De novo adipogenesis parallels follicular stem cell activity
To determine whether changes in individual adipocyte cell size contributes to growth of the intradermal adipocyte layer during the hair cycle (Butcher, 1934
; Chase et al., 1953
; Hansen et al., 1984
)(Figures S1A and S1B
), we analyzed individual adipocytes during the hair follicle cycle by immunostaining skin sections with antibodies against caveolin 1A, which is enriched on the cell surface of mature adipocytes (Le Lay et al., 2010
) and a fluorescent neutral lipid dye, Lipidtox. Morphometric analyses of individual caveolin+
cells quantified the cross-sectional area (XSA) of intradermal adipocytes. Adipocytes progressively increased in size following morphogenesis of the hair follicle (P4–P15) ( and S1C
). Following catagen, intradermal adipocyte XSA decreased to the area of adipocytes during morphogenesis. Thus, intradermal adipose tissue growth during follicle maturation occurs at least in part by hypertrophy of mature adipocytes.
Intradermal adipocytes regenerate via a proliferative precursor cell during the hair cycle
To determine whether anagen induction is associated with changes in intradermal adipocytes, we analyzed intradermal adipose tissue during the 2nd hair cycle when anagen activation is slower than the first hair cycle. At P49 when the follicles are in the second telogen, small intradermal adipocytes exist below the dermis distant from the follicles (). At P56, activation of follicular stem cells is initiated in some follicles, as indicated by an enlarged hair germ. The activated follicles are in close proximity to small caveolin+, Lipidtox+ cells that extend from adipose layer toward the growing hair follicle (), suggesting that follicular stem cell activation is associated with changes in the intradermal adipocytes.
To analyze if de novo
formation of intradermal adipocytes occurs through a proliferative precursor cell during the hair cycle, we determined if proliferative cells expressing perilipin, which is specifically expressed on mature adipocytes (Greenberg et al., 1991
), exist in the skin during the hair cycle by pulsing mice for 3 days with BrdU during different stages of the hair cycle (). When mice were pulsed with BrdU before the first telogen (P18–21), no BrdU positive nuclei were detected within perilipin+
adipocytes. In contrast, when mice were pulsed with BrdU following anagen induction from P21–P24, BrdU positive nuclei were located within perilipin+
cellular membranes ().
We further analyzed de novo
adipocyte formation by examining BrdU incorporation within the nuclei of mature adipocytes (), which were enriched from dermal tissue via enzymatic dissociation and differential centrifugation. Microscopic analysis of isolated cells and analysis of the expression of adipocyte specific mRNAs by real time PCR confirmed the enrichment of mature adipocytes using this isolation procedure (Figure S1D
). FACS analysis of BrdU staining in isolated nuclei from mature adipocytes revealed that when 3-day BrdU pulses were performed during the initiation of anagen, 10% of mature adipocyte nuclei exhibited BrdU localization. In contrast, less than 2% of BrdU+
nuclei were detected when mice were pulsed before anagen induction (). Taken together, these data demonstrate that intradermal adipocytes regenerate through a proliferative precursor during anagen induction.
Adipocyte precursor cells are activated during the hair cycle
Adipocyte precursor cells were recently identified in visceral and subcutaneous adipose tissue depots (Rodeheffer et al., 2008
). To determine if adipocyte precursor cells exist in the skin, we isolated stromal vascular fraction (SVF) cells from the skin dermis at P21, when anagen is induced during the 1st
hair cycle. Similar to visceral adipose tissue, adipocyte precursor cells (Lin-, CD34+
) are present within skin tissue ( and S2A
). To confirm skin-derived adipocyte precursor cells are functional, we cultured FACS-purified adipocyte precursor cells from the skin. After 3 days of culture, skin-derived adipocyte precursor cells form robust adipocytes, as seen by Oil Red O staining (Figure S2B
). In addition, adipocyte precursor cells were able to form caveolin+
cells when injected into the intradermal muscle layer of syngeneic mice (Figure S2B
). Thus, functional adipocyte precursor cells reside in the skin.
Resident skin adipocyte precursor cells display dynamic activity associated with the hair cycle
To analyze the number and proliferation of adipocyte precursor cells during the hair cycle, we pulsed mice with BrdU for 3 days during catagen (P15–P18), anagen initiation (P19–P22) or mature anagen (P22–P25), and analyzed the percentage of adipocyte precursor cells. Few adipocyte precursor cells exist in the skin during catagen (). In contrast, the percentage of adipocyte precursor cells in the CD34+ SVF cell population increased ~four-fold during anagen induction () and returned to baseline during maturation of the hair follicle at P25. Therefore, adipocyte precursor cell number peaks in the skin during follicular stem cell activation.
