Lactate dehydrogenase activity drives hair follicle stem cell activation.

Introduction
The hair follicle is able to undergo cyclical rounds of rest (telogen), regeneration (anagen), and degeneration (catagen). The ability of the hair follicle to maintain this cycle depends on the presence of the hair follicle stem cells, which reside in the bulge (Fig 1). At the start of anagen, bulge stem cells are activated by signals received from the dermal papilla, which at that stage abuts the bulge area 1,2. These stem cells exit the bulge and proliferate downwards, creating a trail that becomes the outer root sheath (ORS). Bulge stem cells are capable of giving rise to all the different cell types of the hair follicle. The ability of HFSCs to maintain quiescence and yet become proliferative for a couple days before returning to quiescence is unique in this tissue, and the precise mechanism by which these cells are endowed with this ability is not fully understood. While significant effort has produced a wealth of knowledge on both the transcriptional and epigenetic mechanisms by which HFSCs are maintained and give rise to various lineages 3,4, little is known about metabolic pathways in the hair follicle or adult stem cells in vivo.
Considering the fact that there are essentially no published data on metabolic states of any cell in the hair follicle, a detailed study of metabolism was necessary to understand the nature of HFSCs and their progeny. Several previous studies employed genetic disruption of the mitochondrial electron transport chain in the epidermis by deletion under the control of a pan-epidermal keratin promoter and found that mitochondrial function was essential for maintenance of the follicle5–8. However, these studies did not explore the metabolic requirements for specific cell types within the tissue, nor did they explore a role for glycolytic metabolism. In this study, we present methods to study the metabolism of HFSCs in vivo, and provide evidence that these cells take advantage of a distinct mode of metabolism not found in their progeny. In the process, we also define small molecules that can take advantage of the unique metabolism of HFSCs to ignite the hair cycle in otherwise quiescent follicles.

