Targeting Perlecan in Human Keratinocytes Reveals Novel Roles for Perlecan in Epidermal Formation*

Heparin-binding growth factors are crucial for the formation of human epidermis, but little is known about the role of heparan sulfate proteoglycans in this process. Here we investigated the role of the heparan sulfate proteoglycan, perlecan, in the formation of human epidermis, by utilizing in vitro engineered human skin. By disrupting perlecan expression either in the dermis or the epidermis, we found that epidermally derived perlecan is essential for epidermal formation. Perlecan-deficient keratinocytes formed a strikingly thin and poorly organized epidermis because of premature apoptosis and failure to complete their stratification program. Exogenous perlecan fully restored epidermal formation. Perlecan deposition in the basement membrane zone correlated with formation of multilayered epidermis. Perlecan deficiency, however, had no effect on the lining and deposition of major basement membrane components as was evident by a continuous linear staining of laminin and collagen IV. Similarly, perlecan deficiency did not affect the distribution of β1 integrin. Addition of the perlecan ligand, fibroblast growth factor 7, protected perlecan-deficient keratinocytes from cell death and improved the thickness of the epidermis. Taken together, our results revealed novel roles for perlecan in epidermal formation. Perlecan regulates both the survival and terminal differentiation steps of keratinocytes. Our results suggested a model whereby perlecan regulates these processes via controlling the bioavailability of perlecan-binding soluble factors involved in epidermal morphogenesis.

Heparin-binding growth factors are crucial for the formation of human epidermis, but little is known about the role of heparan sulfate proteoglycans in this process. Here we investigated the role of the heparan sulfate proteoglycan, perlecan, in the formation of human epidermis, by utilizing in vitro engineered human skin. By disrupting perlecan expression either in the dermis or the epidermis, we found that epidermally derived perlecan is essential for epidermal formation. Perlecan-deficient keratinocytes formed a strikingly thin and poorly organized epidermis because of premature apoptosis and failure to complete their stratification program. Exogenous perlecan fully restored epidermal formation. Perlecan deposition in the basement membrane zone correlated with formation of multilayered epidermis. Perlecan deficiency, however, had no effect on the lining and deposition of major basement membrane components as was evident by a continuous linear staining of laminin and collagen IV. Similarly, perlecan deficiency did not affect the distribution of ␤1 integrin. Addition of the perlecan ligand, fibroblast growth factor 7, protected perlecan-deficient keratinocytes from cell death and improved the thickness of the epidermis. Taken together, our results revealed novel roles for perlecan in epidermal formation. Perlecan regulates both the survival and terminal differentiation steps of keratinocytes. Our results suggested a model whereby perlecan regulates these processes via controlling the bioavailability of perlecan-binding soluble factors involved in epidermal morphogenesis.
Skin is the largest organ of the body. It serves as a shield against microorganism invasion and UV radiation, prevents dehydration, regulates body temperature, and is a part of the immune system (1). It consists of two distinct tissues, epidermis and dermis that are separated by a basement membrane (BM). 4 The dermis is a dense collagen-rich connective tissue that provides the support and nourishment to the overlying epidermis. It is mainly composed of fibroblasts that synthesize and secrete the various extracellular matrix (ECM) components (2). The epidermis is made primarily of keratinocytes that form a stratified squamous epithelium. It consists of multiple layers exhibiting distinct morphology and function. From the innermost to the outermost layers, they are the basal, spinous, granular, and cornified strata (3). Throughout adult life, the epidermis undergoes continuous self-renewal through proliferation of the basal cells, the only cells of the epidermis with the ability to proliferate. The keratinocytes undergo terminal differentiation as they leave the basal layer and move upward through the suprabasal layers toward the tissue surface, where they die and are sloughed off (3). This process makes the skin an excellent model system to study the coordinated regulation of cell proliferation, cell differentiation, and cell death.
The formation of the mature epidermis is regulated by cross-talk with the adjacent connective tissue through a network of cytokines and growth factors that are secreted from keratinocytes and dermal fibroblasts (4,5). In addition, cell-cell and cell-matrix interactions control epidermis proliferation and differentiation (6). Many of the growth factors implicated in formation of the epidermis are heparin-binding molecules, including fibroblast growth factor 7 (FGF7) and granulocytemonocyte colony-stimulating factor (5,7). Little is known, however, about the roles of heparan sulfate proteoglycans (HSPGs), which interact with these growth factors, in epidermal formation.
