Received 6/15/97 Accepted 6/28/97

In the recent past, molecular oncologists focused their studies primarily on the cellular pathways controlling proliferation. Neoplastic disease was typically envisaged as resulting from defects in these pathways leading to excess cell division. By extension, cancer therapies, when successful, were thought to act by selectively targeting rapidly cycling cells. Therapeutic index could, for practical purposes, be conceptualized as a ratio of mitotic rates between normal and malignant cells. During the last five years, however, overwhelming evidence has accumulated which suggests that the other side of the balance - the rate of cell death - is just as important. A central theme in much of cancer research today is the ability of tumor cells to resist apoptosis in response to triggers which typically induce cell cycle arrest or death in their untransformed counterparts. We review the importance of apoptosis in modulating central processes in cancer biology such as tumorigenesis, anchorage-independent survival, tumor-induced angiogenesis, and antineoplastic therapies.

Apoptosis is now widely accepted to play a role in tumorigenesis. In the same way that programmed cell death by apoptosis may have evolved as a mechanism for regulating cell numbers and their interactions during normal development, apoptosis may also serve to eliminate cells gone "awry" in the adult organism. In fact, many proteins which are typically known as regulators of cellular proliferation also are potent inducers of apoptosis. The c-myc protooncogene, an early growth response gene required for G1/S transition upon mitogenic stimulation, has been shown to play apparently divergent roles in modulating both proliferation and apoptosis.

Deregulated Myc expression is observed in a variety of tumors, and its enforced expression in animals leads to cellular transformation and tumorigenesis (139). In culture, fibroblasts transformed with Myc display increased growth kinetics when cultured in the presence of adequate serum growth factors. Upon serum withdrawal, however, they undergo rapid apoptosis, the kinetics of which correlate with cellular levels of Myc protein (133). In contrast, serum starvation in untransformed fibroblasts typically results in growth arrest. This transformation-selective response is also observed following a number of apoptotic triggers such as radiation and chemotherapy (see below). Myc activates apoptosis in other cell types as well (140, 141), and can cooperate with Bcl-2 to transform haematopoietic cells (142, 143). This scenario is observed in vivo during follicular lymphagenesis, whereby Myc overexpression leads to malignant growth instead of apoptosis when placed in the context of Bcl-2's anti-apoptotic effect (144).

While a number of other cellular as well as viral oncoproteins have dual roles in regulating both proliferation and apoptosis (145-147), many of these require p53 to promote cell death. The p53 tumor suppressor was recognized as a regulator of apoptosis following the observation that transfection or activation of wild-type p53 in tumor cells can result in rapid cell death (148, 149). Following these initial observations, a large number of studies demonstrated that most cytotoxic triggers like radiation or chemotherapeutics require functional p53 to induce apoptosis (150-153). Similarly, serum withdrawal-induced death in Myc-overexpressing cells or IL-3-dependent thymocytes is p53-dependent (154-156). Several studies in transgenic mice have confirmed the notion that p53's tumor suppressive role in vivo is closely linked to its ability to induce apoptosis (157, 158).

The adenovirus E1A gene, like Myc, sensitizes cells to apoptosis induced by serum withdrawal, ionizing radiation, and a variety of chemotherapy agents (159, 160). E1A is thought to mediate its effect on cell death at least in part by stabilizing p53 protein levels (160). Similar to the cooperation between Myc and Bcl-2 oncoproteins in follicular lymphoma, transformation of primary cells by adenovirus 5 requires both E1A and E1B (161). E1A's ability to stimulate proliferation through binding and inactivation of pRB is 'balanced' by its apoptosis-inducing activity in the absence of E1B. The E1B gene encodes two transforming proteins - 19K and 55K - both of which disable the p53 pathway to apoptosis (162). Therefore, in the absence of p53 mutations or the E1B proteins, cells overexpressing E1A eventually die by apoptosis (160). The human papilloma virus (HPV), responsible for the majority of cervical carcinomas, triggers an analogous molecular scenario upon infection (163). The E7 protein, like E1A, hijacks the cell cycle machinery through binding pRB, and will trigger apoptosis when expressed alone (164, 165). Coexpression of the E6 protein, which stimulates ubiquitin-mediated degradation of p53, cancels E7's cell death response and leads to cellular transformation (166).

