Hairless is translocated to the nucleus via a novel bipartite nuclear localization signal and is associated with the nuclear matrix

Hairless mice were discovered serendipitously in an aviary in London in 1924, and were first described as a mutation of the wild form of Mus musculus (Brooke, 1926; Summer, 1924). The mouse hairless (hr) locus was later identified as the result of a spontaneous mutation caused by the insertion of an endogenous retrovirus between exons 6 and 7 of the hr gene on chromosome 14 and a subsequent reduction in mRNA levels (Cachon-Gonzalez et al., 1994). The rhino mutation (rh/rh) is allelic to hr/hr, and exhibits an exaggerated phenotype resulting from nonsense mutations in the hr gene, and therefore represents the true null mutation (Ahmad et al., 1998c; Ahmad et al., 1998d; Panteleyev et al., 1998a; Cachon-Gonzalez et al., 1999). The hr gene encodes a putative zinc finger transcription factor, and is one of the candidate genes for the regulation of basic hair follicle function, particularly the dynamic transition to catagen (Panteleyev et al., 1998c). The expression pattern of hr mRNA was first studied by northern blot analysis and revealed that hr was expressed in the skin and brain (Cachon-Gonzalez et al., 1994; Thompson, 1996), and more recent studies report a more widespread tissue distribution, including the colon, retina, inner ear, cartilage and tooth (Cachon-Gonzalez et al., 1999). Furthermore, rats, mice and humans share a similar tissue-specific pattern of hr expression (Ahmad et al., 1998a; Thompson, 1996).

The most striking phenotypic feature of all mice bearing mutations at the hr locus is the rapid onset of a complete and sharply demarcated postnatal wave of hair shedding, which follows a seemingly programmed time course. At the age of 13-14 days, after the development of an apparently normal first hair coat, homozygous mutants begin to lose their hair, starting from the upper eyelids, then around the eyes, and the forelimbs. Loss of hair progresses caudally from the snout, with equal rates of shedding and subsequent alopecia on both the dorsal and ventral sides of the body (Panteleyev et al., 1998c). Histological studies of hairless mutants revealed normal development of the distal part of the hair follicle, including the infundibulum and sebaceous gland, but severe malformation of the germinal, proximal portion of the follicle, leading to the lack of new hair development, follicular degeneration and the formation of dermal cysts (Panteleyev et al., 1998c). The hairless phenotype in mice is characterized by very similar clinical and histological features to a rare form of human atrichia, which led to the suggestion that hr gene mutations may be the underlying cause of this disorder, a notion first put forth by Sundberg and collegues in 1989 (Sundberg et al., 1989).

Atrichia with papular lesions (APL) (OMIM209500) is a rare form of irreversible alopecia inherited in an autosomal recessive pattern (Ahmad et al., 1998b; Ahmad et al., 1999b; Damste and Prakken, 1954; Fredrich, 1950; Lowenthal and Prakken, 1961). In individuals affected with this form of hair loss, hairs are typically absent from the scalp, axilla and body, and patients are almost completely devoid of eyebrows and eyelashes. Histological examination of affected scalp skin shows the absence of mature hair follicle structures. Although alopecia may accompany several different forms of congenital ectodermal dysplasias, APL patients are unique in that, along with total atrichia, papules and follicular cysts filled with cornified material are formed in the skin. These features represent a unique cutaneous abnormality among inherited alopecias. Like the hairless mouse, normal hairs are present at birth in most APL patients, but these neonatal hairs are usually shed within first months of life and are never replaced. Recently, we and others reported linkage of this form of atrichia to chromosome 8p12, in a region syntenic with mouse chromosome 14 (Ahmad et al., 1998a; Cichon et al., 1998). We then elucidated the genetic basis for this disorder by identifying mutations in the human hr gene, which localizes to the same region of chromosome 8 (Ahmad et al., 1998a). Subsequently, thirteen additional mutations have been identified in families from around the world showing a similar phenotype, thereby establishing the molecular basis of this disorder (Ahmad et al., 1998b; Ahmad et al., 1999a; Ahmad et al., 1999b; Aita et al., 2000; Cichon et al., 1998; Kruse et al., 1999; Sprecher et al., 1999; Zlotogorski et al., 1998).