HSPGs are the common constituent of cell surfaces and the ECM, including the BM (8). The basic HSPG structure consists of a protein core to which several linear heparan sulfate chains are covalently attached. At the cell surface, the two major families are the syndecans and glypicans, and the HSPGs found in the ECM are perlecan, agrin, and collagen XVIII (9). HSPGs are implicated in regulating the integrity of BMs, morphogenesis, angiogenesis, tumor metastasis, and tissue repair (10 -14). These activities are attributed to the ability of HSPGs to bind and modulate the biological activities of mitogenic and angiogenic growth factors such as vascular endothelial growth factor and FGF (15,16). HSPGs can protect heparinbinding growth factors from thermal denaturation and proteolytic degradation, and HSPGs of the ECM provide a reservoir from which these growth factors can be rapidly released in response to specific triggering events (15). HSPGs can increase heparin-binding growth factor efficacy by stimulating receptor binding and signaling and influence growth factor/ receptor binding specificity (17,18).
Perlecan, a large HSPG, is present in virtually all BMs (19 -21). It interacts with basement membrane components such as laminin-1 and collagen IV and cell adhesion molecules such as ␤1 integrin (22,23). Inactivation of the perlecan gene in mice results in embryonic lethality and prenatal death (24,25). Similarly, the functional null mutation in the perlecan gene in humans, termed dyssegmental dysplasia, Silverman-Handmaker type (DDSH), is characterized by either stillbirth or death within a few days after birth (26,27). The majority of perlecan null mice die between embryonic days 10 and 12, before the onset of skin differentiation, and those that survive die perinatally (3,24,25). As perlecan is expressed in two-cell embryos, it was rather unexpected that mice null for perlecan will survive beyond this stage. Another unexpected finding was that the major defect in these mice was in cartilage formation, albeit that perlecan protein is present in most basement membranes. It was suggested that other BM components, such as collagen XVIII, can compensate for the lack of perlecan (20). Currently, it is not known whether loss of perlecan function in human has any effect on epidermal formation. Studies in monolayer cultures revealed that perlecan binds and modulates the activity of a variety of growth factors, including FGF7 and GM-CSF that are implicated in human skin formation (28,29). In addition, perlecan is expressed in the skin of human as well as other mammalians (19,21). Taken together, these findings suggest that perlecan may play a role in skin biology. In the present work we utilized in vitro engineered human skin to address the role of perlecan in epidermal formation.
In the engineered skin model (also known as organotypic culture or skin equivalents (SEs)), keratinocytes are cultivated at the air/liquid interphase on various substrates serving as dermal equivalents, including collagen gels populated with dermal fibroblasts (30 -32). The SEs resemble adult human skin in many functional aspects. Unlike monolayer cultures that feature poor differentiation characteristics, keratinocytes in SEs display the architecture, barrier function, and gene expression profile of the normal human epidermis (33,34). Similar to the in vivo situation, formation of mature epidermis in SEs depends on interaction with dermal fibroblasts, and BM constituents are produced in a cooperative manner by the keratinocytes and fibroblasts (35). Moreover, the cells in each of the SE compartments can be individually manipulated, and the effect of these manipulations on epidermis formation can be examined. Finally, because of differences between human and murine skin, and the known differences between the behavior of mouse and human keratinocytes in culture, SEs generated with human keratinocytes currently represent the best relevant system to study human epidermis biology and pathology (36 -39). Hence, skin equivalents are widely used to examine epidermal gene expression, epidermal and dermal interactions, and wound repair and are increasingly employed as an in vitro test in pharmacotoxicologic studies (40 -42).
To address the role of perlecan in the morphogenesis of human epidermis, we disrupted perlecan expression in cells of each of the skin compartments, and we studied the effect of these deficiencies on epidermal formation in the reconstituted skin model. This study revealed novel roles for perlecan in regulating human epidermal morphogenesis.

EXPERIMENTAL PROCEDURES
Cell Cultures-Normal human skin keratinocytes and dermal fibroblasts were isolated from adult skin and cultivated as described (32). 3T3 murine fibroblast cells were used as feeder cells for primary keratinocytes following ␥-irradiation of 5000 rads (43). HaCaT cells were grown as described previously (44). DDSH human embryonic fibroblasts and mouse embryonic fibroblasts from wild type and perlecan-null mice were prepared as described (25,27).
Antisense Vector and Generation of Stably Transfected Clones-A cDNA fragment complementary to nucleotides 3120 -4120 of human perlecan (domain III) was cloned in an antisense orientation into pcDNA3 expression vector (45). HaCaT cells were transfected with this vector using the calcium phosphate method and selected with 800 g/ml G418 (Invitrogen). Clones were isolated from three independent transfections and maintained in growth medium containing 400 g/ml G418. Condition medium from different clones was examined for perlecan expression by slot blot analysis with an anti-human perlecan antibody (Zymed Laboratories Inc.).
