SKPs derive from hair follicle precursors and exhibit properties of adult dermal stem cells.

HIGHLIGHTS
Hair follicles contain Sox2+ dermal precursors that are cells of origin for cultured SKPs

Sox2+ dermal precursors and SKPs share similar function and transcriptional profile

SKPs function as dermal stem cells, clonally generating dermal cells and inducing hair morphogenesis

SKPs home to a hair follicle niche and this niche maintains their stem cell properties

INTRODUCTION
The skin is a unique organ that undergoes continuous cell-turnover and harbors significant regenerative capacity in order to repair environmentally-mediated insults. At least some of this regenerative capacity is due to somatic tissue stem cells, including basal layer and hair follicle epidermal stem cells (Fuchs, 2009) and melanocyte stem cells (Nishimura et al., 2005). However, a dermal stem cell, responsible for maintaining and repairing the dermis, has not yet been described.
The dermis is a complex tissue comprised of many cell types, including dermal fibroblasts, myofibroblasts, adipocytes, blood vessels, nerves, and sensory receptors such as Merkel cells. The dermis also contributes the inductive mesenchymal cells necessary for regulating hair follicle morphogenesis, a cyclic process that occurs continuously throughout the life of many mammals. During embryogenesis, the dermis develops from mesenchymal precursors that generate dermal fibroblasts and adipocytes, and produce the inductive hair follicle cells. Thus, by analogy to other somatic tissue stem cells, one possibility is that a multipotent embryonic mesenchymal precursor persists into adulthood, thereby providing the dermal stem cell activity necessary for homeostatic maintenance and regeneration of this tissue. Support for the presence of such a multipotent dermal precursor comes from work showing that, in some animals, dermal cells play a key role in limb regeneration (Muneoka et al., 1986; Kragl et al., 2009), and can directly differentiate into the skeletal cells necessary to form an exoskeleton (Vickaryous and Hall, 2008).
In this regard, we previously isolated a multipotent precursor cell from rodent and human dermis that differentiated into mesodermal and peripheral neural progeny including adipocytes, skeletogenic cell types and Schwann cells (McKenzie et al., 2006). These cells, termed SKPs for SKin-derived Precursors, displayed properties similar to embryonic neural crest precursors, and within facial dermis were derived from the neural crest (Fernandes et al., 2004). Interestingly, the dermal papillae (DP) of hair follicles appear to comprise one niche for SKPs, based upon coincident patterns of gene expression, and upon the finding that cells with properties of SKPs can be cultured from adult whisker follicle papillae (Fernandes et al., 2004; Hunt et al., 2007; Joannides et al., 2004). Since DP mesenchymal cells are essential for hair follicle induction (Jahoda et al., 1984; Oliver, 1967), and since it has been suggested that DP cells might be dermal precursors (Gharzi et al., 2003), we asked whether SKPs might represent a previously-unrecognized dermal stem cell. Here, we provide evidence in support of this idea, showing that SKPs derive from Sox2+ follicle-associated precursors, and that they can contribute dermal cells for tissue maintenance, wound-healing, and hair follicle morphogenesis.

