Precancer or intraepithelial neoplasia (IEN) has been defined as “a noninvasive lesion that has genetic abnormalities, loss of cellular control functions and some phenotypic characteristics of invasive cancer and that predicts for a substantial likelihood of developing invasive cancer”.1 Animal models of skin carcinogenesis have revealed predictable stages of progression from IEN to invasive cancer, which result from the selective outgrowth of cells with neoplastic potential.2 It is now recognized that the selection and growth of transformed cells is likely to play a dominant role during the earliest stages of cancer progression, while increases in mutation rates are thought to be more common during the later stages of this disease.3 Premalignant cells need to be in the appropriate cellular microenvironment for such selective expansion and progression to occur,4 suggesting that genetic alterations in individual cells are necessary but not sufficient to promote the progression of IEN to invasive cancer. Since it is thought that selection of cells with neoplastic potential precedes the development of foci of morphologically altered, dysplastic cells,3 it essential to further understand the tissue dynamics and mechanisms which direct the selective outgrowth of tumor cells during the early, intraepithelial stage of cancer progression. However, microenvironmental factors that may enable the earliest events in the progression of IEN are not well understood.
It is known that clonal selection drives the emergence of transformed cells in monolayer tissue culture.3 However, conditions permissive for selection of potentially malignant cells in complex tissues with 3-dimensional (3D) architecture have yet to be characterized. For example, the role of microenvironmental factors, such as cell-cell and cell-stroma interactions, in the progression of potentially malignant epithelial cells remain to be elucidated5 and need to be studied in biological systems in which an optimal degree of tissue architecture can be achieved. Unlike their monolayer counterparts 3D, organotypic cultures present a means through which it is possible to mimic the architectural features of the in vivo tissue.6 Fabrication of these tissues provides a biologically meaningful context to study events during morphogenesis and early cancer progression.7, 8, 9, 10, 11 We have previously developed an organotypic, 3D human tissue model that mimics the tissue architecture seen in IEN, in which the fate and phenotype of small numbers of intraepithelial tumor cells could be monitored during the premalignant stage of disease.9, 10, 11 In these studies, we have used clones derived from the immortalized human keratinocyte cell line HaCaT,12 which have been transfected with an activated Ha-ras oncogene13 and exhibit characteristics of distinct and early stages of progression of premalignant and malignant epidermal lesions in 3-dimensional, organotypic cultures and surface transplants in nude mice.9 We have previously incorporated these cell lines into our tissue models of premalignant disease to demonstrate that neoplastic progression of these tumor cells was suppressed in vivo and in vitro by a mechanism mediated by cell-cell interactions.9 In these studies, a hallmark of the IE control of tumor cells was their active displacement from an extracellular matrix (ECM) substrate (Type I collagen gel) that did not contain basement membrane (BM) proteins that was accompanied by the induction of their growth arrest and terminal differentiation. However, it is not known how interactions between BM components and IE tumor cells at an early stage of transformation can contribute to progression of the earliest stages of IEN.
In the current study, we have found that BM proteins at the stromal interface can direct the selective attachment of intraepithelial tumor cells and thus enable their intraepithelial expansion in vitro to generate dysplastic, premalignant epithelium after transplantation in vivo. These proteins serve as a selective template for early cancer progression by providing a permissive environment for intraepithelial tumor cells to adhere to the dermal-epidermal interface, persist in a basal position and proliferate to dominate the tissue. Selective adhesion of II-4 cells was associated with the presence of pre-existing BM components, which normalized deposition of laminin 5 and rapidly organized structured BM. Purified laminin 1 and Type IV Collagen provided a permissive cue for the selection and persistence of IE tumor cells. Using novel, 3D tissue models that mimic human disease, we have demonstrated that potentially malignant cells that acquire the ability to adhere to BM can gain a selective growth advantage and an increased risk of progression to invasive cancer. In this way, the basement membrane microenvironment modulates the phenotype of IE tumor cells by enabling their selection for further cancer progression and by directing their escape from IE growth control.