Results
Numerous studies have uncovered unique gene expression signatures in HFSCs versus other follicle cells or cells of the interfollicular epidermis 9–12. Many of these signatures are regulated by transcription factors that were later shown to play important roles in HFSC homeostasis13. Lactate dehydrogenase is most commonly encoded by the Ldha and Ldhb genes in mammals, the protein products of which form homo- or hetero-tetramers to catalyze the NADH-dependent reduction of pyruvate to lactate and NAD+-dependent oxidation of lactate to pyruvate14. By immunostaining, Ldha appeared to be enriched in quiescent HFSCs in situ (telogen) (Fig 1a), IHC with an antibody that recognizes both Ldha and Ldhb showed that only Ldha appears to be localized to the HFSC niche (Supplementary Figure 1a).
HFSCs are known to go through successive rounds of quiescence (telogen) punctuated by brief periods of proliferation correlating with the start of the hair cycle (telogen-anagen transition) 4,15. Proliferation or activation of HFSCs is well known to be a prerequisite for advancement of the hair cycle. IHC analysis also showed Ldha expression was enriched in HFSCs (Sox9+) at three stages of the hair cycle (Fig 1a). Consistently, immunoblotting of lysates from sorted cells showed strong expression of Ldha in the basal HFSCs (α6HiCD34+), and suprabasal (α6LoCD34+) HFSC populations relative to total epidermis (Fig 1b)9 (Sorting strategy is outlined in Supplementary Figure 1b).
To determine whether Ldha expression patterns correlate with activity of the Ldh enzyme, we used a colorimetric-based enzymatic assay to assess Ldh activity capacity in situ. Typically performed on protein lysates or aliquots with a plate reader16, we adapted the Ldh activity assay to work in situ on frozen tissue sections. Note that since both the in situ and in vitro Ldh activity assays employ use of excess substrate (lactate), the results from these assays reflect the capacity for Ldh activity, and not the steady-state activity.
Applying this assay to skin samples demonstrated that Ldh activity capacity was significantly higher in HFSCs, consistent with the expression pattern of Ldha (Fig 1c). Furthermore, Ldh activity was enriched in HFSCs across the hair cycle (Fig 1c). As a control, assays conducted without the enzymatic substrate (lactate) or on acid-treated tissue yielded zero activity (Supplementary Figure 1c). To further validate these results, we sorted epidermal populations, generated cell lysates on the sorted cells, and performed a similar colorimetric-based enzymatic assay on the sorted cell lysates, which also showed increased Ldh activity in HFSCs (Fig 1d). To better characterize the metabolism of HFSCs, we performed metabolomics analysis on sorted populations from mouse skin by liquid chromatography-mass spectrometry (LC-MS) (Fig 1e). Several glycolytic metabolites, including glucose/fructose-6-phosphate, fructose-bisphosphate, dihydroxyacetone phosphate, 3-phosphoglycerate, and lactate, were routinely higher in HFSCs relative to total epidermis across three independent experiments (isolated from different mice on different days). Conversely, most TCA cycle metabolites were not consistently different between the epidermis and HFSCs (Fig 1e). Collectively these results suggest that while all cells in the epidermis use the TCA cycle extensively to generate energy, HFSCs also have increased Ldha expression, Ldh activity, and glycolytic metabolism.
Measuring metabolism across the hair cycle therefore would capture any dynamic changes that occur in HFSCs that correlate with activation or quiescence. Analysis of RNA-seq data from HFSCs isolated during either telogen or the telogen-anagen transition demonstrated not only that Ldha is the predominant Ldh isoform expressed in HFSCs (Fig 2c), but is also induced during the telogen-anagen transition (Fig 2a and b (NIHGEOGSE67404 and GSE51635). To confirm that the cells analyzed by RNA-seq were indeed either in telogen or the telogen to anagen transition, important markers of this transition were assessed including the Shh and Wnt pathways (Gli1, 2, 3; Lef1, Axin1, Axin2, Ccnd1) as well as proliferation markers (Ki-67, Pcna and Sox4) (Supplementary Figure 2a).
The in vitro Ldh activity assay on lysates from sorted HFSCs uncovered a modest induction of Ldh activity correlating with the telogen to anagen transition (Fig 2d). Hair cycle staging was validated by Ki-67 immunostaining to determine HFSC activation (Supplementary Figure 2b). Additionally, measurements of steady-state metabolites extracted from sorted HFSCs showed an increase in lactate in HFSCs as they transition from telogen to telogen-anagen transition, and then decrease again in anagen as HFSCs return to quiescence (Fig 2e).
To determine whether Ldh activity is functionally related to the ability of HFSCs to remain quiescent or to activate at the start of a hair cycle, we deleted Ldha specifically in the HFSCs. Taking advantage of mice with floxed alleles of Ldha17, this enzyme was deleted in HFSCs by crossing to mice bearing the K15CrePR allele11, known to be inducible by Mifepristone specifically in HFSCs. Deletion of Ldha in HFSCs was initiated by administration of Mifepristone during telogen (day 50) and led to a typically mosaic recombination of the floxed alleles across the backskin11,18. Mice with HFSC-specific deletion of Ldha failed to undergo a proper hair cycle, with most follicles remaining in telogen across at least 33 pairs of littermates 3–4 weeks after Mifepristone treatment (Fig 3a). A complete list of transgenic animals including birthdate, sex, and genotype is provided in Supplemental Table 1.
Histology showed that WT hair follicles entered into the telogen to anagen transition typically by day 70, and this was accompanied by typical expansion of the hypodermis below (Fig 3b). However, in backskin with deletion of Ldha, the hypodermis did not expand, and the telogen to anagen transition was severely abrogated (Fig 3b). In areas of strong phenotypic penetrance, Ldh activity was severely abrogated in the HFSC compartment (Fig 3c), demonstrating that the Ldha allele is critically important for Ldh activity in HFSCs and consistent with the fact that the ‘a’ isoform of Ldh is expressed at the highest level. Quantification of hair cycle progression across numerous animals indicated that most follicles lacking Ldha remained in telogen (Fig 3d).
In addition, to confirm the phenotypes, we also deleted Ldha with an independent HFSC-specific Cre strategy. Lgr5-CreER has been used for lineage tracing in a variety of adult stem cell models, and has been shown to mark cells with high regenerative capacity, including HFSCs19. Lgr5CreER;Ldhafl/fl mice, treated with tamoxifen at post-natal day 50 prior to a synchronized hair cycle, also failed to activate anagen across at least 20 littermate pairs (Fig 3g). in situ Ldh assay and metabolomics confirmed the successful deletion of Ldha in these animals (Fig 3h and 3i).
We also monitored the effect of loss of Ldha activity in K15+ cells over a six month period and found that deletion of Ldha led to a mosaic, but permanent block of HFSC activation in some portions of the backskin (Supplementary Figure 3a). These data confirm that Ldh activity is required for HFSC activation, and is not simply a marker of HFSCs. A closer look at these long term Ldha deletions showed that Ldha-null HFSCs continued expressing typical markers, but lacked Ldh activity, and failed to initiate new hair cycles, while those follicles that escaped deletion continued to express Ldha and to cycle normally (Supplementary Figure 3b and c).