Other proteins which directly regulate cell growth or growth response genes stimulate apoptosis when overexpressed, and point to a role for cell cycle regulation in the life/death decisions of a cell. Deregulated expression of proteins like pRB, cyclin D, or E2F-1 have been shown to affect apoptosis. In fact, characterization of E2F-1 knock-out mice by several groups reveals the paradox whereby this transcription factor can induce tumors as well as promote apoptosis and suppress proliferation in vivo (167, 168)! During the cell cycle the E2F-1 protein is released upon pRB phosphorylation in G1 to transactivate genes involved in S phase entry (169). When expressed alone in the absence of serum or in combination with a growth suppressor gene like p53, E2F-1 induces S phase entry followed by apoptosis (170, 171). Deregulated cyclin D1 expression has been shown to have the same consequences (172). As might be expected, however, the opposite effect on apoptosis has been observed when pRB is overexpressed in cells (173, 174). Therefore, in the context of a growth arrest signal like serum starvation, high p53 levels, or DNA damage (see below), deregulated expression of critical cell cycle progression effectors can trigger the cell into apoptosis. While these observations clearly indicate that signaling pathways to growth and apoptosis are tightly linked, they do not reveal the mechanism by which these proteins induce apoptosis. A key question therefore is whether genes like Myc, E2F, pRB and p53 mediate apoptosis through the same pathways by which they regulate the cell cycle.

Conflicting reports in the literature make it unclear as to whether apoptosis is triggered as a result of conflicting growth signals or whether the regulatory proteins we have discussed so far have distinct apoptosis-inducing activities of their own (175). Myc-induced apoptosis, for example, often correlates with elevated cyclin A expression and activation of cyclin A-dependent kinases cdk2 and cdc2 (176, 177). Inappropriate cell cycle activation of cdc2 leads to cell death (178), which, in other systems, can be suppressed by dominant negative cdks or cdk-inhibitors (179, 180). Similarly, overexpression of either cdc25 or ornithine decarboxylase - two of Myc's transcriptional targets - is sufficient to induce apoptosis (181, 182), and antisense inhibition of cdc25 can block Myc-induced cell death (147). While these studies indicate a role for these proteins in regulating apoptosis, other reports suggest that Myc-induced cyclin A and cell cycle progression does not mediate apoptosis (183, 184). The ability of Myc-overexpressing cells to be triggered into apoptosis throughout the cell cycle also supports this model (133, 185, 186). Studies on the transcription factor p53 and its mechanism of apoptosis-induction have also yielded contradictory results.

p53 regulates the expression of a number of genes in response to extracellular cues such as hypoxia or DNA damage (187). Two alternative cellular responses occur as a result of p53 induction: growth arrest in the G1 phase of the cell cycle or apoptosis. p53's transcriptional targets include Gadd45, a gene thought to be involved in DNA repair (188), p21cip1/waf1, a G1-specific cdk-inhibitor (189), and Mdm-2, a negative feedback regulator of p53 (190, 191). While p53-mediated transactivation of these target genes is thought to be crucial for DNA damage-induced growth arrest (187), it remains unclear whether p53-mediated transcription of these genes is needed to signal apoptosis.

Myc-mediated apoptosis has been shown to require p53 in a manner independent of p21 induction (154), and p21-deficient mice - while defective in G1 arrest - undergo normal development and apoptosis (192). These data support the idea that p53's cell cycle regulatory function may be separate from its role in apoptosis. The bax gene was recently found to be a transcriptional target of p53 and is upregulated in response to a variety of p53-dependent apoptosis triggers (193). However, bax induction does not always correlate with p53-dependent apoptosis (145), and thymocytes from bax knock-out mice undergo normal p53-mediated apoptosis ((56). In some systems, p53-dependent apoptosis can occur without the apparent need of any new transcription or translation (154, 194). Some reports demonstrate that transactivation-defective p53 mutants lose their ability to mediate apoptosis induced by E1A (195), while others find that p53's transactivation activity does not correlate with apoptotic activity (196, 197). A human-tumor derived p53 mutant which retains cell-cycle arrest activity but is defective for apoptosis has also been described (198). p53-interacting proteins which may modulate its transactivation-independent activities have been identified and include WT1, the Wilms tumor suppressor gene product (199), and the XPB and XPD helicases (197). Therefore, p53 may mediate both transactivation-dependent and independent signaling pathways to apoptosis.

