Mouse DESC1 Is Located within a Cluster of Seven DESC1 -like Genes and Encodes a Type II Transmembrane Serine Protease That Forms Serpin Inhibitory Complexes*

We report the identification and functional analysis of a type II transmembrane serine protease encoded by the mouse differentially expressed in squamous cell carci- noma ( DESC ) 1 gene, and the definition of a cluster of seven homologous DESC1 -like genes within a 0.5-Mb re- gion of mouse chromosome 5E1. This locus is syntenic to a region of human chromosome 4q13.3 containing the human orthologues of four of the mouse DESC1 -like genes. Bioinformatic analysis indicated that all seven DESC1 -like genes encode functional proteases. Direct cDNA cloning showed that mouse DESC1 encodes a mul- tidomain serine protease with an N-terminal signal anchor, a SEA (sea urchin sperm protein, enterokinase, and agrin) domain, and a C-terminal serine protease domain. The mouse DESC1 mRNA was present in epi- dermal, oral, and male reproductive tissues and directed the translation of a membrane-associated 60-kDa N -glycosylated sequence analysis were inserted in-frame between the SphI and XbaI sites of the insect cell expression vector pMIB/V5-His (Invitrogen). The resulting expression plasmids were transfected into Sf9 insect cells using Cellfectin (Invitrogen), and several polyclonal cell lines were isolated by selection with blasticidin S as recommended by the manufacturer. Conditioned medium was collected from pools of stably transfected Sf9 cells for subsequent analysis. For purification of the DESC1 serine protease domain, phosphate buffer (pH 7.5) and NaCl were added to the conditioned medium to a final concentration of 20 m M and 0.5 M , respectively. The solution was applied to a cobalt-charged HiTrap chelating column (Amersham Biosciences) using the AKTA fast protein liquid chromatography system. Chelated proteins were eluted with a linear gradient of 20–200 m M imidazole in 20 m M phosphate buffer, 0.5 M NaCl, and the purified mouse DESC1 serine protease domain was aliquoted and stored at (cid:2) 80 °C use. Zymographic Analysis of Mouse DESC1 Serine Protease Do-main— and casein zymography was performed on 11% Tris/ glycine gels containing either 0.5% gelatin or 0.1% (cid:3) -casein (Sigma) as described previously were developed overnight in 0.1 M glycine, pH 8.0, at 37 °C, stained with 0.1% Coomassie destained in a 30% methanol and acetic acid to visualize of substrate medium serine protease full-length Chromogenic Hydrolysis Assay— The recombinant expression in insect cells of the portion of the mouse DESC1 cDNA that was predicted to encode the serine protease domain led to the generation of a protease zymogen that could be converted to a functionally active protease by exposure to trypsin. The zymogen demonstrated both gelatinolytic and caseinolytic activities under zymographic conditions.

Pericellular proteolysis is essential to all aspects of vertebrate life, including development, tissue homeostasis, tissue remodeling, tissue repair, and reproduction, and dysregulated pericellular proteolysis is causally related to a large number of diseases in humans. Within the past century, the most extensive studies of pericellular proteolysis in health and disease focused on defining the contribution of a relatively limited number of proteases belonging to the families of plasminogen activators, matrix metalloproteinases, and cathepsins (1)(2)(3)(4)(5)(6). However, the recent completion of the mouse and human genome sequences, and the generation of extensive mouse and human EST 1 data bases, facilitated an explosion in the discovery of novel candidate protease genes, indicating that a vastly larger repertoire of pericellular proteases may be engaged in these processes than previously anticipated (7,8). Particularly noteworthy in this context was the unveiling of an unexpectedly large family of type II transmembrane serine proteases (TTSPs) (9 -23). This rapidly expanding family of serine proteases is defined by the presence of an N-terminal signal anchor and a C-terminal serine protease domain, separated by a stem region containing an array of protein domains that varies widely between individual TTSPs (24).
The exploration of the physiological function of most members of the TTSP family is in its infancy. An exception is enteropeptidase, whose function as an initiator of a zymogen cascade leading to digestive enzyme activation is well established (25)(26)(27)(28). Likewise, gene targeting studies in mice suggested a function for hepsin in adult liver homeostasis (29,30). Most recently, critical functions of the two TTSPs matriptase/ MT-SP1 and TMPRSS3 in epidermal and inner ear development, respectively, were revealed by gene targeting in mice (31,32) and by the identification of loss of function mutations in TMPRSS3 as the underlying cause of congenital and childhoodonset sensorineural nonsyndromic deafness in humans, respectively (33)(34)(35). Furthermore, several recent studies have demonstrated the aberrant expression of many TTSPs in human cancer, and in some cases, increased TTSP expression correlated with a poor prognosis (15, 17, 36 -44). Collectively, these studies have provided strong evidence that the TTSP family contributes to both human health and human disease.
To facilitate the functional analysis of the TTSP complement of genes in mice, we have identified and characterized the murine orthologue of human DESC1, a putative TTSP gene that was previously identified as an mRNA transcript expressed in certain normal human tissues but not in their corresponding malignancies (19). We show that mouse DESC1 encodes a type II transmembrane glycoprotein, which displays both caseinolytic and gelatinolytic activities, and is a functional serine protease with an Arg-subsite preference. Most interestingly, DESC1 forms inhibitory complexes with plasminogen activator inhibitor-1 (PAI-1) and protein C inhibitor (PCI or PAI-3), indicating that the TTSP family of proteases may be targets for serine protease inhibitor (serpin) inhibition, and suggesting that the two serpins interact with DESC1 to regulate the function of the protease in tissues such as epidermis, salivary gland, and epididymis. Furthermore, we show that the DESC1 gene is located in a cluster containing seven highly homologous genes within a short 0.5-Mb region of mouse chromosome 5E1 with a corresponding cluster of five human DESC1-like genes on human chromosome 4q13.3. All 12 candidate genes are predicted to encode functional serine proteases, thus further expanding the repertoire of functional TTSPs in the mouse.

