To assess whether and how aging affects HF innate immunity, we analyzed available RNA sequencing data of mouse HFSCs (Doles et al., 2012). Pathway analysis revealed that innate and adaptive immunity, as well as TLR signaling, were among the top dysregulated pathways (Figure 1A). Notch, JAK-STAT, TGF-β, and Wnt, and other pathways essential for HF regeneration were also altered by aging (Figure 1A). Notably, the level of Tlr2 mRNA in HFSCs of old mice was 2-fold lower compared to young mice (Keyes et al., 2013). In addition, TLR2 at the protein level was substantially lower in 13-month-old mice compared to 2-month-old mice (Figure 1B and C).

At the same time, the normal hair cycle is marked by an increase in Tlr2 mRNA levels prior to HFSC activation during telogen, according to the analysis of existing RNA microarray data (Figure 1D; Greco et al., 2009). Tlr2 mRNA levels are highest among other Tlrs, with Tlr1 and Tlr4 mRNA showing the opposite pattern. TLR6 may act as a co-receptor for TLR2 as its expression pattern is similar. While the downstream TLR2 signaling molecules, Irak1 and Myd88, are also upregulated, mirroring the Tlr2, IRAK1 inhibitor Tollip is suppressed. Together, the results suggest the critical regulatory role of the entire TLR2/TLR6 pathway in the HF cycle.

High-fat diet-induced obesity causes hair thinning and subsequent loss (Morinaga et al., 2021). In our model, a high-fat diet causes a nearly 2-fold decline in TLR2 protein level in HFSCs, compared to normal diet-fed mice (Figure 1E and F). Further, RNA sequencing data reveal that 3 months of a high-fat diet is sufficient to reduce Tlr2 levels in HFSCs by more than 2-fold compared to control mice (Figure 1G; Morinaga et al., 2021). Thus, our results and analyses of existing datasets demonstrate that conditions causatively associated with hair thinning and loss, such as aging and obesity, result in a dramatic depletion of TLR2 in HFSCs suggesting a possible regulatory role for TLR2 in HFs.

The expression of TLR2 during a normal hair cycle was assessed using a previously characterized TLR2-GFP reporter mouse (Figure 2), one of the best tools for the analysis of TLRs in vivo (Price et al., 2018). The correlation between the reporter and TLR2 protein expression was confirmed (Figure 2—figure supplement 1A and B). The HF cycle was verified by H&E staining (Figure 2—figure supplement 1C). During telogen, a dormant stage for HFSCs, TLR2 was found in the bulge, secondary hair germ (sHG), dermal papilla (DP), and outer root sheath (ORS) (Figure 2A). During regenerative anagen, TLR2 expression was detected in the DP and all sHG-derived progenitor cells (Figure 2B and C) and quiescent bulge stem cells (Figure 2D and E). All cells derived from bulge stem cells (ORS) and sHG (hair shaft and inner root sheath [IRS]) were positive for TLR2 (Figure 2C). In catagen, TLR2 was abundant in the new bulge and sHG formed from ORS cells (Hsu et al., 2011; Figure 2F and G). While TLR2 was expressed in early sHG lineage (including IRS) (Figure 2C), it was absent in mature (Figure 2D and E) and regressing (Figure 2F) IRS. This shows that TLR2 is abundant in stem cells but declines upon differentiation. The second telogen’s old and new bulges expressed TLR2 (Figure 2H). TLR2 was present in sHG and DP during the second telogen (Figure 2H) and increased in the late (competent) telogen compared to the early (refractory) telogen (Figure 2—figure supplement 1D and E). The highest level of TLR2 occurred during active anagen compared to quiescent telogen and catagen (Figure 2I). Quantitative polymerase chain reaction (qPCR) revealed that the Tlr2 mRNA level in HFSCs was 5- and 2.3-fold higher in anagen than in telogen and catagen, respectively (Figure 2J). Notably, analysis of existing RNA sequencing data using FACS-sorted cells (Lorz et al., 2010) confirmed that TLR2 expression was significantly higher in HFSCs than in epidermal or non-stem cells (Figure 2K). Thus, TLR2 is enriched in HFSCs, and its expression increases during activation.

