Keratin 79 identifies a novel population of migratory epithelial cells that initiates hair canal morphogenesis and regeneration

The formation of tubular structures is a morphogenetic process crucial for the development of most organs. In many polarized epithelia, tube formation is initiated by invagination of an epithelial sheet to generate a bud (Andrew and Ewald, 2010; Lubarsky and Krasnow, 2003). Subsequently, this bud can either detach from the overlying epithelium, as occurs during primary neurulation, or remain attached, as occurs during the development of lungs, kidney and pancreas. In either case, a lumen is generated as a by-product of the infolding process.

The hair follicle epithelium forms a tube-like structure that is continuous with the epidermis and houses the hair shaft (Blanpain and Fuchs, 2009; Watt and Jensen, 2009). Unlike the organs described above, developing follicles retain suprabasal cells within their cores (Andl et al., 2002; Chiang et al., 1999; Oro and Higgins, 2003; St-Jacques et al., 1998). Consequently, early hair buds do not possess lumens and require complex, and poorly characterized, cellular rearrangements to generate a hollow interior.

Lumen formation is thought to be mediated by inner root sheath (IRS) cells, which extend toward the distal end of the follicle, reaching the epidermal surface concomitant with the emergence of the hair canal (Fig. 1A) (Paus et al., 1999). During the regenerative growth phase of the hair cycle (anagen), follicular morphogenesis is recapitulated, and lumen formation is also thought to be accomplished through the movement of IRS cells that ensheath the new hair shaft (Fig. 1A) (Müller-Röver et al., 2001). This hair shaft eventually erupts from the same orifice as the club hair, necessitating that the lumens of the new and old follicle fuse into a single canal. How this fusion occurs remains unclear.

In mature resting phase (telogen) hair follicles, the IRS is thought to terminate at the level of the sebaceous glands, which secrete lytic factors that cause desquamation of IRS cells (Tobin et al., 2002). Distal to the sebaceous glands is the infundibulum (INF), which encompasses the mouth of the hair follicle. Few studies have examined the INF directly; however, disruption of the INF is a hallmark of several human skin pathologies, including acne vulgaris (Bellew et al., 2011; Zaenglein et al., 2012).

Here we provide a detailed characterization of the INF and report the expression of a poorly characterized keratin, K79 (or Krt79; also known as keratin 6l), in the suprabasal cells of this domain. K79⁺ cells are specified within the hair germ during development and in the secondary hair germ during anagen. Unexpectedly, K79⁺ cells stream out of their respective compartments during hair follicle morphogenesis and regeneration, suggesting that early outward cell movement initiates lumen formation.

The INF comprises the mouth of the hair follicle and appears continuous with the interfollicular epidermis (IFE) and isthmus, as illustrated by immunohistochemical staining (IHC) for the basal epithelial marker keratin 5 (K5) (Fig. 1B). IHC for early (keratin 10) and late (involucrin, loricrin, filaggrin) epidermal differentiation markers also revealed multiple layers of suprabasal cells extending from the IFE into the INF, suggesting that these domains share some degree of similarity (Fig. 1B).

Previous studies have described a handful of markers for the INF, including Plet1 (also known as MTS24, 1600029D21Rik) (Nijhof et al., 2006; Raymond et al., 2010). In telogen follicles, we observed Plet1 primarily in a narrow domain between the bulge and sebaceous glands, with only occasional extension into the proximal half of the INF, as described previously (Fig. 1C). We further confirmed the localization of two additional proteins, cornifin-α (also known as Sprr1) and cystatin 6 (Cst6; also known as cystatin E/M), to the INF, although Cst6 was additionally observed in the IFE and near the upper bulge (Owens et al., 1996; Zeeuwen et al., 2002) (Fig. 1C). Notably, both markers are enriched in the suprabasal layers of the INF (sINF).

Over the course of our studies, we also identified novel markers of the INF. Keratin 17 (K17) is a wound-inducible protein that is expressed throughout the hair follicle (Bianchi et al., 2005; McGowan and Coulombe, 1998). We confirmed this expression pattern, but additionally observed enrichment of K17 in the sINF and upper bulge (Fig. 1C). Significantly, we found that an antibody generated against the transcription factor Gli2 strongly marks cells along the entire sINF and sebaceous glands in telogen follicles (Fig. 1D). However, IHC on embryonic day (E) 17.5 Gli2^-/- skin revealed that immunostaining is maintained even in the absence of Gli2 (supplementary material Fig. S1), indicating that this antibody non-specifically recognizes an antigen, termed Ag-7195, that localizes to the sINF.

