Restoration of the intrinsic properties of human dermal papilla in vitro

The dermal papilla (DP) is a highly specialized mesenchymal component located at the base of hair follicles (Paus and Cotsarelis, 1999; Stenn and Paus, 2001; Driskell et al., 2011). Previous studies, including hair reconstitution experiments, have demonstrated the indispensable roles of DP in epithelial–mesenchymal interactions that enable hair follicle morphogenesis and cycling (for reviews see Oliver, 1991; Jahoda and Reynolds, 1996; Millar, 2002; Botchkarev and Kishimoto, 2003; Ohyama et al., 2010; Yang and Cotsarelis, 2010; Driskell et al., 2011). In mice, elegant transgenic approaches (Kishimoto et al., 1999; Rendl et al., 2005; Driskell et al., 2009) and newly identified cell surface markers such as CD133 (Ito et al., 2007) and α9-integrin (Rendl et al., 2005) enabled fluorescent labeling and targeted isolation of living DP cells. Sufficient numbers of mouse DP cells are now available for downstream applications such as microarray analysis and tissue regeneration assays (Rendl et al., 2005; Rendl et al., 2008; Driskell et al., 2009). However, isolation of human DP cells mostly depends on manual microdissection, which provides a limited number of DP cells (Messenger, 1984; Magerl et al., 2002; Ohyama et al., 2010). Accordingly, previous studies of human DP predominantly utilized cultured DP cells (Inui et al., 2003; Pflieger et al., 2006; Iino et al., 2007; Park et al., 2007; Shin et al., 2010). A potential disadvantage of this approach is the loss of some intrinsic properties of DP cells during culture (Ohyama et al., 2010; Yang and Cotsarelis, 2010). Loss of biological characteristics, especially hair inductive capacity, has been reported in rodent (Jahoda et al., 1984; Horne et al., 1986; Lichti et al., 1993; Weinberg et al., 1993) and canine (Kobayashi et al., 2011) DP cells, and is likely to occur in human DP cells. Thus, a clear demand exists for an approach that allows in vitro expansion of human DP cells without affecting their in situ characteristics.

Previous studies reported various approaches to maintaining in vivo characteristics in cultured DP cells, including co-culture with keratinocytes (Reynolds and Jahoda, 1996; Inamatsu et al., 1998; Kobayashi et al., 2011), cultivation on extracellular matrices, and sphere formation (Osada et al., 2007; Young et al., 2009; Higgins et al., 2010; Kobayashi et al., 2011). The hair inductive capacity of rodent and canine DP cells was maintained in cultures treated with keratinocyte-derived factors (Inamatsu et al., 1998; Kobayashi et al., 2011) and ligands in signaling pathways involved in hair follicle morphogenesis, including Wnt3a and Bmp6 (Kishimoto et al., 2000; Shimizu and Morgan, 2004; Rendl et al., 2008). However, whether a similar approach is feasible with human DP cells has not been fully assessed.

In the present study, we first elucidated the molecular signature of intact, non-cultured human DP. Subsequent bioinformatics analysis identified biological pathways that are activated in DP. By administering an optimized combination of activators of those pathways, a culture medium that maintained or even restored intrinsic DP characteristics was successfully developed. The current study uncovered previously unreported molecular characteristics of in vivo human DP and described an efficient methodology with which to propagate DP cells while maintaining intrinsic properties.

In our attempt to elucidate the global gene expression profile of intact human DP (Fig. 1A), 15 DPs were microdissected from anagen hair follicles obtained from four volunteers (Fig. 1A). As controls, human DP cells and fibroblasts primarily cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (10% FBS–DMEM) (de Almeida et al., 2005; Wu et al., 2005; Iino et al., 2007) were prepared from the same individuals (Fig. 1A). Real-time PCR analysis of total RNA isolated from freshly isolated human DP (fDP), cultured DP (cDP) cells, and fibroblasts (Fibros) detected upregulation of genes that have been used to confirm successful DP cell isolation (Botchkarev and Kishimoto, 2003; Rendl et al., 2005; Ito et al., 2007; Osada et al., 2007; Driskell et al., 2009; Ohyama et al., 2010), such as alkaline phosphatase (ALPL), noggin (NOG), WNT inhibitory factor 1 (WIF1), and versican (VCAN), in fDP as compared with cDP cells and Fibros, demonstrating successful RNA recovery (Fig. 1B). The same total RNA samples were analyzed using four sets of Affymetrix HG-U133A 2.0 GeneChip microarrays. Despite limitations in the starting materials, the GeneChip data were of high quality (Khanna et al., 2004) and comparable with one another (Ohyama et al., 2006), with 55.5±3.14% of transcripts found to be expressed and an average correlation coefficient of 0.95±0.03 for the microarrays representing each population. The raw GeneChip data files are accessible at NCBI's Gene Expression Omnibus through GEO Series accession number GSE31324 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE31324).