EXPERIMENTAL PROCEDURES
Bioinformatic Analysis of the Mouse Chromosome 5E1 TTSP Cluster-BLAST analysis with the published human DESC1 cDNA sequence was performed by using the NCBI and Celera Discovery System mouse genome data bases (www.ncbi.nlm.nih.gov; www.celera.com; and www.genome.ecsc.edu) and the NCBI mouse EST data base (www. ncbi.nlm.nih.gov/BLAST/Blast.cgi) for the mouse orthologue/homologues of DESC1 using the published human DESC1 cDNA sequence (19). The rat orthologue of mouse DESC1 was identified by screening the NCBI EST data base with the mouse cDNA sequence. Additional mouse DESC1 homologues were identified by using the predicted mouse DESC1 cDNA as in silico probe to scan the mouse chromosome 5E1 region using the Celera Discovery System data base. The exonintron boundaries of DESC1-like candidate genes were identified using FuzzyFinder (www.genome.nci.nih.gov/tools). Signal anchor predictions were done using SignalP version 2.0 (www.cbs.dtu.dk); potential protein coding regions were identified with SMART (www.smart.ox. ac.uk), and potential post-translational modifications were determined using NetPhos2.0 (www.cbs.dtu.dk) and NetNGlyc1.0 (www.cbs.dtu. dk). Alignment of multiple predicted DESC1-like protein sequences was done by using ClustalW.
cDNA Cloning of Mouse DESC1-A mouse skin cDNA library was constructed using the SMART PCR cDNA synthesis kit from Clontech. Gene-specific primers containing SalI linkers were designed to the 5Ј end (5Ј-AGTCGTCGACATGAACACACTCAGTCACGGACC-3Ј) and the 3Ј end (5Ј-ATCAGTCGACGATACCAGTGTTGGAAGCAATC-CAGTG-3Ј) of the predicted mouse DESC1 cDNA. PCR amplification was performed with Platinum High Fidelity Supermix (Invitrogen) using a GeneAmp 9700 thermocycler (Applied Biosystems, Foster City, CA) using 32 cycles of denaturing (94°C, 30 s), annealing (60°C, 30 s), and extension (69°C, 45 s), followed by a 3-min final extension at 69°C. The amplified product was cloned into pBSII-SK Ϫ (Stratagene, La Jolla, CA), and the identity and sequence integrity of the DESC1 cDNA were confirmed by nucleotide sequencing using multiple overlapping internal primers.
Analysis of DESC1 mRNA Expression in the Mouse-Primary mouse keratinocytes and fibroblasts were isolated from the epidermis of 1-dayold mice exactly as described (45). Total RNA from keratinocytes cultured for 3 days, fibroblasts, and 1-day-old mouse skin were isolated using Trizol reagent (Invitrogen), followed by sodium acetate/ethanol precipitation. Total RNA from mouse liver, brain, thymus, heart, lung, spleen, testicles, ovaries, and kidneys was purchased from Ambion (Austin, TX). First strand synthesis was carried out with 1 g of total RNA using RETROscript (Ambion) following the manufacturer's protocol. PCR was performed as described above using 30 cycles of denaturing (94°C, 30 s), annealing (60°C, 30 s), and extension (69°C, 30 s), followed by a 3-min final extension cycle at 69°C. Specific primers (5Ј-CCCAATGTTGATCCTGAGTCAG-3Ј and 5Ј-AGCAGCAGTCACAA-GCCATGTG-3Ј), corresponding to nucleotides 511-532 and 729 -750 of the mouse DESC1 cDNA sequence, or the mouse GAPDH-specific primers (5Ј-ACCACAGTCCATGCCATCAC-3Ј and 5Ј-TCCACCACCCCCT-GTTGCTGTA-3Ј) followed the same PCR parameters except the number of cycles was reduced to 25. PCR amplification products were separated on a 1.5% agarose gel and stained with ethidium bromide for UV visualization. For cDNA array screening, the full-length mouse DESC1 cDNAs or a mouse ubiquitin control cDNA probe was labeled with 32 P using Ready-to-Go random prime DNA labeling beads (Amersham Biosciences). A mouse RNA master blot from Clontech was prehybridized for 1 h with ExpressHyb (Clontech) containing 150 g/ml sheared and denatured salmon sperm DNA. After prehybridization, 1.5 ϫ 10 6 cpm of labeled probe was added, and the blot was hybridized overnight at 65°C. The blot was washed extensively with 2ϫ SSC containing 0.5% SDS and then exposed to a PhosphorImager screen (Amersham Biosciences) for visualization. Quantitation of the hybridization signal was performed with a PhosphorImager and ImageQuant software (Amersham Biosciences).
Expression of Full-length Recombinant Mouse DESC1 and Western Blot Analysis-The mouse DESC1 cDNA was truncated just prior to the predicted TAA stop codon by PCR amplification as described above. The amplification product was verified by sequencing and was subcloned into the PmlI site of the mammalian expression vector pMH (Roche Applied Science) to extend the C terminus of the mouse DESC1 protein with a hemagglutinin (HA) epitope. Human cervical adenocarcinoma (HeLa) cells were maintained in Dulbecco's modified Eagle's medium containing sodium pyruvate and supplemented with 10% fetal bovine serum, penicillin/streptomycin, and 2 mM L-glutamine and were cultured in a 5% CO 2 -buffered tissue culture incubator. Prior to transfection, the cells were seeded on poly-D-lysine-coated 10-cm 2 tissue culture dishes and were transfected with the mouse DESC-HA expression plasmid using Polyfect Reagent (Qiagen, Valencia, CA). After 24 h, the cells were washed twice with ice-cold phosphate-buffered saline (PBS) and lysed by scraping with a cell scraper in a buffer containing 10 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 0.01% SDS, and 1ϫ protease inhibitor mixture (Sigma), followed by a 10-min incubation on ice. The cell lysate was then clarified by centrifugation at 15,000 ϫ g at 4°C. The resulting protein lysate was boiled for 10 min in Laemmli buffer containing ␤-mercaptoethanol. Forty g of protein lysate was separated on pre-cast 4 -12% BisTris SDS-polyacrylamide gels (Invitrogen), and the separated proteins were transferred to a PVDF membrane. The membrane was blocked for 1 h at room temperature in TBS/Tween 20 with 5% (w/v) nonfat dry milk. Primary anti-HA mouse monoclonal antibody (Covance Laboratories, Richmond, CA) was added overnight at 4°C at a dilution of 1:1000 in TBS/Tween 20 with 5% (w/v) dry milk, and the membrane was washed extensively with TBS/Tween 20 and incubated with a secondary alkaline phosphatase-conjugated antibody. Bound antibodies were visualized by using nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate tablets (Roche Applied Science) as recommended by the manufacturer.