Analysis of BrdU incorporation within adipocyte precursor cells revealed that prior to anagen ~50% of adipocyte precursor cells are proliferating. However, once anagen was initiated, the percentage of proliferative adipogenic cells was reduced to ~25% (). Thus, adipocyte precursor cells are stimulated to proliferate during late catagen to generate an increased population of adipogenic cells during anagen induction. These data correlate with the timing of de novo adipocyte generation after anagen induction ().
To further characterize adipocyte precursor cells in the skin, we analyzed the mRNA expression of the adipogenic transcription factor, PPARγ. We find that when compared to expression within SVF cells, PPARγ is enriched in adipocyte precursors and expressed at ~25% of the mRNA levels of mature intradermal adipocytes (). These data confirm the adipogenic nature of resident intradermal adipocyte precursor cells.
Mouse models alter distinct adipocyte lineage cells in the skin
To define the function of intradermal adipocyte lineage cells during hair follicle regeneration, we analyzed mouse skin prior to and during hair regeneration at P21 in two separate mouse models with genetic mutations that affect adipogenesis (Figure S3A
), or of wild-type mice after pharmacological treatments that affect adipocyte lineage cells (Figure S5
In one genetic model, mice lacking Early B cell factor 1
) display a decrease in postnatal intradermal adipose tissue (Hesslein et al., 2009
) (Figure S3A
). Analysis of Ebf1
mRNA expression using in situ
hybridization revealed that Ebf1
is expressed in the DP in mature, growing hair follicles at P4 (Rendl et al., 2005
); however, bulge, hair germ, and DP cells lack Ebf1
expression during the initiation of a new anagen during the hair cycle (Figure S3B
), when adipogenesis is active. This expression pattern was confirmed by real time PCR on isolated DP cells and epithelial cells (Figure S3C
In another genetic model, the lipoatrophic ‘fatless’ Azip/F1 mouse, mature white adipocytes are lacking throughout the animal, including the skin (Figure S3A
), due to the expression of a flag-epitope tagged, dominant-negative form of C/EBP under the control of the aP2 promoter, which normally drives expression of Fatty Acid Binding Protein-4 (FABP4) late in adipogenesis (Moitra et al., 1998
). Immunostaining for the Flag epitope expressed within the Azip transgene detected expression of Flag+
cells within the immature subcutaneous adipose depot below the skin of Azip mice but not within the skin epithelium of Azip mice (Figure S3D
). The lack of Flag+
cells in the intradermal adipose depot of Azip skin suggests that aborted mature adipocytes do not persist in the skin of Azip mice.
While both Azip and Ebf1−/−
mice displayed delayed hair coat formation (to be reported elsewhere), these defects are absent after P15, and catagen proceeded normally in both Azip and Ebf1−/−
mice (Figure S4A
). Both Azip and Ebf1
null mice display normal epidermal and sebaceous gland proliferation at P21 (Figure S4B
) and sebaceous gland size in Azip and Ebf1−/−
skin was similar to WT at P21 (Figure S4C
). Furthermore, Azip and Ebf1−/−
mice displayed normal DP morphology and T cell numbers (Figures S4D and S4E
). These data suggest that these mutant mice do not display any overt changes of non-adipocyte lineage cells within the skin prior to P21.
We next examined the adipocyte lineage in these mutant mice. Using FACS, we find that adipocyte precursors are absent in Ebf1−/− skin but slightly elevated in Azip skin (). To analyze adipocyte precursor activity in Azip and Ebf1 null mice, we defined proliferation within the intradermal adipocytes following 3 days of BrdU injections after P21 (). Due to the lack of mature adipocytes in Azip skin, we analyzed putative adipocytes in the dermis based on their elevated expression of caveolin 1A. Both WT and Azip mice displayed BrdU+, caveolin+ cells surrounding hair follicles (). However, Ebf1 null mice lacked proliferative, caveolin+ cells within the dermis. Similarly, the dermis of WT and Azip mice was filled with PPARγ cells, while the dermis of Ebf1−/− mice exhibited few PPARγ cells (). These data suggest that adipocyte precursor cells are able to proliferate and differentiate into highly expressing PPARγ+ preadipocytes in the dermis of WT and Azip mice, but these early adipogenic events are absent within the skin of Ebf1−/− mice.
