Integrins and anoikis.

The loss of integrin-mediated cell-matrix contact induces apoptosis ('anoikis') in certain cell types. Recently it has been shown that protein kinase signaling pathways control anoikis both positively and negatively. Focal adhesion kinase, when activated by integrins, can suppress anoikis. Phosphatidylinositol 3-kinase and the AKT oncoprotein may mediate the anoikis-suppressing effects of focal adhesion kinase. Conversely, the stress-activated protein kinase/Jun amino-terminal kinase pathway promotes anoikis. Latest results indicate that caspase-mediated cleavage of the first component of this latter pathway, MEKK-1, may trigger activation of this pathway in anoikis. In addition, certain integrins may regulate bcl-2 expression levels, possibly adjusting the threshold for anoikis.

Anoikis mechanisms.

Anoikis is defined as apoptosis that is induced by inadequate or inappropriate cell-matrix interactions. It is involved in a wide diversity of tissue-homeostatic, developmental and oncogenic processes. The central problem of anoikis is to understand how integrin-mediated cell adhesion signals control the apoptotic machinery. In particular, the initiation of the caspase cascade in anoikis remains to be explained.

E1a induces the expression of epithelial characteristics.

Cells closely resembling epithelia constitute the first specific cell type in a mammalian embryo. Many other cell types emerge via epithelial-mesenchymal differentiation. The transcription factors and signal transduction pathways involved in this differentiation are being elucidated. I have previously reported (Frisch, 1991) that adenovirus E1a is a tumor suppressor gene in certain human cell lines. In the present report, I demonstrate that E1a expression caused diverse human tumor cells (rhabdomyosarcoma, fibrosarcoma, melanoma, osteosarcoma) and fibroblasts to assume at least two of the following epithelial characteristics: (a) epithelioid morphology; (b) epithelial-type intercellular adhesion proteins localized to newly formed junctional complexes; (c) keratin-containing intermediate filaments; and (d) down-regulation of non-epithelial genes. E1a thus appeared to partially convert diverse human tumor cells into an epithelial phenotype. This provides a new system for molecular analysis of epithelial-mesenchymal interconversions. This effect may also contribute to E1a's tumor suppression activity, possibly through sensitization to anoikis (Frisch, S.M., and H. Francis, 1994. J. Cell Biol. 124:619-626).

Disruption of epithelial cell-matrix interactions induces apoptosis.

Cell-matrix interactions have major effects upon phenotypic features such as gene regulation, cytoskeletal structure, differentiation, and aspects of cell growth control. Programmed cell death (apoptosis) is crucial for maintaining appropriate cell number and tissue organization. It was therefore of interest to determine whether cell-matrix interactions affect apoptosis. The present report demonstrates that apoptosis was induced by disruption of the interactions between normal epithelial cells and extracellular matrix. We have termed this phenomenon "anoikis." Overexpression of bcl-2 protected cells against anoikis. Cellular sensitivity to anoikis was apparently regulated: (a) anoikis did not occur in normal fibroblasts; (b) it was abrogated in epithelial cells by transformation with v-Ha-ras, v-src, or treatment with phorbol ester; (c) sensitivity to anoikis was conferred upon HT1080 cells or v-Ha-ras-transformed MDCK cells by reverse-transformation with adenovirus E1a; (d) anoikis in MDCK cells was alleviated by the motility factor, scatter factor. The results suggest that the circumvention of anoikis accompanies the acquisition of anchorage independence or cell motility.

Collagenase is a major gene product of induced rabbit synovial fibroblasts.

We have investigated the effects of the tumor-promoting phorbol diester, 12-O-tetradecanoylphorbol-13-acetate (TPA), on rabbit synovial fibroblasts, and found that this agent induced a major switch in gene expression in these cells that was marked by the specific induction of the neutral proteinase, collagenase, and was always accompanied by alterations in cell morphology. Procollagenase synthesis and secretion was first observed 6-12 h after the addition of TPA. The rate of collagenase production (1-5 U, or approximately 0.2-1 micrograms secreted procollagenase protein per 10(5) cells per 24 h) depended on the TPA concentration (1-400 ng/ml) and time of exposure (1-72 h). Procollagenase was the most prominent protein visible by direct silver staining or by autoradiography after SDS PAGE of [35S]methionine-labeled proteins. The two procollagenase bands of Mr 53,000 and 57,000, which migrated as a family of spots on two-dimensional gels and were immunoprecipitated by antibodies to purified rabbit collagenase, accounted for 23% of the newly synthesized, secreted protein in TPA-treated cells. Cell-free translation of mRNA from TPA-treated cells in rabbit reticulocyte lysate produced a single band of immunoprecipitable preprocollagenase (Mr 55,000) as a major product (5% of total) that migrated as a single spot on two-dimensional gels. Secreted procollagenase, preprocollagenase , and active collagenase (purified to homogeneity; specific activity 1.2 X 10(4) U/mg protein) had related peptide maps. Two other major secreted proteins, a neutral metalloproteinase of Mr 51,000 and a polypeptide of Mr 47,000, were also induced by TPA. In contrast to the induction of these four polypeptides, TPA decreased synthesis and secretion of a number of proteins, including collagen and fibronectin. Thus, collagenase is a convenient marker for major alterations in the pattern of protein synthesis and secretion by rabbit synovial fibroblasts treated with TPA.

