Cancer Therapy Vol 1, 343-351, 2003.

Sidney Kimmel Cancer Center, 10835 Altman Row, San Diego, CA 92121

__________________________________________________________________________________

*Correspondence: Ruth, A. Gjerset, Ph.D., Sidney Kimmel Cancer Center, 10835 Altman row, San Diego, CA 92121; Telephone: 858-450-5990; FAX: 858-450-3251; email: rgjerset@skcc.org

Key Words: p14^ARF, p53, mdm2, adenoviral vectors, therapy resistance

ARF plays a key role in the p53/mdm2 stress response mechanism that is central to cancer. This mechanism serves as a point of convergence for a variety of stress signals encountered by the cancer cell, including environmental stress (radiation, exposure to chemotherapeutic drugs, hypoxia), and endogenous oncogenic stress (oncogene activation, genome instability and the ensuing DNA damage) . In response to these signals, the mechanism promotes the accumulation of p53 protein, followed by induction of p53 downstream events leading to cell cycle arrest or apoptosis. We now know that this mechanism is disrupted in virtually all cancers (see below), usually through deletion or mutation of p53 in more than 50% of cancers , or through deletion or promoter methylation of ARF in about 40% of cancers . Disruption of this mechanism impairs the cancer cell's ability to trigger cell cycle arrest or apoptosis in response to oncogenic abnormalities, enabling it to manifest its transformed phenotype, or in response to environmental stress, which can contribute to drug resistance. The high frequency with which this mechanism is targeted for disruption in cancer suggests that an intact mechanism cannot be easily bypassed and is incompatible with cancer cell growth. The mechanism provides therefore an important point of focus for therapeutics development, as treatments that restore a fully functioning p53/mdm2 regulatory mechanism would be expected to be highly suppressive of most cancers, while having minimal consequences for normal cells that lack the endogenous oncogenic stress signals that feed into the mechanism. In fact, ectopic overexpression of wild-type p53 is highly suppressive of cell growth and viability of most cancer cells but is minimally suppressive of normal epithelial cells and fibroblasts . Even in cancer cells that retain expression of endogenous wild-type p53 and would likely have defects in other components of the regulatory mechanism, overexpressed ectopic wild-type p53 can have suppressive effects on growth and viability . P53 gene therapy has therefore been investigated as a promising targeted approach to cancer treatment, applicable to a variety of cancers. Several clinical applications of p53 gene therapy have now entered Phase III trials . With the identification of ARF as a key player in modulating p53 activity, and possibly a critical determinant of p53 activity, it is likely that further refinements in this approach will be possible.

Figure 1 diagrams our current understanding of the interactions between ARF, mdm2 and p53. ARF regulates the interaction of p53 with mdm2, a key negative regulator of p53 . Mdm2 (also termed hdm2 in humans) ubiquitinates p53 and targets it for proteosome-mediated degradation. Through this mechanism, mdm2 ensures low steady state levels of wild-type p53 found in normal cells. Mdm2 expression in turn is induced by p53, establishing a feedback loop that controls expression levels of both proteins . Mdm2 function appears to be largely p53-dependent as mdm2 overexpression in transgenic mice results in embryonic lethality, but lethality can be rescued by p53 deletion . Mdm2 overexpression occurs in about 10% of cancers, predominantly in soft tissue sarcomas . Such tumors generally continue to express wild-type p53, as would be expected if overexpression of mdm2 were the functional equivalent of p53 loss.

Figure 1. Schematic representation of the stress-activated regulatory loop involving p53, mdm2, and p14ARF.

necessary to induce p53-mediated cell cycle arrest in response to these abnormalities. The observation that p53 and ARF abnormalities often occur in a reciprocal manner in human cancer is consistent with this p53-dependent model for ARF activity and suggests that ARF loss is to some extent the functional equivalent of p53 loss . However, unlike mdm2, ARF function does not appear to be confined to a largely p53-dependent role, and is now known to interact with proteins other than mdm2, some of which may mediate important tumor suppressor activities of ARF.

