Differential expression of plasminogen activators and their inhibitors in an organotypic skin coculture system

Plasminogen activators (PAs) are serine proteinases that catalyze the cleavage of plasminogen into the active enzyme plasmin, thus triggering a proteolytic cascade. There are two types of plasminogen activators, the tissuetype (tPA) and the urokinase-type (uPA). Many data indicate that tPA is an important regulator of fibrinolysis and that uPA mediates degradation of extracellular matrix in the pericellular environment (for reviews see Danø et al., 1985; Pollanen et al., 1991; Vassalli et al., 1991). Two major types of plasminogen activator inhibitors [PA inhibitor type 1 (PAI-1) and type 2 (PAI-2)] have been identified and characterized (Loskutoff, 1983; Kruithof, 1986; Wun and Reich, 1987). Although their biochemical properties are distinct, each has the ability to inhibit both uPA and tPA; however, the latter enzyme is inhibited less rapidly by PAI-2 (for review see Andreasen et al., 1990). There is good evidence that PAI-1 plays an important role in the regulation of fibrinolysis through its inhibition of tPA (Wagner et al., 1989; Schneiderman et al., 1992; Fay et al., 1992). A role for PAI-2 has not been defined, but hypotheses include regulation of PA during inflammation and pregnancy (Astedt et al., 1985; Schleef et al., 1988).

Characterizations of the PA cascade in skin have revealed that, under appropriate circumstances, epidermis can produce both of the enzymes and the inhibitors described above (for review see Lyons-Giordano et al., 1993); however, the cutaneous functions of this proteolytic cascade remain unknown.

The primary cell type in epidermis is the keratinocyte. Keratinocytes undergo a specialized process of cell differentiation, leading to the production of the stratum corneum, which serves as a selective barrier between the organism and its environment. Basal cells, which are the least differentiated keratinocytes, reside at the dermal-epidermal junction and constitute the proliferative population. Through mechanisms that are presently unclear, basal cells at some point become post-mitotic, assume a suprabasal position, and begin to sythesize specific differentiation-related products. Continued keratinocyte maturation is accompanied by further vertical migration, terminating in the formation of highly cross-linked squames (stratum corneum), which are shed from the skin surface.

In normal human skin, uPA enzymatic activity can be detected in the basal layer (Lyons-Giordano et al., 1993). In newly re-epithelialized cutaneous wounds, uPA and PAI-1 are present within and beneath the migrating basal cells (Grøndahl-Hansen et al., 1988; Rømer et al., 1991). Tissue PA antigen is also observed in the epidermal outgrowths of murine wounds (Grøndahl-Hansen et al., 1988), but in contrast to uPA and PAI-1, tPA is found suprabasally. In addition, tPA antigen is detected focally in the suprabasal layers in a variety of cutaneous disorders, including psoriasis and blistering diseases (Grøndahl-Hansen et al., 1987; Jensen et al., 1988). In detergent extracts of normal epidermis, small amounts of tPA activity are detected; however, assignment of tPA to particular epidermal layers has not been possible in normal skin. PAI-2 antigen is present throughout normal epidermis, but is concentrated in the suprabasal layers (Hibino et al., 1988; Lyons-Giordano et al., unpublished observations). The PA substrate plasminogen is present in the basal layer of normal epidermis (Isseroff and Rifkin, 1983), but the site of synthesis of cutaneous plasminogen is unclear.

Considered together, the above findings suggest distinct roles and regulatory mechanisms for the PA cascade components found in epidermis. Elucidation of these issues is difficult in intact skin, with its complex epithelial-mesenchymal interactions. Conversely, routine cell culture techniques generally fail to reproduce faithfully the in vivo epithelial architecture and differentiation sequela. Therefore, we have employed an organotypic coculture system (“skin equivalent”; Bell et al., 1983) to investigate further the roles of the epithelial PA cascade.