MATERIAL AND METHODS
Monolayer cell culture
Normal human epidermal keratinocytes (NHK) were cultured from newborn foreskin by the method of Rheinwald and Green14 in keratinocyte medium described by Wu et al.15 Cultures were established through trypsinization of foreskin fragments and grown on irradiated 3T3 fibroblasts. 3T3 cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM) containing 10% bovine calf serum. The HaCaT-ras-II-4 (II-4) cell line16 was grown in DMEM containing 5% fetal calf serum human dermal fibroblasts used for organotypic cultures were derived from newborn foreskins and grown in media containing Dulbecco's Modified Eagle's Medium (DMEM) and 10% fetal calf serum.
Organotypic cultures without intact BM were prepared as previously described.9 Briefly, early passage human dermal fibroblasts were added to neutralized Type I collagen (Organogenesis, Canton, MA) to a final concentration of 2.5 × 104 cells/ml. Three milliliters of this mixture was added to each 35 mm well insert of a 6-well plate and incubated for 4 to 6 days in media containing DMEM and 10% fetal calf serum, until the collagen matrix exhibited no further shrinkage. At this time, a total of 5 × 105 NHK, II-4 cells or mixtures of these cells at a ratio previously shown to result in loss of attachment of II-4 cells to their ECM substrate (9;11) (4:1 ratio/NHK:II-4) were seeded on the contracted collagen gel. Cultures were maintained submerged in low calcium epidermal growth media (EGM) for 2 days, submerged for 2 days in normal calcium EGM and raised to the air-liquid interface by feeding from below with normal calcium cornification medium for 3 days. Organotypic cultures were grown in the presence of pre-existing BM components present on a de-epidermalized dermis derived from human skin (Alloderm™, LifeCell Corp., Branchburg, NJ), which was treated to remove the surface epithelium and stromal cells, while preserving the BM proteins Types IV and VII collagen and laminin 1, on its surface.17 Alloderm was layered on the contracted collagen gel described above to enable fibroblasts to migrate and repopulate the Alloderm from below, and keratinocytes were seeded 1 hr later and cultures were grown as described above. Polycarbonate membranes coated with the purified BM proteins laminin 1 and Type IV collagen or ECM components not found in BM (Fibronectin, Type I collagen or a mixture of Fibronectin and Type I collagen) of murine origin (Becton Dickinson, Billerica, MA) were placed on the contracted, fibroblast-containing collagen gels described above and NHK, II-4 or cell mixtures (4:1 ratio/NHK:II-4) were grown on these substrates as described for organotypic cultures. All organotypic cultures were performed in triplicate. A schematic description of these 3 organotypic culture models is summarized in Figure 1.
Retroviral vectors and transduction of tumor cell lines
The II-4 tumor cell lines were transduced with the MFG-gal vector, which is a moloney murine leukemia virus as previously described.18 Briefly, II-4 cells were transduced 24 hr after plating 1 × 106 cells in a 100 mm dish using fresh, filtered (0.45 μm, Gelman, Ann Arbor, MI) supernatant from confluent amphotropic cells producing the MFG-gal vector (gift of Dr. R. Mulligan). Transduced keratinocytes were passaged at clonal density, and clones were screened for persistence of transgene expression after 3 passages and after transplantation to nude mice. Only II-4 clones maintaining transgene expression in 100% of cells were expanded and used for organotypic culture and grafting experiments. No deleterious effect on cell growth or phenotype were seen after transduction. Retroviral producer lines were maintained in 10% bovine calf serum.
Transplantation of organotypic cultures to nude mice
Organotypic cultures were trimmed using a surgical punch 1.4 cm in diameter. Six-week-old male Swiss nude mice (N:NIHS-nuf DF, Taconic Farms, Germantown, NY) were anesthetized using Xylazine:Ketamine (3:2) and a 1.3 cm diameter of full-thickness dorsal skin was removed. Organotypic cultures were placed onto this area and were covered with petrolatum Gauze (Sherwood Pharmaceuticals, St. Louis, MO) and secured with bandages (Baxter Scientific Products, McGaw Park, IL). Dressings were changed after 7 days and removed completely after 14 days. Animals were grafted in triplicate and sacrificed at 8 weeks after transplantation.
