A Novel Protease Inhibitor of the α2-Macroglobulin Family Expressed in the Human Epidermis*

In the course of a large scale analysis of late-expressed genes in the human epidermis, we identified a new member of the α2-macroglobulin (α2M) protease inhibitor family, A2ML1 (for α2-macroglobulin-like 1). Like A2M and PZP, A2ML1 is located on chromosome 12p13.31. A2ML1 encodes a protein of 1454 amino acids, which fits the characteristics of α2Ms: 1) strong conservation in amino acid sequence including most of cysteine positions with α2M; 2) a putative central bait domain; 3) a typical thiol ester sequence. Northern blot and reverse transcriptase-PCR studies revealed a single 5-kb A2ML1 mRNA, mainly in the epidermis granular keratinocytes. A2ML1 is also transcribed in placenta, thymus, and testis. By Western blot analysis, α2ML1 is detected as a monomeric, ∼180-kDa protein in human epidermis. In vitro keratinocyte differentiation is associated with increased expression levels. By immunohistochemistry, α2ML1 was detected within keratinosomes in the granular layer of the epidermis, and as a secreted product in the extracellular space between the uppermost granular layer and the cornified layer. Recombinant α2ML1 displayed inhibitory activity toward chymotrypsin, papain, thermolysin, subtilisin A, and to a lesser extent, elastase but not trypsin. Incubation with chymotrypsin and the chymotrypsin-like kallikrein 7 protease indicated that α2ML1 binds covalently to these proteases, a feature shared with other members of the family. Therefore, α2ML1 is the first α2M family member detected in the epidermis. It may play an important role during desquamation by inhibiting extracellular proteases.

Regulation of proteolytic enzyme activity is essential for cell and tissue homeostasis. In epidermis, proteolysis of adhesive structures is a prerequisite for desquamation. Several members of the four protease classes with suggested roles in desquamation have been described in epidermis (reviewed in Ref. 1). A wide variety of protease inhibitors is also present in the intercellular spaces of the stratum corneum and participates in the regulation of desquamation-associated proteolysis. Disturbance of the protease-antiprotease balance may have dramatic consequences as demonstrated by the discovery of the serine protease inhibitor Kazal type 5 (SPINK5) also known as lymphoepithelial Kazal type-related inhibitor (LEKTI) as the defective gene in Netherton syndrome (2). The importance of regulated proteolysis in epidermis has also been reported in mouse models. Targeted epidermal overexpression of the serine protease kallikrein 7 (KLK7), 2 also known as stratum corneum chymotryptic enzyme (SCCE), results in a severe phenotype (3). Conversely, a null mutation in the mouse cystatin M/E gene (Cst6), encoding a cysteine protease inhibitor, induces neonatal lethality and abnormalities in cornification and desquamation, highlighting the essential role for protease inhibitors during the final stages of epidermal differentiation (4).
In the course of a large scale search for genes specifically expressed by the last transcriptionally active keratinocytes in the granular layer of the human epidermis, we identified a new member of the ␣-macroglobulin (␣M) superfamily, which we named A2ML1 (␣ 2 -macroglobulin-like 1).
The ␣M superfamily comprises both protease inhibitors and components of the complement. The ␣M protease inhibitor family, typified by the human tetrameric ␣ 2 -macroglobulin (␣2M), is a class of protease inhibitors with broad specificity (for review, see Refs. [5][6][7]. Because of their abundance in plasma (up to 10% of total serum proteins), they are considered as backup protease inhibitors although their precise function remains incompletely defined.
The ability of ␣Ms to inhibit all four protease classes resides in their unique mechanism of inhibition by steric hindrance called the "trap mechanism" (8). The ␣M subunits harbor a central bait domain sensitive to proteolytic cleavage (9). Cleavage of the bait region by a protease induces a major conformational change in the ␣M and as a consequence, the entrapment of the protease. Thereby, the access of possible substrates to the protease active site is hindered (6,10). Because of physical constraints, entrapped protease molecules are unable to hydrolyze large substrates but retain almost full activity against small ones (11,12). Concomitantly to the entrapment, the internal thiol ester bond, distinctive of the ␣M, becomes highly reactive and mediates covalent binding with the attacking protease via ⑀-lysyl-␥-glutamyl bonds (13,14). Moreover, the conformational change of the ␣M exposes their C-terminal domain, allowing binding to specific receptors such as the ␣2M receptor/low density lipoprotein receptor-related protein 1 (15) and hence clearance of the ␣M-protease complexes. In addition to proteases, ␣2M binds non-covalently to various distinct proteins, such as inflammatory cytokines, growth factors, ␤-amyloid peptide, or apolipoprotein E (16 -20). The biological importance of ␣2M in regulating the activity of cytokines and growth factors has been largely documented (16,21).
