Mobilizing Transit-Amplifying Cell-Derived Ectopic Progenitors Prevents Hair Loss from Chemotherapy or Radiation Therapy

Chemotherapy and radiotherapy are widely employed in the treatment of various cancers (1, 2). Both ionizing radiation (IR) and a substantial proportion of chemotherapeutic agents exert their treatment effects against cancer cells by inducing DNA damage (1–3). Despite extensive efforts made to minimize off-target injuries, damage to normal tissues is still inevitable (2, 4–6). Hair loss is a common side effect (7–9). It brings psychosocial stress, compromises patients' sense of personal identity, and even jeopardizes the willingness for treatment (8, 10). Prevention of such hair loss is still an unmet clinical need (7, 8).

Hair follicles (HF) are a dynamic organ that undergo lifelong growth cycles, consisting of anagen (active growth), catagen (regression), and telogen (relative rest) phases (Fig. 1A; ref. 9). Both telogen and anagen HFs share an upper permanent segment, spanning from the follicular infundibulum to the bulge (Fig. 1A; refs. 9, 11–13). The structures below the bulge are not permanent (Fig. 1A; refs. 9, 11–13). In telogen, the lower segment shrinks to a minimum structure of secondary hair germ (SHG; refs. 9, 11, 12). In anagen, the lower segment expands dramatically into a long cylinder where distinct populations of transit-amplifying cells (TAC) reside. Among them, outer root sheath (ORS) cells, located immediately below the bulge, are connected with an enlarged hair bulb where hair matrix germinative cells surrounding the dermal papilla (DP) actively multiply to generate concentric cellular layers of distinct differentiations to support hair elongation (9, 13–15). As at any given time, the majority of human scalp HFs are in anagen (9), this highly proliferative nature makes anagen HFs one of the most sensitive organs to genotoxic injury (7, 16, 17).

The slow-cycling long-lived HF stem cells (HFSC) are constantly preserved in the bulge around hair cycles (11, 12, 18). In addition to these quiescent bulge SCs (BgSC), telogen HFs house another population of more active SCs in SHG [SHG stem cells (ShgSC); refs, 11, 12, 19, 20]. During transition from telogen to anagen, ShgSCs proliferate first to support the formation of the initial hair bulbs (11, 12, 19). This is followed by the activation of quiescent BgSCs a few days later, which contribute to the upper ORS (11, 12, 19). In contrast, the highly expanded lower segment of anagen HFs lacks long-lived SCs (9, 18).

With the presence of TACs only in the lower segment, how are anagen HFs repaired to resume the ongoing anagen hair growth following genotoxic injuries? As quiescent BgSCs are relatively resistant to genotoxic insults (21), such repair might be driven by BgSCs. However, it would require complete HF involution and lengthy resetting of the hair cycle. This possibility is contrasted by the fact that DNA-damaged anagen HFs often bypass telogen involution and resume hair production without cycle resetting (8, 17). Such observations suggest that, in contrast to the telogen-to-anagen regeneration that relies on the activation of BgSCs and ShgSCs, anagen HFs can mobilize other progenitor cells for repair.

In this work, we attempt to explore the mechanisms and map the progenitor sources underlying the regenerative responses of anagen HF repair following ionizing radiation (IR) injury. We provide evidence that HFs are able to employ ectopic progenitor cells from distinct TAC compartments for repair in adaptation to the severity of genotoxic damage. We also demonstrate that efficient mobilization of TAC-derived ectopic progenitor cells for anagen HF repair can be a strategy to prevent hair loss from chemotherapy and radiotherapy.

All animal experiments were approved by the Institutional Animal Care and Use Committee of National Taiwan University. K5^CreER mice were provided by C.M. Chen (22), Lgr5^{EGFP-Ires-CreERT2} mice were from H. Clevers (23), and K19^CreER mice were from G. Gu (24). p53-null mice, Ctnnb1^flox/flox mice (25) and R26LSL^tdTomato mice were from The Jackson Laboratory. C57BL/6 mice were from Taiwan National Laboratory Animal Center (Taipei, Taiwan). For IR and invasive experiments, animals were anesthetized by tiletamine-zolazepam (Telazol).

