|Home | About | Journals | Submit | Contact Us | Français|
Primary cicatricial or scarring alopecias (CA) are a group of inflammatory hair disorders of unknown pathogenesis characterized by the permanent destruction of the hair follicle. The current treatment options are ineffective in controlling disease progression largely because the molecular basis for CA is not understood. Microarray analysis of the lymphocytic CA, Lichen planopilaris (LPP), compared to normal scalp biopsies identified decreased expression of genes required for lipid metabolism and peroxisome biogenesis. Immunohistochemical analysis showed progressive loss of peroxisomes, proinflammatory lipid accumulation, and infiltration of inflammatory cells followed by destruction of the pilosebaceous unit. The expression of peroxisome proliferator-activated receptor (PPAR) γ, a transcription factor that regulates these processes, is significantly decreased in LPP. Specific agonists of PPARγ are effective in inducing peroxisomal and lipid metabolic gene expression in human keratinocytes. Finally, targeted deletion of PPARγ in follicular stem cells in mice causes a skin and hair phenotype that emulates scarring alopecia. These studies suggest that PPARγ is crucial for healthy pilosebaceous units and it is the loss of this function that triggers the pathogenesis of LPP. We propose that PPARγ-targeted therapy may represent a new strategy in the treatment of these disorders.
Cicatricial or scarring alopecia (CA) are a diverse group of hair disorders that cause permanent destruction of the pilosebaceous unit. CAs that result from follicular loss because of thermal burns, metastatic cancer, trauma, and radiation are referred to as secondary (Stenn et al., 1999; Price, 2006). Primary CAs are characterized by a folliculocentric inflammation with the ultimate replacement of the follicle with fibrous tissue and progressive and permanent hair loss (Stenn et al., 1999; Price, 2006). The etiology and pathogenesis of CA remains unclear and they are currently treated as inflammatory disorders. Depending on the inflammatory cells detected during the active phase of the disease, CA are classified as lymphocytic (Lichen planopilaris (LPP), frontal fibrosing alopecia, chronic cutaneous lupus erythematosus, pseudopelade (Brocq), central centrifugal alopecia, alopecia mucinosa, and keratosis follicularis spinulosadecalvans), neutrophilic (folliculitis decalvans, tufted folliculitis, and dissecting cellulitis), and mixed (folliculitis keloidalis and erosive pustular dermatosis; Mirmirani et al., 2005). The clinical features of all these disorders include destruction of hair follicles, progressive hair loss, and permanent replacement of the follicle with fibrous tissue. The destructive inflammatory changes in all primary cicatricial alopecias occur in the infundibular region and, to a more variable degree, the isthmic region of hair follicles (Figure 1a and b). It is hypothesized that the permanent destruction of hair follicles in scarring alopecia may be because of destruction of hair follicle stem cells located in the follicular bulge region (Cotsarelis and Millar, 2001).
The sebaceous glands (SG) are common victims along with the hair follicle in CA (Stenn et al., 1999). SG are appendages connected to the hair follicle to form the pilosebaceous unit (Figure 1a). The function of SG in humans is obscure, although it is known to secrete sebum composed of a unique mixture of lipid metabolic products (Downie and Kealey, 1998). The SGs are thought to facilitate the coordinated breakdown of the inner root sheath (IRS) during the hair cycle, and thus may be critical for follicular regeneration (Stenn, 2001). Spontaneous mouse mutants, Asebia (Josefowicz and Hardy, 1978) and defolliculated (Porter et al., 2002), harbor hypoplastic sebaceous glands that may be the pathological cause of scarring alopecia in these models. Similar observations have been made with sebaceous adenitis with hyperkeratosis in dogs and cats (Stenn et al., 1999). In humans, the extent of sebaceous gland atrophy varies in different patients. Therefore, it is unclear as to whether CA results primarily from an abnormality or loss of sebaceous glands or from a deregulated inflammatory attack on follicular stem cells, although these two possibilities are not mutually exclusive. Thus, a molecular mechanism linking permanent loss of the hair follicle, sebaceous gland atrophy, and inflammation is warranted to develop effective new therapy for CA.
Peroxisome proliferator-activated receptors (PPARγ, α, and δ) are members of the nuclear receptor super-gene family that regulate the expression of genes involved in inflammation and lipid homeostasis (Wahli, 2002). They exhibit unique expression patterns within vertebrate tissues and are central regulators of gene expression and differentiation in several tissues including skin (Kuenzli and Saurat, 2003), sebaceous glands (Rosenfield et al., 1999), and the immune system (Cabrero et al., 2002). Previous studies have shown that PPARδ is the most abundant PPAR in the adult epidermis, although PPARγ is more abundant than PPARδ or PPARα in sebocytes. PPARγ plays a unique role in initiating the differentiation of sebocytes in the sebaceous gland and PPARδ is believed to induce sebocyte maturation (Rosenfield et al., 1999). These expression patterns may change during hyperplasia, differentiation, and inflammation. Previous studies in keratinocytes and epidermis have shown that activation of PPARγ stimulates keratinocyte differentiation, improves permeability barrier homeostasis, and stimulates epidermal lipid synthesis (Schmuth et al., 2008). The activation of all three PPARs is anti-inflammatory and interferes with many components of the inflammatory response by altering the expression of cytokines, receptors, and adhesion molecules (Cabrero et al., 2002). The broad-spectrum regulatory potential of PPARγ in lipid metabolism and in controlling the inflammatory response suggests a crucial role for this nuclear receptor in the maintenance of the pilosebaceous unit.
Here, we report that PPARγ-regulated pathways are deficient in the lymphocytic CA, LPP. PPARγ agonists can induce the expression of these downregulated genes in human keratinocytes. Finally, targeted deletion of the PPARγ gene in the stem cells of the follicular bulge in mice causes scarring alopecia that resembles human disease. These findings reveal a previously unreported role for PPARγ in maintenance of healthy pilosebaceous units and suggest that the loss of this function likely triggers the pathogenesis of LPP.
