Dominant-negative Sox18 function inhibits dermal papilla maturation and differentiation in all murine hair types

Hair follicle (HF) development is a tightly regulated process that relies on interactions between the mesenchymal dermis and overlying epidermis. Hair development is characterised by the formation of a dermal condensate (DC) in response to epidermal signals (Millar, 2002). This DC then induces invagination of epidermal cells in hair placodes, which in turn pushes DC differentiation into the end-point lineage: the dermal papilla (DP). WNT, EDA/EDAR or hedgehog signalling are known to drive the specification of the epidermal component of the hair follicle, but far less is known about the molecular mechanisms that control the dermal differentiation process. WNT signalling has been implicated as being crucial for the first inductive signals for the DC; however, the mechanisms by which the DP develops from this is unknown.

Of significant interest, whereas HF development starts at E14.5, it occurs in three consecutive waves in mice, each giving rise to a different type of HF in dorsal epidermis: the first, starting at E14.5, gives rise to guard hairs; the second wave, at E16.5, gives rise to awl and auchene hairs; and the final wave, at E18.5, gives rise to zigzag hairs, the most abundant hair type in dorsal murine skin (Duverger and Morasso, 2009). Recent work in the field has suggested that the hair type could be determined by the DP. Identification of SOX2 as a specific DP marker and its restricted expression in guard, awl and auchene hairs prompted studies characterising its role at E14.5 in the DC of the first hair wave (Driskell et al., 2009). Unexpectedly, conditional ablation of Sox2 in E14.5 DC allowed DPs and HFs to form. The resulting phenotype was modest and restricted to these three hair types (Clavel et al., 2012; Lesko et al., 2013). This suggests that expression of Sox2 is dispensable for dermal papilla specification, even if it could be important for some aspects of its function. In particular, the ablation of Sox2 had no effect on zigzag hairs (Graham et al., 2003).

DP cells that do not express Sox2 have been shown to be enriched for Sox18 (Driskell et al., 2009). Sox18 is a member of the SOXF family of transcription factors that also includes Sox17 and Sox7 (Dunn et al., 1995). SOXF transcription factors modulate arteriovenous identity and lymphatic endothelial cell specification (Corada et al., 2013; Cermenati et al., 2008; Pendeville et al., 2008; Herpers et al., 2008; Francois et al., 2008). Murine models of Sox18 mutation, including Sox18^−/−, Sox18^+/Ra and Sox18^Op/+, have shown that Sox18 also regulates hair follicle development (Carter and Phillips, 1954; Slee, 1962; Pennisi et al., 2000a,b; James et al., 2003). These studies report a defect that is restricted to zigzag hairs, supporting the view that SOX2 and SOX18 have complementing and non-overlapping functions that drive hair type formation (Slee, 1962; Pennisi et al., 2000a). More recent studies suggest Sox2 and Sox18 function in the dermal papilla, and Slee et. al. even suggest that Sox18 may regulate DP development; however, it remains to be investigated how Sox18 specifically regulates hair growth (Slee, 1962; Woo et al., 2012).

Importantly, in humans, mutation in SOX18 has been reported to cause the human syndrome hypotrichosis-lymphedema-telangiectasia (HLT), a condition that manifests with hair loss, and lymphatic and cardiovascular defects (Irrthum et al., 2003). This condition is mimicked by Sox18^Op/+ mice, which harbour a dominant-negative mutation of Sox18 (Sox18^Op/+) (James et al., 2003). This particular mutation suppresses the compensatory mechanisms by other SOX proteins observed in Sox18^−/− mice (Hosking et al., 2009). The Sox18^Op/+ model is therefore particularly valuable for studying hair development and the role of SOX18 and SOXF in hair development and hair type specification.

In the present study, we report that the Sox18^Op/+ mutation results in abnormal hair follicle morphology of all hair types and we identify the SOXF factor Sox18 as a crucial regulator of the DP differentiation program. We also demonstrate the importance of Sox18-regulated WNT signalling interactions in all hair types between the mesenchymal DP and HF epithelium.