The dorsal hair of female mice at postnatal day 30 was carefully shaved by an electric shaver. Around 2 days later when dorsal HFs were in early full anagen (∼postnatal day 32), single doses (2 or 5.5 Gy) of γ irradiation were given from the dorsal side by a ¹³⁷Cs source (dose rate 3.37 Gy/minute, γ irradiator IBL 637 from CIS Bio International). For comparison, littermate control with the same genetic background was used. Mice were consistently irradiated in the afternoon.

To label basal cells and BgSCs, K5^CreER/+; R26LSL^tdTomato/+ and K19^CreER/+; R26LSL^tdTomato/+ mice received a single intraperitoneal injection of tamoxifen (Sigma; 0.1 mg/g of body weight) 24 hours prior to irradiation. To label cells in the lower segment of epithelial strand at 5.5 Gy of IR, Lgr5^{EGFP-Ires-CreERT2/+}; R26LSL^tdTomato/+ mice received a single dose of tamoxifen (0.05 mg/g of body weight) at 48 hours after radiation.

Inhibition of Wnt/β-catenin signaling in the epithelium was achieved by the conditional deletion of Ctnnb1 in K5^CreER/CreER; Ctnnb1^flox/flox mice by tamoxifen (0.1 mg/g body weight) at 6 and 12 hours after irradiation or by using specific inhibitors IWR1 and IWP2 (Sigma). Both IWR1 and IWP2 were reconstituted in DMSO, and DMSO was used as a control. Mice were subcutaneously injected with IWR1 and IWP2 at 48, 72, and 96 hours postirradiation, totaling 12.5 μg/g body weight for each dose. Skin was harvested at day 5 for examination.

Skin specimens were fixed at 4°C overnight either in formalin–acetic–alcohol solution for paraffin embedment or 4% paraformaldehyde for OCT (Sakura Finetek) embedment. Specimens were sectioned and stained with hematoxylin and eosin (H&E). Apoptotic cells were detected by DeadEnd Fluorometric TUNEL System (Promega). Cryosections were used to visualize tdTomato fluorescence of lineage tracing. IHC and immunofluorescence staining were performed with routine antigen retrieval as suggested by the antibody manufacturers. Super Sensitive IHC Detection Systems (BioGenex) were used for the detection of horseradish peroxidase activity. The antibodies used were described in Supplementary Table S1.

All fluorescent images were acquired on the confocal microscope (SP5, Leica). To quantify bromodeoxyuridine-positive (BrdUrd⁺) and TUNEL⁺ cells, we acquired 1,024 × 1,024 pixels sequential scans with a 63× oil immersion objective lens (1.4 NA). Hair matrix was counted as epithelial cells below the top of DP. Germinative cells are epithelial cells adjacent to DP, and basal hair bulb cells are basal cells located on the outer surface of the hair bulb abutting dermal sheaths.

DNA double-strand breaks were determined by γ-H2AX foci as described previously (26). Briefly, the 3-dimensional fluorescent images were reconstituted by approximately 10 to 15 images of serial two-dimensional Z-stacks of confocal images. Data were analyzed using the software MetaMorph 7.7. Foci within Hoechst-stained nucleus were scored by the value of pixels after the focus threshold was set manually. Five pixels were considered as a standard for single DNA double-strand breaks based on the few discrete γ-H2AX foci generated in the nucleus of unirradiated control specimen. Because extensive γ-H2AX foci appeared after irradiation, individual foci could not be distinguished accurately. We compared the total value of positive pixels within the nucleus in selected cells with the standard 5 pixels/focus to estimate the number of foci in an entire nuclear region.

Keratinocytes of irradiated skin were collected at different time points after irradiation. Total RNA from keratinocytes was extracted using TRI reagent solution (Thermo Fisher Scientific) for the following RNA-sequencing (RNA-seq) library preparation and sequencing. All procedures were carried out according to the protocol from Illumina. The libraries were sequenced on the Illumina NextSeq 500 platform using 75 single-end base pair strategy, and 10 million reads per sample were generated. Gene Ontology (GO) and KEGG analyses were performed using DAVID (https://david.ncifcrf.gov/; ref. 27).