The lymphocytic CA, LPP, is the focus of our study here. Patients with clinical diagnosis of LPP had early active lesions that were judged clinically representative of primary cicatricial alopecia (Otberg et al., 2008). Affected biopsy specimens were taken from the “active border”, an area with inflammation and retained (but decreased) hair follicles. These were paired with samples from clinically unaffected areas in the same patient. At the active border of the affected patches, there is perifollicular erythema and perifollicular scale. Anagen hairs may be easily extracted from the active edge. In contrast, in unaffected scalp, there is no loss of follicular markings, and there is no erythema or scale. A pull test may only result in a few (~2–5) telogen hairs. As previously reported (Mirmirani et al., 2005), we observed that the histopathology of unaffected tissue looked near normal or showed very early histological changes (mild infundibular lymphocytic inflammation and early sebaceous gland atrophy).
In LPP, the scalp is often the only site of involvement with patchy or diffuse hair loss. In the active stage of disease, the histology of affected areas of scalp is very characteristic with dense lymphocytic infiltrate around many follicular epithelia (Figure 1e and f). The end stage is characterized by perifollicular fibrosis, scarring, and replacement of preexisting follicles with fibrotic tracts. Hematoxylin and eosin (H&E) staining of affected LPP sections at the infundibular region showed dense lymphocytic inflammation, atrophy of sebaceous glands, and loss of hair follicles, consistent with LPP histopathology (Figure 1e and f). Normal controls were age and sex matched with LPP patients. Normal controls examined had no evidence of hair or skin disorders, their scalp biopsies showed well-formed sebaceous glands, and were devoid of inflammatory lesions (Figure 1c and d).
Very little molecular information is available about disease initiation and evolution in LPP. We used gene-expression profiling to understand the molecular pathogenesis of LPP. Two sets of microarray experiments were performed. In the first set, we compared gene-expression profiles of affected LPP (N = 20, pooled) with normal scalp biopsies (N = 20, pooled) by hybridization to Affymetrix HG-U133 Plus 2.0 chips. Of the ~47,000 genes and expressed sequence tags represented on these chips, 205 were upregulated and 219 were down-regulated genes with >twofold changes. To identify early or primary events in the pathogenesis, we next compared geneexpression profiles of paired unaffected (N = 10) and affected scalp biopsies (N = 10) from LPP patients with normal controls (N = 10). In this second experimental set where samples were not pooled, 569 differentially expressed genes (182 upregulated and 387 downregulated) were identified in unaffected scalp biopsies (in at least 6 of 10 samples) compared to normal controls (Tables 1 and and22).
In affected scalp tissue from the same patients, 446 genes were differentially expressed in at least 6 of 10 samples (210 upregulated and 236 downregulated) compared to normal controls. Although some patient to patient variation in gene expression was seen, the core set of genes that were differentially expressed in the pooled samples were also differentially expressed in at least 6 of 10 individually analyzed affected samples (Tables 1 and and2).2). The major biological pathways in affected LPP tissue included inflammatory and cell death pathways as most significant among upregulated genes and lipid metabolic and hair follicle cycling and development pathways as most significant among downregulated genes.
In concurrence with the histopathology of unaffected tissue that looked near normal or showed very early histological changes, the microarray data showed increased expression of only three proinflammatory genes in all 10 unaffected tissue samples from LPP patients. These were CD40 (TNFRSF5), SPG21 and ARTS-1, genes required for activation of the proinflammatory cytokine tumor necrosis factor-α (Table 1). Data analysis with Ingenuity Pathways Analysis (Ingenuity Systems, www.ingenuity.com), identified the NF-κB and cytochrome P450 signaling as the most significant pathways among upregulated genes (Figure S1a). The network model of the NF-κB signaling pathway in unaffected tissue (Figure S1c) shows that only proinflammatory factors such as NF-κB, involved in T-cell activation (Baeuerle and Henkel, 1994), are upregulated at this early stage of disease evolution.
In stark contrast to unaffected tissue, a large number of inflammatory genes were upregulated in affected LPP tissue. Figure S1f shows the complex inflammatory gene network in affected LPP tissue compared to the inflammatory network in unaffected tissue (Figure S1c). The majority of upregulated genes in affected LPP were either required for tissue remodeling and apoptosis or were inflammatory genes as anticipated from histopathology. As shown in Table 1, we observed a dramatic increase in gene expression of cytokines/chemokines (MIP1, MCP1, CCL27, MMD, IL6, RANTES), extracellular matrix-associated proteins (OPN, MMP1, MMP9, MMP10, MMP28, TIMP4, ADAMTS1), apoptosis-related genes (CASP1, GADD45B, PDCD6, PDCD4, CASP8)., and cell-surface antigens (CD 68, CD 69), suggesting the activation of macrophages and T-lymphocytes, apoptosis, and substantial matrix remodeling in LPP. Intriguingly, another set of genes upregulated in LPP belong to the arachidonic acid/COX/prostaglandin pathway (PTGER4, PTGS2, ALOX5-AP) that is known to exert numerous immunoregulatory and proinflammatory activities. The microarray-based gene expression changes in key inflammatory genes described in Table 1 have been validated by real-time PCR analysis, using RNAs used to generate microarray data as well as independent sets of paired patient samples (Figure S2).
Thus, our observation that the majority of inflammatory genes are upregulated in affected LPP but not unaffected tissue suggests that although the components of the immune signaling cascade, such as cytokines, chemokines, and adhesion receptors, are important in disease progression, they may not represent primary events in the pathogenesis of LPP.
As shown in Figure S1b and e, the most significant biological pathways downregulated in both unaffected and affected tissue include lipid metabolism. Table 2 shows the comparative expression of the majority of lipid metabolic genes in unaffected and affected LPP tissue (changes observed in at least 6 of 10 samples) compared to normal tissue. The largest group of downregulated genes are those involved in fatty acid metabolism (60 genes), fatty acid desaturation, elongation and transport (10 genes), and cholesterol biosynthesis (10 genes; Table 2). In addition, genes required for hair follicle development and function were also downregulated in affected but not unaffected LPP tissue.