Deglycosylation of Mouse DESC1-Protein lysate (30 mg) was denatured in 1ϫ glycoprotein-denaturing buffer (New England Biolabs, Beverly, MA) containing 0.5% SDS and 1% ␤-mercaptoethanol by heating to 100°C for 10 min. Sodium citrate (pH 5.5) was then added to a final concentration of 50 mM. Endoglycosidase H (2000 units/reaction) was added, and the deglycosylation reaction was performed at 37°C for 1.5 h prior to separation by SDS-PAGE and Western blot analysis as described above.
Confocal Analysis of Mouse DESC1 Membrane Localization and Topology-HeLa cells were seeded on poly-D-lysine-coated coverslips and transfected with mDESC-HA and/or GFP expression plasmids as described above. The transfected cells were washed in PBS 36 h later, fixed in 4% paraformaldehyde in PBS for 30 min at 4°C, and washed three times with PBS containing 3% bovine serum albumin (BSA). After washing, the slides were incubated with a 1:1000 dilution of anti-HA monoclonal antibody (Covance Laboratories) in PBS with 3% BSA for 1 h and washed three times with ice-cold PBS containing 3% BSA. The coverslips were then incubated with AlexaFluor-546-conjugated goat anti-mouse antibodies (Molecular Probes, Inc., Eugene, OR) in PBS with 3% BSA for 1 h at room temperature, washed with ice-cold PBS containing 3% BSA, mounted, and analyzed by confocal microscopy with a Leica SP2 AOBS confocal microscope (Bannockburn, IL).
Expression of Mouse DESC1 Serine Protease Domain and Full-length Soluble Mouse DESC1-Sf9 insect cells were maintained in 420 SFM medium (JRH Biosciences, Lenexa, KA) and cultured at 27°C. The serine protease domain (nucleotides 586 -1329) of mouse DESC1 was amplified by PCR using the forward primer, 5Ј-CTTGCATGCGAAGT-TGTGGGACACGACGGAACAAATC-3Ј, and the reverse primer, 5Ј-CC-TCTAGAAATACCAGTGTTGGAAGCAATCC-3Ј. Full-length soluble mouse DESC1 (nucleotides 184 -1329) was amplified by PCR using the forward primer, 5Ј-ACGCGCATGCACAGATACAATCACAGGAGAAC-CTACAATTAC-3Ј, and the reverse primer, 5Ј-CCTCTAGAAATACCA-GTGTTGGAAGCAATCC-3Ј. The amplification products were verified by DNA sequence analysis and were inserted in-frame between the SphI and XbaI sites of the insect cell expression vector pMIB/V5-His (Invitrogen). The resulting expression plasmids were transfected into Sf9 insect cells using Cellfectin (Invitrogen), and several polyclonal cell lines were isolated by selection with blasticidin S as recommended by the manufacturer. Conditioned medium was collected from pools of stably transfected Sf9 cells for subsequent analysis. For purification of the DESC1 serine protease domain, phosphate buffer (pH 7.5) and NaCl were added to the conditioned medium to a final concentration of 20 mM and 0.5 M, respectively. The solution was applied to a cobalt-charged HiTrap chelating column (Amersham Biosciences) using the AKTA fast protein liquid chromatography system. Chelated proteins were eluted with a linear gradient of 20 -200 mM imidazole in 20 mM phosphate buffer, 0.5 M NaCl, and the purified mouse DESC1 serine protease domain was aliquoted and stored at Ϫ80°C until use.
Zymographic Analysis of the Mouse DESC1 Serine Protease Domain-Gelatin and casein zymography was performed on 11% Tris/ glycine gels containing either 0.5% gelatin or 0.1% ␤-casein (Sigma) as described previously (45). The gels were developed overnight in 0.1 M glycine, pH 8.0, at 37°C, stained with 0.1% Coomassie Blue, and destained in a 30% methanol and 10% acetic acid solution to visualize the zones of substrate lysis.
Proteolytic Activation of the Mouse DESC1 Zymogen-Purified mouse DESC1 serine protease domain or Sf9-conditioned medium containing either the mouse DESC1 serine protease domain or full-length soluble mouse DESC1 was incubated with trypsin-agarose beads (Sigma T-1763) for 10 min at 25°C. The beads were removed by centrifugation, and the activated DESC1 was re-purified by cobalt-charged HiTrap chelating chromatography as above.

