Dual growth arrest pathways in astrocytes and astrocytic tumors in response to Raf-1 activation.

Normal human fibroblasts have been shown to undergo a p16(Ink4a)-associated senescence-like growth arrest in response to sustained activation of the Ras/Raf/MEK/ERK pathway. We noted a similar p16(Ink4a)-associated, senescence-like arrest in normal human astrocytes in response to expression of a conditional form of Raf-1. While HPV16 E7-mediated functional inactivation of the p16(Ink4a)/pRb pathway in astrocytes blocked the p16(Ink4a)-associated growth arrest in response to activation of Raf-1, it also revealed a second p21(Cip1)-associated, senescence-associated, beta-galactosidase-independent growth arrest pathway. Importantly, the p21(Cip1)-associated pathway was present not only in normal astrocytes but also in p53-, p14(ARF)-, and p16(Ink4a)/pRb-deficient high grade glioma cells that lacked the p16(Ink4a)-dependent arrest mechanism. These results suggest that normal human cells have redundant arrest pathways, which can be activated by Raf-1, and that even tumors that have dismantled p16(Ink4a)-dependent growth arrest pathways are potentially regulated by a second p21(Cip1)-dependent growth arrest pathway.

The importance of the Ras/Raf/MEK/ERK pathway in growth is further supported by its alteration in many human tumors. Approximately 50% of human colon carcinomas contain mutant, constitutively active Ras (12), an alteration that is believed to contribute to the proliferative signaling necessary for tumor growth. Tumors that typically lack Ras mutations such as brain tumors (gliomas) also have alterations in the Ras/Raf/MEK/ERK pathway. Approximately 30% of high grade human gliomas have an amplified, rearranged, and constitutively activated epidermal growth factor receptor (13,14), a receptor known to signal through the Ras pathway (15). Additionally, Ras-GTP levels are elevated in a high percentage of high grade human gliomas (16), presumably as a result of signaling from a number of different growth factor receptors. Disregulation of the Ras/Raf/MEK/ERK pathway is therefore a common means by which tumors maintain and amplify mitogenic stimuli and thereby sustain proliferation.
While Ras/Raf/MEK/ERK pathway activation can promote cellular proliferation, it can also paradoxically provoke cell cycle arrest. Sustained activation of oncogenic Ras in normal human fibroblasts (NHF) leads not to proliferation but to permanent arrest in association with elevated levels of p53 and of the cyclin-dependent kinase inhibitors p16 Ink4a and p21 Cip1 (17). This irreversible arrest is also associated with expression of senescence-associated ␤-galactosidase (SA-␤-gal) activity, suggesting that in response to high levels of activated Ras, normal human cells undergo a senescence-like growth arrest. The ability of oncogenic Ras to induce a senescence-like growth arrest in NHF appears to be closely linked to activation of Raf but not to other downstream targets of Ras, such as phosphatidylinositol 3-kinase or RalGDS, since only mutant forms of Ras that selectively activate the Raf pathway induced growth arrest comparable with that induced by constitutively expressed oncogenic Ras (17). This point is further supported by the finding that sustained activation of a conditional form of Raf-1, or overexpression of activated MEK1, in NHF leads to senescence-like arrest comparable with that induced by Ras (17,18). Although the mechanism by which Ras or Raf induces a senescence-like arrest in NHF remains to be fully defined, NHF expressing HPV16 E6 and therefore lacking functional p53 (and p53-dependent induction of p21 Cip1 ) still undergo a p16 Ink4a -associated, senescence-like arrest in response to Raf activation (18). This senescence-like, Raf-induced arrest could also be blocked by the MEK inhibitor PD098059 (18), further suggesting a link between Raf activation, p16 Ink4a induction, and senescence-like arrest. As a whole, these results suggest that the Ras/Raf/MEK/ERK pathway, in addition to playing a key role in cellular proliferation, can also trigger a senescencelike, p16 Ink4a -associated growth arrest that prevents proliferation of normal cells.