Since the formation of a multilayered epithelium is well established in the IFE but frequently overlooked in the INF, we confirmed that the INF is multilayered using transmission electron microscopy (Fig. 1E). Indeed, multiple epithelial layers were observed in the INF, with differentiating sINF cells assuming a flattened appearance. These observations indicate that, although the INF appears continuous with the IFE, sINF cells are biochemically distinct from other compartments in the skin.

To identify Ag-7195, we used immunoelectron microscopy to determine that this protein localizes along intermediate filaments in skin, suggesting that Ag-7195 is a keratin (Fig. 2A). In mice, the keratin family includes at least 51 members (Pan et al., 2013), of which 14 have well-characterized localization patterns distinct from that of Ag-7195. Another 15 keratins were found to be poorly expressed in telogen skin. In nearly all cases, the specialized hair keratins (K31-K40) and the IRS keratins (K25-K28, K71-K74) were expressed at low levels and were not considered further (supplementary material Table S1).

To identify keratins with a similar expression pattern to Ag-7195 in suprabasal hair follicle keratinocytes, we performed gene expression studies on purified basal (integrin α6⁺) and suprabasal (integrin α6^-) hair follicle cells isolated from Shh-Cre mice expressing a ROSA26R-YFP reporter allele (Shh;YFP). Shh;YFP mice possess fluorescent hair follicles (Levy et al., 2005), which aids in the purification of these cells by flow cytometry (Fig. 2B). Of the remaining 11 keratin candidates, six (K4, K74, K76-K79) were upregulated in suprabasal YFP⁺ integrin α6^- cells, relative to basal YFP⁺ integrin α6⁺ cells (Fig. 2C). Importantly, K17 was also enriched in suprabasal cells, consistent with our observation that this keratin is upregulated in the sINF (Fig. 2C).

Since the morphology of the hair follicle INF superficially resembles that of the sweat gland duct, we next assessed the localization of Ag-7195 in eccrine glands from murine paw skin. We observed that Ag-7195 is enriched specifically in the suprabasal cells lining the sweat ducts (Fig. 2D). Lu et al. recently characterized the global gene expression of the different sweat gland compartments (Lu et al., 2012), and we re-analyzed their dataset using GEO2R. Of the 39 keratins represented in their study, only K79 appeared preferentially expressed in suprabasal duct cells (Fig. 2E; supplementary material Fig. S2).

To ascertain whether the Ag-7195 antibody recognizes K79, we overexpressed this keratin in 293FT cells. Immunoblotting revealed that the Ag-7195 antibody detects a protein of the expected size for K79 in transfected cells (Fig. 2F). To confirm that K79 localizes to the sINF, we determined that an independent antibody generated specifically against K79 also demonstrates localization to the sINF and suprabasal sweat duct, but does not recognize Gli2 (Fig. 2G; supplementary material Figs S3, S4). Along with our finding that Gli2 expression is not enriched in suprabasal hair follicle cells (supplementary material Fig. S3), these data strongly suggest that the target of the Ag-7195 antibody in the INF is K79, a largely uncharacterized type II keratin that is primarily expressed in skin (Rogers et al., 2005) (supplementary material Fig. S5). To detect K79, we utilized the Ag-7195 antibody in subsequent experiments. Although we cannot rule out the possibility that this antibody can also recognize Gli2, importantly, we validated our findings using the additional K79-specific antibody as described in supplementary material Fig. S4.

To identify the stem cells that give rise to the INF, we examined the localization of Lrig1, a stem cell marker that is reported to be enriched near the isthmus (Jensen et al., 2009). In telogen hair follicles, cells expressing both Lrig1 and K79 were detected at the isthmus, whereas cells expressing K79 alone were observed in the sINF (Fig. 3A).