The decreased expression of lipid metabolic genes in unaffected LPP, including FADS2 (delta-6 desaturase), whose expression is restricted to differentiating sebocytes in the suprabasal layers of the sebaceous gland, cannot be explained on the basis of sebaceous gland destruction alone. As shown in Table 2, the gene expression of several lipid metabolic genes (ACAA2, ACOT1, ACOX1, SOAT1, HMGCR, HMGCS1, DHCR7, FADS1, FADS2, ELOVL4) is decreased to a greater extent in unaffected compared to affected tissue, although, the histological changes (sebaceous gland atrophy) in unaffected tissue are near normal or mild compared to affected tissue. This suggests that both histological (mild sebaceous gland atrophy) and biochemical (decrease in PPARγ-mediated gene expression of lipid metabolic genes) changes taking place in the pilosebaceous glands contribute to lipid metabolic changes during disease evolution. The microarray-based gene expression changes were independently validated for representative genes by real-time PCR analysis in normal, unaffected, and affected LPP tissue (Figure S3). The effect of altered expression of lipid metabolic genes could be deregulated lipid metabolism in the pilosebaceous glands of LPP patients.
We suspected that the effects of altered expression of lipid metabolism genes would be reflected in the lipid composition of scalp tissue (sebum) in LPP. We therefore analyzed the lipid profiles of paired unaffected and affected scalp biopsies (N = 5) by gas chromatography. As shown in Table S1, lipid analysis demonstrated a 43% decrease in cholesterol esters and a 110% increase in triacylglycerols in affected scalp compared to unaffected scalp biopsies from the same patients. The fatty acid profile in all lipid fractions were also altered with a significant increase in arachidonic acid and a decrease in sapienic acid in affected LPP compared to unaffected tissue (Table S1) in all lipid fractions tested (free fatty acids, triacylglycerols, and phospholipids). Sapienic acid is the major fatty acid in human sebum (Ge et al., 2003) and is synthesized from linoleic acid by peroxisomal and mitochondrial β-oxidation followed by desaturation (Delta 6-and Delta 5-desaturase) pathways. These data validate microarray and real-time PCR observations (Table 2 and Figure S3) and suggest that deregulated lipid metabolism in the pilosebaceous units of LPP patients, results in the increased production of bioactive arachidonic acid that is a precursor of proinflammatory lipids, the leukotrienes, and prostaglandins.
Our data show that although lipid metabolic changes are already seen in unaffected tissue, inflammatory changes take place predominantly in affected tissue (Tables 1 and and2).2). This suggests that lipid metabolic changes represent early or primary events in disease pathogenesis. It also suggests that deregulated lipid metabolism may be the cause rather than the effect of the inflammatory response in LPP.
Another group of genes that are downregulated in LPP are those required for peroxisome biogenesis. Peroxisomes are ubiquitous cell organelles that contain over 50 biochemical pathways required for oxygen, glucose, hydrogen peroxide, and lipid metabolism (Gould and Valle, 2000;Akimoto et al., 2005). Genetic and proteomic studies in yeast and mammalian cell systems have led to the identification of up to 32 proteins (collectively called peroxins or PEX) involved in peroxisome biogenesis. In mammalian cells, three of these peroxins (PEX3, PEX16, and PEX19) are specifically involved in peroxisomal membrane protein (PMP) import (Wanders, 2004; Akimoto et al., 2005). Previous studies have shown that when PEX3 or PEX16 proteins are absent or mutated in cells, peroxisomes disappear (Schliebs and Kunau, 2004; Heiland and Erdmann, 2005). Our microarray data (Table 2) showed that PEX3, PEX16, and PMP22 are downregulated in affected LPP. Again, as in the case of lipid metabolic genes, we show (Table 2) that PEX3 gene is decreased significantly in unaffected LPP tissue as well, suggesting that peroxisomal changes are early events in LPP disease pathogenesis. As many genes required for peroxisomal lipid metabolism are also downregulated in LPP, we speculated that these cell organelles may be defective in LPP. For visualization of peroxisomes, scalp biopsy sections from normal controls (N = 10) and paired unaffected and affected LPP (N = 10) were stained with the peroxisomal membrane protein PMP-70 and Alexa 488-labeled secondary antibodies. Normal scalp sections show PMP-70-positive “punctate” staining pattern characteristic of peroxisomes (Shimozawa et al., 2000; Heiland and Erdmann, 2005). Figure 2 shows abundant staining of peroxisomes specifically in normal sebaceous glands and in the inner (IRS) and outer (ORS) root sheaths of the hair follicles. In contrast, scalp sections from LPP patients lack PMP-70-positive peroxisome staining (Figure 2, LPP). Double staining these tissue sections with the nuclear stain DAPI, verified the presence of sebaceous glands, IRS, and ORS in LPP (Figure 2, LPP, lower panel). Intriguingly, unaffected tissue from LPP patients already begins to show loss of peroxisomes in the IRS and ORS, although, sebaceous glands still show peroxisome staining (Figure 2, uninvolved). This suggests that peroxisomes may be lost before sebaceous glands and demonstrates for the first time that LPP tissue displays peroxisomal deficiency.
As shown in Figure 3, peroxisomal staining and confocal microscopy of the ORS and IRS of hair follicle at higher magnification (× 40) shows numerous PMP-70-positive particles in normal tissue (Figure 3a), but a complete absence of peroxisomal staining in LPP tissue (Figure 3b). A minus primary antibody control of normal tissue (Figure 3c) similarly shows absence of peroxisomal staining. We observed autofluorescence of tissue in the minus primary antibody control (Figure 1c) that is masked by the bright staining of peroxisomes in normal tissue (Figure 1a). The PMP-70 immunoreactivity in normal (Figure 3d) tissue was significantly higher compared to LPP (Figure 3e) and the minus primary antibody control of normal tissue (Figure 3f) when the differences were quantified by Surface Plot Analysis (ImageJ NIH software) that provides a three-dimensional visualization of the intensity of PMP-70-staining particles. Thus, the lack of identifiable peroxisomal structures in LPP scalp tissue together with the downregulation of PEX3, PMP22, and PEX16 suggests that peroxisome deficiency may be because of an impairment of peroxisome biogenesis in LPP.