While normal human cells are programmed to undergo a senescence-like growth arrest in response to excessive stimulation of the Ras/Raf/MEK/ERK pathway, tumor cells could gain a growth advantage by inactivating this tumor-suppressive response. Functional inactivation of the pRb/p16 Ink4a pathway is a common feature of many if not most human tumors, including human gliomas (13,19). Given the role of p16 Ink4a in mediating Ras/Raf-induced senescence, it seems likely that most human tumors lack important growth arrest programs linked to excessive Ras/Raf/MEK/ERK pathway activation. The cellular response to Ras/Raf/MEK/ERK activation in the absence of a functional p16 Ink4a /pRb pathway, however, remains poorly defined. In NHF rendered unresponsive to p16 Ink4amediated cell cycle arrest by expression of cyclin D1 and a mutant Cdk4, expression of oncogenic Ras resulted in growth arrest and alterations in cell morphology similar to those noted in parental cells (20). It is unclear, however, if the arrest pathway activated in the absence of p16 Ink4a was the senescence-like arrest pathway activated in p16 Ink4a -proficient cells and, if so, what proteins were involved. Similarly, Raf activation in pRb-deficient small cell lung carcinoma cells resulted in cell cycle arrest, although again it was unclear if this arrest was associated with markers of senescence or involved either p16 Ink4a or p21 Cip1 (21). To more clearly define Raf-induced growth arrest pathways and, in particular, potential p16 Ink4aindependent alternate growth arrest pathways, we characterized the response of normal human glial cells to activation of a conditional form of Raf-1 and assessed whether this response was similar in pRb-deficient glial cells and in glial tumors lacking p16 Ink4a and/or pRb. The results of these studies show that rather than having a single p16 Ink4a -dependent senescence-associated arrest pathway, astrocytes have a second distinct, SA-␤-gal-independent, p21 Cip1 -associated arrest pathway activated in response to Raf activation. This second pathway is present not only in normal astrocytes but also in glial tumor cells that have lost the function of several key cell cycle regulatory proteins including p16 Ink4a , p53, pRb, and p14 ARF . These results suggest that normal cells (in this case astrocytes) have both primary and secondary mechanisms that can be used to prevent proliferation in the face of sustained Raf/MEK/ERK signaling and that even highly malignant tumors retain at least one pathway in a functional yet dormant state.
Retrovirus Production and Infection-To create cells that express conditional Raf-1, Phoenix A cells were transfected using Lipo-fectAMINE (Life Technologies, Inc.) with EGFP⌬Raf-1:ER retroviral DNA. The pWZLblast3 EGFP⌬Raf-1:ER retroviral vector encodes a protein that is a chimera of the catalytic domain of human Raf-1, the hormone binding domain of the human estrogen receptor, and an enhanced version of green fluorescent protein (EGFP). The vector also encodes blasticidin resistance. A similar retroviral vector for the expression of GFP⌬Raf-1:ER has been described previously (22). Infected cells show little Raf kinase activity until the chimeric protein is acti-vated by the addition of 17␤-estradiol (E 2 ) or the estrogen analog 4-hydroxytamoxifen (4HT) (23). As a control, cells were infected with pBabepuro3⌬Raf301:ER retroviral DNA. The pBabepuro3⌬Raf301:ER retroviral vector encodes a protein that is a chimera of a kinase-inactive version of the catalytic domain of Raf-1 and the hormone-binding domain of the estrogen receptor. The vector also encodes for puromycin resistance. 24 and 48 h after transfection of Phoenix A cells, the medium containing the packaged virus was placed on NHA. Infected NHA were selected for resistance to blasticidin S (25 g/ml for 4 -5 days) (ICN, Costa Mesa, CA) or puromycin (1 g/ml for 4 -5 days) (Sigma-Aldrich), and resistant cells were pooled and expanded for further studies. To ensure successful infection, relevant expanded populations were analyzed by fluorescence-activated cell sorting to verify expression of EGFP⌬Raf-1:ER. Pooled populations of infected NHA were used for all experiments, since the limited life span of astrocytes in culture does not allow for development of clonal populations.
To create pRb-deficient cells with conditional activation of Raf-1, NHA first were infected with HPV16 E7 retroviral DNA packaged in Phoenix A cells, as described above. The pLNCX retroviral vector allows for constitutive expression of the HPV E7 protein (pRb inactivation) and neomycin resistance. Cells expressing E7 were selected for resistance to G418 (0.5 mg/ml for 6 -7 days) (ICN). Infected cells that survived drug selection were then infected with the EGFP⌬Raf-1:ER retroviral DNA, as described. Cells were selected for resistance to blasticidin/G418 and were pooled and expanded for further studies.
U251, SF126, U87, and U87 ϩ E6 glioma cell lines were infected with EGFP⌬Raf-1:ER retroviral DNA as described and selected for resistance to blasticidin. U251 also were infected with pBabepuro3⌬Raf301:ER retroviral DNA as described and selected for resistance to puromycin. Following blasticidin selection, cells infected with EGFP⌬Raf-1:ER were sorted using a Becton-Dickinson FACSVantage SE Cell Sorter, and the cells expressing the highest level of EGFP (upper 5% of expressers) were collected and expanded. All experiments were performed using this population of high expressing cells. These cells were periodically reanalyzed by fluorescence-activated cell sorting to ensure maintenance of EGFP expression level.
Raf Activation and Inactivation-EGFP⌬Raf-1:ER was activated by incubating cells for 4 days with either 1 M E 2 (Sigma-Aldrich) or 100 nM 4HT (Research Biochemicals International, Natick, MA). 1 mM stocks of E 2 and 4HT were made in ethanol. For all studies, both control and E 2 -or 4HT-containing media were changed every 48 h.