Specimens were frozen in embedding media (Triangle Biomedical, Durham, NC) in liquid nitrogen vapors after being placed in 2 M sucrose for 2 hr at 4°C. Tissues were serial sectioned at 6 μm and mounted onto gelatin-chrome alum-coated slides. Tissue sections were washed with PBS and blocked with 10 μg/ml goat IgG, 0.05% goat serum, and 0.2% BSA, vol/vol in PBS without fixation. Sections were incubated with rabbit polyclonal antiserum to bacterial β-glactosidase (Cortex Pharmaceuticals, San Leandro,CA) and detected with Alexa 488™-conjugated goat anti-rabbit IgG (Molecular Probes, Eugene, OR). Double stain immunofluorescence was performed by staining for both β-gal detection and Type IV collagen, by using monoclonal antibodies directed against Type IV collagen (Sigma Chemical Co., St. Louis, MO) or BrdU (Sigma Chemical Co., St. Louis, MO), which were detected with Alexa 594™ conjugated goat anti-mouse IgG (Molecular Probes). Laminin 5 was detected using the GB-3 monoclonal antibody19 directed against the intact heterotrimeric molecule (gift of Dr. G. Meneguzzi). To detect S-phase cells, BrdU (Sigma Chemical Co., St. Louis, MO) was added to organotypic cultures 8h prior to harvesting cultures at a final concentration of 10 μM. Slides were coverslipped with Vectashield containing 1 μg/ml DAPI (Vector Laboratories, Burlingame, CA). Fluorescence was visualized using a Nikon OptiPhot microscope and double exposure photomicroscopy was performed using FITC and Texas Red filters. To extend the observations from immunohistochemical analysis, IE expansion of tumor cells was calculated as the percentage of the epithelial compartment which was β-gal positive in 5 tissue sections (NIH Image, Version 1.61). For routine light microscopy, tissues were fixed in 10% neutral buffered formalin, embedded in paraffin and 4 μm sections were stained with hematoxylin and eosin.
Organotypic cultures were homogenized with mortar and pestle in RIPA buffer supplemented with protease and phosphatase inhibitors to generate total cell lysates. Protein concentrations were quantified by DC Bradford reagent (Bio-Rad, Hercules, CA). To determine the relative levels of the 2 forms of γ2 chain of laminin 5, 15 mg of lysate was boiled in 2× Laemmli sample buffer, resolved on 7.5% sodium dodecyl sulfate/polyacrylamide electrophoresis (SDS-PAGE) gels and immunoblotted onto a PVDF membrane (Bio-Rad, Hercules, CA). The blot was blocked with TBST-5% dry milk, probed with the rabbit, polyclonal antibody SE-144 (20) (gift of Dr. G. Meneguzzi) directed against the γ2 chain of laminin 5.20, 21 and goat, polyclonal antibody directed against actin (Santa Cruz Biotechnologies, Santa Cruz, CA). Immunoreactive proteins were visualized by HRP-linked anti-rabbit (Amersham) and anti-goat (Santa Cruz Biotechnologies) secondary antibodies. Blots were developed in ECL (Amersham) and exposed to autoradiography. The images were then scanned using HP ScanJet 3300C and digitized using the UN-SCAN-IT software (Silk Scientific, Inc., Orem, UT).
Transmission electron microscopy
Organotypic cultures were cut into small pieces of approximately 2 × 2 mm and fixed in 2% Glutaraldehyde in 0.1 M Cacodylate and 0.1 M sucrose at pH 7.2. The samples were then post-fixed in 2% Osmium tetroxide in 0.1 M Cacodylate and 1% Tannic acid in 0.1 M Cacodylate. Samples were then dehydrated in graded ethanol, cleared with propylene oxide and infiltrated with Spurr's resin. Following polymerization of the resin, thick sections were produced using a Reichert Ultracut E microtome and sections were stained with Toluidine blue to determine orientation. The blocks were then thin sectioned at approximately 90 nm and mounted on copper grids. Grids were stained with 5% Uranyl acetate in deionized water and Reynold's lead citrate. Stained grids were examined at various magnifications using a Hitachi H-600 transmission electron microscope.
