Specific MicroRNAs Are Preferentially Expressed by Skin Stem Cells To Balance Self-Renewal and Early Lineage Commitment

Summary

Increasing evidence suggests that microRNAs may play important roles in regulating self-renewal and differentiation in mammalian stem cells (SCs). Here, we explore this issue in skin. We first characterize microRNA expression profiles of skin SCs versus their committed proliferative progenies and identify a microRNA subset associating with “stemness”. Of these, miR-125b is dramatically downregulated in early SC-progeny. We engineer an inducible mice system and show that when miR-125b is sustained in SC-progenies, tissue balance is reversibly skewed towards stemness at the expense of epidermal, oil-gland and HF differentiation. Using gain-and-loss of function in vitro, we further implicate miR-125b as a repressor of SC differentiation. In vivo, transcripts repressed upon miR-125b induction are enriched >700% for predicted miR-125b targets normally downregulated upon SC-lineage commitment. We verify some of these miR-125b targets, and show that Blimp1 and VDR in particular can account for many tissue imbalances we see when miR-125b is deregulated.

Introduction

Stem cells (SCs) possess intrinsic long-term properties that enable them to self-renew and differentiate into the specialized cell types of the tissue in which they reside. They are the basis of tissue homeostasis, wound-repair and likely tumor growth as well (Rossi et al., 2008; Weissman, 2000). Elucidating the transcriptional and posttranscriptional mechanisms that control these properties is essential for understanding the basic principles of animal development, as well as realizing their therapeutic potential for regenerative medicine.

One excellent system to study SCs is the mouse skin, where distinct populations of epithelial stem cells give rise to different tissue structures with distinct turnover rates. SCs within the innermost (basal) layer of the interfollicular epidermis (IFE) and hair follicle (HF) orifice (infundibulum) generate a stratified, terminally differentiated epidermis, which turns over at a relatively constant rate in adult mice. Sebaceous gland (SG) SCs within the junctional zone of the HF outer root sheath (ORS) generate proliferative precursors for the terminally differentiated sebocytes that secrete lubricating oil for the hair and skin surface (Horsley et al., 2006; Jensen et al., 2009; Snippert et al., 2010). Just below this zone is the “bulge”, a niche for the infrequently cycling SCs that fuel the hair cycle (Blanpain and Fuchs, 2009; Hsu et al., 2011). During the growth phase (anagen), the HF below the bulge regenerates and produces hair for several weeks. The HF then triggers a rapid destructive phase (catagen), which is followed by a quiescent resting phase (telogen) which can last for weeks (Cotsarelis et al., 1990; Oshima et al., 2001; Taylor et al., 2000). HF-SCs in the bulge are typically slow-cycling, but they activated at the start of the growth phase. The earliest progeny are thought to be cells along the ORS, which progress to become the transit-amplifying (TA) matrix at the bulb of the HF. After a brief period of active proliferation, matrix cells then terminally differentiate into one of the 6 distinct layers of the hair shaft and its channel, the inner root sheath (IRS) (Blanpain and Fuchs, 2009).

IFE, SG and HF SCs share certain features, which include their expression of transcription factor p63 and keratins K14 and K5, as well as their high levels of integrins, reflective of their adherence to the underlying basement membrane that is rich in extracellular matrix and growth factors. Their close relation is further evidenced by the fact that upon injury, HF-SCs can give rise to IFE and SG progenitors (Blanpain et al., 2004; Claudinot et al., 2005; Horsley et al., 2006; Levy et al., 2005; Morris et al., 2004; Nowak et al., 2008; Tumbar et al., 2004), a feature also suggested for IFE SCs (Ito et al., 2007). However, bulge SCs are distinguished from other skin SCs by their expression of CD34, Lgr5 and K15, and in addition, a group of essential transcription factors (Sox9, Nfatc1, Tcf3, Tcf4 and Lhx2) (Blanpain and Fuchs, 2009). Most of these markers and the slow-cycling properties of bulge HF-SCs are already evident in neonatal mice prior to completion of HF and SG morphogenesis (Nowak et al., 2008).

Posttranscriptional regulation in skin is also important, as evidence by recent studies on microRNAs (miRNAs), which typically dampen expression of proteins by inducing degradation or translational inhibition of their target mRNAs (Ambros, 2004; Bartel, 2004). Most of the evidence comes from K14-Cre mediated conditional loss of function studies on core components of the machinery that generates these small (20- to 24-nucleotide) non-coding regulatory RNAs (Andl et al., 2006; Yi and Fuchs, 2010; Yi et al., 2006; Yi et al., 2009). miRNAs have been cloned from embryonic epidermis and HFs, and temporal analyses at the tissue level suggest that they have distinct patterns of expression in these two epithelia (Yi et al., 2006; Yi et al., 2009). Of these, miR-203 has been identified as an inhibitor of 'stemness' in IFE, where it is thought to act by repressing p63 translation as cells commit to differentiate (Lena et al., 2008; Yi et al., 2008). Conversely, miR-205 is expressed throughout epidermis, where it is thought to function by regulating Akt activity and guard against apoptosis (Yu et al., 2010). The functions of miRNAs in SG and HF SCs remain unexplored, although their significance is suggested by the evagination of hair buds lacking miRNAs (Andl et al., 2006; Yi et al., 2006; Yi et al., 2009) and by the severely restricted HF and SG development seen in the few K14-Cre X Dicer mutant mice that survive into neonatal life (Andl et al., 2006).