After sorting HFSCs from animals with or without Ldha deletion, LC-MS-based metabolomics analysis demonstrated that lactate levels, as well as levels of other glycolytic metabolites, were strongly reduced in the absence of Ldha (Fig 3e), functional evidence that the targeting strategy was successful. The fact that glycolytic metabolites upstream of lactate were also suppressed suggests that HFSCs could be adapting their metabolism to account for the loss of Ldh activity. Immunostaining for markers of HFSC activation and proliferation indicated a failure of HFSC activation. Ki67 and pS6 have been clearly demonstrated to be abundant in the HFSC niche at the start of the hair cycle20, and both of these markers were absent in Ldha deleted backskin (Fig 3f). Immunostaining for Ldha also confirmed successful deletion of this protein, while staining for Sox9, a marker of HFSCs indicated that these cells remained in their niche, but just failed to activate in the absence of Ldha (Fig 3f). Induction of the hair cycle is also thought to be regulated by signaling from the Shh, Wnt and Jak-Stat pathways. We assayed each of these by IHC in normal or Ldha deletion follicles and found that in general these pathways were not activated in Ldha-null HFSCs that failed to enter a telogen-anagen transition (Supplementary Figure 3d).
To determine whether induction of lactate production could affect HFSC activation or the hair cycle, we crossed K15CrePR animals to those floxed for mitochondrial pyruvate carrier 1 (Mpc1) (K15CrePR;Mpc1fl/fl). Mpc1, as a heterodimer with Mpc2, forms the mitochondrial pyruvate carrier MPC, a transporter on the inner mitochondrial membrane required for pyruvate entry into the mitochondria21. Loss of function of Mpc1 has been shown to drive lactate production through enhanced conversion of pyruvate to lactate by Ldh 22.
In animals with Mpc1 deletion in HFSCs, we observed a strong acceleration of the ventral and dorsal hair cycles with all the typical features of a telogen-anagen transition (Fig 4a) (n = 12 littermate pairs). Mifepristone treated K15CrePR;Mpc1fl/fl animals were the only to show any signs of dorsal anagen by day 70. Western blotting on sorted HFSCs validated the loss of Mpc1 protein (Fig 4b). Importantly, purified HFSCs lacking Mpc1 showed a strong induction of Ldh activity (Fig 4c). Quantification of the dorsal hair cycle across three pairs of littermates showed a strong induction of anagen in backskin lacking Mpc1 (Fig 4d, right), and histology showed that the anagen induction was normal in appearance with a typical hypodermal expansion (Fig 4d). Immunostaining demonstrated the induction in Mpc1-null HFSCs of various markers of hair cycle activation such as Ki-67 and pS6, while Sox9 expression was unaffected (Fig 4e). Long term deletion of Mpc1 did not lead to aberrant follicles or exhaustion of HFSCs as judged by pathology and staining for Sox9 (Supplementary Figure 4a). Furthermore, deletion of Mpc1 with Lgr5CreER showed a very similar phenotype as deletion with K15CrePR (Fig 4f and g), validating the fact that deletion of this protein in HFSCs leads to their activation (n = 12 pairs of littermates). Finally, immunofluorescence for the Ires-GFP of the Lgr5CreER transgene along with Ki-67 and lineage tracing with K15CrePR;Mpc1fl/fl;lsl-Tomato mice also demonstrated that the HFSCs were indeed proliferative following induction of Mpc1 deletion by tamoxifen or mifepristone (Supplementary Figure 4b).
On the other hand, deletion of Mpc1 in the top of the follicle (infundibulum, sebaceous gland progenitors) and a limited number of interfollicular cells with Lgr6CreER
23 did not appear to affect the hair cycle (Lgr6CreER;Mpc1fl/fl) (n = 10 littermate pairs) or general skin homeostasis over at least 2 months (Fig S4c). Ldh activity assay on Lgr6+ cells sorted from wildtype or deletion skin demonstrated that the Mpc1 deletion was effective (Supplementary Figure 4d). Together, these results indicate that increasing lactate production through the blockade of pyruvate into the TCA cycle has a strong effect on the ability of HFSCs, but not other cells in the hair follicle, to become activated to initiate a new hair cycle.
UK-5099 is a well-established pharmacological inhibitor of the mitochondrial pyruvate carrier and is known to promote lactate production as a result in various settings24. Topical treatment of animals in telogen (day 50) with UK-5099 led to a robust acceleration of the hair cycle, as well as minor hyperproliferation of the interfollicular epidermis (Fig 5a). Quantification of the hair cycle across at least 6 pairs of animals (vehicle vs UK-5099) indicated a strong acceleration of the hair cycle, in as few as 6–9 days (Fig 5b). Similar to genetic deletion of Mpc1, pharmacological blockade of the mitochondrial pyruvate carrier by UK-5099 for 48 hours during telogen promoted increased Ldh activity in HFSCs and the interfollicular epidermis, consistent with increased capacity for lactate production (Fig 5c). Finally, metabolomic analysis demonstrated that topical application of UK-5099 increases total levels of lactate in sorted HFSCs (Fig 5d).
Because alteration of lactate production in HFSCs appeared to regulate their activation, we attempted to identify other small molecules that could take advantage of these findings to induce the hair cycle. Ldha is known to be transcriptionally regulated by Myc, which has been shown to play an important role in HFSC activation and the hair cycle25–27. RNA-seq on sorted HFSCs indicated that Myc is induced during the telogen-anagen transition (Fig 6a). Western blotting for both c-Myc and n-Myc in sorted HFSCs versus total epidermis showed a strong increase in Myc protein in the nuclei of HFSCs (Fig 6b).
Taking advantage of a molecule with the robust ability to promote Myc expression through binding of GP130 and activation of Jak/Stat signaling, we topically treated mice for 48 hours to determine the effect of RCGD423 on Stat signaling and Myc expression. We found that RCGD423 induced levels of both c-Myc and n-Myc as well as Ldha (Fig 6c), consistent with activation of Stat3 signaling leading to induction of Myc and Ldha protein expression. In vitro measurement of Ldh activity on lysates from total epidermis showed an increase in activity by RCGD423 (Fig 6d). In situ staining for Ldh activity showed a strong induction upon treatment with RCGD423 in both the epidermis and even in the dermis, as expected with topical treatment (Fig 6e). LC-MS-based metabolomics on epidermis isolated from vehicle or RCGD423 showed a large increase in lactate as well, even after just 48 hours (Fig 6f).
RCGD423 binds to GP130, a co-receptor for Jak-Stat signaling, and activates Stat-3. We found that Stat-3 was activated in HFSCs by RCGD423 after topical treatment by immunostaining with phospho-Stat3 antibody (Fig 6g). This also correlated with induction of Ki-67 in HFSCs in the same tissue (Fig 6g). IHC for pStat1 and pStat5 suggested that RCGD423 does not dramatically affect these other Stat family members (Supplementary Figure 5). Topical treatment of animals in telogen (day 50) with RCGD423 led to a robust acceleration of the hair cycle (Fig 6h), as well as minor hyperproliferation of the interfollicular epidermis.