RESULTS

Endogenous hair follicle DS and DP cells express the stem cell gene Sox2 and generate SKPs when cultured
To investigate whether SKPs originate from endogenous hair follicle dermal cells, we took advantage of our finding that Sox2 is expressed by SKPs, as detected by RT-PCR and immunostaining of neonatal murine SKP spheres (Fig. 1A,B). Immunostaining for Sox2 in neonatal murine back skin showed its expression in follicle DP and lower dermal sheath (DS) cells (Fig. 1C). We confirmed this localization in a mouse with EGFP knocked-in to the Sox2 locus (Ellis et al., 2004). Within skin, Sox2:EGFP-expressing cells first appeared in the embryonic dermal condensates that precede hair and whisker follicle formation (Fig. 1D; Fig. S1A,B), as recently published (Driskell et al., 2009). At birth, when hair follicles were in the anagen growth phase, Sox2:EGFP was expressed in all awl, auchene and guard hair follicle DS and DP (Fig. 1E,F). In adulthood, Sox2:EGFP was expressed in DP and DS cells of anagen, but not catagen/telogen follicles (Fig. 1G–I), suggesting that expression was dynamically regulated. Sox2:EGFP was also expressed in dermal cells of adult whisker follicles (Fig. S1C), and in a small number of cells in close proximity to the hair follicle bulge (Fig. S1D). These latter cells did not express K17, K15, nestin, PDGFRα, the melanocyte marker DCT or the Schwann cell marker P0 (data not shown). However, a few expressed the epidermal precursor marker K5 (Fig. S1D).
To determine whether Sox2+ follicle cells gave rise to SKPs, we prospectively isolated them using flow cytometry. Approximately 1–3% and 0.1–1% of neonatal and adult back skin cells, respectively, were Sox2:EGFP positive (Fig. S2A,B). Similar results were obtained with facial skin (Fig. S2C,D). Most Sox2:EGFP+ cells expressed PDGFRα, but not CD34, a marker for bulge epidermal cells and dermal fibroblasts (Fig. 1J). Sorted Sox2:EGFP+ cells from neonatal and adult back and facial skin gave rise to EGFP+ SKP spheres when cultured in SKPs conditions, and they were enriched in this ability relative to Sox2:EGFP− cells (Fig. 1K,N). Quantification showed that adult Sox2:EGFP+ back skin cells were enriched 4-fold and 10-fold relative to unsorted and EGFP− cells (Fig. 1L,M). SKP spheres generated from Sox2:EGFP+ cells expressed the SKPs markers fibronectin, nestin, versican, and α-sma (Fig. 1O), and when differentiated for 14 days, generated βIII-tubulin+ neurons and α-sma+ myofibroblasts (Fig. 1P,Q) that had lost their EGFP expression. Thus, anagen hair follicles are a primary niche for SKP-forming dermal precursors.

Prospectively-isolated Sox2:EGFP+ dermal cells home to hair follicles, induce hair follicle morphogenesis, and differentiate into neural and dermal cell types
Since SKPs generate peripheral neural cells, we differentiated sorted Sox2:EGFP+ cells versus Sox2:EGFP− cells under neural conditions. Only Sox2:EGFP+ cells generated nestin+ precursors (data not shown) or βIII-tubulin+ neurons (Fig. 2A). We then asked whether the sorted cells could induce hair follicle morphogenesis, as predicted by their anatomical localization, using an ex-vivo hair reconstitution “patch assay” (Zheng et al., 2005). These experiments showed that Sox2:EGFP+, cd34− cells induced hair morphogenesis when mixed with neonatal C57/Bl6 epidermal cells and transplanted beneath the dermis of adult nude mice for 12 days (Fig. 2B). In addition, Sox2:EGFP− cells were depleted in this inductive ability relative to total dermal cells (Fig. 2C).
To ask if Sox2:EGFP+ cells could also generate differentiated dermal progeny, we transduced sorted Sox2:EGFP+, PDGFRα+, CD34− backskin cells with an RFP-expressing retrovirus (to allow tracing if Sox2:EGFP expression was lost), and transplanted them into NOD/SCID mouse back skin for 3 weeks. For comparison, we used Sox2:EGFP− cells. Remarkably, the Sox2:EGFP+, but not negative, cells homed back to their DP and DS follicle niche, where they appropriately expressed the DP markers NCAM, cd133 and Sox2:EGFP itself (Fig. 2D–G; Fig. S2E–G). Quantification confirmed this selective homing (Fig. 2H,I). In contrast, Sox2:EGFP+ cells that integrated into the interfollicular dermis expressed the dermal fibroblast markers PDGFRα, PDGFRβ, fibronectin, and cd34 (Fig. 2J,K; Fig. S2I) but no longer expressed EGFP or cd133 (Fig. S2H). Thus, Sox2:EGFP+ follicle dermal cells home to a hair follicle niche, induce hair follicles, generate interfollicular dermal cell types, and differentiate into neural cells that are never normally found within skin.