We performed a comprehensive analysis of Sox18 expression pattern, as determined by the activity of its endogenous regulatory elements during skin embryonic development (from E14.5 to E18.5) using a Sox18-GFP/CreER (Sox18^+/GFPCre) knock-in reporter mouse, which expresses GFP under the control of the endogenous Sox18 promoter (Kartopawiro et al., 2014; McMahon et al., 2008). In the skin, expression of Sox18, as revealed by GFP detection, was restricted at all time-points to the DC structures, apart from some expected remnant expression in vascular structures. At E14.5, Sox18 was expressed in the early DC (Fig. 1A, arrow). Sox18 expression colocalised with the primary and secondary hair follicle marker SOX2, although its detection was not uniform within the DC (Driskell et al., 2009). The expression of Sox18 in the primary HFs was short lived, as at E16.5 and E18.5, GFP could no longer be detected in the DC from primary HFs, which are identified by SOX2 expression. To determine whether all forming primary DC expressed Sox18, we performed whole-mount co-staining for SOX2 and GFP on E15.5 Sox18^+/GFPCre embryos to identify GFP-positive DC. It appeared that all primary HFs were expressing Sox18 within the SOX2-positive mesenchymal region (Fig. 1B). The expression of Sox18 was also transiently detected in secondary follicle DC at E16.5, but not at E18.5 (Fig. 1C). Finally GFP (Sox18) was detected early in tertiary hair follicle development at E18.5 (Fig. 1C), where, as expected, tertiary hairs did not express SOX2 (Fig. 1C). We confirmed SOX18 protein expression at E18.5 using immunofluorescence co-staining with SOX2, which indicated that SOX18 protein expression could be detected in all hair follicle types (Fig. 1D) at E18.5, albeit at much lower levels in primary compared with secondary and tertiary HFs. To further confirm the expression of Sox18 at E14.5 in primary follicles, we used lineage tracing by combining the Sox18^+/GFPCre with an EYFP conditional reporter allele (+/ROSA-EYFP). By induction of Cre activity with tamoxifen at E14.5 in Sox18^+/GFPCre;+/ROSA-EYFP mice, we labelled cells derived from Sox18-expressing cells at E14.5 in primary HFs. We then detected, at birth, GFP/YFP expression via immunofluorescence and found that all visible cells within the DP of primary HFs were GFP/YFP+, along with the secondary and tertiary HFs, indicating overall Sox18 was or is expressed in the DC of all hair waves (Fig. 1E). In addition, these lineage-tracing results confirmed that cumulatively all cells within each DC were labelled with YFP, indicating that, at one point in DC development, all cells express Sox18.

We next explored the expression of other SOXF family members, such as Sox7, using a Sox7-LacZ reporter mouse and SOX17 using immunofluorescence. At E14.5 and E16.5, neither Sox7 (identified by β-galactosidase immunostaining) nor SOX17 could be identified in SOX2-expressing DC cells, although their presence was clearly visible in surrounding vascular structures (Fig. S2A). Similarly, at E18.5, skin sections did not reveal any Sox7 or SOX17 staining in primary, secondary or tertiary HFs (Fig. S2B). Overall, these data indicate the unique profile of expression of Sox18 among SOXF family members in the mesenchymal component of developing HFs.

Having established the mesenchymal expression pattern of Sox18 using endogenous reporters during HF organogenesis, we next investigated its function. The human syndrome HLT and phenotypic analysis on mice carrying the dominant-negative Sox18^+/Op mutation are characterised by alopecia; however, the molecular mechanism that underpins HF defects has yet to be elucidated (Irrthum et al., 2003; Pennisi et al., 2000b; Slee, 1962). Histological analysis of E14.5 Sox18^+/Op epidermis revealed that primary HF induction was indistinguishable from that in control littermates (Fig. 2A). At E16.5, however, a delay in the further growth of primary follicles was evident and the placodes of secondary HFs were not easily detected. By the tertiary wave at E18.5, an even greater delay in hair development was observed and affected all hair types, despite clear formation of placodes (Fig. 2A). Similarly, at birth (P0), there was a marked difference between mutant and wild-type mice regarding outgrowth of secondary and tertiary HFs, whereas primary hair structures could still be observed, although delayed and abnormal. Quantification of hair follicle stage using the method described by Paus et al. at P0 further confirmed that, in Sox18^+/Op mice, the majority of follicles were in an earlier phase of development compared with control (Paus et al., 1999).

