Role of epidermal primary cilia in the homeostasis of skin and hair follicles

The skin performs its protective function by forming a stratified epithelium, the morphogenesis of which begins during embryogenesis when a single layer of ectoderm adopts an epidermal fate and differentiates into basal keratinocytes that express keratin 5 (K5, or Krt5) and K14 (for reviews, see Blanpain and Fuchs, 2006; Blanpain and Fuchs, 2009). Stratification of the epidermis initiates when basal keratinocytes give rise to suprabasal spinous keratinocytes that express K1 and K10. These cells further differentiate into the granular and cornified layers of the epidermis, eventually being shed into the environment and replaced by underlying cells. Thus, skin homeostasis is dependent on a balance between cell proliferation, differentiation and the shedding of cells ultimately derived from basal keratinocytes.

In the case of injury, follicular keratinocytes can exit the differentiation program and respond with increased proliferation and migration to repair the wound (Blanpain and Fuchs, 2009). The main sources of this regenerative capacity are progenitor/stem cells within the interfollicular basal layer and the follicle (for reviews, see Cotsarelis et al., 1990; Fuchs, 2009; Watt and Jensen, 2009), which are characterized by slow proliferation and long-term 5-bromo-2-deoxyuridine (BrdU) label retention (Cotsarelis et al., 1990; Morris and Potten, 1999; Tumbar et al., 2004).

Stratification of the epidermis is attained by self-renewing cells in the basal layer that each give rise to a clone of differentiating cells (Ito et al., 2005; Potten and Morris, 1988; Ro and Rannala, 2005). These clones consist of a few self-renewing progenitor cells in the basal layer together with the overlaying differentiated cells in the suprabasal layer. This stratification process in the skin requires the transcription factor Trp63 (p63) (Candi et al., 2006; Koster et al., 2004; Romano et al., 2007; Romano et al., 2009). p63 positively regulates expression of the basal keratinocyte markers K5 and K14. As such, mutations in p63 result in a single-layered epidermis that lacks hair follicles (Mills et al., 1999), whereas overexpression of a p63 isoform (ΔNp63) in basal keratinocytes causes loss of canonical Wnt signaling, alopecia, and promotes interfollicular epidermal cell fates (Romano et al., 2010).

The formation of hair follicles relies on reciprocal interactions between cells in the epidermis and dermis and is mediated through multiple pathways [e.g. sonic hedgehog (Shh), Wnt] (for reviews, see Fuchs, 2007; Schneider et al., 2009). Briefly, around embryonic day (E)13, hair follicle formation is initiated by a dermal signal that induces the formation of a thickened region in the epidermis (placode), which then induces dermal cell aggregation (dermal condensate). The placode grows downwards, surrounding the dermal condensate, which becomes the dermal papilla. Continuous reciprocal signaling between the epidermal and dermal cells regulates the downgrowth and differentiation of the multiple cell lineages of the adult pilosebaceous unit.

In adults, hair follicles undergo intermittent cycles of growth (anagen), apoptosis-mediated regression (catagen), and quiescence (telogen) (reviewed by Schneider et al., 2009). Follicle cycling involves many of the dermal and epidermal signaling pathways that function during morphogenesis. Lineage-tracing experiments indicate that the regenerative capacity of the follicle in adults is mediated by stem cells in the bulge located just beneath the sebaceous gland.

The Hedgehog (Hh) pathway has important roles in hair and skin development and maintenance. Shh is first expressed in the placode during initiation of follicle formation and again in a subpopulation of cells in the matrix of the mature follicle during the telogen to anagen transition. Mutations disrupting Shh or the downstream transcription factor Gli2 cause hair follicles to arrest during early morphogenesis (Gritli-Linde et al., 2007; Mill et al., 2003). This arrest is rescued in Gli2-deficient mice by transgenic expression of Gli2 in the basal cells using the K5 promoter, suggesting that Hh signaling in the epidermal component of the follicle is sufficient for normal development (Allen et al., 2003; Gu and Coulombe, 2008; Mill et al., 2003).