As a large number of inflammatory, lipid metabolic and peroxisomal genes are differentially expressed in LPP, we hypothesized that these genes may be regulated by a common transcription factor or master regulator. In silico promoter analysis (5 kb upstream of the transcriptional start site) of differentially expressed genes with Multi-genome Analysis of Positions and Patterns of Elements of Regulation, a platform for computational identification of transcription factor-binding sites (Marinescu et al., 2005) revealed PPAR response elements on all downregulated genes (data not shown). This provided preliminary evidence for the possible involvement of PPARs in LPP disease pathogenesis.
Pathway and network analysis of differentially expressed genes from unaffected and affected LPP tissue (Tables 1 and and2)2) were algorithmically generated based on information contained in the Ingenuity Pathways Knowledge Base (Ingenuity Systems, www.ingenuity.com). One such network showed connectivity between PPARγ and several key lipid metabolic genes including FADS1, ACOX1, ACAA1, and ACSL1 suggesting that these genes are directly regulated by PPARγ (Figure S4). The PPARγ network also shows a functional interaction between PPARγ and several of the upregulated genes in affected LPP including CD36, SLC27A1, and PTGS2 (COX2). Interestingly, these data suggest (Figure S4a) a functional negative regulatory loop between PPARγ and PTGS2 (COX2). As shown in Table 1, PTGS2 (COX2) gene expression is significantly increased (~6.5-fold) in affected tissue of LPP patients. Functional analysis of the PPARγ network (Figure S4b) showed that the biological functions and/or diseases that are most significant to the genes in this network include lipid metabolism, arachidonic acid and eicosanoid metabolism, and inflammatory diseases in different systems. Thus, Ingenuity Pathways Analysis (IPA) data suggests a possible role for PPARγ in regulating lipid metabolic and proinflammatory genes and the inflammatory response in LPP.
To confirm the possible role of PPARγ in LPP pathogenesis derived through bioinformatics analysis, we determined the gene-expression levels of the three PPAR isoforms in unaffected and affected LPP and control tissue. Real-time PCR showed that there was a significant decrease in PPARγ in unaffected and affected LPP (Figure 4a). In contrast, the expression of PPARα and δ remained unchanged in LPP compared to normal controls (Figure 4a). Thus, isoformspecific modulation of PPAR expression in LPP is suggestive of a functional role for this master regulator in LPP disease pathogenesis.
We directly assessed the role of PPARγ in modulating peroxisomal gene expression by growing human keratinocyte cell line in the presence of PPAR agonists (1 and 5 µm concentrations in 0.1% DMSO) and monitoring gene expression by real-time PCR (Figure 4b). The agonists tested were Ciglitazone (Cig), Rosiglitazone (Rosi), Pioglitazone (Pio), and Troglitazone (Tro) that are specific for PPARγ, WY-14363 that is specific for PPARα, and GW50516 that is specific for PPARδ. Remarkably, WY-14363 and GW50516 had minimal effect on PEX16 gene expression at any concentration tested. In contrast, Cig, Rosi, Pio, and Tro induced a significant increase in PEX gene expression, respectively, (Figure 4b). We also examined whether expression of the PEX3 gene is modulated by the PPAR agonist Pio in cultured ORS cells. The ORS cells were grown in the presence or absence of Pio and PEX3 gene expression was monitored by real-time PCR. As observed with human keratinocyte cell line (Figure 4b), PPARα (WY-14363) and PPARδ (GW50516) had minimal effect on PEX3 gene expression in ORS cells (Figure 4c). In contrast, Pio induced a significant increase in PEX3 gene expression at a concentration of 1 µm (Figure 4c). To determine, if PPARγ negatively regulates COX2 gene expression, we treated hair follicle outer root sheath cells (ORS cells) with PPAR agonists and measured COX2 gene expression by real-time PCR. As shown in Figure 4d, PPARγ agonists significantly inhibit the expression of the COX2 gene. In contrast, PPARα (WY-14363) and PPARδ (GW50516) agonists tested had minimal effect on COX2 gene expression. These data suggest that in human keratinocytes, it is the activation of PPARγ and not PPARα or PPARδ that induces PEX16 and PEX3 gene expression and represses COX2 gene expression. These observations support our microarray and IPA data and suggest that proinflammatory lipid metabolism and peroxisome biogenesis in LPP are modulated by PPARγ activity.
Together, these results suggest that PPARγ is a master regulator that controls complex functional networks linking lipid metabolism and the immune response in the pilosebaceous units. It is likely that the loss of PPARγ gene expression in the pilosebaceous units, deregulates lipid metabolism and elevates the expression of proinflammatory pathways such as COX2 and 5-lipoxygenase (5-LO;Table 1), thereby inducing an inflammatory response in LPP.
Our microarray data showed that PPARγ gene expression is decreased significantly in unaffected and affected LPP tissue (~27-fold in 6 of 10 unaffected LPP samples and ~19-fold in 6 of 10 affected LPP). As shown in Figure S1a, the most significant biological pathways upregulated in unaffected tissue are cytochrome P450 signaling and xenobiotic metabolism.
To confirm the role of xenobiotic metabolism in LPP, we analyzed the microarray data from unaffected and affected scalp biopsies with IPA-Tox(TM) within IPA application. The IPA-Tox tool enables a mechanistic analysis of the xenobiotic insult (not easily revealed by traditional methods) and helps to rapidly understand biological responses in the tissue of interest. As shown in the Figure S5, this analysis returned several impacted toxicity lists in LPP, the most significant being the Aryl Hydrocarbon Receptor (AhR) Signaling and Xenobiotic Metabolism gene lists. The AhR signaling pathway is a marker of xenobiotic metabolism. The CYP1A1 gene that is a direct target of AhR is upregulated by ~ 1,000-fold in unaffected tissue (in 10 of 10 samples tested) and ~ 500-fold in affected LPP (in 10 of 10 samples tested). These data have been validated by real-time PCR (data not shown).
Accumulating evidence in literature (Hanlon et al.,2003; Cimafranca et al., 2004) has shown that the AhR suppresses PPARγ gene expression in response to dioxin-like compounds. Thus, our data suggests a possible role for AhR as a suppressor of PPARγ gene expression and provides compelling evidence that xenobiotic metabolism may act as an environmental trigger in the pathogenesis of LPP.