Defects in the generation of immature adipocyte lineage cells blocks follicle stem cell activation
In addition to these genetic models that diminish adiposity in the skin, we treated mice with PPARγ antagonists, bisphenol A diglycidyl ether (BADGE) and GW9662 (Bendixen et al., 2001
; Wright et al., 2000
) to inhibit adipogenesis pharmacologically (). Based on the lack of a phenotype in mice lacking PPARγ in the skin epithelium prior to 3 months of age (Karnik et al., 2009
; Mao-Qiang et al., 2004
), we did not anticipate dramatic alterations in the function of epithelial cells with the short use of these drugs in 3-week old mice.
PPARγ antagonists abrogate intradermal adipogenesis and hair follicle stem cell activation
To determine if treatment with PPARγ antagonists altered the regeneration of intradermal adipose tissue during anagen activation, we treated mice with BADGE and GW9662 from P18-P24. BADGE- and GW9662-treated skin exhibited a reduction in skin adipose thickness (Figure S5A
). To determine if intradermal adipocyte precursor cell number was altered with treatment of PPARγ antagonists, we quantified the percentage of adipocyte precursor cells in vehicle, BADGE- and GW9662-treated mice compared to SVF. In mice treated with BADGE and GW9662 from P18–P24, the percentage of adipogenic cells at P24 was elevated compared to the vehicle-treated mice (). Furthermore, intradermal PPARγ expression was decreased in BADGE- and GW9662-treated mice compared to vehicle (). Interestingly, if treatment of BADGE was delayed until after intiation of anagen at P21 (P21–P27), intradermal adipose tissue displayed normal intradermal adipose tissue size and PPARγ expression (). These results demonstrate that inhibition of PPARγ prior to anagen induction blocks intradermal adipose tissue regrowth by blocking the action of adipocyte preadipocytes but not reducing the number of adipocyte precursor cells.
To confirm that BADGE or GW9662 treatment does not alter the homeostasis of sebocytes, which express PPARγ and when aberrant, can alter bulge activity and epidermal homeostasis (Horsley et al., 2006
; Karnik et al., 2009
; Sundberg et al., 2000
), we analyzed Ki67 localization and Lipidtox staining in sebaceous glands of BADGE- or GW9662-treated mice. Treatment of mice with PPARγ antagonists from P18–P24 did not alter the proliferation of cells within the sebaceous gland (Figure S5B
) or the size of sebaceous glands (Figure S5C
). These results confirm that sebaceous gland homeostasis is not dramatically altered during the short-term loss of PPARγ function in the skin. Additional analysis of Ki67 staining in the epidermis revealed that these PPARγ antagonists did not alter epidermal proliferation (Figure S5B
Thus, these three mouse models with diminished or absent intradermal adipocytes affect different stages of adipogenesis in the skin. The Ebf1 null mouse lacks adipocyte precursor cells suggesting that this mutation acts at the adipocyte precursor cell to block postnatal intradermal adipogenesis. PPARγ antagonists do not block the formation of adipocyte precursor cells in the skin but disrupt the formation of PPARγ+, preadipocytes, resulting in a loss of postnatal intradermal adipogenesis. Finally, the Azip transgene blocks late stages of adipocyte maturation after PPARγ+, preadipocyte formation, allowing us to examine the role of mature, lipid-laden adipocytes in the skin.
Adipogenesis defects result in aberrant follicular stem cell activation
Next, we examined the telogen to anagen transition after P19 in WT, Azip, Ebf1
null and mice treated with PPARγ antagonists. Follicles of Ebf1
null mice display telogen or late catagen morphology from P21–P56, suggesting that Ebf1−/−
mice have defects in activation of bulge stem cells (). These defects were evident morphologically and by the lack of BrdU incorporation in hair germ cells after a 24 hr pulse (Figure S3E
). In contrast, Azip mice displayed anagen induction kinetics similar to WT mice (), as evidenced by anagen morphology and proliferation within the hair germ in the majority of Azip follicles at P21 (Figure S3E
). Taken together, these data suggest that immature adipocyte lineage cells, which are absent in Ebf1−/−
mice but present in Azip mice, are necessary for follicular stem cell activation.
