Hairless is a histone H3K9 demethylase

Abstract

The hairless (HR) protein contains a Jumonji C (JmjC) domain that is conserved among a family of proteins with histone demethylase (HDM) activity. To test whether HR possesses HDM activity, we performed a series of in vitro demethylation assays, which demonstrated that HR can demethylate monomethylated or dimethylated histone H3 lysine 9 (H3K9me1 or me2). Moreover, ectopic expression of wild-type HR, but not JmjC-mutant HR, led to pronounced demethylation of H3K9 in cultured human HeLa cells. We also show that two missense mutations in HR, which we and others described in patients with atrichia with papular lesions, abolished the demethylase activity of HR, demonstrating the role of HR demethylase activity in human disease. By ChIP-Seq analysis, we identified multiple new HR target genes, many of which play important roles in epidermal development, neural function, and transcriptional regulation, consistent with the predicted biological functions of HR. Our findings demonstrate for the first time that HR is a H3K9 demethylase that regulates epidermal homeostasis via direct control of its target genes.—Liu, L., Kim, H., Casta, A., Kobayashi, Y., Shapiro, L. S., Christiano, A. M. Hairless is a histone H3K9 demethylase.

Methylation of lysine residues within the N-terminal tails of core histones is recognized as a major epigenetic event in gene regulation during normal development and in human diseases (1–9). The enzymes responsible for the establishment and removal of histone methylation marks have been the subject of intensive research in recent years. More than 50 histone methyltransferases, which can establish a methylation mark based on in vitro and/or in vivo studies, have been reported to date (10). The existence of histone demethylases (HDMs), which remove a methylation mark, however, remained elusive until the isolation of the first HDM in 2004 (11). Currently, two major families of HDMs have been identified, including the LSD1 amine oxidase family and the Jumonji C (JmjC) domain-containing demethylases (12). The JmjC domain is a signature motif conserved among >30 human proteins, many of which have been shown to possess HDM activity (13).

Hairless (HR) is a JmjC protein that is highly expressed in the skin and brain (14). The HR gene is essential for skin homeostasis, since both humans and mice lacking HR activity suffer from congenital hair loss and defects in epidermal proliferation and differentiation (15, 16). The molecular mechanisms underlying these processes, however, remain elusive. Studies from our laboratory and others have demonstrated that that HR is a versatile transcription factor that can act not only as a repressor but also as an activator in a context-dependent manner (14, 17–19). Here, we performed studies to test the potential demethylase activity of HR and showed for the first time that HR can effectively demethylate monomethylated or dimethylated histone H3 lysine 9 (H3K9me1 or H3K9me2, respectively). To uncover the molecular targets and pathways of HR, we performed chromatin immunoprecipitation and deep sequencing (ChIP-Seq) experiments to identify genomic targets with which HR interacts directly. The majority of the HR target genes identified by our ChIP-Seq analysis plays important roles in biological processes, including hair biology, neural activity, cell cycle, and transcriptional regulation, which is consistent with known or predicted functions of HR.

MATERIALS AND METHODS

Cells, recombinant proteins, and expression plasmids

HEK293 cells and HeLa cells were obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA) and maintained according to the provider's protocol. Normal human keratinocytes (NHKs) were isolated from neonatal foreskin and cultured using the CnT-07 culture medium kit (CELLnTEC, Bern Switzerland). Recombinant partial HR protein containing the JmjC domain was obtained from Abnova (Walnut, CA, USA). Recombinant premethylated H3K9me1, H3K9me2, H3K9me3, and K3K4me3 were obtained from Active Motif (Carlsbad, CA, USA). Plasmids encoding Flag-JMJD1A and Flag-HR were constructed by PCR amplification of the complete human cDNAs that were subsequently inserted into a N-terminal Flag-tagged pCMV vector (Sigma, St. Louis, MO, USA). The pCMV-Flag-HR_mutated was constructed by introducing 3-point mutations within the metal-biding motif of the HR JmjC domain (C1007G, E1009G, H1125P). Patient mutation constructs were prepared by introducing single mutations at D1012N or V1056M, respectively.