Total RNA from keratinocytes of irradiated skin collected at different time points was extracted using TRI reagent solution (Thermo Fisher Scientific). The RNA was reverse transcribed into cDNA using RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). In bead implantation experiment, HF epithelial cells were isolated from the HFs surrounding Wnt3a and bovine serum albumin (BSA)-soaked beads, and the RNA was extracted and cDNA was amplified by using REPLI-g WTA Single Cell Kit (Qiagen). Quantitative PCR analysis was performed using SYBR Green qPCR Master Mix (Thermo Fisher Scientific) on an ABI 7900HT Real-Time PCR System (Applied Biosystems). Sequences of gene-specific primers used were described in Supplementary Table S2.

The dorsal skin of 7-week-old female mice were used for the sorting of inactive ShgSCs. For the sorting of activated ShgSCs, dorsal hairs of 7-week-old female mice were plucked to activate ShgSCs 2 days before skin specimen collection (28). The dorsal skin of 32-day-old mice was irradiated with 5.5 Gy of IR, and skin specimen was collected at 72 hours. Cell preparation for FACS was performed as described previously (29). The following antibodies were used: CD34-FITC (eBioscience, 11-0341, 1:50), Sca1-PE-Cy7 (eBioscience, 25-5981, 1:50) and P-cadherin-PE (R&D Systems, FAB761P, 1:100). FACSAria III sorter equipped with Diva software (BD Biosciences) was used for sorting. Total RNA from sorted cells was extracted using MessageBOOSTER cDNA Synthesis from Cell Lysates Kit (Epicentre).

Intracutaneous implantation of protein-coated beads was performed as described previously (30). Human Wnt3a recombinant protein (R&D Systems) was resuspended in 1 mg/mL BSA solution at 1 mg/mL. Affinity affi-gel blue gel beads (Bio-Rad) were then suspended in 5 μL protein solution, either control (1 mg/mL BSA) or experimental (1 mg/mL Wnt3a), at 4°C for 2 hours. Approximately 100 beads were implanted into skin immediately after irradiation. Mice were sacrificed at different time points postirradiation. To visualize HFs, skin specimens were dehydrated in ethanol with graded concentrations and then immersed in xylene until it became transparent. The skin specimens were then further processed for histologic examination and immunofluorescence staining.

Statistical comparison was performed using the software Prism (GraphPad). An unpaired Student t test was used to compare datasets with two groups. To compare three or more groups, we performed one-way ANOVA followed by Bonferroni multiple comparison. Data were presented as mean ± SE. P values were considered statistically significant when less than 0.05.

To characterize the effect of IR, mice were irradiated with 2 and 5.5 Gy on postnatal day 32 when dorsal HFs were in early full anagen (Fig. 1A). In 5 to 10 days, there was a dose-dependent hair loss and HF dystrophy (Fig. 1B and C; Supplementary Fig. S1A). At 2 Gy, the initial reduction of matrix cells and HF shortening were recovered by days 3 and 4 (Fig. 1C–E). At 5.5 Gy, HFs progressively shrank to slender epithelial strands by day 3, followed by restoration of anagen morphology and length between days 5 and 7 (Fig. 1C and D). Around day 10, HFs of the unirradiated, 2 Gy, and 5.5 Gy groups entered catagen, indicated by decreased cell proliferation and increased apoptosis (Fig. 1C and F–I). Afterwards, HFs could resume a new anagen in 2 and 5.5 Gy groups (Supplementary Fig. S1A), indicating the hair loss was transient.

We then analyzed cell death and proliferation. TUNEL staining and cleaved caspase-3 staining were performed to detect apoptosis, and both assays exhibited a similar trend (Fig. 1F, G and J). Apoptosis was more extensive and persistent after 5.5 than 2 Gy (Fig. 1F, G and J). After 2 Gy exposure, cell proliferation almost entirely halted after 6 hours and then increased at 12 and 48 hours (Fig. 1H and I). This early proliferative response successfully restored hair bulb structures and HF length between 72 and 96 hours (Fig. 1C and D). After 5.5 Gy exposure, a similar regenerative response, albeit with lower proliferation, was observed between 12 and 48 hours (Fig. 1H and I). However, proliferation almost entirely ceased again at 72 hours. Progressive HF shrinkage between 0 and 72 hours indicated failure of the early regenerative attempt to compensate for the more severe cell death (Fig. 1C, F, G and J). Importantly, a second late proliferative attempt was observed in the lower tip of HFs at 96 hours (Fig. 1H and I) and led to HF elongation and restoration of hair bulbs (Fig. 1C–E).