BM components enable early stage-transformed, IE tumor cells to form a dysplastic, premalignant lesion in vivo
Normal human keratinocytes (NHK) and low-grade IE tumor cells (II-4) were mixed at a 4:1 ratio (NHK:II-4) and grown in organotypic cultures in the presence or absence of pre-existing BM components. This mixing ratio was chosen to determine if the presence of BM components could rescue tumor cells from loss of ECM attachment and growth suppression seen when II-4 cells and NHK were grown in the absence of BM components in earlier studies.9, 11 Cultures were then grafted to nude mice and the clinical and histologic appearance of transplants was determined after 8 weeks (Fig. 2). When grown in organotypic culture on a Type I collagen substrate that lacked BM components or structures, transplants exhibited clinically and histologically (Fig. 2a) normal epidermal structures. In contrast, when 4:1 mixtures were grown in organotypic culture on AlloDerm in the presence of pre-existing BM components for 7 days and grafted to nude mice, a thickened, white lesion with a verrucous surface was seen at the graft site after 8 weeks. Histologic examination of this lesion demonstrated hyperparakeratosis, moderate to severe dysplasia characterized by disorganization of the basal and suprabasal layers and an irregular epithelial-stromal interface reminiscent of early invasive structures (Fig. 2b, arrows). These findings demonstrated that the presence of BM proteins was a permissive signal for the persistence of IE tumor cells, leading to the formation of dysplastic, premalignant lesions in vivo.
Attachment to BM components is a critical, permissive signal for IE tumor cell selection and expansion
In order to monitor the fate of IE tumor cells in organotypic cultures and transplants, the distribution of genetically-marked II-4 cells was determined by double immunofluorescence stain for β-gal and Type IV Collagen. Mixed organotypic cultures (NHK:II-4/4:1) that were grown on pre-existing BM components (AlloDerm) exhibited expanded clusters of β-gal-positive, II-4 cells that remained attached to the dermal-epidermal interface (Fig. 2d), which was demarcated by the linear distribution of Type IV collagen. These intraepithelial tumor cells persisted at the BM interface after grafting and underwent considerable expansion to occupy a significant proportion of the tissue (Fig. 2f, arrows). In contrast, mixtures grown in the absence of intact BM (Type I collagen gel), as seen by the focal deposition of Type IV collagen, demonstrated II-4 cells that did not attach to the dermal-epidermal interface and were sorted to the superficial layers of the epithelium in vitro (Fig. 2c, arrow). After grafting, the epidermis exhibited no β-gal-positive cells (Fig. 2e), demonstrating that IE tumor cells did not remain in the tissue but were desquamated when grown in vivo for 8 weeks in the absence of BM. These findings demonstrated that the selective attachment and persistence of II-4 cells required the presence of pre-existing BM components in order to generate premalignant lesions in vivo.
Pre-existing BM proteins direct the rapid assembly of BM and attachment of II-4 cells
In order to determine the structural basis for the selective attachment of II-4 cells seen in AlloDerm cultures, the ultrastructural features of the BM zone were studied by transmission electron microscopy when II-4 cells were grown in organotypic culture with and without pre-existing BM components (Fig. 3). BM assembly only occurred when II-4 cells were grown in the presence of pre-existing BM proteins (AlloDerm cultures). Lamina densa and lamina lucida were seen shortly (5 days) after II-4 cells were seeded onto the AlloDerm substrate (Fig. 3a, arrow). Although these structures were discontinuous, they were focally well formed (Fig. 3a, arrow, inset). Further maturation of BM structure was seen when II-4 cells were cultured on AlloDerm for 14 days by the presence of a continuous lamina densa and hemidesmosomes (Fig. 3b, inset, asterisk) consisting of inner and outer plaques and fine bridging structures on their extracellular surface (Fig. 3b, inset, white arrow). In contrast, no lamina densa, hemidesmosomes or other evidence of BM structure were seen when II-4 cells were grown on contracted Type 1 collagen gels in the absence of BM components for either 5 days (Fig. 3c) or 14 days (Fig. 3d). These results showed that BM organization could only occur at an interface that served as a structural template for its assembly. In this light, a microenvironment enriched for BM proteins enabled the rapid formation of structures that allowed II-4 cell attachment and persistence in a basal position.