In the hair follicle, hairless protein (Hr) appears to function in the cellular transition to the first adult hair cycle. In its absence, hair growth completely ceases, a new hair is never induced, and the result is a complete form of inherited atrichia. Furthermore, the hair matrix cells undergo premature and massive apoptosis, together with a concomitant decline in Bcl-2 expression, a loss of NCAM positivity, and disconnection from the overlying epithelial sheath, essential for communication between the dermal papilla and the permanent portion of the hair follicle (Panteleyev et al., 1998b; Panteleyev et al., 1998d). As a consequence, the dermal papilla remains stranded in the dermis, and thus no further hair growth occurs. In hairless mice and in humans with congenital atrichia, it appears that the absence of hairless protein initiates a premature and abnormal catagen due to aberrant signaling that normally controls catagen-associated hair follicle remodeling (Panteleyev et al., 1998b; Panteleyev et al., 1998c). We have previously suggested that Hr may play a crucial role in maintaining the balance between cell proliferation, differentiation, and apoptosis in the hair follicle as well as in the inter-follicular epidermis (Panteleyev et al., 1999; Panteleyev et al., 2000).

The hr gene has been studied extensively at the genetic and molecular level, and these studies have provided information about its expression pattern and temporal regulation. However, at the cellular level, the mechanism of function of the hairless protein remains largely unexplored. In this study, we have investigated the domain organization and subcellular localization of the hairless gene product. Analysis of the predicted hairless amino acid sequence does not show homology to any known protein, however, it disclosed a putative zinc finger motif with homology to the GATA family of nuclear transcription factors (Cachon-Gonzalez et al., 1994). Furthermore, we identified three LXXLL motifs that are implicated in protein-protein interactions among the nuclear receptor gene family (LeDouarin et al., 1996; Voegel et al., 1998). As a first step toward elucidation of the functional role of the hairless protein, we have determined its subcellular localization and demonstrated that the hairless protein resides in the nuclear compartment. We have also identified a novel bipartite nuclear localization signal (NLS). Transfection of hr-green fluorescent fusion genes demonstrated that a region of the N terminus is essential and sufficient for nuclear localization. Deletion analysis further mapped the signal sequence to amino acid residues 409 to 427 in the mouse sequence (Cachon-Gonzalez et al., 1994). Although related to the classical bipartite consensus sequence, this novel motif shows significant differences. Furthermore, immunofluorescence studies together with nuclear fractionation analysis showed that hairless was attached to the nuclear matrix.

NIH 3T3 and COS 7 cells were grown in DMEM supplemented with 10% fetal calf serum (FCS), 2 mM glutamine and antibiotics (100 i.u./ml penicillin; 100 mg/ml streptomycin). Cells were transfected at 60% confluency with plasmid DNA using Lipofectamine reagent (Life Technology, Inc., USA) as recommended by the manufacturer. For immunofluorescence analysis, cells were grown on glass coverslips. After 24 to 48 hours, transfected cells were fixed in 3% paraformaldehyde in phosphate buffer saline (PBS) for 15 minutes, washed twice with PBS, counterstained with 4,6-diamidino-2-phenylindole (DAPI), and mounted with antifade kit (Molecular Probe, Eugene, Oregon USA). The subcellular location of GFP fusion proteins was determined by fluorescence microscopy.