Importantly, final hair formation, of all hair types, was abnormal in postnatal Sox18^+/Op skin. Evaluation of the HF subtype in mutant mice and wild-type littermates clearly revealed a complete lack of zigzag and auchene hair in the Sox18^+/Op mice (Fig. 2C, quantified in Fig. 2D), and compared with Sox18^−/− mice (Pennisi et al., 2000a). The remaining hairs were either guard hairs or had a new hair phenotype somewhere between guard and awl hair types that could not easily be accurately classified. These hairs were larger, like guards; however, they contained a kink, like awls, although the kink was less marked than that of control awl hairs (Fig. 2C). These we quantified as awl hairs in Fig. 2D. We then showed, using histology, that, by P22, secondary and tertiary follicle development was totally disrupted, resulting in thin epithelial strands into the upper dermis (Fig. 2E). These hair remnants seemed to derive from the secondary and tertiary hairs. Overall, these results indicate that a dominant-negative Sox18^+/Op mutation affects HF development of all hairs, halting auchene and zigzag (and likely awl) hair development post-placode development and altering guard hairs.

Although we showed that Sox18 is expressed in E14.5 DC cells, the Sox18^+/Op mutation did not seem to prevent DP formation in primary hairs, which allowed their development. In order to determine whether guard HFs formed from the primary DP after Sox18^+/Op mutation were normal, we explored their morphology and cycling. Interestingly, the remaining primary HFs, a much smaller proportion, also exhibited altered morphology and did not follow the parallel orientation observed in controls (Fig. 3A). When we compared guard hairs specifically (excluding the hairs that did not classify correctly as either guard or awl, mentioned earlier and Fig. 2C), the guard hair shafts obtained from plucking Sox18^+/Op mice were also shorter (Fig. 3B). Furthermore, the hair density on the back skin was certainly decreased in Sox18^+/Op skin (Fig. 3A). Histology showed that the remaining Sox18^+/Op HFs had delayed entry into first catagen and telogen, as long HFs extending into the dermis were seen at all time-points, including P17 and P19 when normal skin displayed only catagen and telogen hairs (Fig. 3C). Ki67 staining further confirmed this, as proliferation could be seen in the hair matrix in the remaining Sox18^+/Op guard hair, even at P19 (Fig. 3D).

To understand this alteration in cycling we next evaluated the effect of the Sox18 mutation in the DP on the epidermal compartment. We first investigated hair differentiation using an AE13 antibody which showed that the differentiated inner root sheath was severely reduced in the remaining guard hairs in the Sox18^+/Op mouse and completely absent from follicle remnants at P24 (Fig. S2), potentially explaining the reduced size of the hair shaft. We then analysed epidermal progenitor cell markers: SOX9 staining indicated that mutant mice maintained a proliferative progenitor compartment comparable with controls (Vidal et al., 2005). NFATc1 and CD34, markers of dormant epidermal bulge stem cells, appeared significantly reduced in Sox18^+/Op compared with control follicles (Fig. S2B-D). The loss of this cell population combined with the observed hyperproliferation and lack of telogen stage suggested an excessive activation of HF stem cells. This led to a permanently maintained anagen phase that resulted in their exhaustion at later stages. In support of this hypothesis, 12-month-old Sox18^+/Op mice showed a total loss of hair shafts and few hair remnants were still observed that did not stain for Sox9 (Fig. S2E-F). In summary, dominant-negative mutation of Sox18 in secondary and tertiary follicles results in complete loss of all hair markers; in primary (or guard) hairs, it resulted in decreased hair stem cell markers and reduced hair differentiation.