Given increasing roles of miRNAs in regulating embryonic and adult SCs (Crist and Buckingham, 2009; Delaloy et al., 2010; Guo et al., 2010; Hatfield et al., 2005; Li and Gregory, 2008; Zhao et al., 2010), the relevance of specific miRNAs to skin SC biology merited further exploration. In the present study, we generated miRNA signatures for four different populations of keratinocytes that possess proliferative capacity within postnatal day 4 (P4) skin. To accomplish this goal, we exploited their differential expression of integrins, K14 and Sox9. We were particularly interested in miRNAs upregulated in P4 GFP+ ORS cells from Sox9GFP transgenic mice, given that lineage tracing revealed that these cells not only give rise to the adult HF and SG SCs, but also contribute to long-term repair of IFE upon wounding (Nowak et al., 2008).

Through miRNA array analysis, we show that surprisingly, the miRNA signature of the early Sox9GFP-expressing cells is closely related to that of IFE basal cells but strikingly different from non-bulge ORS or matrix. We focused on miR-125b as a miRNA that is strongly expressed in Sox9 promoter-active cells, somewhat expressed in basal epidermal cells, and significantly downregulated in ORS and matrix. We show that by sustaining miR-125b throughout K14-expressing cells, the transition from stemness to committed progenitor is markedly retarded both in vivo and in vitro. The consequences in vivo are dire--a hyperthickened epidermis, enlarged SGs and a failure to generate a hair coat--but are fully reversible upon restoring normal miR-125b regulation. We isolate and profile transcripts that are expressed by WT early bulge and non-bulge progeny but repressed when miR-125b is sustained in progeny. Interestingly, transcripts repressed upon miR-125b induction are >700% enriched for conserved predicted miR-125b targets which are normally downregulated in early bulge SCs versus their immediate early progenies. This enabled us to hone in on two key genes encoding transcription factors Blimp1 and VDR. We show that both are bona fide miR-125b targets in skin and can account for many of the phenotypic abnormalities that we see.

Results

Purification of Early Bulge Cells by Fluorescence-Activated Cell Sorting (FACS)

Our purification scheme was based upon transgenic mice co-expressing K14-RFP and Sox9-EGFP (Supplemental Experimental Procedures). Despite some EGFP in suprabasal cells of the upper HF near the hair shaft, only the early ORS region encompassing the future bulge and junctional zones showed strong co-labeling of EGFP/Sox9, α6 integrin and RFP/K14, thereby providing specificity for what we hereafter refer to as “early bulge cells” (Figure 1A). After enzymatically separating epidermis and dermis (including HFs) and generating single cell suspensions, we used FACS to fractionate populations containing early bulge cells (Early Bulge) (RFP^HiGFP^Hiα6^HiLin^neg), ORS cells excluding the early bulge region (RFP^hiGFP^negα6^HiLin^neg) (mostly ORS cells below the bulge), transit-amplifying matrix cells (Matrix) (RFP^LoGFP^negα6^LoLin^neg) and IFE basal cells (Basal) (RFP^Hiα6^HiLin^neg) (Figures 1A–1C).

Figure 1. FACS Purifications of Backskin Populations Possessing Proliferative Capacity.

Mice used are K14-RFP/Sox9-EGFP at P4.

(A) Immunofluorescence for markers used in purification. Bu, early bulge; ORS, non-bulge ORS; Mx, matrix; Epi=IFE basal layer; Inf, infundibulum; SG, sebaceous gland. Color coding is according to secondary antibodies (Abs) used. Scale bar represents 50µm.

(B, C) FACS purifications. (B) HFs: RFP^HiGFP^Hiα6^HiLin^neg early bulge, RFP^LoGFP⁻α6⁺Lin^neg non-bulge ORS, RFP^LoGFP⁻α6^LoLin^neg matrix, and RFP⁺Lin^neg reference cells (RFP All). (C) IFE: RFP^Hiα6^HiLin^neg IFE (basal).

(D) RT-PCR. Vertical axis represents Log2 transformed Fold of enrichment comparing with the RFP All reference population. Error bars represent standard error. GAPDH was used as a reference gene for all RT-PCR.

See also Table S1.

Real-time PCR verified the efficacy of our strategy and lineage assignments (Figure 1D). As expected, Sox9 mRNA levels were highest in the RFP^HiGFP^Hiα6^HiLin^neg cells. This population was also enriched for other early bulge SC markers, including Nfatc1, Tcf3 and Lhx2 (Blanpain and Fuchs, 2009), as well as junctional zone markers, Lgr6 and Lrig1 (Figure 1D, Table S1). Consistent with their known expression patterns (Horsley et al., 2008) (Blanpain and Fuchs, 2009), Nfatc1 was downregulated in the non-bulge ORS population, while Sox9, Tcf3, Lhx2 and Lgr5 persisted (Figure 1D, Table S1). By contrast, the IFE and infundibulum marker Sca-1 was enriched in RFP^Hiα6^HiLin^neg basal cells, while matrix marker Msx2 was prominent in the RFP^LoGFP^negα6^LoLin^neg pool. As negative controls, c-kit (melanocytes) was not detected, and vimentin (fibroblasts) was only found at low levels in the matrix fraction. Taken together, these results confirmed our designations of these populations and their purity.

Defining a miRNA Signature for Early Bulge Cells

To identify miRNAs that are preferentially expressed in each FACS-purified fraction, we performed miRNA microarrays (Experimental Procedures). Two interesting observations emerged upon analyzing these data. First, unsupervised Hierarchical clustering revealed that the miRNA profile of the early bulge fraction was more closely related to the epidermal than ORS or matrix fractions (Figure 2A). This was surprising, since bulge and ORS share more transcriptional similarities than basal epidermis. Second, individual members of an miRNA family that possess a common “seed” sequence tended to exhibit similar expression patterns among different skin cell populations (Figure 2B). The coordinate expression of functionally related miRNAs suggested a possible means of reinforcing the regulation of key miRNA target genes for these SC populations.

Figure 2. Differential Expression of miRNAs.