In addition to its well-described involvement in regulating p53 activity, several lines of evidence now support the involvement of p53-independent activities of ARF in tumor suppression. In some cancers, for example, ARF loss and p53 mutation occur together , suggesting that inactivation of ARF per se constitutes an independent survival advantage for the cancer cell. Evidence for a p53-independent role for ARF in vivo is also suggested by studies of knockout mice, where ARF-mediated suppression of spontaneous tumor formation in p53-null, mdm2-null mice was inferred based on the decreased predisposition of such mice to spontaneous tumor formation compared to mice lacking ARF as well as p53 and mdm2 . This effect is also observed in vitro, where mouse embryo fibroblasts (MEFs) from mice triply nullizygous for p53, mdm2, and ARF are growth suppressed by ectopic re-expression of ARF, indicating that ARF is able to suppress growth independently of p53 in this setting . It has been speculated that in such a setting, ARF may interact with mdmx (discussed below) or another mdm2 homologue to reverse their inhibition of a yet to be identified growth suppressor .

Another study using MEFs implicated both the Rb and p53 pathways in ARF activity . In that study MEFs lacking the p53 pathway could be suppressed by ectopic expression of ARF, but MEFs lacking both p53 and Rb pathways could not be suppressed by ectopic expression of ARF . In our own studies in human tumor cell lines, we see that ectopic overexpression of ARF suppresses growth and viability of human tumor cells lacking expression of endogenous wild-type p53 . Unlike tumor cells expressing endogenous wild-type p53, ectopic overexpression of ARF in tumor cells lacking expression of endogenous wild-type p53 does not lead to p53 accumulation, or to the induction of p53 target gene expression (mdm2, p21waf1, bax), or to an elevated bax to bcl2 ratio . ARF is therefore able to suppress tumor cell growth independently of these key players in p53-dependent growth arrest and apoptosis. Furthermore, since the tumor lines studied lacked a functional Rb pathway as well , ARF appears to have tumor suppressor activity in human tumor cells that is independent of both p53 and Rb.

ARF binding proteins other than mdm2 have been identified through yeast two-hybrid screening techniques, co-immunoprecipitation, and co-localization studies. These proteins, listed in Table 1, include topoisomerase I , E2F-1,2,3 , spinophilin, the regulatory subunit of protein phosphatase I , p120^E4F, a zinc finger transcription repressor , CARF, a novel collaborator of ARF , mdmx, a mdm2 homologue , and HIF-1a , a hypoxia-inducible transcription factor . ARF has also been found to form oligomers with itself . Two of the ARF binding proteins listed in Table 1, mdmx and HIF-1a , may participate in p53 independent functions of ARF. Hypoxia-inducible factor-1a (HIF-1a ), a transcriptional regulator of genes induced during hypoxia, is frequently over expressed in advanced tumors and may play an important role in tumor growth. ARF can inhibit the transcriptional activity of HIF-1a by binding to it and sequestering it into the nucleolus. The interaction between ARF and HIF-1a is independent of both mdm2 and p53, and thus may define a p53 independent role for ARF . Mdmx has been implicated in the p53-dependent activity ofARF, participating through interactions with mdm2, ARF and p53, in the p53/mdm2 regulatory feedback loop, and inhibiting transcriptional transactivation by p53 . Through its association with ARF, mdmx appears to interfere with ARF-mdm2 interactions . In the presence of wild-type p53, and under conditions of oncogenic stress where p53 and mdm2 levels are induced, mdmx may play a minor role.

However, under conditions where p53 and mdm2 are low or absent (i.e., triple knockout MEFS discussed above, or possibly certain human tumor lines), associations between mdmx and ARF may mediate tumor suppressor activities independently of p53 .

Table 1. p14ARF binding proteins

The ARF tumor suppressor is encoded by the ARF-INK4a locus of chromosome 9p21, a region that has already been implicated in human cancer as a very frequent target of genetic alteration . Also encoded by this locus is the p16 (MTS1/CDKN2) tumor suppressor, an inhibitor of cyclin D-dependent kinases CDK4 and CDK6 . Phosphorylation of the retinoblastoma tumor suppressor (pRb) by CDK4 or CDK6 releases the E2F transcription factor from an inactive complex with pRb, and enables downstream events that promote cell cycle progression. Inactivation of either p16 or pRb deregulates cell cycle progression, a key feature of tumor cell metabolism. While p16 and ARF are structurally and functionally unrelated and have distinct first exons (denoted exon 1a and exon 1b , respectively), they are characterized by the unusual property of sharing the same exon 2, although alternate reading frames are used to generate the two proteins.

