis of total cell lysates and culture supernatants of HDFs treated with 200 mM sorbitol (Sorb) for 9 to 24 hours using antibodies against PRSS35 and vinculin (VCL; loading control). Ponceau S staining of the membrane confirmed equal loading of the supernatants. (C) Western blot analysis of nuclear (N), cytoplasmic (CP), and total lysates (TL) of HDFs incubated with 200 mM sorbitol for 18 hours using antibodies against PRSS35, VCL, and the nuclear marker histone H3 (H3). (D) Western blot analysis of total lysates and cell culture supernatants of HDFs, which had been pretreated with 50 μM furin inhibitor I (FI1) or 75 μM furin inhibitor III (FI3) before the addition of 200 mM sorbitol or vehicle for 16 hours. The membrane was probed with antibodies against PRSS35 and VCL. (E) Potential furin cleavage sites as predicted by ProteasePredictor, including R122, R208, and K235. Cleavage at R122 yields a 33-kDa C-terminal fragment that includes the NLS (orange). Amino acid sequence analysis predicted furin as a potential PRSS35-processing enzyme, and a recent study showed that PRSS35 is processed by furin in an immortalized hepatocyte cell line ([140]37). Consistent with these findings, inhibition of furin using two different inhibitors (FI1/3), increased the amounts of sorbitol-induced FL PRSS35 in fibroblasts while decreasing the amounts of the cleavage products ([141]Fig. 4D). Taking into consideration the molecular weight of the cleavage products, we suggest that furin cleaves PRSS35 at R122 in fibroblasts ([142]Fig. 4E). This would render a 33-kDa C-terminal fragment that still contains an intact NLS. These findings provide an explanation for the nuclear versus cytoplasmic/extracellular stainings that we observed in skin wounds using different PRSS35 antibodies. PRSS35 colocalizes with intracellular collagens To determine whether the increase in PRSS35 expression in response to sorbitol is functionally important, we performed siRNA-mediated PRSS35 knockdown (fig. S6A) and analyzed the effect on cell proliferation and cytotoxicity in response to sorbitol treatment. While knockdown of PRSS35 did not affect sorbitol-induced cell death (fig. S6B), it rescued the sorbitol-induced decrease in proliferation ([143]Fig. 5A). Fig. 5. PRSS35 interacts with different ECM proteins. [144]Fig. 5. [145]Open in a new tab (A) 5-Bromo-2′-deoxyuridine (BrdU) incorporation into the DNA of HDFs transfected with scrambled (Scr) or PRSS35 siRNAs (P5 and P7). Cells were treated for 32 hours with 10 μM BrdU and 100 mM sorbitol (N = 2, n = 4). Red bars show sorbitol-treated cells, and blue bars show control cells. (B) Construct used for BioID, containing the FL PRSS35 coding sequence fused C-terminally to BirA-HA under control of a tetracycline-responsive element (TRE; scheme). Western blot of total, nuclear, and cytoplasmic lysates and of cell culture supernatants of HDFs with Dox-inducible expression of PRSS35–BirA-HA using antibodies against PRSS35, the HA epitope, VCL, or H3. Cells were incubated with dimethyl sulfoxide (DMSO) or Dox (20 ng/ml) ± 200 mM sorbitol for 32 hours. (C) Selected putative PRSS35–BirA-HA interaction partners in HDFs identified by BioID with high SAINT probabilistic scoring (SP ≥ 0.99). The Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) representation connects physical networks. Brown color-coding indicates matrisome proteins. Blue indicates ER proteins involved in folding and quality control. (D) Representative coimmunofluorescence stainings of HDFs with Dox-inducible expression of PRSS35-HA using antibodies against the HA epitope (top: red; bottom, green), collagen type I (COLI; green), or MIA3 (red). Scale bar, 10 μm. (E) Proximity ligation assay (PLA) showing proximity of Dox-inducible PRSS35-HA and COLI (top) or MIA3 (bottom). Scale bar, 10 μm. (F) PLA showing interaction of endogenous PRSS35 and COLI in sorbitol-treated (18 hours) HDFs. Red PLA signal is detected with both PRSS35 antibodies and the COLI antibody. Scale bar, 50 μm. (G) Co-IP of lysates from HDFs expressing PRSS35–BirA-HA, which had been treated with Dox (20 ng/ml) or DMSO for 40 hours and incubated with a cross-linker. Beads coupled with HA antibody were used for IP. Input, unbound, and immunoprecipitated fractions were analyzed by Western blot for COLI, HSPA5, PRSS35, and tubulin (TUB). Bar graphs indicate mean with SD. ***P < 0.001 and ****P < 0.0001, two-way ANOVA with Bonferroni post-test (A). Red bars, sorbitol-treated cells; blue bars, control cells. To unravel the function of PRSS35, we searched for proteins, which are in close proximity to and/or directly interact with PRSS35 using the proximity-dependent biotin identification (BioID) technology. For this purpose, a hyperactive biotin ligase coupled to an influenza virus hemagglutinin (BirA-HA) was fused in frame to the C terminus of FL PRSS35 ([146]Fig. 5B, scheme). Because previous studies focused on its extracellular role ([147]29, [148]35), we aimed at identifying intracellular PRSS35 interaction partners and therefore attached the BirA-HA to the C terminus. The expression of the fusion protein was placed under the control of a doxycycline (Dox)–inducible promoter, and the construct was used for stable transfection of HDFs. Similar to endogenous PRSS35 ([149]Fig. 4, B and C), FL PRSS35–BirA-HA was detected in the cytoplasmic lysates that included non-nuclear organelles and was processed into an N-terminal secreted and a C-terminal nuclear protein ([150]Fig. 5B). The processing efficiency of endogenous PRSS35 was, however, higher compared to PRSS35–BirA-HA, and the latter could not be increased by the addition of sorbitol ([151]Fig. 5B). Expression of the fusion protein did not induce obvious ER stress, as demonstrated by the lack of induction/activation of the key effectors of the unfolded protein response (fig. S6C), including heat shock protein family A member 5 (HSPA5), ATF6, and CCAAT/enhancer binding protein homologous protein (CHOP). Following streptavidin-mediated pull-down (fig. S6D), the precipitated proteins were identified by mass spectrometry, and candidates with a significance analysis of interactome (SAINT) probability (SP) score ≥0.99 were further analyzed. A prominent cluster of intracellular proteins identified in the BioID screen (SP ≥ 0.99; [152]Fig. 5C) contained certain fibrillar [collagen type I (COLI) and COLV] and nonfibrillar (COLVI and COLVII) collagens, as well as collagen-associated proteins, such as the collagen receptor subunit integrin β1, and melanoma inhibitory activity protein 3 (MIA3; also known as TANGO1), which plays a key role in collagen secretion ([153]38–[154]40). By contrast, matricellular proteins were not found in the BioID screen, suggesting that the PRSS35-collagen interaction is not mediated by these proteins. The putative PRSS35 interactors also include several components of the protein folding machinery in the ER ([155]Fig. 5C), such as heat shock proteins (HSPA5 and HSP90B1), protein disulfide isomerases (PDIA3, PDIA4, and PDIA6), peroxiredoxin 4 (PRDX4), and proteins involved in N-glycosylation and lectin-assisted folding, such as neutral α-glucosidase AB and calreticulin (fig. S6E and table S1). These proteins had previously been identified as components of the COLI proteostasis network in a screen using immunoprecipitation (IP) followed by mass spectrometry with FL collagen α1(I) or the collagen α2(I) propeptide as baits ([156]41, [157]42). This overlap points toward a role of PRSS35 in collagen folding and maturation in the ER and suggests that PRSS35 might interact with the propeptide of COLI. Consistent with this assumption, consensus cleavage sites for PRSS35 [tandem lysines (KK)] ([158]37) were identified in N- and C-terminal propeptides of several collagens (table S2). Signals from COLI and intracellular PRSS35-HA, which is similarly processed as PRSS35-Bir-HA (fig. S6F), strongly overlapped in immunostained fibroblasts ([159]Fig. 5D, top), and the proximity of both proteins was confirmed by proximity ligation assay (PLA; [160]Fig. 5E, top and fig. S6G, top). Some PRSS35-HA proteins localized to transport vesicles rich in MIA3 ([161]Fig. 5D, bottom; [162]Fig. 5E, bottom; fig. S6G, bottom). In addition, when wild-type fibroblasts were treated with sorbitol for 18 hours, we found COLI and endogenous PRSS35 in close proximity as identified by PLA with both PRSS35-specific antibodies ([163]Fig. 5F and fig. S6H). Using cross-linking followed by co-IP with beads coupled to an HA antibody, we showed that PRSS35–BirA-HA directly binds HSPA5 and COLI ([164]Fig. 5G). Because the proteins identified in the BioID screen point toward a role of PRSS35 in the ECM and because we only identified very few nuclear proteins in the BioID experiment (table S1), such as histone H2B type 1-K (HIST1H2BK), we focused follow-up functional experiments on the role of PRSS35 in the secretory pathway. PRSS35 and collagens type I, V, VI, and XII colocalize in wounds To investigate whether PRSS35 also colocalizes with collagens in vivo, we examined 7-day murine excisional wounds as a model defined by a hyperosmotic stress signature ([165]Fig. 3, D and E). Analysis of spatial transcriptomics data ([166]34) revealed particularly high expression (both in average intensity and in the number of expressing cells) of Prss35 in the granulation tissue edge cluster along with Col1a1, Col1a2, Col5a1, Col6a1, Col6a2, Col6a3, and Col12a1 ([167]Fig. 6A). mRNAs encoding ColI, ColV, ColVI, and ColXII subunits colocalized to the same voxels as Prss35 mRNA ([168]Fig. 6B), and they had high and significant Spearman correlation scores with Prss35, but not with Mia3 and unrelated genes, such as α-smooth muscle actin 2 (Acta2), keratin 17 (Krt17), the immune cell markers Cxcl12 ([169]Fig. 6C) or Cxcl2, which encodes a chemokine that had been identified as a PRSS35 substrate in hepatocellular carcinoma ([170]37). Fig. 6. Prss35 is coexpressed with collagens in murine skin wound fibroblasts. [171]Fig. 6. [172]Open in a new tab (A) Spatial transcriptomics (ST) analysis of murine 7-day splinted wounds ([173]34) showing that Prss35, ColI, ColV, ColVI, and ColXII and Mia3 transcripts are enriched in the granulation tissue edge and the central granulation tissue. (B) Spatial expression of Prss35, collagen genes, and Mia3 at voxel resolution, visualized by color-coding, ranging from blue (low expression) to red (high expression). (C) Spearman correlation scores and their respective P values for Col1a1, Col1a2, Col5a1, Col6a1, Col6a2, Col6a3, Col12a1, Mia3, and Acta2 (myofibroblast marker), Cxcl12 (immune cell marker), Krt17 (epithelial cell marker), and Cxcl2 with Prss35 ([174]34). (D) Uniform manifold approximation and projection (UMAP) plot showing the single-cell transcriptomics analysis of 7-day wounds ([175]43) separating the total wound cell population into six major cell type clusters. (E) Single-cell RNA-seq data showing that expression of Prss35, Col1a1, Col1a2, Col5a1, Col6a1, Col6a2, Col6a3, and Col12a1 is specific to wound fibroblasts. (F) UMAP plots of the whole wound cell population showing expression levels of Prss35 and Col1a1. (G) UMAP plot showing single-cell transcriptomics data of 7-day wounds ([176]43), including only wound fibroblasts and identifying four fibroblast subclusters. (H) UMAP plot of the wound fibroblast population showing expression levels of Prss35. (I) Reactome pathway enrichment analysis of differentially expressed marker genes for cluster #2, which contains Prss35. (J) Expression of Prss35 and genes expressing putative Prss35 interaction partners in radially partitioned murine skin wounds (original wound size, 4 mm) at days 1, 3, 7, and 14 after wounding in the whole cell population ([177]44). At day 1, the open wound size is 4 to 5 mm, and the wound edge is at 4 to 6 mm. At day 7, the open wound size is 2 mm, and the wound edge is at 2 to 4 mm. At day 14, the wound is closed. UW, unwounded. Single-cell transcriptomics analysis of 7-day murine wounds ([178]43) identified six distinct cell clusters ([179]Fig. 6D) on the basis of cell type–specific marker expression (fig. S7A). Prss35 and the collagen genes were exclusively expressed by wound fibroblasts ([180]Fig. 6, E and F). On the basis of the single-cell data, Prss35 had high and significant Spearman correlation scores with Col1a1, Col1a2, Col5a1, Col6a1, Col6a2, Col6a3, and Col12a1, but not with Mia3 and unrelated genes, such as Acta2, Cxcl12, or Krt17 (fig. S7B), which is consistent with the spatial transcriptomics data. Within four identified wound fibroblast subpopulations ([181]Fig. 6G and fig. S7C), Prss35 expression was characteristic for cluster #2 ([182]Fig 6H). The differentially expressed marker genes of this cluster enriched highly for collagen maturation pathways ([183]Fig. 6I). We further analyzed the kinetics of Prss35 expression via another single-cell transcriptomics dataset of murine wounds sampled at several days after injury and at multiple distances from the wound center ([184]44). These data confirmed that Prss35 and the collagen genes are almost exclusively expressed by wound fibroblasts, although a lower expression in keratinocytes was also detected (fig. S7, D to F). After subsetting the dataset to fibroblasts (fig. S7G), we found that Prss35 was up-regulated at days 1 and 7 after injury at the wound edges (i.e., 4 to 6 mm and 2 to 4 mm radial areas, respectively) ([185]Fig. 6J). Within those areas, Prss35 was coexpressed with Col1a1 and Col1a2 at day 1, whereas at day 7, it was coexpressed with Col5a1 ([186]Fig. 6J). Within the fibroblast subpopulations, Prss35 expression was characteristic for cluster #2 (fig. S7G). Its differentially expressed marker genes include Col5a1, Col12a1, tenascin C (Tnc), and lysyl oxidase (Lox; fig. S7H), which are all important in collagen maturation. The fibroblast subtype that selectively