Mechanisms of oncogene cooperation: activation and inactivation of a growth antagonist.

Gene transfer experiments have defined limitations with regard to the ability of individual oncogenes to transform cultured cells to a tumorigenic state. The stable transformation of REF52 cells by either the ras or sis oncogenes requires the continuous expression of a second collaborating oncogene, such as adenovirus-5 E1A or SV40 large T-antigen. Our studies suggest that the function of the nuclear collaborators is to antagonize dominant growth controls which limit the ability of REF52 cells to proliferate in response to mitogenic stimuli.


Introduction
Oncogenes are genes implicated in carcinogenesis by virtue of their associations with oncogenic viruses and tumor-specific chromosome abnormalities and by their ability to transform cultured cells to a tumorigenic state (1,2). Mutations and virus associations enable oncogenes to facilitate autonomous cell growth either by increasing the expression of activities that promote cell proliferation or by interfering with controls that normally serve to restrict cell growth. As carcinogenesis is generally believed to be a multistep process, malignant phenotypes may frequently require multiple alterations affecting several levels of growth control (3). This is supported by gene-transfer experiments in which two or more oncogenes are required to transform normal cells to a malignant state (4)(5)(6).
Collaborations between oncogenes indicate that regulatory mechanisms that preclude transformation by one oncogene can be circumvented by a second oncogene. This implies that certain oncogenes actively influence the way that cells respond to other oncogenes. Gene transfer experiments indicate that expression of an activated ras oncogene is sufficient to transforn cells from a variety of established cell lines (7)(8)(9). In contrast, primary and early passage rodent cells are transformed by ras oncogenes only at very low frequencies unless cotransfected with collaborating genes such as ElA or myc (9,10).
Several observations suggested that in vitro establishment is a necessary prerequisite to transformation by ras oncogenes. First, the susceptibility of normal hamster cells to transformation by ras coincides with in vitro establishment (9). Second, genes such as ElA, p53, SV40 large T-antigen, and myc, which collaborate with ras to transform primary cells, also facilitate immortalization (11)(12)(13)(14)(15). Finally, mutational analysis of ElA revealed a tendency for mutations that affect in vitro establishment functions to also influence the ability of ElA to collaborate with ras in the transformation of primary cells (16,17). However, some ElA genes are defective in immortalization but are nevertheless able to collaborate with ras in transformation (17)(18)(19).
Though informative, the use of primary cells in transformation studies has several limitations. First, a variety of factors including variations in cell type, levels of ras expression, inhibitory effects of normal cells, and other environmental factors can influence transformation by ras alone (20)(21)(22)(23)(24). Second, the fact that immortalization appears to be a necessary prerequisite for oncogenic transformation in vitro may simply reflect difficulties imposing a transformed phenotype over a dominant commitment to senesce. Finally, problems working with nonestablished cell clones complicate any assessment of the phenotypic effects of ras alone. As a result, the mechanisms responsible for oncogene cooperation in nonestablished cells have been difficult to analyze.
We have examined oncogene interactions using an established line of Fisher rat embryo fibroblasts, REF52, which, unlike most established lines, is not stably transformed by activated ras oncogenes. Instead, transformation of REF52 cells by ras strictly requires the presence of a collaborator, such as adenovirus ElA or SV40 large T-antigen (25). The transformation properties of the REF52 line indicate that the role of the viral oncoproteins extends beyond that of promoting establishment and suggests that while immortality may be a necessary prerequisite for transformation by ras, it is not sufficient. While other established cell lines resist transformation by ras (14,26), the REF52 line provides the clearest evidence that oncogenes can complement in a genetic sense to elicit a malignant phenotype.