To address the role of TLR2 in the hair cycle, we generated an HFSC-specific inducible Tlr2 knockout (KO) mouse line (TLR2^HFSC-KO) and deleted Tlr2 during the first postnatal telogen (Xiong et al., 2022b). In TLR2^HFSC-KO mice, the telogen phase was substantially prolonged compared to control (Tlr2^lox/lox) mice, as summarized in the schematic (Figure 3A). Melanogenesis and anagen onset are tightly coupled (Müller-Röver et al., 2001). Thus, the pink skin color at P21 marks the first postnatal telogen. The onset of anagen in control mice was indicated by the change in skin color to gray or black at P26 in control mice. TLR2^HFSC-KO mice at this age did not enter anagen, as evidenced by the delayed darkening of their skin (Figure 3A–C). This was further confirmed by skin section analysis at P21, P26, and P35 (Figure 3D). At P26, control mice displayed pigmented anagen HFs with enlarged bulbs located deeply in the hypodermis, while nearly all follicles of TLR2^HFSC-KO mice were ~5-fold shorter and remained in the dermis on top of adipose tissue, a characteristic of telogen. On day P35, we observe partial entrance into anagen in TLR2^HFSC-KO skin while the skin color of TLR2^HFSC-KO mice remains pink (Figure 3D–F). HFSCs were activated as early as at P24 in control mice based on positive Ki67 staining in sHG and bulge region (Figure 3G), while most cells in TLR2^HFSC-KO sHG (Figure 3H) and bulge (Figure 3I) remained quiescent. At P25, control mice exhibited a large cluster of P-cad+ cells encapsulating DP within the transformed sHG (Figure 3K), whereas the sHG of TLR2^HFSC-KO mice remained small and inactive (Figure 3J and K and Figure 3—figure supplement 1). Despite the substantial delay in anagen onset, the morphology of HFs and expression of established HFSC markers, including Ker15, CD34, and Sox9, were normal in TLR2^HFSC-KO mice (Figure 3—figure supplement 2). Thus, TLR2 in HFSCs is essential for HFSC activation and progression of the hair cycle.

The relationship between Wnt and BMP signaling is central to the cyclic growth of HFs (Plikus et al., 2008). Anagen initiation is triggered by Wnt/β-catenin activation, while BMP signaling suppresses HFSC activation and its reduction is necessary for HFSC activation. Indeed, during the early (refractory) phase of the second telogen, HFSCs exhibit elevated BMP signaling as evidenced by high levels of BMP7 protein and pSMAD1/5/9, downstream targets of BMP signaling, compared to the late (competent) phase (Figure 4A–D). In contrast, Bmp7 and its effectors, Id1 and Id2, are decreased during the late telogen based on our analysis of existing RNA microarray (Greco et al., 2009; Figure 4—figure supplement 1A).

To assess the possible connection between the TLR2 signaling and BMP pathway in human cells, we activated TLR2 and BMP signaling in human epidermal keratinocytes (NHEK) using a canonical TLR2 agonist (Pam3CSK4) and BMP4, respectively. As anticipated, BMP4 promoted the phosphorylation of its downstream target SMAD1/5/9. However, simultaneous co-activation of TLR2 diminished BMP4 signaling (Figure 4—figure supplement 1B and C).

Likewise, stimulation of human HFSC with canonical TLR2 agonist Pam3CSK4 promoted cell proliferation by 1.5-fold compared to controls. Notably, this effect was diminished in the presence of a TLR2-blocking antibody (Figure 4—figure supplement 1D and E). These results reveal that TLR2 activation on human HFSC augments their proliferation.

To gain a deeper understanding of TLR2’s role in HFSC activation, we profiled the transcriptome of HFSCs lacking Tlr2 expression (Figure 4—figure supplement 1F). The results showed that Tlr2 deletion dysregulated 486 genes, many of which were involved in both the hair cycle and innate immunity. The most affected pathways included innate immunity response and TLR2 signaling, with its downstream target NF-kappaB (Figure 4—figure supplement 1G). This profile somewhat resembles changes observed in aging models (Figure 1A).