Global gene expression profiles for fDP, cDP and Fibro GeneChips were subsequently compared to identify genes that are overrepresented in fDP. It should be noted that fDP samples contained microvasculature and possibly minor contaminants such as matrix cells and melanocytes (Magerl et al., 2002; Kobayashi et al., 2011), whereas cDP and Fibro samples contained almost pure cell populations (Fig. 1A). To minimize the influence of unwanted contaminants on gene expression profiles, two respective microarray analyses were performed. In each analysis, DP signature genes were selected by a ‘two-step’ filtering approach comprising primary gene screening by comparing global gene expression profiles of two well-defined subsets, and subsequent selection by assessing the expression patterns of candidate genes among additional cell populations (Ohyama et al., 2006).

In the first analysis (approach 1), cDP and Fibro GeneChip data (n = 4) were compared to identify genes upregulated in cDP. The expression levels of those genes in fDP and cDP cells were then compared to identify genes that were equally or more highly expressed in fDP (Fig. 1C). Among 105 genes detected as upregulated in cDP compared with in Fibros, 70 genes (supplementary material Table S1) were removed because they were downregulated in fDP compared with in cDP. Bioinformatics analysis using Database for Annotation, Visualization and Integrated Discovery software (DAVID, ver. 6.7; http://david.abcc.ncifcrf.gov/) (Huang et al., 2009a; Huang et al., 2009b) revealed that those genes were involved in metabolic processes and responses to external stimulus, suggesting that their upregulation was due to in vitro expansion. No previously described DP gene (Botchkarev and Kishimoto, 2003; Rendl et al., 2005; Ito et al., 2007; Osada et al., 2007; Driskell et al., 2009; Ohyama et al., 2010) was excluded in this step. This analysis identified 35 genes upregulated in fDP (Table 1; supplementary material Table S2). Differential expression of the selected genes was visualized by a heat map (Fig. 1D) and validated by real-time PCR analyses of 11 randomly selected genes (representative data in Fig. 1E). However, this approach did not allow for the detection of genes whose expression was rapidly downregulated in cDP cells to levels similar to those in Fibros (Fig. 1F).

Previous studies reported that DP cell aggregation by centrifugation in low cell binding plates, culturing on synthetic membranes, or hanging drop culture recovered DP functions, including DP biomarker expression. These results suggest that genes upregulated by DP cell aggregation could contribute to the biological distinctiveness of in situ DP (Osada et al., 2007; Young et al., 2008; Higgins et al., 2010). Accordingly, in the second analysis (approach 2), two additional sets of GeneChips were generated from aggregated cDP (agDP) cells and cDP cells used for agDP (cDP for agDP) and compared. Among the genes detected as upregulated in agDP arrays, those more highly expressed in fDP compared with both cDP and Fibros were finally selected as DP signature genes (Fig. 1F). A total of 77 genes were identified in this analysis (Table 2; supplementary material Table S3). The data are displayed in a heat map (Fig. 1G). Real-time PCR analyses of 16 randomly selected genes demonstrated expression patterns consistent with microarray data, although statistical significance between agDP and cDP was not sufficient for two genes (representative data in Fig. 1H). The genes upregulated by DP cell aggregation accounted for approximately 20% of all genes detected as increased in fDP among fDP, cDP, and Fibro subsets. Accordingly, a large proportion of fDP genes (n = 320; supplementary material Table S4) were filtered out in the first step. A total of 36 known hair follicle keratinocyte-related genes, 12 melanocyte-related genes, and 8 vasculature-related genes (supplementary material Table S4) (Rendl et al., 2005; Huang et al., 2009a; Huang et al., 2009b) were found in the removed genes, whereas such genes were not included in those upregulated in agDP cells. These results suggest that the genes expressed by the contaminants in microdissected DP were indeed excluded in the two-step approach.

A total of 108 genes were found to be upregulated in intact human DP cells by two microarray analyses. When the genes detected as increased in fDP prior to microarray generation (Fig. 1B) were combined, 118 human DP signature genes were identified in this study (Fig. 2A).

