|Home | About | Journals | Submit | Contact Us | Français|
The discovery that microRNAs (miRNAs) play important roles in regulating gene expression via post-transcriptional repression has revealed a previously unsuspected mechanism controlling development and progenitor-cell function (reviewed in [1, 2]); however, little is known of miRNA functions in mammalian organogenesis. Processing of miRNAs and their assembly into the RNA-induced silencing (RISC) complex requires the essential multifunctional enzyme Dicer . We found that Dicer mRNA and multiple miRNAs are expressed in mouse skin, suggesting roles in skin- and hair-follicle biology. In newborn mice carrying an epidermal-specific Dicer deletion, hair follicles were stunted and hypoproliferative. Hair-shaft and inner-root-sheath differentiation was initiated, but the mutant hair follicles were misoriented and expression of the key signaling molecules Shh and Notch1 was lost by postnatal day 7. At this stage, hair-follicle dermal papillae were observed to evaginate, forming highly unusual structures within the basal epidermis. Normal hair shafts were not produced in the Dicer mutant, and the follicles lacked stem cell markers and degenerated. In contrast to decreased follicular proliferation, the epidermis became hyperproliferative. These results reveal critical roles for Dicer in the skin and implicate miRNAs in key aspects of epidermal and hair-follicle development and function.
Hair follicles develop during embryogenesis via interactions between the surface ectoderm and underlying dermal cells, and they are first apparent as a regular array of ectodermal thickenings, or placodes (reviewed in ). The descendants of placode cells form a lineage separate from that of adjacent surface epithelium, which develops into stratified epidermis . Dermal cells are induced to cluster by placodal signals and form the hair-follicle dermal papilla, a signaling center that remains intimately associated with follicular epithelium throughout lifelong cycles of hair-follicle growth and regression. Signals from the dermal papilla to surrounding undifferentiated epithelial matrix cells cause their proliferation and differentiation into the hair shaft and inner root sheath . Conversely, maintenance of dermalpapilla function requires hair-follicle epithelial signals . Cyclical growth of the hair follicle is dependent on an epithelial stem cell population located in the follicle's permanent bulge region, which is established soon after birth . Transient proliferation of stem cells at the onset of hair growth phases renews the rapidly proliferating matrix-progenitor-cell population [7, 8]. Hair-follicle stem cells do not contribute to the epidermis under normal homeostatic conditions, indicating that the epidermis and hair follicles remain as separate compartments in adult life [4, 9].
Bioinformatics analyses predict that conserved vertebrate miRNAs target more than 400 regulatory genes , suggesting that they play broad roles in biology. To date, the functions of only a handful of vertebrate miRNAs are understood (referenced in ). Mutations in Dicer, which globally disrupt miRNA processing, cause diverse developmental defects [10-12]. Loss of function of mouse Dicer results in early embryonic lethality . Inducible deletion of Dicer in mouse embryonic stem (ES) cells causes defects in their proliferation, suggesting that, among other functions, Dicer may be essential for expansion of stem cells in the gastrulating embryo ; however, limited information is available regarding Dicer's functions at later stages of mammalian development [10, 15–19]. Because the epidermis and hair follicles are accessible, easy to manipulate genetically, and well characterized with respect to their morphogenesis and stem cell populations, we chose the skin as a model system to investigate the functions of Dicer in mammalian organogenesis.
To begin to determine whether Dicer plays a role in epidermal or hair-follicle development, we first asked whether Dicer is expressed in embryonic and postnatal skin and hair follicles. Whole-mount in situ hybridization of mouse embryos at embryonic day (E) 14.5 revealed a strong signal for Dicer in developing hair-follicle placodes, with less intense signal in the surrounding epidermis (Figure 1A). Section in situ hybridization performed at postnatal day (P) 7 revealed prominent expression of Dicer in the epidermis and hair-follicle outer root sheath (Figure 1B).