The skin equivalent is a three dimensional organotypic coculture system consisting of two types of cells: ker-atinocytes and fibroblasts. The fibroblasts reside in a contracted collagen matrix which forms the dermal portion. Keratinocytes are seeded onto the collagen matrix, where they form an epidermal-like structure that is maintained at the air-liquid interface. Such a system yields excellent epidermal type morphology and differentiation (Asselineau et al., 1986, 1989; Fuchs, 1990).

The aim of our present study is to assess the expression of PA components in the skin equivalent. Using immunohistochemistry and in situ hybridization, we have demonstrated the localizations and sites of synthesis of uPA and tPA as well as PAI-1 and PAI-2 in the skin equivalent. Our data indicate that individual PA components have specific expression patterns related to the state of differentiation of the keratinocyte.

Culture medium and reagents were purchased from the following sources: MCDB 153 base medium powder, DMEM medium powder (for concentrated medium), insulin, hydrocortisone, ethanolamine, phosphoethanolamine, adenine, tri-iodothyronine, transferrin, and Hepes (Sigma Chemical Co., St. Louis, MO); bovine pituitary glands (Pelfreeze Biologicals, Rogers, AK); DMEM with high glucose and L-glutamine, Ham’s F-12 (Hazelton, Lenexa, KS); sodium pyruvate, penicillin, and streptomycin (GIBCO, Grand Island, NY); epidermal growth factor, rat tail collagen (Collaborative Research, Bedford, MA); fetal bovine serum (Gemini Bio-Products, Calabasas, CA); bacterial and tissue culture 60-mm Petri dishes (Fisher Scientific, Pittsburgh, PA and Corning, Corning, NY, respectively).

For histological preparations, JB-4 embedding kit, paraformaldehyde and tissue freezing medium (PolyFreeze) were purchased from Polysciences, (Warrington, PA); Superfrost Plus slides, hematoxylin Gill’s number 1, and Permount were from Fisher Scientific.

Rabbit polyclonal antibodies used in this study were the following: affinity purified anti-human-uPA IgG (Morioka et al., 1985); anti-human-tPA, a gift from Dr Paul Horan from SmithK-line Beecham, King of Prussia, PA (Jensen et al., 1988); anti-human-PAI-1, a gift of Dr Gary L. Davis from Dupont-Merck, Wilmington, DE; anti-human-PAI-2, a gift from Dr Tze-Chein Wun from Monsanto, St. Louis, MO; anti-human involucrin (Biomedical Technologies, Stoughton, MA). Purified IgG from normal rabbit serum was used as a control for the above antibodies. The ImmunoPure Kit from Pierce was used for IgG purification. Murine monoclonal antibodies were the following: anti-human-uPA 2B2 (Jensen and Wheelock, 1992); anti-human-tPA (clone 3) from Monozyme, Virum, Denmark (Miller et al., 1992); antihuman PAI-1 (American Diagnostica no. 380); anti-human-PAI-2 (American Diagnostica no. 3750); anti-human keratin K1/K10 (AE-20, a gift from Dr T.-T. Sun, New York University, NY); and anti-hepatitis IgG as a irrelevant control (a gift from Dr Francee Boches, Baxter, Miami, FL). Biotinylated secondary antibodies and avidin-biotin-peroxidase were from Vector Laboratories (Burlingame, CA). Diaminobenzidine (DAB) substrate was from Sigma Chemical Co.

For in situ hybridization, proteinase K was purchased from Boehringer Mannheim (Indianapolis, In) and ³⁵S-UTP (1000 Ci/mmole) was from Amersham.

Human neonatal foreskin keratinocyte cultures were initiated and propagated in complete MCDB 153 medium with 0.03 mM CaCl2 as previously described (Boyce and Ham, 1983; Shipley and Pit-telkow, 1987; Ando and Jensen, 1993). Keratinocytes from secondary or tertiary passages were used for skin equivalent prep aration.