Next, we examined whether these regenerative attempts led to production of mature hair shafts. At 2 Gy, differentiation toward inner root sheath and hair cortex was only transiently disrupted between 36 and 48 hours (Supplementary Fig. S1B). At 5.5 Gy, a lengthier and more severe disruption was induced, yet differentiation was eventually restored (Supplementary Fig. S1C).

These results show that, depending on the severity of IR damage, HFs activate two distinct regenerative responses: early regenerative attempt between 12 and 48 hours and late regenerative attempt after 72 hours.

Next, we tried to identify cells contributing to early regenerative attempts. In normal hair bulbs, keratin 5 (K5) is exclusively expressed in basal cells (Fig. 2A, 0 hour). However, 12 to 36 hours after 2 Gy exposure, K5⁺ cells extended into the suprabasal positions (Fig. 2A and B). TUNEL showed that, after 2 Gy, apoptosis was more prominent in suprabasal K5⁻ matrix and germinative cells than K5⁺ cells (Fig. 2A, C, and D; Supplementary Fig. S2A) and that apoptosis of K5⁺ cells only slightly increased (Fig. 2C). BrdUrd pulse labeling showed these K5⁺ cells were proliferative (Fig. 2E, yellow arrowheads, and F). During the same period, proliferation of K5⁻ suprabasal matrix cells nearly ceased at 6 hours and was progressively restored only by 48 hours (Fig. 2E and G). Analysis of double-stranded DNA breaks by γ-H2AX expression also showed faster repair of DNA damage in K5⁺ basal hair bulb cells (Supplementary Fig. S2B and S2B′). Together, basal K5⁺ cells are more resistant to IR than suprabasal K5⁻ matrix and germinative cells and become proliferative to support early regenerative attempts.

A total of 5.5 Gy exposure induced a similar increase in suprabasal K5⁺ cells (Fig. 2A and B), yet their proliferation was delayed compared with the 2 Gy group (Fig. 2E and F). Notably, higher apoptosis in both K5⁺ and K5⁻ cells was detected after 5.5 Gy during the early regenerative attempt (Fig. 2A, white arrowheads, C and D). K5⁺ cells also showed more persistent expression of DNA damage markers (Fig. 2H; Supplementary Fig. S2C and S2C′). These data suggest that K5⁺ cells and their progeny sustain catastrophic IR injury. Without sufficient resupply from the K5⁺ compartment, K5⁻ matrix cells, which also proliferate poorly (Fig. 2E and G), undergo eventual collapse, leading to failure of the early regenerative attempt.

To further confirm the cellular source for early regenerative attempts (Fig. 3A), we lineage traced K5⁺ cells in K5^CreER; R26LSL^tdTomato mice (Fig. 3B). Injection with tamoxifen 24 hours prior to IR allowed successful and exclusive labeling of basal, but not germinative cells in the hair bulb (Fig. 3B, 0 hour). After 2 Gy exposure, labeled cells expanded into suprabasal positions at 12 hours (Fig. 3B, yellow arrowhead), replenished the germinative compartment (Fig. 3B, white arrowheads) and contributed to all layers of repaired hair bulbs, including hair shafts (Fig. 3B, white arrow), between 24 and 36 hours. In control HFs, most of the labeled cells remained in the basal position even after 36 hours (Fig. 3B). Quantitatively, progeny of K5⁺ lineage progressively increased, accounting for about 60% of all matrix cells at 36 hours (Fig. 3C). The data underscore that, following IR, K5⁺ basal hair bulb cells display high lineage plasticity.

Next, we determined the contribution of BgSCs. Neither apoptosis (Fig. 3D) nor proliferation (Fig. 3E) was detected in the bulge. Lineage tracing in K19^CreER; R26LSL^tdTomato mice did not show BgSC expansion (Fig. 3F). The results indicate that BgSCs survive 2 Gy IR injury but do not actively contribute to early regenerative attempts.