expresses Prss35 according to this dataset was not limited to the day 7 time point but was more generally enriched at the wound edges during the proliferative phase of healing (days 3 and 7 after injury; fig. S7I). PRSS35 regulates the matrisome under hyperosmotic stress conditions Previous research indicated an involvement of PRSS35 in fibrosis ([187]29, [188]35, [189]45). Furthermore, the secretion of PRSS35 from fibroblasts and its association with collagens and with COLI binding partners points toward a function of PRSS35 in the ECM. Therefore, we determined whether the matrisome deposited by PRSS35 knockdown fibroblasts influences cellular behavior. We focused on the sorbitol-treated matrisome because of the low expression of the PRSS35 protein under baseline conditions ([190]Fig. 4B and fig. S5B). We found that the decellularized matrisome deposited by sorbitol-treated fibroblasts with PRSS35 knockdown promoted the proliferation of wild-type fibroblasts ([191]Fig. 7A), whereas the matrisome deposited by PRSS35 overexpressing fibroblasts had the opposite effect (fig. S8, A and B). This finding suggests that the induction of PRSS35 expression by hyperosmotic stress induces changes in the matrisome, which have functional consequences for cells in contact with the matrix. Fig. 7. Effect of sorbitol-induced PRSS35 on the matrisome. [192]Fig. 7. [193]Open in a new tab (A) Percentage of Ki67-positive nuclei of untreated HDFs seeded on decellularized matrix from HDFs expressing Dox-inducible Scr or PRSS35 shRNAs (Sh8 and Sh10). Matrix-producing cells had been pretreated with mitomycin C for 2 hours, with Dox (100 ng/ml) or vehicle (DMSO) + ascorbic acid (50 μg/ml) for 3 days, and treated with 200 mM sorbitol (Sorb) + DMSO/Dox + ascorbic acid for another 3 days (n = 4). (B) Experimental setup for the proteomics analysis of matrisomes deposited by Scr and Sh8 HDFs. Cells received fresh medium containing ascorbic acid (50 μg/ml) every 1 to 2 days. (C) Volcano plot of more or less abundant proteins in matrisomes of control (Sh8_DMSO) versus PRSS35 knockdown fibroblasts (Sh8_Dox) under hyperosmotic stress (1 ≤ log[2] FC ≤ −1, P < 0.05; red or green). (D) Selected more abundant matrisome proteins in Sh8_Dox versus Sh8_DMSO fibroblasts clustered into basement membrane (pink), core matrisome (beige), or other proteins (blue). (E) Western blot analysis of matrisomes deposited by Scr, Sh8, and Sh10 HDFs treated as shown in (B) and harvested 72 hours after addition of sorbitol using antibodies against TMEM2, NID1, and PTX3 (n = 3). Ponceau S staining of the membranes was used as loading control. *Sample degraded. (F) Western blot analysis of total lysates from Scr, Sh8, and Sh10 HDFs treated as shown in (B) and harvested 72 hours after addition of sorbitol. Membranes were probed with antibodies against TMEM2, NID1, PTX3 and VCL (n = 3). (G) qRT-PCR for TMEM2, NID1, and PTX3 using RNA from Scr, Sh8, and Sh10 HDFs treated as shown in (B) and harvested 48 hours after addition of sorbitol (n = 3). Bar graphs indicate means with SD. *P < 0.05, **P < 0.01, and ****P < 0.0001, one-way ANOVA with Bonferroni post-test (A) or two-way ANOVA with Bonferroni post-test (F). Red and rosé bars, sorbitol-treated cells; blue bars, control cells. To identify relevant components of the matrisome, which are regulated by PRSS35 and may be responsible for the observed effect on cell proliferation, we analyzed the matrisome deposited by sorbitol-treated control (Scr) and PRSS35 knockdown fibroblasts (Sh8) by quantitative mass spectrometry–based proteomics ([194]Fig. 7B). We confirmed a PRSS35 knockdown efficiency of more than 90% (fig. S8C). In the matrisome of sorbitol-treated PRSS35 knockdown cells (Sh8_Dox), 64 proteins were more abundant (1 ≤ log[2] FC, P < 0.05) compared to the matrisome of control cells (Sh8_DMSO) ([195]Fig. 7C and fig. S8D). These proteins are repressed, destabilized, or less efficiently released to the matrix in the presence of PRSS35. On the other hand, 21 proteins were less abundant (log[2] FC ≤ −1, P < 0.05; [196]Fig. 7C and fig. S8D). Consistent with the proteomics data, Western blot analysis of fibroblasts with shRNA-mediated knockdown of PRSS35 (fig. S5B) confirmed that PRSS35 does not affect the amount of COLI in the matrisome (fig. S8E). This was also observed by second harmonic generation (SHG) imaging (fig. S8, F to H). However, subtle changes in collagen macromolecular organization might have been masked by the strong reduction in COLI expression following sorbitol treatment, which was already seen at the transcriptional level (log[2] FC of −1.4 after 9 hours and −1.5 after 24 hours for COL1A1; see RNA-seq data). In the absence of sorbitol, however, we observed a minor increase in collagen density and a slight reduction in the forward-to-backward SHG ratio (F-SHG/B-SHG) in Sh8 cells with PRSS35 knockdown (fig. S8, G and H). Furthermore, analysis of the matrisome proteomics data for Lys and Pro hydroxylation of different collagens revealed that PRSS35 knockdown increased hydroxylation at various sites, whereas only very few sites were more hydroxylated when PRSS35 was induced (fig. S8I). This indicates that PRSS35 affects collagen hydroxylation in addition to matrisome protein abundances. The largest group of proteins that were more abundant in the matrisome of sorbitol-treated PRSS35 knockdown cells includes basement membrane components [nidogen 1 (NID1), laminins, and COLIV] and proteins of the core matrisome (biglycan, COLV, and COLXI; [197]Fig. 7D and fig. S8D). The cell surface hyaluronidase transmembrane protein 2 (TMEM2), also known as cell migration inducing hyaluronidase 2 (CEMIP2), and pentraxin-related protein 3 (PTX3), which is involved in tissue repair ([198]46), were among the more abundant proteins in the matrisome deposited by PRSS35 knockdown cells, while CEMIP, another hyaluronidase, was significantly less abundant ([199]Fig. 7C and fig. S8D). Western blot analysis showed that the amounts of TMEM2 and NID1 were lower in the matrisome of sorbitol-treated versus medium-treated cells ([200]Fig. 7E). This was at least partially rescued by PRSS35 knockdown ([201]Fig. 7E), which is consistent with the proteomics data. The amount of PTX3 in the matrix was not regulated by sorbitol, but PRSS35 knockdown still increased its abundance in the matrisome ([202]Fig. 7E). The increased amounts of TMEM2 and PTX3 in the matrisome of sorbitol-treated PRSS35 knockdown cells were not associated with a reduction of these proteins in the cell lysates ([203]Fig. 7F). There was also no reduction of PTX3 in the conditioned medium (fig. S9A), and TMEM2 was not detectable there. TMEM2 and PTX3 gene expression was either not regulated upon PRSS35 knockdown under sorbitol or even showed an opposite regulation compared to the abundance in the matrisome ([204]Fig. 7G and fig. S9B for validation of the knockdown efficiency and sorbitol-induced PRSS35 