Identification of the Mouse DESC1 Gene within a Gene Cluster That Encodes Seven DESC1-like TTSPs-
We searched for a mouse orthologue and homologues of the human DESC1 gene using the Celera and NCBI genome data bases, and we identified a candidate orthologue that was located in a syntenic region of mouse chromosome 5E1 (hereafter, mouse DESC1) and several homologues (see below). Inspection of the predicted mRNA sequence of mouse DESC1 revealed a 1329-base long open reading frame that contained an in-frame initiation codon and termination codon (Fig. 1A). A consensus polyadenylation sequence was located 340 nucleotides downstream of the termination codon, beginning at position 1669. Translation of the deduced mouse DESC1 mRNA yielded a protein that is 442 amino acids long with a calculated molecular mass of 50,022 kDa (Fig. 1A). Examination of the hypothetical protein using multiple different protein analysis programs demonstrated that the deduced amino acid sequence had all of the necessary components of a TTSP (Fig. 1, A and B). Downstream of the initiating methionine, there was a short hydrophobic region located between amino acid 35 and 62 that was predicted by the SignalP application to encode the transmembrane signal anchor. This putative signal anchor was followed by a SEA (sea urchin sperm protein, enterokinase, and agrin) domain that spanned amino acids 67-188 and a serine protease domain that extended from amino acids 210 to 436, as predicted by the SMART program. A histidine residue was found at position 251 within a conserved Ala-Ala-His-Cys motif. The second amino acid in the catalytic triad, aspartic acid, was located at position 296 in the conserved motif Asp-Ile-Ala-Leu. The active site serine residue was located in the conserved motif Gly-Asp-Ser-Gly-Gly at position 392. The activation domain, Arg-Ile-Val-Gly-Gly, a sequence that is highly conserved in serine proteases, was found at amino acid positions 210 -214. The cytoplasmic tail of the predicted protein was found to contain two putative threonine phosphorylation sites. The phosphorylation status of these residues was not investigated in this report but may play a role in intracellular signaling events. Three potential N-linked glycosylation sites were also identified ( Fig. 1, A and B). All glycosylation sites followed the conserved consensus motif (Asn-X-Ser or Asn-X-Thr) and were found at amino acid positions 93, 184, and 201. Thus, from the bioinformatic analysis of the predicted DESC1 protein, all of the necessary components of a TTSP were identified. This includes the presence of a cytoplasmic domain, N-terminal signal anchor, a stem region SEA domain, and a serine protease domain containing the necessary conserved catalytic residues (His, Asp, and Ser) within the appropriate context.
Comparison of predicted mouse DESC1 protein to the predicted human and rat protein sequences revealed a high level of sequence identity (Fig. 2). Human and mouse DESC1 shared 72% sequence identity over the entire length of the protein. Likewise, the predicted rat and mouse DESC1 sequences displayed a 75% amino acid identity. The predicted rat DESC1 sequence contained a longer cytoplasmic tail than human DESC1, and there was a protein sequence in the putative cytoplasmic tail that was encoded by an exon that was only present in rat DESC1 ( Fig. 2 and data not shown). The high degree of identity between the predicted human DESC1 protein sequence and the protein predicted to be encoded by the identified mouse gene prompted us to designate this gene mouse DESC1.
Further analysis of the chromosomal region that contained the mouse DESC1 gene revealed the close proximity of six additional putative genes with high homology to the mouse DESC1 gene. All seven candidate genes were clustered within a 0.5-Mb region of chromosome 5E1 (Fig. 3A), and they were all located in the same orientation relative to transcription. One gene encoded the murine orthologue of HAT-1 (11,50). ESTs and predicted partial or complete cDNA sequences have been reported previously for the five remaining genes under the 2 S. M. Réhault, manuscript in preparation.  (7,8,50). The comparative analysis of the syntenic region of human chromosome 4q13.3 revealed the presence of five DESC1-like genes. Most interestingly, only four of these candidate genes appeared to be orthologues of mouse DESC1-like genes (MAT, mDESC3, HATlike 4, and mDESC1). The fifth human DESC1-like gene, DESC2 (7), did not have a recognizable mouse orthologue in the 5E1 cluster or other sites in the mouse genome (Fig. 3A, lower  panel). FuzzyFinder, SignalP version 2.0, and SMART programs were used to identify the complete putative coding regions of all seven mouse candidate genes, and the predicted protein sequences were compared by multiple sequence alignment (Fig. 3B). This analysis revealed a high overall sequence homology and indicated that all seven members have an identical modular structure that was distinct from other TTSPs. This included a short, but variable, N-terminal cytoplasmic tail, the putative signal anchor, a single SEA domain, and a C-terminal serine protease domain of the chymotrypsin fold. Moreover, the serine protease activation site, the triad of catalytic amino acids, and the six cysteines required to form the three canonical disulfide bridges of the serine protease domain were present in all seven proteases. This analysis indicated that the mouse DESC1 gene is located in a gene cluster that additionally may encode six functional proteases that are all closely related, thus defining a large cluster of TTSPs in this chromosomal region.
Mouse DESC1 mRNA Is Expressed in Epidermis, Salivary Gland, and Epididymis-Initial in situ hybridization studies using two nonoverlapping mouse DESC1 antisense probes as well as an initial Northern blot analysis indicated that the DESC1 mRNA was present at fairly low levels in tissues, which precluded the convenient analysis of the expression of the DESC1 gene by these methods (data not shown). The expression of the mouse DESC1 gene was therefore analyzed by the more sensitive methods of RT-PCR analysis and dot-blot hybridization. For the RT-PCR analysis, primers spanning exons 5-7 were designed to prevent the amplification of contaminating DESC1 DNA gene sequences. This analysis showed a robust expression of mouse DESC1 mRNA in the skin (Fig. 4A, lane 1) but a low abundance of the mRNA in all other tissues examined, including liver, brain, thymus, heart, lung, spleen, ovary, testis, and kidney (Fig. 4A, lanes 4 -12). To delineate further the expression of DESC1 in mouse skin, total RNA was isolated from either primary cultures of skin fibroblasts or from primary skin keratinocytes, and the RNA was subjected to the RT-PCR analysis. A very robust expression was apparent in the primary keratinocytes (Fig. 4A, lane 3), whereas no mRNA was FIG. 2. Alignment of the deduced amino acid sequences of the mouse, human, and rat DESC1 proteins. The predicted mouse (mDESC1), rat (rDESC1), and human (hDESC1) protein sequences were aligned with the ClustalW program. Identical amino acids are boxed and shaded in gray. The eight conserved cysteine residues in the serine protease domain of the three candidate proteases are highlighted in the consensus sequence. Mouse, human, and rat DESC1 orthologues share a 60% overall amino acid identity.
FIG . 3. The mouse and human DESC1 genes are located in a large cluster of homologous DESC1-like genes. A, gene location of mouse and human DESC1-like genes; B, multiple alignment of predicted amino acid sequences of the corresponding mouse gene products. Bioinformatic analysis of the Celera Discovery System and the public mouse and human genome and EST data bases were used to identify the 12 candidate genes detected in the primary fibroblasts (Fig. 4A, lane 2), suggesting that DESC1 expression in the mouse is confined to the epidermis. The parallel analysis of DESC1 expression by dot blot hybridization using the full-length mouse DESC1 cDNA as a probe (Fig. 4B) confirmed the low level of expression of mouse DESC1 mRNA in most tissues examined. Furthermore, this analysis revealed an additional robust expression of DESC1 mRNA in salivary glands (Fig. 4B, 11th column) and in the epididymis (Fig. 4B, 3rd column), two tissues not represented in the RT-PCR analysis.