The primary cilium has been implicated as a regulator of Hh signaling (Corbit et al., 2005; Gerdes et al., 2007; Huangfu et al., 2003; Haycraft et al., 2005) and in the control of the orientation of cell division (Fischer et al., 2006; Jonassen et al., 2008). Primary cilia are present on most cell types of the mammalian body and are maintained by intraflagellar transport (IFT). IFT mediates bidirectional movement of structural and signaling components between the base and tip of the cilium (reviewed by Goetz and Anderson, 2010), and mutations in genes such as Ift88 and Kif3a disrupt IFT causing a wide variety of developmental and postnatal abnormalities (reviewed by Sharma et al., 2008).

The involvement of primary cilia in the processes utilized in hair and skin development raised the possibility that cilia might have an unappreciated role in the morphogenesis and homeostasis of the skin and skin diseases. We previously demonstrated that the ablation of cilia on dermal cells of the skin results in a phenotype that mimics the loss of Shh or Gli2, with an arrest of follicle development (Lehman et al., 2009). However, the role of epidermal cilia in skin and follicle morphogenesis and maintenance has not been explored. We address this here by examining mice in which the ciliogenic genes Ift88 and Kif3a have been disrupted in the epidermis. Collectively, the results indicate that epidermal primary cilia are not essential for ventral or dorsal hair follicle morphogenesis. Intriguingly, the data suggest that cilia function in a pathway that is involved in epidermal stress responses, homeostasis of the interfollicular epidermis (IFE), and normal keratinocyte differentiation.

The Ift88 (Ift88^tm1Bky, hereafter Ift88^F) and Kif3a (Kif3a^tm2Gsn, hereafter Kif3a^F) conditional alleles were described previously (Haycraft et al., 2007; Marszalek et al., 2000). Tg(KRT14-cre/ERT)20Efu (K14Cre^ERT) mice were obtained from E. Fuchs (Vasioukhin et al., 1999). K5-flx^GFPM2SMO mice were obtained from Dr A. Dlugosz (Allen et al., 2003; Xie et al., 1998). FVB-Tg(K14-Cre)8Brn mice were obtained from Netherlands Cancer Institute (NCI), Amsterdam (Jonkers et al., 2001). Tg(KRT14-cre)1Amc/J, Shh^{tm1(EGFP/cre)Cjt}, Gt(ROSA)26Sor, Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo} (mTmG) and Ptch1^tm1Mps/J (Ptch1-lacZ) mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). Conditional cilia mutant mice were analyzed on a mixed C57BL/6J×129P2/OlaHsd genetic background. Mice were genotyped by PCR using tail or ear DNA (for primers, see Table S1 in the supplementary material). All animals were maintained in AAALAC accredited mouse facilities at UAB or UCSF and procedures were in accordance with the IACUC regulations of each institution.

Preparation of skin sections and immunofluorescence were conducted as described (Lehman et al., 2009). Whole-mount mouse tails were analyzed by immunofluorescence as described (Braun et al., 2003). Immunohistochemistry was performed using the Vectastain ABC Kit (Vector Laboratories, Burlingame, CA, USA) as per manufacturer's instructions. RNA was isolated using the RNeasy Kit (Qiagen, Valencia, CA, USA).

Antibodies used included anti-acetylated α-tubulin (Sigma, St Louis, MO, USA, T6793; 1:2000), anti-K5 (Covance, Berkeley, CA, USA, PRB-160P; 1:1000), anti-K1 (Covance, PRB-165P; 1:500), anti-NuMA (Novus Biologicals, NB500-174; 1:300), anti-laminin (Sigma, GW20044F; 1:200), anti-ΔNp63 (Santa Cruz, sc-8431; 1:50), anti-K17 (a gift of Dr Pierre Coulombe, John Hopkins University, Baltimore, MD, USA; 1:5000), anti-loricrin (Covance, PRB-145P; 1:500), anti-phosphohistone H3 (pH3) (Cell Signaling Technologies, Rab9763; 1:400), and anti-Arl13b (gift of Dr Caspary, Emory University, Atlanta, GA, USA; 1:2000). Nuclei were stained with Hoechst 33258 (Sigma).

Proliferation was measured by BrdU injection as per the manufacturer's instructions [animal weight (g) × 30 μl; Invitrogen, Carlsbad, CA, USA) and skin isolated 2.5 hours post-injection. Long-term label retention was analyzed by injecting P10 mice with BrdU four times at 12-hour intervals.

