Epsin 3 is a novel extracellular matrix-induced transcript specific to wounded epithelia.

Using an in vitro model of keratinocyte activation by the extracellular matrix following injury, we have identified epsin 3, a novel protein closely related to, but distinct from previously described epsins. Epsin 3 contains a domain structure common to this gene family, yet demonstrates novel differences in its regulation and pattern of expression. Epsin 3 mRNA and protein were undetectable in keratinocytes isolated from unwounded skin, but induced in cells following contact with fibrillar type I collagen. The native triple helical structure of collagen was required to mediate this response as cells failed to express epsin 3 when plated on gelatin. Consistent with the reported function of other epsins, epsin 3 was evident in keratinocytes as punctate vesicles throughout the cytoplasm that partially co-localized with clathrin. In addition, epsin 3 exhibited nuclear accumulation when nuclear export was inhibited. In contrast to other known epsins, epsin 3 was restricted to keratinocytes migrating across collagen and down-regulated following cell differentiation, suggesting that expression was spatially and temporally regulated. Indeed, epsin 3 was localized specifically to migrating keratinocytes in cutaneous wounds, but not found in intact skin. Intriguingly, Northern hybridization and reverse transcriptase-polymerase chain reaction experiments indicated that epsin 3 expression was restricted to epithelial wounds or pathologies exhibiting altered cell-extracellular matrix interactions. Thus, we have identified a novel type I collagen-induced epsin that demonstrates structural and behavioral similarity to this gene family, yet exhibits restricted and regulated expression, suggesting that epsin 3 may serve an important function in activated epithelial cells during tissue morphogenesis.

The epidermis consists of a multilayered epithelial sheet that provides a physical barrier against the outside environment and heals in response to injury. In unwounded skin, basal keratinocytes reside on a basement membrane that physically separates these cells from the underlying dermal connective tissue rich in type I collagen. While in contact with this extra-cellular matrix (ECM), 1 keratinocytes express a programmed subset of genes that promotes proliferation and differentiation. Following injury, however, keratinocytes from the surrounding tissue are activated by exposure to ligands released into the wound site and by contact with ECM macromolecules (1)(2)(3). Keratinocyte activation, which typically begins 18 -24 h prior to the onset of migration, occurs as cells at the wound edge enhance the expression of genes that support repair of the tissue defect (4,5). Many of the genes up-regulated in keratinocytes during healing, including secreted proteinases and integrin receptors, enable a fundamental shift in cell behavior that supports sustained and directed migration to re-establish the normal cytoarchitecture of the skin (3). The preponderance of studies to date attempting to identify signals that stimulate keratinocyte activation during wound healing have focused on soluble mediators, whereas the role that alterations in cell-ECM interactions play in this process has received relatively little attention.
Migration of keratinocytes from the wound edge occurs as the cells dissect under a provisional matrix of fibrin and fibronectin (6) and over or through a viable dermis, which includes structural molecules distinct from those in the basement membrane. Loss of contact with the basement membrane and subsequent exposure to the underlying dermal matrix may be a critical determinant that alters gene expression and induces the activated keratinocyte phenotype. Indeed, recent evidence supports this idea as the matrix metalloproteinase collagenase-1 is invariantly expressed in keratinocytes not in contact with a basement membrane, but rather in cells migrating over the dermal matrix and in close apposition to collagen fibers (7)(8)(9)(10). Moreover, expression of this enzyme is selectively induced in keratinocytes following contact with fibrillar type I collagen in vitro and its proteolytic activity is essential for cell migration across this matrix (10,11). Together, these data suggest that keratinocyte contact with dermal ECM, and in particular fibrillar type I collagen, profoundly influences keratinocyte activation following injury by inducing the expression of transcripts required for efficient repair.
We used an in vitro model of keratinocyte activation by collagen to identify novel transcripts associated with migrating keratinocytes and to gain a better understanding of the role that changes in cell-ECM contacts play in keratinocyte behavior during wound repair. We report here the identification of a type I collagen-induced epsin that retains typical structural motifs and behavioral activity common to previously described epsins. However, epsin 3 is profoundly divergent from other members of this gene family, as expression is limited to wounded or pathologic tissues with altered cell-ECM interactions, suggesting that epsin 3 may serve a role during repair of tissues that demonstrate a disrupted basement membrane. Furthermore, our findings underscore the importance of the ECM in stimulating keratinocyte activation following cutaneous injury and suggest that common mechanisms may influence cell activation in other wounded or pathologic epithelial tissues.

EXPERIMENTAL PROCEDURES
Materials-Bovine type I collagen (Vitrogen-100) was purchased from Celltrix Laboratories (Palo Alto, CA). Leptomycin B was purchased from Sigma-Aldrich Chemical.
Isolation and Culture of Human Keratinocytes-Human keratinocytes were harvested from healthy adult skin as previously described (12,13). Under the culture conditions used, keratinocytes proliferate, migrate, differentiate, and cornify similar to cells in vivo (12). Keratinocyte isolates were plated onto tissue culture dishes or coverslips coated with 100 g/ml type I collagen, which is necessary for matrixinduced activation. Some cell suspensions were also plated onto dishes coated with 100 g/ml gelatin (11) to serve as a negative control as this substrate supports attachment and spreading but does not stimulate activation.