Purification of Flag-HR and Flag-JMJD1A

Flag-tagged HR and JMJD1A were expressed in HEK293 cells and were purified using anti-Flag M2-agarose beads (Sigma) following the manufacturer's instructions. Briefly, Flag-HR and JMJD1A plasmids were transfected into HEK293 cells using FuGene 6 (Roche, Indianapolis, IN, USA). At 48 h after transfection, cells were washed with PBS and lysed with flag-lysis buffer (50 mM Tris-HCl, pH 7.3; 137 mM NaCl; 1 mM Na₃VO₄; 1 mM NaF; 10% glycerol; 1 mM EDTA; 0.1% Sarkosyl; and 1% Triton-X100) supplemented with 1× protease inhibitor cocktail (Roche) and 1 mM phenylmethylsulfonyl fluoride. The cell lysate was incubated with M2 beads for 3 h at 4°C. After spinning down, the beads were washed with flag-lysis buffer 3 times and then with BC100 (20 mM Tris-HCl, pH 7.6; 100 mM NaCl; 20% glycerol) 3 times. The recombinant enzymes were eluted from the beads using 0.15 μg/μl of flag peptide solution (in BC100 supplemented with 0.1 mM DTT). The purity of the recombinant protein was assessed by SDS-PAGE gel separation and Coomassie blue staining, followed with a band profile analysis using the Image Lab software (Bio-Rad, Hercules, CA, USA; Supplemental Fig. S1).

In vitro demethylation assays

For in vitro demethylation assays, 1 μg of either recombinant histone peptides or native total histones isolated from HeLa cells were mixed with different amounts of Flag-HR or -JMJD1A protein. For a negative control, Flag peptide in BC100 buffer was used instead of the enzymes. The reaction mixture contained 50 mM HEPES (pH 7.5), 1 mM α-ketoglutarate, 2 mM ascorbic acid, and FeSO₄ (100 μM for HR or 50 μM for JMJD1A), which was incubated at 37°C for 3.5 h, as reported previously (20). Following demethylation reaction, the reaction mixture was mixed with sample buffer and denatured at 95°C for 10 min, separated on 4–15% SDS-PAGE gel (Bio-Rad), and transferred to PVDF membranes following standard procedures. Each target protein was detected using appropriate primary antibodies as specified in Supplemental Table S1. Western blots were detected with SuperSignal West Dura chemiluminescent system (Thermo Fisher Scientific, Rockford, IL, USA) following the manufacturer's instructions. Images were scanned and quantified using the ImageJ program (U.S. National Institutes of Health, Bethesda, MD, USA) to determine the changes in H3K9 methylation. Results from 5 repeats for each demethylation experiment were used for statistical analyses by t test to determine the significant differences between different experimental conditions.

Immunofluorescence staining

For immunofluorescence staining, cells were plated on glass coverslips and cultured for 24 h. After washing with PBS, cells were fixed in 4% paraformaldehyde for 10 min. The cells were then washed once with cold PBS and permeabilized for 5 min with cold PBS containing 0.2% Triton X-100. Permeabilized cells were then washed 3 times with blocking buffer (1% BSA in PBS) and blocked for 30 min before being incubated with primary antibodies for 1 h in a humidified chamber. After 3 consecutive 5-min washes with PBS, cells were incubated with secondary antibodies for 1 h before being washed with PBS and stained with 4,6-diamidino-2-phenylindole dihydrochloride (DAPI) in PBS. Cells were washed again twice with PBS and then mounted in fluorescent mounting medium (Dako, Carpinteria, CA, USA). Images were acquired using a fluorescence confocal microscope (Zeiss, Thornwood, NY, USA).

ChIP, ChIP-Seq, and PCR analysis

The EZ-ChIP kit (Millipore, Billerica, MA, USA) was used in all ChIP studies following the manufacturer's protocol. Briefly, 1 × 10⁷ HEK293 cells (control or Flag-HR-transfected) were cross-linked with 1% formaldehyde and quenched with 0.125 M glycine. Cells were washed with cold PBS and then lysed with ChIP lysis buffer (Millipore) supplemented with protease inhibitor cocktail (Sigma). Cell lysate was sonicated using a Bioruptor sonicator (Diagenode, Denville, NJ, USA) for 8 cycles of 30 s on and 45 s off. Chromatin (∼200 μg) was precleared with Protein A agarose beads and incubated with 4 μg of histone antibodies (anti-H3K9me1 or anti-H3K9me2; see Supplemental Table S1) or rabbit serum overnight at 4°C. Protein A agarose beads (60 μl; Millipore) were then added to the lysate and incubated for 1 h to precipitate the chromatin-antibody complex, which was washed following the manufacturer's protocol. The resulting DNA-protein immunocomplex was reverse cross-linked at 65°C for 6 h. DNA was purified and dissolved in 50 μl H₂O. Input DNA and ChIP DNA were used in subsequent PCR analysis. The primers for ChIP PCR are shown in Supplemental Table S2.