Tail samples were taken 1, 19, 35 and 59 days post-injection and whole-mounts were prepared as above. Proliferation around the infundibulum of follicles was measured in ShhCre mutant and control mice at P23 by immunofluorescence analysis using anti-pH3. Four independent animals were analyzed per genotype (15 random follicles per animal).

For whole-mount analysis, skin biopsies were collected and fixed in 0.2% glutaraldehyde and 2% paraformaldehyde for 30 minutes on ice. After washing in lacZ buffer (2 mM MgCl₂, 0.01% NaDC, 0.02% NP40, in 100 mM sodium phosphate buffer, pH 7.3), tissues were incubated at 37°C in 1 mg/ml X-Gal diluted in 5 mM potassium ferrocyanide and 5 mM potassium ferricyanide (Taulman et al., 2001).

Clone size in K14Cre^ERT;Kif3a^F/Δ;ROSA26 mice was determined from skin biopsies taken from three independent animals per genotype group at ~39 and 70 weeks after tamoxifen injection. For quantification, in each captured image the areas of the 20 most prominent β-galactosidase⁺ epidermal clones were measured using NIS-Elements software (AR 3.2, Nikon). A total of 120 clones were analyzed per genotype.

The spatial activity of the Hh pathway was analyzed using the β-galactosidase assay with the Ptch1-lacZ reporter allele in P31 K14Cre^AMC;Ift88^F/F mutant and control mice.

Hh pathway activity was determined by quantitative real-time (qRT) PCR analysis of RNA isolated from the tail epidermis using iQ SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) with a CFX96 real-time PCR detection system (Bio-Rad). Primer sequences for qRT-PCR analysis of Gli1, Ptch1, Ift88, β-2 microglobulin (B2m) and Gapdh are listed in Table S1 in the supplementary material.

Formalin-fixed paraffin-embedded sections were used for in situ analysis as described (Grachtchouk et al., 2003). Shh probe was obtained from Dr A. P. McMahon (Harvard University, Cambridge, MA, USA). mRNA expression was detected with anti-digoxygenin antibody conjugated to alkaline phosphatase (Roche Applied Science, Branchburg, NJ, USA) and visualized with BCIP/NBT chromogen solution (Roche Applied Science).

Whole-mount tail preparations were probed with NuMA antibody as described (Braun et al., 2003). The spindle angle was calculated relative to the basement membrane using NuMA labeling in confocal image stacks (see Fig. S4 in the supplementary material for details). Analysis was performed using Volocity 5.3 software (PerkinElmer, Shelton, CT, USA) and Microsoft Excel 2007.

To begin addressing the functional importance of epidermal cilia, mice carrying the intraflagellar transport 88 (Ift88) conditional allele Ift88^tm1bky (hereafter Ift88^F) (Haycraft et al., 2007), or the kinesin family member 3A (Kif3a) conditional allele Kif3a^tm2Gsn (hereafter Kif3a^F) (Marszalek et al., 2000) were crossed with transgenic lines expressing Cre recombinase from a K14 promoter (K14Cre). Two K14Cre lines with different temporal Cre activity were used. The Tg(KRT14-cre)8Brn line (hereafter K14Cre^BRN) (Jonkers et al., 2001) does not express Cre during development, has low mosaic Cre activity perinatally (see Fig. S1A in the supplementary material), but is expressed throughout the epidermis and follicle by postnatal day (P) 6, as determined using the Cre reporter line Gt(ROSA)26S (hereafter R26R). The Tg(KRT14-cre)1Amc (hereafter K14Cre^AMC) (Dassule et al., 2000) has higher and more uniform levels of Cre activity in the epidermal components of the skin and follicle as early as E12.5-13.5 (see Fig. S1B in the supplementary material), as determined using the Cre reporter line Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo} (hereafter mTmG). In sections through the skin of K14Cre^AMC embryos at E14.5, Cre activity was uniform in the epidermis and absent in the dermis (see Fig. S1C in the supplementary material). At E16.5, Cre activity was evident throughout the developing follicle and interfollicular epidermis of the K14Cre^AMC embryos (see Fig. S1D in the supplementary material).