Identification of Collagen-regulated Genes in Human Keratinocytes-Total RNA was isolated from keratinocytes immediately following cell harvest (0 h) or 24 h after plating on type I collagen by phenol-chloroform extraction (14). Any potential DNA contamination was removed by DNase (Promega Corp., Madison, WI) treatment as described (15). Differential display reverse transcription polymerase chain reaction (ddRT-PCR) was performed on samples using the RNAimage TM mRNA differential display system (GenHunter Corp., Nashville, TN) according to the manufacturer's instructions. Amplification of the cDNAs was performed with 0.2 M arbitrary primer 5Ј-AAGCTTACGATGC-3Ј and anchored primer H-T 11 G, the amplified samples were resolved on a 6% denaturing polyacrylamide gel, and differentially expressed products were identified. Differential expression was confirmed by Northern hybridization and selection of cDNAs for investigation was limited to genes up-or down-regulated Ͼ5-fold in duplicate RNA samples from two individual donors.
Identification and Cloning of Epsin 3-The 143-bp cDNA clone GAP4G1 was sequenced and compared with genes in public data bases using BLASTn through the NCBI. One resulting significant match, nucleotides 10237-10379 of Homo sapiens chromosome 17 (GenBank TM accession number AC004590), demonstrated 100% sequence identity. To determine if the chromosome 17 genomic clone sequence 5Ј of the GAP4G1 match coded for a novel gene we used the internet-based Genscan 1.0 program (16). To confirm the predicted gene sequence, the full-length cDNA of epsin 3 was amplified using inter-exonic primers. Alignment of the epsin 3 amino acid sequence with other known epsins was accomplished using the ClustalW program (17) and the dendrogram was generated using results obtained from ClustalW and Tree-View software (18).
Northern Analysis and Epsin 3 RT-PCR-Total RNA was harvested from primary keratinocytes, dermal fibroblasts, or umbilical vein endothelial cells as described above. 10 g/sample was denatured and resolved by electrophoresis through a 1% formaldehyde agarose gel. The RNA was then transferred and hybridized with a radiolabeled epsin 3 cDNA probe. To characterize epsin 3 expression in multiple tissues we probed Northern Territory TM Human Normal Tissue Blot II (Invitrogen Corp., Carlsbad, CA) and Human RNA Master Blot TM (CLONTECH Laboratories, Inc., Palo Alto, CA) with a radiolabeled epsin 3 cDNA probe according to the manufacturer's instructions. RNA from both vendors was pooled from multiple individuals of varying sex and age and confirmed to be free of disease. Following hybridization, membranes were washed and exposed to x-ray film for an appropriate duration.
To detect epsin 3 mRNA in various normal, wounded, and pathologic specimens, tissue samples were obtained, immediately immersed in LiN 2 , and homogenized with a Polytron in TRIzol reagent (Life Technologies, Rockville, MD) for RNA isolation. DNase-treated total RNA (1.0 g) was reverse transcribed with random hexamers using kit reagents and under the manufacturer's recommended conditions (GeneAmp RNA PCR kit, PerkinElmer Life Sciences, Norwalk, CT). Epsin 3 cDNA was visualized by amplifying a 601-bp fragment of epsin 3 using the 3Ј-antisense primer, 5Ј-GCTCCAGGTCGGAGGTA-3Ј, and the sense primer, 5Ј-ATGACGACCTCCGCACT-3Ј. These primers are to adjacent exons, and thus, the 601-bp cDNA produced from epsin 3 mRNA would be easily distinguished from products amplified from contaminating DNA or pre-processed mRNA. In addition, the PCR product identity was verified by restriction digestion and sequence analysis. To determine equal loading of RNA we used an established method to amplify GAPDH by RT-PCR (11). PCR for epsin 3 was done for 23 cycles and 25 cycles for GAPDH. The resultant products were detected by Southern hybridization using radiolabeled oligonucleotides and visualized using a Typhoon 8600 variable mode PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA).
Preparation of Antibodies and Immunoblotting Assay for Epsin 3-An 8-chain branching multiple antigenic peptide of 20 amino acids, KQNGTKEPDALDLGILGEAL, corresponding to a unique sequence of epsin 3 (amino acid residues 464 -483, see Fig. 3A), was used as an antigen (Research Genetics, Huntsville, AL). Antibodies were collected as whole serum harvested at 10 weeks after primary injection and purified by affinity chromatography with the antigenic peptide coupled to NHS Sepharose-4B according to the manufacturer's instructions (Amersham Pharmacia Biotech). For immunoblotting, keratinocytes were plated onto type I collagen-coated tissue culture dishes and cultured over a time course (0 -5 days).