Analysis of Perlecan Expression in Human Skin and Skin Equivalents-Total RNA was extracted as described (5). 2 g of total RNA were used for first strand synthesis with random hexamer primers. PCR was performed with human perlecan or glyceraldehyde-3-phosphate dehydrogenase-specific primers. For in situ hybridization, digoxigenin-labeled perlecan RNA probe was prepared from a plasmid containing a cDNA fragment encompassing nucleotides 104 -459 of perlecan, using the digoxigenin RNA labeling kit (sp6/T7), according to the manufacturer's instructions (Roche Applied Science). 4-m paraffin sections were treated with 20 g/ml proteinase K (Roche Applied Science), and hybridization was performed at 45°C for 16 h in a moist chamber. Hybridized probes were detected with a digoxigenin detection kit (alkaline phosphatase; Roche Applied Science).
Erk1/2 Activation and Mitogenic Assays-HaCaT cultures were serum-starved in the presence of AG 1478 (0.5 M; Alexis) as described previously (46). For mitogenic assays, cells were stimulated with increasing concentrations of FGF7 for 16 h. [ 3 H]Thymidine (ICN) was added for 6 h, and thymidine incorporation was determined using a Wallac-1049 liquid scintillation counter (47). At least three independent experiments in duplicates were performed for three independent PDK clones. For Erk1/2 activation assay, cells were stimulated with 10 ng/ml FGF7 for 10 min, and cell lysates were processed as described (48).
Skin Equivalents-HaCaT clones were seeded (1 ϫ 10 6 cells/insert) onto rat tail tendon collagen type I gels containing 3 ϫ 10 5 /ml human dermal fibroblasts. Primary keratinocytes were seeded (1 ϫ 10 6 cells/ insert) onto collagen gels containing 2 ϫ 10 5 /ml DDSH, normal human dermal fibroblast, or mouse embryonic fibroblast cells. The gels were pre-cast in cell culture filter inserts (pore size 3.0 m, polycarbonate; BD Biosciences) placed in either 6-or 12-deep well plates (BD Biosciences) and cultivated as described in detail previously (32). When indicated, 10 -1000 ng/ml recombinant human FGF7 were added to the growth medium (49). At least three independent experiments in duplicates were performed for each SE type, and all experiments with PDKs were carried out with 2-3 independent clones.
For perlecan reconstitution experiments, 10 g/ml human perlecan, purified from human umbilical arterial endothelial cells (50) was added twice to SEs. The first application was on top of the collagen gels before keratinocytes were seeded. Following incubation at 37°C for 16 h, perlecan-containing medium was replaced with fresh medium containing keratinocytes and 10 g/ml perlecan. This medium was aspirated 3 days later, when cultures were exposed to air.
In all experiments, SEs were maintained at the air/liquid interphase for 7-42 days as indicated in the text. Cultures were then fixed in 3.7% phosphate-buffered formaldehyde and embedded in paraffin.
Histology, Immunofluorescence, and Protein Detection-4-m paraffin sections were processed either for hematoxylin-eosin (HE) staining or for immunofluorescence experiments. Immunofluorescence was performed following antigen retrieval according to the manufacturer's instruc-tions, using the following antibodies: mouse monoclonal antibodies directed against K1/10 (clone 8.60, Sigma), involucrin (Sy5, Sigma), perlecan (Zymed Laboratories Inc.), collagen type IV (Sigma), and ␤1 integrin (Santa Cruz Biotechnology). Rabbit polyclonal antibodies were directed against human loricrin (Babco), filaggrin (Babco), and laminin (Engelbreth-Holm-Swarm (EHS) mouse sarcoma cells; Progen). Species-specific fluorochrome-conjugated secondary antibodies were from Jackson Immu-noResearch. Bisbenzimide (Hoechst 33258) DNA dye was used at 0.5 g/ml for nuclear counterstaining. Results representing the expression of differentiation markers and components of BM were reproducibly obtained in at least two separate experiments and using at least two PDK clones. Proliferating cells were visualized using a mouse monoclonal antihuman Ki67/MibI antibody (Dako). Cell proliferation was quantified by counting the ratio of Ki67-labeled cells within the basal layer of the epithelium in sections derived from two to three independent experiments on three PDK clones. For statistical analysis, at least 1000 nuclei were counted. Apoptotic cells were detected by TUNEL staining following the manufacturer's protocol (In situ cell death detection kit TMR Red; Roche Applied Science). The percentage of TUNEL-positive nuclei from the total cell number within the basal layer was calculated from sections derived from two independent experiments on three PDK and two normal clones, and at least 1000 nuclei were counted for each experiment.