Global gene expression analysis of SKPs versus Sox2:EGFP+ dermal cells
To ask whether SKPs maintain properties of their Sox2+ dermal precursor parents, we performed global gene expression analysis, comparing 4 primary passage neonatal back skin SKP preparations with 5 sorted neonatal Sox2:EGFP+CD34− back skin cell isolates. RNA samples were analyzed on Affymetrix GeneChip Mouse Gene 1.0 ST Array. Spearman rank correlation analysis demonstrated that the two groups were very similar, with intersample correlations ranging from 0.86 to 0.94 (Fig. 3B). For comparison, intrasample analysis demonstrated correlations of 0.91–0.99 and 0.96–0.99 for Sox2:EGFP+ and SKP samples, respectively (Fig. 3B). We also analyzed a subset of 36 genes previously associated with follicle dermal cells including versican, BMP4, cd133, alkaline phosphatase, and corin. Sox2:EGFP+ cells and SKPs expressed these genes at similar levels (Fig. 3A).
Some genes were, however, expressed at significantly different levels based on the F-statistic with Benjamini-Hochberg multiple testing correction implemented in the LIMMA Bioconductor package (Smyth, 2004). Of 22,102 non-redundant protein-coding genes assayed by the array, 281 and 310 were significantly upregulated at least two-fold (Fig. 3C). Ingenuity pathway analysis demonstrated that most of these were associated with metabolic pathways or growth factor signaling (Fig. S3A). SKPs were relatively enriched in genes associated with O-glycan, phenylalanine, tyrosine and tryptophan biosynthesis, and the citrate cycle, and Sox2:EGFP+ cells in genes associated with ERK and PI3-kinase-Akt signaling, and glycosphingolipid biosynthesis. These differences were likely due to differences in environment rather than cell identity, since of those transcription factors differentially expressed 2-fold or more (the range was 2 to 4.3-fold), none were cell-type-specific or associated with dermal or stem cell identity (Fig. S3B). Thus, SKPs and Sox2:EGFP+ cells are highly similar at the transcriptional level, arguing that culturing does not fundamentally change the identity of the endogenous dermal precursors.

SKPs differentiate into dermal cell types within intact skin
To ask whether SKPs were also functionally similar to Sox2:EGFP+ dermal precursors, we transplanted neonatal YFP-expressing murine or adult GFP-expressing rat SKPs into adult NOD/SCID mouse back skin. At 2–3 weeks post-transplant, both populations of genetically-tagged cells were located throughout the dermis (Fig. 4A). These cells were morphologically similar to the endogenous fibroblasts, and expressed the dermal fibroblast markers PDGFRα, collagen type I, vimentin, fibronectin (Fig. 4B–D), and fibroblast specific antigen (data not shown) (>90% expressed fibroblast specific antigen or collagen type I, ~73% PDGFRα). Some also coexpressed the hyaluronic acid receptor, CD44 or the myofibroblast protein α-sma (Fig. 4E) (~80% and 7% expressed α-sma in the upper and lower dermis, respectively). Some transplanted cells were present in the adipocyte-rich hypodermis, where they expressed fatty acid binding protein and displayed adipocyte morphology (Fig. 4F). Genetically-tagged transplanted cells were never observed within the epidermis. Thus, SKPs, like Sox2+ dermal precursors, differentiate into dermal cell types.

