Materials and Methods

Cell culture

Primary keratinocytes and fibroblasts were cultured from neonatal foreskin, as described previously (^{Normand & Karasek 1992}). Briefly, newborn foreskin was trimmed, cut into small pieces, and placed in dispase at 4°C overnight. The dispase-treated foreskin was transferred into new Petri dishes and the dispase was neutralized with quenching buffer (containing 10% fetal bovine serum). Epidermis was physically separated from dermis with forceps and treated with 0.3% trypsin at 37°C for 30 min. The trypsin was neutralized with quenching buffer. The epidermis was minced with surgical forceps and the tissue debris was removed by passing the suspension through a metal mesh. The detached keratinocytes were collected by spinning in a clinical centrifuge. The keratinocyte pellet was resuspended in 1:1 serum-free medium (SFM; Gibco BRL, Grand Island, NY) and KMK (Sigma, St Louis, MO) medium, and plated on collagen-coated dishes. The keratinocytes were cultured for two to three passages in a 37°C incubator with 10% CO₂. Primary fibroblast cultures were obtained from the remaining dermis by plating the separated dermis in six well plates in the presence of Dulbecco's modified Eagle's medium (Gibco BRL) with 10% fetal bovine serum for approximately 1 wk. The cells that grew away from the dermal explants were fibroblasts.

Grafting the keratinocyte sheet composite onto SCID mice

Composite graft samples were obtained from P. Khavari's laboratory, Department of Dermatology, Stanford University. Briefly, normal keratinocytes obtained from foreskin biopsies were passaged in tissue culture and then seeded on acellular, devitalized human dermis. After 5 d of growth in tissue culture the composite was grafted directly onto mouse muscle fascia, as described by^{Medalie & Eming (1996)}.

Grafting the cell-sorted skin equivalent (CeSSE) onto SCID mice

A mixed cell slurry containing approximately 6 10⁶ keratinocytes and (6–8) 10⁶ dermal fibroblasts was prepared for each mouse to be grafted. Typically two 100 mm dishes of confluent keratinocytes and four dishes of confluent dermal fibroblasts yielded the needed numbers of cells. Keratinocytes were trypsinized in 0.1% trypsin, and fibroblasts in 0.25% trypsin, and were later neutralized with phosphate-buffered saline (PBS) with 10% fetal bovine serum. The two cell types were mixed in SFM (Gibco), placed in a 15 ml polystyrene conical centrifuge tube (Falcon, Becton Dickinson, Franklin Lakes, NJ), and centrifuged at low speed in a clinical centrifuge for approximately 5 min. Excess medium was removed by aspiration, and the cell pellets were stored on ice until use. Silicone chambers (29–31) implanted on the backs of SCID mice were used as in vivo chambers for the development of reconstituted human skin. Briefly, after anesthetizing the host mice, a circle 1 cm in diameter of upper back skin was removed using curved surgical scissors. The brim and hat of the silicone chamber (CRD culture chambers, Renner, Darmstadt, Germany) were combined and placed under the edge of the skin around the perimeter of the surgical wound incision, and sutured in place. The mixed keratinocyte-fibroblast cell suspension was transferred into the chamber directly onto the mouse muscle fascia through the 3 mm hole in the crown of the silicone hat. In the days following grafting wound fluid built up in the chamber. After 1 wk the silicone hat was removed and the wound was allowed to dry for an additional week before biopsies were taken.

Immunohistochemistry

To examine the CeSSEs cryosections were prepared. Briefly, fixation was with -20°C acetone for 10 min. Samples were rehydrated with five successive PBS washes. Blocking was conducted with mouse IgG diluted 1:400 (Jackson ImmunoResearch, West Grove, PA). A biotin/avidin-peroxidase conjugation system was used. Fifty microliters of an appropriate dilution of the primary antibody were incubated with the sample for 1 h at room temperature, followed by three washes with PBS. The sources of primary antibodies were as follows: antifilaggrin (Biomedical Technologies, Stoughton, MA); antikeratin 10, MAS 445p clone LH1 (Harlan Sera-Lab, Sussex, U.K.); anticytokeratin, MAS 635p clone LL002 (Harlan Sera-Lab); antilaminin-5 3 chain, K-140 (gift from Peter Marinkovich); anticollagen VII, LH7:2 (gift from Irene Leigh); antivimentin (cloneV9, Dako, Glostrup, Denmark); antipropyl-4-hydroxylase, 5B5 (Dako). The secondary antibody was antimouse Ig, horseradish peroxidase (Amersham, Bucks, U.K.) and was incubated with the samples for 40 min. After three washes with PBS the samples were developed with an insoluble peroxidase substrate (Sigma, St Louis, MO) for 20–30 min. For immunofluorescence detection of laminin-5 cryosections were prepared and subjected to a fluorescein isothiocyanate-labeled secondary antibody staining method. In brief, after blocking with goat serum for 20 min at 37°C, the sections were incubated with primary antibody for 1 h at 37°C. Slides were washed between each step in three shifts of PBS for a total of 15 min. All incubations were conducted for 30 min at room temperature, except where otherwise stated above. The slides were lightly counterstained with hematoxylin, dehydrated, and mounted. Negative controls consisted of a nonimmune rabbit IgG applied to adjacent sections at the same concentration.