Changes in cell shape correlate with collagenase gene expression in rabbit synovial fibroblasts.

Induction of the neutral proteinase, collagenase, is a marker for a specific switch in gene expression observed in rabbit synovial fibroblasts. A variety of agents, including 12-O-tetradecanoylphorbol-13-acetate, cytochalasins B and D, trypsin, chymotrypsin, poly(2-hydroxyethylmethacrylate), and trifluoperazine induced this change in gene expression. Induction of collagenase by these agents was always correlated with a marked alteration in cell morphology, although the cells remained adherent to the culture dishes. The amount of collagenase induced was positively correlated with the degree of shape change produced by a given concentration and, to some extent, with the duration of treatment. Altered cell morphology was required only during the first few hours of treatment with inducing agents; after this time collagenase synthesis continued for up to 6 d even when agents were removed and normal flattened cell morphology was regained. All agents that altered cell morphology also produced a characteristic switch in protein secretion phenotype, characterized by the induction of procollagenase (Mr 53,000 and 57,000) and a neutral metalloproteinase (Mr 51,000), which accounted for approximately 25% and 15% of the protein secreted, respectively. Secretion of another neutral proteinase, plasminogen activator, did not correlate with increased collagenase secretion. In contrast, synthesis and secretion of a number of other polypeptides, including the extracellular matrix proteins, collagen and fibronectin, were concomitantly decreased. That changes in cell shape correlated with a program of gene expression manifested by both degradation and synthesis of extracellular macromolecules may have broad implications in development, repair, and pathologic conditions.

A role for Jun-N-terminal kinase in anoikis; suppression by bcl-2 and crmA.

The disruption of interactions between extracellular matrix and specific cognate integrins triggers apoptosis in epithelial cells, in a process termed "anoikis." To understand anoikis, the connections between epithelial cell integrin signaling and the apoptosis-regulatory proteins are being explored. We report herein that early after detachment from matrix, epithelial cells activate Jun-N-Terminal Kinases (JNKs; alternatively known as Stress-activated Protein Kinases), which are also activated by other apoptotic stimuli. The activity of this pathway was required for anoikis. Another early response to cell suspension was the activation of the ICE-related cysteine protease, ICE/LAP3; this activation and anoikis were suppressed by the ICE-protease inhibitor, crmA. The overexpression of bcl-2 suppressed ICE/LAP3 activation as well. Surprisingly, bcl-2 and crmA attenuated the activation of JNKs following cell suspension, suggesting that the JNK pathway is regulated directly or indirectly by proteolysis. In addition, the blockage of the JNK pathway attenuated the activation of ICE/LAP3, suggesting a positive feedback loop between the ICE and JNK systems. These results indicate the following sequence of information flow in anoikis: integrins-->bcl-2/bax-->(ICE-proteases<-->JNK)-->apopt osis. Cell-cell interactions, which were previously shown to sensitize cells to anoikis, caused bcl-2 mRNA to be downregulated, a permissive event for downstream apoptotic signaling.

Control of adhesion-dependent cell survival by focal adhesion kinase.

The interactions of integrins with extracellular matrix proteins can activate focal adhesion kinase (FAK) and suppress apoptosis in normal epithelial and endothelial cells; this subset of apoptosis has been termed "anoikis." Here, we demonstrate that FAK plays a role in the suppression of anoikis. Constitutively activated forms of FAK rescued two established epithelial cell lines from anoikis. Both the major autophosphorylation site (Y397) and a site critical to the kinase activity (K454) of FAK were required for this effect. Activated FAK also transformed MDCK cells, by the criteria of anchorage-independent growth and tumor formation in nude mice. We provide evidence that this transformation resulted primarily from the cells' resistance to anoikis rather than from the activation of growth factor response pathways. These results indicate that FAK can regulate anoikis and that the conferral of anoikis resistance may suffice to transform certain epithelial cells.

The G12/13-RhoA signaling pathway contributes to efficient lysophosphatidic acid-stimulated cell migration.