To uncover the function of PPARγ in the pilosebaceous unit and its linkage to scarring alopecia-associated defects, we used Cre-loxP-mediated gene targeting to delete PPARγ in stem cells of the hair follicle bulge using a stem cell-specific promoter Keratin 15 (Liu et al., 2003). Floxed PPARγ mice (PPARγ (f/f); He et al., 2003) contain loxP sites on either side of the exons 1 and 2 of the PPARγ gene. Cre-mediated deletion of these exons is predicted to result in loss of PPARγ1 and a nonfunctional, N-terminal, 43 amino-acid translational product of PPARγ2 that misses the partial AF1 domain and the first zinc-finger of the DNA-binding domain (Zhu et al., 1995). The floxed PPARγ mice were crossed with a line of mice that express Cre under control of the keratin 15 promoter (K15-Cre) to yield the follicular stem cell-specific PPARγ knockout (KO) mouse, PPARγ (f/f)/Cre. Homozygous PPARγ-stem cell KO mice (PPARγ (f/f)/Cre) were born at the expected Mendelian frequency, suggesting normal early development. Control mice were fl/fl littermates not expressing Cre.
As shown in Figure 5a, floxed mice without Cre (PPARγfl/fl) (male) had normal skin and hair phenotypes. In contrast, PPARγ (f/f)/Cre mice (PPARγ stem cell KO mice; Figure 5a and b) had normal skin and hair phenotype at birth; however, starting at ~3 months the mice displayed progressive hair loss and increasing scratching behavior. The PPARγ-deficient mice also appeared to be slightly smaller than wild-type littermates at ~3 months. The scratching behavior of these mice is reminiscent of pruritus reported by LPP patients. Previous studies (Newton et al., 1987; Mehregan et al., 1992; Tan et al., 2004; Mobini et al., 2005) have shown that patients with active LPP almost always report some degree of scalp itching and that pruritus is the main symptom of human LPP. A close-up of the skin of PPARγ KO mouse (Figure 5c) displays flakiness, mild erythema, and a region with complete loss of follicular orifices. In advanced stage of disease, the skin of PPARγ KO mice appeared flaky and crusty and the remaining sparse hair was matted and often difficult to remove at the time of necropsy (data not shown).
Hematoxylin and eosin-stained paraffin sections of the skin from control littermates showed normal skin and hair follicle histology (Figure 5d). In contrast, the PPARγ KO mice (Figure 5e), showed an obvious difference in the morphology of hair follicles. H&E-stained sections of the skin of these mice showed hyperkeratosis and follicular ostia that appeared dilated and plugged. There was increased interstitial inflammation. As shown in Figure 6, the PPARγ KO mice display several histopathological features of scarring alopecia. Dystrophic hair follicles, follicular plugging, and perifollicular fibrosis were observed. In some cases, the sebaceous glands appeared dystrophic and contiguous with the follicular plugs. The dermis had progressively increasing cellularity with interstitial inflammation. Perifollicular inflammation in the form of a mixed mononuclear infiltrate consisting of lymphocytes, plasma cells, macrophages, and mast cells was also observed (Figure 6).
In advanced disease, H&E staining of skin of PPARγ KO mice showed dystrophic hair follicles with sebaceous gland atrophy (Figure 7a). Follicular scarring, in which fibrous connective tissue strands run perpendicular to the epidermis from remnants of dystrophic hair follicles, was also observed (Figure 7b–d). The scarring seen in PPARγ KO mouse skin (Figure 7) is similar to scarring reported in the skin of Asebia mouse (Sundberg et al., 2000, Skinbase–the mutant mouse skin database http://eulep.pdn.cam.ac.uk/~skinbase/). In contrast, the skin of normal mice (Figure 7a), showed well-formed hair follicles.
Microarray analysis of the skin of PPARγ KO mice confirmed the histological data and showed a dramatic increase in gene expression of chemokines (MIP1a, MIP1b, CCR1, CD14), extracellular matrix-associated proteins (MMP12, MMP8, TIMP2), and apoptosis-related genes (CASP3, DUSP11; Table S2), suggesting the activation and involvement of macrophages and T-lymphocytes. The data also suggest that apoptosis and substantial matrix remodeling may characterize the loss of hair follicles in PPARγ KO mice. In addition, as seen with human LPP tissue, microarray analysis of the skin of PPARγ KO mice showed decreased expression of lipid metabolic and peroxisomal genes (Table S3). These observations in the PPARγ KO mice support our studies with human tissue and suggest that decreased expression of lipid metabolic and peroxisomal genes is likely a direct consequence of loss of PPARγ expression.
Intriguingly, as seen with LPP tissue, microarray data of PPARγ KO mice showed a 64-fold increase in expression of prostaglandin synthase (PTGS2 or COX2) and a 97-fold increase in the expression of the lipid oxidation enzyme 5-LO-activating protein (ALOX5AP; Table S2). These data support the Ingenuity Pathway data in LPP (Figure S4) and suggests the existence of a negative feed-back loop between PPARγ COX2 and ALOX5A.
The animal data confirms our observations in LPP tissue and suggest that the loss of PPARγ expression activates the proinflammatory lipid metabolic pathways that in turn induce the inflammatory response and permanent hair loss in scarring alopecia. The similarity in histopathology between the PPARγ KO mice and human LPP (perifollicular lymphocytic inflammation, fibrosis, scarring, and permanent hair loss) suggests a crucial role for PPARγ in the pathogenesis of scarring alopecia.
Primary CAs are viewed as immune disorders caused by an inflammatory attack in the permanent portion (infundibular and/or isthmic region) of the hair follicle (Stenn et al., 1999; Cotsarelis and Millar 2001; Mirmirani et al., 2005; Price, 2006). To understand the molecular pathogenesis of these poorly understood hair disorders, we carried out global geneexpression analysis of paired unaffected and affected scalp biopsies from LPP patients compared to normal controls. The majority of upregulated genes in affected LPP tissue were either required for tissue remodeling and apoptosis or were inflammatory genes as anticipated from histopathology. The microarray data also revealed decreased expression of multiple genes required for fatty acid β-oxidation, fatty acid desaturation, cholesterol biosynthesis, and peroxisome biogenesis in LPP scalp tissue.