Multistep Transformation of REF52 Cells
In earlier studies, REF52 cells were transfected with T24-Ha-ras linked to the neomycin-resistance gene and clones were selected in G418. The vast majority of the resistant clones were morphologically indistinguishable from the parental line, unable to grow in soft agar, not tumorigenic in nude mice, and expressed low levels of p21T24Ha-ras (30% ofthe level of endogenous Ha-ras p21). While 1 in 1000 clones was morphologically transformed, these rare ras-only transformants experienced morphological crisis and ceased growing. p21T24Ha-ras levels in these abortive transformants were extremely high (greater than 100-fold higher than the endogenous Haras p21), indicating that even high levels ofthe activated ras oncoprotein are not sufficient to induce stable transformation in the absence of a collaborating oncogene (25).
In contrast, cells transfected with T24Ha-ras/neo and either adenovirus E1A or SV40 large T-antigen, gave rise to transformed neo-resistant colonies (25,27). While p21T24Ha-r,S levels in these transformed clones varied considerably, ranging from levels 2to 10-fold greater than the endogenous p21l-Ha-ra8 all of the transformed clones tested formed colonies in soft agar and induced rapidly lethal tumors in nude mice. These studies indicate that relatively low levels of the oncogenic p21 are sufficient to transform in the presence of a collaborator and that the collaborating oncogene supplies activities that are not duplicated by increased ras expression.
Conditional transformation of REF52 cells has also been achieved by cotransfecting a temperature-sensitive allele of SV40 large T-antigen (tsA58) and T24 Haras (27). Thus, two-thirds of the clones transfected with T24 Ha-ras and tsA ceased growing, arresting predominantly in G2 or late S-phase, when transformed from a permissive to a nonpermissive temperature for T-antigen expression. p21i24 Ha-ras levels were unaffected by the temperature shift. This study indicates that transformation by ras requires continuous expression of a collaborating oncogene, and suggests that E1A and SV40 large T-antigen may enable ras to transform REF52 cells by circumventing cellular responses to ras which inhibit cell proliferation.
Oncogenic p21 Induces Quiescent REF52 Cells to Proliferate To investigate early phenotypic effects of ras, activated and normal forms of p21 were introduced into serum-arrested REF52 cells using the bead-loading method (28). The method utilizes small (75-500 ,um) glass beads to mechanically disrupt the plasma membrane to an extent that allows proteins to enter cells without significantly affecting cell viability. p21Ha,-s proteins were purified from insect cells infected with baculovirus vectors expressing high levels of posttranslationally modified T24 and c-Ha-ras p2ls. As shown in Figure 1, up to 31% of the cells loaded in the presence of a 2.2 mg/mL solution of the activated p21 were stimulated to enter S phase within 24 hr, whereas cells loaded with bovine serum albumin failed to respond. Introduction of nonactivated c-Ha-ras p21 was less effective in stimulating cell cycle progression than activated p21 (data not shown). Maximum incorporation of 3H-thymidine was observed 12 to 24 hr after exposure to bead loaded p21T24Ha-ras. At concentrations greater Protein Concentration (mg/ml) FIGURE 1. Bead-loaded T24ras p21 stimulates serum-arrested REF52 cells to progress through the cell cycle. REF52 cells, seeded on glass coverslips, were transferred to Dulbecco's modified Eagle's media containing 1% ITS+ (insulin, transferrin selenium, and oleic acid/bovine serum albumin complex). After 48 hr, growth-arrested REF52 cells were washed with phosphatebuffered saline (PBS), covered with 50 ,L PBS containing varying amounts of T24ras p21, or bovine serum albumin (BSA), and were bead-loaded using 75-to 150-Rm beads, as described (28). BSA loaded at 3.6 mg/mL resulted in 0.1% labeled nuclei. Cells were labeled with 2.5 ,uCi/mL 3Hthymidine for 24 hr post-bead-loading and then fixed and processed for emulsion autoradiography. than 1 mg/mL, p21r24Ha-ras also induced morphological transformation as well as stimulating DNA replication. Similar results have been reported using microinjected p21 (29); thus, although resistant to transformation, REF52 cells are mitogenically stimulated by introduction of p21T24Ha-ras. In this regard, REF52 cells respond to oncogenic p21 in a manner not unlike cells that are susceptible to transformation (29,30). As ras is thought to transduce signals from growth factor receptors, controls that prevent ras from transforming REF52 cells would appear to act at a level subsequent to or independent of a mitogenic response to ras.