Tetrameric, dimeric, and monomeric ␣M protease inhibitors have been identified in a wide variety of organisms including both invertebrates and vertebrates (22)(23)(24). In human, two ␣M have been described, the tetrameric ␣ 2 -macroglobulin (␣2M), and the dimeric pregnancy zone protein (PZP), both being plasma protease inhibitors synthesized predominantly in the liver. Monomeric ␣M inhibitors have been described in rodents (named murinoglobulins) but so far no monomeric form has been described in humans.
Biological Materials-Protein extracts, total RNA, and cDNA minilibraries were prepared from a plastic surgery specimen of normal abdominal human skin. Paraffin-embedded sections were prepared from breast skin. Primary keratinocytes were from human foreskin. All human specimens were kindly provided by Professor J. P. Chavoin (Chirurgie plastique, Hôpital Rangueil, Toulouse) after informed consent of the patients and in accordance with Helsinki principles.
Preparation of a Granular Layer Keratinocyte-enriched Cell Population and EST Production-The procedure to recover total RNA from human epidermis fragments enriched with either basal (sample T1) or granular keratinocytes (sample T4) was described in detail elsewhere (26). Briefly, dermo-epidermal cleavage of the abdominal skin sample was performed after thermolysin incubation. Iterative trypsin incubations of the epidermis led to three samples of dissociated cells named T1, T2, and T3. The T4 sample corresponds to the epidermal fragments left over after the third incubation and is mainly composed of granular keratinocytes and corneocytes. From the latter sample, poly(A ϩ ) mRNA was used to generate mini-libraries of cDNA by the open reading frame EST method, essentially performed as described (26,27). EST were generated by library sequencing.
Analysis of A2ML1 mRNA Expression-Northern blotting of total RNA extracted from human epidermis after dermo-epidermal cleavage was performed with the digoxigenin technology, according to the manufacturer's protocol (Roche Applied Science). An A2ML1 probe was produced with the PCR digoxigenin synthesis kit using primers designed from A2ML1 on exons digoxigenin 18 and 19. After hybridization, detection was performed with an anti-digoxigenin alkaline phosphatase-conjugated antibody (digoxigenin Nucleic Acid Detection kit).
For reverse transcriptase-PCR experiments, a primer pair chosen on different exons to avoid amplification of potential contaminating DNA, generated amplicons of 206 nucleotides (exon 29 to 30 of A2ML1). The primer sequences were designed using Primer3 software (28) and Blast analysis (29) for the absence of similarity to any other human sequence. Human Multiple Tissue cDNA (MTC) panels I and II obtained from BD Biosciences were used as templates for PCR analysis. A control reaction with epidermis cDNA was carried out in parallel. The reactions were conducted for 35 cycles in standard conditions. Normalization of the samples was assessed after 22 cycles of amplification using glyceraldehyde-3-phosphate dehydrogenase primers provided in the kit.
For quantitative real-time reverse transcriptase-PCR experiments, reverse transcription was performed by standard procedures, starting from 100 ng of total RNA of T1 and T4 samples and using a mixture of oligo(dT) and random primers. A2ML1 expression was quantified using two pairs of specific primers amplifying exons 29 to 30 and exons 27 to 28. Assays were performed with the ABI prism 7000 Sequence Detection System and analyzed with the corresponding software (Applied Biosystems, Foster City, CA) using the Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen). The arbitrary defined number of PCR cycles in which each PCR amplification graph is in the linear range corresponds to the cycle threshold (C t ). The relative amount of A2ML1 to LGALS7 internal control and the -fold induction were calculated by using the equation 2 Ϫ⌬⌬Ct , where ⌬C t ϭ C t,A2ML1 Ϫ C t,LGALS7 and ⌬⌬C t ϭ ⌬C t,T4 sample Ϫ ⌬C t,T1 sample . Samples were analyzed in triplicate and differences between the values within triplicates were lower than 0.3 cycles. Amplicons for KRT14, encoding cytokeratin 14 and for KLK7 were analyzed in parallel as additional controls (data not shown).
Production of Rabbit Antiserum-A peptide was synthesized according to the predicted amino acid sequence of human ␣2ML1 in the N-terminal region ( 487 ISFYYLIGKGSLVM 501 ). Anti-peptide antibodies were produced by injecting the synthetic peptide conjugated via an added C-terminal cysteine residue to keyhole limpet hemocyanin. Antipeptide antibodies titer was determined by enzyme-linked immunosorbent assay (Millegen, Toulouse, France). The antiserum was affinitypurified using the peptide coupled to an agarose-activated affinity column (Sulfolink kit, Pierce).