Mouse DESC1 Encodes a 60-kDa Glycoprotein-To analyze the protein product of the mouse DESC1 gene, a full-length mouse DESC1 cDNA was generated by high fidelity RT-PCR from total RNA isolated from newborn mouse skin. The cDNA was sequenced, confirmed to be identical to the predicted cDNA, and was inserted into a mammalian expression plasmid that furnished the expressed protein with an additional 9amino acid HA epitope tag at the C terminus to facilitate immunological detection. HeLa cells were transfected with either an empty expression plasmid or with the HA-tagged mouse DESC1-encoding expression plasmid, and protein extracts from the transfected cells were analyzed by Western blot using antibodies directed against the HA tag. This analysis revealed the presence of a major ϳ60-kDa immunoreactive protein that was present in cells transfected with the HA-tagged mouse DESC1 expression plasmid (Fig. 5, lane 3) but not in cells transfected with the empty expression plasmid (Fig.  5, lane 1). This apparent molecular mass is significantly higher than the predicted molecular mass of 50 kDa, suggesting that mouse DESC1 is subjected to post-translational modification. The coding region of mouse DESC1 contains three potential N-linked glycosylation sites (Fig. 1A). To examine whether the expressed DESC1 may be post-translationally modified by glycosylation, extracts of HeLa cells transfected with the empty expression plasmid and with the HA-tagged mouse DESC1 expression plasmid were treated with endoglycosidase H prior to the Western blot analysis. This enzyme removes N-linked carbohydrate moieties from glycoproteins leaving a single Nacetylglucosamine residue attached to the protein backbone. Endoglycosidase H treatment led to the complete disappearance of the 60-kDa immunoreactive protein (Fig. 5, lane 4) and the emergence of a novel 50-kDa species that was present in extracts from cells transfected with the HA-tagged mouse DESC1 but not in control extracts (compare Fig. 5, lanes 2 and  4). Taken together, this analysis demonstrates that mouse DESC1 encodes a 60-kDa glycoprotein and that one or more of the three potential N-linked glycosylation sites are utilized.
Mouse DESC1 Is a Transmembrane Protein with Type II Topology-Because the predicted protein sequence of mouse DESC1 indicated the presence of a putative transmembrane within the mouse chromosome 5E1 region and human chromosome 4q13.3 and determine their coding potentials. The position orientation of each gene in A is indicated with a box, and the gene orientation is indicated with arrows above each box. The spacing between each gene in the cluster is indicated with arrows. The designation of previously reported partial or complete cDNA sequences that can be unambiguously assigned to the 12 candidate genes is indicated above each gene. Identical amino acids in B are boxed and shaded gray. The position of predicted transmembrane signal anchors (solid underline), SEA domains (weak prediction in DESC3 and HAT-like 5) (striped underline), and serine protease domains (dotted underline) of each protease are indicated. Activation sequences are boxed. signal anchor, we investigated whether the mouse DESC1 protein was localized to the cell membrane and encoded a type II transmembrane protein. HeLa cells were co-transfected with the mouse DESC1-HA expression plasmid and a green fluorescent protein (GFP)-expressing plasmid to visualize the cytosolic compartment of the cell (Fig. 6, A-D), or with a GFP-expressing plasmid alone (Fig. 6, E-H). Nonpermeabilized transfected cells were then stained with anti-HA antibodies and AlexaFluor 546-conjugated secondary antibodies to detect the murine DESC1 protein and were analyzed by confocal fluorescence microscopy. HeLa cells cotransfected with the mouse DESC1-HA expression plasmid displayed an intense red fluorescent signal (Fig. 6C) that did not co-localize with the cytoplasm or the nucleus (Fig. 6D) and was confined exclusively to the plasma membrane. HeLa cells that were transfected with GFP alone did not demonstrate any membrane-localized red fluorescence (Fig. 6, G  and H), demonstrating the specificity of the antibody staining for the murine DESC1 protein. Taken together, the data showed that DESC1 localized to the plasma membrane with the C-terminal serine protease domain being freely accessible to antibodies, compatible with DESC1 being a type II transmembrane protein with a cytoplasmic N terminus and an extracellular C terminus. This conclusion was further supported by the fact that a mouse DESC1 protein that contained an N-terminal epitope tag also localized to the plasma membrane by confocal fluorescence analysis but required membrane permeabilization for visualization with anti-HA antibodies (data not shown).

Recombinant Mouse DESC1 Serine Protease Domain Is a Trypsin-activable Zymogen with Gelatinolytic and Caseinolytic
Activities-To determine whether the murine DESC1 gene encoded a functional serine protease, the DESC1 cDNA sequence encoding the serine protease domain including the predicted propeptide region (amino acid residues 196 -442) was inserted into the Sf9 insect cell expression plasmid pMIB/V5-His. The expression plasmid furnishes the protease domain with a honeybee mellitin signal peptide for efficient secretion and with C-terminal V5 and His 6 epitope tags to facilitate detection and purification. Conditioned medium from Sf9 cells stably expressing the DESC1 protease domain showed a 31-kDa immunoreactive protein by Western blot analysis using anti-V5 antibodies (Fig. 7B, lane 12). A corresponding 31-kDa caseinolytic and gelatinolytic activity was detected in the conditioned medium from transfected Sf9 cells (Fig. 7A, lanes 2 and 5) but not in the conditioned medium from the nontransfected control cells (Fig. 7A, lanes 1 and 4). The recombinant protein was purified from the conditioned medium using cobalt chelation chromatography (Fig. 7B, lanes 7-11 and 12-16), which generated a highly enriched preparation of soluble murine DESC1 (Fig. 7B, lanes 11 and 16). The purified DESC1 serine protease domain displayed prominent gelatinolytic and caseinolytic activities when analyzed by zymography (Fig. 7A, lanes 3 and 6), demonstrating that the mouse DESC1 serine protease domain has proteolytic activity.