NIH 3T3 cells were grown on glass coverslips and transfected as described above. The cells were washed with PBS, fixed with 3% paraformaldehyde in PBS for 15 minutes at room temperature. After quenching with 50 mM NH₄Cl in PBS, cells were permeabilized for 30 minutes in blocking buffer (PBS containing 1% BSA, 0.2% gelatin, 0.2% Triton X-100). Cells were incubated with anti-Flag monoclonal antibodies (Sigma) (diluted 10 μg/ml) and anti-lamin B polyclonal antibodies (diluted at 10 μg/ml) for 30 minutes at room temperature. After appropriate washes, cells were further incubated with secondary antibodies: FITC-coupled goat anti-mouse antibodies or rhodamine-conjugated goat anti-rabbit (Silenus laboratories). All the antibodies were diluted in blocking buffer.

Analysis of hairless and lamin B cytolocalization was carried out using TCS40 (Leica, Germany) confocal imaging system equipped with a ×63 objective (Plan Apo; NA 1.4). For FITC and rhodamine excitation, an argon-krypton ion laser adjusted at 488 and 568 nm was used. For each optical section, double fluorescence images were acquired simultaneously. The two images were merged to check the relative position of the two fluorochromes. A focal series was collected for each specimen, and the focus step between sections was generally 0.8 μm.

High salt isolation of nuclear matrix was carried out as described (He et al., 1990). After transfection, cells were washed in PBS and were extracted in cytoskeleton buffer CSK (10 mM Pipes, pH 6.8, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 1 mM EGTA, supplemented with leupetin, aprotinin and pepstatin (1 μg/ml each), 1 M DTT, and 0.5% Triton. After 3 minutes at 4°C, the cytoskeleton frameworks were separated from soluble proteins by centrifugation at 600 g for 3 minutes. Chromatin was solubilized by DNA digestion with 1 mg/ml of RNAse free DNAse I in CSK buffer plus proteinase inhibitors for 15 minutes at 37°C. Then ammonium sulfate was added from 1 M stock solution in CSK buffer to a final concentration of 0.25 M and after 5 minutes at 4°C, samples were pelleted at 5,000 g for 3 minutes. The pellet was further extracted with 2 M NaCl in CSK buffer for 5 minutes at 4°C, and then centrifuged as above. The remaining pellet was solubilized in urea buffer and represented the nuclear matrix fraction.

A partial cDNA encoding mouse hairless was kindly provided by Catherine Thompson (Thompson, 1996). We then generated the full-length cDNA for hr in the pAlter plasmid (Promega). Other plasmids for protein expression in transfected cells were designed as described below. To produce cDNA inserts for cloning, PCR was performed on the Robocycler Gradient 96 (Stratagene, Inc.). Custom oligonucleotides were obtained from Life Technology. Standard methods (Sambrook et al., 1989) were used for DNA purification, restriction endonuclease digestion, DNA ligation, bacterial transformation, and preparation of plasmid DNA. All recombinant plasmids were analyzed by restriction analysis. Plasmids in which PCR was used to generate coding sequences were also analyzed by DNA sequencing performed on an ABI Prism 310 Genetic Analyzer (PE Biosystems). PCR amplification of segments of hairless from pAlter-hr template were cloned in frame into the XhoI/EcoRI sites of pEGFP-C1 vector (Clontech). The primers used for amplification were as follows (all sequences are oriented in the 5′-3′ direction): EGFP-1F CGCTCGAGCTATGGAGAGTATGCCC and CCGGAA-TTCGAGTGCTTGGCCAGGCC for EGFP-NT; CCGCTCGAGCT-GGAGGGCCTGGC and CCGGAATTCGACCCCACAGCCGC for EGFP-CT; EGFP-1F and CCGGAATTCGACCCAGTGCCTGG for EGFP-Δ 4; EGFP-1F and CCGGAATTCGAGCCAAGATCCCC for EGFP-Δ 5; EGFP-1F and CCGGAATTCGAGGCTTGCTGGAG for EGFP-Δ 6; EGFP-1F and CCGGAATTCGAGCGTTTGGGAGC for EGFP-Δ 41; EGFP-1F and CCGGAATTCGACCTCTTGAGTGC for EGFP-Δ 42; EGFP-1F and CCGGAATTCGAACCTGGGCACTC for EGFP-Δ 43; CCGCTCGAGCTCGGGCACTCAAG and CCGGAAT-TCGAGCCAGGCCCTCC for EGFP-Δ 41a; CCGCTCGAGCTG-CACTCAAGAGG and CCGGAATTCGAGCCAGGCCCTCC for EGFP-Δ 41b; CCGCTCGAGCTGGCCCTTGTAGG and CCGGAA-TTCGAGGGAGCTGGGCC for EGFP-Δ 41c. To generate the full length EGFP-hr fusion protein, the DNA for the carboxyl-terminal section of hr was digested with HindIII and EcoRI and was inserted into the EGFP-NT construct in the corresponding restriction sites.