Retrovirus transfection

The -galactosidase expressing retrovirus (^{Kinsella & Nolan 1996}) was used to infect approximately 6 10⁶ keratinocytes or fibroblasts. The cells were prewashed with SFM (10 ml) with 5 g polybrene per ml for 5 min and then aspirated from the plates prior to infection. Viral supernatant was diluted 1:1 with SFM and contained a final concentration of 5 g polybrene per ml. Plates were centrifuged at 750g for 1 h at 32°C. The plates were incubated overnight at 37°C, and the cells were then used in the CeSSE. -galactosidase staining was conducted 2 wk after the cells were seeded into chambers placed on the backs of SCID mice.

Electron microscopy

Fresh biopsies of normal human skin (breast biopsy), composite graft grown on SCID mice for 3 wk, and CeSSE also grown on SCID mice for 3 wk were prepared for electron microscopy by fixation in 2% paraformaldehyde, 2.5% glutaraldehyde, and 0.1 M cacodylate buffer, pH 7.4. Samples were treated with 2% osmium tetroxide and 2% uranyl acetate, dehydrated, and embedded in epoxy resin as described by^{Hyatt (1986)}.

Results

Reconstitution of full-thickness human skin on SCID mice

We sought to reconstitute full-thickness human skin without artificially sandwiching preformed epidermal and dermal layers together, and found that full-thickness skin could be reconstituted under in vivo conditions by the addition of a mixed cell slurry of primary keratinocytes and dermal fibroblasts placed on a SCID mouse animal host. The cell slurry was added to a commercially available inert silicone bubble chamber (CRD culture chambers, Renner) implanted directly on the mouse muscle fascia, as shown in Figure 1. The chamber contains a brim, or lower chamber, whose primary function is to contain the human cell slurry and prevent overgrowth of mouse tissue onto the graft area. The second component is the hat, or upper chamber, that creates the moist environment required for the survival of the added human skin cells.

Figure 1.

The silicone chamber (CRD culture chambers, Renner) used to contain a mixed cell slurry of keratinocytes and dermal fibroblasts. The lower chamber is placed directly on SCID mouse muscle fascia, after a circular piece of mouse skin has been surgically removed, and serves as a physical barrier to invading mouse skin during wound healing. The upper chamber is placed over the lower chamber and serves to keep other material out of the wound, as well as to maintain humidity. The mixed cell slurry is added by pipette through a hole in the top of the upper chamber.

Full figure and legend (14K)

Approximately 6 10⁶ primary human keratinocytes and 6 10⁶ primary human fibroblasts were added as a thoroughly mixed cell slurry through the hole in the top chamber implanted on a SCID mouse (Figure 2a). This single addition of cells is the sole preparatory step required for a human skin reconstitution using the CeSSE model. Because of the accumulation of wound fluid in the chambers, however, the top part of the chamber (the hat) was removed after 7 d to allow the reconstituted skin to dry. This step is consistent with earlier raft models, where exposure of keratinocytes to an air–liquid interface was a critical step in inducing keratino- cytes to differentiate normally into epidermal layers (^{Green et al. 1979;Bell et al. 1983}). A panel of antibodies specific for standard markers of both dermal fibroblasts and keratinocytes was used. In most cases these reagents were human specific and would not recognize mouse cell antigens, demonstrating that human cells, not mouse cells, are forming the reconstituted skin.

Figure 2.

Reconstituted human skin from the CeSSE model created in situ in silicone chambers on the backs of SCID mice. Complete separation of human keratinocytes and fibroblasts into discrete epidermal and dermal layers from a mixed cell slurry after 14 d. Scale bars in (b)-(f) indicate the extent of epidermal keratinocytes (white) and dermal fibroblasts (black). The region between the bars defines the BMZ, the normal interface of the two skin layers. (a) Implantation of human skin cells pipetted as a cell slurry into a silicone chamber implanted on the back of a SCID mouse. (b) Hematoxylin and eosin staining of a biopsy of the human CeSSE: upper epidermis stains purple, lower dermis stains a lighter violet. Flaking layers at the top of the epidermis are dead differentiated squames, as would be seen in normal skin. (c) Filaggrin antibody immunostaining of a CeSSE. Upper layer differentiated keratinocytes (arrow) express filaggrin, and therefore are stained brown. Biopsies were cryosectioned and detected with a perioxidase linked second antibody in this and subsequent figures (Experimental protocol). (d) Keratin 10 antibody immunostaining of a CeSSE. All keratinocytes express keratin 10 and are stained brown, except the single layer of basal keratinocytes along the DEJ, which remained purple. (e) Keratin 14 antibody immunostaining of a CeSSE. Only basal keratinocytes express keratin 14, and therefore stain dark brown. (f) Laminin-5 antibody immunostaining of a CeSSE. Laminin-5 is a component of hemidesmosomes expressed by basal keratinocytes along the BMZ, and therefore dark brown staining is limited to the BMZ, as indicated by arrows.