To categorize human DP signature genes, a functional classification analysis was performed using DAVID software (Huang et al., 2009a; Huang et al., 2009b), in which DP signature genes were divided into 14 groups according to enrichment scores (Fig. 2A). The clusters with high enrichment scores (>2.0) mainly comprised genes involved in WNT [frizzled-related protein (FRZB), Norrie disease (NDP), WIF1, wingless-related MMTV integration site 5A (WNT5A)], bone morphogenetic proteins 2 and 4 (BMP2, BMP4), NOG, and sclerostin domain containing 1 (SOSTDC1) and fibroblast growth factor [FGF; FGF7 and sprouty homologs 1 and 4 (SPRY1, SPRY4)] signaling pathways, suggesting the significance of these pathways in the maintenance of intrinsic DP properties (Fig. 2A).

DAVID gene ontology analysis annotated 100 of 118 signature genes as those with binding functions and 23 and 17 genes with transcription regulator and enzyme regulator activity, respectively. These results indicate that most DP signature genes were involved in intercellular or intracellular signal transduction or regulation. Previously unreported human DP genes encoding receptor binding molecules [e.g. glia maturation factor gamma (GMFG), lipoprotein lipase (LPL), NDP and metastasis suppressor 1 (MTSS1)], activators or inhibitors of enzyme activity [e.g. apolipoprotein E (APOE); G-protein signaling 2, 24 kDa (RGS2); cyctatin A (CSTA); and stratifin (SFN)], and transcription factors [e.g. ets variant 1 (ETV1)], forkhead box O1 (FOXO1), growth arrest-specific 7 (GAS7), and myocyte enhancer factor 2C (MEF2C)] were listed, which could imply novel mechanisms contributing to human DP function (Fig. 2B). The genes of membrane proteins [e.g. endothelin receptor type A (EDNRA), G-protein-coupled receptor 125 (GPR125), and low-density lipoprotein receptor-related protein 4 (LRP4)] and extracellular matrix proteins [e.g. spondin 1 (SPON1) and matrix Gla protein (MGP)] were also identified (Fig. 2B).

Importantly, 38 of these genes are reportedly over-represented in mouse DP cells (Fig. 2C) (Sleeman et al., 2000; Botchkarev and Kishimoto, 2003; Panteleyev et al., 2003; O'Shaughnessy et al., 2004; Rendl et al., 2005; Driskell et al., 2009). Most of them were established DP signature genes (Botchkarev and Kishimoto, 2003; Rendl et al., 2005; Ito et al., 2007; Osada et al., 2007; Driskell et al., 2009; Ohyama et al., 2010), including those encoding key molecules that enhance [e.g. FGF7, insulin-like growth factor-1 (IGF1), and WNT5A] or inhibit (e.g. FRZB and SOSTDC1) interactions between DP and the hair matrix (Fig. 2B). These findings indicate that a fundamental molecular mechanism is conserved between mouse and human DP cells.

As an application of the results obtained above, we attempted to develop a culture medium that allows human DP cell expansion while sustaining intrinsic properties, including the molecular signature. Because the bioinformatics analysis implied the involvement of WNT, BMP, and FGF signaling pathways in the maintenance of human DP properties, we anticipated that stimulation of these pathways during culture would improve the biological characteristics of cultured DP cells. To test this hypothesis, 6-bromoindirubin-3′-oxime (BIO), recombinant BMP2 (rBMP), and basic FGF (bFGF) were individually added to conventionally used 10% FBS–DMEM to stimulate the WNT, BMP, and FGF pathways, respectively.

To assess the effect of each activator on DP properties, expression of the classic DP genes ALPL, lymphoid enhancer-binding factor 1 (LEF1), and NOG, as well as of RGS2, a DP gene identified in this study, was monitored (Fig. 3A). Interestingly, the expression of these genes was not always upregulated in a dose-dependent manner (Fig. 3A). BIO markedly increased the expression of all tested genes at a concentrations of 1 µM and higher. However, cell proliferation was markedly suppressed in 2 µM BIO. The ameliorative effect of rBMP on DP gene expression did not differ significantly between 200 or 400 ng/ml. bFGF dose-dependently enhanced DP cell proliferation. However, an increase in dead cells, probably due to overproliferation, was observed at 40 ng/ml (data not shown). Although downregulation of ALPL was observed at a higher concentration, upregulation of other genes was maintained at 20 ng/ml. Accordingly, each activator differently maintained DP signature gene expression at an individual optimal concentration of 1 µM for BIO, 200 ng/ml for rBMP, and 20 ng/ml for bFGF (Fig. 3A).