Soluble Protein Preparation-Soluble cell proteins were isolated by scraping cells into 500 l of HB buffer (10 mM HEPES, 1 mM EDTA, 0.32 M sucrose, and 25 g each of aprotinin, leupeptin, and pepstatin). Cells were disrupted by repeatedly passing through a 27-gauge needle, centrifuged 70,000 rpm for 30 min at 4°C, and the supernatant was transferred to new microcentrifuge tubes. 4.0 g of total protein per sample, as determined by Bradford protein assay (Bio-Rad), was mixed with Laemmli sample buffer containing 10% ␤-mercaptoethanol and resolved by SDS-polyacrylamide gel electrophoresis. Proteins were electrophoretically transferred to Immobilon TM -P polyvinylidene difluoride membrane (Sigma), blocked with 5% milk-TBS (20 mM Tris, 150 mM NaCl), and incubated with affinity purified epsin 3 antibodies (1:1000 dilution) in 5% milk, TBS, 1% Tween overnight at 4°C. Bound antibodies were visualized using horseradish peroxidase-linked anti-rabbit secondary antibody (1:2000 dilution) followed by detection using Lumi-GLO TM reagent according to the manufacturer's instructions (Cell Signaling Technology, Beverly, MA). For negative controls, blots were incubated with secondary IgG alone, or specific/nonspecific peptideadsorbed epsin 3 antibodies.
Localization of Epsin 3 in Collagen-activated Keratinocytes and Cutaneous Wounds by Immunofluorescence-Aliquots of keratinocyte cell suspension were plated onto collagen-coated coverslips and cultured for 3 days. For localization in ex vivo cutaneous wounds, punch biopsies of normal human skin (6 mm) were obtained and grown as explant cultures in serum-containing Dulbecco's modified Eagle's medium for 3 days. Following culture, samples were washed, fixed in 4% paraformaldehyde (Polysciences, Inc., Warrington, PA), and stored in phosphatebuffered saline containing 0.1% sodium azide until staining. For cryosectioning, punch biopsies were immersed in 30% sucrose overnight at 4°C, embedded in Tissue-Tek TM O.C.T. compound (Sikura Finetek, Torrance, CA), and 6-m thick sections were cut using a Reichert-Jung cryostat.
For immunolocalization of epsin 3, coverslips and explant sections were permeablized and incubated with 2.5 g/ml affinity-purified epsin 3 antibodies overnight at 4°C followed by incubation with TRITCconjugated AffiniPure donkey anti-rabbit IgG (Jackson Immuno-Research Laboratories, Inc., West Grove, PA). To visualize epsin 3 and clathrin within the same cells, coverslips were treated with clathrin monoclonal antibody X22 (1:1000 dilution) after incubation with antiepsin 3 antibodies. Clathrin antibody localization was visualized using Alexa-Fluor TM 488-goat anti-mouse IgG (Molecular Probes, Eugene, OR). For negative controls, slides were incubated with preimmune serum, secondary IgG alone, or specific/nonspecific peptide-adsorbed epsin 3 antibodies. Digital light photomicrographs were captured with a Polaroid DMC1 camera connected to a Nikon Eclipse E400 microscope. Whole mount immunofluorescence images were captured using a Hammamatsu C4742-9S digital camera connected to a Zeiss Axioplan 2 microscope using OpenLab scientific imaging software (Improvision, Inc., Boston, MA). Confocal images were obtained using a Leica TCS SP microscope with Leica TCS NT capturing software.

Keratinocyte Contact with Collagen Induces Epsin 3-Be-
cause keratinocyte contact with type I collagen specifically induces collagenase-1, a marker of the activated keratinocyte phenotype (10,11,19), we wanted to determine if novel activation-specific transcripts were induced by this ECM protein.
ddRT-PCR was used to compare transcripts expressed in keratinocytes freshly isolated from intact skin or after plating onto fibrillar type I collagen. cDNA clone GAP4G1, an unknown transcript potently induced in keratinocytes following collagen contact, was selected for characterization. As determined by Northern hybridization, cDNA GAP4G1 hybridized to a single 3.9-kilobase mRNA species that was induced Ͼ20-fold in collagen-activated keratinocytes when compared with cells freshly isolated from unwounded skin (Fig. 1). Sequence analysis revealed 100% identity to nucleotides 10237-10379 of human chromosome 17-clone hCIT.22_K_21 (accession number AC004590). Because this genomic clone had no association with a known gene, we used the internet-based Genscan 1.0 program to determine if the sequence surrounding GAP4G1 contained an open reading frame. A putative gene was identified containing an initiation codon, nine exons, and in-frame termination codon that, when translated, coded for a novel 632amino acid protein with a calculated molecular mass of 68 kDa. To confirm the predicted gene structure, we amplified the fulllength coding sequence using multiple pairs of gene specific inter-exonic primers. Sequencing of the resultant cDNA demonstrated a 9 exon, 1896-bp open reading frame as predicted (Fig. 2). Comparison with known proteins in the public data bases using a BLASTp search revealed that the encoded protein had not been previously reported and that it demonstrated remarkable similarity to, but was distinct from, members of the epsin protein family (Figs. 2 and 3A). In addition, and to confirm our initial Northern results (Fig. 1), we also amplified a 400-bp fragment unique to exon 7 of epsin 3 and used this cDNA to probe total RNA from freshly isolated and collagen-activated keratinocytes. As before, strong hybridization to a 3.9-kilobase mRNA was seen only in collagen-activated keratinocytes (not shown).