Production of Recombinant ␣2ML1-The full-length coding sequence of AL832139 (DKFZp686O1010) was cloned into pCEP4 (Invitrogen) in C-terminal fusion with the V5-His epitope and was expressed in stable pools of 293/EBNA cells (Invitrogen). Transfection was performed with JetPEI reagent (QBiogen, Llkirch, France). Cells were grown in the presence of 150 g/ml hygromycin for 2 weeks. After cell washing, serum-free media was conditioned for 48 h, centrifuged at 2000 ϫ g for 5 min, and concentrated on Vivaspin centricons (Vivascience AG, Hannover, Germany). The ␣2ML1 protein was then purified by metal chelate affinity chromatography (Ni-NTA Spin Columns, Qiagen). In parallel, control medium, conditioned from mocked-transfected cells was submitted to identical purification procedures. The purified protein was monitored after SDS-PAGE by staining with Protogold (BB International, Cardiff, UK).
Epidermal Extracts, Keratinocyte Extracts, Immunoblotting Experiments, and Deglycosylation-Epidermal proteins were extracted in TENP-40 lysis buffer (40 mM Tris/HCl, pH 7.5, and 10 mM EDTA containing 0.5% Nonidet P-40 and protease inhibitors). Primary keratinocytes were established from human foreskin and grown either in Rheinwald and Green medium (30) or in KGM medium (Promocell). To induce differentiation, sub-confluent cells were either maintained 48 h in the medium of Green (30) or switched into KGM medium supplemented with Ca 2ϩ at a final concentration of 1.5 mM. Total proteins were extracted in TENP-40 lysis buffer. For Western blotting, Laemmli buffer with or without ␤-mercaptoethanol was added to the extracts. The blots were probed with the purified polyclonal antiserum or mouse monoclonal anti-V5 antibody. For deglycosylation experiments, N-glycosidase F, neuraminidase, O-glycosidase, or both neuraminidase and O-glycosidase were added to purified denatured recombinant ␣2ML1, for 3 h at 37°C in the conditions recommended by the manufacturer.
Immunohistochemistry and Immunoelectron Microscopy-Immunohistochemistry was performed on Bouin-fixed skin samples embedded in paraffin using the peroxidase-labeled streptavidin-biotin amplification method. Immunoelectron microscopy using ultrathin cryosections of normal non-palmoplantar human epidermis were performed as previously described using the purified polyclonal antiserum and gold particle-conjugated anti-rabbit IgG (31). Negative controls consisted in incubations in the presence of secondary antibody alone or with an unrelated primary antibody.
Hide Powder Azure Assay-Human ␣2M and affinity-purified recombinant ␣2ML1 were compared for their ability to prevent the various proteases from digesting the "hide powder azure" substrate, essentially as previously described (12). Concentration of ␣2ML1 was estimated by Protogold staining with respect to various quantities of human ␣2M (in the nanogram range) loaded on the same gel. ␣2ML1 molarity was calculated assuming its monomeric status. Purified recombinant ␣2ML1 protein (0.6 -15 pmol) was reacted with 2 pmol each of trypsin, chymotrypsin, elastase, and papain, 1.9 pmol of subtilisin A, and 3 pmol of thermolysin. At least two different concentrations of inhibitor with respect to each protease were tested. Enzyme solutions were freshly prepared and assumed to be 100% active by weight. In parallel, increasing amounts from 0.2 to 2 pmol of human ␣2M were reacted with each of the enzymes. Papain was activated by adding dithiothreitol (2 mM) and incubating at room temperature for 5 min. A final concentration of dithiothreitol of 0.02 mM was shown not to interfere with ␣2M inhibition. Reactants were preincubated for 5 min at room temperature in 0.2 ml of 50 mM Tris-HCl, pH 8, 100 mM NaCl, 0.1 mM EDTA (or 5 mM CaCl 2 instead of EDTA when thermolysin was used) before 0.8 ml of the same buffer containing 5.5 mg of hide powder azure substrate was added. Incubation was performed under agitation for 30 min at 37°C. The reactions were stopped by placing on ice. The high molecular weight substrate was pelleted by centrifugation at 4°C and proteolytic activity was determined by measuring the A 595 optical density of the supernatants. Each protease was tested with both ␣2M and ␣2ML1 in the same experiment. Reactions were performed in duplicates and repeated at least twice. The data were collected from two independent experiments and analyzed by using the StatSoft STATIS-TICA program, version 6. To compare ␣2ML1 inhibitory capacities to that of ␣2M, we determined R 50% as the molar ratio of each inhibitor with respect to protease, whereby the proteolytic activity of the protease is decreased to 50%.