For ChIP-Seq analysis, Flag-HR-transfected cells were used for ChIP assay using an anti-Flag antibody (Supplemental Table S1) to purify HR-bound chromatin DNA. ChIP'ed DNA fragments were subjected to deep sequencing on the Illumina Genome Analyzer IIx system (Illumina, San Diego, CA, USA) to generate ∼3 × 10⁷ sequence reads, which were mapped to the human genome (hg18 build) using the DNANexus software (DNANexus, Mountain View, CA, USA). All mapped DNA sequences were used for peak analysis using the DNANexus software to identify HR-enriched genomic regions following the provider's guidelines. Mapped reads from the ChIP input control sample were used as background for defining sequence peaks in mapped reads derived from the ChIP samples. The ChIP candidate threshold was set at 30 in order to avoid biases due to low coverage and to ensure a more accurate false discovery rate calculation. A minimum ratio of confident to repetitive mappings in peak was set at 3.0 to avoid peak-calling in locations with too many repetitive mappings. The experiment to background enrichment threshold was set at 3.0.

Structural modeling of HR

The sequence of the JmjC domain from HR was determined from the alignment with Jumonji domains published previously (20, 21) and submitted to homology modeling with the program Modeller (22). Although low sequence identity in some peripheral regions yielded loops with poorly modeled structure, the active site region is well-defined, clearly positioning the putative iron-coordinating ligands close in space.

RESULTS

HR contains a JmjC domain and demethylates H3K9me1 in vitro

The major functional domains of HR include a single zinc finger domain, multiple LXXLL motifs that mediate protein-protein interactions, and, interestingly, a JmjC domain in its C terminus (Fig. 1A and refs. 14, 15), leading us to postulate that HR may possess HDM activity. Because HR is structurally related to JMJD1A (also known as KDM3A), a well-defined JmjC HDM (20), we first modeled the sequence of HR on the basis of the predicted structure and functional domains of JMJD1A using the Modeller homology modeling program (22). Jumonji domains include a metal binding motif, H-X-E/D-Xn-H, which coordinates iron, a critical component of the active site (8, 23). Of the iron ligands, the H-X-E/D segment is located in the loop region between β strands 2 and 3, whereas the final conserved histidine is in β strand 7 (21). On the basis of sequence alignments and its homology model, we found that HR contains a similar C-X-E/D-Xn-H motif that differs by the substitution of a cysteine residue for the first histidine. The known propensity of cysteine to coordinate iron atoms (24) suggests that this site could potentially function as an active metal binding pocket, which could underlie HDM activity by HR (Fig. 1B, indicated by asterisk).

Figure 1.

HR is a JmjC demethylase. A) Schematic depiction of major functional domains of the human HR protein, including a JmjC domain in its C terminus. Two APL patient mutations (D1012N and V1056M) are also indicated. B) Left panel: homology model of the HR protein JmjC domain structure, produced on the basis of homology to JmjC domains of known structure, revealed that the putative active site C-E-H residues are close in space, likely forming an atypical metal-binding pocket (indicated by asterisk). Right panel: aa 1012 and 1056 are indicated on the close-up view of the JmjC domain. C) In vitro demethylation assay using recombinant partial HR protein containing the JmjC domain and premethylated histone H3 peptides: H3K9me1, H3K9me2, H3K9me3, and H3K4me3. D) In vitro demethylation assay using recombinant partial HR and native histone proteins purified from HeLa cells. E) Representative Western blot analysis results illustrating the demethylase activity of HR and JMJD1A. Iron concentration is 50 μM for JMJD1A and 100 μM for HR. Reaction without enzyme input was included as a negative control. F) Graphic illustration of the relative HDM activities of HR and JMJD1A on H3K9me1 at 3 different amounts (10, 20, and 30 μg) of enzyme input. The y axis depicts the relative level of H3K9me1 (determined by the ratio of the intensity between H3K9me1 and total H3), as compared to the no enzyme input control sample (which was assigned a relative value of 1). Results from 5 repeats for each demethylation experiment were used for statistical analyses by t test to determine the significance of differences between different experimental conditions. *P < 0.05, **P < 0.01.