Discussion
Together, these data demonstrate that the production of lactate, through Ldha, is important for HFSC activation, and that HFSCs may maintain a high capacity for glycolytic metabolism at least in part through the activity of Myc. Our data also demonstrate that a genetic or pharmacological disruption of lactate production can be exploited to regulate the activity of HFSCs. It is possible that these results have implications for adult stem cells in other tissues. In an accompanying manuscript, the Rutter lab describes a role for Mpc1 in adult intestinal stem cells28. Consistent with data presented here on HFSCs, deletion of Mpc1 led to an increase in the ability of intestinal stem cells to form organoids.
Previous work showed that hematopoietic stem cells (HSCs) show higher glycolytic activity, but disruption of glycolysis in the HSCs led to activation of their cycling29–32, contrary to what we find with HFSCs. While the distinction could be biological, there are technical reasons for potential discrepancies as well. First, there are no Cre transgenic lines that can delete genes specifically in HSCs, as opposed to HFSCs (K15+ or Lgr5+). Second, to block glycolysis in HSCs, the previous study deleted PDK enzyme, which would only indirectly regulate glycolysis, whereas here we deleted Ldh enzyme specifically. In addition, HSCs and HFSCs are functionally distinct in that HFSCs only cycle at well-defined moments (telogen to anagen transition), while the timing of HSC activation is not as well established or synchronized. Instead, we hypothesize that increased glycolytic rate in HFSCs allows them to respond quickly to the barrage of cues that orchestrate the onset of a new hair cycle. This has also been proposed to be the case for neural stem cells based solely on RNA-seq data33, but as of yet no in vivo functional evidence exists to confirm this possibility.
The fact that small molecules could be used to promote HFSC activation suggests that they could be useful for regenerative medicine. This is not only the case for hair growth, but potentially for wound healing as well. While HFSCs do not normally contribute to the interfollicular epidermis, in a wound setting, HFSCs migrate towards the wound site and make a contribution, as measured by lineage tracing34. Whether activation of Ldh enzyme activity by Mpc1 inhibition (UK-5099) or Myc activation (RCGD423) can promote wound healing will be the subject of intense effort going forward.