Comparison of the epsin 3 amino acid sequence with human epsins 1 and 2 demonstrates an identical tripartite domain structure common to this protein family (Figs. 2 and 3A). Conserved domains include a COOH-terminal consensus sequence of three NPF repeats that bind Eps15 (20), multiple DPW motifs that bind clathrin AP-2 (21,22), and a 150-amino acid protein module, the epsin NH 2 -terminal homology domain (ENTH domain) (22). The region of epsin 3 demonstrating the highest sequence identity is the ENTH domain that contains multiple regions of 100% conservation and 80 -82% sequence identity with human epsins 1 and 2 (Fig. 3A). Although the DPW and NPF domains of epsin 3 possess significantly less sequence identity when compared with epsins 1 and 2 (11-28 and 28 -34%, respectively), the three NPF motifs are 100% conserved. Furthermore, epsin 3 contains 4 conserved and 5 total DPW motifs, which is consistent with the variability demonstrated among each family member.
Intron/exon junctions of the epsin 3 gene conforming to the GT/AG rule for splice sites (23) were mapped by comparing cDNA and genomic sequences (Fig. 2). Exon-intron boundaries and the sizes of exons and introns are summarized in Table I. The most notable structural feature of the epsin 3 gene is that exon 1 codes for the 150-amino acid ENTH domain in its entirety. In contrast, the DPW domain encompasses exons 5-7 and the three NPF repeats are found in the distal sequence of exon 8 and throughout exon 9.
In addition to binding Eps15 and AP-2, epsins 1 and 2 exhibit type I cathrin-binding consensus sequences ( 257 LMDLADV and 283 LLDLMDAL, respectively) proximal to the DPW domain. An additional motif (LV(D/N)LD) was recently identified within the carboxyl-terminal segment of rat epsins 1 and 2 that acts cooperatively with the type I consensus to bind clathrin at an independent site (24). Epsin 3 contains a distal 506 LVNDLD low affinity clathrin-binding sequence as well as 300 ILDLADIF proximal to the second DPW repeat, which is reminiscent of a type I consensus as there are two intervening hydrophobic residues between two acidic amino acids (Figs. 2 and 3A).
Collagen-mediated Induction of Epsin 3 Is Transient and Requires the Native Triple Helical Substrate-To determine the molecular mechanisms regulating epsin 3, we assessed the expression of this transcript in keratinocytes at multiple time points following collagen contact. Northern blotting showed that epsin 3 mRNA was absent in keratinocytes freshly isolated from intact skin (Fig. 4A, 0 h). Expression was stimulated between 8 and 12 h after contact with the collagen matrix and became most prominent at 20 h (Fig. 4A). This collagen-mediated effect was transient, however, as epsin 3 mRNA signal progressively decreased to almost undetectable levels at 72 h, coincident with cells reaching confluence in culture.
The native, triple helical conformation of collagen is requisite for induction of collagenase-1 in keratinocytes as cells plated on gelatin express significantly lower levels of collagenase-1 mRNA and protein, although keratinocyte attachment and spreading is supported by this substrate (11). Keratinocytes were plated on fibrillar type I collagen or gelatin to determine if induction of epsin 3 was dependent on the native substrate. Primary human keratinocytes were isolated from intact skin and ddRT-PCR was performed on RNA obtained from cells immediately following isolation (0 h) or 24 h after plating on type I collagen (100 g/ml). One resultant cDNA exhibiting significant up-regulation (clone GAP4G1) was selected and used to probe a Northern blot containing total RNA (10 g/lane) from 0-and 24-h collagen-activated keratinocytes. The Northern blot confirmed that this cDNA represented an mRNA that was significantly induced in cells following contact with type I collagen. The blot shown is representative of data obtained from at least four separate experiments using RNA from individual donors.
As assessed by Northern hybridization, epsin 3 mRNA was reduced by 85% in cells cultured on gelatin when compared with that expressed in cells plated on collagen (Fig. 4B).
To determine if epsin 3 mRNA was translated, we generated polyclonal antibodies directed against a unique 20-amino acid sequence located between the epsin 3 DPW and NPF domains.
Keratinocytes were harvested from normal skin and cultured on collagen over a time course. Total or soluble cell proteins were isolated and epsin 3 was visualized by immunoblotting with affinity-purified antibodies. An immunoreactive 68-kDa species, the predicted molecular mass for epsin 3, was evident in keratinocytes 1 day after plating onto collagen and became Corresponding epsin 3 DPW motifs are underlined, NPF motifs are indicated by asterisks, and putative clathrin binding consensus sequences are bracketed. B, a dendrogram was generated using pairwise sequence alignments of known epsins using TreeView. Numerical data in the right-hand column represent percent sequence identity when aligned with human epsin 3. increasingly stronger on day 3 (Fig. 4C). In contrast, epsin 3 was absent in cells freshly isolated from intact skin (day 0) and after 5 days on collagen, which is consistent with the transient epsin 3 mRNA expression noted in collagen-activated keratinocytes by Northern hybridization (Fig. 4A). A second, smaller unknown immunoreactive band of ϳ50-kDa was visualized in keratinocyte whole cell lysates, but its expression was not affected by contact with collagen, suggesting that either the antibody bound a nonspecific protein or a currently uncharacterized species (Fig. 4C, left panel). When epsin 3 present in keratinocyte-soluble proteins was visualized, however, only the immunoreactive 68-kDa epsin 3 band was evident. Immunoblots performed following preadsorption of the epsin 3 antibody with antigenic peptide showed no immunoreactive bands (Fig.  4C, Ab ϩ peptide), whereas preadsorption with a nonspecific epsin 3 peptide had no effect (Fig. 4C, Ab ϩ N.S. peptide).
