The INK4a gene encodes two distinct growth inhibitors—the cyclin-dependent kinase inhibitor p16Ink4a, which is a component of the Rb pathway, and the tumor suppressor p19Arf, which has been functionally linked to p53. Here we show that p19Arf potently suppresses oncogenic transformation in primary cells and that this function is abrogated when p53 is neutralized by viral oncoproteins and dominant-negative mutants but not by the p53 antagonist MDM2. This finding, coupled with the observations that p19Arf and MDM2 physically interact and that p19Arf blocks MDM2-induced p53 degradation and transactivational silencing, suggests that p19Arf functions mechanistically to prevent MDM2's neutralization of p53. Together, our findings ascribe INK4a's potent tumor suppressor activity to the cooperative actions of its two protein products and their relation to the two central growth control pathways, Rb and p53.
Growth control in mammalian cells is accomplished largely by the Rb protein regulating exit from the G1 phase (Weinberg 1995) and the p53 protein triggering growth arrest/apoptotic processes in response to cellular stress (Levine 1997). Cross-talk between these two regulatory pathways may be mediated through the p21 cdk inhibitor, which is a target of p53 transactivation as well as a factor that influences the functional status of Rb (Weinberg 1995). An additional level of overlap between p53 and Rb is provided by the MDM2 protein that can physically associate with both proteins and prevent their growth suppression (42 and 65). In tumorigenesis, Rb and p53 appear to serve collaborative roles as evidenced by the observations that many tumor types exhibit mutations in both Rb and p53 ( Williams et al. 1994), and mice that are Rb (+/−) and p53(−/−) develop a wider range of tumors at earlier ages than mice that are either Rb (+/−) orp53 (−/−) ( Williams et al. 1994). Moreover, the ability of several viruses to transform cells in culture and cause tumors in mice is due to viral oncoproteins that bind to and inactivate both Rb and p53 ( 40, 23, 44 and 60). The mechanistic basis for this dual requirement stems in part from the deactivation of a p53-dependent cell suicide program that would normally be brought about as a response to unchecked cellular proliferation resulting from Rb-deficiency ( 19, 32 and 37).
p53 mutation is thought to be the most frequent genetic alteration in human cancers (Hollstein et al. 1991; Levine et al. 1991). In proliferating normal and neoplastic cells, the consequences of p53 overexpression are context-dependent, resulting in either cell cycle arrest or induction of apoptosis ( Ko and Prives 1996). These biological end points provide a basis for p53's antioncogenic actions ( 12 and 17) and have been shown to relate to its capacity to function as a sequence-specific transcription factor ( 10 and 49) and to interact with key cellular proteins. The critical role served by p53 in these diverse physiological processes necessitates that p53 activity be subject to stringent multilevel regulation. One crucial level of regulation involves the MDM2 protein, whose direct interaction with p53 blocks p53-mediated transactivation ( Chen et al. 1995) and targets the p53 protein for rapid degradation ( 22, 33 and 37). MDM2 itself has been shown to be amplified in primary tumors ( Oliner et al. 1992), to act as an immortalizing oncogene in cell culture ( Finlay 1993), and to directly repress basal transcription ( Thut et al. 1997).
In human cancers, disruption of the Rb pathway can result from inactivation of Rb itself through gene mutation/deletion, viral sequestration or hyperphosphorylation ( Weinberg 1995), or through disregulation of the components controlling the degree of Rb phosphorylation. The latter can take place through activating mutations in the G1-specific cyclin-dependent kinase 4 (CDK4) catalytic unit, up-regulation of D-type cyclin levels, and/or elimination of INK4 s (for inhibitors of cyclin-dependent kinase 4) ( Sherr 1996). The products of INK4 family genes have been shown to bind to CDK4 and inhibit CDK4-directed phosphorylation of Rb ( 56 and 50), thereby blocking exit from the G1 phase of the cell cycle ( Sherr 1996). One member of the INK4 family, INK4a, has been shown to exhibit loss of function in a wide spectrum of tumor types; this pathogenetic event appears to be exceeded in frequency only by p53 inactivation. The basis for the prominence of INK4a, as opposed to other members of the INK4 family, in tumor suppression is not fully understood but may relate to its unusual capacity to encode two distinct proteins—the cyclin-dependent kinase inhibitor, p16INK4a, and a novel protein of unknown function, p19ARF. This special feature of INK4a results from an unusual gene organization in which the two INK4a gene products are encoded by different first exons and alternative reading frames residing in a common second exon. The fact that both gene products are often eliminated or mutated in many cancers has raised questions regarding their relative contributions to INK4a-mediated tumor suppression.
Compelling support for p16INK4a as a critical target of tumorigenesis includes germline mutations/deletions exclusively affecting the p16INK4a ORF in melanoma-prone kindreds and a tumor-associated CDK4 mutation rendering this kinase insensitive to p16INK4ainhibition ( Zuo et al. 1996). With regard to p19ARF, although direct evidence linking loss of p19ARF function with human tumorigenesis has been lacking, many INK4amutations/deletions map to the exon 2 region that is shared by p19ARF, and a p19ARF-specific knockout leads to spontaneous tumor formation in mice ( Kamijo et al. 1997).