Typical of the serine proteases, DESC1 is predicted to be synthesized as a single chain proenzyme with activation effected by proteolytic cleavage within a highly conserved activation motif (Arg 210 -Ile-Val-Gly-Gly), which is susceptible to cleavage by trypsin-like serine proteases. To determine whether the murine DESC1 serine protease domain was secreted into conditioned medium as an inactive proenzyme, an ␣ 2 -macroglobulin (␣ 2 -M) capture experiment was performed. Recombinant purified mouse DESC1 was unable to form a stable complex with an excess of ␣ 2 -M both prior to and after affinity purification (data not shown). This suggested that the murine DESC1 serine protease domain was synthesized as a zymogen but was able to undergo proteolytic activation under the conditions associated with the zymography procedure, as has been found with urokinase-type plasminogen activator (51) and matrix metalloproteinase-9 (52). The mouse DESC1 zymogen could be converted to the active enzyme form following transient exposure to immobilized trypsin, as evidenced by the small increase in mobility on SDS-polyacrylamide gels (Fig. 7C,  lane 2). Moreover, the active enzyme formed SDS-stable complexes with ␣ 2 -M (Fig. 7C, lane 3) that were abolished by preincubation with the generic active site serine protease inhibitor aminoethylbenzenesulfonyl fluoride (AEBSF) (data not shown).
Mouse DESC1 Preferentially Hydrolyzes Peptides with Arg in the P 1 Ј Position-The substrate specificity of serine proteases for their target substrates is determined, in part, by an optimal amino acid sequence surrounding the peptide bond cleavage site, designated P 1 -P 1 Ј (53). To provide insight into the substrate specificity of the recombinant murine DESC1, peptide assays with synthetic chromogenic peptides were performed using the recombinant DESC1 protease domain. As illustrated in Fig. 8, DESC1 showed a clear substrate preference for P 1 -Arg-containing peptides, with the peptide Suc-Ala-Ala-Pro-Arg-pNA showing the highest hydrolytic activity (227 nM pNA/ min). P 1 -Lys was less effective than P 1 -Arg, and no hydrolytic activity was detected when P 1 was substituted with Val (as preferred by leukocyte elastase) or with Phe (chymotrypsin preference). The hydrolytic activity of DESC1 against Suc-Ala-Ala-Pro-Arg-pNA was abolished by several generic inhibitors of serine proteases (AEBSF, aprotinin, and leupeptin) but was not affected by E64 or EDTA, inhibitors of cysteine-and metalloproteinases, respectively (Fig. 8B). Egg white trypsin inhibitor, although an effective inhibitor of trypsin activity, only partially inhibited DESC1 hydrolysis of the Suc-Ala-Ala-Pro- Arg-pNA peptide. This substrate specificity of the DESC1 protease domain is consistent with the structural features of this protease. The S1 specificity of serine proteases is largely determined by the amino acid residue positioned six amino acids N-terminal to the active site serine residue. The Asp 386 residue is at this position in DESC1 and points to a cleavage specificity for substrates with Arg or Lys at the P 1 position, although the Ala 387 predicts that DESC1 prefers a P 1 -Arg over Lys.

Mouse DESC1 Serine Protease Domain Forms Inhibitory
Complexes with the Serpins PAI-1 and PCI-The potential role of serpin-type serine protease inhibitors in regulating TTSP activity has not been explored previously. To study this, we incubated trypsin-activated purified DESC1 with purified PAI-1 and PCI, two serpins that are present in DESC1-expressing tissues, as well as with a range of purified recombinant serpins that are not present in these tissues (␣ 1 -plasmin  4) and DESC1-expressing Sf9 cells (lanes 2 and 5) was analyzed by gelatin and casein zymography as described under "Experimental Procedures." B, purification of recombinant mDESC1 serine protease domain by metal chelation affinity chromatography. The mouse DESC1 serine protease domain-containing conditioned medium (lanes 7 and 12) was applied to a cobalt chelation column, and bound protein was eluted with an imidazole gradient (lanes 8 -10 and 13-15) to produce a highly enriched (lanes 11 and 16) preparation of the mouse DESC1 serine protease domain, which eluted at ϳ100 mM imidazole. The purification procedure was monitored by SDS-PAGE and silver staining (lanes 7-11), or by Western blot analysis using an anti-V5 antibody (lanes [12][13][14][15][16]. The purified product displayed prominent gelatinolytic and caseinolytic activities when analyzed by zymography (A, lanes 3 and 6). The position of molecular mass markers (kDa) is indicated on the left. C, mouse DESC1 serine protease domain is a zymogen that can be activated by trypsin. The purified mouse DESC1 serine protease domain (lane 1) was activated using trypsin-immobilized beads, re-purified by chelation chromatography (lane 2), and then incubated with  inhibitor, ␣ 1 -antichymotrypsin, anti-thrombin III, or heparin co-factor II). Serpin-protease complex formation was analyzed by SDS-PAGE and Western blot using serpin antibodies and antibodies recognizing the V5 epitope-tagged DESC1 serine protease domain (Fig. 9). The mechanism of serpin-serine protease inhibition is well characterized and involves a branched pathway wherein the protease recognizes the reactive center loop of the serpin and, in particular, the reactive bond as a potential substrate (54). Following cleavage of the peptide bond at the P 1 residue of the reactive center loop, proteolysis may continue with subsequent release of the cleaved form of the serpin or the acyl intermediate may be trapped, resulting in a loop-insertion conformational change within the serpin and distortion of the protease active site. The latter pathway represents "suicide substrate inhibition" and is characterized by the formation of a lower mobility, SDS-resistant complex following analysis by SDS-PAGE. DESC1 was unable to cleave or form SDS-stable complexes with ␣ 1 -plasmin inhibitor, ␣ 1 -antichymotrypsin, anti-thrombin III, or heparin co-factor II (data not shown). Most interestingly, SDS stable complexes were efficiently formed between both DESC1 and PAI-1 and between DESC1 and PCI (Fig. 9A, lanes 2 and 6). The PAI-1-DESC1 complex was completely abolished as follows: 1) by mutation of the P 1 amino acid of PAI-1 creating a mutant defective in the serpin reactive site loop cleavage required for complex formation (PAI A ) (47) (Fig. 9A, lane 5); 2) by a PAI-1 mutant that is defective in loop insertion and the associated conformational changes required for complex formation (PAI-1 R ) (47) (Fig. 9A,  lane 4); or 3) by preincubation of the serine protease domain with AEBSF (data not shown). Similarly, the PCI-DESC1 complex was completely abolished by a P 14 mutant of PCI that is defective in loop insertion (55) (Fig. 9A, lane 8), by a mutation of the P 1 amino acid (49) (Fig. 9A, lane 9), or by preincubation of the serine protease domain with AEBSF (data not shown). Both serpin mutants defective in loop insertion, PAI-1 R and P 14 -PCI, were found to act as substrates for DESC1, as evidenced by the presence of a faster migrating band in reducing SDS-polyacrylamide gels (Fig. 9, B, lane 5, and C, lane 12), consistent with cleavage of the reactive center loop. A minor cleavage of wild type PCI by DESC1 was also observed (Fig. 9B,  lane 11).