Whole cell extracts were harvested in urea buffer (8 M urea, 0.1 M NaHPO₄, 0.01 M Tris, pH 8) from the transiently transfected NIH 3T3 cells 48 hours after transfection. Total extracts were subjected to 10% SDS-polyacrylamide gel electrophoresis. After separation, the proteins were transferred to nitrocellulose membranes and were incubated with blocking buffer for several hours as described previously (Djabali et al., 1991). The membranes were then incubated with the monoclonal anti-GFP antibody (Clontech), and further incubated with the corresponding second antibody. The proteins were visualized using the enhanced chemiluminescence detection system (Amersham Pharmacia Biotech).

NIH 3T3 cells were transfected with the full length mouse hr cDNA that had been cloned into the mammalian expression vector, pAlter, in-frame with the Flag-Tag epitope sequence. 24 hours after transfection, cells were fixed and processed for indirect immunofluorescence microscopy. Hr was found to be localized in the nuclear compartment, and using confocal microscopy, hr labeling revealed a strictly nuclear signal (Fig. 1A). Hr shared a similar nuclear localization with lamin B (Fig. 1B), however, they showed distinct distribution patterns, as seen on the merged image of both signals (Fig. 1C). Furthermore, Hr showed a diffuse nucleoplasmic signal that was excluded from the nucleoli and from regions of highly condensed heterochromatin, which are stained brightly by DAPI (data not shown).

To investigate the possibility that Hr might bind to the nuclear skeleton or nuclear matrix, we isolated the nuclear matrix by the sequential extraction procedure as described previously (He et al., 1990). In the first fractionation step, soluble proteins are removed by extraction with Triton X-100. Chromatin proteins are then released by DNAse I digestion and extraction with 0.25 M ammonium sulfate. After washing with 2 M NaCl, the last fraction is composed of structural proteins and nuclear-associated proteins. Supernatants from each extraction step and the final matrix pellet were analyzed by SDS-PAGE and immunoblotting. Surprisingly, Hr was not released with the chromatin proteins, as no Hr signal was detected in the different supernatant fractions (Fig. 2, middle panel, lanes 2-4), while Hr was tightly associated with the nuclear matrix proteins (Fig. 2, middle panel, lane 5). This was compared with the distribution of another nuclear protein, lamin B. We used an antibody against lamin B, one of the major components of the nuclear matrix, as a marker to verify that cytoplasmic and chromatin fractions were not contaminated with nuclear matrix proteins. As expected, lamin B was localized solely in the nuclear matrix fraction (Fig. 2, right panel, lane 5). Furthermore, we observed that the Hr protein was partitioned in the same fraction as lamin B.

We then investigated whether the distribution of Hr is similar in whole nuclei and in nuclear matrices. We used antibodies in indirect immunofluorescence analysis of the nuclear material remaining after in situ sequential DNAse I digestion and high salt extraction. As shown in Fig. 1D-F, the labeling pattern was similar for both Hr and lamin B before and after extraction (Fig. 1A-C and Fig. 1D-F, respectively), but the intensity of the signal did not decrease in extracted nuclei (Fig. 1D-F). As expected, the DAPI signal disappeared completely in the nuclear matrix preparation after DNAse digestion (data not shown). Lamin B antibodies gave a strong signal in the nuclear lamina and a much weaker signal in the nucleoplasm. The intensity of the signal of Hr and lamin B were similar in whole nuclei (Fig. 1A-C) and in extracted nuclei (Fig. 1D-F).