The presence or absence of cilia on the epidermis and hair follicle of adult K14Cre mutant and control mice was analyzed by immunofluorescence microscopy (see Fig. S2A,B in the supplementary material). In both K14-cre^BRN;Ift88^F/F and K14Cre^AMC;Ift88^F/F mutants, cilia were deleted from the cells in the epidermis of the follicle and in the interfollicular region (see Fig. S2A,C versus S2B,D in the supplementary material). Genotyping of DNA isolated from the epidermis or dermis of the tail of P31 K14Cre^AMC;Ift88^F/F conditional mutants indicated a substantial decrease of the Ift88 allele in the epidermis (see Fig. S2E in the supplementary material) as compared with the whole tail or dermis. The presence of a deleted allele in the dermis is likely to reflect contamination with epidermal cells. Western blot analysis of protein from the tail epidermis of the P31 K14Cre^AMC;Ift88^F/F conditional mutant also showed near complete loss of Ift88 (see Fig. S2F in the supplementary material).

To evaluate Ift88 protein depletion, skin samples from E18.5 K14Cre^AMC;Ift88^F/F mutant and control embryos were analyzed by immunofluorescence microscopy using antibodies against Ift88 and acetylated α-tubulin as a cilia marker. Although cilia could be detected in the epidermis of both K14Cre^AMC;Ift88^F/F mutant and control embryos, in the conditional mutants these epidermal cilia did not contain Ift88, in contrast to dermal cilia (see Fig. S2G,H in the supplementary material). The persistence of cilia for a short period after inducing Ift88 deletion is seen in vivo in the skin and kidney as well as in vitro using tamoxifen-inducible Cre Ift88^F/F cell lines (B.K.Y., unpublished). Collectively, these data indicate that the K14Cre^AMC line disrupts IFT during development and would be useful for evaluating epidermal cilia function in hair follicle morphogenesis, whereas the K14Cre^BRN line would address the role of cilia in hair follicle cycling and epidermal homeostasis in adults. However, the caveat of this analysis is that Ift88 loss in the K14Cre^AMC line was not necessarily complete at the onset of follicle morphogenesis.

In newborn epidermal cilia mutants generated with either of the K14Cre lines, there were no defects evident in the skin or follicle (data not shown); however, mutants began to exhibit a ruffled hair pattern on the ventrum at ~3 weeks of age. The phenotype became more severe over time and led to regions of alopecia in older animals (Fig. 1D). Identical phenotypes were obtained using the Kif3a^F/F or Ift88^F/F lines and in most cases data are shown for Ift88^F/F mice (Fig. 1).

Histological abnormalities in the ventral skin of mutants included a thickened epidermis and basaloid hyperplasia (Fig. 1B). The hyperplasia was confined to the basal layer of the epidermis and there were no signs of hyperkeratosis, suggesting a healthy stratum corneum. In aged K14Cre^AMC;Ift88^F/F mutants (P730), the basal epidermis showed morphological signs of palisading (Fig. 1D, inset), which is also seen in disorders associated with increased Shh activity (Nieuwenhuis et al., 2007; Nieuwenhuis et al., 2006). There were also epidermal ingrowths into the dermis that resembled ectopic hair buds (Fig. 1D, arrow). No tumors or blistering were ever seen in mutant (K14Cre^BRN;Ift88^F/F) mice.

Although the phenotype in the ventral skin was the same in each of the K14Cre mutant lines, the tails of K14Cre^AMC cilia mutants alone showed significant pilosebaceous dysgenesis compared with controls (Fig. 2B). In control animals, the pilosebaceous units of the tail were in groups of three equally spaced follicles positioned in rows along the skin, with each follicle having two sebaceous gland lobules (Fig. 2C). In the K14Cre^AMC;Ift88^F/F cilia mutants, the follicles were in abnormal clusters, had five or more sebaceous gland lobules and were often misoriented (Fig. 2D). The phenotype in the tail was evident by at least P3 (Fig. 2H), when follicle-like structures were often seen in a swirled pattern distinct from that of littermate control mice (Fig. 2G,I).

To characterize the thickened epidermis, we crossed mice carrying the Kif3a^F and Ift88^F conditional alleles with Tg(KRT14-cre/ERT)20Efu (hereafter K14Cre^ERT) (Vasioukhin et al., 1999) and R26R mice. Cilia loss was induced on the basal cells at 4.5 weeks of age by tamoxifen injection. Clone size in cilia mutant and control samples was then analyzed at 39 (Fig. 3A,B) and 70 (Fig. 3C,D) weeks post-injection. Interestingly, mutant clones were markedly larger than those of control mice (Fig. 3E,F).