Primary human dermal fibroblasts were initiated from trypsin-treated and epidermis-stripped neonatal foreskin using DMEM plus 20% fetal bovine serum; these cells were passaged in DMEM medium supplemented with 10% fetal bovine serum, 5 mM Hepes, 1 mM sodium pyruvate, penicillin (100 i.u./ml), and streptomycin (100 mg/ml). Fibroblast cultures between the 6th and 10th passages were used for the preparation of the collagen lattice (dermal equivalent).

The procedure for skin equivalent preparation was similar to that reported by Bell et al. (1983) and Asselineau and Prunieras (1984). Nearly confluent cultures of human dermal fibroblasts from neonatal foreskin were trypsinized, and approximately 10⁶ cells were suspended in 4.2 ml of DMEM plus 20% fetal bovine serum. A mixture of 0.4 ml of 0.1 N NaOH, 1.4 ml of 3× concentrated DMEM (prepared from powder), and 2.4 ml of rat tail collagen (mostly type 1 collagen, 3.0 mg/ml) was added to the above cell suspension; the mixture was immediately cast into a 60-mm bacterial Petri dish and incubated at 37°C. The fibroblast/collagen mixture geled in 10 min. The collagen gel started to contract away from the wall of dish in a few hours and became an opaque, free-floating disc with a diameter of 1.5 cm after 7 d of incubation.

The collagen gel (dermal equivalent) was then transferred to a 60mm tissue culture dish and anchored to the bottom with 1 ml of cell-free collagen suspension. A stainless steel ring with inner area of 1.5 cm² was then placed on top of the collagen gel. Approximately 5 ×10⁵ human neonatal foreskin keratinocytes were seeded inside the ring in a total of 0.5 ml of keratinocyte culture medium (DMEM/Ham’s F-12 (3:1) plus 5% fetal bovine serum, supplemented with 1.8×10⁻⁴ M adenine, 5 μg/ml insulin, 5 μg/ml trans - ferrin, 2 nM tri-iodothyronine, 0.5 μg/ml hydrocortisone, 1 mM sodium pyruvate, 3 mM Hepes, 100 i.u./ml penicillin, and 100 μg/ml streptomycin). The stainless steel ring was removed 3 h later and the culture was kept submerged for one week. The skin equivalent culture was then raised to the air-liquid interface on a stainless steel grid placed in a 60-mm Petri dish. Medium was changed every 2-3 d. The skin equivalents were harvested after 10 days of air exposure.

For morphological evaluation, the skin equivalent was fixed with 4% paraformaldehyde and embedded in JB-4 compound as described by the manufacturer. Cellular proliferation in the skin equivalent was demonstrated by incubation with bromodeoxyuridine (BrdU) or [³H]thymidine, followed by immunohistochemical (Cell Proliferation Kit, Amersham) or autoradiographic (Cotsarelis et al., 1989) detection, respectively. Immunohistochemical localization of differentiation markers was conducted as for PA components described below.

We used the procedure previously described (Grøndahl-Hansen et al., 1987; Miller et al., 1992) with the following slight modifications. Skin equivalents were embedded in PolyFreeze, snap frozen in liquid nitrogen, and stored at -70°C. Frozen sections (6 μm) were cut, applied to Superfrost Plus slides, and air-dried for 30 min before storage at -70°C. All antibody incubations were for 1 hour at room temperature. For primary polyclonal and monoclonal antibodies, the concentrations used were respectively 2.5-5.0 μg/ml and 5.0-10.0 μg/ml. For the detection of uPA, tPA, PAI-1, and PAI-2, both monoclonal and polyclonal antibodies were employed, with comparable results. Visualization of antibody binding was performed with avidin-biotin-peroxidase and the substrate DAB. Slides were rinsed with water and counterstained with hematoxylin Gill’s number 1, dehydrated through graded ethanols, cleared with xylenes and then mounted with Permount.