Since TLR2 suppresses BMP signaling and promotes HFSC proliferation, we assessed whether the delayed anagen in TLR2^HFSC-KO mice might be associated with the BMP pathway. qPCR analysis reveals that several key components of the BMP pathway were dysregulated in HFSCs lacking Tlr2 (Figure 4E). Among those, the most notable changes were observed for Bmp7, which was upregulated by ~4-fold in TLR2-null HFSCs compared to controls (Figure 4E). This was substantiated by co-staining of tissue sections for BMP7 and CD34, which demonstrated an ~2-fold increase in BMP7 on HFSCs of TLR2^HFSC-KO mice as compared to the control (Figure 4F and G). Activation of BMP signaling was assessed by pSMAD1/5/9 positive staining in HF. Quantification revealed an ~15-fold higher level of pSMAD1/5/9 in TLR2-null mice (TLR2^KO) as compared to controls (wild type [WT]) (Figure 4—figure supplement 1H and I). The significant increase in BMP signaling observed was attributed to the absence of TLR2 in HFSCs. This was evidenced by a comparable 14-fold increase in BMP signaling in follicles of TLR2^HFSC-KO mice compared to control during the first postnatal telogen phase, thereby ensuring the preservation of follicles in the dormant telogen stage (as shown in Figure 4H and I). Simultaneously, the Wnt signaling and β-catenin stabilization within HFSCs, known to trigger their activation (Deschene et al., 2014), remained unchanged between control and TLR2^HFSC-KO mice (as shown in Figure 4—figure supplement 1J).

To demonstrate that an altered BMP pathway is, indeed, responsible for the phenotype of TLR2 KO in HFSCs, we utilized intradermal injection of noggin, a well-known inhibitor of BMP signaling (Botchkarev et al., 2001; Botchkarev et al., 1999), to block the upregulated BMP signaling in TLR2^HFSC-KO mice. As a result, noggin injection diminished activation of BMP signaling by >10-fold in TLR2^HFSC-KO mice as assessed by pSMAD1/5/9 staining of HF (Figure 4J and K). Moreover, noggin promoted activation of TLR2^HFSC-KO HFs while the HFs in BSA-treated TLR2^HFSC-KO mice remained quiescent. Noggin treatment of TLR2^HFSC-KO mice dramatically upregulated cell proliferation within sHG as evidenced by Ki67⁺ cells (Figure 4L and M), promoting ~2.5-fold increase in activated follicles (Figure 4N and O), thereby contributing to nearly 2-fold larger sHG (Figure 4P and Q) as compared to BSA-treated TLR2^HFSC-KO mice. Thus, curbing suppressive BMP signaling in TLR2^HFSC-KO mice can reactivate their HFs, demonstrating a causative connection between TLR2 and BMP pathways in the hair cycle.

High expression of TLR2 and its critical role in HFSC activation during the hair cycle prompted us to test the role of HFSCs TLR2 in an injury model where cells are more likely to be exposed to TLR2 ligands. First, we compared TLR2 levels in HFs in wounded and healthy skin using TLR2-GFP reporter mouse (Figure 5A). In healthy skin, HFSCs upregulated TLR2 during their transition from middle to late telogen (day 5 to day 10) (Figure 5B, upper panels, and gray bars in Figure 5C), consistent with RNA sequencing results (Greco et al., 2009). This increase in TLR2 precedes HFSCs activation during the normal cycle. However, in wound HFSCs, TLR2 was upregulated immediately after an injury resulting in 1.5-fold higher expression compared to normal unwounded skin (Figure 5B and C).

Hair regeneration after injury represents a substantial part of the healing process (Abbasi and Biernaskie, 2019; Chen et al., 2015; Wang et al., 2017). We next assessed the role of TLR2 using age- and gender-matched WT and TLR2^KO mice. The lack of Tlr2 visibly impaired hair regeneration after wound healing (Figure 5D). On day 14 post-injury, HFs in WT mice entered precocious anagen judged by a spot of pigmented skin, which, on day 21, developed into a black hair patch (Figure 5D). In contrast, the follicles of TLR2^KO mice remained quiescent lacking regenerated HFs around wounds even after 21 days post-injury (Figure 5D and E). At this point, the pigmented skin area in WT was ~9-fold larger than in TLR2^KO mice. Skin flaps showed substantial pigmentation and growing hair bulbs around wounds in WT, indicative of active anagen. In contrast, the TLR2^KO skin flap was devoid of pigmentation, consistent with telogen (inner skin flap in Figure 5D). Ki67 staining confirmed an increase in HFs’ activation in WT but not in TLR2^KO skin (Figure 5F and G). The resulting density of regenerated HFs based on Ker17 staining in WT was 2-fold higher than in TLR2^KO mice (Figure 5H and I). Most importantly, this effect was dependent on TLR2, specifically on HFSCs, since TLR2^HFSC-KO mice exhibited a similar phenotype with a dramatic reduction in pigmentation and hair growth compared to control mice (Figure 5J and K). The upregulation of pSmad1/5/9 in TLR2^HFSC-KO wounds compared to controls demonstrates that similar to the HF cycle scenario, increased BMP signaling might contribute to diminished HF regeneration (Figure 5L and M). Thus, the TLR2-BMP axis in HFSCs governs HF regeneration after injury.