To assess whether the increased clone size was associated with proliferation, we injected BrdU into adult K14-Cre^AMC;Ift88^F/F and control mice. BrdU incorporation into basal (K5; Fig. 4A,B) and spinous (K1; Fig. 4C,D) cells was quantified in three-dimensional stacks of the epidermis in whole-mount tail samples. All of the animals analyzed were 4 to 6 months old and comparisons were made only between littermates. The number of proliferating cells/μm² was significantly increased in the IFE of mutants compared with controls (Fig. 4E; Student's t-test, P<0.01; there was no change in cell size in the mutants compared with controls). A modest, but insignificant increase, was also observed in the follicles (Fig. 4E; Student's t-test, P=0.23).

Markers for the different skin layers were used to characterize differentiation in the K14Cre^Brn;Ift88^F/F line. Analysis of loricrin (Fig. 5A,B) and K1 (Fig. 5C,D) expression indicated that the granular and spinous layers are not hyperplastic. Expression of K1 (Fig. 5D) was discontinuous in the skin of conditional mutants, which was not evident in controls (Fig. 5C). By contrast, the K5 (Fig. 5F) basal layer was expanded in mutants compared with controls (Fig. 5E). Normally, K6 and K17 are expressed in sublayers of the hair follicle (Fig. 5G,I) (Gu and Coulombe, 2008; Panteleyev et al., 1997); however, in mutants, both K6 and K17 were detected ectopically in the IFE (Fig. 5H,J).

ΔNp63 is a transcriptional regulator that is normally expressed in basal keratinocytes of the proliferative layer and at a lower level in suprabasal differentiated layers of mature epidermis. It is required for commitment of the immature ectoderm to epidermal lineages, for epidermal maturation and maintaining the proliferative potential of basal keratinocytes, and its expression is induced by wounding (Koster et al., 2004; Mills et al., 1999; Romano et al., 2009; Romano et al., 2010; Yang et al., 1999); (Ichikawa et al., 2008). In addition, overexpression of ΔNp63 in the epidermis causes epidermal thickening, promotion of interfollicular epidermal fate and rapid depletion of follicular stem cells (Romano et al., 2009). Many of these phenotypes are seen in the epidermal cilia mutants (see below). In K14Cre^Amc;Ift88^F/F mutants, there was a significant increase in ΔNp63 in the basal keratinocytes at P3 and P28 (Fig. 6A,C,D). ΔNp63 was also elevated in the K1 suprabasal layers of the P28 mutants (Fig. 6B), although this was not statistically significant (P=0.6). It is not yet known whether this effect on ΔNp63 is a direct or secondary consequence of cilia loss.

The phenotypes of the epidermal cilia mutants are reminiscent of those of mice with a truncated form of the Shh receptor patched 1 (Ptch1) (Nieuwenhuis et al., 2007) and of mice in which the Shh pathway is ectopically activated in basal cells (Adolphe et al., 2004). These shared phenotypes, along with the connection between Hh and cilia, prompted a comparison between the epidermal cilia mutants and mice expressing activated smoothened (Grachtchouk et al., 2003; Wong et al., 2009). To accomplish this, we generated K14Cre^BRN;K5-M2SMO (K5-flxGFP-M2SMO) (Xie et al., 1998) mice in which Cre-mediated excision of a stop sequence between the K5 promoter and a M2SMO (activated) minigene induces expression of smoothened in the basal cells. Both the K5-M2SMO and cilia mutant mice had a ‘scruffy’ ventral appearance, an expanded basal layer, ectopic epidermal ingrowths (Fig. 7B,C, arrows) (Hoseong Yang et al., 2008) and ectopically expressed K17 (Fig. 5J and see Fig. S3 in the supplementary material). Generally, the phenotype of the epidermal cilia mutants was less severe than that of the activated smoothened mice.