Fig. 3A shows a schematic representation of the different EGFP-hr constructs used in this study. Total extracts from NIH 3T3 cells transfected separately with each construct were analyzed by western blot to verify that the corresponding GFP fusion protein was of the expected size (Fig. 3B).

GFP alone is uniformly distributed throughout the cell, thus GFP is a suitable tag to study hairless localization (Fig. 4A). As expected, the GFP-tagged construct expressing the full length hr transfected into NIH 3T3 cells gave a nuclear localization (Fig. 4C). Hr fusion protein had a similar nuclear localization whether fused to a Flag epitope or to the GFP protein, suggesting that the nuclear localization of hr is dictated by the Hr protein sequence itself and that the tagged protein (Flag or GFP) did not contribute to restricted nuclear compartment localization of the Hr protein. In both COS 7 and HeLa cells, GFP-hr has an identical localization pattern as in NIH 3T3 cells (data not shown). Thus, only NIH 3T3 were used in further localization studies. The full-length protein was divided into two fragments corresponding to the amino- and carboxyl-terminal part of the protein, and each was fused to the GFP protein (Fig. 3A). The C-terminal hairless protein fused to GFP was localized in the cytoplasm (Fig. 4E). This result showed that the carboxyl-terminal region of hairless is not necessary for its nuclear localization. However, cells transfected with the N-terminal region of hairless protein showed a nuclear localization of the GFP fusion protein, which indicated that this region contains the putative nuclear localization signal (Fig. 4G).

In order to identify the NLS region within the amino-terminal region of hairless, we designed several deletion constructs of the N-terminal fragment of the hairless protein (Fig. 3A). Immunofluorescence studies of cells transfected with EGFP-Δ4 showed a nuclear localization of the corresponding GFP fusion protein (Fig. 4I) while proteins corresponding to EGFP-Δ5 and EGFP-Δ6 were no longer restricted to the nucleus, but localized instead to the cytoplasm (Fig. 4K). These results suggest that the putative NLS sequence resides between amino acids 337 and 436 which were deleted in the construct EGFP-Δ5 (Fig. 3A).

We were unable to identify any previously reported cluster of basic amino acids or a bipartite NLS in this region. To search for a novel NLS, we designed serial deletion constructs shortened by an average of 20 amino acids, starting from the carboxyl-terminal end of the EGFP-Δ4 construct (aa436), and studied their cellular localization after transfection. A schematic representation of the different deletion constructs is summarized in Fig. 5A. Furthermore, we verified that each construct produced a GFP chimera of the expected size by western blot analysis (Fig. 5C). We observed that EGFP-Δ41 showed a restricted nuclear localization (Fig. 6A), while the fusion protein corresponding to construct EGFP-Δ42 accumulated in the nucleus but was also found in the cytoplasm (Fig. 6C). This observation suggested that EGFP-Δ42 contained a truncated NLS sequence. The EGFP-Δ43 fusion protein was localized primarily in the cytoplasm with some fusion protein localized in the nuclear compartment (Fig. 6E). We concluded from these studies that the NLS sequence must reside between amino acids 406 and 427, which corresponded to the following amino acid sequence: RALKRAGSPEVQ-GASRGPAPKR. Since the construct EGFP-Δ42 was not restricted to the nucleus, we hypothesized that the last KR residues were probably essential for the NLS. To further narrow this sequence, we made different constructs to delineate more precisely the minimal sequence required for nuclear localization (Fig. 5B).