SKPs actively integrate into a hair follicle niche
Intriguingly, some transplanted SKPs homed back to the follicle DP and DS (Fig. 4A,G), where they appropriately expressed the DP markers versican (Fig. 4H; Fig. S4A), p75NTR, and NCAM (data not shown), or the DS marker α-sma (Fig. 4I). Some transplanted cells in the DS also expressed the proliferation marker Ki67 (Fig. 4J), consistent with the fact that DS but not DP cells proliferate. Transplanted cells did not differentiate into Pax3+ or tyrosinase+ melanocytes (Fig. 4H; Fig. S4B). The specificity of this integration was shown by transplanting neonatal murine YFP+ forebrain neurospheres (NSCs), adult rat GFP+ bone marrow MSCs, and dermal fibroblasts from non-hairy plantar skin. The NSCs survived poorly, and never associated with follicles, while MSCs and plantar skin dermal fibroblasts survived, but were never found within the DP (Fig. 4K,L; Fig. S4C–E).
To determine whether follicle cycling alters this homing behavior, we depilated or shaved back skin of NOD/SCID mice (depilation induces follicles to enter anagen), prior to transplantation. Depilation caused adult rat SKPs to integrate into 2–3-fold more hair follicles than did shaving (28.5±4.9 versus 8.0±1.6 follicles), with the DP of each positive follicle containing 6-fold more transplanted cells (10.16±1.0 versus 1.52±0.14) (Fig. 4N,O). Similar results were obtained with neonatal murine SKPs (Fig. 4M).
Finally, we asked whether SKPs hair follicle integration was modified by skin wounding. Neonatal murine YFP+ SKPs were transplanted adjacent to punch wounds on NOD/SCID mouse back skin. 3–4 weeks later, many transplanted cells were present within the scar, where they expressed fibroblast-specific antigen, collagen type 1, fibronectin and the myofibroblast protein α-sma (Fig. 4P–R; Fig. S4F,G) (16% expressed high α-sma). Interestingly, transplanted cells were also present in DP and DS of immature-appearing hair follicles (Fig. 4S,T), with the DP cells appropriately expressing versican (Fig. 4T) and p75NTR (data not shown), suggesting that SKPs contribute to new follicle formation in wounded skin.

SKPs regulate hair follicle morphogenesis
To ask if SKPs induce hair follicle morphogenesis, as do Sox2:EGFP+ cells, we performed patch assays. As positive controls, we used neonatal dermal cells, which generated many hair follicles (Fig. 5A,B), and as negative controls MSCs or NSCs, which generated no or very few follicles (Fig. S5A,B). Interestingly, adult rat GFP+ SKPs were highly enriched for hair follicle inductive ability (Fig. 5E,F), generating follicles where the entire DS and DP was comprised of genetically-tagged cells (Fig. 5C,D). Neonatal YFP-expressing murine SKPs also induced follicle formation (data not shown).
We next performed in vivo experiments, taking advantage of the finding that the size of the DP determines follicle size (Ibrahim and Wright, 1982). GFP+ adult rat SKPs were injected into depilated adult NOD/SCID mouse back skin and analyzed after 8 weeks. Many transplanted cells differentiated into interfollicular dermal fibroblasts (Fig. 5G), and many comprised the entire DP and DS of correctly-oriented follicles (Fig. 5G,H). Remarkably, relative to endogenous murine hairs, hairs induced by rat SKPs were longer (10.41 mm ± 0.23 versus 7.96 mm ± 0.11; p < 0.0001) and had increased follicle bulb diameter (107.24 μm ± 4.99 versus 82.27 μm ± 2.51; p <0.01) and hair fiber width (49.26 μm ± 0.87 versus 44.60 μm ± 0.83; p < 0.001) (n=8, with 3 animals quantified) (Fig. 5I,J). Although many of these follicles were in anagen (Fig. 5H), some were in catagen/telogen phase (Fig. 5K), indicating that follicle-associated SKPs cycle with their newly-induced hairs.
To confirm that rat SKPs induced larger follicles, we performed patch assays. Adult rat SKPs instructed mouse epidermal cells to generate larger follicles than did mouse dermal cells, with an almost 2-fold increase in bulb diameter (Fig. 5L–N). Confocal microscopy demonstrated that the increased rat SKP follicle size correlated with almost 4 times more NCAM+ DP cells (Fig. 5O–Q) (88% of these cells were GFP+).

SKPs clonally differentiate into dermal cell types and induce hair follicle morphogenesis
To ask if single SKPs could both induce follicle morphogenesis and differentiate into dermal cell types, as expected for a dermal stem cell, we generated clones of adult rat SKPs. Patch assays showed that, of 7 clonally-derived rat SKP lines passaged at least 6 times (approximately 8–12 weeks in culture), 5 induced hair follicle formation (Fig. 6A,B). Indeed, when 50 spheres from one clone were mixed with 5 × 105 total neonatal murine skin cells, 30 ± 2 hair follicles had DP entirely comprised of GFP+ SKPs. This activity was persistent; one clone induced follicle formation after 11 months in culture, albeit inefficiently (Fig. 6C). Consistent with these results, 3 of 3 clones transplanted into adult NOD/SCID mouse skin reconstituted hair follicle DP and DS in vivo (Fig. 6D). Transplanted clone cells also differentiated into fibronectin− and vimentin+ interfollicular dermal fibroblasts and α-sma+ myofibroblasts (Fig. 6D,E). Thus, single SKPs were multipotent with regard to these dermal activities.