To determine whether miRNAs are expressed in developing mouse skin and hair follicles, we carried out microarray analyses of miRNA expression at birth, a time point at which new hair follicles are still developing and primary hair follicles are beginning to undergo terminal differentiation. miRNA expression profiles were compared in skin from control newborn mice and skin from littermate mice that had been engineered to ectopically express the potent secreted WNT inhibitor Dickkopf 1 (DKK1) in the epidermis, resulting in complete absence of hair follicles [20, 21] (Table 1; see also Tables S1 and S2 in the Supplemental Data available online). The ten miRNAs giving the strongest hybridization signals in control newborn skin are listed in Table 1A. Several miRNAs were identified for which hybridization signals were on average more than 2.5-fold higher in control samples than in Dkk1-expressing samples (Table 1B), suggesting that these may be upregulated in hair follicles and/or are direct or indirect targets of WNT inhibition in the skin. Expression of mmu-mir-200b, mmu-mir-196a, mmu-mir-29b, mmu-mir-27b, and mmu-mir-203 in control newborn skin, as well as decreased expression of mmu-mir-200b and mmu-mir-196a in Dkk1-expressing compared with control newborn skin, was confirmed by semiquantitative RT-PCR with primers that selectively amplify mature miRNAs (Figures 1C and 1H). Expression of all of these miRNAs and an additional miRNA, mmu-mir-24, was also detected in the skin at subsequent postnatal stages (Figures 1C, 1G, and 1H). Relative levels of mmu-mir-200b and mmu-mir-196a varied with the hair growth cycle, consistent with possible predominant expression of these miRNAs in hair follicles.
To determine whether Dicer is required for development of the hair follicles or epidermis, we generated an epidermal-specific deletion of the Dicer gene. ES cells carrying a conditional allele of Dicer (Dicerflox) in which essential exons 22 and 23 encoding the majority of both RNase III domains are flanked by loxP sites  were used to generate Dicerflox mice. Cre-mediated recombination of the loxP sites in Dicerflox yields a non-functional allele . Mice homozygous for the floxed allele were viable and fertile with no apparent pheno-type. Dicerflox mice were crossed to a transgenic mouse line in which Cre recombinase is expressed under the control of a keratin 14 (K14) promoter, resulting in efficient deletion of target genes in the surface ectoderm, basal epidermis, and hair follicles by E14.5, the stage at which development of primary hair follicles is initiated .
K14-Cre; Dicerflox/flox mice were identified by geno-typing of tail-biopsy DNA. Recombination of the floxed allele was confirmed by PCR analysis of DNA extracted from isolated newborn K14-Cre; Dicerflox/flox epidermis (Figures 1D and 1E). Western-blot analysis demonstrated that Dicer protein was absent from the epidermis in newborn Dicer mutant mice and was substantially reduced in P63 full-thickness skin (Figure 1F). Low levels of protein detected in mutant skin at this stage could be due to expression of Dicer in the dermis, or to less efficient Dicer deletion in long-surviving mutants. To determine whether lack of Dicer protein impacted mature-miRNA production, levels of selected skin miRNAs identified as described above were compared in Dicer mutant and control skin by northern-blot analysis and semiquantitative RT-PCR. Northern-blot analysis of mmu-mi-200b detected its precursors in control and Dicer mutant epidermis at P1 and P6; however, the mature miRNA was present in control but absent from Dicer mutant epidermis at both stages (Figure 1G). Semiquantitative RT-PCR analyses revealed reduced or absent production of multiple mature miRNAs, including mmu-mi-200b, in isolated skin epithelium at P6 (Figure 1H). Additionally, expression of mmu-mi-200b and mmu-mir-203 was reduced or absent in full-thickness mutant skin at P1 and P63 (Figure 1H). Expression of mmu-mir-27b was detected at reduced levels in full-thickness mutant compared with control skin at P1 and P63 (Figure 1H). Persistence of mmu-mir-27b in full-thickness mutant skin at these stages might reflect inefficient Dicer deletion or dermal expression of this miRNA.