(A) Unsupervised Hierarchical clustering and Heat map display of miRNA expression profiles of FACS-purified populations. Comparisons are normalized to the RFP-All reference cell population. Each row represents an individual miRNA. Each column represents an independent sample. 2–3 biological replicates were analyzed for each population. Red represents downregulation while green represents upregulation compared to RFP-All.

(B) Histogram of Pearson’s correlation coefficient of expression patterns between all possible miRNAs (Blue Bars) or between miRNAs that belong to the same miRNA family (Intra-family, Red Bars). The correlation of expression was calculated based on the miRNA expression profiles presented in (A).

(C) miRNA-microarray results and RT-PCR verification of expression patterns of the miRs indicated. U6 was used as the reference gene in RT-PCR. Error bars represent Standard Error.

(B) (D,E) miR-125b in situ hybridizations of WT P4 and P17 backskins. In situ signals (black) were developed by BM purple substrate. K5 (green) in ORS and basal IFE. Scale bars represent 50µm.

We selected four miRNAs with divergent expression patterns--miR-125b, miR-24, miR-205 and miR-31--and employed Taqman probes to verify their expression by RT-PCR (Figure 2C). MiR-125b was notable for its strong enrichment in the early bulge fraction. In situ hybridization confirmed its localized expression in the upper ORS (Figures 2D, 2E).

MiR-125b was particularly interesting in that it is a mammalian ortholog of C. elegans lin-4 miRNA, the first animal miRNA characterized. In C. elegans, lin-4 targets lin-14 and lin-28 to regulate temporally dependent developmental programs (Lagos-Quintana et al., 2002; Lee et al., 1993; Moss et al., 1997; Wightman et al., 1993). The lin-4(miR-125)-lin-28 regulation pair has been evolutionarily conserved and functions critically in mammalian embryonic SC differentiation (Rybak et al., 2008; Wu and Belasco, 2005). miR-125b’s localization in the upper ORS made it a candidate for regulating early bulge SCs. However, since lin-28 was not expressed in post embryonic skin, miR-125b must control other targets in the HF.

Maintaining miR-125b in Skin SC Progeny Causes Striking Pleiotropic Phenotypes

To explore miR-125b’s role in skin SCs, we first devised a strategy to sustain its expression in non-bulge ORS, where it is normally downregulated, and to upregulate it in basal epidermal cells, where it is normally expressed at lower levels than in bulge. For this purpose, we engineered three lines of transgenic mice to express miR-125b under the control of a tetracycline regulatory element (TRE) (Figure 3A). The TRE-miR-125b mice were then mated with our previously described transgenic line that expresses the tetracycline-sensitive transactivator, rtTA2S-M2-VP16 (rtTA) controlled by a keratin 14 (K14) promoter that is active in ORS and basal epidermis (Nguyen et al., 2006). Within 24hrs after tetracycline (doxycycline, Dox) induction, K14-rtTA/TRE-miR-125b double transgenic mice (DTG) exhibited robust miR-125b expression in all K14(+) cells, including basal epidermis and ORS (Figure 3B). MiR-125b induction was confirmed by RT-PCR on FACS-isolated DTG ORS cells (Experimental Procedures). Strong and specific induction of miR-125b was observed without significant disturbance of other unrelated miRNAs (Figure 3C).

Figure 3. Sustained miR-125b Blocks Hair Growth.

(A) Schematic of constructs used to generate DTG mice.

(B) miR-125b in situ hybridizations of P4 backskin sections harvested after 24hr Dox induction. In situ signals (black); DAPI (blue). TRE, TRE-miR-125b single transgenic. DTG, double transgenic. Scale bar represents 50µm. Note that high miR-125b levels are no longer confined to HF-SCs, but are now expanded to early progeny.

(C) Taqman RT-PCR analysis of miRNAs in FACS-sorted P4 stage backskin ORS cells. Error bars represent standard error.

(D) Appearance of DTG and WT littermates given Dox at P3 (anagen) or P21 (telogen) and photographed at ages indicated.

(E) Mice were given Dox at P56, and at P58, the upper backskins were waxed and lower parts were shaved. Animals were photographed at ages indicated.

See also Figure S1, Figure S2

Without Dox, the mice developed normally and were morphologically indistinguishable from wild type (WT) littermates (Figure 3D). When Dox was administered to neonatal mice during the growth phase of the first hair coat (P3), hair growth was inhibited, a feature which became obvious 10d later (Figure 3D). When Dox was given during the first resting phase (P21) and then monitored through one entire hair cycle to the second resting phase (P48), the new hair coat never emerged (Figure 3D). As long as miR-125b was sustained, baldness persisted. All three DTG lines behaved analogously, although the progression to baldness varied among them, likely reflective of their different miR-125b expression levels (Figure S1).

Hair growth inhibition was best visualized by administering Dox at P56 during the prolonged telogen phase of the 2^nd postnatal hair cycle, and then waxing backskins 2d later to eliminate existing club hairs, and promote precocious, synchronous initiation of a new hair cycle. In WT mice, the coat regrew in the depilated area by P71, while miR-125b-induced mice still displayed the bald patch (Figure 3E). Intriguingly, while hair growth was markedly impaired in DTG mice, growth of some other skin tissues and appendages was noticeably accentuated. This was particularly true for the lips, nails and anogenital areas in which the K14 promoter is active (Figures S2A, S2B). Moreover, when Dox was administered to DTG mice, the hair coat became greasy prior to its loss. Closer inspection revealed greatly enlarged SGs, whose sebocytes stain with Oil-red-O (Figure 4A).

Figure 4. Enlarged Sebaceous Glands in DTG Mice Traced to miR-125b-Induced Suppression of Blimp1.

(A) Whole mount Oil Red O staining of tail skin of P72 DTG and WT littermates (+Dox a/P56).

(B) Anti-Blimp1 immunofluorescence of backskin sections of P13 mice (+Dox a/P3).