A 610 bp fragment (positions 1364-1973) of the human uPA cDNA (pHUK-8) subcloned in pSP64 and pSP65 was the template for synthesis of antisense and sense uPA RNA probes respectively (Verde et al., 1981). A 614 bp fragment (positions 188-810) of the human tPA cDNA (pW349F) subcloned into pSP65 and pSP64 provided a template for antisense and sense tPA cRNA probes respectively (Fisher et al., 1985). These clones were the kind gifts of Dr D. Belin (University of Geneva, Switzerland). RNA probes were generated by in vitro transcription from the SP6 promoter. A 1.3-kilobase human PAI-1 cDNA fragment, subcloned in the vector pGEM-3, was used as a template for in vitro transcription of antisense or sense probes by SP6 or T7 RNA polymerase respectively (Schneiderman, 1992). A full-length PAI-2 cDNA, excised from pJ7/PAI-2 and inserted between the T3 and T7 promoter sites of Bluescript M13+ (Schleuning et al., 1987), was used as a template for the synthesis of antisense cRNA probe using the T7 promoter and sense probe using the T3 promoter. These PAI-1 and PAI-2 clones were generously provided by Dr David Loskutoff (Scripps Clinic and Research Foundation, La Jolla, CA) and Dr Egbert Kruithof (University of Lausanne, Switzerland), respectively.

Transcription reactions were conducted in the presence of ³⁵S-UTP (1000 Ci/mmole). Unincorporated radiolabel was removed from the probes by column purification. The specificity of the antisense cRNA probes for uPA, tPA, PAI-1 and PAI-2 was verified by northern analyses of total RNA prepared from keratinocyte cultures (Baird et al., 1990; Lyons-Giordano, unpublished results).

Skin equivalents for in situ hybridization were fixed in 4% paraformaldehyde in PBS at 4°C overnight, dehydrated through graded ethanols, washed twice in xylene, and paraffin embedded. 6 μm sections were applied to Superfrost Plus slides.

Sections were deparaffinized in xylenes, rehydrated in 50% ethanol in PBS, and then fixed in 4% paraformaldehyde in 0.1 M sodium phosphate pH 7.4 for 10 min at 4°C. Sections were washed in 0.5× SSC (1× SSC is 0.015 M sodium citrate, 0.15 M sodium chloride), incubated with 10 μg/ml proteinase K in 10 mM Tris-HCl pH 8.0 containing 0.5 M sodium chloride for 10 min at room temperature, and washed in 0.5× SSC. The sections were prehybridized for 1 h at 55°C in a mixture of 50% formamide, 0.3 M sodium chloride, 20 mM Tris-HCl pH 8.0, 5 mM EDTA, 1× Den-hardt’s solution (0.02% bovine serum albumin, 0.02% Ficoll, 0.02% polyvinylpyrolidone), 10% dextran sulfate, 10 mM DTT, and 500 μg/ml transfer RNA. Hybridizations were conducted in the same buffer overnight at 55°C in the presence of ³⁵S-labeled cRNA probes (6×10⁶ cpm/ml). Non-specifically bound probe was removed by digestion with RNase A (20 μg/ml) in 10 mM Tris-HCl pH 8.0 containing 0.5 M sodium chloride for 45 min at room temperature. Sections were washed at high stringency in 0.1 × SSC containing 1 mM EDTA and 10 mM 2-mercaptoethanol for 2 h at 68°C for hybridizations using uPA and tPA probes and at 60°C for those with PAI-1 and PAI-2 probes. The sections were dehydrated and exposed to Polysciences K 2 autoradiographic emulsion for 10-14 days. After development, they were stained with hematoxylin and eosin. Sections were examined and photographed using epipolarization microscopy. The specificity of each hybridization was verified by the absence of detectable signal when radiolabeled sense probes for uPA, tPA, PAI-1 and PAI-2 replaced the antisense probes. Appropriate sense controls for each specimen were included in all hybridization experiments.