Elevated Hh activity might occur in epidermal cilia mutants as a consequence of the requirement for cilia in Gli repressor formation (Haycraft et al., 2005; Huangfu and Anderson, 2005; Huangfu et al., 2003; Liu et al., 2005). Alternatively, repression of the Hh pathway could also be responsible as cilia are required for the activation of Gli (Haycraft et al., 2005; Huangfu et al., 2003). To distinguish between these scenarios, expression of the Hh target genes Gli1 and Ptch1 was examined by qRT-PCR in RNA isolated from tail epidermis of P31 K14Cre^AMC;Ift88^F/F mice (Fig. 7D). Expression of both Gli1 and Ptch1 was decreased along with that of Ift88 in the mutants. The spatial activity of the Hh pathway was analyzed in the skin of P31 K14Cre^AMC;Ift88^F/F mutant and control mice using the Ptch1-lacZ reporter line. Similar to the qRT-PCR data, the Hh pathway was found to be repressed in the follicles of cilia mutants as compared with controls (Fig. 7E). There was no pathway activity evident in the IFE of either mutant or controls (data not shown).

To more specifically analyze primary cilia in the hair follicle, we crossed the Ift88^F/F and Kif3a^F/F mice onto the Shh^{tm1(EGFP/cre)Cjt} (hereafter ShhCre) (Harfe et al., 2004; Levy et al., 2005). ShhCre is expressed in the placode during morphogenesis and in the lateral disc of matrix cells in anagen follicles (Levy et al., 2005). Descendants of ShhCre-expressing cells contribute only to the follicle during development. Similar to the results obtained in the K14Cre mutants, ShhCre cilia mutants did not exhibit defects in hair follicle development, despite the follicles being largely composed of descendants from Cre-expressing cells (Fig. 8B). In part, this might reflect stability of the cilium after Ift88 or Kif3a has been deleted, although most cells that expressed Cre did not have primary cilia, and those that were retained were severely truncated (Fig. 8C). To conduct lineage-tracing studies we crossed the ShhCre mice onto the R26R line and found that the mutants showed that the cells in which Cre had been expressed were markedly expanded (Fig. 8D) and were present abnormally in the IFE (Fig. 8B,E,F, arrows).

The presence of follicular cells in the IFE normally occurs only in response to wounding (Ito et al., 2005; Levy et al., 2005; Levy et al., 2007). In situ analysis showed that Shh is expressed normally in the lateral disc within the matrix cells (Fig. 8G) (St-Jacques et al., 1998) and not in the IFE. Thus, the ectopic cells in the IFE are likely to result from the migration of follicular cells into the IFE and not from ectopic Shh expression (Fig. 8B,D-F). Furthermore, in ShhCre cilia mutants, there was ectopic expression of K17 (Fig. 8H) and increased ΔNp63 (Fig. 8H, left) in these clones. This was similar to what was seen in the IFE of K14Cre cilia mutants (Fig. 5J), although we did not observe a population that co-expressed K6 (Fig. 8H, right).

The expanded K5-positive basal cell layer, altered keratinocyte differentiation and the epidermal thickening suggested that the phenotype might be associated with an abnormal maintenance of progenitor cell populations. This was analyzed using long-term label-retention assays (Cotsarelis et al., 1990; Morris and Potten, 1999; Tumbar et al., 2004). In the hair follicle, retention of BrdU label after a long chase period is used as an indicator of progenitor cells. For this analysis, P10 control and epidermal cilia mutant mice (K14Cre^AMC;Ift88^F/F) were injected with BrdU four times over a 48-hour period, which labels most cells in the follicle and IFE (Fig. 9A,B, insets). The depletion of label, relative to the day after BrdU injection, was evaluated in whole-mount tail epidermis 19, 35 and 59 days post-injection. The number of cells retaining label after each chase period was not markedly different in the IFE of mutants and controls (Fig. 9C,D). By contrast, in the follicle there was a greater and more rapid decrease in the number of cells retaining label in the mutants compared with controls. This was especially evident at day 19 (16.3% in controls versus 5.8% in mutants; Student's t-test, P<0.05; Fig. 9A,B,E). At the end of the chase period there were fewer labeled cells in the bulge region of mutants, with many follicles lacking labeled cells (Fig. 9B,D).