One of the most important endogenous ligands for TLR2 is CEP, which is a naturally occurring product of PUFA oxidation shown to be accumulated during inflammation and wound healing (West et al., 2010; Xiong et al., 2022a). Healthy tissues are typically devoid of this product, which is mainly associated with inflammation and pathologies (Yakubenko et al., 2018). However, in contrast to other tissues, healthy HFs exhibited high levels of CEP accumulation (Figure 6A). During anagen, CEP is present within the proximal part of the follicle, while in telogen the entire follicle is encased by this PUFA metabolite (Figure 6B–E). Generation of CEP from PUFA is directly aided by myeloperoxidase (MPO) (Xiong et al., 2022a; Yakubenko et al., 2018). MPO is present in abundance in sebaceous glands, possibly as a part of immune defense (Figure 6—figure supplement 1A). Even more surprising, in contrast to other organs and tissues, CEP in HFs is substantially depleted with age (Figure 6F and G), and this decline coincides with the reduction in the regenerative potential of HFs. This is likely due to a decreased level of MPO during aging (Figure 6—figure supplement 1B and C).

The connection between CEP levels and hair thinning and loss in aging prompted us to test whether exogenous CEP can activate TLR2 in HFSCs and stimulate their proliferation. Our in vitro experiments revealed that CEP increases the proliferation of human HFSC in a TLR2-dependent manner since the blockade of TLR2 abrogates the CEP effect (Figure 6—figure supplement 1D and E). In another model, CEP promotes cell proliferation of human hair follicle dermal papilla cells (HFDPCs) by ~2-fold compared to control (Figure 6—figure supplement 1F).

Next, we show that CEP promotes hair regeneration in injury in a TLR2-dependent manner. CEP administration promoted HF regeneration in WT wounds. However, it was ineffective in global TLR2^KO mice (Figure 6—figure supplement 1H). CEP promoted a 55% increase in the number of HF and cell proliferation in WT wounds, and at the same time, there was no effect in TLR2^KO wounds (Figure 6—figure supplement 1I,J).

To ensure independence from immune cells, WT and TLR2^KO mice were irradiated and transplanted with WT bone marrow prior to wounding. Applying CEP on wounds in WT/WT chimeras promoted cell proliferation, thereby dramatically increasing the density of HFs (Figure 6H and I, Figure 6—figure supplement 1G). At the same time, CEP was not effective in TLR2^KO/WT mice (Figure 6H and I, Figure 6—figure supplement 1G), demonstrating the TLR2-dependent mechanism.

These CEP effects were mediated by TLR2 on HFSCs. In control mice, CEP effectively initiated regeneration of HFs around the wound (Figure 6J), resulting in ~3-fold higher density of HF (Figure 6K and L) and dramatic acceleration of cell proliferation by >10-fold (Figure 6M and N) as compared to TLR2^HFSC-KO mice where CEP was mainly ineffective. A similar stimulatory effect of CEP was observed in a primary keratinocyte culture (Oshimori and Fuchs, 2012). CEP dramatically promoted WT but not TLR2^KO keratinocyte proliferation (Figure 6O and P). CEP was an effective stimulator of TLR2 signaling as judged by augmented Nfkb2, Il1b, and Il6 expression in HFSCs upon treatment with CEP (Figure 6Q). Consistent with the key role of BMP signaling in TLR2-dependent HF regeneration, CEP treatment suppressed inhibitory Bmp7 expression by ~2.5-fold (Figure 6R), demonstrating that endogenous and natural TLR2 ligand can counteract an inhibitory effect of BMP7 to stimulate HFSCs’ activation (Figure 6S).