In summary, many oncogenes which stimulate cellular proliferation also potently induce apoptosis. Although it still remains unclear whether these genes kill cells by mechanisms related to the manner in which they promote proliferation, tumors which exhibit deregulated expression of these oncogenes typically select genetic aberrations which inactivate apoptosis. p53 loss or high Bcl-2 levels are common examples of such aberrations, and often correlate with poor prognosis in the clinics (see below). These observations have provided insights on the in vivo process of 'multi-step' tumorigenesis, whereby loss of growth control as well as inactivation of the apoptosis pathway are central for the survival and proliferation of tumors.

The extracellular matrix (ECM) mediates a number of pivotal processes from cellular migration during gastrulation to tissue homeostasis in the adult (200, 201). Only recently, however, has the ECM's importance as a suppressor of apoptosis been recognized (202). 'Anchorage-dependence' describes the requirement of certain cell types for ECM adhesion in order to proliferate. In the same way that trophic factors prevent apoptosis of "factor-dependent" cells, adhesion to the ECM - a process largely mediated by the integrin family of cell surface receptors (203) - has also been shown to be crucial for the survival of certain anchorage-dependent cells. The ability to grow imbedded within a semi-solid matrix has been extensively studied as a property cells can acquire during tumorigenesis, and es in integrin expression patterns and levels during progression of tumors from benign to more malignant phenotypes have been described (204, 205). The realization that ECM adhesion is important for cell survival as well as proliferation has broadened our understanding of anchorage-dependence and motivated researchers to decipher the adhesion-mediated signaling pathways which inhibit apoptosis. Triggering adhesion-specific death pathways in transformation-selective ways may allow us to prevent anchorage-independent proliferation, a likely first step in the progression of tumors to metastasis.

Similar to the cellular responses to DNA damage or serum starvation, loss of matrix adhesion in primary cells can lead to one of two fates: growth arrest or apoptosis. Primary fibroblasts which are detached from ECM slow their rate of protein synthesis and eventually growth arrest in the G1 phase of the cell cycle (206, 207). Indeed, integrin-mediated signals are known to regulate the cell cycle machinery at the level of the transcription and translation of various cyclins and cdk inhibitors (reviewed in 208). The behavior of primary fibroblasts contrasts with that of epithelial and endothelial cells which undergo rapid apoptosis when denied ECM adhesion - a process termed 'anoikis' (209).

Human umbilical vein endothelial cells plated on agarose or MDCK cells incubated in saturating amounts of either soluble ECM components or RGD inhibitory peptides are prevented from adhering and rapidly undergo anoikis (209-211). Several instances of developmentally-regulated programmed cell death triggered as a result of ECM degradation have also been reported. Mammary epithelial cells which require beta1 integrin-mediated adhesion to survive are typically eliminated by apoptosis upon matrix degradation during mammary gland involution (212, 213). Similarly, during frog metamorphosis, the transcriptional activation of the matrix metalloproteinase stromelysin-3 by thyroid hormone results in apoptosis of primary intestinal epithelial cells due to remodeling of the ECM (214). During cavitation of the vertebrate embryo, inner endodermal cells which fail to contact the basement membrane are triggered into apoptosis, while those in direct contact with it are rescued (215). The mechanism by which integrin-mediated adhesion suppresses anoikis in these cell-types, however, remains unclear.