In the kidney, ciliary dysfunction can cause the misorientation of cell divisions (Fischer et al., 2006; Jonassen et al., 2008; Patel et al., 2008). The expansion of the basal layer in the epidermal cilia mutants raised the possibility that altered cell divisions might contribute to the phenotypes. This possibility was evaluated in whole-mount tail preparations from control and K14Cre^AMC;Ift88^F/F adult mice by staining with antibodies against nuclear mitotic apparatus protein (NuMA; Numa1 – Mouse Genome Informatics) (see Fig. S4 in the supplementary material) (Lechler and Fuchs, 2005). The angle of the spindle was measured relative to the plane of the basal membrane. There were no significant differences in either the IFE or follicles of mutants versus controls in the distribution of the angles of cell division (Student's t-test; follicle, P=0.97; IFE, P=0.72), which were typically within 25 degrees of the basal membrane (see Fig. S4 in the supplementary material).

Primary cilia are present on most cell types in the skin (Elofsson et al., 1984; Lehman et al., 2009; Lehman et al., 2008; Wandel et al., 1984; Warfvinge and Elofsson, 1988), yet their functional importance in the morphogenesis and maintenance of the skin and follicle is only now being investigated. Loss of dermal cilia results in follicular arrest due to attenuated Hh signaling (Lehman et al., 2009). Here, we demonstrate that the loss of cilia on the epidermis has a minimal or no effect on morphogenesis of the ventral and dorsal hair follicles, but has a marked impact on skin homeostasis. For example, we observed progressive epidermal ridging, ventral alopecia, basaloid hyperplasia, abnormal keratinocyte differentiation and ectopic epidermal ingrowths into the dermis. In the tail of the early cilia mutant line (K14Cre^AMC), the phenotype is more severe: the follicles are dysmorphic, abnormally clustered and possess excess sebaceous gland lobules, demonstrating that cilia are required for tail follicle morphogenesis.

The minimal follicular phenotype is an intriguing finding considering the known importance of Hh in hair follicle morphogenesis and cycling (Corbit et al., 2005; Gritli-Linde et al., 2007; Haycraft et al., 2005; Liu et al., 2005; May et al., 2005; St-Jacques et al., 1998). The alopecia in Gli2 mutant skin can be rescued by expressing Gli2 specifically in the K5-positive basal cells, further suggesting that Shh activity in the epidermis is crucial for normal follicle development (Mill et al., 2003). However, we find that phenotypes associated with the loss of Hh signaling are not evident in the epidermal cilia mutants, suggesting that Hh signaling through epidermal cilia is not required for follicle morphogenesis and cycling. Indeed, our data indicate that the pathway is attenuated in the epidermis. However, it remains possible that a sufficient number of cells with cilia might remain in the epidermis to allow normal development and cycling to proceed.

In adults, two-thirds of the follicle is replenished each cycle largely from stem cells in the bulge (for reviews, see Blanpain and Fuchs, 2006; Blanpain and Fuchs, 2009; Fuchs, 2009). Intriguingly, epidermal cilia mutants have a marked reduction in the number of cells retaining BrdU label, suggesting a more rapid turnover of the stem cell population (Cotsarelis et al., 1990; Morris and Potten, 1999; Tumbar et al., 2004). This reduction implies that cilia have a role in stem cell homeostasis and helps explain the mild alopecia that is observed in older mutants.

In the K14Cre epidermal cilia mutants there is an increase in the number of cells expressing ΔNp63, particularly in the perinatal period. Increased ΔNp63, along with ectopic K17 expression, was observed in patches of cells located just outside of the follicles of the cilia mutants generated using ShhCre. Recently, a mouse expressing ΔNp63 in the epidermal basal layer was generated (Romano et al., 2010) and exhibits phenotypic overlap with the epidermal cilia mutant mice. For example, in both lines the expression of several keratins (K5, K17 and K6) was expanded and there was a reduction in the number of progenitor cells within the bulge. ΔNp63 is thought to promote interfollicular epidermal cell fates and our analysis in ShhCre lines suggests that a similar effect might be occurring in these mutants. However, there are also several key differences between the epidermal cilia mutants and the ΔNp63-overexpressing line. The cilia mutants do not show changes in terminally differentiated keratinocytes (loricrin staining), nor was there a decrease in proliferation within any domain of the follicular unit. Finally, the phenotypes of the ΔNp63-overexpressing mice were a consequence of reduced Wnt signaling, which was not evident in the epidermal cilia mutants (data not shown). Collectively, this indicates that although the activity of ΔNp63 is altered by the loss of cilia, this is not solely responsible for the phenotypes.