Studies on murine ARF (p19) have localized the p53-dependent tumor suppressor activity of ARF to the N-terminal exon 1b encoded sequences, which are completely non-overlapping with p16 sequences . In murine ARF the regions important for nucleolar localization, mdm2 sequestration, and p53-dependent apoptosis reside entirely within the first 37 N-terminal residues and these 37 residues are sufficient to carry out these effects . With regard to this activity, the C-terminal sequences might appear therefore to constitute a largely non-functional and dispensable domain. Because the N-terminal exon 1b -encoded sequences of murine and human ARF are highly conserved (particularly in the first 14 amino acids involved in mdm2 binding), it seemed likely that this conclusion would extend to human ARF as well. However, while the N-terminal exon 1b -encoded region of human ARF (ARF1b ) appears sufficient to bind mdm2 and to mediate the p53-dependent activity of ARF, several studies now raise the possibility that ARF C-terminal sequences, encoded by ARF exon 2, also contribute to its tumor suppressor properties. For example, human ARF has nucleolar localization sequences in both N terminal (exon 1b ) and C-terminal (exon 2) domains , and both may be required for efficient nucleolar localization . An example of the differences in subcellular distribution of ARF and ARF1b is shown in Figure 2 (Y Huang and R Gjerset, unpublished). MCF-7 breast cancer cells, which express endogenous wild-type p53 and lack endogenous expression of ARF were treated with replication defective adenoviral vectors encoding ARF1b or full length ARF (vectors described in . 48 hours post-vector treatment cells were fixed and stained with anti-p14ARF (Zymed laboratories) and goat anti-rabbit Alexa 488 (Molecular Probes) plus Hoechst 33345. The Hoechst stain binds DNA and facilitates the identification of nuclei and subnuclear structures. The results show that over expressed nuclear ARF is concentrated in nucleoli (Figure 2, first column, bottom panel), consistent with published studies . In contrast, nuclear ARF1b showed far less nucleolar concentration and distributed more evenly throughout the nucleus (Figure 2, second column, bottom panel). In addition, several studies on melanoma-prone kindreds suggest the involvement of ARF C-terminal sequences in ARF functions . In contrast to an earlier study that reported that

Figure 2. Subcellular distribution of ectopic ARF and ARF1ß in MCF7 breast cancer cells (endogenous ARF null). Cells were treated with adenoviral vectors encoding ARF1ß or full length p14ARF, followed 24 hours later by fixing and staining. Nucleoli, indicated by arrows in top panels, are evident by DIC (Differential Interference Contrast) and by Hoechst staining as darker regions (middle panels). Bottom panels show immunofluorescent staining for p14ARF. Magnification = 60x.

several cancer associated mutations affecting C-terminal (exon 2-encoded) sequences altered p16 function but not ARF function , the melanoma study identified mutations in the C-terminal nucleolar localization domain encoded by exon 2 in certain melanoma-prone kindreds that altered the subcellular distribution of ARF and/or its binding to mdm2 . Another study also reported a C-terminal mutation in ARF in a melanoma kindred that gave rise to a protein carrying an 8 residue deletion that was less potent than wild-type ARF at stabilizing p53 and inducing cell cycle arrest . Interestingly, the physical properties of these mutants were often altered as well, as several of them (R81Q, R82L, R88ter) showed greater solubility in the absence of urea than did full length wild-type ARF . C-terminal sequences therefore contribute to the proper folding of ARF and conceivably influence the intermolecular interactions and activity of ARF.