Some clues addressing p19ARF's mechanism of action have been provided by the requirement for p53 in p19Arf-induced G1 arrest and by an absence of p53 mutations in postcrisis p19Arf (−/−) MEF cultures ( Kamijo et al. 1997) and in RAS-induced melanomas arising in the Ink4a null mice ( Chin et al. 1997). Additionally, studies reported here suggest that p19Arf requires p53 function to suppress cellular transformation. All of these observations have led to the intriguing possibility that the INK4a gene is linked not only to the Rb pathway through p16INK4a but also to the p53 pathway through p19ARF. Along these lines, our studies demonstrate that p19Arf engages the p53 pathway through physical interactions with the MDM2 oncoprotein. In addition, we show that p19ARFinhibits the oncogenic actions of MDM2, blocks MDM2-induced degradation of p53, and enhances p53-dependent transactivation. Finally, we demonstrate that loss of Ink4aattenuates apoptosis brought about by Rb deficiency. These studies provide physical and mechanistic insight fortifying Ink4a's position at the nexus of the two most important tumor suppressor pathways governing the development of neoplasia and provide an explanation for the frequent involvement of Ink4a in tumorigenesis.
Distinct and Cooperative Effects of p16Ink4a and p19ARF in the Suppression of Primary Cell Transformation
The antioncogenic potencies of the two Ink4a gene products were tested in the rat embryo fibroblast (REF) cotransformation assay ( Land et al. 1983) against various oncogene combinations (e.g., Myc/RAS, E1a/RAS, or SV40 Large T Antigen (T-Ag)/RAS). This approach has been used extensively to provide a quantitative measure of antioncogenic activity and allows for placement of these activities along known growth control pathways ( 54, 1 and 35). In the first series of experiments, we investigated the degree of inhibition of E1a/RAS- versus Myc/RAS-induced foci formation by p19Arf,p16Ink4a, or both. As shown in Figure 1, addition of mouse p16Ink4a induced a 1.7- to 3-fold reduction in foci number when added to c-myc/RAS transfections (panel A, p16Ink4a) and failed to cause a statistically significant decrease in E1a/RAS foci counts (panel B, p16Ink4a); these results are identical to our previous report for the human p16Ink4a (Serrano et al. 1995). Since E1A inactivates the Rb protein, the failure of p16Ink4a to suppress E1/RAS transformation is as expected ( 39, 41 and 57). In the same cotransfection experiments, addition of p19Arf resulted in marked foci reductions in c-myc/RAS (5- to 10-fold) as well as E1a/RAS (4- to 5-fold) cotransfections (panels A and B, p19ARF). E1a/RAS inhibition by p19Arf was not further augmented by the addition ofp16Ink4a (panel B, compare p19Arf and p16Ibk4a + p19Arf). In contrast, coaddition ofp16Ink4a and p19Arf resulted in a complete inhibition of c-myc/RAS transformation. Thus, the distinct activity profiles of p16Ink4a and p19Arf (i.e., E1a/Ras transfections), together with their additive effects in the c-myc/RAS transfections, suggests that these proteins suppress neoplasia through separable but cooperative mechanisms of action (see below).
p19Arf Suppression of Transformation in Primary Rodent Cells
(A) Cooperative effects of the mouse p16Ink4a and p19Arf expression constructs in Myc/RAScotransformation assays. Histogram of a representative REF cotransformation assay showing the average number of foci per 10 cm plate following cotransfections with 2 μg mouse c-myc, H-RASval12, and the various expression constructs listed above the error bars.
(B) Distinct actions of p16Ink4a and p19Arf expression constructs in E1a/RAS cotransformation assays. The same experimental design as described in (A) except that each plate was cotransfected with 2 μg E1a, H-RASval12, and the various expression constructs listed. In this particular experiment, the p16Ink4atransfection point exhibited an unusually low number of foci relative to the empty vector. Although this decline is not statistically significant, in all other experiments the addition of mouse p16Ink4a had no inhibition against E1a/RAS transformation, similar to our previous studies with the human p16Ink4a (Serrano et al. 1995). Support for the lack of an effect also derives from the lack of additional suppression by p16Ink4a in the p16Ink4a + p19Arf transfection point compared to p19Arf alone. The general cyclin-dependent kinase inhibitor p21Cip1 served as a positive control for an inhibitory agent acting downstream of Rb.
(C) Antioncogenic activity of p19Arf in T-Ag/RAS or dominant-negative p53/RAS REF cotransformation assays. On the left, histogram of a representative REF cotransformation assay showing the average number of foci per 10 cm plate following cotransfections with 2 μg each of T-Ag, H-RASval12, and empty vector or p19Arf. On the right, histogram showing the average number of foci per 10 cm plate following cotransfections with 2 μg each of p53KH215 (encoding a dominant-negative mutant p53) and H-RASval12with or without p19Arf.
(D) Antioncogenic activity of p19ARF in Myc/RAS MEF cotransformation assays. The early passage MEFs used for each experiment were either null for Ink4a (left panel) or null for both Ink4a and p53 (right panel). The bars represent the number of foci generated in the presence of p19Arf relative to control plates receiving 2 μg c-myc, 2 μg RAS, and 2 μg empty vector. These assays were performed on an Ink4a null background because wild-type MEFs do not give clear, countable foci in Myc/RAS cotransformation assays.