Given the central role of integrins in mediating adhesion to the ECM, their putative function in transducing survival signals has come under scrutiny. Transfection of certain integrins, for example, is sufficient to rescue melanoma cells from anoikis in a three-dimensional dermal collagen matrix (216). Focal adhesion kinase (FAK), which is known to bind and transmit signals from the integrins (217), has recently been linked to anoikis in primary cells (218, 219). The ECM has also been suggested to regulate the activity of common apoptosis signaling intermediates like p53 (220, 221), Bcl-2 (222), the Jun-N-terminal kinase (223), and ICE (212). Alternatively, integrins, like oncogenes, may signal through their effect on anchorage-dependent cell-cycle progression (208). It also remains to be seen which specific integrin heterodimer combinations mediate protective effects in distinct cell-types. Whatever the precise molecular signals may be, oncogenic transformation in epithelial and endothelial cells has been suggested as a mechanism for escaping this regulation, and achieving anchorage-independent survival and growth characteristics.

Oncogenic transformation of endothelial or epithelial cells as well as treatments which reduce cell-cell contacts between these cells rescues them from anoikis (202). Transfection of MDCK or human endothelial cells with v-Ha-ras or v-src oncogenes abrogates anoikis (209, 211). Similarly, inducible c-H-ras expression rescues rat intestinal epithelial cells cultured in three-dimensional growth conditions (224). Although the protective mechanism of transformation in these cell types is not yet fully understood, Ras-induced PI3-K leading to Akt activation may be a crucial ECM-mediated survival signal (138). Ras has been suggested to harbor both pro- and anti-apoptotic potentials depending on the downstream cellular pathways it activates (135), and therefore differences in the effects of Ras-transformation on anoikis propensity may be more complex than anticipated. In certain contexts such as overexpression of nuclear oncogenes like Myc or E1A combined with serum starvation, Ras' pro-apoptotic signaling capacity may be 'dominant' over its ability to activate Akt. In fibroblasts which lose matrix adhesion, Ras-transformation in combination with either Myc or E1A does not rescue from anoikis, but instead profoundly activates it (221). Despite differences between cell types, matrix contact in primary cells may promote cell survival via integrin-mediated Ras activation of PI3-K/Akt (138).

Treatment of MDCK cells prior to matrix detachment with either scatter factor or the phorbol ester TPA - both of which lead to disassembly of epithelial intercellular junctions (225, 226) - also confers resistance to anoikis (209). Other developmental signals, in addition to scatter factor, which induce epithelial-mesenchymal transitions might similarly affect a cell's propensity to anoikis. Thus regulation of the survival signals from matrix attachment may themselves be subject to regulation during development. In contrast, a colon carcinoma cell line required intercellular contact for survival in suspension (227); integrins instead of intercellular junctions appeared to mediate this rescue. These observations suggest that a cell's response to the loss of ECM may depend on the status of cell-cell junctions and the signaling pathways they trigger. The loosening of intercellular contact is a likely prerequisite for migration as well as survival in suspension - two characteristics typically shared by primary fibroblasts. The effect of cell-cell contact on sensitivity to anoikis is likely to be cell-type specific. Unlike epithelial/endothelial cells which typically form tight cellular sheets in vivo as a result of intercellular connections, fibroblasts are highly motile and migratory and do not typically display extensive cell-cell contacts. Therefore, the apparent disparities in anoikis sensitivity observed between different cell-types in culture may reflect the in vivo behaviors and intracellular pathways of the cells.

In summary, detachment from the ECM triggers apoptosis in cell types such as epithelial and endothelial cells, while inducing cell cycle arrest in primary fibroblasts. Cell-type differences may also extend to transformed epithelial/endothelial cells, which may resist anoikis, and transformed fibroblasts, which are sensitized to it. The existence of ECM-mediated survival pathways suggests that tumor cells may acquire anchorage-independent growth and survival characteristics by circumventing an otherwise 'default' apoptotic pathway (228). This may be the case for the alpha5beta1 integrin, for example, which has been suggested to send growth arrest/apoptosis signals in fibroblasts or colon carcinoma cells when not ligated to fibronectin (221, 229). Interestingly, there is evidence to suggest that this same receptor induces apoptosis in hematopoietic cells when ligated to fibronectin (230). Therefore tumors arising from a variety of cell types may employ different strategies for circumventing anoikis: selection of mutations which result in constitutive activation of integrin-mediated survival pathways, or loss-of-function mutations in genes required to complete apoptosis in response to the loss of matrix adhesion. It will be interesting to see which signal transduction pathways underlie the basic behavioral difference between different cell types, and which genetic alterations are required to alleviate anoikis and allow for metastatic growth of tumors.