In order to determine the compartmental origin of the HF defect in Sox18^+/Op in either epidermal or mesenchymal skin, we employed hair regeneration assays. Wild-type or Sox18^+/Op mutant neonatal dermis and keratinocytes were mixed in multiple combinations to assess HF regeneration (Fig. 4). Wild-type keratinocytes combined with wild-type dermis resulted in significant hair regeneration 2 weeks post-injection (Fig. 4A), whereas a combination of Sox18^+/Op keratinocytes and Sox18^+/Op dermis recapitulated the Sox18^+/Op phenotype of greatly decreased hair formation (Fig. 4B). Interestingly, Sox18^+/Op dermis combined with wild-type keratinocytes (Fig. 4C) could not induce hair regeneration to anywhere near the same extent as wild-type dermis combined with either wild-type or Sox18^+/Op keratinocytes (Fig. 4A,D). These data suggest the Sox18^+/Op defect was of dermal origin.

We next developed neonatal dermal-derived spheres, adopting the skin-derived precursor (SKP) culture method for both wild-type and Sox18^+/Op mutant mice (Biernaskie et al., 2007). SKPs were easily produced from neonatal dermis and unanimously expressed SOX2 and Sox18 (Fig. S4). SKPs from Sox18^Op/+ mice did not change the phenotype observed in hair regeneration assays when added to Sox18^Op/+ dermis and keratinocytes (Fig. 4E). However, wild-type dermal-derived SKPs increased the number of hairs regenerated in the same assay mixing mutant dermis and keratinocytes (n=3) (Fig. 4F). This indicates that wild-type dermal SKPs have intrinsic Sox18-dependent properties that can rescue HF formation inhibited by dominant-negative Sox18 mutation in the dermal component.

In accordance with our initial findings of Sox18 expression in the DC, we sought to determine whether the Sox18 mutation was affecting DC specification. Interestingly, secondary and tertiary hair remnants showed inhibited DC differentiation, as very little alkaline phosphatase and no CD133 staining could be identified at P0 in the extremity of invaginating epithelial structures (Fig. 5A,B). Similarly, no α-smooth muscle actin could be identified in the mesenchyme surrounding each follicle (Fig. 5C). Specification of DC and DP was occurring, indicated by SOX2, which marks primary and secondary follicle DC cells (Fig. S1) and LEF1 (Fig. 5D), which, although not specific to the DC, is expressed in all hair types. We showed that the WNT signalling responsible for DC specification was induced because LEF1, a downstream target of WNT, could be detected in the condensate of developing HFs at all stages (Fig. 5D). These findings indicate that the dominant-negative defect of SOX18 allowed formation of a mesenchymal condensate with LEF1 expression. However, this DC failed to differentiate into a dermal papilla and to generate the second dermal signal for HF development.

In order to better understand how dominant-negative Sox18 mutation is affecting the DP, we performed gene expression microarray on dermal spheres (SKPs) isolated from Sox18^+/Op and wild-type littermate neonates that demonstrated differential hair inductive capacity. Comparisons were made within the same litter and animals were matched for sex. Unsupervised clustering analysis classified samples according to their genotype (Fig. S4). 2570 genes were differentially expressed with a P-value less than 0.05 after correction for multiple testing. Among these, 576 showed over twofold differences between control and Sox18^+/OP SKPs. Among the dysregulated genes, Wnt5a (2.4 fold, P<0.05), Fgf7 (2.2 fold, P=0.014), Wnt7b (2.2 fold, P<0.05), Mmp9 (18 fold, P<0.05) and Tnc (1.8 fold, P<0.01), all known as important contributors to the regulation of hair cycling or hair development by the DP, were significantly downregulated in Sox18^+/OP mutants. Other genes, such as Epha7 (a known target of SOX18) were also downregulated (2.4-fold, P<0.05) (Hoeth et al., 2012). WNT5A, MMP9 and TNC downregulation was confirmed by real-time PCR in Sox18^+/OP compared with control dermal spheres (Fig. 6A). Furthermore, WNT5A and TNC immunofluorescence showed a significant reduction of expression in the Sox18^+/OP skin, in part explaining the observed phenotype (Fig. 6B,C) (Youssef et al., 2012; Kaplan and Holbrook, 1994).