ARF C-terminal sequences have now been shown to be crucial for ARF interaction with topoisomerase I, for the ability of ARF to stimulate topoisomerase I relaxation activity, and for the co-localization of ARF and topoisomerase I in the nucleolus . Furthermore, two ARF C-terminal point mutations that are highly conserved among species were unable to stimulate topoisomerase I activity . Taken together the data argue for an important role for the ARF C-terminal sequences in regulating the structure and function of ARF, although the reason why p14ARF and p16 overlap remains unknown.

The disruption of the p53/p14ARF/mdm2 feedback loop is likely to be an essential survival mechanism common to most cancers, as it renders cancer cells unable to undergo growth arrest or apoptosis in response to the intracellular oncogenic stress signals that are universal features of cancer. One would predict that restoration of normal functioning of the loop would be highly suppressive of cancer cell growth and viability, in that it exploits the cancer cell's intrinsic fragility and predisposition to apoptosis, but would have minimal consequence for normal cells that lack the endogenous oncogenic stress signals that feed into the loop. The importance of ARF in regulating this loop, and the high frequency with which ARF function is lost in cancer, makes ARF an important therapeutic target.

Numerous therapeutic applications of the p53 tumor suppressor are now in an advanced stage of development, with clinical trials ongoing or completed that attest to the safety and efficacy of adenovirus-mediated gene transfer of p53 for a variety of tumors, including prostate cancer, lung cancer, and head and neck cancer, brain cancer, ovarian cancer, and bladder cancer (see references ). The vector of choice for these applications has been a replication defective vector based on adenovirus serotype 5, a serotype that lacks serious pathogenicity, has a broad host cell range, displays a high infection efficiency for most cell types, and is relatively easy to prepare in high yield and high titer, and is stable. Unlike retroviral vectors, adenoviral genomes do not integrate into the host cell DNA, thereby avoiding the risk of insertional mutagenesis, which is a concern with retroviral vectors. To minimize potential vector toxicity and to maximize the exposure of the target tumor cells to the vector, these trials have generally employed direct administration of the vector into a patient's tumor or into the region where a tumor is localized, with doses as high as 10¹¹ plaque forming units per injection being well tolerated . While patients may develop neutralizing antibody to the vector, expression of the p53 gene in tumors is still detectable, and antibody does not preclude multiple administrations of vector at weekly or biweekly intervals . An extension of such approaches to include ARF therefore seems feasible, and approaches that exploit both p53 and ARF may prove to be far more effective than approaches based on p53 alone.

The ultimate challenge for cancer treatment is metastatic disease, for which systemic delivery approaches are required. Systemic delivery of p53 using viral or non viral delivery strategies can have antitumor efficacy in mice and reduce the incidence of metastatic spread when combined with conventional therapy. However, the clinical application of systemic delivery approaches will likely require the development of improved second generation viral or non viral vectors, including vectors consisting of polyethylenimine (PEI)-DNA complexes , and possibly incorporating a tumor specific targeting moiety to improve delivery efficiency, a critical factor in the success of any gene transfer strategy. Because systemic gene delivery is likely to be far less efficient than direct intratumoral gene delivery, there is also a need to optimize gene combinations so as to improve therapeutic efficacy at low gene expression levels.

The activity of p53 is attenuated through a feedback mechanism whereby p53 induces the expression of its own inhibitor, mdm2 (Figure 1). This feedback mechanism, which may be exacerbated by certain cancer associated abnormalities such as loss of p14ARF or overexpression of mdm2, may compromise the outcome of p53-based strategies for cancer, including p53 gene transfer. Full activation of the pathway may therefore require additional steps to modulate the activity of mdm2. One approach is to supply both p53 and p14ARF, either as a combination of single gene vectors or as a bicistronic vector . Co-expression of p14ARF and p53 leads to a dramatic enhancement of tumor suppression compared to that achieved with p53 alone, as a result of increased accumulation of p53 protein followed by increased expression of p53 target genes, including mdm2, p21waf1, and bax, an elevated bax to bcl2 ratio, and apoptosis . P14ARF also appears to promote increased recruitment of p53 message into polysomes and increased synthesis of p53 protein, through a mechanism that is not yet understood and Huang and Gjerset, unpublished). The cooperative effects of ectopically expressed p53 and p14ARF are observed even in cells expressing endogenous ARF, indicating that the presence of endogenous ARF is likely to be insufficient to counterbalance the effects of induced levels of mdm2. The use of an optimized gene combination involving p53 together with p14ARF, could therefore enhance efficacy under conditions where gene delivery was inefficient or resulted in low levels of transgene expression.