Binding Assays-For zymography assay, 20 ng of chymotrypsin were incubated with 15 ng of purified ␣2ML1 or control medium for 10 min at room temperature in phosphate-buffered saline. ␣2ML1 or control medium were incubated in parallel without addition of chymotrypsin. Non-reducing Laemmli buffer was added to the samples and the reactants were separated by electrophoresis on an SDS-polyacrylamide gel containing 1% casein. The gel was renatured in 0.1 M Tris-HCl, pH 8, 2.5% Triton X-100 for 1 h, then incubated in 0.1 M Tris-HCl, pH 8, 0.2% Triton X-100 overnight at 37°C, and finally stained with Coomassie Blue (Bio-Safe Coomassie, Bio-Rad). Casein digestion was evidenced as white, unstained bands. Covalent binding to KLK7 was checked using affinity-purified recombinant ␣2ML1 (50 ng or 0.3 pmol) incubated with either active KLK7 or the inactive pro-KLK7 (10 ng each or 0.3 pmol) for 10 min at 22°C or 5 min at 37°C in 40 mM Tris-HCl, pH 7.4, 150 mM NaCl. The reactions were stopped by adding 0.1 mM phenylmethylsulfonyl fluoride and Laemmli buffer. Reactants incubated on ice were used as controls. Control medium was used as an additional control. The reactants were separated by SDS-PAGE and transferred onto nitrocellulose membrane. The membrane was probed with a polyclonal antibody recognizing both pro-KLK7 and KLK7 (32).    MARCH 3, 2006 • VOLUME 281 • NUMBER 9

RESULTS
Characterization of A2ML1-To identify epidermis late-expressed genes, successive rounds of proteolysis of normal human skin were performed to isolate four batches of cells that roughly represent the successive steps of keratinocyte differentiation. The last batch, mostly containing granular keratinocytes and corneocytes, was used to produce EST mini-libraries by the open reading frame EST method (27). One EST showed complete identity with a full-length cDNA (GenBank TM accession number AL832139) encoding a hypothetical protein displaying strong conservation with ␣2M. Thus, in agreement with the HUGO gene nomenclature committee, we named this new gene A2ML1 (for ␣ 2 macroglobulin-like 1). The alias name of A2ML1 is CPAMD9 (for C3 and PZP-like ␣ 2 -macroglobulin containing domain 9). According to the May 2004 human genome assembly (33), the A2ML1 gene is located on chromosome 12p13.31, telomeric to the two other human genes of the ␣M family, A2M and PZP (Fig. 1). The locus of A2ML1, A2M, and PZP encompasses a 385-kb region. The predicted A2ML1 exon-intron organization (36 exons spanning 54 kb) is similar to that of A2M (36 exons spanning 48 kb) and PZP (36 exons spanning 59.5 kb). A notable difference is the opposite orientation of A2ML1. Twelve mRNA matched the A2ML1 gene (Fig. 1, box), among which AK057908, referred to as the predicted gene FLJ25179, spans only the 3Ј part of the A2ML1 transcript. The corresponding hypothetical protein of 158 amino acids (NP653271) was recently classified as a member of the ␣2M family (34). To note, AK122624 mRNA lacks the 3Ј part of exon 20, exons 21-26, and the 5Ј part of exon 27. The gene has two alternative, non-coding last exons.
Searching for mRNA and EST originating from putative orthologs of A2ML1 resulted in the identification of three mRNA from chimpanzee, dog, and pig that appeared closely related to ␣2ML1 by using the Multalin algorithm (35) (Fig. 2). No mRNA or EST from rat or mouse was identified as transcribed from a potential A2ML1 ortholog. As illustrated by the phylogenetic tree, even if ␣2ML1 and the cluster ␣2M/PZP arose from a common ancestor, they diverged long ago. The cluster of the monomeric murinoglobulins, which evolved after divergence of the rodents and therefore are not present in the human genome (34), were also distinct from ␣2ML1.
The A2ML1 mRNA encodes a putative protein of 1454 amino acids that displays a similar organization to that of the prototyped ␣2M (Fig.  3). The ␣2ML1 protein displays a putative signal peptide of 17 amino acids, a divergent bait domain near the center of the molecule (residues 695-726), and a typical thiol ester sequence ( 969 Gly-Cys-Gly-Glu-Gln 973 with the putative thiol ester bond formed by Cys 970 and Gln 973 ). Of the 10 potential N-glycosylated residues, eight showed striking con-servation of position with ␣2M. The disulfide bridge pattern of ␣2M as determined by Jensen and Sottrup-Jensen (36) contains 25 cysteinyl residues, 23 of which were found in conserved positions in ␣2ML1, as illustrated in Fig. 4. However, the two cysteinyl residues involved in the formation of interchain disulfide bridges for ␣2M were missing in ␣2ML1. The conservation scores between ␣2ML1 and the main prototypes of the ␣2M family assessed by the Clustal W algorithm (37) are presented in Table 1. The best overall amino acid sequence identity was found with human ␣2M (40% of identity), a score higher than, albeit very similar, to those obtained with human PZP, mouse ␣2M, or rat ␣1I3. As expected, the identity score was greatly reduced when the bait domains were compared, in agreement with the typical divergence of these regions.