Full figure and legend (196K)

Normal morphology of human reconstituted skin

Hematoxylin and eosin staining of a paraffin-embedded thin section (5 m) from a biopsy taken at 2 wk showed that the morphology of the reconstituted skin is very similar to that of normal skin (Figure 2b), including an epidermal top layer and a dermal bottom layer. The epidermal architecture appears normal and includes a basal layer, stratum spinosum, stratum granulosum containing keratohyalin granules, and stratum corneum, seen as flaking layers of orthokeratotic cells at the top. The DEJ is very clean, with the dermal fibroblast population distinctly separated from the keratinocytes in the epidermis. The dermis is more tightly packed with fibroblasts than is seen in the morphology of normal skin, which is characterized by more sparsely distributed fibroblasts in an extensive extracellular matrix. As the extracellular matrix is slowly deposited by fibroblasts over time, we speculate that with increased reconstitution time greater deposition of extracellular matrix would occur. The fact that a mixed population of keratinocytes and fibroblasts added as a slurry to the chambers can cleanly separate into two separate populations within 2 wk shows that these cells have the necessary signaling and motility to migrate apart, thereby demonstrating a phenomenon known as cell sorting. The formation of two distinct layers suggests that cell sorting of human skin cells might be harnessed to create full-thickness skin for grafting.

Immunohistochemistry was conducted with antibody reagents specific for keratinocyte markers both to document the normal differentiation of the epidermis in the reconstituted model and to show that the keratinocytes are human derived, and not murine in origin. A range of epithelial phenotypes is displayed by keratinocytes as they terminally differentiate, with concomitant changes in gene expression patterns. For example, peroxidase staining using an antibody specific for human filaggrin, present predominantly in the stratum granulosum, serves as a late stage differentiation marker for keratinocytes (Figure 2c). Characteristic brown staining is seen in the upper layer of keratinocytes, representing human keratinocytes (not mouse) in the later stages of differentiation. The lower layers closer to the basement membrane zone (BMZ) are not stained brown by the peroxidase but remain purple, indicating that they do not express filaggrin, a pattern characteristic of normal skin.

Likewise, keratins can be used to document changes in differentiation of keratinocytes as they move to the outer layers. For example, genes encoding basal cell keratins 5 and 14 are switched off whereas those encoding differentiation-specific keratins 1 and 10 are induced. Therefore, a monoclonal antibody against keratin 10 stained the differentiated epithelial layers (most of the area within the white bar) (Figure 2d), with a single cell layer of unstained keratinocytes remaining along the BMZ (arrow, and note the violet epidermal cells between the black and white bars), as is also characteristic of a normal differentiated epidermis. Conversely, a single layer of keratinocytes along the BMZ would be expected to stain with keratin 14 antibody, as keratin 14 is made exclusively by basal keratinocytes. Staining with keratin 14 antibody did, in fact, show dark brown staining along the single layer of keratinocytes along the BMZ (Figure 2e, arrow). Thus, these markers suggest that normal differentiation is occurring in the epidermis of the reconstituted skin, and that the cells are human.

To establish that the DEJ of the reconstitution indeed expresses components unique to the BMZ, we used a monoclonal antibody that recognizes the human laminin-5 3 chain, but has no cross-reactivity with the mouse protein (Figure 2f). Laminin-5 is a component of the hemidesmosomes that attach basal keratinocytes to the BMZ, and is expressed uniquely by basal keratinocytes. In immunochemistry experiments dark brown staining was found along the BMZ, confirming normal localization of laminin-5 to the BMZ in reconstituted skin.

To document that the reconstituted dermis is composed of human fibroblasts, and that these cells also express normal markers, thin sections were stained with antibodies specific for dermal fibroblasts. Immunoperoxidase staining was conducted using the monoclonal antibody LH 7:2 that recognizes human collagen VII (^{Leigh & Purkis 1985}) (Figure 3a). Normally, collagen VII deposition occurs just below the BMZ, and is considered to be contained at the top of the dermis, although it is expressed by both keratinocytes and fibroblasts. Collagen VII forms the anchoring fibrils that serve as dermal attachments for hemidesmosomal components. The absence of normal collagen VII, which can occur in rare cases of a primary genetic lesion in the collagen VII genes, is responsible for the chronic blistering disease dystrophic epidermolysis bullosa (^{Fine et al. 1991}). The observed staining near the DEJ reflects the normal expression pattern. Significantly, others have observed a severe decline in anchoring fibrils, made of collagen VII, when epidermal autografts were grown on burn patients, and have postulated that the observed skin fragility in these patients is the direct result of this defect (^{Woodley et al. 1988}). Although the immunohistochemistry does not indicate whether the collagen VII is organized into anchoring fibrils, its correct localization to the BMZ is indicative of proper formation in the CeSSE.

Figure 3.