mice may display defects in the skin based on Ebf1
expression in the DP at P4, we determined if the lack of adipocyte lineage cells are the primary defect that results in hair cycling defects in Ebf1−/−
mice using skin grafting experiments. Skin was isolated from P18 female WT or Ebf1−/−
mice, scraped to remove intradermal adipocytes, and grafted onto full thickness wounds of male Ebf1−/−
or WT littermates, respectively. Three weeks after grafting, hair growth was evident in the grafts from Ebf1−/−
mice on WT recipients, whereas WT grafts lacked external hair follicles when grafted onto male Ebf1−/−
mice (Figure S3F
). We verified that dermal cells in these grafts were derived from the male recipients using in situ
hybridization for the Y chromosome (Figure S3F
). Importantly, the epithelium and DP in anagen follicles were derived from the female Ebf1−/−
donor skin, suggesting that inherent defects in hair follicle cells of Ebf1−/−
mice do not prohibit hair growth induction.
To further confirm if adipocyte lineage cells are able to rescue hair cycling defects of Ebf1−/−
mice, we transplanted WT adipocyte precursor cells, which were FACS isolated from skin total dermal SVF, into Ebf1−/−
skin at P21. As a control, the contralateral side of the backskin was injected with total WT SVF cells, which consists of unfractionated cells isolated from the dermis. Three days post-injection, follicles within WT SVF-injected Ebf1−/−
mice remained in telogen as indicated by follicle morphology and by the lack of Ki67+
hair germ cells, which indicates anagen at early stages of activation (). In contrast, regions of Ebf1−/−
backskin injected with WT adipocyte precursor cells displayed Ki67+
cells within the hair germ of follicles and were adjacent to Y chromosome+
cells when WT male cells were injected into Ebf1−/−
female recipient mice (). When cell transplantations were followed for 2 weeks, follicles in Ebf1
null skin injected with WT adipocyte precursor cells were in full anagen, while the SVF injected skin remained in telogen (). Together with the skin grafting experiments (Figure S3F
), these data strongly suggest that the lack of adipocyte precursor cells in Ebf1
null mice at P21 is the likely cause for the lack of follicular stem cell activation in Ebf1−/−
mice, and the function of Ebf1 in other skin cell types, such as DP cells, is likely not responsible for the hair cycle phenotype.
Adipogenic cells are sufficient to induce hair follicle regeneration
Next, we examined whether PPARγ+ preadipocytes in the skin were necessary to induce follicular regeneration. To do so, we analyzed mice treated with BADGE and GW9662 during the transition from telogen to anagen from P18–P24. As controls, we treated mice with vehicle from P18–P24 or with BADGE from P21–P27 after anagen induction. The hair follicles of both mice treated with vehicle and mice treated with BADGE from P21–P27 generated anagen follicles normally with almost 100% of the follicles in anagen after 6 days of treatment (). In contrast, mice treated with BADGE or GW9662 from P18–P24 did not enter into anagen and remained in the telogen phase of the hair cycle (). These data indicate that preadipocytes with functional PPARγ nuclear receptors are necessary for regeneration of the hair follicle.
Adipocytes are sufficient to induce follicular stem cell activation
To determine if adipocyte lineage cells are sufficient to alter follicular stem cell activity, we intradermally grafted adipocyte precursor cells derived from the SVF of subcutaneous adipose tissue from mice expressing luciferase under the leptin
promoter (Rodeheffer et al., 2008
). We used 6–8 week old mice since murine hair follicles enter into an extended telogen phase that lasts for 3–4 weeks around 7 weeks of age. When shaved mice were injected with adipocyte precursor cells into the ventral region of WT mice, luciferase activity was identified at the injection site after 6 weeks (). Interestingly, mice with robust adipocyte formation displayed external hair growth in the injected area ().
To further determine if the hair growth-inducing activity of adipocyte lineage cells is enriched compared to unfractionated SVF cells, we injected SVF or FACS-isolated subcutaneous adipocyte precursor cells into the dermis of shaved, murine backskin at 7 weeks of age. Both cell populations were injected into the same region of the backskin to avoid differences in the hair follicle stage due to regional differences in the skin (Plikus et al., 2008
). Two weeks following cell engraftment, hair growth was evident at the adipocyte precursor cell injection site but not on the adjacent side injected with SVF cells (). Histological analysis of skin from these mice revealed morphological anagen induction in the adipocyte precursor injected skin but not in the skin injected with SVF cells (). These data suggest that adipocyte lineage cells are sufficient to induce precocious hair follicle regrowth.