Inducible K15-CrePR1 mice (Stock No. 005249), TLR2-GFP reporter mice (Stock No. 031822), and TLR2^KO mice (Stock No. 004650) were purchased from the Jackson Laboratory. Tlr2^flox/flox mice with Exon3 of the Tlr2 gene flanked by two loxP sites were described elsewhere (McCoy et al., 2021). HFSC-specific Tlr2 KO (TLR2^HFSC-KO) mouse line was described previously (Xiong et al., 2022b). Briefly, Tlr2^flox/flox mice were crossed with K15-CrePR1 mice to generate the inducible HFSC-specific Tlr2 KO mouse line. To induce Cre recombinase activity, RU486 (Sigma) was used topically on shaved dorsal skin (1% mixed with Neutrogena Hand Cream) or via intraperitoneal injection (10 mg/ml in corn oil, 75 µg RU486 per 1 kg body weight) during the first postnatal telogen. To block BMP signaling in mouse HFs, at first postnatal telogen after applying Ru486, 200 ng of recombinant mouse Noggin (BioLegend) reconstituted in 30 µl of PBS were injected intradermally into a dorsal skin for 3–5 consecutive days. BSA in PBS was used as vehicle control.

For high-fat diet feeding studies, male WT C57BL/6J were purchased from Jackson Laboratories (Bar Harbor, ME, USA), and at 7 weeks of age, mice were either maintained on standard rodent chow or switched to a high-fat diet containing 60% of kilocalories from fat (Research Diets D12492) for an additional 15 weeks prior to tissue collection. For all animal experiments mice were randomly assigned to the groups (if it had been required), and results were evaluated in a blinded manner. All procedures were performed according to animal protocols (00002319) approved by the Cleveland Clinic IACUC committee. All surgical procedures were performed under ketamine/xylazine anesthesia followed by subcutaneous injection of a single dose of buprenorphine SR after surgery. According to veterinarian recommendations, water with acetaminophen was provided for the next 5-7 days to minimize suffering.

Mouse keratinocytes and HFSCs were isolated from the mouse dorsal skin as described previously (Xiong et al., 2022b). Briefly, isolated dorsal skin samples were trypsinized, the epidermis was scrapped, minced, and filtered through a 70 µm cell strainer to prepare primary keratinocytes single-cell suspension. To isolate HFSC, the single-cell suspension was incubated with CD34-FITC antibody (eBioscience, 11-0341-82), Alexa 647-conjugated CD49f antibody (BD Biosciences, 562494), 7-AAD (BD Biosciences, 559925), and different fluorescence minus one was used as a control. Cells were then sorted by BD FACS Aria and analyzed by Flow Jo. All the primary cells were used within 48 hr for experiments.

Human HFDPCs, mycoplasma tested, were purchased from Cell Applications, Inc (cat.# 602-05a). Human HFSCs, mycoplasma tested, were purchased from Celprogen (cat.# 36007-08). Human epidermal keratinocytes, neonatal, pooled, mycoplasma tested, were purchased from Lonza Reagents (cat.# 192906).

Mouse skin samples were harvested at indicated ages and fixed in 4% paraformaldehyde, kept in 30% sucrose for 2–3 days, followed by snap-freezing at –80°C in OCT (Fisher HealthCare, 4585). 10 µm skin sections were permeabilized, blocked, and incubated with primary antibodies followed by incubation with the corresponding secondary antibody, and mounted with an antifade mounting medium with DAPI (Vector Laboratories, H-1500-10). Images were captured on a Leica DM2500 confocal microscope and analyzed using Bitplane Imaris software (version 9.7.2) or ImageJ. Briefly, image z-stacks were loaded into Imaris to reconstruct three-dimensional images, and surface rendering was performed with default settings using the surface tool. The same background subtraction was performed on each z-stack. The green channel was used as a source channel to create surfaces for GFP⁺ cells in the area of interest in HFs, and other channels were created based on the expression of different cell markers (e.g. CD34, Ker5) in the HFs. The overlap between GFP surface and other maker surfaces was created and visualized as the co-localized area with the co-localization module. At least 10 HFs from each mouse were used for quantification.