Newborn Dicer mutant mice were grossly indistinguishable from control littermates, but by P7 were stunted and lacked external hair growth (Figure 1I). Viability of the mutants was poor, with severely affected mutants dying within a few days of birth and less affected mice, including a mouse with a mosaic phenotype (Figure 1J), surviving for up to 2.5 months.
Analysis of whole-mounted Dicer mutant skin at P6 revealed absence of external hair shafts and apparent evaginations of the epidermis (Figures 1K–1M). Hair follicles viewed from the dermal aspect of the skin were misangled and failed to display the normal anterior-posterior polarity seen in control skin (Figure 1M). Histology of Dicer mutant newborn skin revealed that hair-follicle growth was stunted compared to Dicerflox/flox or K14-Cre; Dicerflox/+ littermate controls (Figures 1N–1O′). By P7, Dicer mutant hair follicles were misangled and wavy, weaving in and out of the plane of section (Figures 1P-1Q′). Hair-shaft structures were present but underdeveloped, and hair shafts did not extend beyond the level of the epidermis. The hair bulbs, which are usually populated by proliferating matrix cells, were much smaller than those seen in control littermate skin. Secondary, later-developing hair follicles failed to extend into the dermis (Figures 1R and 1S). A highly unusual feature of the skin at this stage was the appearance of evaginating dermal cells that became engulfed by epidermal cells (Figure 1T). The Dicer mutant epidermis was expanded compared to normal at P7 (Figures 1P–1T). By P49, control hair follicles were in the resting, telogen stage of the hair-growth cycle and appeared as regularly spaced uniformly oriented structures (Figures 1U and 1U′). In contrast, hair follicles had degenerated in large stretches of Dicer mutant skin and were replaced by cyst structures or disorganized clumps of epithelial cells within the dermis (Figures 1V and 1V′). The epidermis remained expanded at this stage (Figures 1U–1V′).
To begin to determine the basis for the defects in hair-follicle morphogenesis and maintenance observed in Dicer mutant skin, we analyzed proliferation and expression of markers and regulators of hair-follicle and epidermal differentiation in newborn Dicer mutants and littermate controls. Proliferation, assayed by either Ki67 staining or the presence of phosphohistone H3, was significantly reduced in mutant hair follicles compared with controls (1.42 ± 0.48 phosphohistone-H3-positive cells per hair follicle in control mice versus 0.71 ± 0.39 in the mutants; p = 0.003) (Figures 2A–2D). Epidermal proliferation at this stage was not significantly different in Dicer mutant and control skin (0.58 ± 0.67 phosphohistone-H3-positive cells per field at 20× magnification in the control versus 0.33 ± 0.49 in the mutant) (Figures 2C and 2D). Consistent with decreased hair-follicle proliferation, levels of phosphohistone H3 assayed by western blot of separated epidermis (containing hair-follicle epithelia) were decreased in the mutant (Figure 2M). Levels of SOX9 protein, a marker of the hair-follicle outer root sheath, were also decreased, reflecting the smaller average size of mutant follicles (Figure 2M). Epidermal stratification, judged by expression of keratins 1 and 10, was similar in Dicer mutant and control skin at this stage (Figures 2A and 2B and data not shown). Consistent with these data, transmission electron microscopy (TEM) of newborn Dicer mutant and control skin revealed that basal and suprabasal layers and stratum corneum were formed in the mutant. TEM also revealed the presence of occasional apoptotic cells in the basal layer of the mutant epidermis (Figure 2L) and in some mutant hair follicles. Quantitation of apoptotic cells revealed by TUNEL staining revealed that increased apoptosis in the developing hair-follicle bulbs did not reach the level of statistical significance (0.06 ± 0.04 TUNEL-positive cells per hair follicle in the control versus 0.12 ± 0.13 in the mutant; p = 0.134). However, levels of cleaved caspase 3, analyzed by immunoblotting, were markedly increased in mutant epithelial preparations (Figure 2M).