Laminin 1 and Type IV collagen support attachment, basal retention and proliferative expansion of IE tumor cells
In order to determine which BM proteins or ECM component enabled the selective attachment and expansion of IE tumor cells that occurred on AlloDerm cultures, cell mixtures (NHK:II-4, 4:1) were grown on polycarbonate membranes coated with purified ECM proteins, layered on a contracted Type I collagen gel containing fibroblasts and grown in organotypic culture. Cultures were grown for 7 days and analyzed by double stain immunofluorescence for β-gal and BrdU to determine the distribution and growth of IE tumor cells. When cell mixtures were seeded onto membranes coated with fibronectin (Fig. 4b), a mixture of fibronectin and Type I collagen (Fig. 4c) or Type I collagen (Fig. 4d), II-4 cells did not attach to the substrate and were seen as single cells and small clusters in the middle to upper layers of the epithelium (arrows) in a pattern similar to that seen when mixtures were grown directly on a contracted Type 1 collagen gel (Fig. 4a). S-phase, BrdU-positive nuclei were not seen in these suprabasal II-4 cells and were limited to basal NHK. This demonstrated that tumor cells withdrew from cell cycle when they were not attached to the connective tissue substrate, as was previously described.9 In contrast, mixtures seeded onto membranes coated with laminin 1 (Fig. 4e) and Type IV collagen (Fig. 4f) demonstrated clusters of II-4 cells that adhered to these substrates and underwent a significantly greater degree of IE expansion. This expansion of II-4 cells was a consequence of the active proliferation of adherent cells, as seen by the colocalization of BrdU-positive nuclei in β-gal-positive cells (Fig. 4e,f). Similarly, mixtures grown on AlloDerm demonstrated similar attachment and proliferation of IE tumor cells (Fig. 4g). The degree of II-4 cell expansion was determined on each substrate by quantification of the percentage of β-gal stained area in numerous tissue sections (Fig. 5). Mixtures grown on laminin 1 (21.9%), Type IV collagen (27%) and AlloDerm (24.4%) demonstrated a 2–3 fold increase in the percentage of the total tissue area composed of β-gal-positive cells when compared to mixtures grown on other substrates (Fig. 5). This demonstrated that contact with laminin 1 or Type IV collagen-coated membranes rescued II-4 cells from the loss of attachment seen when these cells were grown on other ECM proteins in a manner similar to AlloDerm (Fig. 2). Thus, in a microenvironment enriched for BM proteins, such as laminin 1 and Type IV collagen, IE tumor cells could compete with NHK for basal attachment in a substrate-specific manner, thus creating permissive conditions for anchorage-dependent growth.
Pre-existing BM proteins (alloderm) and laminin 1 mediate attachment of II-4 cells and normalize deposition and organization of laminin 5
In order to determine how BM proteins and AlloDerm provided an appropriate microenvironment for II-4 cells to attach and persist in a basal position, pure cultures of NHK and II-4 cells were grown in organotypic culture on Type I collagen or laminin 1-coated polycarbonate membranes or AlloDerm or Type I collagen gels and the distribution, synthesis and proteolytic processing of laminin 5 were analyzed by immunohistochemical staining and immunoblot analysis. Since laminin 5-mediated adhesion is known to be an early event in keratinocyte attachment,22 we reasoned that the linear deposition of laminin 5 would be indicative of a permissive environment for II-4 cell attachment and a useful marker of this event. Cultures were grown for 8 days in organotypic culture and the distribution of the heterotrimeric form of laminin 5 was determined by staining with the GB-3 antibody. In the absence of BM proteins, II-4 cells grown on either contracted collagen gels (Fig. 6a) or on Type I collagen-coated membranes (Fig. 6f) did not deposit laminin 5 along the BMZ and expression was limited to the cytoplasm of suprabasal cells (Fig. 6, arrows). However, when NHK were grown on these substrates (Fig. 6a,e), laminin 5 deposition was restricted to the BM zone and no cytoplasmic or suprabasal expression was seen. This suggested that the inability of II-4 cells to normalize laminin 5 deposition at the BMZ in comparison to NHK, enabled NHK to preferentially attach to Type 1 collagen-coated substrates or gels when grown as cell mixtures. In contrast, when II-4 cells were grown in the presence of AlloDerm (Fig. 6d) or laminin 1 (Fig. 6h), laminin 5 deposition was seen in a linear and polarized distribution that was restricted to the BM zone. This showed that microenvironmental signals, such as those directed by the presence of BM proteins such as laminin 1, redirected the normalized, polarized deposition of laminin 5 at this interface by II-4 cells and prevented its ectopic expression in suprabasal layers. This linear pattern of laminin 5 deposition was similar to that seen when NHK were grown on AlloDerm (Fig. 6c) and laminin 1 (Fig. 6g). This showed that the ability of II-4 cells to normalize their laminin 5-mediated adhesion in a manner similar to NHK when grown in the presence of BM proteins had allowed the selective attachment and persistence of II-4 cells when these cells were mixed with NHK.