The following antibodies or reagents were used: Ker17 (Santa Cruz Biotechnology, sc-393002), Ker15 (ABclonal, A2660), MPO (Santa Cruz Biotechnology, sc-390109), GFP (Thermo Fisher Scientific, CAB4211), TLR2 (Santa Cruz Biotechnology, sc-21759), Ki67 (Abcam, ab16667), P-cadherin (R&D Systems, AF761-SP), CEP (Pacific Immunology), pSmad1/5/9 (Cell Signaling Technology, 13820S), β-catenin (Cell Signaling Technology, 8480), CD34 (eBioscience, 11-0341-82), CD49f (BD Pharmigen, 562473), Ker5 (BioLegend, 905903), Sox9 (Cell Signaling Technology, 82,630T), BMP7 (Proteintech, 12221-1-AP), and Nile Red (ATT BioQuest, 250730). As a negative control, we used appropriate isotype match nonimmune antibody: normal mouse IgG2b-PE (Santa Cruz Biotechnology, sc-2868), normal goat IgG control (R&D, AB-108-C), normal mouse IgG (Santa Cruz Biotechnology, sc-2025), normal rabbit IgG (Cell Signaling Technology, 2729S), normal rat IgG (Santa Cruz Biotechnology, sc-2026).

Mouse wound healing procedure was performed as previously described (West et al., 2010; Xiong et al., 2022b). Briefly, an intraperitoneal injection of a ketamine/xylazine cocktail was used to anesthetize 7- to 8-week-old mice. After shaving, full-thickness wounds were made into the dorsal skin using a 6 mm biopsy punch. To examine the effect of CEP on hair regeneration after wound healing, CEP (CEP in polyethylene glycol) or vehicle (polyethylene glycol) was applied to the wounded area every day for 2 weeks. Pictures were taken at different time points to record hair regeneration around the wounded area.

The primary keratinocytes after isolation were plated on rat tail collagen-coated plates with Epilife medium (Gibco, MEPI500CA) supplemented with EDGS (Gibco, S0125). Cells were co-cultured with CEP or control (BSA or PBS) for 48 hr. The cell counting kit 8 (APEXbio, 269070) has been used to measure cell proliferation according to the manufacturer’s protocol.

HFDPCs (Cell Applications, Inc cat.# 602-05a) were cultured in HFDPC Growth Medium (Cell Applications, Inc cat.# 611-500) for 60 hr and then transferred into collagen-coated 48-well plate for 24 hr. After 24 hr cells were washed with 1× D-PBS and incubated in HFDPC Basal Medium contains no growth supplement (Cell Applications, Inc cat.# 610-500) for the next 24 hr. After starvation, cells were incubated with CEP 5 µM or with HFDPC Growth Medium (positive control), or in HFDPC Basal Medium (negative control) for 48 hr. Absorbance was read using cell counting kit 8 (ApexBio, cat.# K1018) on a microplate reader.

Human HFSCs (Celprogen cat.# 36007-08) were cultured in HFSC Un-differentiation Media with Serum (Celprogen cat.# M36007-08US) for 48 hr and then transferred into Undifferentiated ECM 96-Well Plates (Celprogen cat.# UD36007-08-96Well) for 24 hr. After 24 hr cells were washed with 1× D-PBS and incubated in HFSC Serum Free Un-differentiation Media (Celprogen cat.# M36007-08U) overnight. After starvation, cells were incubated for 2 hr with or without TLR2 blocking antibody (Invivogen cat.# mab2-mtlr2) followed by incubation with Pam3CSK4 (Invivogen cat.# tlrl-pms) for 24 hr. HFSC Serum Free Un-differentiation Media was used as a negative control. Absorbance was read using cell counting kit 8 (ApexBio, cat.# K1018) on a microplate reader.

Human epidermal keratinocytes (Lonza Reagents cat.# 192906) were cultured in KGMTM Gold Keratinocyte Growth Medium (Lonza Reagents cat.# 192060) for 60 hr and then transferred into a six-well plate for 24 hr. After 24 hr media was changed, and cells were incubated with Pam3CSK4 10 µg/ml for 1 hr followed by incubation with BMP4 10 ng/ml for 1 hr.