Sonic hedgehog (Shh) plays a key role in regulating follicular proliferation and down growth [23, 24]. In control newborn skin, Shh was expressed in discrete populations of matrix cells on the anterior aspects of hair follicles. Shh was expressed at similar levels in Dicer mutant newborn hair follicles (Figures 2G and 2H), indicating that decreased follicular proliferation is not a consequence of aberrant Shh expression. However, expression of Shh in mutant follicles appeared less polarized than in controls. Notch1 regulates maintenance of inner-root-sheath precursor cells and epidermal proliferation [25, 26]. Notch1 was expressed similarly in control and Dicer mutant newborn epidermis and hair follicles (Figures 2I and 2J). The signaling molecules IKKα and PI3 kinase were also expressed at similar levels in P1 control and mutant epidermis, as was RAC1, a Rho guanosine triphosphatase required for maintenance of epidermal stem cells  (Figure 2M).
Keratin 15 is a specific marker for hair-follicle stem cells after approximately P20, but in newborn skin it is expressed broadly in the basal layer of the epidermis [28, 29]. Expression of K15 was reduced in Dicer mutant newborn skin (Figures 2E and 2F). Although the significance of epidermal K15 expression is not known, expression of other keratins as well as epidermal Notch1 in Dicer-deficient newborn epidermis suggests that Dicer is specifically required at this stage either for K15 expression or for the survival of a subpopulation of epidermal cells that express K15. Consistent with decreased K15 expression, quantitative RT-PCR assays for CD34 mRNA, another marker for hair-follicle stem cells, demonstrated reproducible reduction in P1 Dicer mutant skin to 40% ± 8% of the levels detected in litter-mate control skin.
Two-dimensional gel electrophoresis of epidermal proteins from control and Dicer mutant littermates at P1 revealed an overall similar pattern of protein expression in the mutant, consistent with specific effects of Dicer deletion on particular cell types or pathways, rather than broad alterations in gene expression (Figure S1).
By P7, the Dicer mutant epidermis displayed a marked elevation in the numbers of both basal and suprabasal cell layers compared with control littermate epidermis, indicated by immunofluorescence for K14 and K1, respectively (Figures 3A–3D). The hair-follicle outer root sheaths, revealed by immunofluorescence for K14, K17, and SOX9 [22, 30, 31], appeared to differentiate relatively normally in the Dicer mutant (Figures 3A and 3B and data not shown). Differentiation of the hair-shaft cortex and cuticle, assayed by immunofluorescence for S100A3 and AE13 [32, 33] and the inner root sheath, assayed by fluorescence for GATA3  (Figures 3E, 3F, 3S, and 3T and data not shown), occurred in the Dicer mutant hair follicles, and these cell layers developed in a normal concentric pattern within the follicle. These data indicate that Dicer function is dispensable for the differentiation of hair-follicle matrix cells. However, the number of cells positive for each of these markers was substantially reduced in the mutant hair follicles (Figures 3E and 3F and data not shown). Proliferation in the bulb regions of P7 Dicer mutant hair follicles, assayed by BrdU incorporation and immunofluorescence for phosphohistone H3, was significantly reduced compared with controls (10.2 ± 1.3 BrdU-positive cells per hair follicle in control skin versus 6.8 ± 1.4 in Dicer mutants [p = 0.0078]; 3.3 ± 0.7 phosphohistone-H3-positive cells per hair follicle in control skin versus 0.41 ± 0.18 in Dicer mutants [p = 0.0005]) (Figures 3O and 3P and data not shown). Apoptosis was significantly increased in mutant hair-follicle bulbs by P7 (0 ± 0 TUNEL-positive control versus 1.2 ± 1.5 TUNEL-positive Dicer mutant hair-bulb cells per field in control skin at 10× magnification; p = 0.03). Thus the decrease in numbers of differentiating cells is most likely accounted for by failure of proliferation and survival of matrix cells in the hair bulb. Expression of Shh and its target and effector gene Gli1 disappeared from Dicer mutant hair follicles by P7 (Figures 3G–3J). Similarly, Notch1 expression was substantially reduced in Dicer mutant hair follicles (Figures 3K and 3L). Expression of cellular retinoic acid binding protein 1 (Crabp1), which normally shows specific expression in the dermal papilla , was absent from Dicer mutant follicles, indicating that loss of epithelial Dicer compromises epithelial signaling to the dermal papilla (Figures 3M and 3N). Whereas reduced Shh and Notch signaling could account for the observed defects in follicular proliferation and developmental arrest, Dicer function could be required either to maintain expression of these genes or to maintain the expansion or survival of the cells that express Shh and Notch1.