Analysis of the Expression of A2ML1 Transcripts in Epidermis and Other Human Tissues-The expression of A2ML1 was analyzed by Northern blot performed on total RNA extracted from normal human epidermis. A single 5-kb band was detected, consistent with the size expected from the AL832139 cDNA sequence (Fig. 5A). To further investigate the transcription of A2ML1 during epidermal differentiation, we quantitatively analyzed the expression of A2ML1 in epidermal samples representative of the basal (T1) and granular (T4) layers. The cell separation method was validated by a previous work (26) through the analysis of two genes highly specific to the basal and granular layer, KRT14 and KLK7, respectively. Expression levels were calibrated using LGALS7, a gene previously shown by in situ hybridization to be expressed in all epidermal layers (38). A2ML1 displayed a 10-fold T4/T1 expression ratio (Fig. 5B), and is thus clearly an epidermis late-expressed gene.
Human cDNA were used as PCR templates to study the expression of A2ML1 in adult tissues. A representative amplification after 35 cycles is shown in Fig. 5C. A2ML1 was detected in epidermis, placenta, testis, and thymus, but not in epithelia of kidney, lung, small intestine, or colon. The expression of the A2ML1 transcript variant AK122624 was checked on both epidermis and the MTC panels. Using three different specific pairs of primers, we failed to detect any mRNA corresponding to AK122624 (originally cloned from tongue tumor tissue) in any of the tested tissues (data not shown).
␣2ML1 Is Secreted as a Monomer and Is Induced in Keratinocytes Differentiated in Vitro-We transfected 293/EBNA cells with a fulllength, V5/His-tagged A2ML1 cDNA and analyzed the conditioned medium by Western blot using an anti-V5 monoclonal antibody. A band with 180 kDa apparent molecular mass was detected, most probably corresponding to the full-length protein (Fig. 6A, left panel). This result is consistent with the theoretical molecular mass of 160 kDa tak- ing into account possible glycosylation of ␣2ML1. A minor band of ϳ55 kDa corresponding to a COOH-terminal fragment of ␣2ML1 (containing the V5 tag) was also detected (denoted by an arrow) and is discussed below. To investigate the presence of the ␣2ML1 protein in epidermis, we immunized a rabbit with an ␣2ML1 peptide corresponding to the NH 2 -terminal domain of ␣2ML1 (amino acids 487-501). Western blot analysis of total protein extracts from human epidermis revealed again a single band of ϳ180 kDa, demonstrating the presence of ␣2ML1 (Fig.  6A, right panel). An additional band of ϳ116 kDa (denoted by the arrow) was inconsistently detected. We next analyzed the expression of ␣2ML1 in human primary keratinocytes grown on a feeder layer of irradiated 3T3 fibroblasts according to the method of Rheinwald and Green (30). In this system, keratinocyte differentiation is initiated by confluence. As expected, increased amounts of involucrin, an early marker of terminal differentiation, were observed in the post-confluent keratinocytes (Fig. 6B, bottom panel). Similarly, increased amounts of ␣2ML1 were also observed (Fig. 6B, left panel). Likewise, the expression of ␣2ML1 was studied during keratinocyte differentiation induced by Ca 2ϩ concentration switch in a serum-free culture system (39). Increased amounts of ␣2ML1 were observed 48 h after Ca 2ϩ addition (Fig. 6B, right panel). In contrast, the protein was not detected in cervix carcinoma HeLa cells.
As we noticed, the two cysteinyl residues involved in ␣2M dimer formation (36) are not conserved in the ␣2ML1 sequence. This suggested that ␣2ML1 is monomeric, as the lack of one or two of the corresponding residues has been reported in other monomeric ␣M (40,41). However, to confirm the absence of dimer formation by disulfide bridge, recombinant ␣2ML1 and endogenous ␣2ML1 from keratinocytes were analyzed by SDS-PAGE in non-reducing conditions. The ϳ180-kDa form was still observed, whereas no 360-kDa dimers could be detected (Fig. 6C), demonstrating the absence of dimer formation by disulfide cross-linking. The members of the ␣ 2 -macroglobulin family, such as the rat ␣1I3 and the murinoglobulins lack one or two of the cysteinyl residues involved in dimer formation and are expressed as monomers (42,43). This is most probably the case for ␣2ML1, even if noncovalent interactions between subunits have not been definitively excluded.