(C) RT-PCR of P79 tail skin mRNAs of P79 mice (+Dox a/P21). GAPDH, reference.

(D) Pairing between predicted miR-125b binding site on Blimp1 3’UTR and mmu-miR-125b. The seed region of miR-125b is highlighted.

(E, F) Relative luciferase activity of WT and Mutated (Mut) Blimp1 3’UTR reporters in cultured MK when co-transfected with Pre-miRNA precursors (E) or antisense LNA Knockdown probes (F) compared with their corresponding controls. The vector control is CMV-PGL4. The effectiveness of miR-125b overexpression by Pre-miR-125b precursor and miR-125b knockdown by Anti-miR-125b LNA was confirmed by the activity of miR-125b sensor.

All error bars represent standard error unless otherwise indicated. All scale bars represent 50µm unless otherwise indicated.

MiR-125b Regulates SG Development By Targeting Blimp1 Through a Conserved 3’UTR Binding Site

The enlarged SGs accompanying miR-125b induction were reminiscent of those in mice conditionally targeted for Blimp1, which encodes a transcriptional repressor normally expressed by K14-promoter-active sebocyte progenitors (Horsley et al., 2006). Indeed, immunofluorescence microscopy revealed a marked reduction in Blimp1 protein following induction of miR-125b (Figure 4B). RT-PCR confirmed that Blimp1 mRNA levels were also strongly downregulated in Dox-induced DTG skin. Analogous to the Blimp1 cKO animals, the sebocyte marker PPARγ was strongly upregulated(Figure 4C).

The 3’UTR of Blimp1 mRNA is predicted by TargetScan (Lewis et al., 2005) to contain a single miR-125b binding site that is well-conserved from amphibian to mammals. To test the hypothesis that Blimp1 mRNA may be a direct target of miR-125b in skin, we constructed a luciferase reporter for WT Blimp1 3’UTR, as well as a control reporter in which the miR-125b binding site of the Blimp1 3’UTR was mutated (Figure 4D). Luciferase reporter assays were then carried out in primary keratinocytes (MK) cultured from P0 epidermis of WT skin and transfected with synthesized miR-125b precursor (Pre-miR-125b) or control Pre-miR.

Elevated miR-125b resulted in strong repression of the WT Blimp1 3’UTR reporter but not a mutant Blimp1 3’UTR reporter lacking the miR-125b binding site (Figure 4E). Furthermore, when miR-125b was repressed by transfecting anti-sense LNA probes against miR-125b into cultured MK, WT Blimp1 3’UTR reporter activity was significantly upregulated in contrast to the control reporter (Figure 4F). Taken together, these data indicate that Blimp1 mRNA is a direct and functionally relevant target of miR-125b in skin keratinocytes.

MiR-125b Represses Hair Differentiation

While the link between miR-125b and Blimp1 was sufficient to account for the SG phenotype, it did not explain the hair growth defect observed in our mice. To decipher how sustained miR-125b expression might induce baldness, we first examined backskin histology from strain-, sex- and age-matched littermates of Dox-induced DTG versus WT (or single transgenic) mice. Surprisingly, near the end of the first anagen, when WT littermate HFs had reached maximal length (P13, Dox-induced since P3), DTG HFs were much longer, exhibiting a highly elongated ORS (Figure 5A). Immunofluorescence microscopy showed proper expression of ORS markers K14 and β4 integrin, but a marked reduction in markers of matrix-derived hair shaft and inner root sheath (IRS) cells (Figure 5B).

Figure 5. Repressed Differentiation and Expansion of Early Bulge ORS Region in DTG HFs.

DTG and WT littermates were treated with Dox a/P3 and backskin sections were either examined at P13 (A, B) unless subjected to further experimentation (C–H). (A) Hematoxylin and eosin staining and quantification of average HF length. (B) Immunofluorescence for ORS, IRS and hair shaft markers.

(C) Schematic of BrdU experiment. Immunofluorescence with Abs indicated (color coding according to secondary Abs). Quantifications: #BrdU+ cells per matrix (section) 4hrs a/BrdU; %BrdU+ cells in pre-cortex (early hair shaft) region and IRS at times indicated a/BrdU.

(D) Immunostaining for activated caspase 3 shows little or no change in DTG vs WT littermate. (E,F) Quantifications of BrdU+ cells in lower vs upper ORS.

(G) Immunofluorescence reveals hyperthickened upper ORS bulge region in DTG HF.

All error bars represent standard error unless otherwise indicated. All scale bars represent 50µm unless otherwise indicated.

The dearth of hair shaft and IRS cells could arise from repressed terminal differentiation, an increase in cell death or an impaired ability of matrix to fuel production of these lineages. To distinguish between these possibilities, we examined matrix size and proliferative potential. When P13 mice were pulsed by a single injection with the nucleotide analog BrdU and sampled after 4 hrs, S-phase cells were primarily in matrix (Figure 5C). Compared to WT, DTG HFs showed a >30% reduction in labeled matrix cells. By 11hrs, many of these labeled cells in WT HFs had moved into the non-dividing precortex, reflecting their normal progression of differentiation. By contrast, an ~3X reduction was seen in labeled DTG precortical cells (Figure 5C). Taken together, these data revealed reductions both in the number of dividing transiently amplifying (TA) cells in the matrix and their rate of differentiation into hair shaft and IRS lineages.