Collagen-activated Epsin 3 Partially Co-localizes with Clathrin and Shuttles to the Nucleus-To assess the subcellular distribution of epsin 3, keratinocytes were plated onto collagencoated coverslips and processed for immunofluorescence stain-

. Contact with native type I collagen induces epsin 3 expression by keratinocytes.
A, keratinocytes were plated onto dishes coated with type I collagen (100 g/ml). Total RNA was isolated at the indicated times and 10 g RNA/lane was resolved by gel electrophoresis. The RNA was transferred to Hybondϩ membrane and probed using random-primed 32 P-labeled epsin 3 and GAPDH cDNA probes. B, keratinocytes were plated on dishes coated with type I collagen (100 g/ml) or heat-denatured collagen (Gelatin, 100 g/ml). 10 g of total RNA/lane was resolved by gel electrophoresis, transferred, and probed using random-primed 32 P-labeled epsin 3 and GAPDH cDNA probes. C, keratinocytes were cultured on collagen over a time course and total cell or soluble proteins were isolated at the indicated times. Equal amounts of cell lysate were separated by electrophoresis in 10% SDS-polyacrylamide electrophoresis gel and electrophoretically transferred to polyvinylidene difluoride membrane. Membranes were immunoblotted with epsin 3 antibodies alone, or with antibodies preadsorbed with antigenic or nonspecific peptides. Equal loading was verified by Coomassie Blue staining (not shown). The data shown in each panel is representative of results obtained from three separate experiments using cells from individual donors. ing using affinity purified epsin 3 antibodies. Following 3 days on collagen, keratinocytes exhibited prominent cytoplasmic immunoreactivity for epsin 3 that concentrated in the perinuclear region (Fig. 5A). These findings are consistent with previous reports documenting the localization of epsins 1 and 2 in other cell types (21,24). Processing cells with preimmune serum resulted in nonspecific staining as only the outlines of suprabasal cells were highlighted (Fig. 5B).
Epsin 1 binds to clathrin directly and intracellular distribution studies have shown that a portion of the total epsin pool associates with clathrin-coated structures (21,24). To determine if epsin 3 demonstrated a similar pattern of localization, we double labeled cells with affinity purified epsin 3 antibodies (Fig. 5C) and monoclonal antibody X22 directed against the clathrin heavy chain (Fig. 5D). Consistent with the behavior of epsins 1 and 2, merging of the epsin 3 and clathrin images indicated that a subpopulation of epsin 3 co-localized to clathrin-coated structures (Fig. 5E, arrows). Vesicles that exhibited double labeling were predominantly located at the cell periphery where clathrin-coated pits would be assembling, whereas epsin 3 immunoreactivity was absent from a pool of clathrin located near the nucleus (Fig. 5, C-E, arrowhead).
Resolution of the epsin 1 ENTH domain crystal structure has revealed remarkable structural similarity to ␤-catenin armadillo repeats (25). In addition to its involvement in endocytosis, epsin 1 is capable of interacting with the transcription factor promyelocytic leukemia Zn 2ϩ finger protein and translocating to the nucleus where it may regulate transcription (25). Because the epsin 3 ENTH domain is highly conserved, we hypothesized that epsin 3 may exhibit similar behavior. To determine the ability of epsin 3 to shuttle between the nucleus and cytoplasm, collagen-activated keratinocytes were treated with leptomycin B, an antifungal antibiotic that blocks the Crm1dependent nuclear export pathway and induces the nuclear accumulation of proteins that have shuttling activity (26 -28). As shown in Fig. 5F, epsin 3 in keratinocytes treated transiently with leptomycin B accumulated in the nucleus (arrowheads), whereas control cells treated with vehicle alone exhibited prominent cytoplasmic localization (not shown).