For studies done to assess the reversibility of the Raf-induced growth arrest, cells were incubated for 4 days with 1 M E 2 . Cells then were washed extensively with PBS, grown in control medium for an additional 4 days, and analyzed.
[ 3 H]Thymidine Incorporation-The extent of proliferation was assessed by monitoring incorporation of [methyl-3 H]thymidine (PerkinElmer Life Sciences). Cells were seeded in at least quadruplicate in 96-well plates, at a density of 16,000 cells/ml (NHA, NHA/E6, NHA/E7, and SF126 cells) or 12,000 cells/ml (U251 and U87 cells). Cells were grown for 4 days and then incubated with 20 Ci/well of [methyl- Analysis of Senescence-associated ␤-Galactosidase Activity-SA-␤gal activity was determined by the method of Dimri et al. (24). Cells were fixed in 2% formaldehyde, 0.2% glutaraldehyde and incubated with a 1 mg/ml X-gal solution (40 mM citric acid/sodium phosphate buffer (pH 6), 5 mM potassium ferrocyanide, 5 mM ferricyanide, 150 mM NaCl, 2 mM MgCl 2, 1 mg/ml X-gal in dimethylformamide) for 16 -24 h. Cells were then rinsed with PBS and methanol and air-dried. The percentage of SA-␤-gal-positive cells was assessed by counting the number of blue cells in a ϫ 10 microscopic field and dividing that number by the total number of cells in the field. The percentage of positively stained cells in each plate was the average of four fields per plate.

Raf Activation Results in a Phenotype
Resembling Senescence in NHA-4HT-or E 2 -mediated activation of Raf in proliferating NHA retrovirally infected with EGFP⌬Raf-1:ER (NHA/ RafER) led to elevated expression of the chimeric Raf:ER protein. Western blot analysis showed that E 2 incubation led to a sustained high level of EGFP⌬Raf-1:ER expression in NHA/ RafER cells (Fig. 1a), consistent with increased stability of the Raf-1:ER fusion protein, as reported previously (23). EGFP⌬Raf-1:ER induction was noted within 24 h of E 2 addition and was sustained for 4 days (Fig. 1a). E 2 incubation also led to an increase in phosphorylated ERK1/2 in NHA/RafER cells, but not NHA or NHA containing the kinase-inactive version of Raf-1 (NHA/Raf301:ER), as determined by Western blot analysis using a phosphospecific p42/p44 antibody ( Fig. 1a and data not shown). This increase in phosphorylated ERK1/2 is consistent with p42/p44 ERK1/2 pathway stimulation following Raf activation. The level of phosphorylated ERK1/2 remained high throughout the experiment. As shown in Fig. 1a, the addition of 4HT resulted in induction of EGFP⌬Raf-1:ER and phosphorylated ERK1/2 similar to that seen with E 2 .
Activation of EGFP⌬Raf-1:ER also led to changes in cell proliferation and morphology. Beginning ϳ48 h following EGFP⌬Raf-1:ER activation, cells became smaller and refractile but remained adherent and viable (determined by trypan blue exclusion, data not shown). [ 3 H]thymidine incorporation was reduced by 90% after 4 days of EGFP⌬Raf-1:ER activation (Fig.  1b). Incorporation of [ 3 H]thymidine by uninfected NHA and by NHA/Raf301:ER was not affected by the addition of E 2 to the medium (Fig. 1b). Additionally, 4 days after EGFP⌬Raf-1:ER activation by the addition of either 4HT or E 2 , 45-50% of cells stained positively for SA-␤-gal, a common marker of senescent cells, while control NHA/RafER cells and parental NHA had fewer than 10% SA-␤-gal positive cells (Fig. 1c).
Senescent cells have been shown to have increased expression of p16 Ink4a , p21 Cip1 , and p53 (20,25,26), and human fibroblast studies have demonstrated that elevation of p16 Ink4a expression is associated with the maintenance of senescence (18,26). Therefore, we assessed the effect of Raf activation on the expression of these cell cycle regulatory proteins in NHA. As shown in Fig. 1d, incubation of uninfected NHA with E 2 had no effect on expression of p16 Ink4a , p21 Cip1 , or p53. Similarly, incubation of NHA/Raf301:ER with 4HT did not affect p16 Ink4a expression (data not shown). EGFP⌬Raf-1:ER activation in NHA/RafER cells similarly did not significantly increase levels of p21 Cip1 expression or p53 expression 4 days following Raf activation (54 Ϯ 23 and 114 Ϯ 31% of control values, respectively, by quantitative Western blot analysis; Fig. 1d). EGFP⌬Raf-1:ER activation in NHA/RafER cells did, however, result in increased expression of the cyclin-dependent kinase inhibitor p16 Ink4a , at 72 h after Raf activation (Fig. 1d).