SKPs maintain their multipotency and capacity to self-renew within the hair follicle niche and serially reconstitute hair follicles
One of the most striking assays of in vivo stem cell functionality is serial repopulation by hematopoietic stem cells. We therefore asked whether follicle-derived DS and DP cells could self-renew as SKPs and sequentially induce and repopulate hair follicles (Fig. 6F). Hair follicles were generated in patch assays with adult rat GFP+ SKPs, and cells from these follicles (Fig. 6G) were cultured in SKPs conditions; 10–14 days later, genetically-tagged spheres were observed (Fig. 6H). When these spheres were passaged, and put back into patch assays, they again induced follicle formation (Fig. 6I). Using this approach, we serially generated SKP spheres and reinduced hair follicles up to three times (Fig. 6J).
To ask whether these follicle-derived SKPs also maintained their multipotentiality, we took four different approaches. First, we transplanted them into adult NOD/SCID mouse back skin for 4 weeks. Many transplanted cells differentiated into PDGFRα and collagen type I+ interfollicular dermal fibroblasts (Fig. 6K–M), and some integrated into the DS and DP of endogenous follicles (Fig. 6K), where they expressed versican and NCAM in the DP and fibronectin in the DS (data not shown). Second, we differentiated cells in culture conditions that support neural and mesodermal differentiation of neonatal SKPs (Fernandes et al., 2004). Follicle-derived SKPs generated adipocytes, α-sma+ myofibroblasts or smooth muscle cells, nestin+ precursors and βIII-tubulin+ neurons (Fig. 6N–Q). Third, we transplanted follicle-derived SKPs distal to a sciatic nerve crush in NOD/SCID mice, an environment that instructs naïve SKPs to differentiate into Schwann cells (McKenzie et al., 2006). Immunocytochemistry of the distal nerve 6 weeks post-transplant identified GFP+ cells closely aligned with axons expressing the Schwann cell markers P0 and p75NTR (Fig. 6R,S). Finally, we transplanted follicle-derived SKPs into the embryonic chick neural crest migratory stream at HH stage 18, as we have done for total neonatal SKPs (Fernandes et al., 2004). Analysis 3 days later showed that follicle-derived SKPS migrated out of the neural tube into neural crest targets such as the spinal nerve and DRGs, similar to murine SKPs and adult rat clonal SKPs (Fig. 6T). Intriguingly, some total and follicle-derived SKPs migrated to the presumptive dermis, and at late timepoints, some of these expressed the DP marker versican (Fig. 6U). Thus, follicle-associated SKPs display properties of dermal stem cells.

Lineage tracing of transplanted follicles demonstrates that follicular DP and DS cells contribute to dermal maintenance and wound healing
To ask if the endogenous Sox2+ dermal precursors are recruited out of their niche to contribute differentiated dermal cells, we performed punch wounds on Sox2:EGFP mice. At 36 hours, Sox2:EGFP+ cells were limited to hair follicles. However, by 72 hours EGFP+ cells were seen exiting hair follicles and streaming into the interfollicular dermis toward the wound site (Fig. 7A–H). Some of these emigrating cells coexpressed the DP/DS marker NCAM, but many were negative (Fig. 7C,E), indicating that they were losing their follicle-associated phenotype.
To further define this phenomenon, we used a transplantation-based lineage tracing approach. Genetically-tagged hair follicles were made in patch assays with either rat GFP+ SKPs or neonatal dermal cells (Fig. 7I) and transplanted under adult nude mouse back skin. Over the ensuing 4 weeks, tagged follicles integrated and grew to generate small patches of hair (Fig. 7J). Analysis at this timepoint showed that many GFP-tagged cells were retained within engrafted hair follicles, but that many had also migrated into the adjacent interfollicular dermis where they expressed fibroblast-specific antigen, cd34 and collagen type I (Fig. 7K–N; Fig. S7). Similar results were obtained with both follicle populations. In some animals, we also injured the skin adjacent to transplanted SKP-derived hairs with a punch wound (Fig. 7O). Analysis 4 weeks later showed that GFP+ cells had migrated into the regenerated, newly-healed dermis where they expressed cd44 and collagen type I (Fig. 7P–R). Thus, DP/DS cells can contribute differentiated cells to the dermis.