In order to determine if differences in laminin 5 synthesis or proteolytic processing could further explain the selective attachment of II-4 cells in mixtures cultured on BM proteins, immunoblot analysis was performed from cell extracts grown in organotypic culture. Relative amounts of the unprocessed (155 kDa) and processed (105 kDa) forms of the γ2 chain of laminin 5 were determined using the polyclonal antibody SE144, which recognizes the COOH-terminal domains of the γ2 chain.20, 21 As seen in Figure 7, both the total amount of laminin 5 synthesized and the relative amounts of the processed and unprocessed forms of the γ2 chain were similar for both NHK (Fig. 7, Lane 1) and II-4 cells (Fig. 7, Lane 3) grown on the Type 1 collagen gel. Although more synthesis and proteolytic processing occurred when these cell types were grown on AlloDerm, no differences in the relative amounts of the 2 forms of γ2 chain were seen when NHK (Fig. 7, Lane 2) and II-4 cells (Fig. 7, Lane 4) were compared. These findings demonstrated that factors other than the total amount of laminin 5 synthesized or the degree of laminin 5 processing, such as the rapid spatial organization of laminin 5 at the BMZ, were associated with the selective attachment of II-4 cells in the presence of AlloDerm or laminin 1 when grown in the presence of NHK.
Intraepithelial neoplasia (IEN) is a near-obligate precursor to invasive cancer and places patients manifesting this condition at a significantly elevated risk for developing squamous cell carcinoma.1 Development of IEN from phenotypically normal tissues to increasingly severe dysplasia is associated with the selective expansion of IE tumor cells and is followed by the migration of these cells through BM after its proteolytic modification.24, 25 While attachment to BM is a prerequisite for this to occur, the role of adhesive interactions between intraepithelial tumor cells and BM proteins during the earliest, intraepithelial stage of cancer progression has not been directly studied in human stratified squamous epithelium. We have developed novel, human tissue models of premalignant disease, which mimic the 3D architecture of this condition, to further understand interactions between dysplastic, intraepithelial tumor cells and their connective tissue interface at the BM zone. We found that the presence of pre-existing BM components at this interface promoted the earliest stages of neoplastic progression in this tissue. Specifically, we found that laminin 1 and Type IV collagen were permissive for the selective attachment and intraepithelial growth of early-stage, tumor cells. In the absence of adhesive interactions mediated by BM proteins, tumor cells did not persist in a premalignant tissue and were desquamated from the tissue. Thus, the epithelial-stromal interface is a critical microenvironmental determinant that can promote the selective attachment, persistence and expansion of initiated cells in IEN.
These findings show for the first time that premalignant cells are responsive to their environmental context and that BM-mediated adhesion directs their fate in IEN of stratified squamous epithelium. This supports previous studies in other tissue types, such as breast and prostate, showing that interactions between early-stage tumor cells and stromal factors could modulate early tumor progression. It was shown that the stromal microenvironment of early-stage tumor cells, defined by diffusible factors26 and structural components,27, 28 could modulate the phenotype of potentially-malignant epithelial cells. Integrin-mediated interactions between premalignant breast epithelium and adjacent BM proteins have been shown to restore anchorage-dependent adhesion and revert the malignant phenotype in a 3D model of early breast cancer progression.27, 28 It was also demonstrated that “initiated” tracheal epithelia required contact with extracellular matrix proteins to persist and expand when mixed with normal tracheal cells after inoculation onto denuded tracheas in vivo.29 Together with our findings, this demonstrates that in addition to the genetic alterations present in individual, initiated cells, multi-step carcinogenesis involves the environmental modulation of IE tumor cell behavior by factors in the basement membrane microenvironment.