expression). NID1 expression was also not consistently regulated by PRSS35 knockdown ([205]Fig. 7G). However, the abundance of this protein was reduced in the total lysates of PRSS35 knockdown cells ([206]Fig. 7F), suggesting that PRSS35 knockdown promotes NID1 secretion and its deposition in the matrisome. Together, we identified PRSS35 as a protein that affects the matrisome composition ([207]Fig. 8), particularly in the presence of hyperosmotic stress, which limits proliferation of fibroblasts. Fig. 8. Scheme showing PRSS35 gene activation and PRSS35 protein expression, processing, and function. [208]Fig. 8. [209]Open in a new tab PRSS35 gene expression in HDFs is induced by hyperosmotic stress via p38/JNK-ATF2 and JNK-NFAT5. The PRSS35 protein is cleaved by furin at R122, generating an N-terminal, secreted fragment and a C-terminal, nuclear fragment. The FL fragment localizes mainly to the cytoplasm (including organelles). Intracellularly, PRSS35 is in proximity to several collagens. Knockdown of PRSS35 in sorbitol-treated HDFs causes functional alterations of the matrisome and increases the abundance of different collagens, NID1, TMEM2, and PTX3 in the matrisome. Created with [210]Biorender.com. DISCUSSION We characterized the transcriptional response of HDFs in 3D cultures to hyperosmolarity, which revealed a rapid activation of various cytoprotective genes. This allows the cells to adapt to the hyperosmotic condition and to restore isotonicity. At later stages, cells regulate genes involved in the biosynthesis of cholesterol, the main sterol present in the plasma membrane of mammalian cells that defines crucial membrane properties ([211]47). In analogy to findings obtained in Saccharomyces cerevisiae ([212]48, [213]49), the sorbitol-induced increase in cholesterol levels is likely to stabilize the plasma membrane and to prevent permeabilization. The most strongly and robustly osmoregulated gene was PRSS35. We showed that sorbitol-induced PRSS35 expression is controlled by classical osmosensitive pathways that involve JNK, p38, ATF2, and NFAT5 activation ([214]6, [215]50, [216]51) ([217]Fig. 8). Our study also provides insight into the sub- and extracellular localizations of PRSS35, for which conflicting results had previously been reported ([218]37, [219]52–[220]54). We show that the N-terminal cleavage product is secreted, whereas the C-terminal cleavage product that includes an NLS localizes to the nucleus of fibroblasts in vitro and in wounds in vivo ([221]Fig. 8). Our and previously published data ([222]37) demonstrate that PRSS35 enters the secretory pathway and is processed by furin, a protease that localizes to the trans-Golgi network and the extracellular space ([223]55). The PRSS35 N terminus continues along the secretory pathway and accumulates in the cell supernatant, while the C terminus is directed toward the nucleus. Because our BioID screen identified only very few nuclear proteins, we focused the functional analysis on the non-nuclear protein (fragment), while the role of the nuclear fragment and its shuttling mechanism should be determined in future studies. Our work identified a role of PRSS35 in cell proliferation in response to hyperosmotic stress because PRSS35 knockdown rescued the sorbitol-induced decrease in fibroblast proliferation. This might overrule cellular quality control mechanisms ([224]20, [225]50), leading to the propagation of osmostressed cells. The effect of PRSS35 on cell proliferation is, at least partially, matrix dependent, as suggested by the changes in fibroblast proliferation when plated on ECM deposited by sorbitol-treated PRSS35 knockdown versus control cells. An effect of PRSS35 on the matrix is consistent with reports, which functionally link PRSS35 to fibrosis. For example, PRSS35 expression was up-regulated in fibrosis-associated fibroblasts ([226]35), and cancer-associated fibroblasts from wound-induced and chemically induced skin tumors showed a profibrotic gene expression profile and expressed Prss35 ([227]29), suggesting a role of PRSS35 in the promotion of fibrosis. By contrast, administration of antibodies neutralizing the serine protease inhibitor human epididymis protein 4, which targets PRSS35, PRSS23, and matrix metalloproteinases, led to a reduction in renal fibrosis ([228]35). Although the inhibitors were not specific for PRSS35, this suggests an antifibrotic role for this protein. In line with this hypothesis, red light used for fibrosis therapy induced PRSS35 gene expression and reduced fibroblast proliferation ([229]45). Furthermore, global knockout of Prss35 increased the collagen density in late skin wounds and skin tumors ([230]29), which was associated with a higher tumor incidence. Our BioID screen identified COLI and other collagens as putative binding partners of PRSS35. While this may implicate either pro- or antifibrotic effects, our ECM proteomics experiment points to the latter, as we found increased abundance of different collagens in the matrisome of sorbitol-treated PRSS35 knockdown cells. The amounts of COLI in the ECM were only mildly affected by PRSS35 knockdown under homeostatic conditions, and no effect was seen in sorbitol-treated cells. However, minor differences may be masked by the strong reduction of COLI deposition in the presence of sorbitol. Several data obtained in this study point toward a role of PRSS35 in collagen folding and maturation. These include the selective expression of Prss35 in the fibroblast cluster enriching for collagen maturation in wounds, the identification of key components of the COLI proteostasis network as putative PRSS35 interaction partners, and the alterations in Lys and Pro hydroxylation in PRSS35 knockdown fibroblasts. Therefore, PRSS35 seems to affect not only the abundance of some collagens but also different steps of their maturation. PRSS35 was originally termed “inactive serine protease 35” because it includes a threonine in the predicted catalytic center at the C terminus, whereas an active enzyme was expected to have a serine at this site ([231]56, [232]57). Nevertheless, a previous study reported cleavage of COLI in vitro ([233]35). An explanation for this discrepancy was provided in a recent study, which showed catalytic activity of the N-terminal fragment of PRSS35. It was found to cleave the chemokine CXCL2 at dibasic KK residues in the secretome of liver cells, which was required for the tumor-suppressive effect of PRSS35 in the liver observed in this study ([234]37). TMEM2 harbors two, and NID1 and PTX3 harbor one each of these potential cleavage sites for PRSS35. In addition, we detected several KK motifs in the proregions of different collagens. Therefore, these proteins might be direct substrates of PRSS35, which should be tested in the future. Independent of this open question, PRSS35 influences the matrisome composition through its effect on the stability, maturation, and/or secretion of selected