Intriguingly, the increase in expression of inflammatory genes was seen in affected and not in unaffected tissue. In contrast, the decreased expression of lipid metabolic genes was seen to a greater extent in unaffected compared to affected LPP tissue. These data suggest that the lipid metabolic changes likely represent early or primary events in disease pathogenesis. Our data also suggests that the lipid metabolic changes may be the cause rather than the effect of the inflammatory response in LPP. Lipid analysis by gas chromatography showed a significant increase in arachidonic acid in affected LPP compared to unaffected tissue, thereby raising the possibility of arachidonate metabolites such as leukotrienes and prostaglandins acting as proinflammatory signals in LPP.
Indeed, biochemical pathway and promoter analysis of the differentially regulated genes identified PPARγ as an upstream regulator of the changes in LPP. These data also revealed a negative regulatory loop between PPARγ and prostaglandin endoperoxide synthase 2 (COX2). The observation that PPARγ agonists induce the expression of the COX2 gene in ORS cells, provided experimental proof to support the bioinformatics data. We identified a similar negative feed-back loop between PPARγ and 5-LO (data not shown). Interestingly, both COX2 and ALOX5AP are significantly upregulated in both LPP and in the PPARγ KO mouse, suggesting a role for these pathways in the pathogenesis of scarring alopecia. The first rate-limiting step in the conversion of arachidonic acid to prostaglandins is catalyzed by PTGS2 (COX2), an enzyme that is associated with biologic events such as injury, inflammation, and proliferation. The 5-LO activating protein (ALOX5AP) is necessary for activation of 5-LO that converts arachidonic acid into leukotrienes, which are eicosanoid lipid mediators of inflammation. Thus, elevated COX2 and 5-LO may lead to the increased production and secretion of prostaglandins and leukotrienes.
We have shown that there is a significant decrease in expression of PPARγ but not PPARα or PPARδ in LPP. In vitro studies in cultured human keratinocytes showed that specific PPARγ agonists induced the expression of peroxisomal genes that are downregulated in LPP. Finally, hair follicle stem cell-specific deletion of PPARγ in mice causes scarring alopecia with symptoms that parallel human disease. It is interesting to note that in mice with a PPARγ deficiency localized to the epidermis, patchy hair loss was noted in older animals (Mao-Qiang et al., 2004). Our data shows that targeted KO of PPARγ in the stem cells of the bulge causes scarring alopecia. These results demonstrate that it is the dysfunction of stem cells caused by loss of PPARγ signaling rather than the deletion of stem cells per se that likely triggers the pathogenesis of CA. Together, these data provide compelling evidence that PPARγ deficiency leads to the accumulation of proinflammatory lipids generated by 5-LO and COX2 pathways that trigger the pathogenesis of scarring alopecia.
A likely model for pathogenesis of primary CA is shown (Figure 8). In normal pilosebaceous units, PPARγ binds to PPAR response elements and regulates peroxisome biogenesis and lipid metabolic genes thereby maintaining lipid homeostasis. PPARγ also has anti-inflammatory effects and modulates the inflammatory response by regulating the expression of proinflammatory lipid synthetic enzymes (COX2, 5-LO), cytokines, chemokines, and adhesion molecules. In primary CA, environment, diet, or genetic factors likely suppress PPARγ expression. The PPARγ deficiency or dysfunction in LPP patients induces peroxisome loss, disturbs lipid homeostasis, and deregulates lipid metabolism in the pilosebaceous unit. This causes the accumulation of proinflammatory lipids that in turn trigger chemokine/cytokine expression, recruit lymphocytes and macrophages, and causes tissue damage (lipotoxicity) and activates a lipid-mediated programmed cell death (lipoapoptosis), thereby contributing to permanent hair loss and scarring in LPP. Altered sebaceous and epidermal lipids have been suggested to be the cause of skin lesions seen in the Asebia mouse (Wilkinson and Karasek MA, 1966; Brown and Hardy, 1988; Sundberg et al., 2000). Thus, the accumulation of proinflammatory lipids in the pilosebaceous units of LPP tissue and PPARγ KO mice may induce an inflammatory response because of lipotoxicity and contribute to CA pathogenesis. Recent studies (Wan et al., 2007) have shown that PPARγ deficiency causes lipid accumulation in the lactating mammary gland. These studies showed that PPARγ deficiency induces the production of inflammatory lipids in milk that causes hair loss in nursing pups.
Accumulating evidence suggests that environmental pollutants disrupt the body’s homeostatic controls through deregulation of critical pathways involved in lipid metabolism or energy balance. The genetic or environmental mechanisms that initiate loss of PPARγ signaling in LPP are not fully understood. However, the skin as the outermost barrier of the body is exposed to various sources of environmental toxins such as dioxin or dioxin-like compounds that are known to inhibit the expression of PPARγ and all lipogenic genes that are transcriptionally activated by PPARγ (Liu and Jefcoate, 2006). Dioxin-like compounds exert their biologic effects via the AhR, a ligand-dependent transcription factor. The AhR is a ligand responsive transcription factor that belongs to the basic helix-loop-helix Per-Arnt-Sim (bHLH-PAS) superfamily. Upon binding ligands such as dioxins, the AhR mediates an adaptive metabolic response by upregulating the transcription of a battery of xenobiotic metabolizing enzymes, including the cytochromes P450, CYP1A1, CYP1A2, and CYP1B1 (Schmidt and Bradfield, 1996). Although epidemiologic links between dioxins and scarring alopecia are lacking, it is interesting to note that our microarray data showed the increased expression of dioxin-inducible Cytochrome P1-450 (CYP1A1) gene in both unaffected and affected LPP tissue, suggesting the constitutive activation of AhR. In addition, toxicity analysis of unaffected and affected LPP tissue with IPA-Tox(TM) identified the Aryl Hydrocarbon Receptor Signaling and Xenobiotic Metabolism as the most significant toxicity pathways involved in LPP.