ElA Enables the sis Oncogene to Transform REF52 Cells Similar growth controls may also regulate transformation of REF52 by c-sis, the B chain of the plateletderived growth factor (PDGF) (31,32). Transformation by sis is thought to result from autocrine stimulation resulting from binding to the growth factor to PDGF receptors.
Quiescent REF52 cells were stimulated to enter S phase by recombinant p28c"s, indicating that these cells have receptors for PDGF (Fig. 2). In addition, p28'-ss induced, with expected kinetics, the transcription of several growth factor-induced early response genes Ha-ras p21 inhibits cellular responses to mitogenic stimuli. REF52 cells expressing neo (N6) or T24 Ha-ras genes (RN14) were seeded on cover slips at a density of 400 cells/cm2 and grown for 48 hr. At this time, monolayers were washed twice with phosphate-buffered saline and transferred to Dulbecco's modified Eagle's medium ITS+ (insulin, transferrin selenium, and oleic acid/bovine serum albumin complex). After 48 hr, the indicated growth factors were added to the GO-arrested cells. Samples were labeled for 12 hr at 12-hr intervals with 5 Ci/mL 3Hthymidine and then processed for emulsion autoradiography. Results for the time period 12 to 24 hr after growth factor addition (i.e., the interval with maximum response) are plotted. The growth factor concentrations were as follows: serum (10% fetal calf); recombinant platelet-derived growth factor B-chain (PDGF, 40 ng/mL); epidermal growth factor (EGF 100 ng/mL). RN14 cells also failed to respond to phorbol 12-myristate 13-acetate (TPA) at 10 ng/mL (data not shown).
(i.e., igc,fos, and JE), suggesting that the signal transduction pathway from the cell surface to the nucleus is intact (data not shown). However, when REF52 cells are cotransfected with c-sis and the gene for neomycin resistance, all of the resistant colonies were indistinguishable from REF52 cells transfected with the neomycin-resistance gene alone. In each case, similar numbers of neo-resistant colonies were obtained, suggesting that the sis gene was not toxic to REF52 cells. E1A introduced together (Table 1) or in a secondary transfection (data not shown), enabled c-sis to transform REF52 cells to anchorage independence. These results suggest that the ability of cells to be mitogenically stimulated by sis growth factor, while perhaps necessary, was not sufficient for stable transformation. As with ras, ElA appears to circumvent controls that prevents sustained exposure to a mitogenic stimulus from transforming.

Collaborating Oncogenes Have Complementary Effects on Cellular Responses to Growth Factors
Since oncogenic transformation of REF52 cells can require the activities of two oncogenes acting in concert, these cells provide an opportunity to study the biological effects of individual oncogenes in the absence of secondary phenotypic changes resulting from transformation. REF52 cells expressing low levels of activated Ha-ras p21 (T24REF) are morphologically normal and fail to form either colonies in soft agar or tumors in nude mice. However, these cells express functional ras oncoproteins, since ElA and SV40 large T-antigen each efficiently transform when transfected subsequently (clone N6) (25).
When transferred to serum-free media, T24REF (clone RN14) cells cease growing with kinetics indistinguishable from the neo-resistant REF52 control (clone N6) (Fig. 3). Nevertheless, a marked difference was evident between the cell lines upon restimulation by individual mitogens. As shown in Figure 2, the presence of even low levels of activated p21 interfered with the ability of REF52 cells to progress into S phase upon stimulation by phorbol 12-myristate 13-acetate, epidermal growth factor, PDGF or PDGF and epidermal growth factor together. The response to serum, however, was unaffected or even moderately enhanced by T24Ha-ras.