RESULTS
The Effect of Perlecan Deficiency on Epidermal Formation-In situ hybridization was employed to verify that the expression pattern of perlecan, in SEs, resembles that of adult human skin. In human skin specimens from healthy adult donors, perlecan transcripts were specifically detected in the dermis and epidermis (Fig. 1, A and E). Perlecan transcripts were detected in both dermal fibroblasts and epidermis of SEs ( Fig. 1, C and F). In the epidermis of both normal human skin and SEs, perlecan transcripts were present in all layers except for the stratum corneum. RT-PCR analysis using perlecan-specific primers confirmed perlecan expression in primary human dermal fibroblasts, primary and immortalized human keratinocytes, in both monolayer cultures and SEs (Fig. 1G).
To investigate the role of perlecan in skin organogenesis, we examined the effect of targeted disruption of perlecan expression, either in epidermal keratinocytes or in dermal fibroblasts, on formation of the epidermis. First, we examined the role of dermally derived perlecan by constructing SEs using perlecan null fibroblasts and human primary keratinocytes. Fibroblasts were derived either from human fetus carrying the DDSH syndrome or from perlecan-null mice (25,27). Both fibroblast types fully supported epidermal formation, as evident from the appearance of a well organized, terminally differentiated epidermis containing spinous, granular, and stratum corneum layers. The normal morphology of the epidermis was also reflected in the regular expression and localization of proliferation and differentiation markers. Thus, proliferating cells were confined to the basal layer as determined by Ki67 staining (51); the spinous layers expressed the intermediate differentiation marker involucrin, and the late differentiation markers filaggrin and loricrin were readily detected in the stratum corneum (Fig. 2, and data not shown).
Next, we examined the role of epidermally derived perlecan by targeting its expression in the spontaneously immortalized human keratinocyte cell line, HaCaT (44,52,53). Cells were transfected with an antisense construct targeting domain III of the human perlecan gene. . Bar, 100 m. EP, DR, and SC denote epidermis, dermis, and stratum corneum, respectively. Arrows point to dermal fibroblasts. G, total RNA was extracted from epidermal layers of human SE or monolayer cultures of primary (Pr) keratinocytes, primary dermal fibroblasts, or immortalized human keratinocytes (HaCaT). Perlecan mRNA levels were determined by RT-PCR using a specific primer set from domain I of the human perlecan cDNA. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA was used as internal standard. 8d and 14d denote days in air/liquid interphase. NC indicates negative control. FIGURE 2. Dermally derived perlecan is dispensable for epidermal formation. Epidermal tissue morphology of SEs constructed with primary human keratinocytes and wild type (wt) or perlecan null (Ϫ/Ϫ) mouse embryonic (MEF) fibroblast. Cross-sections were either stained with hematoxylin and eosin or analyzed by immunofluorescence staining for proliferation (Ki67, red), intermediate (involucrin, red), and late (loricrin, green) differentiation markers. Nuclei were counterstained with Hoechst DNA dye (blue). Keratinocytes-collagen matrix boundaries are indicated by a dotted line. Bar, 100 m The antisense strategy has been applied successfully to suppress perlecan gene expression in both neoplastic and normal cells (45, 54 -56). The reduction in perlecan levels following gene targeting was determined by measuring the amounts of secreted perlecan in individual clones or by RT-PCR (Fig. 3A, and data not shown). We obtained clones that express relatively normal, intermediate, or barely detectable levels of perlecan. The proliferation rate of the different clones in monocultures was similar (Fig. 4, and data not shown). Clones that expressed the lowest levels of perlecan (Fig. 3A, lanes 2-4) were designated PDKs, for perlecan-deficient keratinocytes. Next, we examined the ability of PDKs to form multilayered epidermis when seeded on top of dermal equivalents containing human primary dermal fibroblasts. Although control clones formed a well structured multilayered epidermis, PDKs formed a strikingly thin and poorly organized epidermis with only a single or two layers of flattened cells (Fig. 3B).
To investigate further the mechanism underlying the observed morphological defects in perlecan-deficient epidermis, we first examined the expression pattern of epidermis differentiation markers in SEs grown for 2 or 3 weeks in the air/liquid interphase. Similar to control clones, the epidermis of PDKs stained positive with antibodies directed against the early differentiation markers keratins 1 and 10, suggesting that the keratinocytes successfully initiated their differentiation program (57). In contrast, only the control cultures stained positive for the intermediate and late differentiation markers involucrin and filaggrin (Fig. 3B). These results suggested that PDKs failed to complete their differentiation program.
To confirm that the defect in epidermal formation results from perlecan deficiency, we performed reconstitution experiments with purified human perlecan. Exogenous addition of perlecan, at a concentration of 10 g/ml, restored the ability of PDKs to form a multilayered, well differentiated epidermis (Fig. 3C). Therefore, our findings indicate that perlecan derived from keratinocytes is essential for these cells to form multilayered and fully differentiated epidermis.