Several serpins are activated as inhibitors of target proteases by binding to heparin or other linear negatively charged glycosaminoglycans. In the cases of PAI-1 and PCI, the glycosaminoglycan is thought to act as a bridge, simultaneously bringing the protease and inhibitor together in an appropriate orientation for productive interaction (54). Although the addition of heparin to the reactions of DESC1 with anti-thrombin III or heparin co-factor II did not lead to any inhibitory complexes or serpin cleavage (data not shown), heparin markedly enhanced inhibitory complex formation between DESC1 and PCI (Fig. 9A, lane 7). Heparin did not appear to affect inhibitory complex formation between PAI-1 and DESC1; however, formation of an inhibitory complex was complete in the absence of heparin under the same reaction conditions (Fig. 9A, lane 3).
DESC1, like other TTSPs, is a multidomain serine protease that contains an extracellular SEA domain in addition to the serine protease domain. To determine whether exosite interactions mediated by the SEA domain affect serpin inhibitory complex formation, we generated a soluble full-length version of pro-DESC1 furnished with a C-terminal V5 epitope tag by deleting the membrane anchor and cytoplasmic tail. The soluble full-length pro-DESC1 migrated as an ϳ52-kDa protein under both reducing and nonreducing conditions when analyzed by Western blot using anti-V5 antibodies (Fig. 9D, lanes  1 and 10, and data not shown). The soluble full-length pro-DESC1 could be activated by trypsin, as evidenced by the appearance of a prominent 32-kDa V5 epitope containing protein under reducing conditions (Fig. 9D, lanes 7 and 16). The activation by trypsin was incomplete, as shown by the presence of a residual 52-kDa full-length protein under reducing conditions (Fig. 9D, lanes 7 and 16). Trypsin exposure also resulted in the formation of a secondary cleavage product that contained the C-terminal V5 epitope, as shown by the appearance of an ϳ32-kDa protein under nonreducing conditions (Fig. 9D, lanes  4 and 13). Based on the apparent molecular weight of the cleavage product and the ability of this protein to form serpin inhibitory complexes (see below), the secondary cleavage appeared to take place in the linker region between the SEA domain and serine protease domain just N-terminal to the activation cleavage site to liberate the intact serine protease domain. Whether the cleavage of the linker region was mediated by trypsin or by activated DESC1 could not be determined. The rate of formation of this cleavage product varied from experiment to experiment (data not shown). However, the amount of V5 immunoreactive material in the 32-kDa region was always lower under nonreducing conditions as compared with reducing conditions (Fig. 9D, compare lanes 4 with 7 and 13 with 16), demonstrating that activated full-length DESC1 was one protein product generated by trypsin exposure of fulllength pro-DESC1. The various V5 epitope-containing DESC1 proteins formed after trypsin exposure are shown schematically in Fig. 9E.
Incubation of trypsin-treated full-length soluble DESC1 with either PAI-1 or PCI led to the formation of SDS-stable inhibitory complexes with apparent molecular masses of ϳ90 kDa compatible with the formation of full-length soluble DESC1serpin complexes (Fig. 9D, lanes 5 and 14). As predicted for a full-length DESC1-serpin inhibitory complex, the 90-kDa complexes were observed under nonreducing conditions but were absent under reducing conditions that break the disulfide bridge flanking the activation cleavage site thus releasing the SEA domain ( Fig. 9, D, lanes 8 and 17, and E). Furthermore, the 90-kDa DESC1-serpin inhibitory complexes were not formed in the absence of trypsin activation of DESC1 (Fig. 9D, lanes 2 and 11) and did not form with either the PAI-1 A or the P1-PCI mutant (Fig. 9D, lanes 6 and 12), showing that complex formation required the active DESC1 protease and reactive loop cleavage. In addition to these full-length DESC1-serpin complexes, lower molecular weight serpin inhibitory complexes (65-70 kDa) were formed with the DESC1 serine protease domain liberated by cleavage of the linker region (Fig. 9D,  lanes 5 and 14). Taken together, the data show that both PAI-1 and PCI form inhibitory complexes with soluble full-length DESC1 and that exosite interactions mediated by the SEA domain do not prevent the formation of DESC1-serpin inhibitory complexes.
In summary, this analysis shows that the murine DESC1 is synthesized as an inactive protease zymogen that can be proteolytically converted by trypsin to an active serine protease that displays both gelatinolytic and caseinolytic activities, and can be inhibited by both ␣ 2 -M and serpin-type serine protease inhibitors. The latter finding suggests that DESC1-type TTSPs and perhaps other types of TTSPs may be novel targets for inhibition by serpins. DISCUSSION The recent dramatic expansion in the number of known genes with the potential to encode pericellular serine proteases has necessitated a degree of rethinking of the specific reper-toire of proteolytic activities that may be engaged in tissue development, tissue homeostasis, and tissue repair, as well as in cancer and other tissue-destructive diseases. Particularly dramatic in this context has been the explosive increase in the number of identified genes that have the potential to encode TTSPs. This report further increases the number of functional TTSP genes, and with a few exceptions (see Introduction), the contribution of each of these genes to human health and disease remains essentially undefined.