The presence of a JmjC domain, albeit with an atypical metal binding motif, prompted us to test whether HR possesses functional HDM activity. Given the structural similarity between HR and JMJD1A, which is a known H3K9me1 and H3K9me2 HDM, we postulated that HR might utilize the same substrates for its putative HDM activity. To test this possibility, we first performed in vitro demethylation assays using a recombinant partial HR protein containing the JmjC domain (Abnova) and tested its ability to demethylate various premethylated recombinant histone H3 substrates. Among the different forms of premethylated H3K9, we found that the partial HR protein demethylated H3K9me1 in a dose-dependent manner, while having a relatively weaker effect on H3K9me2 and no detectable demethylation effect on H3K9me3 or H3K4me3 (Fig. 1C and Supplemental Fig. S2). The H3K9me1 demethylase activity of HR was further verified in separate experiments using total native histones from HeLa cells in place of the recombinant H3 peptides (Fig. 1D).

To further interrogate the spectrum of HR HDM activity, we performed additional demethylation experiments by using full-length HR to recapitulate the intact enzymatic structure. Flag-tagged full-length HR protein was isolated from Flag-HR-overexpressing HEK293 cells. Similarly, Flag-JMJD1A protein was prepared and used in parallel demethylation experiments as a positive control for demethylase activity. Total native histones from HeLa cells were used as substrates in demethylation reactions. We determined that the demethylase activity of HR appeared to be more effective at a slightly higher concentration of 100 μm of Fe²⁺, whereas 50 μm of Fe²⁺ concentration was optimal for JMJD1A activity (unpublished observations). The results confirmed that full-length HR can effectively reduce the level of H3K9me1 (Fig. 1E) and also H3K9me2 (Supplemental Fig. S2A) in total native histones. While JMJD1A displayed dose-dependent demethylation activity toward H3K9me1, the demethylase activity of HR seemed to depend on an optimal enzyme to substrate ratio (illustrated in Fig. 1F). Excess HR protein may be rate-limiting for the reaction because they might compete for cofactors. Alternatively, the atypical metal-binding motif in the HR JmjC domain might require other cofactors for its enzymatic activity, which was not present in our in vitro demethylation reactions. Notably, however, HR displayed little demethylation activity toward H3K9me3 (Supplemental Fig. S2B), consistent with its putative function as an H3K9me1 and H3K9me2 demethylase, as predicted by its structural similarity to JMJD1A.

Ectopic expression of hairless removes H3K9 methylation marks in cultured human cells

Our in vitro demethylation results prompted us to ask whether HR can demethylate H3K9 in cultured cells. To address this question, we transfected HeLa cells with either wild-type HR (pCMV-HR-Flag) or HR with mutations in its metal-binding motif (pCMV-HR_mut-Flag). At 48 h post-transfection, we performed immunofluorescence staining to detect colocalization of HR (green) and H3K9me1 (red). As shown in Fig. 2A, we observed a dramatic loss of H3K9me1 in cells expressing wild-type HR compared with nonexpressing cells (top panel, white arrowhead). In contrast, cells expressing the mutant HR maintained their level of H3K9me1 (Fig. 2A, bottom panel, yellow arrowhead), suggesting that mutations in the metal-binding motif of HR JmjC abolished its demethylase activity. Similarly, overexpression of wild-type HR led to active demethylation of H3K9me2 in cultured HeLa cells, as well (Supplemental Fig. S2C). In further support of the in vivo H3K9 demethylation activity of HR, we found that HR-expressing dermal papilla (DP) cells in anagen human hair follicles harbor little H3K9me1 (Fig. 2B), confirming our in vitro demethylation assays showing that HR is a bona fide H3K9 demethylase.

Figure 2.