Reconstituted human skin after 14 d immunostained for markers of human dermal fibroblasts. Scale bars on the right of each figure indicate the extent of epidermal keratinocytes (white) and dermal fibroblasts (black). The grey bar indicates the region of mouse fibroblasts. (a) Collagen VII antibody immunostaining. Collagen VII is expressed primarily by dermal fibroblasts and is localized to the upper dermis along the BMZ, and therefore brown staining is along the BMZ as indicated by arrows. (b) Vimentin antibody immunostaining. Vimentin is uniformly expressed by dermal fibroblasts, and therefore brown staining is seen in the entire dermal layer. (c) Human fibroblast specific monoclonal antibody 5B5 immunostaining. This antibody recognizes an enzyme intermediate in collagen synthesis, propyl-4-hydroxylase, and recognizes only human, but not mouse, fibroblasts. Brown staining seen in reconstituted dermis. In this particular reconstitution fibroblasts of mouse origin are present below the human dermis, but do not stain with 5B5.

Full figure and legend (115K)

A marker uniformly expressed by dermal fibroblasts is vimentin. Staining using an antibody to this marker showed strong and ubiquitous expression in the reconstituted dermis (Figure 3b). Another monoclonal antibody specific to human fibroblasts is 5B5, which recognizes the beta subunit of propyl-4-hydroxylase, an enzyme required for collagen synthesis. 5B5 stained the reconstituted dermis, indicating that a human dermal layer was present beneath the reconstituted epidermis. This particular section also demonstrates a lack of staining in another cell layer, located directly below the reconstituted dermis and probably composed of mouse fibroblasts (Figure 3c, indicated by the stippled bar). The presence of this nonstaining layer was variable in different reconstitutions, and was often absent (data not shown). We conclude from our reproducible immunohistochemistry experiments that the CeSSE human skin reconstitutions are capable of recapitulating many aspects of normal full-thickness human skin.

Targeting expression to epidermal or dermal layers

In a different set of experiments, a -galactosidase expressing retrovirus (^{Kinsella & Nolan 1996}) was used to infect the cells of either the epidermal component or the dermal component used in the CeSSE (Figure 4a). Primary fibroblasts or primary keratinocytes were infected with an amphotrophic retrovirus produced in the Phoenix helper cell line jNX-A (^{Kinsella & Nolan 1996}) and were used to reconstitute human skin. -galactosidase staining is shown 2 wk after the cells were seeded into the chambers placed on the backs of SCID mice. As predicted, -galactosidase staining was confined to dermal (Figure 4b) or epidermal (Figure 4c) layers when fibroblasts or keratinocytes, respectively, were infected. These experiments again confirm that human cells reconstitute these skin samples, rather than mouse cells, as only the human cells were infected with retrovirus expressing -galactosidase. This highlights another useful aspect of this model, as targeted gene introduction can be conducted when the component cells are still in tissue culture, an advantage for the efficient introduction of any transgene.

Figure 4.

Retroviral -galactosidase expression vector targeted to keratinocytes or fibroblasts, and used in the CeSSE model to reconstitute human skin. (a) Diagram of retroviral expression vector used to express -galactosidase. Long-terminal repeat (LTR) contains promoter used to drive expression (bent arrow) of inserted -galactosidase gene, and + indicates retroviral packaging sequence (^{Kinsella 1996}). (b) Retroviral -galactosidase expression vector was used to infect dermal fibroblasts in tissue culture, and subsequently used in the CeSSE model to reconstitute human skin. Blue staining is limited to the dermis. (c) Retroviral -galactosidase expression vector was used to infect keratinocytes in tissue culture, and subsequently used in the CeSSE model to reconstitute human skin. Blue staining is limited to the epidermis.

Full figure and legend (61K)

Comparison of CeSSE and composite models by electron microscopy

A direct comparison of the standard composite model with the CeSSE and normal skin was conducted by transmission electron microscopy (Figure 5). Under low power examination of normal skin a complex network of keratin intermediate filaments is visible within basal keratinocytes along the DEJ (Figure 5a, arrows). Also conspicuous is the largely acellular nature of the dermis (region below epidermis, indicated by black bar) containing numerous collagen and elastin filaments. By comparison, the standard composite model under low power shows few, if any, keratin intermediate filaments (Figure 5b, no arrows). Although we do not exclude the possibility of keratin intermediate filaments forming in the composite model, we have consistently observed a paucity of these structures. Further comparison with a CeSSE reconstitution shows that the basal keratinocytes in the epidermis of the CeSSE contain a more normal appearing network of keratin intermediate filaments (Figure 5c, arrows).

Figure 5.