WNT5A and TNCC have been previously implicated as regulators of WNT signalling in hair and whisker (Reddy et al., 2001; Hendaoui et al., 2014). To determine how the secreted factors WNT5A and TNC may be affecting WNT signalling in the skin, we crossed Sox18^+/Op mice with a Flash reporter line expressing luciferase in cells with active canonical WNT signalling (Hodgson et al., 2014). WNT signalling was reduced in Sox18^+/Op mice compared with controls globally, at all phases of the hair cycle. Luminescence analysis in Sox18^+/Op:Flash^+/− mice indicated that the WNT signalling was at a baseline activity that did not strongly cycle (Fig. 7A,B). Interestingly, these data also supported our findings of delayed hair development and cycling in Sox18^+/Op mice (Fig. 3). However, as global WNT signalling decrease may reflect the lower density of hair follicles, we also investigated at the individual follicle level. LEF1 immunofluorescence performed on P24 skin further supported that the remaining WNT signalling from the epidermal compartment was reduced in intensity in Sox18^+/Op mice (Fig. 7C). This was further confirmed by β-catenin immunofluorescence. Nuclear staining, specifically in the epidermal compartment, of hair follicles confirmed fewer cells with active WNT signalling in Sox18^+/Op hair follicles when compared with control (Fig. 7D).

The molecular events leading to the development and function of the dermal clusters and dermal papillae remain unclear. Here, we have identified the essential role of Sox18 in DP development and function, especially in secondary and tertiary hairs but also in primary hairs. Indeed, Sox18 was expressed in a highly temporally restricted manner in the developing DP of all HF types. The absence of other SOXF family members in the DC of dorsal skin further supported the important role of Sox18 in this process. Dominant-negative mutation of Sox18 inhibited DP formation and differentiation without affecting early stages of dermal condensation. Skin-derived precursor (SKP) dermal spheres from wild-type mice restored the Sox18^+/Op phenotype in hair regeneration assays, further demonstrating that the defect in Sox18^+/Op mice is dermal and that Sox18 function is cell-autonomous. Finally, gene expression arrays using skin-derived precursor (SKP) dermal spheres, identified WNT5A and TNC as a potential downstream effectors of Sox18 in the dermal papilla that we hypothesise are mediating decreases in WNT signalling in the neighbouring epidermal HF cells.

A first contradiction that motivated this study was the proposed role of Sox18 in only zigzag hairs, which is opposed to its reported expression in the DC at E14.5 in a single study (Pennisi et al., 2000a). Sox18 reporter activity, SOX18 immunofluorescence and lineage tracing undoubtedly indicate that Sox18 is expressed at the RNA level in the mesenchymal component of all HFs limited to stages 1-2 of HF development (Paus et al., 1999), but persists at protein level beyond this stage until birth. When the follicle developed beyond stage 3, when the DP is engulfed by the epidermal matrix, Sox18 expression could no longer be detected in primary or secondary follicles. Despite the GFP reporter expression in Sox18^+/GFPCre only labelling some DP cells, lineage tracing showed that all cells within each dermal cluster express SOX18, indicating that all DP cells express Sox18 at one point within their development.