Epsin 3 Is Expressed by Migrating Keratinocytes in Vitro and ex
Vivo-Keratinocytes grown on collagen under high calcium (1.8 mM) conditions recapitulate the epidermal wound healing response by forming subpopulations of migrating, proliferating, and differentiating cells (12). When viewed from above, differentiating keratinocytes are seen as islands of confluent cells (asterisks, Fig. 6, B and C) surrounded by hyperproliferative cells. Bordering the hyperproliferative cells, and often detached from them, are migrating keratinocytes (arrows, Fig. 6, B and C). Because collagen-induced epsin 3 mRNA and protein expression by keratinocytes was transient (Fig. 4, A and C), we examined whether expression of this protein was spatially restricted. Keratinocytes were isolated from intact skin and suspensions of low or high cell density were plated onto collagencoated coverslips. The low density cell suspension prolonged individual migrating keratinocytes for the duration of the experiment, whereas the higher cell density suspension promoted cell:cell contact that resulted in the formation of differentiating islands. In cultures predominantly made up of individual, migrating keratinocytes epsin 3 was prominently expressed in all cells examined (Fig. 6A). In differentiation promoting cultures, however, epsin 3 was spatially confined to keratinocytes migrating away from and bordering islands of differentiating cells (Fig. 6, B and C, arrows). Clathrin expression was not restricted to a particular cell population as positive staining was noted in both migrating and differentiating keratinocytes (Fig.  6D). Co-localization of epsin 3 with clathrin was evident in the migrating cells as seen previously (Fig. 6, C and D,  arrowheads).
Because epsin 3 was predominantly expressed by migrating keratinocytes in vitro, we assessed if this protein was expressed during cutaneous repair. Punch biopsies of normal human skin were cultured ex vivo over a time course and processed for immunofluorescence staining using affinity purified epsin 3 antibodies. During the culture period, keratinocytes at the cut edge of the biopsy undergo a wound healing response and are activated to migrate off of the basement FIG. 5. Epsin 3 is expressed by collagen-activated keratinocytes, partially co-localizes to clathrin-coated structures, and shuttles to the nucleus. A-E, keratinocytes were isolated from normal, intact skin and cultured on collagen-coated coverslips for 3 days. Following fixation, cells were processed for immunofluorescence staining using affinity-purified epsin 3 antibodies (A and C), preimmune serum (B), or monoclonal antibody X22 directed against clathrin heavy chain (D). Panel E represents the merged epsin 3 and clathrin images. Nuclei were visualized by staining with Vectashield TM -DAPI mounting medium. Keratinocytes grown on collagen displayed prominent staining for epsin 3 that was visualized as punctate vesicles throughout the cytoplasm and concentrated around the nucleus. Double labeling with the clathrin antibody revealed that epsin 3 partially co-localized to clathrin-coated structures at the cell periphery (arrows), whereas a pool of clathrin located near the nucleus (arrowhead) was devoid of epsin 3 immunoreactivity (C-E). No specific immunoreactivity was noted in keratinocytes processed with preimmune serum (B). F, collagen-activated keratinocytes were treated for 3.5 h with vehicle control (not shown) or leptomycin B (10 ng/ml, F), an inhibitor of Crm1-dependent nuclear export. Following treatment, cells were processed for epsin 3 immunofluorescence staining. Treatment with leptomycin B induced an accumulation of epsin 3 in the nucleus (F, arrowheads), whereas control cells exhibited prominent cytoplasmic staining (not shown), illustrating the ability of epsin 3 to shuttle between the cytoplasm and the nucleus. membrane and onto the denuded dermis (Fig. 7A, arrowheads). We have shown in previous studies that expression of collagenase-1 in this model mirrors that seen in vivo (15).
Following 3 days after wounding, epsin 3 immunoreactivity was predominantly localized to basal keratinocytes within the migrating epithelial tongue (Fig. 7, B and C, arrowheads). Paralleling our in vitro data, epsin 3 was specifically localized to basal keratinocytes at the cell-ECM interface, whereas suprabasal differentiating cells in the migrating tongue exhibited only background staining (Fig. 7, B and C). Epsin 3 signal progressively diminished away from the wound edge and was detectable at low levels in only a few basal cells of intact epidermis (Fig. 7B). Immunoreactivity was also noted in a subpopulation of cells scattered throughout the dermis, most of FIG. 6. Epsin 3 is expressed in a spatially distinct manner. Keratinocytes were isolated from normal, intact skin and increasing concentrations of cells (50 -150 l) were cultured on coverslips coated with 100 g/ml collagen for 3 days. After fixation, cells were processed for immunofluorescence staining using affinity purified epsin 3 (A-C) or clathrin X22 monoclonal (D) antibodies and visualized by confocal microscopy. Keratinocytes plated in a single cell suspension (50 l/ coverslip, panel A) displayed prominent staining for epsin 3 in the cytoplasm that concentrated in the perinuclear region. Keratinocytes plated at higher concentrations (150 l/coverslip, panels B-D) formed islands of differentiating and stratifying cells (*). Bordering these islands, and often detached from them are migrating keratinocytes (arrows). Epsin 3 was restricted to keratinocytes migrating away from and bordering the islands of differentiating cells.  were cultured ex vivo for 3 days and processed for epsin 3 immunofluorescence staining with affinity purified antibodies. A, hematoxilin and eosin-stained sections demonstrating migration of the epithelial tongue (arrowheads) away from the wound edge (arrow) (Ep, epidermis; D, dermis). B and C, prominent immunoreactivity for epsin 3 was seen in basal keratinocytes at the leading edge of migration (arrowheads) and in some basal keratinocytes adjacent to the wound edge. Staining intensity progressively diminished away from the wound edge and was reduced to background levels in intact, unwounded skin. A subpopulation of dermal cells also demonstrated a low level of immunoreactivity (B and C). Serial sections of those shown in panels A-C were processed with preimmune serum (D). The dashed line represents the border between the epidermis and dermis. Images B-D were generated by confocal microscopy. Bars, 100 m (A and B), 50 m (C and D).
which had a fibroblast-like appearance. No specific staining was detected in samples processed with preimmune serum (Fig. 7D) or epsin 3 antibodies preadsorbed with antigenic peptide (not shown).