Previous studies tracking individual cell clones have suggested that Lrig1-expressing cells contribute to the INF, IFE and sebaceous glands (Jensen et al., 2009). To examine whether Lrig1⁺ stem cells give rise to differentiated K79⁺ sINF cells, we performed lineage tracing using tamoxifen-inducible Lrig1-Cre^ERT2 mice harboring either a ROSA26R-lacZ or ROSA26R-YFP reporter allele (Lrig1;lacZ or Lrig1;YFP, respectively) (Powell et al., 2012). Two days after induction with tamoxifen, Lrig1;lacZ mice expressed lacZ (β-galactosidase) near the follicular isthmus, as expected (Fig. 3B). Labeled cells contributed initially to the proximal half of the INF, before filling the entire INF over a period of 3 weeks (Fig. 3B,C). lacZ⁺ cells were detected in both basal and suprabasal layers of the INF, and were occasionally observed in the IFE and sebaceous glands, but rarely found in the bulge or lower anagen bulb (Fig. 3B-D; supplementary material Fig. S6). Further labeling of the INF or other compartments was not observed over longer trace periods, indicating that the replacement of unlabeled INF cells by lacZ-expressing progenitors in these mice had reached homeostasis (Fig. 3C; supplementary material Fig. S6).

Since hair follicle bulge stem cells have been reported to contribute to sebaceous glands (Petersson et al., 2011), we also tested the contribution of bulge cells to the INF using RU486-inducible K15-Cre^PR1 mice harboring the YFP reporter allele (K15;YFP). Upon induction, K15;YFP mice exhibited fluorescent bulge cells; however, these cells did not contribute substantially to the INF (Fig. 3E). Wounding can mobilize bulge-derived cells to enter the IFE (Ito et al., 2005), and we observed that hair follicles adjacent to wounds also contained labeled cells in the INF (Fig. 3E). These YFP-expressing cells were maintained at least 50 days after wounding, and were in greatest abundance in follicles closest to the wound site.

Overall, our fate mapping studies reveal that the INF is ordinarily replenished by the upward movement of Lrig1⁺ stem cells, but not by bulge stem cells. Upon wounding, bulge-derived cells are mobilized up the follicle and can establish a permanent niche, probably taking on the role of departed Lrig1⁺ stem cells to stably repopulate the INF after the wound has healed. Similar findings have recently been reported by Page et al. (Page et al., 2013) and by C. Wong and A. Balmain (personal communication).

In mice, hair follicle development initiates at E14.5. Differentiated IRS cells first appear during stage 4 of hair follicle morphogenesis, when hair pegs develop a concave follicular base that envelopes the dermal papilla (Paus et al., 1999). Subsequently, these IRS cells migrate to the top of the hair follicle during stage 6, coinciding with the emergence of the hair canal.

We examined the expression of K79 during embryogenesis, and observed the appearance of K79⁺ cells early within the hair germ (Fig. 4A). These cells were exclusively suprabasal and, surprisingly, formed a continuous stream extending from nascent hair buds into the suprabasal epidermis (Fig. 4A). These follicles were scored as stage 2 hair germs, as indicated by the absence of a concave follicular base (Paus et al., 1999). These findings suggest that the specification of cells lining the lumen of the future hair canal occurs significantly earlier than previously reported.

The appearance of K79⁺ cells extending out into the epidermis might reflect either the movement of K79⁺ cells out of the hair germ or the upregulation of K79 in a continuous column of stationary suprabasal cells. To differentiate between these possibilities, we performed lineage tracing using Shh;YFP embryos. Since Shh expression is restricted to keratinocytes fated to become hair follicles (Levy et al., 2005), all YFP⁺ keratinocytes in Shh;YFP embryonic epidermis must have been derived from follicles. Indeed, we observed that individual K79⁺ cells originated from within early hair germs and later formed streams that expressed YFP (Fig. 4B; supplementary material Figs S7-S9). These streams were also enriched for Plet1, K17 and K10, although expression of these markers was additionally observed outside of this domain (Fig. 4C; supplementary material Fig. S10).

Importantly, in E18.5 hair pegs, cells specifically at the distal tips of migratory streams upregulated cornifin-α as well as the matrix metalloproteinase Mmp9 (Fig. 4D,E). Sharov et al. previously noted the appearance of Mmp9⁺ cells above the presumptive hair canal (Sharov et al., 2011). K79⁺ cell streams persisted until approximately postnatal day (P) 3 and displayed weakened expression of K10 (Fig. 4F,G). In more mature follicles, K79⁺ cells were lost from the epidermis and became restricted to the sINF, leaving behind gaps in the epidermis above sites of future hair canals (Fig. 4G). These findings suggest a novel mechanism for hair follicle lumen formation, whereby K79⁺ cells initially migrate outwards, extending from the hair germ into the epidermis. Subsequently, distal cells within K79⁺ cell streams express proteolytic enzymes and cornification proteins, possibly serving to weaken suprabasal cell-cell junctions or to cause cellular deterioration to generate the hair follicle orifice.