Demethylation activity of HR in human cells. A) Detection of demethylase activity by wild-type HR (pCMV-Flag-HR) or mutant HR (pCMV-Flag-HR_mutated, with mutations in the putative metal-binding motif in the JmjC domain) in cultured human HeLa cells by immunofluorescence staining and colocalization of HR (green) and H3K9me1 (red). White arrowheads and circles highlight cells overexpressing wild-type HR with reduced genomic H3K9me1. Yellow arrowheads and circles highlight cells overexpressing mutant HR and normal genomic H3K9me1. B) HR-expressing dermal papilla (DP) cells in human anagen hair follicle (HF) display reduced levels of genomic H3K9me1. a–c) HR expression detected by immunofluorescence staining with DAPI (blue; a), anti-HR (green; b), and merged image (c) in HF. e–g) H3K9me1 methylation detected by immunofluorescence staining with DAPI (blue; e), anti- H3K9me1 (green; f), and merged image (g) in human HF. Dotted white line separates DP cells from surrounding matrix cells. d, h) Graphic illustration of the relative intensity of HR (d) and H3K9me1 (h) staining (green) vs. DAPI staining (blue) from a representative population of DP and matrix cells (indicated by the yellow dotted arrow in a and e) for each antibody. DP cell population is highlighted by a dotted yellow border. C) H3K9me1 demethylation by wild-type HR and patient-specific mutant HR found in patients with APL (D1012N and V1056M). Total histones from HeLa cells were used as the demethylation substrate. *P < 0.05.

Mutations in the HR JmjC domain abolish its HDM activity

Previous studies from our laboratory and others have established that various missense mutations in the HR JmjC domain are associated with atrichia with papular lesions (APL; refs. 25, 26). We were intrigued to ask whether such mutations exerted their pathological effects via the HDM activity of HR. After introducing a single-point mutation in the HR JmjC domain (to generate the mutations D1012N or V1056M, respectively, as depicted in Fig. 1A), we purified full-length Flag-HR harboring each mutation and performed demethylation assays as described above. The results shown in Fig. 2C revealed that both patient mutations markedly diminished the HDM activity of mutant-HR compared to wild-type HR, suggesting that loss of HR demethylase activity may underlie the clinical findings in a subset of APL patients with missense mutations in the HR JmjC domain.

Demethylation of target genes by HR in HEK293 cells

In previous reports and our expression profiling studies, Wise, Caspase14, and Mxi1 have been shown to be regulated by HR (16, 17). It is unknown, however, how HR modulates the expression of these target genes. In light of the novel function of HR as an H3K9 demethylase, we postulated that HR overexpression could lead to H3K9 demethylation in the target gene promoters or transcription regulatory elements (TREs). To test this hypothesis, we performed ChIP studies to compare the levels of H3K9 methylation in the promoter or TREs of Wise, Mxi1, and Caspase14 between HR-expressing and nonexpressing control HEK293 cells. We found that overexpression of HR resulted in a complete loss of H3K9me1 and a partial loss of H3K9me2 in the Wise promoter (Fig. 3A), suggesting that Wise is a direct target of HR demethylation. Likewise, HR overexpression led to a partial demethylation of H3K9 within the TREs of Mxi1 and Caspase14 (Fig. 3A).

Figure 3.

Demethylation targets of HR in human cells. A) Comparison of H3K9me1 and H3K9me2 levels in the promoter or TREs of Wise, Mxi1, and Caspase14 between HR-expressing and nonexpressing HEK293 cells by ChIP assays. For each gene, 4 different fragments (as schematically depicted on each gene map) were assessed by ChIP PCR. Intensity of each PCR product correlates positively with the level of H3K9 methylation in each fragment assessed by ChIP PCR. B) Effect of ectopic HR expression on its target gene regulation in HEK293 cells and NHKs, respectively. C, empty-vector transfected cells; HR, HR-transfected cells. *P < 0.05, **P < 0.01.