Morphological examination of epidermis in the skin equivalent revealed cuboidal keratinocytes in the basal layer, several suprabasal layers of flattened keratinocytes, and partially anucleate cornified keratinocytes on the surface (Fig. 1A; Asselineau et al., 1986; Kopan et al., 1987). Proliferating keratinocytes were detected only in the basal layer of epidermis, in agreement with previous investigations (Kopan and Fuchs, 1989). The dermal fibroblasts were evenly distributed in the collagen matrix, mostly in stellate shape; no keratin-positive cells were detectable in the dermis.

Keratinocytes in the skin equivalent expressed differentiation markers characteristic of human epidermis, including K1/K10 keratin, involucrin, transglutaminase, and pro-filaggrin/filaggrin (Asselineau et al., 1986, 1989; Kopan et al., 1987; Choi and Fuchs, 1990). As in normal epidermis, all suprabasal keratinocytes expressed K1/K10 keratin (Fig. 1B). In normal epidermis, involucrin and transglutaminase are found only in the more superficial layers; however, in skin equivalent, these markers were expressed by ker-atinocytes immediately upon leaving the basal layer (Fig. 1C). This expression pattern for involucrin and transglutaminase resembles that found in healing wounds (Mans-bridge and Knapp, 1987) or in the hyperproliferative disease psoriasis (Murphy et al., 1984; Bernard et al., 1988; Thacher, 1989) and reflects the many similarities that have previously been noted between these two in vivo states and normal keratinocytes in culture (Mansbridge and Knapp, 1987).

Our characterization of the skin equivalent is similar to previously published data (Asselineau et al., 1986, 1989; Kopan et al., 1987; Choi and Fuchs, 1990) and indicates that, although the skin equivalent does not exactly reproduce the epidermal differentiation sequela, it is a useful model for the study of differentiation-related events in the epidermis.

Antigen and mRNA for both uPA and PAI-1 were localized to the keratinocytes of the basal layer and, in some fields, one layer above (Figs 2, 3). Strong and consistent staining for PAI-1 was also found beneath the basal ker-atinocytes, extending into the collagen matrix (Fig. 3A). In many fields there was a faint deposition of uPA antigen beneath the basal cells (Fig. 2A). Thus the basal ker-atinocytes synthesize uPA and PAI-1, but soon after assuming a suprabasal position, the keratinocytes cease production of both of these PA cascade components.

In contrast to uPA and PAI-1, tPA was localized predominantly to the suprabasal layers (Fig. 4). Tissue PA mRNA was detected in the cells that were one or two layers above the basal layer, but it did not usually extend to the uppermost layers. Staining for tPA antigen was sometimes faintly detected close to the basal keratinocytes, but it was most intense in the upper half of the epidermis, with the exception of the cornified layers, which did not stain. The tPA distribution was focal; i.e. some suprabasal fields were intensely positive, while adjacent ones were much fainter or negative. These results indicate that keratinocytes begin to synthesize tPA when they leave the basal layer. Synthesis of tPA mRNA appears to be transient, only occurring in a few epidermal layers; however, the antigen is retained as the cell differentiates further.

PAI-2 mRNA and antigen gave strong signals in most epidermal layers, with the exception of the basal cells where the signals were usually weaker or undetectable (Fig. 5). In contrast to the other PA cascade components, PAI-2 appeared concentrated around the cell periphery (Fig. 5A). This pattern resembled that of involucrin, a protein that is incorporated into the cell envelope of the terminally differentiated squames, and transglutaminase, an enzyme necessary for cross-linking of the envelope (compare Figs 5A and 1C).

Our localization data for all of the epidermal PA components are summarized in diagrammatical form in Fig. 6.

Neither antigen nor mRNA for tPA was detected in the cells in the collagen matrix. Occasionally, fibroblasts embedded in the collagen matrix stained for uPA and PAI-2 antigens but their mRNAs were never detected in these cells. PAI-1 antigen and mRNA were detected in some dermal fibroblasts. The collagen matrix itself also consistently stained for PAI-1, suggesting that this inhibitor was distributed throughout the matrix. PAI-1 is known to bind to vitronectin (Declerck et al., 1988; Seiffert and Loskutoff, 1991). Since the collagen gel is contracted and maintained in the presence of serum, it is very possible that vitronectin is present in the collagen gel and mediates binding of PAI-1 therein.