Hair follicle morphogenesis is recapitulated during anagen, when the secondary hair germ (SHG) extends into the dermis to regenerate a new hair shaft (Greco et al., 2009). Similar to during development, lumen formation during regeneration is thought to be mediated by the specification and migration of IRS cells beginning in anagen III (Müller-Röver et al., 2001). By anagen IV, the tip of the IRS growth cone reaches the old hair canal, at which point the lumens of the new and old hair follicles fuse into a single cavity through an unknown mechanism.

We examined the localization of K79 in P23 hair follicles, which have re-entered anagen. Whereas K79 is not expressed in the SHG during telogen, we observed early specification of individual K79⁺ cells in the SHGs of anagen I follicles, reminiscent of embryonic hair germs (Fig. 5A). In slightly later stage anagen follicles, streams of suprabasal K79⁺ cells were observed along the anterior face of the club hair bulge (Fig. 5B). By mid-anagen III, these streams had formed a continuous line of K79⁺ cells connecting the future companion layer (CL) of the new anagen follicle with the sINF of the pre-existing hair canal (Fig. 5C,D).

As in the case of embryonic hair germs, we sought to determine whether these streams involved cell movement. Brownell et al. previously demonstrated that Gli1-Cre^ERT2 activates reporter expression in the upper and lower bulge/SHG in telogen follicles, with a prominent gap of unlabeled bulge cells situated between these two domains (Fig. 5E) (Brownell et al., 2011). Using Gli1-Cre^ERT2 animals harboring a YFP reporter allele (Gli1;YFP), we reasoned that if K79⁺ cells in the SHG migrate upwards into the club hair bulge, then the zone of unlabeled bulge cells during telogen would be infiltrated by YFP⁺ cells during early anagen. To test this hypothesis, we induced Gli1;YFP mice with tamoxifen during telogen, depilated the skin, and examined the fate of labeled cells after 5-6 days. Individual K79⁺ YFP⁺ cells were observed within the SHG during anagen I (Fig. 5F; supplementary material Fig. S11), and in anagen II follicles K79⁺ YFP⁺ cells extended along the anterior face of the bulge (Fig. 5G; supplementary material Fig. S11). In anagen III follicles, continuous columns of K79⁺ YFP⁺ cells extended from the lower bulb up to the isthmus, eventually giving way to K79 single-positive cells in the sINF (Fig. 5H). Although this labeling pattern might also arise from the downward migration of labeled upper bulge cells, given that K79 expression originates in the SHG, the most parsimonious explanation for the YFP⁺ K79⁺ cell streams is that SHG cells migrate upwards during early anagen.

In summary, our data provide a mechanism for how the new and old hair follicle lumens fuse during regeneration. At early anagen, SHG-derived K79⁺ cells migrate up along the anterior side of the bulge. These migratory cells terminate just below the isthmus, where they join with pre-existing K79⁺ cells which reach up to the sINF, together generating a continuous layer of suprabasal cells that line a single lumen extending from the anagen bulb to the top of the hair canal.

To gain further insight into the regulation of K79, we examined its expression in pathological situations in which the INF is perturbed. In the skin, differentiation is mediated by Notch signaling, which is activated by cleavage events that generate the Notch receptor intracellular domain (NICD) (Schroeter et al., 1998; Watt et al., 2008). NICD subsequently translocates into the nucleus, where it forms a transcriptional complex that includes CSL/RBP-Jk and a mastermind-like coactivator to induce the expression of target genes, including Hes1.

In telogen follicles, we observed nuclear NICD in the sINF, suggesting that upstream Notch signaling is activated in these cells (Fig. 6A). To confirm that downstream Notch signaling occurs in the sINF, we utilized Hes1-Cre^ERT2 knock-in mice harboring either the YFP or lacZ reporter allele to monitor Notch activity (Kopinke et al., 2011). We observed YFP⁺ cells along the sINF that disappeared within 3 weeks, consistent with their rapid renewal by Lrig1⁺ stem cells (Fig. 6B). Together, these findings suggest that the sINF consists of a transient population of cells that activate Notch.