Epsin 3 Expression Is Restricted to Epithelial Wounds and Pathologies with Altered Cell-Extracellular Matrix
Interactions-Our findings demonstrating the spatially confined expression of epsin 3 in migrating keratinocytes suggested that, unlike the previously described members of this protein family, epsin 3 was expressed in a restricted manner. To analyze expression in various tissues, we hybridized total RNA from multiple tissues using an epsin 3 cDNA probe. Surprisingly, epsin 3 was not expressed in multiple fetal or most adult tissues examined (summarized in Table II). A low level of epsin 3 mRNA (0.3-fold over background) was detected in adult stomach, but only after extended exposure of the x-ray film (Fig.  8A). This finding was variable, however, as hybridization was not found in stomach total RNA obtained from a separate pool of RNA donors (not shown). We also screened dermal fibroblasts and endothelial cells, two additional cell types activated during wound repair. Epsin 3 was not detected in total RNA harvested from resting or activated cells (not shown).
Because epsin 3 was not expressed in any resting tissues or cultured cells examined (with the exception of collagen-activated keratinocytes), we hypothesized that expression may be specific to wounded or pathologic tissues in which epithelial cells undergo altered cell-ECM interactions. This may explain our results demonstrating variable and low levels of epsin 3 mRNA in stomach, as the pooled samples may have been contaminated with wounded tissue (i.e. gastric ulceration). Expression of epsin 3 in wounded or pathologic tissues was determined by amplifying a 601-bp fragment of epsin 3 by RT-PCR. Interestingly, epsin 3 mRNA was expressed in total RNA isolated from chronic cutaneous wound, basal cell carcinoma, and ulcerative colitis, all wounds or pathologies that undergo altered cell-ECM interactions (Fig. 8B). In contrast, expression was not found in normal, unwounded skin (Fig. 8B). DISCUSSION The epsins are constitutively expressed genes that demonstrate a wide tissue distribution and are purported to provide a role in mediating the assembly and internalization of clathrincoated pits (21,22). Originally identified by screening for proteins that interact with the EH (Eps15 homology) domains of Eps15 (20, 29) and clathrin AP-2 (30), epsin functions as a molecular bridge to facilitate the ordered assembly of these molecules into the clathrin lattice at the plasma membrane. In addition, epsin contains two clathrin binding consensus sequences that have been shown to act cooperatively in clathrin interaction (24). Together, these data suggest that epsin is requisite for coordinating a specific molecular architecture prior to coated pit invagination. In fact, perturbation of epsin function in fibroblasts by overexpression or antibody injection potently inhibits this process (21,22).
The findings described in this report identify a novel epsin that retains the conserved structural features common to this gene family. Epsin 3 exhibits a significant degree of sequence identity when compared with other epsins and contains the ENTH protein module that is highly conserved from yeast to humans. Epsin 3 demonstrates greater sequence diversity distal to the ENTH domain, yet multiple DPW and three NPF motifs shown to be required for Eps15 and AP-2 binding (21,22), respectively, are fully conserved. We therefore hypothesize that epsin 3 provides a similar function in activated epithelial cells that is consistent with previously described members of this gene family.
The assumption that epsin 3 functions similarly to other epsins is strengthened by our in vitro data demonstrating co-localization with clathrin and shuttling between the cytoplasm and the nucleus. When collagen-activated keratinocytes were double-labeled with epsin 3 and clathrin antibodies, we  8. Expression of human epsin 3 mRNA in resting, wounded, and pathologic tissues. A, a Northern blot containing total RNA from multiple tissues was hybridized with 32 P-labeled epsin 3 and GAPDH cDNA probes. Following an extended exposure, faint hybridization to a single 3.9-kilobase mRNA band representing epsin 3 was detected in stomach. B, epsin 3 mRNA expression was detected in wounded and pathologic tissues by RT-PCR. The epsin 3 PCR product identity was verified by restriction digestion and sequence analysis. GAPDH RT-PCR was also accomplished to verify equal loading of RNA in each tissue sample. Epsin 3 was detected in chronic cutaneous wound, basal cell carcinoma, and ulcerative colitis. In contrast, epsin 3 was not present in normal skin. RNA from collagen-activated keratinocytes was used as a positive control.
found that a subpopulation of epsin 3-labeled structures partially co-localized with clathrin as has been shown for epsins 1 and 2 in other cell types (22,24). We also found that epsin 3 has the capacity to shuttle between the cytoplasm and the nucleus as inhibition of the Crm1-dependent nuclear export pathway resulted in a prominent accumulation of epsin 3 in the nucleus. Thus, our data demonstrate that epsin 3 retains the structural and functional behavior previously described for other members of this family and it is reasonable to hypothesize a similar function for epsin 3 in activated keratinocytes.