Concomitant with the demethylation by HR, we found that overexpression of HR in HEK293 cells led to a significant increase in the expression of Mxi1, a cell growth regulator, but almost complete repression of caspase14, which regulates keratinocyte differentiation and epidermal cornification (ref. 27 and Fig. 3B). In keratinocytes, ectopic expression of HR also repressed caspase14 expression, albeit to a lesser degree compared to HEK293 cells, likely because these cells have lower transfection efficiency than HEK293 cells, which is also supported by a much lower fold increase in HR expression in the keratinocytes (12-fold) than in the HEK293 cells (>600-fold) following transfection (data not shown). More intriguingly, ectopic HR expression had opposite effects on Wise gene expression between HEK293 cells (up-regulation) and keratinocytes (down-regulation) (Fig. 3B), exemplifying that HR functions in a context-dependent manner, which may involve other unknown cofactors. In contrast, siRNA-mediated knockdown of HR in keratinocytes augmented caspase14 expression (Supplemental Fig. S3). However, no significant effect was observed on Wise or Mxi1, probably due to a partial knockdown of HR activity (∼50%) rendered by the low transfection efficiency, as mentioned above (Supplemental Fig. S3). Taken together with prior studies showing that HR can exert opposing effects on target gene activity, where it sometimes activates and sometimes represses (18), these results confirm that HR is a versatile transcription factor functioning through H3K9 demethylation, and its impact on target gene activity may vary in a gene-specific and/or context-dependent manner.

Genomic target genes of HR

HR contains a single zinc finger domain that could facilitate its DNA-binding ability, which led us to postulate that HR may also modulate gene expression via interaction with its target genes (15, 28). To identify genomic targets with which HR physically interacts, we employed a ChIP-Seq approach using the Illumina Genome Analyzer IIx system. ChIP-Seq uncovered ∼46 enriched genomic regions, 25 of which are associated with the genes summarized in Table 1. These genes are clustered into 3 major functional categories, including hair biology, neural activity, and cell cycle and transcriptional regulation, which are consistent with the predicted biological functions of HR. Using real-time RT-PCR, we confirmed that ectopic expression of HR in HEK293 cells led to either up-regulation or down-regulation of the majority of these new HR target genes (Table 1). Further studies are warranted to elucidate whether HR may bind to these target genes directly or through its association with other transcription factors.

Table 1.

Summary of HR target genes identified by ChIP-Seq and validated by real-time RT-PCR

Target gene functional category	Name	Con	HR
Hair biology
COL6A1	Collagen, type VI, α 1	+	++++
COL25A1	Collagen, type 25, α 1	+	++++
FGF13	Fibroblast growth factor 13	+	++++
KDSR	KDS reductase	+	0.5
KRTAP9-9	Keratin-associated protein 9	ND	ND
Brain/neural
CTNND2	Catenin δ-2	+	++++
KCNIP1	K+ channel-interacting protein 1	+	+++
DYM	Dymeclin	+	++
MAGI2	Membrane-associated guanylate kinase	+	++
Transcription regulation/cell proliferation/cancer
PREX2	Phosphatidylinositol 3,4,5-trisphosphate Rac exchanger	+	++++
TAF7L	TAF7-like RNA polymerase II	+	++++
DLEC1	Deleted in lung and esophageal cancer	+	+++
CSNK2A2	Casein kinase II subunit α (Wnt molulator)	+	+++
FADD	FAS-associated death domain protein	+	+++
UBR2	Ubiquitin-protein ligase E3-α-2	+	+++
IBTK	Inhibitor of Bruton's tyrosine kinase	+	+
ANO1	Anoctamin 1	+	−
PVT1	Noncoding RNA gene	ND	ND
Other
GPD1L	GAPDH-like protein	+	++++
AGPAT4	1-acylglycerol-3-phosphate O-acyltransferase 4	+	+++
BOC	Brother of CDO	+	+++
NACC2	Nucleus accumbens-associated protein 2	+	0.5
IL1R2	Interleukin-1 receptor β	+	−
CYP2F1	Cytochrome P450, subfamily IIF, polypeptide	+	−
NCRNA00164	Nonprotein-coding RNA 164	ND	ND

Con, control HEK293 cells; HR, HR-overexpressing HEK293 cells; ND, not determined; +, normalized gene expression in Con cells; ++, 1.5- to 2-fold increase in HR cells; +++, 2- to 8-fold increase in HR cells; ++++, >8-fold increase in HR cells; −, nearly complete repression in HR cells; 0.5, >50% repression in HR cells.