We next prepared 10% FBS–DMEM containing optimal concentrations of two or all three factors and cultured fDP cells. The combination of rBMP and bFGF enhanced DP cell proliferation, whereas the mixture of BIO and bFGF or all three factors reduced DP cell growth to approximately 30% that of controls (Fig. 3B). In addition, DP cells hardly proliferated when exposed to the combination of BIO and rBMP (data not shown). Importantly, the combination of all three factors caused statistically significant (P<0.05) upregulation of all tested DP signature genes compared with controls (Fig. 3C). Thus, 10% FBS–DMEM containing 1 µM BIO, 200 ng/ml rBMP, and 20 ng/ml bFGF provided the most favorable conditions for maintaining proliferative capacity and intrinsic gene expression in cultured human DP cells. We named this medium ‘dermal papilla activation culture’ or DPAC.

Expression of other classic DP signature genes in DP cells primarily cultured in DPAC (dpacDP) was comparable with or higher than that in fDP (Fig. 3D). In addition, pentraxin 3 (PTX3), a Fibro signature gene upregulated in cDP compared with fDP, was significantly underrepresented in dpacDP versus cDP, further confirming the effect of DPAC (Fig. 3D).

Both the conventional medium (10% FBS–DMEM) and DPAC allowed DP cell expansion for more than five passages (Fig. 3E). Early-passage dpacDP cells predominantly demonstrated a small polygonal appearance distinct from the fibroblast-like morphology of cDP cells. However, numbers of spindle-shaped cells increased in higher-passage dpacDP (Fig. 3E). Consistent with this observation, ALPL, NOG, and BMP4 gene expression and alkaline phosphatase enzyme (ALP) activity, which correlates with the hair inductive capacity of DP cells (McElwee et al., 2003; Rendl et al., 2008), were maintained in early-passage dpacDP at levels comparable with those in fDP (Fig. 3F,G). Thus, the effects of DPAC persisted for several passages.

Human DP cells primarily cultured in DPAC sustained the expression of previously undescribed human DP signature genes (Fig. 4). LRP4, which is reportedly a specific marker of murine DP (Lrp4/corin) (Enshell-Seijffers et al., 2008; Enshell-Seijffers et al., 2010), was almost identically expressed in fDP and dpacDP (Fig. 4). Likewise, BMP and activin membrane-bound inhibitor homolog (BAMBI) and SPRY4, antagonists of TGF-β and FGF pathways (Fürthauer et al., 2001; Sekiya et al., 2004), respectively, and NDP, an activator of WNT signaling (Seitz et al., 2010), were similarly expressed in fDP and dpacDP cells (Fig. 4). Expression of those not directly related to the activated pathways, including ETV1; GPR125; guanylate cyclase 1, soluble, alpha 3 (GUCY1A3); and semaphorin-4C (SEMA4C), were also well maintained (Fig. 4). Trichorhinophalangeal syndrome I (TRPS1), a gene responsible for human congenital hair anomalies (Momeni et al., 2000), and sex-determining region Y-box 2 (SOX2), a mouse DP gene that defines a hair shaft phenotype (Driskell et al., 2009), were highly expressed in dpacDP compared with cDP cells (Fig. 4). These findings indicate that DPAC provides a useful tool to probe the links between potential DP signature genes and human DP biology.

When DP cells initially cultured in 10% FBS–DMEM were passaged into DPAC (yielding dpacDP P1 cells; Fig. 5A), their morphology changed from spindle-shaped to a polygonal form resembling that of DP cells primarily cultured in DPAC (Fig. 5B). In line with this observation, DP signature genes such as ALPL, LEF1, and NOG were upregulated after passaging into DPAC (Fig. 5C). In addition, ALP activity was much higher in dpacDP P1 cells compared with passage one cDP cells (cDP P1 cells; Fig. 5A,D). Thus, DPAC ameliorated DP biomarker expression reduced in conventionally cultured DP cells.

To assess whether exposure to DPAC improved DP functions, either dpacDP P1 or cDP P1 cells were co-cultured with normal human hair follicle keratinocytes (Fig. 6A) (Inui et al., 2002). In the bulb of the hair follicle, DP cells interact with hair matrix cells, actively dividing keratinocytes that generate the hair shaft (Rendl et al., 2005; Rendl et al., 2008; Ohyama et al., 2010). Intriguingly, dpacDP P1 cells significantly increased the expression of hair matrix signature genes LEF1 and msh homeobox 2 (MSX2) (Rendl et al., 2005) in keratinocytes compared with cDP P1 cells, suggesting that exposure to DPAC ameliorated the capacity of DP cells to provoke epithelial–mesenchymal interactions in the hair follicle (Fig. 6B).