To delimit the amino-terminal boundary, the construct EGFP-Δ41a began with the R at position 406 and extended downstream. We observed that the corresponding GFP fusion protein resided within the nucleus (Fig. 6G). We concluded from this observation that not only did the EGFP-Δ41a contain the nuclear localization signal, but importantly, that the upstream R at position 406 was not part of the NLS sequence, since EGFP-Δ41b that began at position 407 was also restricted to the nucleus (Fig. 6I). However, the GFP signal was no longer restricted to the nucleus in cells transfected with EGFP-Δ41c in which the last KR at position 426 and 427 were deleted to delineate the carboxyl-terminal boundary (Fig. 6K). This result indicated that residues KR at position 426 and 427 were included in the NLS sequence, and necessary for nuclear localization. We concluded that the NLS sequence corresponded to the sequence KRAGSPEV-QGASRGPAPKR (409-427), since deletion of the first or last basic amino acids KR (EGFP-Δ42 and EGFP-Δ41c) abolished the nuclear localization (Fig. 6C and K). Furthermore, sequence comparison between human, rhesus monkey, mouse and rat hairless genes indicated that the R (421) was not conserved between the different species, suggesting that this residue is not necessary for the nuclear localization signal. Using a series of deletion constructs and analyzing their subcellular compartmentalization, we have identified a new bipartite NLS in hairless that is denoted as KRA(X13)PKR.

All transport into the nucleus occurs through the nuclear pore complex (Gorlich and Mattaj, 1996), which allows small molecules (under 45 kDa) to diffuse freely into and out of the nucleus (Baker et al., 1990; Siomi and Dreyfuss, 1995). The targeting of larger proteins to the nucleus requires the presence of a nuclear localization signal (NLS) (Mattaj and Englmeier, 1998). NLS motifs are short sequence motifs necessary for nuclear import of the respective proteins, and two major types of NLSs are recognized. The first type consists of a single cluster of basic amino acids, for example, PKKKRKV in the SV40 large T antigen NLS (Dingwall and Laskey, 1991). The second type is the bipartite NLS composed of two clusters of basic amino acids separated by 8-16 unspecific residues, for example, KRPAATKKA-GQAKKKK found in nucleoplasmin (Dingwall and Laskey, 1991; Gorlich and Mattaj, 1996). By cDNA deletion analysis and transfection studies, we showed that the amino-terminal region of hairless is necessary and sufficient for nuclear localization (Fig. 4).

More specifically, we identified a fragment of 19 consecutive residues constituting a novel nuclear localization signal: KRAGSPEVQGASRGPAPKR (amino acids 409-427, Fig. 6). To confirm that the KR residues at the N- and C-termini were critical for nuclear localization, we performed computer-assisted mutagenesis using the program PredictNLS (Cokol et al., 2000), and found that mutating either of the two K residues or the two R residues completely aboloshed recognition of the sequence as an NLS motif, and further, the sequence no longer bore homology to any nuclear protein (not shown). Comparing this region in mouse, human and rat, we found that all residues were conserved except for ‘SR’ (420-421) which was ‘R’ in rat and ‘MG’ in human. We therefore concluded that the hairless NLS consists of one cluster bearing a lysine, an arginine and an alanine, a spacer of 13 amino acids, and a second cluster of a proline, lysine and an arginine: KRA(X13)PKR, constituting a novel NLS motif (Cokol et al., 2000).

Searching the SWISS-PROT database (Bairoch and Apweiler, 1999) using the program PredictNLS (Cokol et al., 2000) with the motif KRA(X13)PKR matched only the hairless proteins. Extending the motif to KR[A/G](X13)PKR, matched the nuclear RNA annealing protein YA1 in yeast (yra1_yeast; Bairoch and Apweiler, 1999); extending to KR[A/G](X10-16)PKR matched three other nuclear proteins: (1) the SNAI protein from Drosophila (snai_drome; Bairoch and Apweiler, 1999), (2) the high-mobility gene C (hmgc_mouse; Bairoch and Apweiler, 1999), and (3) p53 in hamster (p53_mesau; Bairoch and Apweiler, 1999). The latter was particularly interesting since the match was in a region of an experimentally verified NLS which itself was too specific to match in hairless (Bairoch and Apweiler, 1999). Extending the motif further to KR(X14-17)KR matched a number of non-nuclear proteins, indicating that the basic flanking residues were not sufficient to specify the NLS. Thus, the generalized novel NLS we propose is KRA(X10-16)PKR.