Disruption of Notch activity in the skin, either by genetic ablation of Notch pathway components or by expression of dominant-negative GFP-tagged mastermind-like 1 (dnMAML), can induce widespread cyst formation (Blanpain et al., 2006; Demehri and Kopan, 2009; Dumortier et al., 2010; Pan et al., 2004; Proweller et al., 2006; Vauclair et al., 2005; Yamamoto et al., 2003). To determine whether Notch is required to maintain the INF, we generated mice expressing Lrig1-Cre^ERT2 and a Cre-inducible dnMAML cassette under the control of the ROSA26 promoter (Lrig1;dnMAML). Upon inducing dnMAML in adult telogen skin, the INF developed keratinized cysts specifically in the INF after 20 weeks (Fig. 6C; supplementary material Fig. S12). The progeny of Lrig1⁺ stem cells with intact Notch signaling are normally confined to the INF (Fig. 3B), whereas dnMAML-expressing cells extended out into the epidermis (Fig. 6C), which is reminiscent of previous findings that perturbing Notch causes bulge cells to depart from their niche (Demehri and Kopan, 2009). Notably, infundibular cysts contained dnMAML-expressing K79⁺ GFP⁺ suprabasal cells, suggesting that canonical Notch signaling is dispensable for K79 expression.

Lastly, we examined K79 in human acne, which is characterized by perturbation of the INF (Thiboutot, 2008; Zaenglein et al., 2012). Early acne lesions develop cyst-like structures known as comedones (Fig. 6D,E), and we observed that K79 was lost in five of seven closed comedones (‘whiteheads’) removed from the faces of different patients (Fig. 6E). By contrast, all comedones retained K5, K10 and K17 (supplementary material Table S2). Although our sample size was limited, loss of K79 was generally associated with comedones that possessed larger lumenal areas. Although further studies will be needed to extend these findings, our results suggest that K79 might be lost during acne disease progression.

In polarized epithelia, tube morphogenesis is typically initiated by the infolding of a cell sheet, leading to the formation of a hollow lumen. Hair follicle primordia also invaginate into the dermis during development, but uniquely maintain suprabasal cells in their cores that require subsequent removal. Our observation that these cells migrate out of nascent hair germs prior to lumen formation suggests a unique mechanism for hair canal morphogenesis, a process that has remained poorly studied.

Indeed, early anatomical studies have suggested a variety of mechanisms for lumen formation. In the opossum, Gibbs observed that a plug of cells connected to the sebaceous glands is initially pushed into the epidermis (Gibbs, 1938). Subsequently, these cells deflect horizontally beneath the stratum corneum before keratinizing to form the canal. A similar mechanism has also been described in sheep (Wildman, 1932). By contrast, using mouse skin explants, Hardy suggested that hair canal morphogenesis begins with the keratinization of cells within the stratum spinosum of the epidermis (Hardy, 1949). This mechanism has also been proposed for canal formation in human skin (Maximow and Bloom, 1934).

In mice, the earliest sign of hair canal specification is thought to occur during stage 3 of hair follicle morphogenesis, when epidermal keratinocytes above the hair bud appear to reorient perpendicularly to the skin surface (Pinkus, 1958). By contrast, we show here that by stage 2 the nascent follicles already possess continuous streams of hair germ-derived K79⁺ cells extending out into the epidermis. In later stage hair pegs, the tips of these streams upregulate Mmp9, suggesting that the hair canal might subsequently be created through proteolysis at the distal end, leading to epidermal gaps above nascent hair follicles. Consistent with this, Mmp9^-/- mice possess slightly narrower hair canals (Sharov et al., 2011); however, additional functional evidence will be needed to clarify the role of proteolysis during hair canal formation.

During early anagen, K79⁺ streams reappear, recapitulating embryonic hair canal morphogenesis. These streams also move outwards, threading along the anterior club hair bulge and linking the future CL of the anagen follicle with pre-existing sINF cells. During homeostasis, K79⁺ cells line the sINF, and translocate upwards over time. Although these observations suggest that migratory behavior might be a common feature of K79⁺ cells, as supported by our lineage tracing studies using Shh-Cre, Lrig1-Cre^ERT2 and Gli1-Cre^ERT2 animals (Fig. 7), we cannot rule out the alternative possibility that these cells are passively pushed outwards.