DISCUSSION
These studies support a number of major conclusions. First, our studies identify a Sox2+ endogenous dermal precursor within the follicle DP and DS that, when prospectively isolated, homes back to a hair follicle niche, induces hair follicle morphogenesis, and displays neural and dermal potential. Second, we show that these endogenous precursors generate SKPs when cultured, and that these two cell populations are very similar with regard to their transcriptome and their functional properties when transplanted into the dermis. Third, we demonstrate that SKPs can clonally reconstitute the dermis and induce hair follicle morphogenesis, properties predicted of a dermal stem cell. Fourth, the serial reconstitution experiments show that SKPs maintain their multipotentiality and their ability to self-renew within their hair follicle niche, and that they can serially induce hair follicle formation. Finally, our follicle graft/lineage tracing studies argue that Sox2+ dermal precursors can be recruited out of their niche to contribute differentiated dermal cells. Together, these experiments provide evidence for a hair follicle-associated dermal precursor that functions to regulate hair morphogenesis, and to maintain the intact or injured dermis and that, when cultured, generate SKPs that display all the predicted properties of multipotent dermal stem cells.
On the basis of these findings, we propose that these adult dermal precursors derive from a subset of embryonic mesenchymal precursors within the presumptive dermis that interact with epidermal precursors and undergo a transition to a Sox2+ precursor state, as recently suggested (Driskell et al., 2009). We also propose that hair follicles provide a niche for maintenance of the mesenchymal precursor phenotype, and that in adult life, these Sox2+ follicle cells both regulate follicle morphogenesis, and serve as a reservoir of dermal precursor activity, contributing dermal cells for the normal and injured dermis. The broader mesenchymal potential displayed by SKPs (Lavoie et al., 2009) and whisker follicle papillae (Jahoda et al., 2003; Wojciechowicz et al., 2008) suggests that they may also serve as precursors for adipocytes and/or blood vessel smooth muscle cells.
That DP mesenchymal cells induce hair follicle morphogenesis is well-established (Jahoda et al., 1984; Oliver, 1967). A precursor function for DP cells has also been previously suggested based upon their ability to contribute cells to healing wounds upon transplantation, and to generate non-dermal cell types when cultured (Waters et al., 2007). The data we present here provide strong evidence that this is indeed the case. Moreover, our finding that both DP and DS cells express Sox2 suggests that these anatomically-distinct cells may in fact represent a common precursor population. In support of this idea, previous studies showed that cells move between the DS and DP (Tobin et al., 2003), and that lower DS cells can regenerate the DP (McElwee et al., 2003). We suggest that since DP cells rarely divide, they may represent a quiescent stem cell population, with DS cells functioning as a transit amplifying population that ultimately provides differentiated cell types. Alternatively, DS cells may represent dermal stem cells, and provide cells both for dermal differentiation, and for the DP, where the cells acquire their inductive properties and function to regulate follicle morphogenesis.
One of the surprising findings reported here is the ability of transplanted Sox2:EGFP+ cells and the SKPs they generate to home back to their hair follicle niche. Interestingly, a hint of this activity was previously obtained in experiments transplanting GFP+ DP/DS cells into ear skin (McElwee et al., 2003). These findings may reflect a previously-undescribed mechanism for recruiting DS cells into the DP and/or for allowing re-establishment of the DS/DP when hair follicles re-enter the hair growth cycle. Clearly, SKPs are attracted to cells that are similar, as seen in patch assays where single dissociated SKP cells aggregated to form structures that nucleated hair follicle formation. While the relevant attractive signal is unknown, evidence that it is cell-intrinsic comes from our data showing that more rat than mouse cells aggregate to form the DP, even within the same tissue environment.
A second surprising finding is that Sox2:EGFP expression was regulated during the hair cycle; DP and DS cells did not express Sox2 during catagen or telogen. While some cells exit the DP during hair follicle regression (Tobin et al., 2003), many are retained, implying that at least some precursors regulate Sox2 expression in response to their external micro-environment. Since Sox2 is associated with maintenance of a stem cell state (Avilion et al., 2003) then this may suggest that the niche regulates the “stemness” of these dermal precursors. However, a recent study also showed Sox2:EGFP expression in the DP of developing hair follicles, and provided evidence that the level of expression was associated with follicle type, with small zig zag hairs expressing low or no Sox2:EGFP (Driskell et al., 2009). Thus, Sox2 expression may be associated with multiple functional properties, and it will be important to determine whether Sox2 is a readout for different functional states and/or whether it actually regulates the properties of these dermal precursors.
One key question in the stem cell field has been whether multipotent precursors like SKPs actually reflect endogenous precursor cells, or whether they are cells that dedifferentiate in culture. Data presented here showing that prospectively-isolated Sox2+ follicle DP and DS cells give rise to SKPs, and that these Sox2+ cells are very similar to SKPs with regard to their transcriptome and functional properties, argue that SKPs accurately reflect this endogenous precursor population. This finding, together with our data showing that SKPs display all of the properties predicted of dermal stem cells, provide strong support for the idea that the Sox2+ precursors we have identified here are endogenous dermal stem cells that serve to induce hair morphogenesis and maintain the dermis.