Cell lines individually expressing either ElA or SV40 large T antigen, although morphologically altered, were not transformed, as assessed by their failure to grow in soft agar or from tumors in nude mice. However, as shown in Figure 2, the presence of either SV40 large T antigen or ElA in REF52 cells allows for extended proliferation in the absence of serum growth factors. Thus, there is an apparent functional complementation: ras inhibits proliferative responses to mitogenic stimuli, while ElA and large T each enable REF52 cells to proliferate in the absence of mitogenic stimuli. In other words, REF52 cells appear to possess growth controls  (20), pSMRlMoNeo, a derivative of pSMl (42), c-sis and neo genes expressed from the SV40 early promoter and moloney murine leukemia virus LTR, respectively; plA, advenovirus-5 ElA (25); and plAhgm, Ad5 ElA and hygromycin resistance gene. Following selection with 0.4 mg/mL G418 (Gibco, 50% active) and/or 0.1 mg/mL hygromycin B (Calbiochem), cells from at least 100 colonies were trypsinized, pooled, and plated in media containing 0.37% agarose. Fresh agar-containing media was added to soft agar cultures once a week for 3 to 4 weeks and then the cultures were stained with p-iodonitrotetrazolium violet (Sigma). Soft agar plating efficiency refers to percent of cells forming colonies 70 jLm in diameter or larger. . REF52 cells expressing adenovirus ElA or SV40 large T-antigen continue to proliferate in serum-free media. Cell lines were seeded at a density of 3000 cells per 35 mm well and allowed to attach and proliferate for 48 hr. Cells were then washed twice with phosphate-buffered saline and transferred to media containing Dulbecco's modified Eagle's medium (DME) and 1% ITS + (insulin, transferrin selenium, and oleic acid/bovine serum albumin complex). This is the zero time point. Cells were labeled for 12 hr prior to each of the indicated time intervals in 1 mL of DME/ITS + containing 5 ,u Ci 3H-thymidine. Trichloroacetic acid-precipitable counts were determined and expressed as a percent of the time zero (i.e., label added from -12 to 0 hr) cpm. Results with REF52 clones expressing the following genes are plotted: neo, N6; T24 Ha-ras, RN14; adenovirus-5 ElA, Al; and SV40 large T-antigen, WS1. which limit the ability of the cells to proliferate in response to mitogenic stimuli. These controls are activated by ras and are circumvented by ElA and SV40 large T-antigen.

A Dominant Antagonist Regulates Transformation by ras and El A
A series of cell fusion experiments were performed to determine whether the growth controls that prevent ras from transforming REF52 cells are dominant or recessive. In principle, REF52 could differ from other established rodent lines, which are transformed by ras alone, in one of two ways: a) REF52 cells lack activities that are required for transformnation by ras, and ElA supplies the missing cellular functions; b) REF52 cells possess an antagonist that prevents ras from transforming, and ElA inactivates of circumvents the antagonist. In order to distinguish between these two possibilities, cell lines, shown in Table 2, containing either hygromycin or neomycin resistance markers, were fused using polyethylene glycol and selected in the presence of both antibiotics (33). Hybrid colonies were pooled and plated in soft agar to assess whether any expressed transformed phenotypes (Table 3).
When NRK cells transformed by ras were fused with REF52 cells, 20-fold fewer colonies were obtained than when normal NRK cells were fused with REF52. Furthermore, the hybrid colonies obtained were morphologically nonnal and failed to produce any soft agar colonies, indicating that transformation of NRK cells by ras was suppressed by fusion with REF52 cells. In contrast, the fusion of ras-transformed NRK cells with ElA-expressing REF52 cells yielded transformed colonies, which formed colonies in soft agar with a frequency of 4%. Similar results were obtained with Rat2 cells were substituted for NRK cells (Table 3). Thus, the growth controls that prevent ras from transforming and act to inhibit proliferation in the presence of ras are dominant, trans-acting, and are either circumvented or negated by function(s) of ElA.