Perlecan-deficient Epidermis Exhibits an Increased Apoptosis Rate-The differentiation defect in PDK-derived epidermis could not explain  2-7). The volume of conditioned medium was normalized to cell number. B, epidermis morphology in SEs constructed with control or PDK clones grown for 14 or 21 days (14d or 21d) in air/liquid interphase. Cross-sections were fixed and stained with HE. 21-Day cultures were analyzed for the differentiation markers K1/10 (red), involucrin (red), and filaggrin (green) as described under "Experimental Procedures." Nuclei were counterstained with Hoechst DNA dye (blue). C, exogenous human perlecan restores epidermal formation. Skin equivalents were generated with PDK clones and primary dermal fibroblasts in the presence of 10 g/ml purified perlecan, as described under "Experimental Procedures." SEs were fixed and processed 18 days following air exposure. Bar, 100 m. Similar results were obtained with two additional PDK clones. FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 the reduced number of epidermis cell layers. Thin epidermis could have resulted from a failure to maintain the progenitor cell population necessary for epidermis regeneration. This could be due to the reduced proliferation rate of basal keratinocytes or premature cell death. To distinguish between these two possibilities, we compared the proliferation and apoptosis rates of keratinocytes in the epidermis of control and PDK SEs. In addition, we compared the response of quiescent monocultures of PDKs and control cells to FGF7, a known mitogen for keratinocytes (58). This study was performed with three independent PDK clones, and representative results are shown in Fig. 4. The number of proliferating cells in the basal layer, determined by Ki67 staining, was identical in the epidermis of both SE types (Fig. 4, A and B). Likewise, monocultures of both normal and PDKs responded equally well to FGF7 in a mitogenic assay, and FGF7 activated Erk1/2 mitogen-activated protein kinases to a similar extent in both cell types (Fig. 4C). A dramatic increase in the rate of apoptosis, however, was detected in the epidermis of PDKs (ϳ40% TUNEL-positive nuclei compared with 4 -8% in basal layer of control SEs (p Ͻ 0.007)) (Fig. 4, D and E). These data strongly suggest that keratinocyte-derived perlecan is required for the survival of the regenerative cell population required to form a multilayered epidermis. Fig. 1). Yet primary dermal fibroblasts did not support normal epidermal formation from PDKs. One plausible explanation may be that fibroblast-derived perlecan does not accumulate in the basement membrane zone (BMZ) in sufficient amounts to support formation of the epidermis. This possibility was addressed by determining the location of perlecan gene product in SEs constructed with various precursor cells. In SEs formed with primary human keratinocytes and fibroblasts null for perlecan expression (DDSH fibroblasts), perlecan was readily detected in the BMZ of fully differentiated epidermis (Fig. 5A). Similarly, in SEs generated with control HaCaT cells and primary dermal fibroblasts, perlecan was observed in the BMZ of fully differentiated epidermis. In these SEs perlecan was also detected in the dermal fibroblasts (Fig. 5B, left). In PDK-derived SEs grown for 28 days in culture, perlecan was readily detected in the dermal fibroblasts but was absent from the BMZ (Fig. 5B). Because dermal fibroblasts express perlecan, and because PDKs express low levels of perlecan, we examined whether cultivating PDK-SEs beyond 28 days may lead to the accumulation of sufficient amounts of perlecan in the BMZ, and whether formation of the epidermis will be improved under such conditions. To test this, PDK-SEs were maintained for 6 weeks in culture. Each week SEs were examined morphologically as well as tested for perlecan protein expression, for apoptosis rate, and for the expres-sion of skin differentiation markers. In agreement with the previous results, cultures grown for 1-4 weeks in the air/liquid interphase were negative for perlecan staining in the BMZ and exhibited the character- FIGURE 5. Perlecan accumulation in the BM correlates with normal epidermal formation. A, SEs were generated with DDSH fibroblasts and primary human keratinocytes. Seven days after air exposure, cultures were fixed and stained with HE or anti-human perlecan monoclonal antibody (red). Nuclei were stained with Hoechst DNA dye (blue). Bar, 100 m. B, SEs were generated with human primary dermal fibroblasts and either control or PDK clones. Perlecan staining was performed with cultures grown for 28 days (28d) (control, and PDK middle panel) or 35 days (35d) in the air/liquid interphase (PDK, right panel). Bar, 100 m. C, longer cultivation time of PDKs in SEs improves epidermis morphology and expression of late differentiation markers. SEs of control and PDKs were fixed at the indicated times after air exposure. Paraffin-embedded sections were either stained with HE or analyzed for the late differentiation marker filaggrin (green), as described under "Experimental Procedures." Cells undergoing apoptosis were detected as described in the legend to Fig. 4. D, deposition and distribution of cell adhesion and BM components in the absence of epidermally derived perlecan. SEs of control and PDK clones grown for 14 days in air/liquid interphase were immunostained for collagen IV (red), laminin (red), or ␤1 integrin (red) as indicated and counterstained with Hoechst DNA dye (blue). Bar, 100 m.