To analyze its glycosylation state, the recombinant ␣2ML1 protein produced by transfected 293/EBNA cells was treated with N-glycosidase F, neuraminidase, and O-glycosidase. Treatment with N-glycosidase F caused a significant mobility shift of the full-length 180-kDa form to ϳ160 kDa (denoted by the arrowhead in Fig. 6D). Neither neuraminidase, O-glycosidase, nor a combination of both enzymes affected the mobility of the protein (data not shown). Thus, in agreement with the FIGURE 5. Detection of A2ML1 transcripts in adult human tissues. A, detection of A2ML1 transcripts in human epidermis by Northern blot. Ten g of total RNA from normal human epidermis was analyzed using a central A2ML1 probe. A single transcript of 5 kb was detected. B, detection of A2ML1 transcripts by quantitative real-time reverse transcriptase-PCR analysis in epidermal cell samples T1 and T4. Two sets of primers were used to generate amplicons, noted as 1 and 2. The relative amount of A2ML1 to LGALS7 internal control was calculated as ⌬C t . The difference of A2ML1 amount between T4 and T1 was calculated as ⌬⌬C t . The T4/T1 -fold induction of A2ML1 was calculated by using the Equation 2 Ϫ⌬⌬Ct . C, detection of A2ML1 transcripts by reverse transcriptase-PCR in human tissues. Epidermal cDNA were produced as described under "Experimental Procedures." Other cDNAs were from MTC panels. Besides epidermis, A2ML1 transcripts were detected in testis, placenta, and thymus. PBL, peripheral blood leukocytes. Bottom panel, glyceraldehyde-3-phosphate dehydrogenase (G3PDH) control amplification leading to a 983-bp fragment.  MARCH 3, 2006 • VOLUME 281 • NUMBER 9 presence of 10 potential N-glycosylated residues, ␣2ML1 is an N-glycosylated protein.

Characterization of the Human A2ML1 Gene
When ␣2M is heated in the presence of denaturant the internal ␤-cysteinyl-␥-glutamyl thiol ester bonds react with the polypeptide backbone causing scission into two proteolytic fragments of 120 and 60 kDa detected by SDS-PAGE only under reducing conditions (6). Although this point was not specifically addressed here, the ϳ55-kDa C-terminal fragment detected by the anti-V5 antibody on reducing SDS-PAGE and the ϳ116-kDa band detected by the polyclonal antibody (Fig. 6, A and D, arrows) are of the expected size for being the products of the heat-induced hydrolysis of the internal ␤-cysteinyl-␥glutamyl thiol ester bond of ␣2ML1. The ϳ55-kDa C-terminal fragment observed in reducing conditions was not detected in nonreducing conditions (Fig. 6C, lane 2), suggesting that disulfide bonds connect the two fragments generated upon heating as it is the case for ␣2M.
Expression Analysis of ␣2ML1 in Vivo-To localize ␣2ML1 in the human epidermis, sections of Bouin-fixed skin were analyzed by immunohistochemistry. The labeling signal was predominantly observed within the granular layer at the apical edge of the keratinocytes (Fig. 7A). On parallel control sections incubated in the absence of primary antibody, no significant reactivity was observed (Fig. 7B). The subcellular location of ␣2ML1 was determined by cryoimmunoelectron microscopy analysis of normal non-palmoplantar human epidermis. ␣2ML1 localized in small cytoplasmic vesicles containing lamellar structures, corresponding to keratinosomes (Fig. 8, A and B). ␣2ML1 labeling was also detected in the extracellular space between the uppermost granular   cell and the lowermost corneocyte (Fig. 8, C and D), confirming that the protein is secreted in vivo.
␣2ML1 Inhibits Various Proteases and Associates with Chymotrypsin and KLK7 in Vitro-The inhibition of several model proteases by ␣2ML1 was demonstrated using the high molecular weight substrate hide powder azure. The serine proteases chymotrypsin, subtilisin A, pancreatic elastase, the metalloprotease thermolysin, and the cysteine protease papain were inhibited by ␣2ML1, although its protease inhibitory capacity was found to be lower than that of prototype ␣2M ( Table  2). As an example, the activity of 1 pmol of chymotrypsin toward the high M r substrate was reduced by 50% in the presence of 0.77 pmol of ␣2ML1 (i.e. R 50% ϭ 0.77) as compared with 0.24 pmol of ␣2M (i.e. R 50% ϭ 0.24). In contrast, trypsin was not inhibited by ␣2ML1 even with a molar ratio of inhibitor/protease as high as 7.5:1. Altogether, the results showed that ␣2ML1 appears as a less efficient protease inhibitor with a more restricted spectrum of inhibition than ␣2M. Similar findings were previously reported with the rat ␣1I3, another monomeric member of the family. It was proposed that monomeric ␣2M, because of their smaller size, shield less access of proteins to the bound sterically hindered protease than tetrameric ␣2M. Consequently, more monomeric inhibitor is needed for complete inhibition of the protease (12).
Monomeric and multimeric ␣M have the ability to form covalent complexes with attacking proteases. To investigate the formation of similar complexes by ␣2ML1, chymotrypsin was incubated with affinity purified ␣2ML1 and the active protein species were analyzed by zymography with casein as the substrate (Fig. 9A). A broad band with higher molecular weight appeared specifically in the presence of ␣2ML1, revealing a covalent binding of chymotrypsin with a fragment of ␣2ML1. Moreover, this result shows that bound chymotrypsin is still able to degrade casein. This was expected according to the low molecular weight of the casein substrate. The observed size of the complex (75 kDa) suggests that a ϳ50-kDa fragment of ␣2ML1 was bound to chymotrypsin. If bait region cleavage was the only step leading to covalent binding of the protease, we would have expected the formation of a 110-kDa fragment consisting in the ϳ80-kDa COOH moiety of ␣2ML1 linked to chymotrypsin. Additional ectopic proteolysis by unbound proteases may thus account for the observed size of the ␣2ML1 fragment bound to chymotrypsin.