In contrast to decreased hair-follicle proliferation, epidermal proliferation, assayed by phosphohistone H3 immunofluorescence or BrdU incorporation, was elevated at P7 (0.5 ± 0.8 control versus 5 ± 2.4 Dicer mutant epidermal cells positive for phosphohistone H3 per field at 20× magnification; p = 0.008) (Figures 3O–3P′). Consistent with epidermal hyperproliferation, expression of K17  was markedly increased in Dicer mutant epidermis (Figures 3Q, 3R, 3W′, and 3X′). Apoptosis was significantly elevated in Dicer mutant compared with control epidermis at P7 (0.2 ± 0.6 control versus 1.3 ± 0.95 Dicer mutant TUNEL-positive epidermal cells per field at 10× magnification; p = 0.0067) (Figures 3Q and 3R). Similar to the situation in the newborn, expression of K15 protein was reduced in Dicer mutant compared with control skin at P3, P6, and P7 (not shown). Interestingly, expression of Notch1 was reduced in Dicer mutant epidermis as well as in the hair follicles (Figures 3K and 3L). Deletion of Notch1 in the epidermis causes hyperproliferation and tumor development , suggesting that the observed decrease in Notch1 expression in the Dicer mutant could contribute to the epidermal phenotype.
A unique feature of Dicer mutant skin at P7 was the appearance of clusters of dermal cells apparently in the process of being surrounded by epidermal cells (Figures 3V, 3X, 3X′, Z, and Z′). Consistent with engulfment of dermal cells by the epidermis, nests of cells that failed to express K14 or K17 were detected within the expanded basal epidermis (Figures 3X and 3X′). Occasionally these cells could be detected apparently being extruded from the epidermis. Because K14-Cre does not cause recombination of the Dicerflox allele in dermal cells (Figure 1E), this phenotype is due to Dicer deficiency in the epidermis or hair follicle epithelium.
The abnormal, engulfed dermal cells were tightly clustered within the epidermis and were positive for the dermal-papilla marker alkaline phosphatase (Figures 3Y–3Z′), suggesting that they could comprise dermal papillae from later developing hair follicles that migrate outwards, rather than into the dermis in association with hair-follicle epithelium. Alternatively, these dermalpapilla-like cell clusters could be induced by abnormal signaling from the epidermis. Epithelial cells surrounding the dermal structures expressed SOX9, a marker for the hair-follicle outer root sheath (Figures 3U and 3V), suggesting either a partial fate change in the epidermal cells or migration of outer-root-sheath cells into the epidermis in association with the condensed dermal structures. Evagination of tooth and whisker-follicle epithelium, which would normally invaginate into the dermis, occurs in embryos lacking IKKα via an unknown signaling pathway that does not involve NF-κB . However, IKKα protein levels were unaffected or slightly elevated in Dicer mutant skin at P1 and P6 (Figure 2M and data not shown).