In contrast to its structural and behavioral similarity to the epsin protein family, epsin 3 is the first member to diverge from a constitutive pattern of expression. Prior studies have speculated that epsin provides a housekeeping function in cells for clathrin-mediated endocytosis. We found, however, that epsin 3 was not expressed in keratinocytes isolated from normal, unwounded skin, but induced specifically and invariantly in migrating cells following contact with fibrillar type I collagen.
Previous studies from our group and others have shown that keratinocyte contact with collagen induces collagenolytic activity in migrating cells, a gene expression pattern identical to activated keratinocytes engaged in re-epithelialization in vivo (11,31). The results reported here further underscore the profound influence that collagen contact has on keratinocyte behavior during wound healing and supports the hypothesis that in addition to soluble factors, contact with dermal ECM induces keratinocyte activation by stimulating the expression of transcripts during repair.
In addition to collagen-induction of epsin 3 in keratinocytes, we further determined that expression of this transcript is restricted to tissues that demonstrate alterations in cell-ECM contacts as occurs during wound healing and tissue morphogenesis. In an exhaustive tissue survey we found that epsin 3 was not expressed in resting, homeostatic tissues, but rather in epithelial wounds or pathologies that typically exhibit a disrupted basement membrane. This is in marked contrast to other epsins, which are widely distributed in resting tissues, and suggests that epsin 3 may play a role in regulating the function of epithelial cells in which cell-ECM interactions have been altered.
During the preparation of this manuscript, a full-length cDNA was identified in human colon (accession number AK000785) coding for a putative protein having nearly 100% sequence identity to epsin 3 described here. In addition, following the release of our sequence (accession number AF324241), a coding sequence predicted from cDNA AK000785 was submitted and named human epsin 3 (accession number NM_017957). Interestingly, cDNA NM_017957 was cloned from colon total RNA, yet our probing of multiple tissue blots for epsin 3 only produced faint and variable hybridization in the stomach (Fig. 8A). Furthermore, we were unable to verify hybridization in total RNA from colon, except in samples obtained from ulcerative colitis (Table I). We hypothesize that the variable expression of epsin 3 found in stomach resulted as a consequence of wounded epithelium present in the donor tissue used. Pooled RNA from multiple donors is blotted onto each tissue Northern used in our studies and RNA from an injured epithelial lining of a single donor would produce a faint band due to dilution of the transcript. Perhaps the colon sample used to generate cDNA NM_017957 included RNA from wounded tissue, which would explain our inability to identify epsin 3 in normal colon.
As keratinocytes complete re-epithelialization they restore cell:cell contacts, establish an intact basement membrane, and revert to a gene expression program that supports proliferation and differentiation. Interestingly, we found that epsin 3 ex-pression was limited to keratinocytes actively involved in migration and absent in differentiating cells in vitro and ex vivo. The formation of polarized cell:cell contacts and synthesis of a new basement membrane are the most likely candidates that down-regulate epsin 3 expression in collagen-activated keratinocytes. Because we assessed expression in higher cell density cultures over 3 days, a time in which both cell:cell contacts are established and basement membrane proteins are synthesized, we cannot yet separate the effects of either in determining how down-regulation of epsin 3 expression is mediated. Because epsin 3 expression parallels the spatial and temporal pattern of collagenase-1 in migrating keratinocytes, we speculate that both phenomena contribute to epsin 3 down-regulation in confluent cells and in intact skin after healing has completed.
The spatially and temporally distinct expression pattern of epsin 3 noted in keratinocytes in vitro and ex vivo indicates that this protein serves an important function in activated, migrating cells. The ability of keratinocytes to respond to injury requires the processing of multiple extracellular signals, including soluble ligands binding to cognate receptors and exposure to new ECM. Signaling through receptor tyrosine kinases by growth factors such as epidermal growth factor, keratinocyte growth factor, and transforming growth factor-␣ elicits a number of transient cell responses including enhanced cell proliferation and migration (2). As healing progresses, however, epidermal cells lose their responsiveness to these signals and revert to a gene expression program that promotes differentiation. Activated growth factor receptors, such as the epidermal growth factor receptor (32) and keratinocyte growth factor receptor (33), are rapidly internalized via endocytosis as cell migration ceases and this pathway serves as an important regulatory mechanism to control cell surface receptor expression and downstream signaling events. A mechanism by which activated receptors are internalized via epsin 3-enhanced endocytosis would have profound effects on the ability of keratinocytes to respond to extracellular stimuli and mediate the wound healing response.
In summary, our results demonstrate that epsin 3 is induced in wounded or pathologic epithelial tissues exhibiting altered cell-ECM interactions as occurs following basement membrane disruption and contact with underlying collagen. Epsin 3 demonstrates a regulated and specific pattern of expression, unobserved for any other known epsin, and indicates that the cell machinery required for epithelial repair differs fundamentally from that required to maintain tissue homeostasis. Although a function for epsin 3 has yet to be determined, its novel temporal and spatial expression pattern suggests an role in regulating cell responses to the extracellular environment during morphogenetic events.