Further evidence that REF52 cells possess a dominant antagonist of transformation has come from the analysis of spontaneous transformants induced by activated ras oncogenes. We pooled 102 to 104 to neomycin-resistant colonies isolated after transfecting REF52 cells with T24Ha-ras, and 5 x 106 cells were injected into nude mice. Tumors arose within 4 to 6 weeks, and the cell lines derived from each tumor were aREF52, Rat2, and NRK49F cells were transfected with the indicated plasmids. Plasmids expressed the following genes: pT24neo, an activated Ha-ras gene derived from T24 human bladder carcinoma cells (25); plAhgm, the adenovirus 5 ElA gene linked to hph, the gene of hygromycin B phosphotransferase; Homer6, neo (20); pY3, hph (43); HO6T1, T24 H-ras inserted into Homer6 (20); and pSVrasC, a T24 ras cDNA replacing the neo of pKOneo. re-established in culture and used in fusion experiments (Table 4). Fusions between tumor cells and the parental REF52 cells failed to produce transformed hybrids, whereas fusions between tumor cells and NRK or Rat2 lines yielded transformed colonies which grew in soft agar. In addition, fusions between some of the tumor lines and REF52 cells expressing ElA produced transformed hybrids, indicating that the antagonist was neutralized by ElA. This series of fusions suggest that the ras-induced tumors arose as a result of the loss of antagonist functions which normally prevent transformation and not from the activation of cellular functions analogous to ElA. While the biochemical mechanisms are yet unknown, this genetically defined antagonist resembles an antioncogene, a gene whose functional loss leads to transformation. This is provocative, since the ability of ElA proteins to bind the retinoblastoma gene product, p1Q5Rb, is tightly linked to ElA activities that enable ras to transform primary baby rat kidney cells (34)(35)(36).
This raises the possibility that p1O5Rb is a component of the transformation antagonist of REF52 cells. If so, one would expect to observe differences in the levels of plO5Rb in normal REF52 cells, and ras-transformed REF52 tumor cells. However, the eight ras-induced tumor lines examined thus far maintain a normal number of intact Rb genes as well as unaltered levels of Rb message. These results do not exclude the possibility of point-mutations or posttranslational alterations that adversely affect the activity of p105Rb.

Summary
In summary, REF52 cells possess growth controls that prevent a sustained mitogenic stimulus from transforming. These growth controls are dominant and can act in trans to antagonize transformation of rat cells previously transformed by ras. Given that ras and sis initially provoke REF52 cells to proliferate, the antagonist may itself be activated by mitogens as a part of a feed-back mechanism which limits cell proliferation. ElA and SV40 large T are able to circumvent or inactivate the antagonist. The extent to which this relates to the mechanism by which the viral oncoproteins enable REF52 cells to grow in the absence of serum is currently under investigation. Future studies will focus on genes, such as Rb and p53, known to be involved in the negative regulation of cell proliferation in an effort to understand at a biochemical level the information obtained in genetic analyses. In addition, we will examine the effects of ras on various cell-cycle genes. It is interesting that ras causes REF52 and other cells to arrest in G2 upon removal of SV40 large T-antigen (27,37). The control of the cells cycle in G2 has been described in a variety of organisms and can involve a protein kinase identified as: a) a component of maturation promoting factor (MPF), b) a 34 kDa protein in mammalian cells, c) a starfish histone Mphase kinase, and d) products of the CDC2 and cdc28 genes of S. pombe and S. cerevisiae, respectively (38,39). That ElA both induces p34 in baby rat kidney cells (40) and binds to a cyclinlike molecule (41) as well as the observation that ras-expressing REF52 cells arrest in G2 suggests that p34 may influence complementation between these oncogenes. For example, ElA may protect against the repression of p34 expression or activity by ras. This work is supported by National Cancer Institute Public Health Service grants R01CA40602 and P01CA42063 (to H.E.R.) and partially supported by a Cancer Center Core grant P30CA14051 to P.A. Sharp.