Where do K79⁺ cells originate from? During anagen, matrix cells give rise to the differentiated layers of the hair follicle and hair shaft, and it is possible that an early matrix cell population might also differentiate into K79⁺ cells. However, it is important to note that K79⁺ cells appear significantly earlier than matrix descendants such as IRS cells, both during development and regeneration. K79⁺ cells also display outward movement earlier than do IRS cells, which are thought to extend distally only after they have ensheathed the growing hair shaft (Müller-Röver et al., 2001; Paus et al., 1999). Thus, if early, single K79⁺ cells within the hair germ and SHG during anagen I are indeed derived from a matrix population, this would suggest that matrix cells are specified even earlier. Clearly, this is subject that deserves further study.

Since the signaling pathways that regulate K79 remain ill defined, we examined the expression of this keratin during pathological situations involving INF perturbation, both experimentally and clinically. Upon disruption of Notch, K79 is maintained, even in spite of cyst formation in the INF, suggesting that Notch signaling is unnecessary for K79 expression. By contrast, K79 is frequently lost in human acne lesions, suggesting that defective INF differentiation might be associated with disease progression. As the causes of acne are likely to be multifactorial (Thiboutot, 2008), one or several collaborating factors might contribute to downregulating K79.

Finally, although the formation of a multilayered epithelium is well established in the epidermis, the presence of multiple layers in the INF is frequently overlooked. In the opossum, Gibbs noted that although the close juxtaposition of the suprabasal layers of the hair canal and epidermis lends the appearance of continuity between these layers, these cells are, in actuality, distinct (Gibbs, 1938). We have reached a similar conclusion by examining a series of markers (K79, K17, Plet1, cornifin-α and Cst6) that we and others have found to be enriched in the sINF (Nijhof et al., 2006; Owens et al., 1996; Raymond et al., 2010; Zeeuwen et al., 2002).

In summary, we have identified K79 as a marker of a unique population of cells that appears to migrate throughout the life cycle of the hair follicle (Fig. 7). Future studies are likely to shed light on how K79⁺ cells sculpt the hair follicle lumen during development and how these cells are disrupted during disease.

The following strains were used: Lrig1^{tm1.1(cre/ERT2)Rj} (Lrig1-Cre^ERT2) (Powell et al., 2012); Tg^{(Krt-15-cre/PGR)22Cot} (K15-Cre^PR1) (Morris et al., 2004); Shh^{tm1(EGFP/cre)Cjt} (Shh-Cre) (Harfe et al., 2004); Gli1^{tm3(cre/ERT2)Alj} (Gli1-Cre^ERT2) (Ahn and Joyner, 2004); Hes1^{tm1(cre/ERT2)Lcm} (Hes1-Cre^ERT2) (Kopinke et al., 2011); Gt(ROSA)26Sor^tm1(EYFP)Cos (ROSA26A-YFP) (Srinivas et al., 2001); Gt(ROSA)26Sor^tm1Sor (ROSA26A-lacZ) (Soriano, 1999); ROSA26-dnMAML1-GFP (dnMAML) (Tu et al., 2005); and Gli2^tm2.1Alj (Gli2^-/-) (Bai and Joyner, 2001).

Mice were induced with tamoxifen or RU486 during telogen (∼7.5 weeks of age) as described (Wong and Reiter, 2011) at the following doses: 5 mg tamoxifen per 40 g body weight (Gli1-Cre^ERT2 and Hes1-Cre^ERT2) or 1 mg tamoxifen per 40 g body weight (Lrig1-Cre^ERT2).

Antibodies used for IHC included: rabbit anti-Gli2 (Ag-7195; ab7195, 1:500, Abcam); goat anti-K79 (Y-17, 1:400, Santa Cruz Biotechnology); chicken anti-GFP/YFP (GPF-1020, 1:2000, Aves Labs); guinea pig anti-K5 (GP5.2, 1:500, American Research Products); rabbit anti-K10 (PRB-159P, 1:500, Covance); rabbit anti-involucrin (PRB-140C, 1:1000, Covance); rabbit anti-loricrin (PRB-145P, 1:500, Covance); rabbit anti-filaggrin (PRB-417P, 1:1000, Covance); rat anti-Plet1 (MTS24, 1:25, gift of Dr A. Sonnenberg, The Netherlands Cancer Institute, Amsterdam, The Netherlands) (Raymond et al., 2010); rabbit anti-cornifin-α (1:250, gift of Dr A. Jetten, NIEHS, Research Triangle Park, NC, USA) (Owens et al., 1996); rabbit anti-Cst6 (ARP53533_P050, 1:25, Aviva Systems Biology); rabbit anti-K17 (4543, 1:250, Cell Signaling); goat anti-Lrig1 (AF3688, 1:20, R&D Systems); rat anti-β4 integrin (346-11A, 1:400, BD Pharmingen); rabbit anti-NICD (4147, 1:1000, Cell Signaling); rabbit anti-Mmp9 (AB19016, 1:500, Millipore); rat anti-P-cadherin (PCD-1, 1:200, Life Technologies); guinea pig anti-K75 (GPK6hf, 1:200, American Research Products); and mouse anti-K6 (LHK6B, 1:100, Lab Vision). For whole mounts, we used the protocol of Braun et al. (Braun et al., 2003).