The hr gene encodes a putative zinc-finger transcription factor, and the hr protein has a predicted molecular mass of 127 kDa. The zinc-finger motif contains a potential DNA-binding domain with six cysteines located at the carboxyl-terminal region of hr (Cachon-Gonzalez et al., 1994; Thompson, 1996; Thompson and Bottcher, 1997; Ahmad et al., 1998a). The zinc-finger domain of hairless is homologous to those found in the GATA family of transcription factors, which have the following distinctive structure: C-X-X-C-(X17)-C-X-X-C (Arceci et al., 1993; Ko and Engel, 1993; Weiss and Orkin, 1995). Previous studies indicate that a single zinc-finger domain of the type found in GATA-family members is sufficient for DNA binding (Ko and Engel, 1993). The hr putative zinc finger domain resides within a loop region at aa 590-625 in the human hr sequence and contains a cluster of six cysteine residues with a novel spacing that is conserved among rat, mouse and human (Ahmad et al., 1998a). Although the zinc finger motif is usually present in multiple copies, several proteins in a variety of species contain only a single zinc finger (Evans and Hollenberg, 1998), many of which have been implicated in transcriptional regulation. Whether or not this region of the hr protein binds directly to specific DNA sequences is currently under investigation.

Analysis of the secondary structure of Hr using the PhD program (Rost and Sander, 1994), revealed that the N-terminal domain has no clearly defined secondary structure, and the folding of this part of the protein remains to be determined (Fig. 7). From amino acid 378 to the C-terminal end of Hr, however, we noted several stretches of α helix interrupted by long loop regions. The carboxyl-terminal part of the protein presented a highly structured sequence that may fold into several structural domains. Furthermore, we identified three LXXLL motifs within two separate α-helical regions of Hr. This LXXLL motif, also termed the NR box (nuclear receptor), has been shown to be required for interaction with nuclear receptors. NR boxes are conserved in several transcriptional coactivators, including a family of genes known as TRIPs, or thyroid hormone interacting proteins (LeDouarin et al., 1996; Voegel et al., 1998) as well as DRIPs, or vitamin D receptor interacting proteins (Rachez and Freedman, 2000).

Nuclear receptors (NR) including vitamin D3, thyroid hormone, and retinoic acid receptors, are part of a large family of transcription factors that can transduce the signals of a small lipophilic hormonal ligands by binding to target DNA sequences, and thereby regulate gene transcription in direct response to such ligands (Mangelsdorf and Evans, 1995). Furthermore, NRs all share similar functional regions that include a DNA binding domain, and a ligand binding domain (LBD). The LBD contains a short α helical motif called AF-2 at its carboxyl terminus (Barettino et al., 1994; Danielian et al., 1992; Durand et al., 1994), that is required for ligand-dependent transactivation and is involved in the interaction with coactivators. Interestingly, coactivators such as SRC-1 and GRP/TIF2 have LXXLL motifs that are arranged as sets of three NR boxes which are required for direct interaction with the nuclear receptor AF-2 domain (Glass et al., 1997). Several combinations of this motif appear to direct the specificity of interaction with nuclear receptor heterodimers (Darimont et al., 1998; McInerney et al., 1998). By analogy to these coactivators, the hairless protein shares three LXXLL motifs, suggesting a role in a specific interaction with member(s) of the nuclear receptor family. Hairless may belong to this large family of nuclear receptor coactivators, which could reveal a new pathway of gene activation due to its restricted tissue expression pattern (Panteleyev et al., 2000; Thompson, 1996). In further support of this notion, the hairless gene was identified as a direct early response gene in a study to identify genes that are regulated by thyroid hormone (Thompson, 1996). The thyroid hormone exerts its effect through nuclear receptors (thyroid hormone receptors or TRs), which modulate the transcription of downstream genes in response to hormone binding. The same investigators identified a potent thyroid hormone response element (TRE) in an upstream regulatory region of the rat hr gene. They reported that thyroid receptor binds to a fragment mapped within 106 bp located 9 kb upstream of the first exon of hr (Thompson, 1996). Studies demonstrating a biochemical interaction between Hr and the thyroid receptors (Thompson, 1996) and the presence of three NR boxes provide additional evidence for this hypothesis. In addition, phenotypic similarities between the alopecia in vitamin D receptor knock out mice and hairless mutants (Kato et al., 1999; Sakai and Demay, 2000), as well as human vitamin D-resistant rickets patients (Hochberg et al., 1984) and individuals with papular atrichia, suggest a potential association between VDR and hairless as well. Collectively, these observations raise the possibility that hairless may be part of a coactivator complex which interacts with partners such as VDR and TR, among others, in a tissue-restricted manner.