The membrane redistribution and phosphorylation of focal adhesion kinase (FAK) have been reported to be important for cell migration. We previously showed that Lysophosphatidic acid (LPA) induced FAK membrane redistribution and autophosphorylation in ovarian cancer SK-OV3 cells and the signaling pathway consisting of Gi-Ras-MEKK1 mediated LPA-induced FAK membrane redistribution but not FAK autophosphorylation. We also showed that the disruption of the Gi-Ras-MEKK1 pathway led to a significant reduction in LPA-stimulated cell migration. These findings raised the question of whether LPA-induced FAK autophosphorylation was required for LPA-stimulated cell migration and what signaling mechanism was involved in LPA-induced FAK autophosphorylation. In this study, we expressed the membrane anchored wild-type FAK (CD2-FAK) in SK-OV3 cells and found that the expression of CD2-FAK greatly rescued LPA-stimulated cell migration in Gi or Ras-inhibited cells. However, Gi inhibitor pertussis toxin or dominant-negative H-Ras still significantly inhibited LPA-stimulated cell migration in cells expressing the membrane anchored FAK containing a mutation in the autophosphorylation site [CD2-FAK(Y397A)]. These results suggest that FAK autophosphorylation plays a role in LPA-stimulated cell migration. With the aid of p115RhoGEF-RGS, G12 and G13 minigenes to inhibit G12/13, we found that the G12/13 pathway was required for LPA-induced FAK autophosphorylation and efficient cell migration. Moreover, LPA activated RhoA and Rho kinase (ROCK) in a G12/13-dependent manner and their activities were required for LPA-induced FAK autophosphorylation. However, Rho or ROCK inhibitors displayed no effect on LPA-induced FAK membrane redistribution although they abolished LPA-induced cytoskeleton reorganization. Our studies show that the G12/13-RhoA-ROCK signaling pathway mediates LPA-induced FAK autophosphorylation and contributes to LPA-stimulated cell migration.

Evidence for a function of death-receptor-related, death-domain-containing proteins in anoikis.

Normal epithelial cells undergo apoptosis if integrinmediated matrix contacts are lost, in a process termed 'anoikis'. Anoikis prevents shed epithelial cells from colonizing elsewhere, and is thus essential for maintaining appropriate tissue organisation. Aberrant oncogenes or tumor suppressor genes can cause resistance to anoikis, thereby contributing substantially to malignancy. Apoptosis is mediated by a well-ordered signaling cascade, which involves activation of intracellular proteases known as caspases. However, the mechanism by which the caspase cascade is initiated following cell-matrix detachment is unknown. We have hypothesized that death receptor activation might be involved in anoikis. To test this hypothesis, we developed a transient assay for anoikis and used it to assay the effects of proteins that block the function of domains found within death receptors known as death domains. In this assay, silencer of death domains (SODD) and dominant-negative FAS-associated death domain protein (FADD) efficiently inhibited anoikis in Madin-Darby canine kidney (MDCK) cells. The protective activity of SODD required its BAG domain, which interacts with the heat shock proteins hsp70 and hsc70, and inhibits the chaperone activity of the latter. Both caspase 8, which physically associates with death receptors, and cleavage of the caspase-8 substrate BID, were activated by cell-matrix detachment. These findings indicate a role for death receptors or proteins with related death domains in triggering anoikis.

Positive and negative transcriptional elements of the human type IV collagenase gene.

Proteolysis by type IV collagenase (T4) has been implicated in the process of tumor metastasis. The T4 gene is expressed in fibroblasts, but not in normal epithelial cells, and its expression is specifically repressed by the E1A oncogene of adenovirus. We present an investigation of the transcriptional elements responsible for basal, E1A-repressible, and tissue-specific expression. 5'-Deletion analysis, DNase I footprinting, and gel mobility shift assays revealed a strong, E1A-repressible enhancer element, r2, located about 1,650 bp upstream of the start site. This enhancer bound a protein with binding specificity very similar to that of the transcription factor AP-2. A potent silencer sequence was found 2 to 5 bp downstream of this enhancer. The silencer repressed transcription from either r2 or AP-1 enhancer elements and in the context of either type IV collagenase or thymidine kinase (tk) gene core promoters; enhancerless transcription from the latter core promoter was also repressed. Comprising the silencer were two contiguous, autonomously functioning silencer elements. Negative regulation of T4 transcription by at least two factors was demonstrated. mcf-7 proteins specifically binding both elements were detected by gel mobility shift assays; a protein of approximately 185 kDa that bound to one of these elements was detected by DNA-protein cross-linking. The silencer repressed transcription, in an r2 enhancer-tk promoter context, much more efficiently in T4-nonproducing cells (mcf-7 or HeLa) than in T4-producing cells (HT1080), suggesting that cell type-specific silencing may contribute to the regulation of this gene.