For TEM analysis, excised skin was fixed in 3% paraformaldehyde and 2.5% glutaraldehyde in Sorensen’s buffer for 90 minutes at 4°C. For immunogold analysis, skin was similarly fixed except with reduced glutaraldehyde (0.1%). Stainings were performed on standard grids, using Aurion BSA-C for blocking, Ag-7195 antibody (1:100 for 1 hour at room temperature) and donkey anti-rabbit secondary antibody conjugated to 10 nm gold particles (1:30 for 30 minutes at room temperature). Samples were post-fixed in 2% glutaraldehyde. All reagents were purchased from Electron Microscopy Sciences.

Excision wounds and depilation on ∼7.5-week-old telogen back skin were performed as previously described (Wong and Reiter, 2011). All studies were performed in accordance with regulations established by the University of Michigan Unit for Laboratory Animal Medicine.

Epidermal cells were obtained by overnight trypsinization of mouse back skin as described (Jensen et al., 2010). Cells were passed through a 40 μm cell strainer, blocked with rat anti-mouse DC16/CD32 Fc block (553141, BD Biosciences) for 5 minutes, and incubated with 50 ng Alexa 647-conjugated rat anti-human CD49f (α6 integrin) (MCA699A647T, ABD Serotec) per 1×10⁶ cells for 1 hour on ice. Cells were analyzed using a MoFlo Astrios cell sorter (Beckman Coulter); 3-5×10⁴ sorted cells were collected directly into RLT buffer (Qiagen) containing β-mercaptoethanol for RNA isolation.

RNA was purified using the RNeasy Micro Kit (Qiagen) and cDNA was synthesized using the High Capacity cDNA Reverse Transcription Kit (Invitrogen). Quantitative RT-PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems); for primers see supplementary material Table S3.

293FT cells were cultured in DMEM containing 10% fetal bovine serum and were transfected using Lipofectamine 2000 (Invitrogen). Mouse K79 overexpression plasmid (MMM1013-202767000) was purchased from Thermo Scientific, and pcDNA3.1-Flag-Gli2 plasmid was kindly provided by Dr M. Grachtchouk (University of Michigan). Three days after transfection, protein lysates were harvested using Laemmli buffer, resolved on 12% SDS-PAGE, and transferred to PVDF membranes, which were probed with antibodies against K79 (1:400), Ag-7195 (1:2000), FLAG (2368, 1:1000, Cell Signaling) and β-actin (A5316, 1:4000, Sigma-Aldrich).

Gene expression data from Lu et al. [accession number GSE37274 (Lu et al., 2012)] and Dezso et al. [accession number GSE7905 (Dezso et al., 2008)] were accessed through NCBI GEO2R. Data visualization was performed using CIMminer.

Human acne samples were obtained with informed consent under IRB #HUM003422 in accordance with procedures approved by the Institutional Review Board of the University of Michigan Medical School. Biopsy samples were collected for frozen sections and de-identified, and thus not regulated as per IRB guidelines.

We thank Dr R. J. Coffey (Vanderbilt University, Nashville, TN, USA) for sharing Lrig1-Cre^ERT2 mice; Drs C. Wong and A. Balmain (UCSF, San Francisco, CA, USA) for distributing Lrig1-Cre^ERT2 mice; Dr G. Fisher (University of Michigan, Ann Arbor, MI, USA) for human acne samples; Drs M. Grachtchouk and M. Verhaegen (University of Michigan) for sharing reagents; Dr A. M. Jetten (NIEHS, Research Triangle Park, NC, USA) for cornifin-α antibody; Dr A. Sonnenberg (The Netherlands Cancer Institute, Amsterdam, The Netherlands) for Plet1 antibody; and Laura Melroy (Santa Cruz Biotechnology) for K79 antibody.