Low levels of dioxin exposure have become a focus of interest in the context of other PPARγ involved diseases such as adult-onset diabetes (Remillard and Bunce, 2002). Chronic low-dose exposure may cause the accumulation of dioxins in lipid-rich regions such as sebaceous glands and at a certain threshold level (which may be reached at middle-age) may cause the loss of PPARγ expression and scarring alopecia in susceptible individuals.
Whether the decrease in PPARγ expression in LPP is indeed the result of exposure to an environmental toxin or is induced by dietary or genetic factors will require further study. However, activation of PPARγ signaling by PPAR agonists could be effective in alleviating the deleterious effects of inflammatory lipid accumulation in the pilosebaceous unit. Thiazolidinediones, that influence free fatty acid flux, are known to activate PPARγ (Berger et al., 2005). Our data show that Rosi, Cig, Tro, and Pio induce the expression of peroxisomal gene expression in LPP. Thus, it seems likely that the stimulation of PPARγ activity by specific agonists could potentially inhibit the deleterious effects of proinflammatory lipids such as inflammation, loss of hair follicles, and scarring seen in LPP. Synthetic PPARγ ligands are currently used therapeutically in the treatment of dyslipidemias, type 2 diabetes, cardiovascular disease, and metabolic syndrome (Berger et al., 2005). Alternatively, specific inhibitors of 5-LO pathway or specific COX2 inhibitors may provide previously unreported therapeutic strategies for the treatment of scarring alopecias.
In summary, we show here that the loss of PPARγ expression in the stem cells of the bulge results in progressive hair loss, sebaceous gland atrophy, scarring, and inflammation in a mouse model. These observations clearly implicate primary defects of PPARγ in the generation of scarring alopecia. We believe the human disorder is a consequence of PPARγ deficiency that in turn induces a series of changes in key metabolic pathways that induce the production of proinflammatory lipids. We have shown that perturbation of lipid metabolism induced by PPARγ deficiency, most likely an acquired condition, results in inflammation-induced destruction of the pilosebaceous gland in CA. These effects reveal a crucial role for PPARγ in the maintenance and normal functioning of the pilosebaceous unit and suggest that loss of this signaling pathway may be responsible for the pathogenesis of CA. To our knowledge, a link between PPARγ deficiency, deregulated lipid metabolism and hair disorders in humans is previously unreported. Ongoing studies in the PPARγ KO mice should help to further refine this hypothesis to genetically define disease progression in primary cicatricial alopecia.
These observations provide a previously unreported framework for understanding the role of PPARγ in the pathophysiology of primary cicatricial alopecia. Indeed, PPARγ agonists may represent a potential new therapeutic strategy in the treatment of these disorders.
Scalp biopsies were obtained from “the active border” (an area with inflammation and retained but decreased hair follicles) from patients with a clinical diagnosis of lymphocyte-mediated LPP and who were seen at the clinics and University Hospitals of Cleveland or University of California at San Francisco. All patients had active disease with symptoms of itching, burning, or pain, and with progressive hair loss, positive pull test, and evidence of inflammation. Patients were 18 years or older and were able to give informed consent. These patients were evaluated in a standard manner. This evaluation included a medical history, detailed hair questionnaire, treatment history, examination of hair, scalp, and skin, scalp photographs, and two 4 mm scalp biopsies—one from affected and another from clinically unaffected scalp. Scalp biopsy specimens from healthy volunteers were included as controls. All biopsies were done under the approval of the University Hospitals Case Medical Center Institutional Review Board and with the written, informed consent of patients and volunteers. The study was conducted according to Declaration of Helsinki Principles. All tissue samples were stored at −80 °C until processed. These biopsies were utilized for total RNA extraction, microarray analysis, real-time PCR, and immunofluorescence.
Total RNA from each biopsy was extracted using TRIzol (Life Technologies Inc., Gaithersburg, MD) as per the manufacturer’s instructions, followed by purification using RNeasy Mini columns (Qiagen Inc., Valencia, CA). The RNA was quantitated by spectrometry and used for microarray and real-time PCR experiments.
Two sets of microarray experiments were performed. In the first set, we analyzed samples of fresh frozen scalp tissue biopsied from patients with LPP (n = 20, pooled) and compared the pattern of gene expression against normal (control) scalp tissue (n = 20, pooled) by interrogating the Affymetrix GeneChip oligonucleotide array Human U133A 2.0 (Affymetrix, Santa Clara, CA). To confirm our observations with the first dataset and to identify early changes in LPP, a second microarray experiment was performed with paired unaffected (N = 10) and affected (N = 10) LPP scalp biopsies and compared with normal scalp tissue (N = 10). Unlike the first experiment, this set of samples were not pooled but were individually analyzed with the Human U133A 2.0 array. This array represents 18,400 transcripts and variants, including 14,500 well-characterized human genes. Fluorescent Cy3- or Cy5-labeled cDNA (Amersham Pharmacia Biotech, Piscataway, NJ) was synthesized from 50 to 100 µg total RNA, using oligo-dT-primed polymerization with SuperScript II reverse transcriptase (Life Technologies). Hybridization to the oligonucleotide arrays and subsequent washing and detection was performed as described in the Affymetrix Expression Analysis Technical Manual (Affymetrix). Array images were acquired using a GeneChip Scanner 3000 (Affymetrix) and analyzed with Genechip Operating Software (GCOS). The image from each GeneChip was scaled such that the average intensity value for all of the arrays is adjusted to a target intensity of 500 to take into account the inherent differences between the chips and their hybridization efficiencies. The Affymetrix program Netaffyx and the Online Mendelian Inheritance In Man were used to identify the functional significance, cellular location, and the role of genes in various biological and metabolic processes.