Existing models used to study early tumor progression have been limited by the nature of monolayer culture in vitro and heterotypic transplantation in vivo, neither of which provide the proper tissue architecture to study the earliest events in tumor development as they occur in vivo. Biologically meaningful signaling pathways, mediated by the coupling of adhesion and growth, function optimally when cells are spatially organized in a 3D tissue and are uncoupled and lost in 2-dimensional culture systems.30 It is therefore essential that 3D cultures, which display the architectural features seen in in vivo tissues, be adapted to model and further understand the biological behavior of tumor cells in their appropriate tissue context.31 Since the earliest events in cancer progression in stratified squamous epithelium are manifested as a clone of dysplastic cells expands in the context of more normal cells, tissue models of IEN must incorporate this interrelationship. Our findings have been facilitated by the development of novel in vitro and in vivo tissue models for premalignant disease in which the fate and phenotype of small communities of IE tumor cells was monitored during the earliest stages of disease progression. Using these models, we previously determined that normal cell context and tissue architecture controlled clonal expansion and expression of the neoplastic phenotype of early-stage transformed cells9 and that this suppression could be overcome by altering interactions between IE tumor cells and cells adjacent to them.10, 11 Since cell growth and survival are restricted to matrix-attached keratinocytes in normal stratified epithelia,32, 33, 34 it is possible that anchorage-dependent growth regulation may control the degree of epithelial dysplasia35 and that loss of adhesion between early-stage tumor cells and ECM components can restrict cell growth.36
We have found that the polarization of laminin 5 and organization of well-structured BM were associated with the selective adhesion of IE tumor cells only in the context of pre-existing BM components. Laminin 5 is secreted by basal keratinocytes21, 37, 38 and polymerizes into a network that stabilizes the BM39 by bridging α6B4 integrin in hemidesmomes to Type VII collagen40 in anchoring filaments. Basal polarization of laminin 5 at the connective tissue interface is a very early step in epidermal morphogenesis and the normalization of tissue architecture.7, 20, 41, 42 We have determined that laminin 5 polarization at the stromal interface is a marker of IE tumor cell attachment and is required for the earliest stages of cancer progression. In the absence of pre-existing BM components, the distribution of laminin 5 was restricted to the cytoplasm of suprabasal II-4 cells, while the presence of BM components directed the deposition of laminin 5 to the dermal-epidermal interface. This substrate-dependent modulation of laminin 5 distribution reflects the environmental plasticity of these early stage tumor cells. These findings showed that the normalized deposition and organization of laminin 5 by II-4 cells was controlled and enhanced by a microenvironment enriched for BM proteins and provided the means through which IE tumor cells may selectively compete with their normal neighbors for basal attachment. When II-4 cells could not normalize laminin 5 organization, as occurred in the absence of BM proteins, these cells were displaced to a suprabasal position. In this light, laminin 5, which is synthesized only by the epithelium, may play an important role as a marker of early tumor cell attachment. Differences in the expression of the processed or unprocessed forms of laminin 5 have been seen in a variety of tumor types43 and decreased laminin 5 processing of the γ2 subunit has been associated with the lack of mature hemidesmosomes seen in cylindromas.44 The increased processing of this subunit seen when II-4 cells were cultured on AlloDerm was associated with the maturation of BM assembly and may be associated with the capacity of these cells to selectively adhere to this substrate. However, our immunblot analysis showed that the relative amounts of the processed and unprocessed forms of the γ2 chain were similar for both II-4 cells and NHK grown either on the Type 1 collagen gel or AlloDerm, demonstrating that the degree of processing of this laminin 5 subunit did not direct the selective attachment of II-4 cells when grown with NHK.