matrisome components, which, in turn, affects fibroblast proliferation. This may also involve differential binding of growth factors and cytokines to the matrix proteins as suggested by the increased abundance of fibroblast growth factor 2 in the matrisome of PRSS35 knockdown cells. Our data also suggest that PRSS35 induces alterations in the fibroblast matrisome that counteract the consequences of hyperosmotic stress. For example, TMEM2 was significantly more abundant in the matrisome of sorbitol-treated PRSS35 knockdown compared to control fibroblasts, while CEMIP was less abundant. Both proteins cleave high molecular weight hyaluronic acid, but they generate fragments of different sizes that have different functions ([235]58, [236]59). Hyaluronic acid is the major glycosaminoglycan in mammalian ECM ([237]60). It contains multiple negatively charged residues, thereby contributing to the maintenance of the tissue’s water content. Alterations in the processing of hyaluronic acid by PRSS35-mediated regulation of the amounts of TMEM2 and CEMIP could therefore modify the water retention capacity of the ECM. Together, our results provide comprehensive information on the response of fibroblasts to hyperosmotic stress. They identify PRSS35 as a major osmoregulated protein, which functionally alters the matrisome to adapt to hyperosmotic stress conditions. MATERIALS AND METHODS Experimental design The goal of this study was to determine the molecular and functional consequences of hyperosmotic stress on the skin. As a model system, we used human primary skin fibroblasts in 2D monolayer, 3D organotypic skin cultures, and 3D hydrogels and exposed them to hyperosmotic stress. Differentially expressed genes were analyzed by RNA-seq, and the signaling pathway that resulted in the up-regulation of a major osmoregulated gene (PRSS35) was characterized. Last, we determined the expression of these osmoregulated genes in skin wounds and the function of PRSS35 in fibroblasts. For all experiments, we used the largest possible sample size, and there was no exclusion of any data point. Wound healing experiments in mice Mice (C57Bl/6 genetic background) were housed under specific pathogen–free conditions and received food and water ad libitum. After anesthesia with ketamine/xylazine and shaving, four full-thickness excisional wounds (diameter, 5 mm) were generated on either side of the back midline as previously described ([238]61). All animal experiments had been approved by the veterinary authorities of the Swiss canton of Zurich (Kantonales Veterinäramt Zürich, Switzerland). Human wound samples Biopsies from acute human wounds were obtained anonymously from a healing study with healthy volunteers (trial number 002WH99) undertaken by SWITCH BIOTECH GmbH (Neuried, Germany) at the Clinic and Polyclinic for Dermatology and Allergology (Biederstein) of the Technical University of Munich, after approval by the local research ethics committee. Briefly, 6-mm-diameter punch biopsies were taken from the margins of acute skin wounds 15 days after injury. Biopsies were fixed and paraffin-embedded, and 8-μm sections were prepared for histological analysis. Cell culture in 2D HDFs and human primary keratinocytes originating from foreskin tissue of three different donors were provided by H.-D. Beer, University Hospital Zurich, Switzerland. The foreskin had been collected with informed written consent of the parents in the context of the Biobank project of the Department of Dermatology, University Hospital Zurich, and upon approval by the local and cantonal research ethics committee. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) high glucose (#D6429; Sigma-Aldrich, St. Louis, MO) containing 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (Sigma-Aldrich) at 37°C with 5% CO[2]. Medium for fibroblasts expressing shRNA against PRSS35 was supplemented with puromycin (2 μg/ml). Medium for PRSS35-(BirA)-HA fibroblasts was supplemented with G418 (1 mg/ml). Passage numbers did not exceed P20. Cells were routinely tested for mycoplasma contamination using a PCR mycoplasma test kit I/C (PromoKine, Heidelberg, Germany). Human immortalized, but nontransformed, HaCaT keratinocytes were cultured in DMEM/10% FBS. Encapsulation of fibroblasts into hydrogels GelMA, PEG–norbornene (PEG-Nb), and lithium phenyl-2,4,6-trimethylbenzoyl-phosphinate (LAP) were synthesized according to established protocols ([239]62–[240]64) using porcine skin gelatin (#G2500; Sigma-Aldrich), methacrylic anhydride (#276685; Sigma-Aldrich), and 20-kDa 4-Arm PEG-Amine (#SKU: 4ARM-NH2; JenKem, Plano, TX). GelMA hydrogel mixes contained 0.1% LAP, 4% GelMA, and 1 million fibroblasts. PEG hydrogel mixes contained 0.125% LAP, 3.25% PEG-Nb, 2.35 mM matrix metalloproteinase–degradable peptide (KCGPQGIWGQCK; GenScript), 1.8 mM adhesion peptide RGD (CRGDS; GenScript), and 1 million fibroblasts. All percentages represent w/v. Gels were formed between two Sigmacote-functionalized glass slides at a thickness of 0.5 mm (#SL2; Sigma-Aldrich) and polymerized for 90 s using ultraviolet light (405 nm, 14 mW/cm^2). They were rinsed with phosphate-buffered saline (PBS), and cells were cultured in the gels for 3 days before experimental treatment. Scaffold-free 3D organotypic skin cultures 3D organotypic skin cultures were generated according to the protocol from Berning et al. ([241]26). HaCaT keratinocytes and HDFs were used. For separating epidermal and dermal equivalents, cultures were placed, with the epidermis facing up, on dispase II (2 mg/ml; #17105-041; Gibco, Carlsbad, CA) in DMEM at 37°C without CO[2] for 30 min. Layers were peeled apart and snap-frozen in TRIzol reagent (#15596; Life Technologies, Carlsbad, CA). Transient transfection HDFs were seeded at 70% confluency in 24- or 6-well plates. siRNAs (2.5 pmol/24-well or 15 pmol/6-well) against NFAT5 (#J-009618-06 and #J-009618-07; Dharmacon, Lafayette, CO), ATF2 (#J-009871-07 and #J-009871-08; Dharmacon) or PRSS35 (#J-008433-05 and #J-008433-07; Dharmacon) were used for transfection using Lipofectamine RNAiMAX (#13778150; Thermo Fisher Scientific, Waltham, MA). The transfection mixture was left on the cells for 1.5 to 2 days. Sorbitol was added to the transfection mixture at a final concentration of 200 mM. Generation of genetically modified fibroblasts For cloning of the PRSS35-(BirA)-HA plasmid, the FL PRSS35 coding sequence amplified from the intronless PRSS35 genomic DNA of fibroblasts was fused in frame with (BirA)-HA from MCS-BioID2-HA (#74224; Addgene, Watertown, MA) and introduced into pInducer20 (#44012; Addgene). The PRSS35Pro-FFLuc construct was generated by replacing the constitutive MSCV promoter from the BLIV 2.0 Lentivector (#BLIV713PA-1; System Biosciences, Palo Alto, CA) by the osmoresponsive PRSS35 promoter (PRSS35Pro, −1000 bp upstream of the transcription start site). PRSS35Pro had been amplified from fibroblast genomic DNA. Plasmids with