KLK7 is a chymotrypsin-like serine protease that plays a pivotal role in desquamation (25). Because both ␣2ML1 and KLK7 are secreted by granular keratinocytes (44), we hypothesized that KLK7 was a target for ␣2ML1. To address this question, either KLK7 or pro-KLK7 were incubated with affinity-purified ␣2ML1, and analyzed for the formation of a complex by Western blot using an antibody recognizing both forms of KLK7. Incubation of KLK7 but not Pro-KLK7 with ␣2ML1 leads to the presence of an additional, slower migrating SDS-PAGE-resistant band (Fig. 9B), reminiscent of that observed in zymography assays. This band was not observed at 0°C or in the absence of ␣2ML1 using control medium (not shown). Similarly to the results obtained with chymotrypsin, the observed size of the complex (75 kDa) suggests it is not com-posed of KLK7 bound to the entire ␣2ML1 molecule, but rather to a ϳ50-kDa fragment of ␣2ML1. The additional ectopic proteolysis observed with both chymotrypsin and KLK7 are likely to occur at the same sites.

DISCUSSION
Our approach, aiming to characterize the transcriptome of the most differentiated epidermal keratinocytes, led to the description of a novel gene, A2ML1, and its protein expression in the epidermis. Indeed, FIGURE 9. Formation of a complex between ␣2ML1 and chymotrypsin or KLK7. A, analysis of the association between affinity-purified ␣2ML1 and chymotrypsin. Chymotrypsin was incubated alone (lane 1), with control medium (lane 2), or with purified ␣2ML1 (lane 3) for 10 min at room temperature in phosphate-buffered saline. Purified ␣2ML1 (lane 4) or control medium (lane 5) were incubated alone. The proteins were then subjected to casein zymography. After staining, the activity was visualized as white bands. Free chymotrypsin was detected as a band of 25 kDa. Two additional bands (ϳ42-46 kDa) probably reflect the activity of chymotrypsin contaminants. A slight band was observed in control lanes 4 and 5, probably resulting from the activity of a contaminant present in the culture medium. In lane 3, the arrow indicates an additional, active band resulting from the interaction of chymotrypsin with ␣2ML1. B, analysis of the association between ␣2ML1 and KLK7. Affinity-purified ␣2ML1 was incubated with the recombinant active form of KLK7 (lanes 1, 3, and 5) or the inactive pro-KLK7 form (lanes 2, 4, and 6) for 10 min at 0°C (lanes 1 and 2), for 10 min at 22°C (lanes 3 and 4), or for 5 min at 37°C (lanes 5 and 6). The proteins were then analyzed by Western blot with a polyclonal anti-KLK7 antibody. In each lane, the doublet corresponds to glycosylated and unglycosylated forms of pro-KLK7 or KLK7, as described previously (25). Incubations of KLK7 but not pro-KLK7 with ␣2ML1 result in an additional, slower-migrating band (arrow) indicating the formation of a covalent complex between KLK7 and ␣2ML1. Proteases were titrated in a hide powder azure assay with recombinant ␣2ML1 or human ␣2M. The R 50% values (in bold) were determined as the molar ratio of the inhibitor with respect to protease whereby the proteolytic activity of the protease is decreased to 50%. Data plots were analyzed by using the StatSoft STATISTICA program, version 6. In parentheses are given the 95% confidence interval values. a ND, not determined. 50% of the proteolytic activity was not reached in the assay. The proteolytic activity was decreased to 70% for a molar ratio of 2. b No inhibition was observed for a molar ratio of 7.5.

Inhibitor
A2ML1 encodes a new member of the ␣2M family mainly differing from A2M by its restricted expression pattern and its monomeric conformation. A2ML1 expression in the epidermis is restricted to the granular layer that corresponds to the last keratinocyte differentiation step still exhibiting transcriptional and translational activity. ␣2ML1 thus constitutes a new late marker of epidermal differentiation. As expected for epidermis late-expressed genes, A2ML1 expression is up-regulated during the differentiation of normal human keratinocytes induced by Ca 2ϩ or by confluence.
We detected A2ML1 transcripts in epidermis, placenta, testis, and thymus. Among the 12 A2ML1 mRNA sequences from GenBank, five arise from tongue tumor, three from cervix carcinomas, two from brain, and one each from normal tongue and testis. Among the 38 EST reported in the UniGene cluster Hs.334306, 17 arise from tissues expressing a squamous epithelia-like differentiation program: nine from tongue, seven from hypopharynx, and one from skin squamous cell carcinoma. Nine additional EST were from colon carcinoma, and the rest were cloned only once or twice from a variety of other tissues. Thus, the origin of mRNA and EST from GenBank confirms the restricted transcription pattern of A2ML1, mainly in normal and tumorous stratified epithelia.