Superficially, there appears to be considerable redundancy to the PA cascade; i.e. the two PAs catalyze the same reaction (the conversion of plasminogen to plasmin) and the two inhibitors are capable of blocking both of the enzymes. However, many biochemical studies have revealed large differences in kinetic and regulatory properties between the two enzymes and between the two inhibitors. These results have suggested that the apparently redundant PA cascade components have in fact evolved to fulfill different functions (for reviews see Danø et al., 1985; Pollanen et al., 1991; Vassalli et al., 1991).

In the present study we have used a differentiating model of human epidermis to evaluate the changes that occur in the PA cascade as a function of keratinocyte differentiation. Our data clearly demonstrate that the two PA enzymes, as well as the two inhibitors, are synthesized by keratinocytes at specific and distinct stages of differentiation (Fig. 6).

uPA is detected by in situ hybridization and immunohistochemical staining in the basal keratinocytes, but little or no signal is present in the suprabasal cells. In contrast, tPA mRNA is not detectable in the basal cells, but is present only in a few layers of immediately suprabasal cells. tPA antigen is found in the upper half of the non-cornified epidermis. Our results indicate that when basal keratinocytes begin to differentiate further and stratify, their synthesis of uPA greatly declines and their synthesis of tPA dramatically increases. This switch in the type of PA produced by the keratinocyte strongly argues for distinct roles, related to epidermal differentiation state, for the two enzymes. Similar observations apply to the two inhibitors. PAI-1 mRNA and antigen are present selectively in the basal ker-atinocytes; suprabasal cells make little or no PAI-1. In contrast, PAI-2 mRNA and antigen are primarily and most consistently localized in the suprabasal layers.

Many biochemical and morphological changes occur in the keratinocyte during terminal differentiation. Our data indicate that alterations in the biosynthesis of PA enzymes and inhibitors represent an additional aspect of differentiation-related changes in the epidermis.

The distinct localizations of the PA cascade components invite comparisons to previously published in vivo data and provoke several speculations concerning modes of regulation and biological roles. Using northern analysis, mRNA for uPA is observed in normal epidermal extracts (Baird et al., 1990), although its levels are apparently too low to be detectable by in situ hybridization (Lyons-Giordano et al., 1993). However, uPA enzymatic activity is observed selectively in the basal layer of normal epidermis (Lyons-Giordano et al., 1993), consistent with our localization of uPA in the skin equivalent. As the basal keratinocytes constitute the only proliferative epidermal layer, these data raise the possibility that expression of uPA may be correlated with proliferative capacity in the epidermis. Several previous reports have suggested that uPA can enhance proliferation and that uPA expression is correlated with transition to the proliferative state (Kirchheimer et al., 1987a,b; Rabbani et al., 1990, 1992; Grimaldi et al., 1986).

The basal keratinocytes are also unique in that they constitute the only layer directly attached to the underlying extracellular matrix. Hence the basal and sub-basal localization of uPA, as well as of the inhibitor PAI-1, suggests that these molecules may play a role in regulation of basal keratinocyte-matrix interaction. As for uPA, PAI-1 is not detectable by immunostaining or by in situ hybridization in normal human or murine epidermis (Rømer et al., 1991; Lyons-Giordano et al., 1993). However, uPA and PAI-1 are expressed in re-epithelializing cutaneous wounds in vivo, where keratinocyte migration is occurring to begin the repair process (Grøndahl-Hansen et al., 1988; Rømer et al., 1991). The localizations of uPA and PAI-1 in the skin equivalent are remarkably similar to those observed in epithelial outgrowths in vivo. Both in re-epithelializing wounds as well as in the skin equivalent, it is of interest that PAI-1 is often found beneath the epithelium near the basal aspect of the basal cells (Rømer et al., 1991). The adhesive interactions between basal keratinocytes and the extracellular matrix must be carefully regulated as the cells migrate through the provisional matrix during the early stages of wound repair; our results are consistent with the hypothesis that PAI-1 and uPA are involved in such regulation. The detachment of basal keratinocytes from the basement membrane during normal stratification and differentiation also may require proteolytic activity; basal cell uPA may aid in this detachment.