An in vivo hair induction assay was subsequently performed. DP-like cell aggregates were generated from dpacDP P1 or cDP P1 cells and placed between enzymatically separated afollicular murine sole epidermis and dermis. The ‘sandwich’ composites were then inserted subcutaneously into immunodeficient mice (Fig. 6C). Intriguingly, developments suggestive of hair morphogenesis, including epidermal invagination or papillary mesenchymal body formation (Osada and Kobayashi, 2000; Osada et al., 2009) (Fig. 6D), were observed in 9 of 21 mice in the DPAC-treated group, but only in 2 of 18 mice in the control group (P<0.05). These developments were surrounded by dermal cells positive for ALP and human-specific vimentin staining (Fig. 6D,E), suggesting that they were induced by transplanted human DP cells.

These findings demonstrate that DPAC restores some aspects of DP cell function in conventionally cultured DP cells.

The success of microarray analysis greatly depends on the purity of the tested cell populations (Rendl et al., 2005). Microdissected DP predominantly comprised DP cells, as demonstrated by high expression of established DP signature genes. However, the vascular structure within DP could not be removed, and complete detachment of the surrounding hair matrix cells and melanocytes was hardly achieved by the microdissection technique (Ohyama et al., 2010; Kobayashi et al., 2011). Because the isolation of a pure DP cell population is currently not feasible in humans (Ohyama et al., 2010), a methodology that allows for the exclusion of non DP-related genes from fDP upregulated genes must be developed to obtain a bona fide intact DP gene expression profile.

We previously successfully identified the markers of human hair follicle putative stem cells by ‘two-step’ microarray analyses (Ohyama et al., 2006). We adopted this approach to filter out unwanted genes. Because no signal from possible contaminants was detected on cDP, Fibro, or agDP arrays, comparisons between cDP and Fibro GeneChips and agDP and cDP GeneChips should detect the differential expression of the genes expressed in DP cells. These preceding microarray analyses provided the criteria with which to link a gene to DP fate in each subsequent step. In approach 1, the second step could be crucial because some DP signature genes might be detected as downregulated in fDP compared with cDP cells and discarded due to the lower occupancy of DP cells in fDP subsets. However, previously reported DP genes (Botchkarev and Kishimoto, 2003; Rendl et al., 2005; Ito et al., 2007; Osada et al., 2007; Driskell et al., 2009; Ohyama et al., 2010) were not among the genes excluded in this step, suggesting that this loss is unlikely.

Theoretically, approach 1 could not detect any DP gene downregulated in cDP cells to a level similar to that in Fibros. In truth, cDP cells exhibited a global gene expression pattern resembling that of Fibros (Fig. 1G), indicating the need for additional analyses. Although approach 2 identified as many as 77 DP signature genes, excluding those expressed in contaminants in microdissected DP, some DP signature genes likely remained undetected because of the limitation intrinsic to this approach that not all DP signature genes were upregulated by cell aggregation. Indeed, 12 established DP genes (Botchkarev and Kishimoto, 2003; Rendl et al., 2005; Ito et al., 2007; Osada et al., 2007; Driskell et al., 2009; Ohyama et al., 2010) were upregulated in fDP compared with cDP cells and Fibros, but were not detected by approach 2. Importantly, half of those genes were selected by another analysis and finally listed as DP signature genes. This result, together with the observation that only four genes were common to both analyses, suggests that the combination of individual approaches was beneficial.

The current study identified 118 human DP signature genes, whereas global gene expression profiling of FACS-sorted pure mouse DP cells returned 225 DP signature genes (Rendl et al., 2005), most of which overlapped with the genes recently identified in CD133+ mouse DP cells (Driskell et al., 2009). Thus, a considerable number of human DP signature genes were not detected in the present study. Our two-step criteria were sufficiently stringent to identify selective markers, but may have been too restrictive to fully unveil the molecular signature of tested populations. Successful genetic fluorescent labeling of living human hair follicle cells through specific promoter expression has been reported (Tiede et al., 2009). A similar approach utilizing a promoter for a DP signature gene identified in this study or the use of potential cell surface markers described below might enable the isolation of more homogenous human DP cell samples and allow for the detection of unidentified DP genes.