Functional p53 Is Required for Full Oncogenic Suppression by p19Arf
The cell cycle inhibitory effects of p19Arf in primary MEF cultures have been shown to be p53-dependent (Kamijo et al. 1997). To examine the possibility that p19Arf may also act in a p53-dependent manner to suppress cellular transformation, we employed cells rendered functionally (T-Ag or dominant-negative p53) or genetically [p53 (−/−)] deficient for p53 in transformation assays. The addition of p19Arf to T-Ag/RAS cotransfections was found to have no effect on the number of foci generated in the REF assay ( Figure 1C) or on the morphological/growth characteristics of these foci (data not shown). Since T-Ag is known to engage many pathways beyond p53 ( 15 and 62), we next assessed the ability of p19Arf to suppress transformation in two other contexts. First, in comparison to the addition of an empty vector control, the addition of p19Arf did not affect the number of foci-generated in cotransfections of a dominant-negative mutant form of p53 (p53KH215) andRAS in the REF assay ( Figure 1C). Second, potent p19Arf-induced suppression ofMyc/RAS foci formation was observed in early passage Ink4a (−/−) mouse embryonic fibroblasts (MEFs), but this suppression was completely eliminated in MEFs doubly null for Ink4a and p53 ( Figure 1D). These results strongly suggest that p19Arf does not act in a nonspecific cytotoxic manner to reduce foci formation in the Myc/RAS and E1a/RASexperiments described above. Instead, these results appear to assign specificity to the antioncogenic actions of p19Arf. More specifically, in accord with the recently reported cell cycle studies ( Kamijo et al. 1997), these findings support the hypothesis that p19Arfacts in a p53-dependent manner to inhibit cellular transformation.
p19Arf Associates with MDM2 In Vivo
To gain insight into the mechanistic basis for the functional link between p19Arf and the p53 pathway, coimmunoprecipitation experiments were performed to assess potential physical interactions between p19Arf and p53 or the p53-associated protein, MDM2. Since endogenous levels of these proteins are very low in normal primary cells (Levine 1997), the composition of the p19Arf complexes was determined following cotransfection of various expression constructs (including one encoding a Flag epitope-tagged p19Arfprotein, p19Flag) or through the use of different tumor cell lines expressing some or all of these proteins. As shown in Figure 2A, IP-Western blot assays readily detected p19Flagin anti-p53 immunoprecipitates following cotransfection with p53, MDM2, and p19Flag(lane 2) but not with p53 and p19Flag (lanes 3 and 4). The requirement of MDM2 overexpression to reveal a p53–p19Flag interaction was also observed following either anti-p53 or anti-Flag immunoprecipitations of metabolically labeled transfected cells (data not shown). These results demonstrate that p53, MDM2, and p19Flag can exist as components of a multiprotein complex in vivo. Moreover, the requirement for abundant MDM2 to detect p53–p19Flag interaction suggested that MDM2 serves as a bridging molecule, or that MDM2 induces changes in steady-state levels of p19Flag, among other possibilities. The possibility that MDM2 overexpression stabilizes the level of p19Flag was ruled out by Western blot analysis showing equivalent levels of p19Flag in 293T cells following transfection of MDM2 and p19Flag or of p19Flag alone (data not shown). Moreover, although MDM2 can target p53 for degradation in some cell types (22 and 33), the levels of endogenous p53 in 293T cells remain constant following cotransfection and overexpression of MDM2 (data not shown) due to the presence of T-Ag ( Henning et al. 1997).
To examine more directly whether p19Flag can associate with MDM2, coimmunoprecipitation studies were conducted in metabolically labeled 293T cells and in 3T3DM (amplified for Mdm2) and SAOS2 cells (null for p53). In the 293T cells (Figure 2B and Figure 2C), MDM2 was readily detected in anti-Flag immunoprecipitates following cotransfection with p19Flag and MDM2 (lane 8) but not with either empty vector (lane 6), p19Flag alone (lane 5), or MDM2 alone (lane 7). Correspondingly, anti-MDM2 immunoprecipitations confirmed the MDM2–p19Flag association in the p19Flag andMDM2 cotransfections (lane 12). In addition, the endogenous p19Arf band was present in the anti-MDM2 immunoprecipitates (lane 9) and the signal intensity of this band diminished upon cotransfection of p19Flag (lane 12), this likely due to competition for a common binding site in the MDM2 complex. In each of these experiments, Western blot analyses of lysates that were run in parallel confirmed the identity of p19Flag and MDM2 bands (data not shown). The interaction between p19Flag and MDM2 in 293T cells was also demonstrated by coimmunoprecipitation in both low- and high-stringency conditions yielding identical results ( Figure 2C, lanes 18 and 19, respectively).
To address whether the interaction of T-Ag with MDM2 and p53 (Brown et al. 1993) alters the composition of MDM2/p53/p19Arf complexes in T-Ag-expressing 293T cells, we examined the interactio