Electron micrographs of basal keratinocytes and the BMZ from normal human skin, reconstituted human skin created by the standard composite model, or reconstituted human skin created by the cell-sorting method. (a) In this normal skin sample basal keratinocytes located along the BMZ form a well-developed network of keratin intermediate filaments as shown (arrows). (b) In this composite model human skin keratinocytes were seeded directly onto devitalized dermis, with the basal keratinocytes showing no clear presence of intermediate filaments (no arrows). (c) In the CeSSE model intermediate filaments are present (arrows). (d) In the normal skin sample at higher magnification the well-developed network of intermediate filaments (arrows) connect to hemidesomosomes, electron-dense ''buttons'' located along the BMZ, as seen in the regions labeled h. (e) In the composite skin sample at higher magnification no intermediate filaments are seen, but hemidesmosomes are visible, as seen in regions labeled h. (f) In the CeSSE skin sample at higher magnification intermediate filaments are seen (arrows) connecting to hemidesmosomes, as seen in regions labeled h. Scale bars: (a–c) 2.4 m; (d–f) 0.4 m.

Full figure and legend (261K)

Under high power examination of normal skin the network of keratin intermediate filaments (arrows) is seen to connect down to the hemidesmosomes (dark, electron-dense structures located along the DEJ, for example, just to the right of the labeled h in Figure 5d). In addition, fine threads termed anchoring fibrils located just beneath the hemidesmosomes are plentiful, and larger collagen filaments in the dermis are clearly evident. In contrast, high power examination of composite skin epidermis shows a lack of keratin intermediate filaments connected to clearly visible hemidesmosomes (located above and to the left of the labeled h in Figure 5e).

Two positive characteristics of dermis in the composite model, the serrated interface at the BMZ, visible as numerous interdigitations along the BMZ, and the overall normal appearance of the dermis, are evident in the electron micrographs. One possible problem, however, is seen in the lamina lucida, the clear region between the basal keratinocytes and the lamina densa (the darker line on the dermal side that parallels the edge of the keratinocyte lower membrane). The lamina lucida appears thicker than normal (clear region between white and black bars, extending diagonally across the micrograph, Figure 5e), possibly indicating reduced dermal-epidermal cohesion. Although the deep dermis retains the morphology of the cadaver dermis it is derived from (Figure 5b), under higher power the dermis just beneath the BMZ tends to lack clearly identifiable anchoring fibrils and collagen filaments (Figure 5e). One possibility is that the dermal collagen is partially degraded, in keeping with earlier observations that collagenase is induced in keratinocytes as they migrate across dermal collagen during wound healing (^{Saarialho-Kere et al. 1993}) and when keratinocytes are plated onto collagen matrices in tissue culture (^{Inoue et al. 1995}).

In the CeSSE the keratin intermediate filaments (arrows) are seen to connect down to hemidesmosomes located along the BMZ, as seen directly above the labeled h in Figure 5f. We note that the dermal layer consists of tightly packed fibroblasts in the CeSSE, in contrast to normal skin where the dermis is largely acellular. At the dermal–epidermal interface, under high power, clearly visible collagen filaments do develop, as seen by the long filaments just below the labeled h in Figure 5f. In addition, numerous anchoring fibrils were noted on the dermal side just below the hemidesmosomes, seen as thin threads just below the hemidesmosomes.

Discussion

The purpose in the most commonly used method in grafting, the composite model, is to recreate the two distinct skin layers by preforming each separately and then sandwiching them together. In contrast, we have found that a mixed population of keratinocytes and fibroblasts placed in an inert silicone chamber implanted on SCID mouse muscle fascia spontaneously cell sorts to form distinct dermal and epidermal layers, and thereby each achieves its own correct positioning. Our data indicate important advantages of harnessing cell sorting to recreate the two cell layers, especially the morphology of the DEJ.

We evaluated this novel cell-sorted model (CeSSE, pronounced ''cease'') by immunostaining for human cell specific markers of the differentiated epidermis, and for the dermis. Our results showed that filaggrin was expressed in only the more differentiated outer layer keratinocytes, whereas keratin 10 was expressed in all layers above the basal layer, as expected. The newly formed interface of the epidermal and dermal layers appeared to be normally formed as judged by the staining of keratin 14, which was confined to the basal layer, and the staining for laminin-5, a component of hemidesmosomes, localized to the interface. Likewise, markers for dermal components were also normally localized, with collagen VII staining specifically along the dermal side of the DEJ, and vimentin and propyl-4-hydroxylase present throughout the dermis. Correct positioning of keratinocytes to the epidermis and fibroblasts to the dermis was further confirmed by tagging the human grafted cells with a -galactosidase reporter gene and noting their localization to the correct cell layer.

Ultrastructure analyses by electron microscopy indicated that the composite model set up on SCID mice did not have a normal structure along the DEJ, whereas the CeSSE model did. First, keratin intermediate filaments, normally plentiful in the basal keratinocytes that line up along the DEJ, were greatly reduced in the composite model, especially filaments connecting to hemidesmosomes. The intermediate filaments are crucial in providing rigidity to basal keratinocytes to counter additional shear forces due to their location as the first keratinocyte layer at the DEJ.

One explanation for this deficiency is that the partial differentiation of keratinocytes needed to form a graftable sheet, as used in the composite model, results in a loss of basal characteristics. Thus, the grafted cells are not able to express normal levels of the specific proteins that make up the intermediate filaments, keratins 5 and 14. In contrast, keratinocyte populations used in the CeSSE are not exposed to differentiating reagents, such as high Ca²⁺ or serum concentrations, and thereby retain a more normal basal intermediate filament architecture (compare epidermis seen in Figure 5e,f.