In addition to the elucidation of the roles of WNT, BMP, and FGF signaling in the maintenance of DP properties, the bioinformatics analysis discovered attractive gene groups implying future applications. For instance, the modulation of enzyme activators, including APOE and RGS2, could augment DP cell functions. Among transcription regulator genes, those involved in cell differentiation, such as FOXO1 (Takano et al., 2007), GAS7 (Moorthy et al., 2005), and MEF2C (Lin et al., 1997), are of particular interest. The introduction of exogenous transcription factors could reprogram the biological properties of target cells (Takahashi and Yamanaka, 2006; Tsai et al., 2010; Tsai et al., 2011). Likewise, forced expression of these factors may facilitate the conversion of other dermal cells into cells with DP-like properties. Membrane proteins, represented by GPR125 and LRP4, may provide useful cell surface markers for living human DP cell isolation. Of note, Lrp4/corin is reportedly specific for mouse DP cells (Enshell-Seijffers et al., 2008; Enshell-Seijffers et al., 2010). The development of antibodies against these molecules for immunohistochemical- and/or fluorescence-activated cell sorting analyses could represent an important next step. Interaction with extracellular matrix plays a key role in the maintenance of DP cell properties (Ohyama et al., 2010). Upregulation of DP biomarker was observed when DP cells were cultured on laminin (Kobayashi et al., 2011). SPON1 or MGP may provide a supportive environment for human DP cells in culture.

For the development of DPAC, we adopted BIO, rBMP, and basic FGF to stimulate biological pathways downregulated during conventional culture. BIO is a small molecule commonly used to activate canonical WNT signaling in cultured cells (Sato et al., 2004; Yamauchi and Kurosaka, 2009). Although BMP6 improved mouse DP cell properties more efficiently than BMP2/BMP4 (Rendl et al., 2008), we did not detect upregulation of BMP6 in human fDP. Based on the observations that BMP2 was markedly suppressed during conventional culture and upregulated during DP cell aggregation, we selected this molecule as a BMP signaling activator. bFGF upregulated some DP signature genes, suggesting that active FGF signaling contributes not only to cell proliferation, but also to the maintenance of DP properties. As suggested by decreased expression of NOG, SOX2, and TRPS1 in dpacDP compared with that in fDP cells, dpacDP cells probably did not perfectly reproduce the molecular signature of in situ DP. Once a pure human DP cell population is isolated, microarray comparison between intact DP and dpacDP cells must be performed to more precisely assess the ability of DPAC to sustain DP properties. The findings obtained by this analysis could enable further improvement of DPAC.

In previous studies, supplementation of exogenous Wnt or Bmp molecules maintained mouse DP cell properties in culture, but did not affect their proliferative capacities (Rendl et al., 2008). For clinical application, the use of BIO may be advantageous because Wnt is not widely available in large quantities and may exhibit unwanted effects via non-canonical pathways (Shimizu and Morgan, 2004). However, considering that DP growth suppression due to BIO (Yamauchi and Kurosaka, 2009) represents a major drawback of DPAC, BIO may be substituted by recombinant WNT3A (Kishimoto et al., 2000; Shimizu and Morgan, 2004) to maintain the intrinsic properties and proliferative capacity of dpacDP cells. Because past reports suggested that activation of multiple biological pathways is necessary to fully maintain DP characteristics in vitro (Kishimoto et al., 2000; Shimizu and Morgan, 2004; Rendl et al., 2008), DPAC could be more potentiated by the addition of activators of other signaling pathways, including the Sonic hedgehog, Notch, and Endothelin pathways (Gao et al., 2008; Driskell et al., 2009; Hu et al., 2010). In contrast to DP cells, fibroblasts cultured in DPAC reduced ALP expression (data not shown). Thus, the conversion of other dermal cells into the DP cell lineage using DPAC is not feasible.

A unique feature of DPAC is that it restored the DP signature gene expression and functions normally reduced in conventionally cultured DP cells. The gene expression recovery seems to be imperfect by, for instance, moderate upregulation of RGS2. For full assessment of functional restoration, additional examinations, including the evaluation of the response to androgen (Inui et al., 2002) or adenosine (Iino et al., 2007), may be performed. However, the co-culture experiment and in vivo hair induction assay demonstrated that the most characteristic DP property, communication with the epithelial component (Ohyama et al., 2010; Driskell et al., 2011), was at least partly restored by DPAC. These observations imply a promising strategy for the efficient propagation of functional human DP cells by primary expansion under conventional culture conditions, followed by functional restoration. In fact, the yield of DP cells after the second passage was more than 10 times higher than that obtained through DPAC expansion from primary cultures, and the time required was approximately 50% less, indicating substantial advantages. The phase of isolated DPs might have affected the efficiency of the restoration. In the present study, DPs were mostly microdissected from full-anagen hair follicles. The role of DP in such hair follicles is to support hair shaft elongation rather than hair induction (Oliver, 1967; Paus and Cotsarelis, 1999). Early-anagen DP cells may possess more potent biological capacities, including resilience, than late-anagen DP cells. Although the use of such DP cells is technically very challenging, it could improve the outcome of the restoration experiments.