The results above show that the remaining short epithelial strand at 72 hours serves as a platform for the late regenerative attempt. To determine the origin of the epithelial strand, we performed lineage tracing in K5^CreER; R26LSL^tdTomato mice (Fig. 4A) following tamoxifen administration 24 hours prior to IR. This protocol labeled basal cells in the ORS in addition to basal cells of the bulb (Fig. 4A). As HFs regressed between 24 and 72 hours, a larger portion of the epithelial strand became composed of the K5⁺ progeny (Fig. 4A). As K5⁺ basal hair bulb cells fail to survive 5.5 Gy injury, we conclude that the epithelial strand originates from the K5⁺ ORS cells that are more radioresistant and that basal ORS progeny fuel successful HF regeneration during late regenerative attempts (Fig. 4A).

After 5.5 Gy, cell proliferation largely halted at 72 hours (Figs. 1H and I and 4B). By 84 hours, proliferation was first resumed in P-cadherin⁺ lower tip cells (Fig. 4B, white arrowheads). At 96 hours, more lower tip cells proliferated as they continued to regenerate new hair bulbs (Fig. 4B). We did not detect apoptosis of BgSCs (Fig. 4C). Continuous BrdUrd labeling showed that BgSCs did not start proliferating until day 5 post-IR (Fig. 4D, white arrowheads), when the new hair bulb already regenerated (Fig. 1C). Lineage tracing of BgSCs in K19^CreER; R26LSL^tdTomato mice showed that BgSCs contributed to the suprabulbar ORS cells right below the bulge, but not to the regenerated hair bulbs (Fig. 4E). These results indicate that the lower tip cells fuel the initial late regenerative attempt. This parallels the stepwise activation of epithelial progenitor populations during physiologic telogen-to-anagen transition, when ShgSCs, rather than quiescent BgSCs, fuel early HF growth (12, 19).

We considered that cells in the lower segment of the epithelial strand acquire SC-like characteristics. We compared the epithelial strand structure of 5.5 Gy–irradiated anagen HFs with normal telogen HFs. Typical BgSC markers, CD34 and K15 (31, 32), were still maintained in the bulge (Supplementary Fig. S3A). However, cells at the lower end of the epithelial strand were CD34/K15 double-negative, yet positive for P-cadherin, Sox9, and Lgr5 promoter activity, markers of the normal telogen ShgSCs (Fig. 4F and G; Supplementary Fig. S3A; refs. 12, 20, 33). Lack of AE13, AE15, and K75 keratin expression shows that lower epithelial strand cells do not prematurely differentiate toward inner anagen HF structures (Supplementary Fig. S3A). We then FACS-isolated these lower tip cells using a Sca1^low CD34^low P-cadherin^high marker profile (Supplementary Fig. S3B; ref. 12) and examined the expression of known ShgSC genes (12). We found Ccnb1, Clca1, Sox4, and Sox6 were also upregulated in the sorted lower tip cells at 72 hours (Fig. 4H). Taken together, these results show that following 5.5 Gy IR injury, ORS cells were remodeled into the epithelial strand whose lower tip cells acquired a ShgSC-like progenitor property.

To strengthen the evidence for the cellular dynamics noted above (Fig. 4I), we performed additional lineage tracing using Lgr5^{EGFP-Ires-CreERT2} mice, which allowed for specific tracing of the lower tip cells (Fig. 4J; Supplementary Fig. S4). Consistent with the prior report (34), in unirradiated Lgr5^{EGFP-Ires-CreERT2} mice, ORS cells and suprabasal hair bulb cells were labeled after tamoxifen induction (Fig. 4J, left). In the 5.5 Gy group, although the epithelial strand was negative for Lgr5 by immunostaining (Supplementary Fig. S3A), its lower portion was distinctly positive for the Lgr5 promoter activity (Fig. 4G). Twenty-four hours after tamoxifen injection to Lgr5^{EGFP-Ires-CreERT2}; R26LSL^tdTomato mice at 48 hours after 5.5 Gy, labeled cells were exclusively found in the lower portion of the epithelial strands of all HFs examined (n = 50; Fig. 4J, right), with higher incidence in the lower tip. These labeled Lgr5 progeny contributed to the ORS, the entire hair bulb, and hair shafts of the repaired HFs (Fig. 4J, right; Supplementary Fig. S4).

Further tracing showed that the progeny of labeled Lgr5 cells homed back to the SHG and bulge when repaired HFs eventually transitioned to telogen (d17, Fig. 4J, right), and that in the next cycle, they formed the new lower segment of the anagen HFs (d35, Fig. 4J, right). These results reveal expanded plasticity of ORS cells following IR and underscore IR-induced acquisition of SC properties.