One theory concerning the regulation of differentiation and homeostasis in stem cell populations relies on control of the axis of division of basal keratinocytes (Lechler and Fuchs, 2005). Symmetric division parallel to the basement membrane is thought to result in two equivalent cells that replenish the basal stem cell population. By contrast, in an asymmetric division, the newly formed daughter cell is positioned into the suprabasal layer, initiating the differentiation program (Lechler and Fuchs, 2005). Thus, defects in the regulation of this process could cause an expansion of the basal population, as observed in the epidermal mutants. Furthermore, a precedent connecting cilia and the regulation of the orientation of cell division already exists in the kidney (Fischer et al., 2006; Jonassen et al., 2008; Patel et al., 2008). However, our analysis suggests that there were no differences between epidermal cilia mutants and controls with regard to their axis of division.

Lineage-tracing studies using a tamoxifen-inducible K14Cre^ERT line revealed that loss of cilia leads to an increase in clone size in adults. These clones ectopically express K17, suggesting that ciliary functions are involved in cell differentiation decisions or cell behavior, such as the migration of follicular cells into the IFE. This is further supported by the results from the ShhCre line that disrupts cilia only in the follicle, yet labels cells located in the mutant IFE. These ectopically positioned cells also express K17 and show elevated ΔNp63. Descendants of the Shh-expressing cells are normally restricted to the follicle but can contribute to the IFE under conditions of wounding (Levy et al., 2007). Based on these data and the depletion of label-retaining cells in the follicle, we propose that the loss of cilia on the epidermis might mimic conditions of epidermal damage or affects pathways involved in regulating a response to epidermal stress.

An association between the injury response and cilia has been noted previously. Primary cilia are temporarily elongated after obstruction of the pancreatic duct as well as after ureteral obstruction or ischemic reperfusion injury in the kidney (Hamamoto et al., 2002; Hellman et al., 2010; Verghese et al., 2009; Verghese et al., 2008; Wang et al., 2008). In addition, recent studies in the kidney have revealed that cyst development associated with cilia dysfunction is highly dependent on when cilia loss occurs (Davenport et al., 2007; Piontek et al., 2007). In adult induced cilia mutants, renal cyst development is greatly enhanced by a subsequent injury. Together, these data support the idea that in the absence of the cilium, there is an inappropriate activation or impaired deactivation of a wound response that contributes to the pathogenesis. In the skin, this might cause inappropriate recruitment of cells out of the follicle into the interfollicular epidermis leading to epidermal thickening, ectopic expression of K17 and ΔNp63 and a more rapid turnover of the stem cell population.

In summary, these data indicate that cilia on epidermal cells are not essential for ventral or dorsal follicle morphogenesis, but are required for normal homeostasis of the IFE, normal label retention in the bulge and are possibly involved in regulating the stress response. Loss of epidermal cilia caused hyperplasia, increased ΔNp63, expanded clone size and altered keratinocyte differentiation. There were also changes in cell behavior, such that cells that should have been restricted to the follicle were present in the IFE and expressed ectopic keratins and other markers associated with a wound response. The phenotypes were not correlated with an alteration in the orientation of cell division or Wnt pathway activity; however, Shh signaling was attenuated. Interestingly, no Hh-related phenotypes were noted, suggesting that the epidermis is not the primary Hh-responsive component of the skin. Importantly, these data indicate that the function of primary cilia on epidermal cells is distinct from that on dermal cells and that the phenotype of the epidermal cilia mutants might involve novel pathways not previously associated with the cilium.

We thank the UAB Comparative Pathology Core for histology services; Pierre Coulombe and Michael Ruppert for K17 antibodies; Tamara Caspary for the Arl13b antibody; and Andrezj Dlugosz for the K5-flx^GFPM2SMO mice. This work was supported by a T32 predoctoral award (AR047512-07, Dr McDonald) to A.K.O.; postdoctoral fellowships from the American Cancer Society and the NIH (1K99AR059796) to S.Y.W.; by NIH RO1s (AR052792 and HD056030) to B.K.Y.; ES015323 to M.A.; and AR054396, March of Dimes, Burroughs Wellcome Fund, Packard Foundation and the Sandler Family Supporting Foundation to J.F.R. Deposited in PMC for release after 12 months.