Normal Epidermal Formation Correlates with Accumulation of Perlecan in the BM Zone-Perlecan is expressed in dermal fibroblasts in normal and engineered human skin (see
istic features of perlecan-deficient epidermis. By contrast, in SEs cultivated for 5 or 6 weeks, perlecan staining could be detected in the BMZ, and concomitantly epidermis thickness and morphology were improved. In addition, late differentiation markers were expressed and properly localized, and the apoptotic cells were mainly confined to the upper cornified layer, similar to the situation in normal skin (representative results are shown in Fig. 5, B and C). These findings underscore a strong correlation between the presence of perlecan in the BMZ and its absolute requirement for the formation of multilayered and fully differentiated epidermis.
Perlecan interacts with major BM components and is involved in the adhesion of different cell types to the BM (22,59,60). To investigate how perlecan controls epidermal formation, we examined whether its absence from the BMZ impairs the deposition of BM proteins or affects the expression and distribution of cell-matrix adhesion receptors. To this end, the expression pattern of collagen IV, laminin, and ␤1 integrin were analyzed. Collagen IV and laminin, similar to perlecan, are major constituents of BMs, and ␤1 integrin is involved in the anchorage of keratinocytes to the BM (1, 6, 61). The expression pattern of these proteins was similar in control and perlecan-deficient epidermis cultivated for 14 or 28 days (representative results are shown in Fig. 5D). Taken together, these findings imply that the absence of perlecan from the BMZ did not perturb the deposition and lining of BM components nor did it affect cell-BM adhesion.
Exogenous FGF7 Protects PDKs from Apoptosis in the Reconstituted Skin-The above described studies suggested that perlecan does not modulate the response of keratinocytes to endogenous mitogens and that the defect in epidermal formation does not result from disruption of the BM or keratinocytes-matrix adhesion. An important function of HSPGs of the BM is to sequester growth factors and provide protection from proteolytic degradation (62,63). By virtue of this capacity, HSPGs can increase the local concentration of growth factors. Hence, one consequence of perlecan deficiency could be reduced levels of growth factors that bind perlecan and are required for keratinocyte survival and differentiation. If this hypothesis is correct, exogenous addition of such factors should restore the ability of PDKs to form a multilayered epider-mis. Therefore, we examined the effect of exogenous FGF7 on the ability of PDKs to form multilayered epidermis. FGF7 was chosen because of its known capacity to bind perlecan and the major role it plays in the formation of human epidermis (5,28,64). Furthermore, FGF7 protects a variety of epithelial cells, including keratinocytes, from apoptosis (58,(65)(66)(67)(68). Small quantities of FGF7 (10 -30 g/ml) were added to PDK-SEs for a period of 18 days. Thereafter, SEs were fixed and examined for morphological appearance, the number of proliferating and apoptotic cells, and expression of skin differentiation markers. Representative results are shown in Fig. 6. Addition of FGF7 markedly improved epidermis thickness. This effect was accompanied with a dramatic reduction in apoptosis rate, from 40 to 4 -8%, without an effect on basal keratinocytes proliferation (about 12% of basal cells were positive for Ki67 staining for PDK or control keratinocytes). In addition, FGF7 had little or no effect on the differentiation capacity of PDKs. The epidermis was less organized compared with control cultures, and except for a modest and discontinuous staining for involucrin, late differentiation markers were not detected. Increasing the concentrations of FGF7 did not improve the differentiation of PDKs (data not shown). The inability of FGF7 to fully restore epidermis differentiation is in agreement with previous findings that FGF7 is not directly involved in the differentiation program of keratinocytes (5).

DISCUSSION
A variety of heparin-binding growth factors are involved in skin organogenesis and homeostasis, yet little is known about the HSPGs that regulate these processes. In the present work, we studied the role of perlecan in the morphogenesis of human epidermis by using an engineered human skin model. By genetically disrupting perlecan expression in the dermal or epidermal compartments of the engineered skin, we found that epidermally derived perlecan is essential for human epidermal formation. To the best of our knowledge, this is the first demonstration of perlecan involvement in the morphogenesis of human epidermis.