The inhibition of hair development in Sox18^+/Op skin supports that SOX18/SOXF are required for the normal formation of the DP. The phenotype observed in Sox18^+/Op mutant mice is much more severe compared with previous reports of Sox18^−/− mice (Pennisi et al., 2000a). Previous studies focusing on the cardiovascular phenotype of these same mice have reported a possible compensatory mechanism by other SOXF factors, notably SOX17 when SOX18 is fully lost (Hosking et al., 2009). By contrast, the Sox18^+/Op mutation, a single-point cytosine mutation at residue 1048, creates a missense transcript and a truncated version of the protein that prevent this transcription factor from recruiting normal SOX18 or its interacting partner MEF2C (James et al., 2003). The Sox18^+/Op mutation results in dominant-negative activity of SOX18 that abrogates its own activity as well as the role of its binding partners. Although we could not find SOX17 or Sox7 in the DC/DP during hair development, other SOX proteins, such as SOX2 or other binding partners could be involved in the observed phenotype of Sox18^+/Op mutation. As such, given the wide expression of Sox18, the Sox18^+/Op phenotype observed here highlights the activity of the complex formed by SOX18 with its binding partners rather than the activity of SOX18 alone. However, an important partner of this transcription factor complex is SOX18 itself.

The epidermis was not affected by the Sox18^+/Op mutation as it could be induced to form HFs. Moreover, SOX2 expression was normally detected in primary and secondary DC cells of mice with Sox18^+/Op. This suggested that DC formation and the first ‘inductive’ dermal signal was independent of SOX18 activity. Rapidly, after DC formation in Sox18^+/Op mice at all stages, there was a defect in epidermal invagination and proliferation, suggesting that the second dermal signal inducing HF growth was absent. Our gene expression array and further validation identified WNT5A as a potential candidate ‘second signal’ affected by Sox18 mutation that might be necessary for hair development. In addition, Sox18^Op/+ skin did not display normal DP formation, as witnessed by the loss of CD133, α-smooth muscle actin and alkaline phosphatase activity, the most commonly used markers of differentiated DP and dermal sheath. When combined, these data suggest that, after DC induction, SOX18, despite its restricted expression in time, allows for the differentiation of function of DP cells and their hair inductive properties.

The presence of dominant-negative SOX18 at E14.5 in primary DC cells had long-term consequences, despite the expression of SOX2 resulting in a loss of quiescent epidermal stem cells, extended anagen periods, poor differentiation (likely due to reduced number of cells with active WNT signalling) and long-term exhaustion of epidermal stem cells. These indirect effects on the epidermis in guard hairs further highlight the importance of SOX18 and its binding partners in the long-term regulatory function of the dermal papilla. This also implies that Sox18 might have different functions in guard versus awl/auchene/zigzag hairs. This may be related to stronger activity of its interaction partners in guard and awl hairs. SOX2 is an obvious candidate for interaction with SOX18 and this role remains to be established. This is reminiscent of observations in the equivalent human disease HLT that occurs through homozygous or dominant-negative SOX18 mutation (Irrthum et al., 2003). Indeed, individuals with HLT are born with no eyebrows or eyelashes, and never develop any pubic or axillary hair at puberty. However, they have some scalp hair that is rapidly lost within the first few months to years after birth.

In summary, Sox18 expression is spatiotemporally regulated and detected in the mesenchymal component of all developing hair types. Its function in association with its binding partners is required to mediate DP differentiation and promotion of epidermal changes that will form the hair, possibly through production of WNT ligands and modulators. SOX18 activity in combination with its binding partners at E14.5 is further needed for proper regulation of quiescence and cycling of guard hair epidermal stem cells. These findings further elucidate the development of the dermal papilla, as well as the mechanism of the human HLT syndrome. This will aid the identification of the molecular cues in the dermal component that lead to the development of HFs.

+/Sox18GCre mice were kindly provided by Dr Andy McMahon (Kartopawiro et al., 2014; McMahon et al., 2008). Sox18^+/Op mice were obtained from Prof. Peter Koopman (Institute for Molecular Bioscience, The University of Queensland, Australia) and +/Flash reporter mice were as described previously (Hodgson et al., 2014). Sox7-LacZ knock-in mice were generated using Velocigene targeting of Sox7 gene using a BAC clone with a LacZ insert (Regeneron, Tarrytown, NY, USA). These mice were kindly provided by Dr Nicolas Fossat. All murine work was performed to institutional ethical requirements (University of Queensland Animal Ethics Committee).