nontargeting (#[242]VSC11657; Horizon Discovery, Waterbeach, UK) and PRSS35-targeting (#V3SH11252-229744426 and #V3SH11252-230227744; Horizon Discovery) shRNAs were commercially obtained. Lentiviruses packaged with PAX2 (#12260; Addgene) and vesicular stomatitis virus glycoprotein (#8454; Addgene) were collected from human embryonic kidney 293T cell supernatants 2 days after transfection. Viral supernatant was filtered through a 0.45-μm filter, supplemented with polybrene (8 μg/ml; #sc-134220; Santa Cruz Biotechnology, Santa Cruz, CA) and used for overnight (O/N) treatment of HDFs. The next morning, cells were washed and incubated with fresh DMEM supplemented with polybrene. Selection with puromycin or G418 was started 2 days after infection. RNA isolation For RNA isolation from cells in 2D, cells were lysed in RB buffer from the Total RNA Mini Kit (#IB47320; IBI Scientific, Dubuque, IA) and scraped off from the plate using a cell scraper. RNA was purified using the Total RNA Mini Kit according to the manufacturer’s instructions. For RNA isolation from hydrogels and 3D organotypic skin cultures, samples were homogenized in TRIzol reagent (#15596; Life Technologies), and RNA was extracted according to the manufacturer’s instructions for tissue extraction. Reverse transcription and qRT-PCR RNA was reverse-transcribed using the iScript cDNA Synthesis Kit (#1708890; Bio-Rad, Hercules, CA) according to the manufacturer’s instructions, and qRT-PCR was performed as described previously ([243]65). Primer sequences are listed in table S3. Preparation of protein lysates and Western blot For nuclear-cytoplasmic fractionation, cells were lysed on ice using 0.1% NP-40/PBS supplemented with PhosSTOP (#4906845001; Roche, Rotkreuz, Switzerland) and protease inhibitor cocktail (#05056489001; Roche). The supernatant following centrifugation of the total lysates at 17,000g for 3 min at 4°C was collected as cytoplasmic fraction. The nuclear pellet was washed on ice six to eight times using the lysis buffer. For isolation of secreted proteins, cells were cultured in Opti-MEM (#31985062; Thermo Fisher Scientific). Harvested supernatants were incubated with 10% trichloroacetic acid O/N at 4°C. Proteins were precipitated by centrifugation at 17,000g for 12 min at 4°C and washed using ice-cold acetone. Pellets were solubilized in 1% SDS. For whole-cell lysate preparation, cells were lysed on a 95°C hot plate with Laemmli buffer. All samples were boiled in Laemmli buffer, and dithiothreitol was added. Western blot was performed as described previously ([244]65). Primary and secondary antibodies and respective concentrations are depicted in tables S4 and S5. Coimmunoprecipitation PRSS35-BirHA–expressing HDFs were harvested and processed in the presence of 0.2 mM dithiobis(succinimidyl propionate) (Lomant’s reagent; #22585, Thermo Fisher Scientific) as previously described ([245]41). PRSS35–BirA-HA was pulled-down using HA-coupled beads (#88836; Thermo Fisher Scientific). Pulled-down proteins were analyzed by Western blot. Immunofluorescence staining of cultured cells Fibroblasts were fixed for 20 min at room temperature in 4% paraformaldehyde (PFA). Subsequently, they were permeabilized with 0.5% Triton X-100 in PBS for 5 min. Unspecific binding sites were blocked for 1 hour using 2% bovine serum albumin (BSA)/0.05% Triton X-100 in PBS at room temperature. Samples were incubated with primary antibodies (table S4) in blocking solution O/N at 4°C. The next day, they were incubated with secondary antibodies (1:200) and Hoechst (1:500) in 0.05% Triton X-100 in PBS and mounted with 1% N-propyl gallate. Washing steps in PBS (3 × 5 min) were performed after each antibody incubation step. Images were taken on an Axio Imager.A1 microscope equipped with an AxioCam MRm camera (both from Carl Zeiss AG, Jena, Germany) at 20×, 40×, and 63× magnification. SHG microscopy SHG imaging of ECM deposited for 6 days ([246]Fig. 7B) was performed using a Leica TCS SP8 confocal laser scanning microscope equipped with a 25× 0.95 numerical aperture L Water HCX IRAPO objective, and a Mai Tai XF (Spectra-Physics, Milpitas, CA) MP laser, tunable from 710 to 950 nm. F-SHG signal was collected by a Köhler aligned condenser (0.55 NA), SHG filter (435 to 455 nm), and photomultiplier tube (PMT), while B-SHG signal was collected by an external nondescanned Hybrid detector (HyD7) right after the objective (also 435 to 455 nm). Laser power was kept constant throughout each experiment, as were PMT voltage and gain and HyD gain. Leica SP8 LAS X was used to control the instrument and for image acquisition. Pixel size was set to 253 nm by 253 nm; images were acquired with 48-μs pixel dwell time and 3× line averaging; fields were zoomed to a size of 259 μm by 259 μm. Two by two fields were stitched together with a 10% overlap to cover an area of 494 μm by 494 μm. Image analysis was performed using QPath ([247]66). Proximity ligation assay Fibroblasts were seeded at 20% confluency in 16-well chamber slides (#178599; Thermo Fisher Scientific). Following treatment, cells were fixed for 20 min at room temperature in 4% PFA and permeabilized in 0.5% Triton X-100 in PBS for 5 min. PLA was performed using the Duolink In Situ Red Kit (#DUO92101; Merck, Darmstadt, Germany) according to the manufacturer’s instructions. Duolink In Situ PLA Anti-Goat MINUS (#DUO92008; Merck) and Anti-Rabbit PLUS probes (#DUO92002; Merck) were used for ligation. Antibodies and concentrations are listed in table S4. Photos were taken on an Axio Imager.A1 microscope equipped with an AxioCam MRm camera at 20×, 40×, and 63× magnification. Immunohistochemistry staining of tissue sections PFA-fixed paraffin sections of human and mouse wounds were dewaxed and rehydrated using a xylene/ethanol gradient followed by antigen retrieval using citrate buffer (10 mM citric acid, pH 6.0) at 95°C for 1 hour. Unspecific binding sites were blocked with 12% BSA in PBS for 1 hour at room temperature. Incubation with primary antibodies (table S4) at 4°C O/N was followed by incubation with biotin-conjugated secondary antibodies (table S5). After each antibody incubation step, extensive washing steps in 0.1% Tween in PBS were performed (3 × 10 min). The VECTASTAIN ABC and DAB peroxidase substrate kits (#PK-6100 and #SK-4100; Vector Laboratories, Burlingame, CA) were used for signal visualization according to the manufacturer’s instructions. Sections were counterstained with hematoxylin and eosin ([248]65) and mounted with Eukitt. Immunohistochemistry sections were imaged using a 3DHistech Pannoramic 250 slide scanner (3DHistech, Budapest, Hungary). Sample preparation for matrisome proteomics The experimental setup used for matrisome proteomics is depicted in [249]Fig. 7B. Briefly, 220,000 fibroblasts were seeded in 10-cm dishes and treated with ascorbic acid (50 μg/ml), Dox (100 ng/ml), or vehicle [dimethyl sulfoxide (DMSO)] at the same dilution and 200 mM sorbitol