In normal human skin and in the control SEs, perlecan transcripts FIGURE 6. FGF7 prevents apoptosis of PDKs in SEs. SEs of control and PDK clones were grown for 14 days in air/liquid interphase. When indicated, 30 ng/ml FGF7 was added, every other day, to the growth media of SEs constructed with PDK clones. Cross-sections were either stained with HE or analyzed for the differentiation markers as described in the legend to Fig. 3. Proliferating cells were detected by Ki67 staining (red), and apoptotic cells were detected by TUNEL staining (red). 〉ar, 100 m.
were detected in both dermal fibroblasts and epidermal keratinocytes. Perlecan from dermal fibroblasts, however, was not required for formation of the epidermis because the epidermis developed normally in skin equivalents constructed with perlecan null fibroblasts (either mouse or human) and primary human keratinocytes. By contrast, the epidermis was thin and poorly differentiated in SEs constructed with perlecandeficient keratinocytes and primary human dermal fibroblasts. Apparently, perlecan deficiency in the keratinocytes had no effect on their proliferation rate in monocultures or in SEs nor on their ability to express the early differentiation markers keratins K1 and K10 (see Figs.  3 and 4). These cells, however, failed to express intermediate and late differentiation markers such as involucrin and loricrin. These findings suggest that perlecan-deficient keratinocytes became committed to terminal differentiation but failed to complete their stratification program.
In addition to the defect in differentiation, perlecan deficiency in keratinocytes had a dramatic effect on their survival in the SEs. In normal skin, at the last step of their differentiation, keratinocytes undergo apoptosis as they reach the stratum corneum layers. Programmed cell death associated with terminal differentiation of keratinocytes is considered a special type of apoptosis. In typical apoptosis, the cells round up, shrink, and are eventually phagocytosed. Terminally differentiated keratinocytes, however, become bigger, flattened, and are ultimately shed when reaching the stratum corneum (69). In the epidermis formed by control keratinocytes, as expected, apoptotic cells exhibited this morphology and were mostly confined to the cornified layer. In SEs constructed with PDKs, nearly half of the keratinocytes were apoptotic, and their morphology was similar to that observed in typical apoptosis. Obviously, as PDKs do not complete their stratification program, apoptotic cells typical of terminally differentiated keratinocytes were absent from these SEs. Because the proliferation rate of PDKs in SEs was similar to that of the parental keratinocytes, their failure to form multilayered epidermis must be due to the apparent increase in their apoptotic rate, which depleted the pool of cells required for epidermis regeneration. The ability of FGF7 to restore thickness of perlecan-deficient epidermis, by preventing PDK apoptosis, is in line with this conclusion. Taken together, our findings revealed that perlecan from keratinocytes is critical for normal epidermal formation and that keratinocyte-derived perlecan is required for keratinocyte survival and terminal differentiation. Our findings and recent reports from other groups emphasize a key role for perlecan in regulating cell survival in different tissues. For example, perlecan heparan sulfate deficiency induced apoptosis of lens epithelial cells. Similar to the PDK, these cells exhibited a normal proliferation rate (70). In addition, domain V of the perlecan core protein was reported to protect vascular smooth muscle cells from apoptosis in vitro (71). Unlike the present findings, gross changes in epidermis morphology were not observed in perlecan null mice. This may be due to functional compensation by other BM components (20). This possibility is supported by recent reports where transgenic mice engineered to express perlecan core alone survived and displayed defects that were not observed in mice lacking the entire perlecan product. These mice also exhibited defects in wound repair (70,72). Another possibility is that perlecan may not fulfill a similar function in human and mice skin. It is well known that murine and human skin differ in several aspects, including thickness, sensitivity to carcinogens, and susceptibility to proapoptotic agents (36,39). Therefore, they may also differ in individual HSPGs requirements for epidermal formation.
Different studies suggested that perlecan protein is present in the BMZ and dermis in human skin (19,60,73,74). This staining pattern is identical to that observed in our control SEs. However, Murdoch et al. (19) could not detect perlecan transcripts in human epidermis and therefore concluded that perlecan in the BMZ is derived from the dermal fibroblasts. The present findings clearly show that epidermal keratinocytes express perlecan transcripts and are capable of producing and secreting perlecan product. This conclusion is based on several lines of evidence. 1) Perlecan transcripts were specifically detected in epidermal keratinocytes by in situ hybridization in both native human skin and in SEs. 2) RT-PCR analysis confirmed the expression of perlecan transcripts in keratinocytes. 3) Perlecan protein is produced and secreted from human keratinocytes grown in monocultures and in SEs. Hence, perlecan protein was readily detected in BMZ of SEs engineered with primary human keratinocytes and perlecan null fibroblasts. Because in these SEs the keratinocytes are the only source for perlecan, our findings unequivocally prove that human keratinocytes can produce and secrete perlecan. At present it is not clear why epidermally derived perlecan protein can be detected in the BMZ but not in the epidermal cells that produce it. A possible explanation is that the epitope recognized by the antibody is not accessible in the intracellular perlecan. Alternatively, perlecan synthesis and secretion into the BMZ may be tightly linked, and the amount remaining inside the cells is below the threshold for detection by the antibody.