In this light, our findings suggest that microenvironmental control of IE tumor cell attachment was independent of synthesis or processing of laminin 5, but the distribution of this protein suggests that it may play a role in the spatial organization of other BM components that are important in cell adhesion to the stromal interface. This suggested that BM can serve as a pre-existing template that enables the organization of BM structure required for II-4 cell attachment. In the current study, we have used a tissue substrate (AlloDerm) that serves as such a template. AlloDerm is a de-epidermalized, human cadaver dermis that is processed to eliminate all cells while preserving laminin 1 and Types IV and VII collagen on its surface. We have recently shown that the presence of these proteins directs the rapid assembly of well-structured BM at the dermal-epidermal interface.17 It is known that the cell surface assembly of laminin 1 and laminin 5 forms a scaffold that serves as a nucleation site for the earliest initiation of BM-mediated adhesion of normal keratinocytes.39 and that laminin 5 can accelerate lamina densa formation in skin equivalents.45 The early, focal invasion of colorectal carcinoma cells has been shown to be dependent on the connective tissue microenvironment, which plays a role in regulating the expression and organization of laminin 5.46 Another recent study showed that both pancreatic carcinoma cells47 and gastric carcinoma cells48 were associated with the expression of laminin 1 and laminin 5, which provided newly synthesized basement-membrane-like material to enable tumor cell invasion. Our findings also suggest that these BM proteins may interact to direct the attachment and persistence of IE tumor cells in a basal position during an earlier stage of progression that precedes invasion by enabling the attachment and intraepithelial expansion of tumor cells in our tissue models. Furthermore, subpopulations of carcinoma cells with increased proliferative potential have been associated with their augmented adhesion to laminin-149, 50 and with an enhanced migratory phenotype and metastatic potential.51, 52 In a similar manner, the elevation of II-4 cell proliferation that we have seen on laminin-1-coated substrates suggests that appropriate matrix attachment, as directed by BM proteins,53 is required for IE tumor cell growth. This confirms the relationship between integrin-mediated extracellular matrix attachment and cell cycle progression that has previously been described for fibroblasts.54 It may therefore be possible that in addition to augmenting their growth, attachment of IE tumor cells to BM may promote their progression by selecting cells that will have a migratory advantage during early invasion. Determination of mechanisms through which this anchorage-dependent growth of IE tumor cells occurs and the role of integrins in this process55 requires further study.
While it is known that clonal selection drives the emergence of transformed cells in simple, monolayer tissue culture,3, 56, 57 permissive conditions for selection of potentially malignant cells in stratified epithelium with 3D tissue architecture have not been studied until now. We have shown that the ability of IE tumor cells to attach to BM components at the BM interface is a meaningful selective pressure that yields tumor cells with an enhanced advantage for tissue retention that may predispose these cells to the accumulation of additional genetic change over time. In contrast, the lack of attachment to BM may function as an “innate anticancer mechanism,”31 since potentially malignant IE tumor cells are anchorage-dependent for persistence and growth and need to be in appropriate tissue context in order to compete with more normal keratinocytes for attachment to BM components and enable disease progression. The initiation of tumor cell invasion may first require such binding of IE tumor cells at the BM zone. Our findings offer mechanistic insights into this event by demonstrating that there is a requisite, adhesion-dependent stage that creates permissive conditions for IE tumor cell selection and growth. Microenvironmental selection of potentially-malignant cells is therefore a driving force in the progression of IEN. By understanding the role of epithelial-stromal interactions in progression of precancer to malignancy and by exploring signaling pathways associated with this transition in future studies, new therapeutic modalities designed to abrogate these events may be formulated to prevent cancer occurrence.
We thank Dr. S. Griffey and Dr. J. Harper of LifeCell, Inc. for AlloDerm used for organotypic culture and grafting; L. Taichman, S. Ghazizadeh and A. Margulis for critical discussions; R. Harrington, S. Pawagi, L. Bertolotti, L. Pfeiffer, M. Scalia, N. Segal, L. Nguyen and N. Lin for technical assistance; and K. Henrickson, S. Suchit and C. Dowling for preparation of illustrations and R. Mulligan for the MFG vector.