The 180-kDa ␣2ML1 protein was detected by Western blot and/or immunohistochemistry in keratinocytes and epidermis. However, it was not detected in commercially available extracts from placenta or testis (data not shown). We assume that in these tissues the ␣2ML1 protein, if any, is present in scant amounts even though transcripts were evidenced by PCR. Thus, ␣2ML1 exhibits expression features similar to those of known markers of the terminally differentiating keratinocyte, such as corneodesmosin (45) and SPINK5 (46). Dimeric and tetrameric ␣Ms are known to be more efficient protease inhibitors than monomeric ␣Ms. Indeed, the inhibitor/trypsin stoichiometry needed for full inhibition is 1:2 for the tetrameric ␣2M, but is as high as 10:1 for the rat monomeric prototype ␣1I3 (12). For all tested proteases, higher inhibitor/protease molar ratios were necessary to reach 50% inhibition of the proteases using ␣2ML1 as compared with ␣2M, although the exact binding ratio of the reactions remains to be quantified. However, our data clearly establish that ␣2ML1 preferentially inhibits chymotrypsin rather than trypsin, suggesting that chymotrypsin-like proteases may be physiological targets of ␣2ML1 in vivo. Moreover, ␣2ML1 also clearly inhibits papain and subtilisin. Subtilisins are chymotrypsin-like serine proteases expressed by various Bacillus species and are known to contribute to host cell invasion. The activity of ␣2ML1 may be involved in the mechanism of defense against invading pathogens, a role earlier suggested for ␣Ms (10).
Along with the trypsin-like kallikrein 5 (KLK5) also known as stratum corneum tryptic enzyme (47), KLK7 is a pivotal protease synthesized by the terminally differentiating keratinocytes (25). KLK7 is thought to be involved in the proteolysis of the desmosomal proteins corneodesmosin and desmoglein (32). We report that ␣2ML1 can form specific complexes with KLK7 in vitro, suggesting covalent binding via reaction of the internal thiol ester of ␣2ML1 with KLK7 lysine residues. This hypothesis is supported by previous reports demonstrating that, unlike tetrameric ␣Ms, monomeric ␣Ms absolutely require covalent binding to target proteases (12). In vivo, both ␣2ML1 (this study) and KLK7 (44) localize in the keratinocyte secretory vesicles (keratinosomes) before secretion into the extracellular space. Overall, these findings suggest that ␣2ML1 physically interacts with KLK7. A possible consequence could be the inhibition of proteolysis activity with respect to large substrates because of steric hindrance, whereas accessibility to small sub-strates would still be effective. Alternatively, binding to ␣2ML1 could protect KLK7 either from hydrolysis by other proteases or from inhibitors such as the serine protease inhibitor SPINK5, which plays a pivotal role in controlling KLK5-and KLK7-like activities in the upper epidermis (48,49). Beside KLK7, cathepsin L2, also called stratum corneum thiol protease and cathepsin L-like are papain-like cysteine proteases that are thought to be involved in desquamation (50). The high efficient inhibition of papain by ␣2ML1 suggests that cathepsin L2 and cathepsin L-like may also be physiological targets of ␣2ML1 in vivo. Binding to KLK7 and preferential inhibition toward chymotrypsin-like and papainlike proteases support the hypothesis that ␣2ML1 is a key regulator of desquamation.
Similarly to ␣2M, ␣2ML1 is also expected to bind growth factors such as TGF-␤1 (19). The growth factor-binding site in ␣2M for TGF-␤1, platelet-derived growth factor-BB, and nerve growth factor-␤ is located within a short peptide adjacent to the bait domain (51). A peptide, derived from human ␣2M sequence, has been shown to directly bind to TGF-␤1 and block its interaction with TGF-␤1 type I and II receptors (52). This peptide sequence (ETWIWDLVVVN), rich in hydrophobic and acidic residues, is quite conserved in ␣2ML1 (ETWLWDLFPIG). TGF-␤1 is synthesized by the upper differentiated layers (53) and is a potent inhibitor of keratinocyte proliferation (54). Therefore, we speculate that ␣2ML1 binding to such growth factors might play a role in epidermis homeostasis. Moreover, binding of ␣2M to proteases, growth factors, or cytokines can induce clearance and catabolism of the entire complexes via binding to the low density lipoprotein receptor-related protein 1 receptor (16), although its expression in the epidermis is still controversial (55,56).
Overall, our findings reveal the presence of a new member of an important class of protease inhibitors at the stratum granulosum-stratum corneum interface. Other members of this class are known to regulate not only protease activities but also the catabolism of proteases, growth factors, and cytokines. ␣2ML1 might thus play a central and regulatory role not only in the desquamation process but also in the homeostasis of the human epidermis.