The influence of microenvironment on the carcinogenesis and progression of solid tumors is now well recognized. Transformed cells may not only accumulate genetic mutations which allow for unchecked cell cycle progression and anchorage-independent survival, but also undergo selective pressure to adapt to low nutrient/oxygen conditions. Hypoxia, which commonly occurs as a result of increasing tumor mass and lack of vasculature, is sensed by cells and leads to activation of the hypoxia inducible factor 1 (HIF-1) (231). This basic-helix-loop-helix transcription factor has been shown to have several transcriptional targets including the erythopoeitin (EPO) gene and vascular endothelial growth factor (VEGF) (232). The p53 tumor suppressor is also induced by hypoxia, although it remains unclear whether this occurs via HIF-1 (233).

Hypoxia is also a trigger for apoptosis (as well as necrosis) in transformed cells in vitro (234, 235), and correlates well spatially with areas of apoptosis in vivo (236). Likely as a result of its ability to modulate p53 protein levels and trigger apoptosis, hypoxia has been suggested to provide a selective pressure for expansion of apoptosis-resistant/p53-deficient cells at the core of solid tumors in vivo (236). Interestingly, solid tumors which have hypoxic regions typically carry a poor prognosis suggesting a possible correlation between hypoxia's ability to select for apoptosis-deficient cells and subsequent tumor resistance to chemotherapy or radiation.

In addition to conferring a survival advantage to p53-deficient tumor cells, hypoxia also induces a number of other cellular es in order for cells to adapt to low nutrient/oxygen conditions (237). This adaptive response is carried out in a number of ways including upregulation of metabolic pathways (238, 239) as well as stimulation of angiogenic signaling pathways to increase tumor perfusion (240). The 'angiogenic switch' during tumorigenesis has been suggested to result from a shift in the local balance of angiogenic activators and inhibitors. Hypoxia has been shown to induce a number of genes whose expression may 'trip' the angiogenic switch (241). The VEGF promoter, for example, contains HIF-1 binding sites and is activated by low oxygen as well as glucose starvation (242-244). Furthermore, oncogenic transformation by c-Ha-ras acts synergistically with the effects of hypoxia to stimulate VEGF reporter expression in vitro (232). Hypoxia's role in selecting for p53-deficient cells may also indirectly influence the local angiogenesis balance since mutation of p53 is typically associated with the loss of the angiogenic inhibitor thrombospondin-1 (245).

The effects of angiogenesis on the growth and metastatic potential of tumors have been extensively studied. Light microscopic and immunohistochemical studies have revealed a direct correlation between the invasive and metastatic potential of a tumor and the number and density of microvessels surrounding it (246, 247). The discovery of molecules such as angiostatin (248) and endostatin (249) have demonstrated how depletion of antiangiogenic factors following removal of a primary tumor can directly affect the angiogenic balance and stimulate the growth of local and distant metastases (250). Since preventing angiogenesis can block the growth of both the primary tumor and its metastatic outposts, it has been the focus of much attention in the development of new cancer therapies. Recent studies aiming to understand the mechanisms of angiogenesis and potential ways of interfering with tumor-induced vascularization have revealed that matrix adhesion of proliferating vascular endothelial cells is required for their survival (251).

Endothelial cell proliferation is stimulated by a number of soluble factors like the acidic or basic FGFs, VEGF, and other factors, most of which can be released by hypoxic tumor cells (240). Neovascularization, however, requires that endothelial cells accomplish a number of other tasks such as matrix degradation and migration to appropriate sites (252). These processes involve extensive es in cell-matrix recognition pathways. For example, two cytokine-dependent angiogenic pathways in endothelial cells have been shown to induce expression of distinct integrin receptors. Basic FGF- or TNFalpha-induced angiogenesis was found to be dependent on the alphavbeta3 integrin, while VEGF- or TGFalpha-induced angiogenesis depended on alphavbeta5 (253).