To determine if immature adipocyte lineage cells or mature adipocytes are sufficient to induce hair follicle growth, we determined if adipocyte precursor cells derived from Azip mice could induce anagen in syngeneic WT mice at P49. Since mature adipocytes cannot be transplanted by current methods without adipocyte precursor cell engraftment, induction of anagen by Azip adipocyte lineage cells would indicate that mature adipocytes are not the primary adipogenic cell type involved in the induction of stem cell activity in hair follicles. When we injected SVF cells derived from Azip mice, Flag+ cells were absent from the skin and hair follicles remained in telogen (). However, in the areas of skin injected with adipocyte precursor cells from Azip mice, Flag+ cells were evident within the skin and were adjacent to hair follicles entering into anagen, as indicated by the enlarged hair germ morphology and Ki67 staining in the hair germ (). Taken together, these data suggest that immature adipocyte lineage cells initiate hair growth through the activation of follicular stem cell activity.
Defective PDGF signaling in follicles without adipocyte regeneration
To characterize potential molecular mechanisms by which adipocytes regulate hair follicle cycling, we analyzed skin sections in WT and Ebf1−/− mice for activation of signaling
pathways that regulate follicular homeostasis and regeneration (Blanpain and Fuchs). Specifically, we immunostained skin sections with antibodies against phospho-SMAD1/5/8, phospho-42/44 MAP kinase, and β-catenin to analyze bone morphogenetic, growth factor, and Wnt signaling, respectively. While nuclear β-catenin and phospho-SMAD1/5/8 were localized to the nuclei of cells within hair follicles in P7 Ebf1 null mice, as is observed in WT mice, phosphorylation of MAP kinase (p42/44) was diminished in Ebf1 null follicles compared to WT follicles (). This lack of MAP kinase activation extended to anagen induction, where phospho-MAPK+ nuclei were found in WT follicles in the hair germ and DP, but Ebf1−/− follicles lacked phospho-MAPK localization in both of these cell types ().
PDGF signaling in the skin requires intradermal adipocyte precursor cells
To define candidate molecules expressed by adipocyte lineage cells that could mediate cell signaling, we analyzed mRNA expression for molecules that have been implicated in hair follicle cycling in skin-derived adipocyte precursor cells and mature adipocytes. As described previously, BMP expression is enriched in mature adipocytes (Plikus et al., 2008
) (). Interestingly, the expression of PDGFA in adipocyte precursor cells was elevated almost 100 fold over the expression in SVF cells. Mice lacking PDGFA
display phenotypic similarities with Ebf1
null mice, including a delay of follicle stem cell activation that blocks anagen induction (Karlsson et al., 1999
; Tomita et al., 2006
To determine if mice with defects in intradermal adipocyte regeneration display defective PDGF signaling, we analyzed the expression and activity of the PDGF receptor (PDGFR) by immunofluorescence in WT, Ebf1−/−
and BADGE-treated skin. We find that during telogen and anagen, PDGFR is expressed below the bulge in the DP as described previously (Rendl et al., 2005
) (). Analysis of phospho-PDGFR demonstrated that during anagen induction, PDGFR is activated in the DP and the lower part of the hair germ (). To determine if intradermal adipose regeneration is required for activation of the PDGFR, we analyzed hair follicles from BADGE-treated and Ebf1
null mice for activation of the PDGFR at P21. As seen in , PDGFR activation was diminished in the DP of both BADGE-treated and Ebf1
Based on the data above, we hypothesized that PDGF signaling may be defective in Ebf1 null mice, which lack adipocyte precursor cells. Thus, we tested whether elevated PDGFA could trigger the activation of stalled hair follicle regeneration in Ebf1 null mice. To this end, we injected PDGFA-coated beads intradermally into Ebf1 null mice at P21. Three days after bead implantation, a majority of follicles adjacent to PDGFA-coated beads displayed morphologies characteristic of anagen follicles (). This growth induction increased with elevated concentrations of PDGFA with 100ng/µl activating ~86% of adjacent follicles, demonstrating a dose dependency of activation of Ebf1 null hair follicles. By contrast, follicles in Ebf1 null mice that were adjacent to BSA-coated beads remained in telogen. Taken together, these data suggest that intradermal adipocyte precursor cells activate PDGF signaling in the DP in a dynamic manner.