Because HF regeneration from ORS-derived progenitors represents a novel repair mechanism, we aimed to clarify its molecular basis. First, we compared epithelial cell transcriptomes at different time points before and after 5.5 Gy of IR by RNA-seq. GO enrichment analysis revealed several biological processes altered following IR (Fig. 5A). We screened for signaling pathways downregulated between 0 and 24 hours and upregulated between 24 and 72 hours. We found that WNT and hedgehog pathways fit this trend (Supplementary Table S3). Hedgehog signaling has been shown to be downregulated by cyclophosphamide in growing HFs (35).

WNT signaling is activated in ShgSCs at the onset of physiologic anagen and is crucial for normal hair cycle progression (36–38). The transcript and protein levels of Lef1, a downstream mediator of WNT signaling in HFs, were prominently suppressed after IR and restored only by 96 hours post-IR (Fig. 5B and C). This was accompanied by an increase in nuclear β-catenin in the lower tip cells 72 to 96 hours post-IR (Fig. 5D). WNT ligands are secreted by HF epithelium and help to maintain proper epithelial–mesenchymal interaction during anagen (36, 39–41). We found that Wnt3a levels were diminished between 24 and 48 hours post-IR, but later rebounded from 72 hours onward at the base of the epithelial strand (Fig. 5E and F). These results show that 5.5 Gy disrupts WNT signaling and that its reactivation correlates with late regenerative attempts. In comparison, after 2 Gy exposure, levels of Wnt3a and Lef1 were suppressed only transiently and partially (Supplementary Fig. S5A–S5D). Therefore, higher IR doses induce more severe suppression of WNT signaling.

To determine whether WNT signaling reactivation is required for late regenerative attempts, we pharmacologically disrupted WNT ligand secretion with IWR1 and IWP2 inhibitors and found that the late regenerative attempt was attenuated (Fig. 5G). We also disrupted β-catenin in K5^CreER; Ctnnb1^flox/flox mice after 5.5 Gy of IR and found that the late regenerative attempt was abolished (Fig. 5H). Therefore, reactivation of WNT signaling is essential for the late regenerative attempt. This parallels the requirement for WNT signaling during normal HFSC activation upon telogen-to-anagen transition (36, 38).

p53 has been shown to mediate pathologic changes of chemotherapy-induced hair loss and IR-induced injury to other organs (42, 43). Its role in IR-induced HF dystrophy remains unclear. We found that p53 was induced by IR in hair bulbs, but not DP, and that p53 expression was more extensive and persistent at 5.5 Gy than 2 Gy (Fig. 6A). Furthermore, matrix apoptosis, suppression of proliferation, HF shrinkage, and hair loss were not induced in p53-null mice by IR (Fig. 6B–E; Supplementary Fig. S6A–S6C). In wild-type mice, IR increased expression of downstream target genes of p53, Noxa, Bax, and p21, whereas Puma expression was slightly decreased from 6 to 48 hours and transiently elevated at 72 hours (Fig. 6F). In p53-null mice, the baseline expression of Bax and p21 was higher than wild-type mice, while Noxa and Puma were lower than wild-type mice (Fig. 6F). After IR, compared with wild-type mice, decreased expression of Bax, Noxa, and Puma was observed in p53-null mice, and p21 was only slightly increased at 24 hours. This might help to explain the decreased apoptosis and unsuppressed cell proliferation after IR in p53-null mice.

Comparison of 2- and 5.5 Gy–induced injuries revealed a correlation between the duration of p53 activation and more severe and persistent WNT signaling suppression, suggesting that p53 might be involved in mediating IR-induced WNT suppression. We found that the basal mRNA expression of Wnt3a and Lef1 at 0 hour was significantly higher in p53-null mice than in p53 wild-type mice (Fig. 6G). The higher WNT signaling activity in p53-null mice is consistent with the prior report showing the inhibition of WNT signaling by p53 (44). In p53 wild-type mice, both Wnt3a and Lef1 expressions were prominently suppressed by 5.5 Gy of IR and were restored again at 72 and 96 hours, respectively. In p53-null mice, although the Wnt3a expression was decreased after IR, it was still higher than that in p53 wild-type mice at 6, 24, and 96 hours. Lef1 expression was not downregulated until 24 hours, but its levels in p53-null mice were much higher than that of p53 wild-type at all time points. Consistent with this, we found that protein expression of Wnt3a and Lef1 was not suppressed by IR in p53-null mice (Fig. 6H and I). Judging from Lef1 expression, WNT signaling was maintained at a higher level in p53-null mice, and this might contribute in part to the attenuated HF dystrophy after IR.