In contrast to the basal localizations of uPA and PAI-1, tPA and PAI-2 are predominantly localized in the upper keratinocyte layers, suggesting that these molecules may be involved in some aspect(s) of differentiation. In the skin equivalent, tPA mRNA is expressed in a rather narrow layer immediately above the basal cells; however, tPA antigen is retained up to the cornified layers. In a variety of human cutaneous lesions as well as in murine wounds, tPA antigen is similarly detected in the suprabasal layers (Grøndahl-Hansen et al., 1987, 1988; Jensen et al., 1988). Although it has not been possible to localize tPA antigen or mRNA in normal epidermis, small amounts of tPA activity are detectable in some normal human epidermal extracts (Jensen et al., 1988; Baird et al., 1990; Lyons-Giordano et al., 1993; Grøndahl-Hansen et al., 1987). Although the role of epidermal tPA is unknown, its preferential location in the superficial layers suggests a differentiation-related function.

PAI-2 is appropriately placed for regulation of tPA, both in the skin equivalent and in epidermis (Lyons-Giordano et al., 1993). This correlation is intriguing, since there is good evidence from other tissues that PAI-1, which rapidly inhibits both the single-chain and two-chain forms of tPA, is the primary regulator of this enzyme (Erickson et al., 1985; Loskutoff, 1988). Although PAI-2 is a rather slow inhibitor of single chain tPA, it is a more rapid inhibitor of tPA that has been cleaved to its two-chain form (Andreasen et al., 1990). Both single-chain and two-chain tPA have potent enzymatic activity (Rijken et al., 1982).

Hence cleavage of tPA might provide a mechanism by which this enzyme becomes more susceptible to an endogenous epidermal inhibitor. The distribution of PAI-2 along the cell periphery in the same region as involucrin and transglutaminase suggests the possibility that PAI-2 may be involved in cornified envelope formation in the terminally differentiating keratinocyte. For example, PAI-2 may protect the newly-forming envelope from proteolytic damage. Alternatively, we cannot discount the possibility that PAI-2 may have a role independent of its ability to inhibit proteolysis, even though this molecule was initially identified as an inhibitor of PA. PAI-2 belongs to a class of inhibitors known as serpins, and precedent exists among other members of this class for roles not directly related to proteolytic inhibition (Kilpatrick et al., 1991; Joslin et al., 1992).

Although much further work is required to establish the biological roles of epidermal PA, the present report clearly demonstrates that the two PA enzymes, as well as the two inhibitors, are synthesized by keratinocytes at distinct stages of differentiation. The specific localizations suggest possible sites and modes of action, which can be tested further with the skin equivalent model system.

The authors thank Drs Norman Schechter and Robert Lavker for critical reading of the manuscript. Donations of antibodies by the following investigators are gratefully acknowledged: Dr Francee Boches (Baxter); Dr Gary L. Davis (Dupont-Merck); Dr Paul Horan (Smith-Kline Beecham); Dr T.-T. Sun (New York University); Dr Tze-Chein Wun (Monsanto). cDNA clones were very kindly provided by Dr Dominique Belin (University of Geneva, Switzerland); Dr Egbert Kruithof (University of Lausanne, Switzerland); Dr David Loskutoff (Scripps Clinic and Research Foundation). The authors also thank Dr Loskutoff for assistance with the in situ hybridization procedure. This work was supported by grant AR-39674 from the NIH.