Blockage of tropoelastin secretion by monensin represses tropoelastin synthesis at a pretranslational level in rat smooth muscle cells.

The blockage of protein secretion in the R22 cultured rat aortic smooth muscle cell strain with monensin repressed tropoelastin gene expression at the mRNA level by ca. 50-fold as measured by biosynthetic pulse-labeling, in vitro translation, and hybridization with a tropoelastin genomic DNA probe. These results suggest that tropoelastin gene expression is autoregulated, and they represent the first reported effect of monensin on gene expression.

Roles of Grainyhead-like transcription factors in cancer.

The mammalian homologs of the D. melanogaster Grainyhead gene, Grainyhead-like 1-3 (GRHL1, GRHL2 and GRHL3), are transcription factors implicated in wound healing, tubulogenesis and cancer. Their induced target genes encode diverse epithelial cell adhesion molecules, while mesenchymal genes involved in cell migration and invasion are repressed. Moreover, GRHL2 suppresses the oncogenic epithelial-mesencyhmal transition, thereby acting as a tumor suppressor. Mechanisms, some involving established cancer-related signaling/transcription factor pathways (for example, Wnt, TGF-ß, mir200, ZEB1, OVOL2, p63 and p300) and translational implications of the Grainyhead proteins in cancer are discussed in this review article.

Adenovirus E1a-mediated tumor suppression by a c-erbB-2/neu-independent mechanism.

We reported previously that the adenovirus E1a gene reversed the transformed phenotype of one human melanoma and one fibrosarcoma cell line (S. Frisch, Proc. Natl. Acad. Sci. USA, 88: 9077-9081, 1991). To determine the generality of the tumor suppression effects of E1a, a diversity of tumor cell lines, including A204 rhabdomyosarcoma, RD rhabdomyosarcoma, Saos-2 osteosarcoma, NCI-H23 non-small cell lung carcinoma, MDA-MB435S breast carcinoma, and ras-transformed MDCK kidney epithelial cells, were infected with a retrovirus bearing the 12S E1a coding sequence. We demonstrate here that the expression of E1a severely reduced the anchorage-independent and tumorigenic growth of these cell lines without affecting their growth under normal culture conditions. The parental tumor cells used in this study did not overexpress c-erbB-2/neu, and E1a did not affect its expression in these cells. Thus, tumor suppression by E1a can operate in a wide variety of human tumor cells by c-erbB-2/neu-independent mechanisms. E1a also sensitized these cell lines to the cytotoxic effects of the anticancer drugs etoposide and cisplatin. The results suggest that E1a could prove useful for the gene therapy of a wide variety of human cancers.

Evidence for a function of CtBP in epithelial gene regulation and anoikis.

Previously, we reported that adenovirus E1a protein behaves as a tumor suppressor in human cells. It apparently functions by transcriptionally inducing an array of epithelial cell adhesion genes, while repressing other cell-type specific genes, thus producing an epithelial phenotype. Concomitantly, the cells become sensitive to anoikis (apoptosis of epithelial cells detached from extracellular matrix), potentially causing tumor suppression. E1a protein interacts with the nuclear acetylases p300, CBP and P/CAF, and also with the co-repressor protein CtBP. In this study, we have determined the role of these interactions in E1a's phenotypic effects on human tumor cells. The results indicate that E1a's interaction with CtBP activates at least three epithelial cell adhesion gene promoters. The E-cadherin repressor appeared to be the CtBP-interacting protein delta EF1/ZEB, which bound the ras-repressible E-boxes of the E-cadherin promoter. The E1a-CtBP interaction also contributed to anoikis-sensitization. E1a's interactions with the nuclear acetylases conferred epithelial morphologies but did not activate epithelial genes. These latter interactions did not sensitize tumor cells to anoikis but nevertheless conferred tumor suppression. These results implicate CtBP as an antagonist of the epithelial phenotype and anoikis. They also indicate a new but undefined role for nuclear acetylases in maintaining the transformed phenotype.

Adenovirus E1A represses protease gene expression and inhibits metastasis of human tumor cells.