As WNT ligands produced by anagen HF epithelium are known to maintain anagen progression (36, 38, 39) and as WNT signaling undergoes early suppression after IR, we posited that maintaining WNT signaling might promote HF regeneration, in part by bypassing the suppressive effect of p53 on cell proliferation. To test this, we delivered Wnt3a-soaked beads into the 5.5 Gy irradiated skin (Fig. 7A). We found that hairs continued to emerge in the Wnt3a-treated skin (Fig. 7A; Supplementary Fig. S6D) with reduced HF dystrophy and faster restoration of anagen structures (Fig. 7B; Supplementary Fig. S6E and S6F). Wnt3a treatment preserved Lef1 expression (Supplementary Fig. S7A), but did not prevent p53 activation and apoptosis (Fig. 7C and D; Supplementary Fig. S7B). The latter was only slightly decreased at 24 hours post-IR (Fig. 7C and D). Analysis of gene expression of HF epithelial cells from Wnt3a-treated and control (BSA)-treated skin showed that Wnt3a treatment significantly inhibited the upregulation of Bax, Noxa, and p21, but increased the expression of Puma after IR (Fig. 7E). The upregulation of proapoptotic Puma with downregulation of proapoptotic Bax and Noxa might help to explain that Wnt3a did not largely suppress IR-induced apoptosis (Fig. 7D). The suppression of p21 expression, an important p53 target that induces cell-cycle arrest, by Wnt3a treatment (Fig. 7E) might promote regenerative proliferation. Indeed, compared with control, an early increase of basal K5⁺ cell proliferation was observed at 24 hours, and cell proliferation within the hair matrix was maintained at a higher level for at least 72 hours (Fig. 7F, white arrows, and G; Supplementary Fig. S7C). Importantly, the number of matrix cells was increased in the Wnt3a-treated HF (Supplementary Fig. S7D).

We also tested whether this approach can prevent chemotherapy-induced alopecia. Cyclophosphamide administered on postnatal day 32 also induced extensive hair loss on day 5 (Fig. 7H). Similarly, local delivery of Wnt3a-soaked beads reduced hair loss after cyclophosphamide treatment (Fig. 7I). These results demonstrate that maintaining WNT signaling can prevent hair loss from genotoxic injury by enhancing the regenerative cell proliferation of K5⁺ basal hair bulb progenitors.

Our results show that, rather than deploying long-lived bulge SCs, IR-damaged anagen HFs mobilize extra-bulge TAC-derived progenitor cells for repair without having to reset their hair cycle (Fig. 7J). This ability of HFs to mobilize distinct extra-bulge progenitors for anagen HF repair also suggests a new therapeutic strategy for hair loss that relies on TACs.

In hair bulbs, germinative and basal bulb cells are thought to represent two distinct lineages within the TAC population (14). Proliferative germinative cells fuel hair shaft elongation (14, 45), while basal hair bulb cells (also referred to as lower proximal cup cells) are speculated to maintain the shape of hair bulbs and not to contribute to the inner HF structures (14). We show that K5⁺ basal cells serve as SCs for rapid repair of the HF bulb and that their concealed plasticity is quickly unveiled upon injury. Previously, colony-forming epithelial SCs were shown to be present in the hair bulbs (46). It is likely that K5⁺ basal hair bulb cells can be a source of colony-forming cells. The employment of local multipotent progenitors within the TACs shortens the time needed for regeneration, and this strategy is advantageous over regeneration via BgSCs. The employment of BgSCs would require intricate signaling relays and create a significant time delay during their downward migration toward injured hair bulbs.