The second problem seen along the DEJ in the composite model appears on the dermal side just below the lamina densa. At low power, the dermis in the composite model appears similar to that of normal skin simply because in this case actual cadaver dermis was used (compare Figure 5a,b). Upon careful examination of the interface at higher magnification, however, a cleared region is observed below the lamina densa with far fewer anchoring fibrils (Figure 5e). In normal skin a network of anchoring fibrils is seen connecting the lamina densa to collagen filaments (Figure 5d); with fewer numbers of these fibrils the composite skin would be prone to blister.

A reasonable explanation for this cleared region is that collagenase (and other matrix metalloproteinases) is induced when keratinocytes come into direct contact with collagen as in the composite model. Hence collagen IV in the lamina densa, collagen VII in the anchoring fibrils, and collagens I/III in the collagen filaments are potentially subject to degradation by collagenase (Figure 5e). This is in agreement with extensive work from other laboratories showing that keratinocytes cultured on collagen I and IV express significant levels of collagenase (^{Woodley et al. 1986;Peterson et al. 1990,1992;Sudbeck et al. 1994}). In addition, the clinical observation that autologous keratinocyte sheets grafted onto muscle fascia lack anchoring fibrils (^{Woodley et al. 1988}) may also be relevant to our results.

As the CeSSE is relatively simple in its design and implementation it is perhaps somewhat surprising that it was not harnessed earlier as a method of reconstituting human skin. In fact, earlier work on characterizing the interplay between different cell types leading to the formation of a fully differentiated epidermis, including hair follicle formation, was conducted using fractionated mouse cells. These studies helped to lay down many of the currently accepted principles concerning the role of dermal–epidermal interactions (^{Worst et al. 1982;Mackenzie & Fusenig 1983};^{Boukamp et al. 1985}). These studies differed from our study in several important aspects, however. First, we took advantage of the availability of SCID mice, which allow for the grafting of human skin cells onto mice without concerns about graft rejection; these were not available at the time of the earlier studies. Thus, we set out to specifically generate human full-thickness skin for future potential grafting applications, whereas the earlier studies were focused on the question of how mouse dermal cells may provide soluble factors needed for proper epithelial differentiation. Second, we exclusively utilized tissue culture passaged human primary keratinocytes and fibroblasts, as opposed to mixed cell fractions separated from disrupted mouse skin separated on Ficoll gradients. The gradients were a cruder method of cell type separation that gave rise to subpopulations of cells specifically relevant to the fur-bearing mouse, such as epidermal cells from the lower parts of hair follicles and interfollicular epidermal cells (^{Worst et al. 1982}). In these experiments a mixed population of keratinocytes and fibroblasts also referred to the initial layering of dermal cells on the wound before the addition of keratinocytes, and in no instance is the morphology of a mixed cell slurry shown. Third, the site of graft implantation was in preinduced granulation tissue deep in the panniculus carnosus in the mouse studies, whereas we grafted our mixed cell slurry directly onto fresh muscle fascia. The choice of graft site reflects a difference in the purpose of the studies, and has substantial effects on the result. Likewise, in the interpretation of the resulting tissue morphology, a correct orientation of the dermal and epidermal layers is not clear in an embedded graft, whereas we found significance in the normal orientation of dermis and epidermis in the CeSSE. It was also possible for us to remove the upper chamber after 1 wk to expose the reconstituted skin to an air–liquid interface, important in obtaining normal differentiation of the reconstituted epidermis by allowing drying out of the wound surface.

The CeSSE model can also be broadly applied to the study of various dermatologic diseases. For instance, accurate modeling of genetic skin diseases can be achieved by mixing normal cells with the patient cell type expressing a mutant gene. In cases where the defective cell type is not known, skin reconstitutions can shed light on the basic pathophysiology of the disease. In fact, the development of the CeSSE model originally stemmed from our interest in accurately modeling junctional epidermolysis bullosa (^{Matsui et al. 1995;Matsui et al. 1998}). Epidermolysis bullosa is a genetic blistering disease localized to the DEJ and can be caused by primary genetic lesions in any of the hemidesmosome components (^{Fine et al. 1991}).

In conclusion, the CeSSE recapitulates many aspects of the normal morphology of human skin, and has advantages over a comparable composite model also set up on SCID mice. As currently available skin grafts are variations of composite grafts, the problems we found with the composite model may also be underlying causes of the fragility of grafted human skin reconstitutions.