according to the scheme shown in [250]Fig. 7B. The ECM was harvested at day 7. Cells were washed in PBS and lysed in 1 ml of 0.5% Triton X-100 in PBS for 10 s. The ECM was washed with PBS and incubated with 1 ml of 20 mM NH[4]OH in PBS for 10 s, followed by five washing steps in PBS. The ECM was scraped off using 250 μl of 4% SDS in 0.1 M Tris-HCl (pH 7.6). Samples were boiled in Laemmli buffer and dithiothreitol at 95°C for 5 min and stored at −80°C. Sample processing was continued as described ([251]67), separating proteins by SDS–polyacrylamide gel electrophoresis and in-gel digestion using trypsin. The resulting peptide mixtures were desalted using STAGE tips ([252]68) before liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis. Mass spectrometry analysis of matrisome proteins LC-MS/MS measurements were performed on an Exploris 480 mass spectrometer coupled to an Easy-nLC 1200 nanoflow-HPLC (all Thermo Fisher Scientific) as described ([253]69). Briefly, after each mass spectrometry scan (mass range mass/charge ratio= 370 to 1750; resolution, 120,000), a maximum of 20 MS/MS scans was performed using an isolation window of 1.3, a normalized collision energy of 28%, a target AGC of 50%, and a resolution of 15,000. Mass spectrometry raw files were analyzed using MaxQuant software using a UniProt FL human database and common contaminants, such as keratins and enzymes used for digestion ([254]70). Carbamidomethylcysteine was set as fixed modification; protein N-terminal acetylation and oxidation of methionine, lysine, and proline (oxMKP) were set as variable modifications. MaxQuant results were analyzed using Perseus ([255]71). Identification of statistically significant differentially abundant proteins between two groups was performed using t test (FDR < 0.05, minimally two valid values per group). To identify significantly regulated lysine and proline hydroxylation sites (P < 0.05, FC > 2), sites had to be identified in minimally three samples (FDR < 0.01) and normalized to protein levels. Missing values were imputed (width, 0.3; down shift, 1.8) and filtered on a localization probability >0.75. Biotin labeling and sample preparation for BioID mass spectrometry PRSS35–BirA-HA–expressing HDFs were pretreated with Dox (20 ng/ml) or vehicle (20 ng/ml; DMSO) for 16 hours, followed by 24 hours of incubation with Dox (20 ng/ml) or vehicle (20 ng/ml) + 50 μM biotin (#B4501; Sigma-Aldrich). Biotinylated proteins were pulled down from total lysates using Dynabeads MyOne Streptavidin C1 (#65001; Thermo Fisher Scientific) as described previously ([256]72). The pull-down efficiency of biotinylated proteins was monitored by Western blot. Trypsin digestion and mass spectrometry analyses were conducted by the Functional Genomics Center Zurich. On-bead digestion was performed using trypsin (100 ng/μl) in 10 mM Tris/2 mM CaCl[2] buffer at pH 8 for 30 min at 60°C. Residual peptides were collected using 0.1% trifluoroacetic acid/50% acetonitrile and pooled with the supernatant from the digestion step. Mass spectrometry analysis of biotinylated proteins For LC-MS/MS analysis, the peptide mixtures were dried and dissolved in ddH[2]O + 0.1% formic acid, transferred to autosampler vials, and injected to a nanoACQUITY UPLC coupled to a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific). Mass spectrometry data were processed using the MaxQuant search engine (version 2.0.1.0.). Spectra were searched against the PRSS35 sequence merged with the Homo sapiens background database (Swiss-Prot, Lausanne, Switzerland) using the following variable modifications: acetyl (Protein N-term), oxidation (M), deamination (NQ), and no fixed modifications. Protein identification results were imported into the Scaffold software (Proteome Software, Portland, OR), and the following settings were applied: protein FDR = 1%, minimum number of peptides per protein = 2, and peptide FDR = 0.1%. The candidate interaction partners were uploaded to the Contaminant Repository for Affinity Purification database for further analysis using SAINT and filtered for proteins with SP of ≥0.9. Proteins identified in the PRSS35–BirA-HA BioID experiment are listed in table S1. Luciferase assay FFLuc activity in PRSS35Pro-FFLuc fibroblasts following hyperosmotic stimulation was measured using the Dual-Luciferase Reporter Assay System (#E1910; Promega, Madison, MA). 5-Bromo-2′-deoxyuridine incorporation assay Cells at 50 to 70% confluency were transfected in Opti-MEM and treated with sorbitol and 10 μM 5-bromo-2′-deoxyuridine (BrdU) for 32 hours. They were fixed with 4% PFA for 20 min at room temperature and washed twice with PBS containing 0.1% Triton X-100, followed by incubation with 2 M HCl/0.1% Triton X-100 for 30 min at room temperature. Washing steps with 0.1 M sodium decahydrate tetraborate at pH 8.5 for 5 min and 0.1% Triton X-100 in PBS followed. Unspecific binding sites were blocked with 10% BSA in PBS for 10 min at room temperature. Cells were incubated with rabbit anti–BrdU–fluorescein isothiocyanate (#1202693; Roche) in 1% BSA in PBS at 4°C O/N. Nuclei were stained using Hoechst 33342 (Thermo Fisher Scientific) in 1% BSA in PBS. Cells were mounted onto microscope slides using Mowiol (Sigma-Aldrich). Images were acquired using an Axio Imager.A1 microscope equipped with an AxioCam MRm camera (Carl Zeiss AG). Further image processing and analysis were performed using ImageJ software (National Institutes of Health, Bethesda, MD). MTT viability assay Cells subjected to different treatment regimens were incubated with MTT at 5 mg/ml in PBS for 30 min. They were lysed in 0.33% HCl in isopropanol for 10 min. The reaction was stopped by adding an equal volume of ddH[2]O. The absorption at 590 nm was measured using a Promega GloMax Plate Reader (Promega). Lactate dehydrogenase cytotoxicity assay Lactate dehydrogenase (LDH) activity in the supernatant of treated cells was determined using the Pierce LDH Cytotoxicity Assay Kit (#88953; Thermo Fisher Scientific) according to the manufacturer’s instructions. Absorption at 490 and 600 nm was measured using a Promega GloMax Plate Reader (Promega). Isolation of matrisomes for functional assays A total of 60,000 fibroblasts were seeded in a four-chamber slide system (#177437; Thermo Fisher Scientific). Cells were treated as depicted in [257]Fig. 7B and fig. S8A and left for ECM production for 6 days. For harvesting the matrisome, cells were washed with PBS at day 7 and lysed in sterile 0.5% Triton X-100 in 20 mM NH[4]OH for 30 s. The matrisome was washed carefully with PBS and treated with sterile deoxyribonuclease I (100 U/ml; #10104159001; Merck) in PBS for 60 min at room temperature. The decellularized matrisome was washed with PBS, and cells were directly added on top. Analysis of fibroblast proliferation on decellularized matrisome A total of 20,000 fibroblasts were seeded in 1% FBS/DMEM on decellularized matrisome. The next day, they were fixed