Dicer mutant epidermis analyzed at P49 or P63 was thickened, with increased numbers of both basal and suprabasal layers, indicated by immunofluorescence for K14 and K1 (Figures 4A and 4B and data not shown). Involucrin was expressed in mutant epidermis, indicating that terminal differentiation of the epidermis was relatively normal (Figures 4C and 4D). The epidermis was hyperproliferative, evident histologically (Figure 1V′) and by assaying for BrdU incorporation (1 ± 1 control versus 4.5 ± 3.4 Dicer mutant epidermal cells positive for BrdU per field at 20× magnification at P63; p = 0.021) (Figures 4E and 4F). TUNEL staining revealed apoptotic cells in Dicer mutant skin, both in the epidermis and associated with hair-follicle remnants (Figures 4G and 4H), but the overall levels of epidermal apoptosis at P63 were not statistically significantly different than those observed in control skin (0.33 ± 0.82 control versus 0.83 ± 1.17 Dicer mutant TUNEL-positive epidermal cells per field at 10× magnification; p = 0.076). Nuclei positive for the outer-root-sheath marker SOX9 were detected in hair-follicle epithelial remnants as well in some regions of epidermis (Figures 4I and 4J). However, immunofluorescence for the hair-follicle epithelial stem cell marker K15 [28, 29] and the hair-follicle bulge markers S100A4 and S100A6 [37, 38] was absent from hair-follicle remnants and cysts (Figures 4K–4P). Similarly, CD34 mRNA expression, assayed by quantitative RT-PCR, was reduced in Dicer mutant skin to 15% ± 7% of control levels at P63. S100A4 also marks hair-follicle dermal papillae (Figure 4M). Dermal papillae were not detected associated with the hair-follicle remnants, suggesting that the epithelial signals required to maintain these structures  were not produced. Reduced or absent expression of K15, CD34, and other bulge-cell markers in Dicer mutant skin despite persistent expression of the outer-root-sheath markers SOX9 and K17 is consistent with failure of specification or apoptosis of the hair-follicle stem cell population.
Loss of epithelial Dicer produces several distinct defects in the skin, affecting both the epithelium and epithelial-mesenchymal signaling. These phenotypes include absence of hair-follicle stem cell marker expression, failure of dermal papilla and hair-follicle maintenance, and epidermal evagination of clusters of dermalpapilla-like cells. In addition, hyperproliferation in the absence of significantly increased apoptosis in older mutant epidermis suggests that aging Dicer mutant skin might be prone to developing tumors. Our findings that multiple miRNAs are expressed in postnatal mammalian skin and that production of mature miRNAs is reduced or absent in Dicer-deficient epidermis suggest that failure of miRNA processing contributes to the Dicer mutant skin phenotype. Our data thus reveal the existence of a previously unsuspected mechanism for post-transcriptional regulation in the skin and pave the way for identification of specific miRNAs required for critical aspects of hair-follicle and epidermal morphogenesis and maintenance.
The authors thank Dr. Kenji Kizawa for S100A3 antibody, Dr. Witold Filipowicz for anti-Dicer antibody, Leroy Ash for histology, Dr. Don Baldwin for miRNA array analysis, and Dr. Qian-Chun Yu for TEM. 2D-DIGE analysis was carried out by the Proteomics Core Facility of Genomics Institute and Abramson Cancer Center, University of Pennsylvania. We are grateful to Drs. George Cotsarelis and Mark Kahn for critical reading of the manuscript and helpful discussions. S.E.M. is supported by National Institutes of Health (NIH) grants R01-AR47709 and R01-DE015342. E.P.M. is an Engelhorn Scholar of the Watson School of Biological Sciences and is supported by a predoctoral fellowship from the Department of Defense (DOD) Breast Cancer Research Program (X81XWH-05-1-0256). G.J.H. is supported by funds from the DOD Breast Cancer Research Program and the NIH.
Supplemental Data Supplemental Data include one figure, two tables, and detailed Experimental Procedures and can be found with this article online at: http://www.current-biology.com/cgi/content/full/16/10/1041/DC1/.
Primary miRNA microarray data reported in this paper may be accessed in the Gene Expression Omnibus (GEO) database (accession number GSE4723).