DISCUSSION

Since the discovery of the first HDMs in recent years, rapid progress has been made toward unraveling the cellular and molecular functions of HDM. The JmjC HDMs play important roles in normal development and have also been implicated in complex human diseases, including cancer and neurological disorders (9, 13, 29, 30). Our findings add a new facet to the emerging HDM field by showing that an atypical JmjC domain-containing protein functions as a HDM for the first time. Multiple missense mutations in the JmjC domain of HR have been reported in patients with APL (25, 26). Our finding that two of the patient mutations in the HR JmjC domain can abolish its HDM activity adds novel insights into the significance of histone methylation in epidermal biology. Concurrent with our efforts to understand epigenetic regulation of epidermal development, a recent study has elucidated how hair follicle stem cells employ dynamic patterns of histone modification to coordinate the activity of a host of hair genes to achieve self-renewal and lineage-specific differentiation and to form hair cells during hair follicle growth and cycling (31). These studies altogether reveal the crosstalk between genetic and epigenetic mechanisms during epidermal development and homeostasis.

With a few exceptions, H3K9me1 is often found in the promoter or 5′ end of intragenic DNA elements of actively expressed genes, whereas H3K9me2 is generally associated with repressed genes or heterochromatin regions (13, 32). HR displays demethylase activity on both H3K9me1 and H3K9me2, suggesting that HR can potentially exert opposing effects on its target gene expression. These findings explain the context-dependent mode of gene regulation by HR, as reported previously (16–18). For example, a recent study shows that HR acts not only as a transcriptional repressor but also as an activator in keratinocyte-derived HaCaT cells in a gene-selective manner (18). This dynamic role of HR in gene regulation is further supported by our studies showing that overexpression of HR results in a significant increase in Mxi1 expression accompanied with a nearly complete repression of caspase14 in HEK293 cells (Fig. 3B). Our findings showing that HR repressed Wise expression in human keratinocytes but drastically increased its expression in HEK293 cells demonstrate the context-dependent mode of gene regulation by HR (Fig. 3C). We speculate that demethylation by hairless probably alters the chromatin environment of its target genes to allow the binding of either an activator or a suppressor to its target gene to dictate the expression status of its target genes. Taken together, our findings argue against a uniform correlation between H3K9 methylation and gene expression. The ultimate outcome of how HR affects its target gene expression may vary in a context-dependent manner.

The spectrum of HR targets in the human genome may not be limited to the list of genes uncovered by our current ChIP-Seq analysis. In addition to the 25 genes shown in Table 1, our ChIP-Seq studies also revealed that HR interacts with 21 genomic sites that are not associated with known genes (unpublished observations). The biological significance of such interactions remains to be elucidated. The small number of genomic sites that were identified by our ChIP-Seq experiment may be attributable to the fact that the ChIP-Seq experiments were conducted using HEK293 cells, which do not express HR. Some candidate HR target genes may adopt cell type-specific chromatin conformations, thereby preventing the binding of HR in HEK293 cells. DLX3, for example, was shown to be a direct target of HR in the skin (19). However, our ChIP-Seq studies did not reveal the interaction between HR and DLX3 in HEK293 cells, although HR overexpression in HEK293 cells dramatically decreased DLX3 mRNA expression (unpublished observations). It is possible that HR is a relatively weak DNA-binding transcription factor, and its interaction with DNA may require other cofactors. The resolution of the ChIP-Seq technique in uncovering target genes of a given transcription factor is often affected by the hit-and-run nature of most transcription factors, which may also contribute to an incomplete coverage of other potential HR target genes by our ChIP-Seq analysis.

Taken together, our data show for the first time that HR is a bona fide HDM, despite an atypical metal-binding motif in its JmjC domain. We also show that HR can regulate the activity of its target genes, such as caspase14, Mxi1, and Wise through H3K9 demethylation in their promoters or TREs. The newly identified list of HR target genes will greatly facilitate future investigations to elucidate the molecular mechanisms by which HR exerts its regulatory role in epidermal development and its dysregulation in human skin pathogenesis. Furthermore, our findings highlight the significance of epigenetic mechanisms such as histone methylation in epidermal homeostasis and will raise new therapeutic opportunities for treating skin and hair diseases by targeting histone methylation.