The underlying basis for these lineage defects did not appear to reside in enhanced cell death (apoptosis), as evidenced by the relative lack of activated caspase 3 staining, except for the core of the hair shaft, also positive in WT HFs (Figure 5D). We therefore focused on earlier steps along the lineage, namely in the progression of activated HF-SC progeny along the ORS to matrix. Further analysis of BrdU+ cells revealed a modest reduction in labeled cells within the lower DTG ORS during both 4 hr and 11 hr labeling periods (Figure 5E). Unexpectedly however, the upper ORS, including the bulge region, showed an ~6–7X enhancement in S-phase cells in these HFs (Figure 5F). This was accompanied by hyperthickening of this zone (Figure 5G), a feature not seen in mid and lower ORS (Figure 5B). Taken together with the marked elongation of the mid/lower DTG ORS, these data suggest that sustained miR-125b in the ORS increases the number of cell divisions that early bulge cells and/or their progenitors make, which in turn restricts their commitment and progression along the lineage.

MiR-125b Inhibits the Differentiation of Skin SCs Without Affecting Their Maintenance or Activation

Thus far, our data suggested that stemness may still be maintained when miR-125b cannot be downregulated in HF-SC progenies. To examine this possibility, we first looked at bulge markers Lhx2, Sox9, Tcf4, Nfatc1 and CD34 in Dox-induced DTG skins of adult mice. With the exception of CD34, all these markers were still present 17d after miR-125b induction (Figure 6A, Figure S3). To determine whether DTG skin still possessed functional HF-SCs, we administered Dox at P21, waited until DTG mice were bald at P58, and then either kept mice on Dox or withdrew Dox until P94. For comparative purposes, an area of skin from WT mice was shaved so we could monitor hair regeneration.

Figure 6. Evidence for Expansion of Stem Cell Populations in DTG Skin.

(A) Immunofluorescence reveals presence of most HF-SC markers in DTG skin (Dox a/P17; analyses a/P35).

(B) Hair coat recovery after Dox withdrawal in DTG mice. Mice were given Dox a/P21; a/P58, Dox was either continued or withdrawn as indicated.

(C) HF-SCs are activated comparably in DTG and WT by wax-induced depilation of resting phase follicles. BrdU was administered beginning 1d a/waxing and HFs were analyzed and quantified 2d later.

(D–E) 10⁴ FACS sorted ORS cells from Dox-induced P33 DTG and WT littermates were plated and a/2wk, cultures were either fixed and stained (D) or individual large clones were picked, serially passaged and quantified for long term survival (E). Note: WT bulge cell colonies typically have <10% survival rate in serial passage experiments (Greco et al., 2009), and WT ORS consists of bulge and non-bulge ORS. DTG ORS is highly enriched for holoclones.

(F, G) FACS sorted ORS cells from P59 K14H2BGFP DTG and WT HFs (Dox a/P21) attach comparably 36hr a/plating onto feeder-coated dishes, but colony forming efficiency is higher for DTG cells.

(H) Immunofluorescence reveals expansion of basal cells in IFE of P13 DTG backskin (Dox a/P3).

(I, J) RT-PCR for basal vs differentiation markers. mRNAs were from MKs in either Low Ca²⁺ (0.05nM) media or a/switching to Medium Ca²⁺ (0.3nM) media for 24hrs. (I), DTG MK ± Dox induction as indicated.

(J), WT MK treated w/anti-miR-125b or control scramble antagomirs as indicated.

All error bars represent standard error unless otherwise indicated. All scale bars represent 50µm unless otherwise indicated.

See also Figure S3, Figure S4

The results of the experiment are shown in Figure 6B. As expected, only WT mice grew their hair when Dox was maintained. Interestingly however, by 5 wk after Dox withdrawal, DTG hair coats had largely regenerated. This happened even when mice were kept on Dox for 4 months prior to withdrawal (data not shown). Since the hairless phenotype was reversible, the defect appeared to be a delay in the rate of bulge→hair lineage progression rather than a loss of HF-SCs.

To determine whether sustained miR-125b prevented HF-SC activation, we stimulated resting (telogen-phase) HFs by depilation (waxing) to initiate a new hair cycle, and then examined the response by administering a 48hr pulse of BrdU. Under these conditions, >50% of Sox9+ cells in both WT and Dox-induced DTG HFs showed actively cycling (Figure 6C). Moreover, signs of a Sox9(−) early matrix were also apparent at this time. Thus, even though hair coat formation was impaired in Dox-induced DTG mice (Figure 3E), HF-SCs were still able to respond to a hair growth stimulus.

If miR-125b inhibits hair lineage differentiation without impairing the maintenance and activation of the HF-SCs, then either relatively un-differentiated SC-derived progeny and/or HF-SCs themselves should accumulate in DTG HFs after Dox induction. To test this hypothesis, we FACS-purified and cultured equivalent numbers of ORS cells from HFs of DTG and control mice and compared their ability to form large (>2mm) colonies (holoclones), a characteristic of skin SCs (Barrandon and Green, 1987). Holoclone-forming efficiency was at least 3X higher in ORS cells isolated from induced DTG vs WT HFs (Figure 6D). Moreover, >30% of holoclones formed by DTG ORS cells could be serially passaged >3X in culture, a characteristic of adult bulge HF-SCs but not their progeny (Greco et al., 2009) (Figure 6E). The difference in holoclone numbers was not attributable to differences in cell attachment, but rather to enhanced colony initiation efficiency (where a colony is defined as ≥4 cells) (Figures 6F, 6G).

The increased ability of DTG ORS to generate holoclones could either reflect an expanded HF-SC pool or alternatively, an increase in the proliferation rate of early bulge progeny. To distinguish between these possibilities, we overexpressed miR-125b in cultured WT MK, which resulted in a significant decline in both overall proliferation rate and the percentage of S phase cells (Figure S4). Thus while the DTG ORS contained a higher percentage of cells with holoclone forming and passaging ability, elevated miR-125b in vitro made cells grow more slowly. Together, these findings suggest that miR-125b’s function resides in enhancing SC self-renewal rather than promoting proliferation in general.