and stained for the proliferation marker Ki67 (table S4) as described previously ([258]65). RNA-seq data quantification and statistical analysis Kallisto software (version 0.44.0) ([259]73) was used for reads alignment and gene quantification with the following parameters: quant -t 8 --bias --bootstrap-samples 10 --seed 42 --rf-stranded and reference genome version GRCh38.p13 from the GENCODE project with gene model annotation version 32. To exclude extremely weakly expressed genes, the data were filtered by estimated counts with the threshold of 20. The edgeR software (version v3.28.) ([260]74) was used for differential expression analysis with the following parameters: Normalization method – TMM (trimmed mean of m-values), data modelling – glm (generalised linear model), statistical test – QL (quasi likelihood), and Benjamini-Hochberg (BH) multiple testing correction. Ingenuity pathway analysis IPA of differentially regulated genes (1 ≤ log[2] FC ≤ −1, P < 0.01) was performed using the IPA software (#26127183; QIAGEN, Hilden, Germany). Prediction of transcription factor binding sites The Eukaryotic Promoter Database was used to predict transcription factor binding sites within the PRSS35 core promoter (1000 bp upstream of the transcription start site). Sites with a binding probability of P < 0.001 were considered significant hits. GSEA and leading edge analysis Sets of significantly up- and down-regulated genes were generated by reanalyzing relevant published transcriptomic datasets related to osmotic stress and wound healing, which were downloaded from Gene Expression Omnibus (GEO) or from the respective publication (table S6). The gene sets were uploaded to GSEA (version 4.2.1) ([261]75) and filtered to those mapped by gene symbol and those present in the tested datasets. Two GSEA runs were completed, testing the skin fibroblast hyperosmotic stress signatures following sorbitol treatment for 9 and 24 hours against gene sets related to osmotic stress and wound healing (table S6). GSEA results were organized into tables, and the NES and FDR values were color-coded for visualization. Leading edge analyses were extracted to identify commonly enriched genes between gene sets. The GSEA run settings, gene sets, and all GSEA output data, including leading edge analysis details, are provided in table S6. Analysis of spatial RNA-seq datasets of skin wounds The spatial RNA-seq datasets from 7-day splinted mouse skin wounds and unwounded skin were downloaded from GEO ([262]GSE178758) ([263]34) and reanalyzed using the Seurat R package (version 4.2.0) ([264]76). The raw data were uploaded using the Load10X_Spatial function. The wound dataset was cropped to include only the undamaged area of the main wound section via the “subset” function. The cropped dataset was normalized via SCTransform function using standard settings ([265]77), followed by RunPCA, RunUMAP, and FindNeighbors functions with dims = 1:30. Spatial voxel clustering for major skin wound morphological areas was done via the FindClusters function with resolution = 0.5, resulting in six major spatial clusters roughly defining the following areas: normal dermis, epithelium, hypodermis, edge granulation, upper granulation, and central granulation. Relative expression of genes of interest between spatial clusters was visualized by DotPlot. Gene coexpression analysis was performed by first extracting gene count data from Seurat objects and using the cor.test function with method = “spearman.” Coexpression scatter plots were performed by ggplot with a linear regression fit line. Gene expression signature scoring was done via the AddModuleScore function using standard settings and the osmotic stress gene set consisting of the validated osmosensitive genes in fibroblasts: Prss35, Spp1, Gper1, Adora2a, Arg2, Me1, Dhcr7, Gdf15, and Mmp13. The signature score was overlaid on the histological section via the SpatialDimPlot function. Analysis of single-cell RNA-seq datasets of skin wounds The single-cell RNA-seq datasets from 7-day mouse skin wounds were downloaded from GEO [[266]GSE153596 ([267]43); [268]GSE20477 ([269]44)] and reanalyzed using the Seurat R package (version 4.2.0). For the first study, Loom files were first imported via the connect function of the loomR package (version 0.2.0) and converted to Seurat objects via the as.Seurat function of the SeuratDisk package (version 0.0.0.9020). For each study, all datasets were first analyzed separately for quality control by eliminating low-quality cells with low RNA features, as well as those with relatively high mitochondrial and ribosomal RNA content via the subset function, followed by SCtransform using the vars.to.regress modifier to additionally regress out the mitochondrial and ribosomal RNA features ([270]77). For the second study, time point and wound location specific barcodes were determined for each cell using the HTODemux hash tag oligo demultiplexing function. The datasets were then merged and integrated using the merge, SelectIntegrationFeatures, PrepSCTIntegration, FindIntegrationAnchors, and IntegrateData functions, according to the published data analysis and integration protocol ([271]78). The integrated datasets were analyzed via the RunPCA function, followed by ElbowPlot to visualize the SD of the principal components and to select the most significant dimensions for further analyses. The RunUMAP and FindNeighbors functions were then run using standard settings and the dims = 1:6 and 1:30 modifier for the first and second study, respectively. This was followed by FindClusters with resolution = 0.1 to cluster the dataset on major cell types. Marker genes of skin cells were obtained from CellMarker 2.0 ([272]79) and mapped onto the dataset by DotPlot and FeaturePlot. For the first study, the following major cell types were identified and annotated: differentiated keratinocytes, activated keratinocytes, melanocytes, immune cells, endothelial cells, and fibroblasts. For the second study, additional cell types were identified, including immune cell and other cell clusters. For each study, wound fibroblasts were analyzed in depth by first subsetting the (myo)fibroblast clusters via the subset function, followed by the downstream steps outlined above. Marker genes for fibroblast subtypes were determined by the FindAllMarkers function. Gene coexpression analysis was performed by first extracting gene count data from Seurat objects and using the cor.test function with method = “spearman.” Coexpression scatter plots were performed by ggplot with a linear regression fit line. Statistical analysis Statistical testing was performed using GraphPad Prism (GraphPad, San Diego, CA). One- or two-way analysis of variance (ANOVA) with Bonferroni, Dunnett, or Tukey correction was used for multiple comparisons. Student’s t test was used for comparisons between two groups (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). Acknowledgments