alphavbeta3 and alphavbeta5 integrins may affect not only the migratory capabilities of cells but also their requirement for matrix in order to block apoptosis (251). This anchorage-dependent requirement appears to be specific to proliferating vascular endothelial cells implying that agents which target these integrin-ECM interactions would be quite specific and not affect normal vasculature. Short cyclic RGD-containing peptides which mimic the integrin's target recognition sequence on ECM proteins have been used in several mouse models and shown to be effective in blocking neovascularization (254, 255). Specifically, their mechanism of action in inhibiting angiogenesis has been linked to their ability to induce apoptosis of the newly forming vasculature. Therefore, the discovery of "anchorage-dependent" survival signals suggest new targets not only for blocking processes such as anchorage-independent tumor cell growth, but also for interfering with tumor-induced angiogenesis by initiating anoikis of proliferating vascular endothelial cells.

Accrued interest in deciphering apoptotic signaling pathways has, in part, been sparked by the recent understanding that successful antineoplastic therapies induce tumor cell apoptosis rather than killing them as a result of direct insult to DNA. Traditional radiation biology has hypothesized that ionizing radiation kills cancer cells by overwhelming their mitotic and/or metabolic needs, resulting in disarray and death. The same reasoning applied to chemotherapeutics which cause DNA damage or inhibit DNA modifying enzymes. These explanations, however, do not account for the fact that certain slow-growing tumors respond well to therapies while rapidly growing ones can be completely immune to the same treatments (175). Instead, examination of tumor samples using immunohistochemical assays specific for apoptotic cell death has revealed that death by apoptosis, not necrosis, often follows radiation or chemotherapy (256, 257). In addition, clinical evidence has suggested prognostic links between treatment outcome and distinct molecular genetic alterations which are known to disable the cellular apoptosis pathway (258). As noted above, the foremost example of such a mutated gene is the tumor suppressor p53.

Mutations in p53 are found in over 50% of all malignancies making it one of the most frequently aberrant genes in human cancer (259, 260). Although a number of different mutants of p53 are observed in human cancers, they cluster as missense mutations throughout the DNA binding region and in four 'hotspots' (261). Most of these mutations impair p53's ability to bind DNA and result in a loss of transactivation potential. Mice deficient in p53 show minimal developmental defects, but have a higher incidence of spontaneous tumors (262, 263), similar to humans with Li-Fraumeni syndrome (with p53 germline heterozygosity (264)). In addition to its undisputed role in tumorigenesis, p53 also has prognostic value in terms of response to therapy since most triggers of apoptosis induce and may require p53. From a clinical standpoint, mutations in p53 are usually a poor prognostic indication in a variety of tumor types including gastrointestinal, hematopoietic, breast, and genito-urinary cancers (152, 265-277). Conversely, highly curable tumors like pediatric ALL, testicular, and Wilms tumors often correlate with wild-type p53 status (175). In other tumors such as cervical cancer or sarcomas, although p53 status is wild-type, the protein may be sequestered in an inactive or degraded state due to the effects of viral oncoproteins such as HPV E6 and Mdm-2.

Cell culture studies using E1A/ras-transformed mouse embryo fibroblasts derived from either wild-type or p53 deficient mice have demonstrated that classical therapeutics including etoposide and adriamycin as well as ionizing radiation employ p53 to induce apoptosis (150). These observations were subsequently confirmed in vivo in a nude mice solid tumor model (278). Similarly, transgenic mice carrying a p53-responsive lacZ construct confirm the in vitro data and demonstrate that p53 modulates radiation and drug sensitivity in vivo (279, 280). A number of genes in addition to p53 have been demonstrated to affect the cytotoxicity of drugs and radiation in vitro, and are suspected to have the same effects in vivo. Bcl-2 and Bcl-xL, for example, inhibit chemotherapy-induced apoptosis in neuroblastoma cells (281, 282). The relevance of this observation is confirmed by the in vivo data that a significant number of primary neuroblastomas harbor elevated Bcl-2 levels which correlate with aggressive tumor behavior and poor prognosis (283, 284). Bcl-2 has been shown to modulate drug-induced apoptosis in other systems as well (285, 286) although, surprisingly, high Bcl-2 expression has been correlated with improved prognosis in breast cancer (287-291).