The effects of miR-125b on SC behavior extended to IFE progenitors. In vivo, the IFE of Dox-induced DTG skin was hyperthickened and signs of basal cell expansion were evident (Figure 6H). In vitro, cultured MK from DTG versus WT epidermis displayed a significantly dampened response to calcium-induced terminal differentiation when exposed to Dox (Figure 6I). Conversely, when exposed to an antagomIR specific for miR-125b, WT epidermal cells displayed an accentuated response to calcium-induced differentiation (Figure 6J).

Identification of Additional miR-125b Target Genes

miRNAs often downregulate the mRNA level of their target genes (Baek et al., 2008; Lim et al., 2005; Selbach et al., 2008). Therefore, to gain further mechanistic insights underlying miR-125b’s effects, we conducted microarray analysis on ORS cells from P4 WT and DTG mice that had been treated with Dox 1d earlier. We conducted our analyses on independent, duplicate matched sets of male and female mice.

miRNAs only modestly regulate their target mRNA levels (Baek et al., 2008; Selbach et al., 2008) and our aim was to identify early changes occurring shortly after miR-125b induction. Hence, we chose a relatively modest cutoff, Log2(DTG/Control Fold of change) <−0.3, for scoring a gene as being downregulated in each array. Notably however, under this criterion, TargetScan-predicted, conserved miR-125b targets were prominent within the downregulated genes of both datasets (Figure S5). Moreover, the DIANA-mirExTra algorithm (Alexiou et al.) identified miR-125b’s seed sequence “CUCAGG” as the most overrepresented, conserved hexamer motif within 3’UTRs of these downregulated genes. Thus our strategy to detect miR-125b targets was yielding meaningful specificity and sensitivity.

Since sustained miR-125b resulted in analogous HF phenotypes in male and female mice, we reasoned that the majority of bona fide miR-125b targets in ORS cells should be downregulated in both sexes. Consistent with this idea, ~12% of genes scoring as downregulated in both sexes upon miR-125b induction displayed conserved miR-125b binding sites in their 3’ UTRs (Figure 7A). This represented an ~6X enrichment compared to that of all ORS-expressed genes and an ~3X enrichment compared to that of genes downregulated in either sex alone. This indicated that the specificity for miR-125b target detection would be greatly improved by focusing on the overlap in downregulated genes in our two datasets. Hereafter, we refer to these genes as “miR-125b downregulated genes” (Table S2).

Figure 7. Identifying and Testing Putative miR-125 Targets in Skin SCs.

(A) mRNAs from DTG vs WT ORS from P4 male and female mice were subjected to microarray analyses to identify a list of mRNAs downregulated by miR-125b (see text). Shown are the %predicted targets for miR-125b and let-7 among the list.

(B) mRNAs from P4 WT early bulge and non-bulge ORS were subjected to microarray analyses. Shown is an accumulative percentile map for the distribution of the miR-125b downregulated genes from (A) within these new array data.

(C) Comparisons reveal that miR-125b downregulated genes tend to be downregulated in early bulge vs non-bulge ORS.

(D) Luciferase reporter assays validate the predicted conserved miR-125b binding sites in the 3’UTRs of 3 miR-125b downregulated ORS genes. Mut indicates constructs with mutated miR-125b binding sites.

(E) VDR protein is elevated in lower WT ORS, but reduced in DTG ORS.

(F) RT-PCR of Vdr and VDR targets in P79 tail skins (Dox a/P21). GAPDH, reference gene.

(G, H) Accumulative percentile maps for the distribution of VDR targets in P4 ORS microarray results. VDR targets tend to be downregulated a/24hr Dox induced miR-125b expression in DTG ORS.

(I,J) Relative activities of VDR-responsive element (VDRE) luciferase reporters in (I) DTG MK ± EB1089 ligand (Lig) (10⁻⁷M) ± Dox induced miR-125b or (J) WT MK knocked down by anti-miR-125b or scrambled antagomirs (± Lig).

(K) Partial rescue of hair coat defect in DTG mice treated with VDR ligand. DTG and K14-rtTA littermates were treated with Dox a/P48, waxed a/P50 and either injected with PBS or ligand (5 ng/g body weight 3X/wk). Backskins were photographed a/P67.

All error bars represent standard error unless otherwise indicated. All scale bars represent 50µm unless otherwise indicated.

See also Figure S5, Figure S6, Table S2, Table S3.

Expression of a miRNA often inversely correlates with that of its targets (Farh et al., 2005; Stark et al., 2005). Since miR-125b is naturally enriched in early bulge but strongly downregulated in non-bulge ORS, we posited that physiologically relevant miR-125b target mRNAs should be expressed at reduced levels in early bulge compared with non-bulge ORS. To explore this possibility, we performed microarray analyses on mRNAs isolated from our FACS-purified early bulge and non-bulge ORS populations of P4 stage backskin HFs. Interestingly, when WT early bulge transcripts were compared with those in ORS, mRNAs that scored as being preferentially downregulated in WT early bulge encompassed many mRNAs that were also downregulated in miR-125b-induced DTG ORS (Figure 7B). These data suggest that a significant portion of miR-125b-downregulated genes in the DTG ORS are endogenous miR-125b targets in WT early bulge cells.

Thus far, our data revealed that miR-125b downregulated genes were significantly enriched for both Targetscan-predicted miR-125b targets and also downregulated genes in WT early bulge vs non-bulge ORS. Remarkably, the highest enrichment (>700%) was observed for predicted conserved miR-125b targets that were also downregulated naturally in miR-125b+ early bulge vs miR-125b(−) non-bulge ORS (Figure 7C). To derive a list of high-confidence potential miR-125b targets, we overlapped these criteria and identified those miR-125b downregulated genes that also a) show downregulation in WT early bulge compared to non-bulge ORS cells and b) possess at least one Targetscan-predicted conserved miR-125b binding site in the 3’UTR. By applying these criteria, we generated a short-list of 33 genes as putative miR-125b targets in the early bulge region (Table S3). From this list, we picked three potentially interestingly targets for further verification: Vdr encodes VDR (vitamin D receptor), involved in HF differentiation; Trp53Inp1 encodes a positive regulator of p53, a known miR-125b target; and Scarb1 harbors two conserved miR-125b binding sites in the 3’UTR. We validated their miR-125b binding sites by luciferase reporter assays (Figure 7D).