We also revealed a novel role for ORS cells of the TAC pool for regeneration. HFs suffering from a more dystrophic change regenerate from the surviving ORS cells. Although our lineage studies are not able to directly differentiate between K5⁺ basal bulb cells and K5⁺ basal ORS cells, the two cell populations are known to be distinct in origin and function during anagen (14). K5⁺ hair bulb cells broadly apoptose after 5.5 Gy, indicating that radioresistant K5⁺ ORS cells are the most likely origin for the epithelial strand. Although targeted depletion of SCs can allow differentiated cells to gain SC-like properties upon migration into the physiologic SC niche (47, 48), it remains unknown whether acquisition of SC-like properties can occur ectopically outside the physiologic SC niche. Here, we showed that ORS cells can directly acquire a SC-like state ectopically after IR. As DP cells are able to convert interfollicular keratinocytes into follicular cells for HF neogenesis upon close contact (49), the close proximity of lower tip cells to DP suggested that DP cells might play an important role here by either directly reprogramming ORS cells into a SC-like state or providing signals for the dedifferentiation of ORS cells through short-range interaction.

The lower tip cells in the epithelial strand not only express markers characteristic of ShgSCs, but also behave like them. Similar to the telogen-to-anagen transition (12), the late regenerative attempt displays the step-wise activation of progenitor cells: lower tip cells are activated first to regenerate hair bulbs, followed by the activation of BgSCs to resupply upper ORS. Their contribution to the entire hair bulb and all bulb-derived differentiated structures confirms their multipotency. Furthermore, their progeny home back to the SC niche at the end of the repaired hair cycle and contribute to the new hair bulb and ORS during the following physiologic hair growth cycle.

The activation of p53, a central mediator of pathologic changes following IR injuries in other organs (43, 50), is required for chemotherapy-induced HF dystrophy (42). We revealed its key role in the response of HFs to IR and also uncovered a complicated cross-talk between IR, p53, and WNT signaling. We found that IR induces p53 expression in the hair bulb in a dose-dependent manner. In p53-null mice, neither apoptosis nor dystrophy was induced, indicating the essential role of p53 in IR-induced hair loss. As WNT signaling is not inhibited by IR in p53-null mice, our results suggest that p53 may have dual roles in both inducing apoptosis and suppressing regeneration-promoting WNT signaling following IR injury. We demonstrated that augmenting WNT signaling could attenuate the suppressive effect of p53 on cell proliferation, likely through inhibition of p53-responsive upregulation of p21, to enhance the regenerative program of ectopic progenitors from TACs following the injury of IR and cyclophosphamide. Because the effect of applied Wnt3a protein is transient and localized, it potentially represents a new, low-risk clinical strategy to reduce hair loss after genotoxic injury.

No potential conflicts of interest were disclosed.

Conception and design: W.-Y. Huang, S.-F. Lai, H.-Y. Chiu, M. Chang, S.-J. Lin

Development of methodology: W.-Y. Huang, S.-J. Lin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.-Y. Huang, C.-C. Chan, T.-L. Yang, S.-J. Lin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W.-Y. Huang, H.-Y. Chiu, P.-N. Tsao, T.-L. Yang, S.-J. Lin

Writing, review, and/or revision of the manuscript: W.-Y. Huang, H.-Y. Chiu, M.V. Plikus, P.-N. Tsao, P. Chi, S.-J. Lin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.-F. Lai, H.-Y. Chiu, M. Chang, C.-C. Chan, Y.-T. Chen, S.-J. Lin

Study supervision: H.-Y. Chiu, S.-J. Lin

Other (supporting specific transgenic mice): H.-S. Lee

We thank the staff of the imaging core and the Flow Cytometric Analyzing and Sorting Core at the First Core Lab, National Taiwan University College of Medicine, and the staff of the 8th Core Lab, Department of Medical Research, National Taiwan University Hospital for technical support. The authors also thank the members of the S.J. Lin laboratory for their discussion and Drs. Hironobu Fujiwara, George Cotsarelis, and Ralf Paus for their discussion.

This work was supported by Taiwan Bio-Development Foundation (TBF; to S.J. Lin), a Taiwan National Health Research Institutes grant (EX104-10410EI to S.J. Lin), Taiwan Ministry of Science and Technology grants (105-2627-M-002-010 to S.J. Lin; 105-2314-B-002-073-MY4 to P. Chi; and 106-2314-B-002-133-MY3 to S.F. Lai), National Taiwan University Hospital grants (105-S3010, 104-P04 to S.J. Lin), NIH NIAMS grants (R01-AR067273 and R01-AR069653 to M.V. Plikus) and Pew Charitable Trust (to M.V. Plikus).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.