The structure and synthesis of CEP have been described elsewhere (West et al., 2010). To prepare CEP for wound healing, 250 µl CEP in PBS was mixed with 1.1 g polyethylene glycol with sonication in a 45°C water bath for 15 min followed by a strong vortex to mix well. This mixture was stored at 4°C after preparation, warmed to room temperature, and mixed again before use.

Total RNA from primary keratinocytes or HFSCs was isolated with RNeasy Mini Kit (QIAGEN, 74104) and reverse-transcribed into cDNA with PrimeScript RT Master Mix (Takara, RR036A). The real-time PCR was performed using iQ SYBR Green Supermix (Bio-Rad, 1708882) on the Bio-Rad cfx96 qPCR system. Target gene expression levels were normalized to internal control Rps16, and the ΔΔCt method was used to calculate fold change in gene expression. Primers can be found in Supplementary file 1.

Cells were lysed with RIPA Lysis and Extraction Buffer (Thermo Scientific cat.# PI89900) buffer with protease/phosphatase inhibitor cocktail. The lysate was centrifuged at 12,000×g at 4°C for 15 min, boiled with Laemmli buffer for 7 min at 95°C, and transferred to PVDF membranes (Millipore). After blocking, membranes were incubated with primary antibody at 4°C overnight followed by incubation with corresponding secondary HRP-linked antibody. The following antibodies were used for western blotting: Smad1 (D59D7) XP Rabbit mAb (Cell Signaling Technology cat.# 6944), Phospho-Smad1 (Ser463/465)/Smad5 (Ser463/465)/Smad9 (Ser465/467) (Cell Signaling Technology cat.# 13,820P), NF-κB p65 (D14E12) XP Rabbit mAb (Cell Signaling Technology cat.# 8242), Phospho-NF-κB p65 (Ser536) (93H1) Rabbit mAb (Cell Signaling Technology cat.# 3033), Anti-GAPDH antibody EPR16884 Loading Control (Abcam cat.# ab181603).

We performed bone marrow transplant (BMT) as previously described (West et al., 2010) Briefly, 2-month-old male WT or TLR2^KO mice were lethally irradiated with 9 Gy followed by tail vein injection with 10⁷ bone marrow cells isolated from the WT donor femurs. Eight weeks after BMT, mice were subjected to wound healing assay (described above).

First telogen mouse dorsal skin was used for HFSC isolation by FACS. Total RNA was extracted using the RNeasy Mini Kit (QIAGEN, 74104). Sample quality assessment was performed on a Fragment Analyzer electrophoresis system (Agilent). Total RNA was normalized prior to oligo-dT capture and cDNA synthesis with SMART-Seq v4 (Takara). The resulting cDNA was quantified using a Qubit 3.0 fluorometer (Life Technologies). Libraries were generated using the Nextera XT DNA Library Prep kit (Illumina). Medium-depth sequencing (50 million reads per sample) was performed with a NextSeq 550 (Illumina) on a High Output flow cell using 75 base pairs, Paired-End run. Raw demultiplexed fastq paired-end read files were trimmed of adapters and filtered using the program skewer to throw out any with an average Phred quality score of less than 30 or a length of less than 36. Trimmed reads were then aligned using the HISAT2 aligner to the Mouse NCBI reference genome assembly version GRCm38 and sorted using SAMtools. Aligned reads were counted and assigned to gene meta-features using the program featureCounts as part of the Subread package. These count files were imported into the R programming language and were assessed for quality control, normalized, and analyzed using an in-house pipeline utilizing the limma-trend method for differential gene expression testing and the GSVA library for gene set variation analysis. The pathway analysis for differentially expressed genes with adjusted p-value<0.05 was performed using Enrichr web server https://maayanlab.cloud/Enrichr.

Statistical analyses were performed using GraphPad Prism 9. All results are mean ± s.e.m. Shapiro-Wilk normality and lognormality test was used with n≥6. For normally distributed data, we use an unpaired two-tailed t-test to compare two groups and the one-way ANOVA followed by Dunnett’s or Tukey’s post hoc analysis to compare more than two groups. For non-normally distributed data and small sample size (n<6), we appraised statistical differences with the non-parametric Mann-Whitney test to compare two sample datasets and the Kruskal-Wallis test with Dunn’s post hoc test for three or more groups. A p-value ≤ 0.05 was considered to be statistically significant. The sample size was calculated based on a significance level of 0.05 and power 80% (0.8).