MiR-125b Regulates Hair Differentiation by Repressing VDR Expression

We focused on Vdr since it is required for HF differentiation (Palmer et al., 2008) and its loss results in a hairless phenotype (Li et al., 1997). Our microarray data revealed higher levels of Vdr (~1.7X) in non-bulge ORS than in early bulge ORS, and this was confirmed by immunofluorescence microscopy (Figure 7E). In DTG ORS, VDR expression was markedly diminished, and transcripts of genes previously identified as putative HF targets for the VDR transcription factor (Palmer et al., 2008) were significantly downregulated both in DTG tail skin and in purified ORS cells (Figures 7E–7H).

We confirmed the negative regulation of miR-125b on the activity of VDR pathway by testing a VDR-responsive element (VDRE) luciferase reporter construct in cultured DTG MK. When miR-125b expression was induced by Dox, the expression of the VDRE reporter in response to the Vitamin D Analog EB1089 (Palmer et al., 2008) was dramatically reduced (Figure 7I). Conversely, in cultured WT MK, anti-miR-125b antagomir treatment significantly enhanced VDRE reporter expression in response to EB1089 (Figure 7J).

Finally, if the downregulation of VDR pathway activity contributes to the miR-125b-dependent inhibition of hair differentiation, stimulation of the pathway by ectopic VDR ligand should dampen the phenotype. Consistent with this hypothesis, IP injection of EB1089 partially rescued the hair growth defect in DTG mice during depilation (waxing) initiated hair cycle (Figure 7K). Similar rescue effects were achieved via topical application of EB1089 solution (Figure S6).

Discussion

A miRNA Signature of “Stemness” in the Hair Follicle

In the present study, we identified a miRNA expression signature which specifically associates with stemness in HFs, and at least for miR-125b, this signature is functionally relevant to tissue homeostasis. It is tempting to speculate that by coordinately fine-tuning the expression of a large number of proteins, miRNAs may confer greater accuracy than transcription factors in acting as a rheostat to regulate decisions involving the switch from stemness and commitment (Stark et al., 2005).

In the hair lineage, quiescent bulge SCs are known to be activated by external Wnt promoting and BMP inhibitory signals. As progeny move downward along the ORS, they are thought to progressively adjust their fate from that of SCs to committed progenitors. Eventually, these cells will pass a point of no return and end up as TA matrix (Hsu et al., 2011). The dramatic switch in miRNA expression profile between early bulge and non-bulge ORS and the similarities between ORS and matrix miRNAs suggest that miRNAs may function in regulating fate progression through this transition point. In this regard, it was particularly interesting that miR-125b exhibited a sharp downregulation early in the lineage between bulge and adjacent ORS. Moreover, when this switch was blocked by enforcing miR-125b expression throughout the ORS, a gross distortion arose in the relative ratios of cells with stemlike characteristics and those further along the hair lineage. Notably, miR-125b did not seem to affect the maintenance, self-renewing or activation of HF-SCs.

Within the DTG HF lineage, the upper ORS region was hyperthickened, while mid/lower ORS was merely elongated and the progression to a TA cell fate was actually delayed. Given the absence of defects in bulge SC maintenance or activation, our findings best suit a model whereby miR-125b acts as a rheostat whose downregulation controls the rate at which bulge progenies transit through the commitment step along the lineage. This view is further supported by the dramatic miR-125b-dependent increase in holoclone-forming cells within the ORS (including bulge). Importantly, miR-125b does not promote proliferation per se, as when miR-125b is overexpressed in vitro, cells grow at a significantly slower rate.

When our in vivo and in vitro data are taken together, they point to a role for high miR-125b as a faithful definer of self-renewing, slow-cycling SCs, while reducing miR-125b was required to transition to a state of rapid proliferation, fate commitment and differentiation. Our results help to explain a myriad of seemingly paradoxical findings about miR-125b, where in some cell types, it has been reported to be an oncomir and in others, a tumor suppressor. Our studies show how miR-125b can suppress overall proliferation while still acting to expand stem cell numbers and tissue/organ size.

miR-125: a Global Balancer of Stemness and Commitment in Adult Tissues?

Another key finding was that miR-125b’s rheostat function in governing the switch between stemness and commitment was not limited to the HF lineage. In the normal IFE, which also modestly expresses miR-125b in the basal layer, miR-125b elevation resulted in an expansion in the number of layers possessing features of undifferentiated cells in vivo. Moreover in cultured epidermal MK, calcium-induced differentiation was enhanced when miR-125b was knocked down and repressed when miR-125b was elevated. Finally, since Blimp1 is known to limit SG progenitors by repressing c-Myc (Horsley et al., 2006), miR125b’s ability to repress Blimp1 and yield enlarged SGs was likely rooted in an expansion of its progenitors.

In B cells, Blimp1 has been reported to induce differentiation (Turner et al., 1994). Moreover, in vitro studies recently showed that overexpression of miR-125b leads to Blimp1 repression and inhibition of B cell differentiation (Gururajan et al., 2010; Malumbres et al., 2009). One model that could explain these seemingly paradoxical effects of Blimp1 in the SG versus B cell lineages is that by downregulating Blimp1 in SG progenitors, miR125b makes the cells go through extra rounds of proliferation before they can progress into a more differentiated cell fate. This would effectively delay differentiation and expand the pool of less differentiated SG cells.