Primary antibodies used on frozen tissue were SOX2 (1/100, Abcam ab97959, lot#196138-7), GFP (1/100, Life Technologies A10262, lot#1729643), CD34 MEC 14.7 (1/50, Abcam ab8158) and CD133 (1/50, Ebioscience 14-1331-80, lot#E04033-1631). Primary antibodies used on paraffin wax-embedded tissue were LEF1 C12 A5 (1/150, Cell Signalling CST2230), CD26 H194-112 (1/50, R&D AF954), Dlk (1/50, Abcam ab21682, lot#GR192235-1), αSMA 1A4 (1/100, Abcam ab7187), anti-hair cytokeratin AE13 (1/500, Abcam ab16113), SOX9 H-90 (1/200, Santa Cruz sc20095), Nfatc1 7-A6 (1/100, Santa Cruz sc7294), SOX18 H-140 (1/50, Santa Cruz sc20100) and WNT5A 3D10 (1/100, Abcam ab86720). Secondary antibodies used were anti-rabbit A568, anti-rabbit A488, anti-goat A488 and anti-chicken A647 (Invitrogen), all used at 1/500 dilution. Immunofluorescence was performed on both 4% paraformaldehyde (PFA) paraffin wax-embedded skin or 1% PFA-fixed cryoembedded skin; all sections were counterstained with DAPI. Images were acquired on either a Zeiss LSM 710 or Olympus FV1200 confocal microscope.

Luciferase analyses were performed as previously described (Hodgson et al., 2014). Briefly, mice were injected with 150 mg/kg D-Luciferin Firefly (Caliper) and imaged after 10 min using a Xenogen IVIS Spectrum live animal imaging system. Analysis was performed using Living Image Software.

Microarray experiments were performed on n=3 control and n=3 Sox18^+/Op sex-matched littermate SKP samples. Preparation and analysis was performed as previously established (Rodero et al., 2013).

RNA was isolated from ∼1×10⁶ SKP cells, separated by dispase digestion for 2 h at 37°C, using a Qiagen RNeasy kit. RNA (1 µg) was used for cDNA preparation using superscript III, according to the manufacturer's instructions. RT-PCR was performed using SYBR Green (Applied Biosystems) on a StepOne Plus real-time PCR machine (Applied Biosystems). Ct values were exported and data analysed using Excel and GraphPad Prism. Statistical significance was determined using an unpaired t-test between the relative RNA species amounts in control and Sox18^+/OP samples. Data represent the fold change.

Dermal and epidermal cells were prepared after skin was removed from neonate mice, washed in antibacterial and anti-mycotic solutions. Epidermis was separated overnight in Dispase II (Sigma Aldrich) (5 mg/ml at 4°C). Epidermal sheets were then incubated on TrypLE (Gibco) for 20 min at room temperature and epidermal cells pelleted. For each patch assay, between 1×10⁶ and 2×10⁶ dermal cells and 1×10⁶ epidermal cells were injected subcutaneously in nude mice. Patch assays were analysed after 2 weeks.

Hair development was analysed by counting number of hairs at each hair development phase using Haematoxylin/Eosin-stained tissue (Paus et al., 1999). Hair type and length was quantified from the plucked back skin of 3-week-old wild-type and Sox18^+/OP littermates observed using light microscopy. Type counts were then averaged and compared with previously reported values (Pennisi et al., 2000a). Statistical tests performed were Student t-test on both hair stage (Fig. 2) and hair length (Fig. 3).

We thank Patrick Tam, Nicolas Fossat and Tania Radziewic (CMRI, UNSW, Sydney) for the kind gift of the Sox7-LacZ mouse line. We thank Peter Koopman and Andy McMahon for access mouse lines.