Eventually, small molecule approaches targeting specific molecular structures or interactions may provide alternatives to gene delivery when systemic delivery is needed for metastatic disease. One possibility is the use of peptides corresponding to regions of the p53 protein or to segments of p53-binding proteins so as to correct the folding of p53 mutant proteins and reactivate function . A better understanding of how ARF contributes to the pathology of cancer, including a mapping of regions necessary for function and for interaction with itself and other proteins may make it possible to design ARF peptide mimetics or other small molecule therapeutics, to mimic protein-protein interactions critical for ARF activity, or disrupt interactions that impede its activity . Such molecules could be administered systemically, possibly tagged with moieties to facilitate cellular uptake .

A variety of anticancer agents, including cisplatin (cis-diamminedichloroplatinum, CDDP) and doxorubicin (adriamycin), generate lesions that promote p53-mediated apoptosis, and that these agents are more effective in vitro and in vivo in animal tumor models, if p53 function is also restored in these tumors . There is also evidence that loss or mutation of p53 in cancer contributes to therapy resistance , currently the major obstacle to successful treatment of cancer. The fact that many chemotherapeutic agents use the p53 pathway to kill their target cells has now been exploited in clinical trials of several cancers, including advanced head and neck cancer, where p53 gene transfer was seen to improve responses to cisplatin or radiation . These observations are consistent with the well-described involvement of p53 in the DNA damage response , which leads to phosphorylation of mdm2 and disruption of the interaction between p53 and mdm2. In this way, p53 is stabilized and activated as a transcription factor, followed by induction of downstream target genes involved in growth arrest (p21waf1) or apoptosis (bax) . P53-mediated senescence, or irreversible loss of proliferative potential, has also been associated with the outcome of therapy, and correlates with increased expression of senescence-associated b -galactosidase .

In light of the important role played by ARF in the opposing mdm2-mediated degradation of p53, it is likely that ARF gene transfer will be required to optimize the outcome of p53-chemotherapy combination approaches. In fact, deletion of the INK4A/ARF locus in Em -myc murine lymphoma cells impairs apoptosis, compromises p53 function, and reduces the response to the DNA alkylating agent cyclophosphamide in vitro and in vivo . Other studies have shown that ARF expression, though dispensable for initiating a p53 response to DNA damage, is required for a sustained accumulation of p53 following DNA damage . ARF gene transfer alone can sensitize breast cancer cells to doxorubicin, a drug that acts in part through the induction of DNA damage , or to cisplatin, which forms bifunctional adducts on DNA . A role for ARF in the cellular response to these chemotherapeutic agents would greatly expand its potential application to cancer treatment, as it could be used to help reverse therapy resistance in advanced cancers.

ARF has emerged as a critical human tumor suppressor with relevance to the underlying mechanism of cancer and to cancer treatment. Through its involvement in the p53/mdm2 feedback loop, which shuts down cell growth in response to oncogenic abnormalities, ARF plays a key role in suppressing a broad array of cancers that retain wild-type p53 expression. ARF may also be necessary to optimize p53-based therapies and to fully activate the p53-mediated response to conventional chemotherapy. In addition, through pathways that are independent of p53 and less well understood, but which presumably involve ARF interactions with proteins other than mdm2, ARF may contribute additional tumor suppressor activities that add to its p53-dependent activity. As we gain a better understanding of these activities and how they are regulated we will certainly improve and refine our paradigms to understand cancer and find new opportunities for the development of highly targeted cancer therapeutics.

Our work in this field has been supported by grants from the California Tobacco-Related Disease Research Program (11RT-0074), the California Cancer Research Program (99-00517V-10140), the California Breast Cancer Research Program (6JB-0077), and the National Cancer Institute (CA69546).