In an attempt to define the physiological function of Hr, we sought to determine the subcellular localization of the Hr protein within the nucleus. To further investigate the association of Hr with the nucleus, several different procedures were tested in an attempt to release the hr protein from transfected NIH 3T3 cells. Rather than behaving as a soluble and easily extractable factor, we were surprised to find that Hr was resistant to extraction with detergent and NaCl, even after total DNA digestion with DNAse I. Such treatments are normally used to extract chromatin components including groups of proteins, transcription factors, or histones. Thus, we demonstrated that not only is Hr localized in the nucleus, but furthermore, that it is tightly associated with the nuclear matrix (Fig. 1).

The extensive array of filaments in the remnant nuclei following the extraction of nuclei under such aggressive conditions has historically been named the nuclear matrix (Barrett and Spelsberg, 1999; Brown, 1999; Pederson, 1998; Pederson, 2000; Stein et al., 1998; Stenoien et al., 1998). While the existence of a nuclear matrix was the source of some debate, in recent years, the nuclear matrix has been shown to play a role in DNA organization, replication, gene transcription, RNA processing, and potentially intranuclear signaling for regulating gene transcription (Barrett and Spelsberg, 1999), suggesting a structural basis for nuclear architecture. It has also been implicated in the activation of gene expression by steroid hormones, as steroid receptors, such as TR and VDR have been localized to the nuclear matrix in a variety of target tissues (Barrett and Spelsberg, 1999; Nangia et al., 1998). Several lines of evidence support the hypothesis that the association of certain transcription factors with the nuclear matrix may be required for accuracy of gene expression and maximum transcriptional activity (Stein et al., 1998). Finally, the nuclear matrix is believed to function in the local recruitment of coregulator complexes in a cell type specific manner to potentiate fidelitous gene expression. Morphological studies of the hairless phenotype demonstrate an absolute requirement for hr in maintaining the fidelity of the hair cycle. While histological examinations implicate hr in the regulation of apoptosis during the remodeling of the catagen hair follicle (Panteleyev et al., 1999; Panteleyev et al., 2000), the molecular basis of this regulation remains unknown. The demonstration of the nuclear localization of Hr, together with the identification of the NLS, provides additional evidence in support of Hr functioning as a transcriptional regulator. Furthermore, the association of Hr with the nuclear matrix may indicate one possible function of Hr as a nuclear scaffolding protein. We posit that Hr may regulate gene expression by recruiting and physically organizing other transcriptional cofactors to effect the coordinated regulation of the dynamic cellular changes associated with the catagen stage of the hair cycle.

We express our sincere thanks to Dr Burkhard Rost (Columbia University) for valuable insights and assistance in using his new NLS identification program. We appreciate the assistance of Dr Andrei A. Panteleyev in preparation of the images, Dr Howard Worman for stimulating discussions and Dr Andrew Engelhard for help with homology analysis. This work was supported in part by the National Alopecia Areata Foundation, The Kirsch Foundation and the Skin Disease Research Center, Department of Dermatology, Columbia University (P30 AR44535).