Experimental Procedures

Tissue culture
SKPs were generated from dorsal back skin of neonatal (P1-7) YFP-expressing mice (Hadjantonakis et al., 1998) (Jackson Laboratory) or P0–P3 and adult (5–10 week) GFP-expressing Sprague Dawley rats (SLC, Japan) and cultured as described (Fernandes et al., 2004) at 1,000–20,000 cells/ml. Spheres were passaged at least twice for transplants. For clones, secondary SKPs were grown at 1,000 cells/ml, and single clonal spheres isolated and expanded at least 5 weeks. MSCs were isolated from adult GFP-expressing rat bone marrow (generous gift of Fabio Rossi, U.B.C.), and cultured at 50,000 cells/ml in Mesencult human MSC medium plus 10% FBS (Stem Cell Technologies). Neurospheres were made from P1 lateral ventricles as described (Reynolds and Weiss, 1992). SKPs were differentiated in vitro under previously-defined conditions for neurons, Schwann cells, and SMA+ cells (Biernaskie et al., 2006; McKenzie et al., 2006). Adipocytes were differentiated in DMEM-F12 containing 1% penicillin streptomycin, 10% FBS, dexamethasone (1μM, Sigma), isobutylmethylxanthine (1mM, Sigma), and insulin (20μg/mL, Gibco/Invitrogen).

In vivo experiments
For transplants, 2×105 to 106 passaged, dissociated murine (n=8) or rat (n=12) SKPs were injected into depilated (n=11) or shaved (n=10) back skin of adult NOD/SCID mice (Charles River Laboratories). Alternatively, a full thickness 4 mm wide biopsy punch was made, and SKPs were injected adjacent to the wound (n=10). Controls were performed with MSCs (n=4) or NSCs (n=4). For in vivo hair follicle morphogenesis experiments, adult rat SKPs were transplanted (n=8), and quantification was performed with 30–50 hairs from each of 3 experiments. Awl hairs were used for width measurements. Punch wounds were performed on Sox2:EGFP mice (n=4) as for transplants. Nerve transplants were performed as described (McKenzie et al., 2006). In ovo transplants were performed as described (Fernandes et al., 2004) at Hamilton/Hamburger stage 18 with analysis at Stage 30–35. All procedures were approved by the Hospital for Sick Children Animal Care Committee and were within the guidelines of the Canadian Council of Animal Care.