Materials and Methods

Mice
Several of the animal strains came from Jackson Labs (K15-CrePR, Lgr5-CreER and Lgr6-CreER), while others were generated in the Rutter (Mpcfl/fl) and Seth Labs36 (Ldhafl/fl) and maintained under conditions set forth by IUCUC and UCLA ARC. For experiments that include analysis of the telogen stage of the hair cycle, animals were harvested at post-natal day 50, for telogen-anagen transition animals were harvested at day 70, and for anagen animals were harvested at post-natal day 90. For experiments that include analysis of transgenic animals, K15-CrePR animals were shaved and treated by injection of mifepristone and Lgr5-CreER and Lgr6-CreER animals were shaved and treated with tamoxifen (10 mg/ml dissolved in sunflower seed oil, 2 mg per day for 3 days) during telogen (post-natal day 50), and monitored for hair regrowth following shaving. For Figs. 5 and 6, wildtype C57BL/6J animals were shaved at post-natal day 50 and treated topically with Transderma Plo Gel Ultramax Base (TR220) (vehicle), UK-5099 (Sigma PZ0160) (20uM) or RCGD423 (50uM) for indicated periods of time. Both male and female animals were used in this study in approximately equal numbers with no apparent difference in phenotype between genders. All animal experiments were done in compliance with ethical guidelines and approved by the UCLA Animal Research Committee (ARC) according to IACUC guidelines in facilities run by the UCLA Department of Laboratory Animal Medicine (DLAM).

Histology, Immunostaining and Immunoblotting
Tissues were isolated from the indicated genotypes and embedded fresh in OCT compound for frozen tissue preparations, or fixed overnight in 4% formalin and embedded in paraffin. For frozen tissue, sectioning was performed on a Leica 3200 Cryostat, and fixed for 5 minutes in 4% paraformaldehyde. Paraffin embedded tissue was sectioned, de-paraffinized, and prepared for histology. All sections prepared for staining were blocked in staining buffer containing appropriate control IgG (Goat, Rabbit etc.). Immunohistochemistry was performed on formalin-fixed paraffin-embedded tissue with citrate or Tris buffer antigen retrieval with the following antibodies: Ki67 (Abcam ab16667, 1:50), p-S6 (Cell Signaling CST2215, 1:50), Sox9 (Abcam ab185230, 1:1000), Ldha (Abcam ab47010, 1:100), Ldh (Abcam ab125683, 1:100), p-Stat3 (Abcam ab68153, 1:200), p-Stat1 (Abcam ab109461, 1:200), p-Stat5 (Abcam ab32364; 1:50), Gli3 (Abcam ab6050; 1:100), β-catenin (Abcam ab32572; 1:500). The DAKO EnVision+ HRP Peroxidase System (Dako K400911-2) and Dako AEC Substrate Chromogen (Dako K346430-2) was used for detection. Images were collected on an Olympus BX43 Upright Microscope and Zeiss Model Axio Imager M1 Upright Fluorescence Microscope. Protein samples for western blots and enzymatic assays were extracted from FACS sorted epidermal populations in RIPA lysis buffer (Pierce) with Halt protease and phosphatase inhibitors (Thermo-Fisher) and precipitated in acetone for concentration. The following antibodies were used: β-actin (Abcam ab8227; 1:1000), β-actin (Santa Cruz sc-47778; 1:1000), C-Myc (Abcam ab32072; 1:1000), N-Myc (Santa Cruz sc-53993; 1:200), H3K27Ac (Abcam ab177178; 1:200), Mpc1(Sigma HPA045119).

Cell isolation and FACS
Whole dorsal and ventral mouse skin were excised and floated on trypsin (0.25%) for 1 h at 37° or overnight at 4°. The epidermis was separated from dermis by scraping and epidermal cells were mechanically dissociated using a pipette. Epidermal cells were filtered with a 70 μM cell strainer into 20% BCS, collected at 300g and washed twice with PBS. The cells were then filtered through a 40 μM cell strainer and stained for FACS processing with CD34 Monoclonal Antibody (RAM34), FITC, eBioscience™ (Catalog #:11-0341-82) and CD49d (Integrin alpha 4) Monoclonal Antibody (R1-2), PE, eBioscience™ (Catalog #:12-0492-81). Gating strategy shown in Suppl Fig 1b. Cells sorted using BD FACSAria high-speed cell sorters. Single positive and double positive populations were collected into 20% BCS, RIPA lysis buffer (Thermo Scientific, Pierce), or 80% methanol for enzymatic assays, western blots or mass spec analyses respectively.