The analysis presented here has assigned a number of recently reported predicted or cloned mouse transcripts with homology to human DESC1 as all being the product of seven potential TTSP genes located within a 0.5-Mb region on mouse chromosome 5E1. Most interestingly, the seven candidate genes are in close proximity (14 -104-kb spacing) and are all transcribed in the same orientation, suggesting a common origin and, possibly, a coordinated regulation of their expression. Multiple alignment of the predicted transcripts indicated that all seven mouse genes may indeed encode functional TTSPs. This conclusion is based on the presence of a predicted signal anchor and a predicted serine protease domain containing all the essential conserved catalytic residues and activation sequences. One of these seven genes is the mouse orthologue of the gene encoding the human airway trypsin protease (HAT), which has been demonstrated previously to be a functional enzyme (56). Human orthologues for four of these genes (MAT, mDESC3, HAT-like 4, and DESC1) were identifiable in a syntenic region of human chromosome 4q13.3. However, no human orthologues of HAT-like 2, HAT-like 3, and HAT-like 5 could be identified at this locus or at other chromosomal locations when screening both the Celera and NCBI genome data bases. Conversely, no orthologue of human DESC2 was apparent in the mouse genome. Taken together, this analysis suggests a considerable divergence in the repertoire of DESC1-like genes between humans and the mouse.
In addition to the bioinformatic analysis described above, the direct molecular and biochemical analysis presented in this paper provides strong evidence that the mouse DESC1 gene encodes a functional TTSP. The mRNA predicted to be encoded by the mouse DESC1 candidate gene would translate into a protein that contains both a conserved signal anchor motif and an intact serine protease domain with the chymotryptic (S1) fold. The presence of this predicted mRNA in the mouse was confirmed by direct cDNA cloning, and the DESC1 mRNA was demonstrated to be present in several mouse tissues. The expression of the mouse DESC1 cDNA in mammalian cells led to the production of a plasma membrane-associated recombinant glycoprotein with a type II topology, compatible with the presence of a functional signal anchor. The recombinant expression in insect cells of the portion of the mouse DESC1 cDNA that was predicted to encode the serine protease domain led to the generation of a protease zymogen that could be converted to a functionally active protease by exposure to trypsin. The zymogen demonstrated both gelatinolytic and caseinolytic activities under zymographic conditions. Like other serine proteases of the chymotrypsin (S1) fold (57), the protease domain of DESC1 is characterized by a triad of His, Asp, and Ser residues in the active site that are necessary for catalytic activity, as well as a binding pocket whose size, shape, and charge are determinants for the substrate specificity. Activated DESC1 efficiently hydrolyzed peptide substrates containing P 1 -Arg and formed specific inhibitory complexes with the serpins, PAI-1, and PCI, among a range of representative serpins from different clades ( Table I). The inability to form inhibitory complexes with ␣ 1 -proteinase inhibitor, ␣ 1 -antichymotrypsin. and heparin cofactor II is consistent with the P 1 -Arg subsite specificity. However, other determinants and/or exosites are also clearly important, as neither inhibitory complex formation nor cleavage was observed with anti-thrombin III or ␣ 2 -antiplasmin, which have the same P 1 -P 1 Ј as PCI and PAI-1, respectively (Table I). Together these data strongly suggest that the human DESC1 gene encodes a functional TTSP.
The expression analysis carried out in this paper documented the presence of mouse DESC1 mRNA in epidermal tissues, salivary glands, and the epididymis, with low levels of mRNA being present in other tissues. A previous survey (19) of human DESC1 expression also reported the presence of the mRNA in oral epidermal tissues and in the male reproductive tract, as well as the limited expression in other tissues. Although more detailed expression analysis awaits future studies, these findings do suggest that the human and mouse DESC1 orthologues are expressed at similar anatomical locations, and likely serve common functions in mice and in humans.
The role of serpin inhibition of TTSPs has not been explored prior to the study presented here. Most interestingly, in a screen of an array of serpins for activity toward DESC1, we found that both PCI and PAI-1 efficiently formed bona fide serpin-reactive loop insertion-dependent inhibitory complexes with the novel protease. This finding dramatically extends the potential range of known serpin inhibitory activities in physiology and disease. The TTSP matriptase was previously demonstrated to be inhibited by the Kunitz-type serine protease inhibitors aprotinin and hepatocyte growth factor activator inhibitor-1 (58,59). These prior data combined with the current study now suggest that TTSPs are likely to be the targets of both classes of serine protease inhibitors. Moreover, with respect to the delineation of selective DESC1 expression in epidermal, oral, and male reproductive tissues presented here, it is conspicuous that the novel protease efficiently formed inhibitory complexes with PCI and PAI-1, two serpins that are abundantly expressed in these tissues (60 -62), but did not interact with a range of serpins expressed at other anatomical sites. The co-localization of PAI-1 and PCI with DESC1 in the mouse and the complex formation of the two serpins with the purified protease at physiologically relevant serpin concentrations in vitro suggest that both PAI-1 and PCI could be functional inhibitors of DESC1 in vivo. Thus, the two serpins could be critical regulators of DESC1-dependent developmental or physiological processes in these tissues.
The expression of both the mouse and the human DESC1 in oral epidermal and male reproductive tissues and the co-expression of inhibitory serpins suggest that DESC1 has a function in the development or maintenance of these tissues. Moreover, the uniform absence of human DESC1 mRNA in malignant tumors derived from oral epithelial tissues has prompted the speculation that elimination of DESC1 expression may be causally related to the genesis of oral squamous cell carcinoma (19). Mouse genetics experiments have been initiated to delineate the specific function of DESC1 in both physiological processes and in cancer.