Hair follicle induction assay
106 SKPs (n=6 adult rat, 4 neonatal murine), neonatal (n=2) or adult (n=3) rat dermal cells, neonatal murine NSCs (n=3) or adult rat MSCs (n=3) were mixed with 10,000 neonatal epidermal aggregates (5 × 105 to 2 × 106 single cells), and injected into adult nude mouse (nu/nu; Charles River) back skin for 10–12 days as described (Zheng et al., 2005). SKPs, MSCs and NSCs were passaged 1–5 times. For sorted cells, 2–5 × 105 cells were mixed with 10,000 epidermal cells. Individual bulb diameters (50 follicles/graft; n=2 grafts each) were measured using Volocity acquisition software (Improvision) and a Leica MZ16F stereomicroscope. To quantify DP cell number, confocal image stacks spanning the entire DP were collected and all NCAM+ and NCAM+, GFP+ cells were counted. For serial reconstitutions, patches were digested in collagenase (Type XI) at 37 °C for 1 hour or follicles with GFP+ DP (n=40–60 hairs/experiment) were individually dissected and digested with 0.25% trypsin-EDTA; similar results were obtained with both approaches (n=3 adult, 2 neonate rat SKPs isolations). Single cells were liberated by gentle trituration and cultured in SKPs medium at 2,000 to 20,000 cells/ml for 14 days.

Follicle grafting experiments
Follicles generated in patch assays were dissected as above, transplanted into a back skin incision on adult Nude or NOD/SCID mice and those that grew were analyzed at 3–4 weeks (n=4 adult rat SKPs, 2 neonatal rat dermal cells). Alternatively, punch wounds were made adjacent to the graft at 3 weeks, and skin analyzed 4 weeks later (n=3).

Cell sorting
Back and facial skin from neonatal (n=15) or adult (n=9) Sox2:EGFP mice (Ellis et al., 2004) were dissociated to single cells, sorted for EGFP expression on a MoFlo (Dako) or FACsAria (Becton Dickinson) with viable cells identified by propidium iodide exclusion. In some experiments, cells were stained with APC-conjugated anti-PDGFRα (1:100; eBioscience) and/or biotin conjugated anti-cd34 (1:100; eBioscience) followed by a PECy7 conjugated streptavidin secondary antibody (1:1000; BD Biosciences). Gates were set according to single stained positive controls and negative (secondary only/isotype) controls. For transplants, sorted cells were incubated in medium containing RFP retrovirus and 2 ug/ml polybrene for 18 hours (a gift from Drs. Akitsu Hotta and James Ellis, HSC), washed extensively, and 300,000 cells injected (n=4 Sox2:EGFP+PDGFRα+cd34− cells and 5 Sox2:EGFP− cells).

Microarray and bioinformatics
P4 Sox2:EGFP+, cd34− cells were sorted directly into RNAlater (Ambion). RNA was extracted and quality assessed from sorted cells and primary passage P4 SKPs using Trizol (Invitrogen), and an Agilent BioAnalyzer. RNA samples were analyzed on Affymetrix GeneChip Mouse Gene 1.0 ST Arrays, and background was corrected and normalized using standard RMA procedure implemented in the Affymetrix Expression Console software. Preprocessed data were analysed for significant differential expression using the LIMMA Bioconductor package, with the F-statistic with Benjamini-Hochberg (BH) multiple testing correction implemented in the eBayes function. Genes with BH-corrected p-value < 0.01 were considered statistically significant (Smyth, 2004). Differentially expressed genes were analyzed using Ingenuity Pathway Analysis software (www.ingenuity.com) for functional enrichment. Microrray data are deposited in the NIH GEO repository (accession number GSE18690): http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE18690.

Immunocytochemistry and histology
Antibodies are listed in Supplementary Experimental Procedures. Immunocytochemistry was performed as described (Fernandes et al., 2006; Fernandes et al., 2004; McKenzie et al., 2006), and acquisition and co-localization was confirmed by collecting z-stacks comprised of 0.2μm to 1μm optical slices using a Hamamatsu spinning disk confocal microscope fitted to a Zeiss Axiovert 200 inverted microscope (Quorum Technologies).

Statistics
All data are represented as mean ± SEM. Data (with the exception of the microarray analyses) were analyzed using two-tailed t-tests or one-way ANOVA where appropriate. All experiments were done at least in triplicate, unless otherwise noted.

Supplementary Material
01