This same logic can potentially be applied to other lineages. When the differentiation process is delayed and/or repressed by miR-125b, actively cycling undifferentiated progenitors will accumulate, generating an enlarged tissue/organ, as seen for the lips, anogenital epithelium, nails, ORS and epidermis in our miR-125b-induced mice. This would also explain why in epidermis, ORS and SG, the most dramatic changes in proliferation seen in Dox-induced DTG skin occurred at the juncture between SCs and their immediate progeny. Notably, this is also the site of dramatic downregulation of miR-125b in normal skin.

Overall, our studies implicate miR-125b as a governor of the relative numbers of divisions that a progenitor undergoes prior to committing to a lineage. Given that controlling this property is likely to be of fundamental importance for maintaining the correct size of most organs/tissues, it is intriguing that miR-125a, a related family member, was recently found to be a positive regulator of SC number in the hematopoietic lineage (Guo et al., 2010). Taken together, our findings suggest that the differentiation rheostat role for miR-125b that we’ve identified here may be a general feature of somatic SCs.

MiR-125g’s Target Genes

Our analysis of the global gene expression changes that occur when miR-125b is sustained in HF ORS cells represents the first effort to systematically identify the downstream targets of miR-125b in a postnatal somatic tissue. Our data identified ~800 transcripts that were downregulated within 24 hr of miR-125b induction in both male and female mouse ORS (Table S2), and importantly they are highly enriched for predicted conserved miR-125b targets that we also found to be downregulated in WT early bulge versus their progenies. These findings suggest that a considerable portion of these putative miR-125b targets identified in our study are likely to be physiologically relevant.

In the current study, we focused on two miR-125b targets, Blimp1 and Vdr. While their altered expression accounts for many of the miR-125b-induced defects we observed, some perturbations remain unaccounted for by these two miR-125b targets. Notable examples are the remarkable lengthening of HFs and enlargement of lips in DTG mice, which were not observed in either Blimp1 or VDR knockout animals. By contrast, other phenotypes e.g. the enlarged SGs and blockage of hair growth were more severe in miR-125b induced mice than in Blimp1 or Vdr knockout animals. Moreover, since Blimp1 repression in vivo (Figure 4C) was appreciably greater than that displayed by reporter constructs in vitro (Figure 4E), indirect repression of Blimp1 transcription might also result from miR-125b induction. When taken together with the fact that excess VDR ligand only partially rescued miR-125b-induced hair loss, it seems likely that additional miR-125b target genes unearthed in our study but as yet untested will be necessary to fully account for the complexities of the DTG phenotype.

In closing, it is notable that like Blimp1, Vdr has also been reported to be a miR-125 target in a breast carcinoma line (Mohri et al., 2009), raising the possibility that there may be a common core of key miR-125 targets that are coordinately repressed in progenitors and whose activation results in proliferation and commitment. Of the handful of miR-125b targets described to date in other cell types, it is intriguing that Bmf, a pro-apoptotic miR-125b target in human glioma cells, also appeared on our shortlist (Xia, 2009 #1380), as did the p53 regulator Trp53inp1. While p53 itself is known to be a miR-125b target in several systems (Le et al., 2009; Xia et al., 2009; Zhang et al., 2009; Zhou et al., 2010), this merits further investigation in skin, as it did not surface in our list of putative targets (Table S3). As future studies are conducted, it will be interesting to evaluate the extent to which miR-125’s targets will be tailored to suit the general vs specialized needs of SCs in adult tissues.

Experimental Procedures

Mouse Strains and Genetics

Transgenic mice expressing EGFP from a 260Kb bacterial artificial chromosome (BAC) surrounding the Sox9 locus were obtained from the GENSAT project (GENSTAT, 2008; Gong et al., 2003). The BAC contained 75Kb genomic DNA upstream of the Sox9 coding region, which had been replaced with the EGFP coding sequence. To create K14-RFP/Sox9-EGFP double transgenic mice, Sox9-EGFP mice were bred to mice regulating RFP expression under the control of the K14 promoter (Vasioukhin et al., 1999). To generate TRE-miR-125b, the mmu-miR-125b coding sequence was amplified by PCR from mouse genomic DNA (Primers 5’-TGCAAAGCTTAGACAGAAATGAATCTAAATTTTGTG-3’ and 5’-TGCATCTAGACTCAAGGGTGTATTACCATCACTTC-3’), cloned into HindIII/Xba1 sites of the pTRE2 plasmid (Clontech) and sequenced. Transgenic mice were generated as described (Vasioukhin et al., 1999).

MicroRNA Microarray and Data Analysis

For both miRNA and mRNA microarrays, total RNA samples were extracted from FACS-sorted cells using the miRNeasy Mini Kit (Qiagen). For mRNA microarray, RNA samples were submitted to the Genomics Core Laboratory of the Memorial Sloan Kettering Cancer Center for cRNA and amplification, and analyzed on Affymetrix mouse 430_2 arrays. For microRNA arrays, samples were submitted to Exiqon. In brief, samples were labeled using the miRCURY™ Hy3™/Hy5™ power labeling kit and hybridized on the miRCURY™ LNA Array (v.11.0) (Exiqon). A sample of Early Bulge, ORS, Matrix or Basal cells and a reference sample of the corresponding RFP All Cells were differentially labeled and co-hybridized on the same array. Dye switches were included. The normalized signal ratio between sample and reference was used to represent the relative expression level of each miRNA in each sample. Unsupervised Hierarchical clustering analysis was conducted using the Cluster software (Eisen et al., 1998). Low expression miRNAs (Raw signal intensity ≤100) were excluded from analyses. miRNA family data were downloaded from the TargetScan website (Lewis et al., 2005; Lewis et al., 2003).