In summary, we have elucidated the molecular signature of intact human anagen DP cells. This guided the development of DPAC for the efficient in vitro expansion of human DP cells that retain their intrinsic properties. The present study sheds new light on the biological properties of human DP cells and implies a novel strategy for supplying sufficient functional human DP cells for the development of molecules acting on in vivo DPs or for regenerative medicine-based approaches to treat intractable hair loss disorders.

Human hair follicles were manually isolated from intact scalp skin obtained during surgical removal of benign skin tumors. DPs were microdissected from the bulbar portion of hair follicles under a dissection microscope as previously described (Williams et al., 1994). All experimental procedures were officially approved by the Institutional Review Board of Keio University. All human donors provided written informed consent in accordance with Declaration of Helsinki guidelines.

Five or six DPs were placed on the bottom of a 60-mm dish and cultured in DMEM (Invitrogen, Carlsbad, CA, USA) containing penicillin and streptomycin (Roche, Basel, Switzerland) and 10% FBS. For fibroblast collection, the dermis of scalp samples was cut into small blocks, 2–3 mm square. The blocks were attached to the bottom of 60-mm culture dishes and cultured in 10% FBS–DMEM. All cell culture was performed at 37°C in 5% CO₂. When confluent, primary cultured cells were harvested and re-seeded (20,000 cells/dish). After primary culture, all cells were passaged on a weekly basis.

Cultured DP cells were incubated with 2% trypsin and then transferred into low-cell-binding 96-well plates and centrifuged for several seconds to spin down the DP cells, and then incubated overnight at 37°C in 5% CO₂.

Total RNA was isolated from microdissected DP, DP cell aggregates and cultured DP cells using an RNeasy Micro Kit (Qiagen, Hilden, Germany). For real-time PCR analyses, cDNA was synthesized using SuperScript III First-Strand Synthesis SuperMix for qRT-PCR (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Real-time PCR analysis was performed using Power SYBR Green PCR Master Mix (Applied Biosystems, Santa Clara, CA, USA) and a StepOne Real-Time PCR system (Applied Biosystems). All primers were designed using Primer Express® software (Applied Biosystems) and reactions were performed under the following cycling conditions: 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 60 s at 60°C. The primer sets used are shown in supplementary material Table S5. Target mRNA expression levels normalized to the GAPDH mRNA using the ΔCt method were calculated as 2^−ΔCt. GAPDH was used as the reference gene because of its similar expression in microdissected and cultured DP cell populations, as shown by consistent Ct values. Three independent experiments were performed in duplicate. The results are presented as the means ± standard error of the mean (s.e.m.).

Isolated total RNA was amplified and biotin-labeled using a T7 primer based GeneChip® Two-Cycle cDNA Synthesis Kit and an IVT Labeling Kit (Affymetrix Inc., Santa Clara, CA, USA), respectively. Then, 20 µg of each labeled cRNA sample was fragmented and hybridized to an Affymetrix HG-U133A 2.0 GeneChip microarray and scanned using GeneChip Scanner 3000 7G (Affymetrix Inc.) according to the manufacturer's protocol.

To identify genes that are differentially expressed in intact human DP, two data analyses individually consisting of two-step microarray comparisons were performed using Gene Chip Operating Software ver. 1.4 (Affymetrix Inc.). For all microarrays, chip files (CHIP files) were generated according to the following parameters: target signal scaling, 500; Alpha1, 0.05; Alpha2, 0.065; and Tau, 0.015. Gene expression files were generated based on the following parameters: Gamma1H, 0.0045; Gamma1L, 0.0045; Gamma2H, 0.006; Gamma2L, 0.006; and Perturbation, 1.1. Only probe sets detected as being expressed (present call) in these files were analyzed. In the first analysis, global gene expression for the cDP and Fibro chips was compared (n = 4), and probe sets whose targets were upregulated ≥2-fold in cDPs were selected. Next, the expression levels of the targets for these probe sets in fDP and cDP GeneChips were compared in order to identify genes whose expression was upregulated in fDP or equally expressed in fDP and cDPs. In the second analysis, two sets of GeneChips were used to analyze total RNA samples from agDPs and cDPs in order to identify probe sets whose targets were >1.5-fold upregulated by DP cell aggregation. Among the probe sets upregulated in agDPs, those found to be >3-fold more highly expressed in fDP, as compared to both cDPs and Fibros, were selected. The Affymetrix IDs of these probe sets were uploaded into the gene batch viewer of DAVID (http://david.abcc.ncifcrf.gov/) (Huang et al., 2009a; Huang et al., 2009b) to generate a list of genes whose expression is upregulated in intact human DP.