Stable transfection of human tumor cell lines with the adenovirus-5 E1A gene repressed the expression of the secreted proteases, type IV collagenase, interstitial collagenase and urokinase. In addition, E1A blocked the 12-O-tetradecanoyl phorbol acetate (TPA) induction of interstitial collagenase transcription in HT1080 fibrosarcoma cells. Plasmids bearing the interstitial collagenase or type IV collagenase 5' flanking regions linked to a chloramphenicol acetyl transferase coding sequence were constructed and analysed for expression by transient cotransfections into HT1080 cells. Cotransfection with a plasmid bearing a functional E1A gene repressed transcription of the type IV collagenase promoter and blocked the TPA induction of the interstitial collagenase promoter. Furthermore, E1A repressed transcription from a TK promoter driven by AP-1 complex binding sites (TRE), suggesting that E1A interferes with the AP-1 trans-activation pathway. This effect was not, however, due to the repression of c-jun gene transcription by E1A. In fact, the expression of E1A rendered the c-jun gene hypersensitive to TPA induction. Concomitant with reduction in expression levels of secreted proteases, stable E1A transfectants showed reduced metastatic activity in vivo and reduced ability to traverse a reconstituted basement membrane in vitro. Monospecific anti-type IV collagenase antibodies inhibited invasive activity of parental tumor cell lines in the in vitro assay, suggesting a possible causal relationship between the repression of secreted proteases and loss of metastatic properties of the transformants.

Tumor suppression activity of adenovirus E1a protein: anoikis and the epithelial phenotype.

Adenovirus E1a proteins reverse-transform diverse human tumor cells in culture. This has stimulated interest in the arenas of clinical and basic cancer research. Clinically, cancer gene therapy trials on E1a are in progress, and drug discovery strategies based on E1a are being considered. Biologically, the effect of E1a is unique in that it overrides most or all oncogenic signaling pathways to yield nontumorigenic cells. Apparently, this is a consequence of the ability of E1a to reprogram transcription in tumor cells so as to produce an epithelial phenotype that is refractory to oncogenic growth stimulation. The molecular basis for this effect is emerging.

Secreted proteases. Regulation of their activity and their possible role in metastasis.

Extracellular matrix metalloproteases are secreted by the resident cells of the tissue in a proenzyme form, and their extracellular activity is regulated at the level of gene expression, proenzyme activation, and interaction with inhibitors. To understand the molecular mechanisms that control the activity of ECM metalloproteases and their effect on the cellular phenotype, we have established cell lines in which the transcription of the protease genes is repressed. We also have undertaken a detailed study of the pathway of extracellular activation of interstitial procollagenase. Stable transfection of three human tumor cell lines--H-ras-transformed bronchial epithelial cells TBE-1, fibrosarcoma cells HT1080, and melanoma cells A2058--with the adenovirus E1A gene dramatically repressed the expression of the secreted proteases, type IV and interstitial collagenases, and urokinase-type plasminogen activator. Concomitantly, E1A-expressing cells showed reduced metastatic activity in vivo and reduced ability to traverse a reconstituted basement membrane in vitro. Monospecific anti-type IV collagenase antibody inhibited the invasive activity of parental tumor cell lines in the in vitro system, suggesting a possible causal relationship between the effect of E1A on the expression of secreted proteases and the reduced metastatic potential of the E1A-expressing transformants. We have also studied the mechanism of regulation of metalloprotease activity at the level of extracellular activation by investigating the cascade of proteolytic events that results in the activation of interstitial procollagenase. Cocultivation of the major cellular components of skin, dermal fibroblasts, and epidermal keratinocytes induces activation of interstitial procollagenase and prostromelysin in the presence of plasminogen. This activation occurs through a uPA-plasmin-dependent pathway in which plasmin catalyzes the first step in activation of both collagenase and stromelysin by amino-terminal processing. Activated stromelysin can in turn convert plasmin-activated collagenase into a fully active enzyme by removal of approximately 15 amino acid residues from the carboxyl end of the enzyme. This second step of activation results in a 5-8-fold further increase in specific activity of collagenase. This cascade of proteolytic events may constitute a major physiologic pathway of collagenase activation.

Reversal of malignancy by the adenovirus E1a gene.

Tumor suppressor genes such as Rb and p53 usually kill tumor cells when overexpressed ectopically. This is a consequence of their normal cell cycle regulatory functions. By contrast, the E1a gene of adenovirus, a common cold virus, converts tumor cells into viable normal cells. This has advantages for investigation and control of cancer. In particular, E1a is a master programmer of the epithelial phenotype. This provides a new tool for understanding the molecular basis of the epithelial-mesenchymal transition, and how it goes awry in cancer cells. Furthermore, epithelial cells are sensitive to a form of apoptosis - 'anoikis' - that is induced by detachment from extracellular matrix. This property confers strict anchorage-dependence. Transcriptional programming, by E1a or the formation of cell-cell junctional complexes, programs epithelial cells to be sensitive to anoikis.