Since tumors accumulate mutations which increase their resistance to certain environmental apoptosis triggers like hypoxia as well as to classical antineoplastic therapies, it is crucial to learn how to manipulate the downstream apoptosis machinery in new, perhaps more direct, ways. Finding new therapeutic triggers which induce tumor cell apoptosis in a manner independent of p53 has important clinical implications and remains a central focus in this area of apoptosis research. Similarly, shifting the balance of Bcl-2 family proteins in tumor cells may also trigger cell death. Paclitaxel (also called taxol or taxotere), a drug which acts p53-independently in patients, may exert its cytotoxic effects through its ability to phosphorylate and inhibit Bcl-2 downstream of p53 (71). Interestingly, Bax has been shown to enhance the cytotoxicity of chemotherapeutics like paclitaxel, vincristine and doxorubicin (but not etoposide or hydroxyurea) in a p53-independent fashion (292). In addition, new drugs like Apoptin are being discovered which, at least in vitro, appear to kill transformed cells which lack p53 and overexpress Bcl-2 (293). Understanding how a tumor cell's environment makes it more or less susceptible to certain drugs may also add therapeutic strategies. For example, both hypoxia and the ECM have been shown to affect the cell's response to cytotoxic drugs or radiation.

How might ECM affect a cell's capacity to respond to classical antineoplastic therapies? An early study shows that intercellular contact can modify the cell's ability to repair DNA following radiation-induced damage (294). In some instances, blocking intercellular contact using integrin-targeted monoclonal antibodies is sufficient to induce apoptosis (227). Similarly, cells in contact with basement membrane ECM are more resistant to conventional therapies than ones lacking adhesion (295-298). The survival pathways activated upon ECM adhesion which have recently been the focus of anoikis research may therefore provide a broad resistance to cytotoxic treatments. Topotecan - a topoisomerase I inhibitor - is significantly more effective in inducing apoptosis in a variety of tumor lines when cells are treated in suspension (299). Therefore, therapeutic agents which specifically target tumor cell adhesion mechanisms similar to the cyclic RGD peptides used as angiogenesis inhibitors could potentially synergize with other apoptotic triggers. In combination with 'classical' therapies such as radiation or chemotherapeutics, these may potentiate their cytotoxic effects in vivo by downregulating survival signals like Akt (138) or Bcl-2 (222).

Hypoxia is another clear example of a microenvironement which directly determines the efficacy of certain treatments like radiation in vivo. Radiation is known to require oxygen in order to fix the DNA damage, a process referred to as the "Oxygen Effect" (300). Free radicals produced on the DNA can either decay or, in the presence of oxygen, form DNA-O2 intermediates which typically result in double-strand breaks. Therefore, areas of low oxygen tension in solid tumors (caused by chronic or acute hypoxia) have been observed to be more resistant to the cytotoxic effects of radiation (301). Radiosensitizers - compounds which act as oxygen mimetics and potentiate the effects of radiation in such hypoxic environments - have been used in order to circumvent this problem (302, 303), and are under active clinical investigation.

Another approach whereby therapies may halt tumor cell proliferation and induce apoptosis has been to target the activity of dominant oncogenes. Farnesyltransferase inhibitors block the activity of the enzyme farnesyltransferase which modifies a number of small G proteins like Ras and trigger localization to the plasma membrane (304). Since this localization is required for Ras' signaling activities, inhibiting this process was hoped to abrogate its signaling potential and inhibit Ras-induced transformation. These inhibitors may also block Rho, another small G-protein involved in actin cytoskeletal organization (305). Other studies have confirmed a role for Rho in Ras-induced transformation (306) and shown that Rho can induce apoptosis via the production of ceramides (307). These studies indicate that other downstream effectors of Ras may also provide good therapeutic targets. The fact that fibroblasts forced into suspension appear more sensitive to farnesyltransferase inhibitors than adherent ones (308) again points to the importance of combining drugs with different targets in order to more potently induce apoptosis in a transformation-selective way.