References

1. Archer, SB, Henke, A, Greenhalgh, DG, Warden:: GD. The use of sheet autografts to cover extensive burns in patients. J Burn Care Rehab 1998 19:33–38,
2. Bell, E, Ehrlich, HP, Buttle, DJ, Nakatsuji, TA: Living tissue formed in vitro and accepted as skin-equivalent tissue of full thickness. Science 1981 211:1042–1054,
3. Bell, E, Sher, S, Hull, B et al.: The reconstitution of living skin. J Invest Derm 1983 81:2S–10S,
4. Ben-Bassat, H, Eldad, A, Chaouat, M, Livoff, A, Ron, N, Ne'eman, Z, Wexler, MR: Structural and functional evaluation of modifications in the composite skin graft: cryopreserved dermis and cultured keratinocytes. Plast Reconstr Surg 1990 89:510–520,
5. Boukamp, P, Rupniak, HT, Fusenig, NE: Environmental modulation of the expression of differentiation and malignancy in six human squamous cell carcinoma cell lines. Cancer Res 1985 45:5582–5592,  | PubMed | ISI | ChemPort |
6. Boyce, ST, Greenhalgh, DG, Kagan, RJ et al.: Skin anatomy and antigen expression after burn wound closure with composite grafts of cultured skin cells and biopolymers. Plast Reconstr Surg 1993 9:632–641,
7. Burke, JF, Yannas, IV, Quinby, WC, Bondoc, Cc Jr, Jung, WK: Successful use of a physiologically acceptable artificial skin in the treatment of extensive burn injury. Ann Surg 1981 194:413–420,  | PubMed | ISI | ChemPort |
8. Cooper, M, Hansbrough, J: Use of a composite graft composed of cultured human keratinocytes and fibroblasts on a collagen-GAG skin matrix to cover full-thickness wounds on athymic mice. Surgery 1991 109:198–207,
9. Cuono, C, Langdon, R, McGuire, J: Use of cultured epidermal autografts and dermal allografts as skin replacement after burn injury. Lancet 1986 1:1123–1124,
10. Fine, JD, Bauer, EA, Briggaman, RA, Carter, DM, Eady, RAJ, Esterly, NB: Revised clinical and laboratory criteria for subtypes of inherited epidermolysis bullosa. J Am Acad Dermatol 1991 24:119–135,  | PubMed | ISI | ChemPort |
11. Fusenig, NE, Amer, SM, Boukamp, P, Worst, PK: Characteristics of chemically transformed mouse epidermal cells in vitro and in vivo. Bull Cancer 1978 65:271–280,  | PubMed | ISI | ChemPort |
12. Fusenig, NE, Breitkreutz, D, Dzarileva, RT, Boukamp, P, Bohnert, A, Tilgen, W: Growth and differentiation characteristics of transformed keratinocytes from mouse and human skin in vitro and in vivo. J Invest Dermatol 1983 81:168s–175s,
13. Green, H, Kehinde, O, Thomas, J: Growth of cultured human epidermal cells into multiple epithelia suitable for grafting. Proc Natl Acad Sci USA 1979 76:5665–5668,  | PubMed | ChemPort |
14. Hansbrough, JF, Boyce, ST, Cooper, ML, Foreman, TL: Burn wound closure with cultured autologous keratinocytes and fibroblasts attached to a collagen-glycosaminoglycan substrate. JAMA 1989 262:2125–2130,
15. Hyatt, MA: Basic Techniques for Transmission Electron Microscopy.1986 Orlando FL: Academic Press, pp. 56–68,
16. Inoue, M, Kratz, G, Haegerstrand, A, Stahle-Backdahl, M: Collagenase expression is rapidly induced in wound edge keratinocytes after acute injury in human skin; persists during healing and stops at re-epithelialization. J Invest Dermatol 1995 104:479–483,  | Article | PubMed | ISI | ChemPort |
17. Kinsella, TM, Nolan, GP: Episomal vectors rapidly and stably produce high-titer recombinant retrovirus. Hum Gene Ther 1996 7:1405–1413,  | PubMed | ISI | ChemPort |
18. Krejci, NC, Cuono, CB, Langdon, RC, McGuire, J: In: Vitro Reconstitution of Skin: Fibroblasts Facilitate Keratinocyte Growth and Differentiation on Acellular Reticular Dermis. J Invest Dermatol 1991.91:478–485,
19. Langdon, RC, Cuono, CB, Birchall, N, Madri, JA, Kukinska, E, Meguire, J, Moelmann. G. Reconstitution of structure and cell function in human skin grafts derived from cryopreserved allogenic dermis and autologous cultured keratinocytes. J Invest Dermatol 1988 91:478–485,
20. Leigh, IM, Purkis, PE: LH7: 2 a new monoclonal antibody to a lamina densa protein. J Invest Dermatol 1985 84:448–449,
21. Mackenzie, IC, Fusenig, NE: Regeneration of organized epithelial structure. J Invest Dermatol 1983 81:189s–194s,  | Article | PubMed | ChemPort |
22. Matouskova, E, Vogtova, D, Konigova, R: A recombined skin composed of human keratinocytes cultured on cell-free pig dermis. Burn 1993 19:118–123,