The list of DP human signature genes was uploaded in DAVID, and the gene clustering analysis was performed with the following settings: similarity term overlap, 3; similarity threshold, 0.25; initial group membership, 2; final group membership, 2; and multiple linkage threshold, 0.50. Ontological functional annotation for each gene was obtained by DAVID. In addition, gene expression profiles for approaches 1 and 2 were analyzed by GeneSpring GX 11.5 software (Agilent Technologies, Santa Clara, CA, USA). The MAS without baseline transformation was applied for normalization. The genes were analyzed by hierarchical clustering and drawn as heat maps using the GeneSpring default setting.

DP cells were cultured in 10% FBS–DMEM containing BIO (1, 2, or 4 µM; Sigma, St. Louis, MO, USA) rBMP2 (100, 200, or 400 ng/ml; R&D, Minneapolis, MN, USA), or bFGF (10, 20, or 40 ng/ml; PEPROTECH, Rocky Hill, NJ, USA). The growth rates of DP cells under these different culture conditions were monitored every 4–5 days by assessing the ratio between the size of a starting DP and an outgrowing colony. Four weeks later, DP cells were collected and total RNA was extracted. Then, ALPL, NOG, LEF1, and RGS2 expression levels were measured by real-time PCR as described above. Subsequently, DP cells were cultured in 10% FBS–DMEM containing two activators (1 µM BIO and 200 ng/ml rBMP, 1 µM BIO and 20 ng/ml FGF, or 200 ng/ml rBMP and 20 ng/ml FGF) or all three activators (1 µM BIO, 200 ng/ml rBMP, and 20 ng/ml FGF). DP cell growth and signature gene expression were then assessed as described above.

Detection of ALP activity in DP cells was performed using an ALP detection kit (Chemicon Millipore, Temecula, CA, USA), according to the manufacturer's protocol.

DP cells primarily cultured in 10% FBS–DMEM were seeded on the upper compartment of 6 well Transwell culture dishes (Corning, Lowell, MA, USA) and cultured either in DPAC or 10% FBS–DMEM for 1 week. At the same time, human middle follicle outer root sheath keratinocytes were cultured on the collagen type I coated six-well plates in defined keratinocyte serum free medium (Invitrogen, Carlsbad, CA, USA). At day 7, Transwell dishes with DP cells were inserted into the plates with keratinocytes and cultured for additional 2 days in DMEM without FBS. Then, total RNA was isolated from keratinocytes for gene expression analyses.

Freshly isolated human DP cells were conventionally cultured in 10% FBS–DMEM and passaged either in DPAC or 10% FBS–DMEM. After 4 weeks in culture, aggregates of DPAC-treated and control DP cells were generated as described above. Afollicular footpads were surgically removed from 8-week-old ICR mice under a dissection microscope and incubated with dispase I (Godo Syusei, Tokyo, Japan; 1500 U/ml) for 30 min at 37°C. The epidermis was separated from the dermis and cell aggregates were placed between them. The footpad skin composites were incubated on Millicell-CM culture plate inserts (Millipore, Billerica MA, USA) for 2 h at 37°C in 10% FBS–DMEM and grafted subcutaneously into BALB/c^nu/nu nude mice. Skin samples were harvested 2 weeks later for histological analysis.

Frozen sections were fixed through incubation with acetone for 20 min at −20°C and incubated overnight at 4°C with a monoclonal anti-human vimentin-Cy3 antibody (Sigma). Cell nuclei were counterstained with 4′,6-diamidino-2-phenylindole.

Statistically significant differences were identified by two-sided Student's t-test or Mann–Whitney U-test, depending on the size of the sample population. A P-value of less than 0.05 was considered significant.

We thank the late Jonathan C. Vogel (Dermatology Branch, CCR, National Cancer Institute, NIH, Bethesda, USA) for his technical advice and discussion.