Biologically relevant pathways were constructed using IPA application (Ingenuity Systems, www.ingenuity.com). Detailed methods and guide to interpreting IPA data are provided in the Supplemental Methods section. The differentially expressed genes containing Affymetrix identifiers and associated fold-change values were uploaded for IPA analysis. The genes that had a fold change greater than 2.0 were included in this analysis. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. These genes, called focus genes, were overlaid onto a global molecular network developed from information contained in the Ingenuity Pathways Knowledge Base. Networks of these focus genes were then algorithmically generated based on their connectivity. The Functional Analysis of a network identified the biological functions and/or diseases that were most significant to the genes in the network. Fischer’s exact test was used to calculate a P-value determining the probability that each biological function and/or disease assigned to that dataset is because of chance alone. The program also computes a score for each network according to the fit of the network to the set of focus genes.
FAM-labeled PCR primers and TaqMan hydrolysis probes for all target genes and 18S rRNA was purchased from Applied Biosystems (Foster City, CA). Real-time PCR was performed on an ABI Prism 7700 Sequence Detection System (Applied Biosystems) according to the recommendation of the manufacturer. The target gene expression in LPP and control samples was quantitated by the comparative computed tomography method as described in the ABI Prism 7700 Sequence Detection System manual (PE Biosystems).
Scalp tissue specimens were cut horizontally or vertically and serial sections were prepared using a cryostat (Leica Microsystems Inc., Bannockburn, IL). The slides were fixed in acetone and stored at −80 °C until immunostaining was performed. For the morphological detection of peroxisomes, horizontal and vertical sections of scalp biopsies were stained using the SelectFX Alexa Fluor 488 peroxisome labeling kit (Invitrogen-Molecular Probes, Carlsbad, CA), following the recommendation of the manufacturer and visualized by indirect immunofluorescence light microscopy. The kit utilizes rabbit antibodies directed against the peroxisomal membrane protein 70 (PMP-70), which is a high abundance integral-membrane component of peroxisomes. In some instances, the slides were counterstained with the nuclear stain DAPI. Antigen-antibody complexes were detected under a Carl Zeiss Axioskop FL microscope, using Alexa Fluor 488 goat anti-rabbit IgG antibody. The approximate absorption and fluorescence emission peaks of the Alexa Fluor 488 dye conjugate are 495 and 519 nm and the labeling was observed using standard fluorescein filter sets. The slides were cover-slipped with Vectashield mounting medium (Vector Labs Inc., Burlingame, CA).
Tissue lipids were extracted by the Folch method (Folch et al., 1957). The chloroform phase containing lipids was collected, dried under nitrogen, and subjected to methylation. Fatty acid methyl esters were prepared by standard methods using BF3/methanol reagent (14% Boron Trifluoride). Fatty acid methyl esters were analyzed by gas chromatography using a fully automated HP5890 system equipped with a flame-ionization detector (Morrison and Smith, 1964). The chromatography utilized an Omegawax 250 capillary column. Peaks were identified by comparison with fatty acid standards (Nu-chek-Prep, Elysian, MN), and the area and its percentage for each resolved peak were analyzed using a PerkinElmer M1 integrator.
Human hair follicle ORS cells were obtained from ScienCell (San Diego, CA) and grown in mesenchymal stem cell medium consisting of 500 ml of basal medium, 25 ml of fetal bovine serum, 5 ml of mesenchymal stem cell growth supplement, and 5 ml of penicillin/ streptomycin solution. Human keratinocyte cell line were cultured in Dulbecco’s modified Eagle’s medium containing 10% fetal bovine serum under 5% CO2 at 37 °C. The different PPAR agonists were added in triplicate in dimethylsulfoxide (<0.1% by volume) for 48 hours to evaluate their effects on PEX and COX2 gene expression by real-time PCR.
PPARγ stem cell KO mice (PPARγ (f/f)/Cre) mice were generated by intercrossing mice carrying floxed alleles of PPARγ (He et al., 2003) were crossed with a Cre-transgenic line K15-CrePR1 (Liu et al., 2003), expressing Cre recombinase under the control of mouse keratin complex 1, acidic, gene 15 promoter. Both mouse strains were purchased from The Jackson Laboratory, Bar Harbor, Maine. Littermates lacking the K15-Cre transgene were used as controls. All experimental procedures were conducted in accordance with the Guide for Care and Use of Laboratory Animals of the National Institutes of Health, and were approved by the Case Western Reserve University IACUC.
PCR genotyping was carried out by using the following primers. For the Cre transgene, primers oIMR1084 (5′-GCGGTCTGGCAG TAAAAACTATC-3′) and oIMR1085 (5′-GTGAAACAGCATTGCTGT CACTT-3′) yield a 100 bp fragment. For identifying the floxed allele, the following primers were used: oIMR1934 (5′-TGTAATGG AAGGGCAAAAGG-3′) and oIMR1935 (5′-TGGCTTCCAGTGCA TAAGTT-3′) amplify a 214 bp product from the wild type and a 250 bp product from the mutant (floxed) allele. Genomic DNA was amplified by 35 cycles of 94 °C for 20 s, 60 °C for 30 s, and 72 °C for 55 s. Total RNA was isolated from mouse tissues by using TRIzol (Invitrogen-Molecular Probes). Reverse transcription was performed with SuperScript (Invitrogen-Molecular Probes, Carlsbad, CA). Sense (5′-GTCACGTTCTGACAGGACTGTGTGAC-3′) and antisense (5′-TATCACTGGAGATCTCCGCCAACAGC-3′) primers were designed to anneal to regions in exons A1 and 4 of PPARγ1, respectively, which distinguish the full-length (700-bp) and recombined (300-bp) transcripts. PCR was performed by 40 cycles of 94 °C for 20 s, 60 °C for 30s, and 72 °C for 60s.
This work was funded by the Cicatricial Alopecia Research Foundation (CARF), North American Hair Research Society, Dermatology Foundation, the American Skin Association, and a pilot and feasibility grant from the Case Western Reserve University Skin Diseases Research Center (NIH-NIAMS; P30-AR-39750). We acknowledge the contribution of the Gene Expression Array Core Facility of the Comprehensive Cancer Center of Case Western Reserve University and University Case Medical Center (P30 CA43703). We thank Dr Kord Honda (University Hospitals Case Medical Center) for interpretation of histopathology slides and Dr Larry L Swift (Vanderbilt University School of Medicine) for lipid analysis.
CONFLICT OF INTEREST
The authors state no conflict of interest.