23. Matsui, C, Pereira, P, Wang, CK et al.: Extent of laminin-5 assembly and secretion affect junctional epidermolysis bullosa phenotype. J Exp Med 1998 187:1273–1283,  | Article | PubMed | ISI | ChemPort |
24. Matsui, C, Wang, CK, Nelson, CF, Bauer, EA, Hoeffler, WK: The assembly of laminin-5 subunits. J Biol Chem 1995 270:23496–23503,  | Article | PubMed | ISI | ChemPort |
25. Medalie, DA, Eming, SA, Tompkins, RG, Yarmush, ML, Krueger, GG, Morgan, JR: Evaluation of human skin reconstituted from composite grafts of cultured keratinocytes and human acellular dermis transplanted to athymic mice. J Invest Dermatol 1996 107:121–127,  | Article | PubMed | ISI | ChemPort |
26. Moscona, A, Moscona, H: Dissociation and aggregation of cells from organ rudiments of the early chick embryos. J Anat 1952 86:287–301,
27. Normand, J, Karasek, MA: A method for the isolation and serial propagation of keratinocytes, endothelial cells, and fibroblasts from a single punch biopsy of human skin. In Vitro Cellular Dev Biology 1992 31:447–455,
28. Odyssey, R: Multicenter experience with cultured epidermal autographs for treatment of burns. J Burn Care Rehabil 1992 13:174–180,
29. Peterson, MJ, Woodley, DT, Stricklin, GP, O'keefe, EJ: Enhanced synthesis of collagenase by human keratinocytes cultured on type I or type IV collagen. J Invest Dermatol 1990 94:341–346,
30. Peterson, MJ, Woodley, DT, Stricklin, GP, O'keefe, EJ: Synthesis and regulation of keratinocyte collagenase. Matrix 1992 1:192–197,
31. Philips, TJ, Gilchrest, BA: Clinical applications of cultured epithelium. Epith Cell Biol 1992 1:39–46,
32. Phillips, TJ: New skin for old. Arch Dermatol 1998 134:344–349,  | Article | PubMed | ISI | ChemPort |
33. Phillips, TJ, Bhawan, TJ, Leigh, IM, Baum, HJ, Gilchrest, BA: Cultured epidermal allografts: a study of differentiation and allograft survival. J Am Acad Dermatol 1990 23:189–198,
34. Saarialho-Kere, UK, Kovacs, SO, Pentland, AP, Olerud, JE, Welgus, HG, Parks, WC: Cell–matrix interactions modulate interstitial collagenase expression by human keratinocytes actively involved in wound healing. J Clin Invest 1993 92:2858–2866,  | PubMed | ChemPort |
35. Shimoyama, Y, Hirohashi, S, Hirano, S, Noguchi, M, Shimosato, Y, Takeichi, MM, Abe, O: Cadherin cell-adhesion molecules in human epithelial tissues and carcinomas. Cancer Res 1989 49:2128–2133,  | PubMed | ISI | ChemPort |
36. Sudbeck, BD, Parks, WC, Welgus, HG, Pentland, AP: Collagen-stimulated induction of keratinocyte collagenase is mediated via tyrosine kinase and protein kinase C activities. J Biol Chem 1994 269:30022–30029,  | PubMed | ISI | ChemPort |
37. Townes, PL, Holtfreter, J: Directed movements and selective adhesion of embryonic amphibian cells. J Exp Zool 1955 128:53–120,  | Article | ISI |
38. Weiss, P, Taylor, AC: Reconstitution of complete organs from single-cell suspensions of chick embryos in the advanced stages of differentiation. Proc Natl Acad Sci USA 1960 46:1177–1185,
39. Woodley, DT, Briggaman, RA, Herzog, SR, Meyers, AA, Peterson, HD, O'keefe, EJ: Characterization of ''neo-dermis'' formation beneath cultured human epidermal autografts transplanted on muscle fascia. J Invest Dermatol 1990 95:20–26,
40. Woodley, DT, Kalebec, T, Banes, AJ, Link, W, Prunieras, M, Liotta, L: Adult human keratinocytes migrating over nonviable dermal collagen produce collagenolytic enzymes that degrade type I and type IV collagen. J Invest Dermatol 1986 86:418–423,  | Article | PubMed | ISI | ChemPort |
41. Woodley, DT, Peterson, HD, Herzog, SR: et al.: Burn wounds resurfaced by cultured epidermal autografts show abnormal reconstitution of anchoring fibrils. JAMA 1988 259:2566–2571,  | Article | PubMed | ISI |
42. Worst, PKM, Mackenzie, IC, Fusenig, NE: Reformation of organized epidermal structure by transplantation of suspensions and cultures of epidermal and dermal cells. Cell Tissue Res 1982 225:65–77,  | Article | PubMed | ISI | ChemPort |

Acknowledgments

This work was supported by National Institutes of Health grant AR 41045–01. C.K. Wang was supported by a postdoctoral training grant from the National Institutes of Health We thank G.P. Nolan for providing the retroviral delivery system expressing -galactosidase, Irene Leigh for monoclonal antibody LH7:2, Peter Marinkovich for monoclonal antibody K-140, and Sabine Kohler for her expertise in electron microscopy of the skin.