In perlecan-deficient epidermis, the staining pattern of major BM constituents, such as collagen IV and laminin, was similar to that observed in SEs with control clones, suggesting that perlecan accumulation in the BMZ is not essential for proper assembly of the BM. These findings are in agreement with reports that in perlecan null mice most BMs appeared normal (24,25). In addition, our study suggests that perlecan may not be required for adhesion of basal keratinocytes to the BM. Apparently the levels and distribution of ␤1 integrin were not affected in perlecan-deficient epidermis. Moreover, we did not observe an increased tendency for splitting at the epidermis/collagen interphase, further indicating that the epithelial-matrix junctions have not been weakened in the perlecan-deficient BM. ␤1 integrin was implicated in cell attachment to perlecan (33), as well as to other BM components such as collagen IV and laminin (6,73,75). These additional interactions may explain why perlecan deficiency had no apparent effect on ␤1 integrin distribution. Alternatively, perlecan may not interact with ␤1 integrin in the skin. Altogether, these findings suggest that the defect in epidermal formation by PDK does not result from disruption of the BM or keratinocyte-matrix adhesion.
The present findings highlight a strong correlation between the presence of perlecan in the BMZ and normal formation of the epidermis. This conclusion is supported by the following observations. 1) In SEs generated with primary keratinocytes and perlecan null fibroblasts, where epidermal formation was normal, perlecan was readily detected in the BMZ of fully differentiated epidermis. 2) In PDK-SEs cultivated for up to 4 weeks, where epidermal formation was clearly abnormal, perlecan staining was absent from the BMZ. 3) In SEs grown for 5-6 weeks, the morphology of the perlecan-deficient epidermis was significantly improved with time, concomitant with the appearance of perlecan in the BMZ. The keratinocytes in these cultures resumed the expression of late differentiation markers, and their apoptotic rate was reduced to normal. In a previous study, where a nidogen-binding laminin fragment (L␥F) was added to human SEs, perlecan was not observed in the BMZ of long term cultures, and epidermis morphology appeared normal (76). In this study, keratinocytes that are fully capable of producing perlecan were utilized. Moreover, the laminin fragment was added after stratification has begun, and several consecutive applications of the fragment, for a period of 15 days, were required to displace perlecan from the BM. We assume that the amount of perlecan secreted by the keratinocytes, prior to addition of the fragment, and until perle-can was displaced from the BM, was sufficient to fully support epidermal formation. This suggestion is reinforced by the present finding that epidermal formation from PDKs can be fully restored by relatively small amounts of exogenous perlecan added for only 2 days and prior to induction of differentiation. In addition, recent studies indicate that epidermal formation is impaired if the laminin fragment is first applied to SEs before the onset of terminal differentiation. 5 Thus, the lack of an effect on formation of the epidermis in the previous study is because of the timing at which the laminin fragment was added.
HSPGs enhance the formation of growth factor-receptor complexes and subsequently increase cellular responses to heparin-binding growth factors (9). HSPGs of the ECM sequester heparin-binding growth factors and as a consequence regulate their local concentration and protect them from inactivation (62,63). It is evident from the present results that perlecan does not regulate the response of keratinocytes to endogenous mitogens because PDKs proliferated normally in the SEs. In PDK monocultures as well, perlecan was not required for cellular responses to FGF7. In this respect, keratinocytes differ from colon carcinoma cells that do not respond to FGF7 in the absence of perlecan (77). This apparent difference may be due to the tissue-specific nature of HSPG functions (70). The strong correlation between the presence of perlecan in the BMZ and proper epidermal formation suggests that perlecan may function as a reservoir for soluble factors involved in the survival and differentiation of keratinocytes. The levels or biological activities of such factors may therefore be reduced in the absence of perlecan from the BMZ. This can explain why exogenous FGF7 protected perlecan-deficient epidermal cells from apoptosis. It remains to be determined whether perlecan directly modulates the bioavailability of endogenous FGF7. The high affinity of perlecan to FGF7 (28), the previous observations that depletion of endogenous FGF7 from SEs results in the formation of thin epidermis, and the known capacity of FGF7 as an epithelial cell survival factor support this possibility (28,58,64,65). The inability of exogenous FGF7 to restore full differentiation capacity of PDKs further suggests that perlecan modulates additional factor(s) required for keratinocytes to complete their differentiation program. A stimulatory molecule whose activity is perturbed in the absence of perlecan or an inhibitory molecule whose expression or activity is enhanced in the absence of perlecan could be such a factor. Studies are underway to address this question.
In summary, by utilizing in vitro engineered human skin we demonstrated for the first time that perlecan is essential for the morphogenesis of human epidermis, controlling the survival and differentiation of human keratinocytes. Our data suggest a model whereby perlecan controls these processes by regulating the bioavailability of survival and differentiation factors involved in epidermal formation. The system described here can lead to the identification of such factors and to a better understanding of the mechanisms that control the differentiation and survival processes in human epidermis.