Cell lines
No cell lines were used in this study.

Plate-reader Ldh assay
Ldh activity was determined in cell lysates by measuring the formation of soluble XTT formazan in direct relation to production of NADH over time at 475 nm at 37°C using a Synergy-MX plate reader (Biotek Instruments). Lysates were prepared in RIPA Buffer (Thermo Scientific Pierce). Protein content was determined using the BCA Protein Assay Kit (Thermo Scientific Pierce). 10 μg of protein were used per well. The staining solution contained 50 mM Tris buffer pH 7.4, 150 μM XTT (Sigma), 750 μM NAD (Sigma), 80 μM phenazine methosulfate (Sigma) and 10mM of substrate lactate (Sigma). Ldh activity was determined in cell lysates by measuring the change in absorbance of their common substrate or product, NADH, over time at 340 nm at 25°C using a Synergy-MX plate reader (Biotek Instruments).

In situ Ldh assay
Cryostat sections of mouse skin were briefly fixed (4% formalin for 5 min), washed with PBS pH 7.4, and then incubated with the appropriate solution for LDH activity. Staining medium contained 50 mM Tris pH 7.4, 750 μM NAD (Sigma), 80 μM phenazine methosulfate (Sigma), 600 μM Nitrotetrazolium Blue chloride (Sigma), 10 mM MgCl2 (Sigma) and 10mM of the substrate lactate (Sigma). Slides were incubated with staining medium at 37°C until they reached the desired intensity, then counterstained using Nuclear Fast Red (Vector, Burlingame, CA) and mounted using VectaMount (Vector, Burlingame, CA). Control reactions were performed by using incubation medium that lacked the substrate mixture or NAD.

Mass spectrometry-based metabolomics analysis
The experiments were performed as described in 36. To extract intracellular metabolites, FACS sorted cells were briefly rinsed with cold 150 mM ammonium acetate (pH 7.3), followed by addition of 1 ml cold 80% MeOH on dry ice. Cell suspensions were transferred into Eppendorf tubes and 10 nmol D/L-norvaline was added. After rigorously mixing, the suspension was pelleted by centrifugation (1.3*104 rpm, 4 °C). The supernatant was transferred into a glass vial, metabolites dried down under vacuum, and resuspended in 70% acetonitrile. For the mass spectrometry-based analysis of the sample, 5 ul was injected onto a Luna NH2 (150 mm × 2 mm, Phenomenex) column. The samples were analyzed with an UltiMate 3000RSLC (Thermo Scientific) coupled to a Q Exactive mass spectrometer (Thermo Scientific). The Q Exactive was run with polarity switching (+3.50 kV / −3.50 kV) in full scan mode with an m/z range of 65–975. Separation was achieved using A) 5 mM NH4AcO (pH 9.9) and B) ACN. The gradient started with 15% A) going to 90% A) over 18 min, followed by an isocratic step for 9 min and reversal to the initial 15% A) for 7 min. Metabolites were quantified with TraceFinder 3.3 using accurate mass measurements (≤ 3 ppm) and retention times. Normalized metabolite data is available at figshare.com (https://figshare.com/s/e2d8445b3ec37b5f2c33).

Statistics and Reproducibility
Experiments were performed on male and female animals in approximately equal numbers with no apparent difference in phenotype between sexes. All phenotypes described are representative of a minimum of n = 3 littermate pairs as indicated in the description of each experiment. For analysis of the hair regrowth phenotype no statistical measure was used to determine the sample size beforehand, nor were statistics used to measure effects, as the results were essentially positive or negative as represented in the figures. The results described include data from all treated animals. Investigators were not blinded to allocation during the experimental data collection. Experiments were not randomized. All results shown were representative images from at least three independently treated animals, and genotyping was performed both before and after animal treatment for confirmation. For graphs, all comparisons are shown by Student’s two-tailed unpaired t-test and all graphs, bars or lines indicate mean and error bars indicate Standard error of the mean (s.e.m).

Data Availability
Previously published transcriptomics data that were re-analysed here are available under accession code GSE67404 and 5163537,38.

Supplementary Material
1234567