To address whether the Raf-induced growth arrest in NHA/ RafER cells was permanent and irreversible, NHA/RafER cells were incubated with E 2 for 4 days, washed extensively with PBS to remove E 2 from the culture, and grown in control medium for an additional 4 days. E 2 was used to activate EGFP⌬Raf-1:ER in these experiments, since it is more easily removed from the cells than 4HT. As shown in Fig. 1a, hormone removal resulted in decreases of EGFP⌬Raf-1:ER levels and phosphorylated ERK1/2 levels to those seen in control cells. Following E 2 removal, however, cells maintained an altered morphology, and thymidine incorporation remained reduced by 90% (Fig. 1b). Additionally, there was no change in the percentage of cells that stained positively for SA-␤-gal (Fig. 1c). Finally, consistent with the senescence phenotype, p16 Ink4a expression remained elevated (Fig. 1d). Irreversible growth arrest, maintenance of SA-␤-gal expression, and sustained elevation of p16 Ink4a expression suggest that activation of Raf and the ERK1/2 pathway is able to induce a p16 Ink4a -associated senescence phenotype in NHA.
SA-␤-gal Activity, but Not Growth Arrest, of NHA following Raf Activation Is Dependent on Functional pRb-Because of the association between increased p16 Ink4a expression and Rafinduced senescence in NHA, studies were initiated to determine if a functional p16 Ink4a /pRb pathway was required for Raf-induced growth arrest in NHA. NHA constitutively expressing E7 and conditional EGFP⌬Raf-1:ER (NHA/E7/RafER) were derived by retroviral infection. Expression of the HPV16 E7 protein resulted in decreased protein levels of pRb, as shown in the Western blot in Fig. 2a, lower panel. While 4HT had no effect on protein expression in parental NHA/E7 cells (data not shown), 4HT-mediated activation of EGFP⌬Raf-1:ER (in NHA/E7/RafER cells) led to increased expression of both chimeric EGFP⌬Raf-1:ER protein and phosphorylated ERK1/2 (Fig. 2a). EGFP⌬Raf-1:ER activation also resulted in a 60% inhibition of [ 3 H]thymidine incorporation, as compared with control cells (Fig. 2b). EGFP⌬Raf-1:ER activation did not, however, alter cell morphology, and after 4 days of 4HT exposure fewer than 5% of cells stained positively for SA-␤-gal.
To better understand the difference in Raf-induced arrest in NHA/E7/RafER cells compared with NHA/RafER cells, Western blot analysis was done to assess changes in protein expression following EGFP⌬Raf-1:ER activation. Expression of p16 Ink4a was elevated in NHA/E7/RafER cells as compared with NHA/RafER cells (Fig. 2a, lower panel), which is consistent with the destabilization of pRb by the E7 protein (27,28). Unlike what was noted in NHA/RafER cells with a functional p16 Ink4a /pRb pathway, however, there was no further increase in p16 Ink4a levels, relative to control cells (average p16 Ink4a levels were 65 Ϯ 9% of control values) in the pRb-deficient NHA/E7/RafER cells following Raf activation (Fig. 2c). There was, however, an increase in p21 Cip1 expression in NHA/E7/ RafER cells following EGFP⌬Raf-1:ER activation. This p21 Cip1 induction was apparent within 48 h (250 Ϯ 43% of control values), persisted at least 4 days (250 Ϯ 16% of control values), and appeared to be independent of p53, the levels of which remained unchanged throughout the experiment (Fig. 2c). These results indicate that activated Raf is able to induce growth arrest in NHA in the absence of a fully functional p16 Ink4a /pRb pathway. This arrest, however, is associated with increased p21 Cip1 expression, is not associated with increased SA-␤-gal, and appears to be functionally Activation of Raf Results in Growth Arrest in High Grade Gliomas-Although NHA have two pathways leading to growth arrest following sustained Raf expression, glial tumors often have genetic defects that might influence their responses to Raf activation. For example, the vast majority of high grade (grade IV) gliomas have defects in the p16 Ink4a /pRb pathway (13). To assess whether glioma cells retained cell cycle arrest pathways in response to Raf activation, high grade glioma cells were infected with EGFP⌬Raf-1:ER, incubated with 4HT or E 2 to activate EGFP⌬Raf-1:ER, and monitored for cellular alterations and activation of cell signaling pathways. While [ 3 H]thymidine incorporation by U251 cells containing the kinase-inactive form of Raf-1 (U251/Raf301:ER) was not affected by incubation with 4HT (Fig. 3a), [ 3 H]thymidine incorporation was reduced by 80% in U251/RafER cells, by 90% in SF126/ RafER cells, and by 85% in U87/RafER cells following 4 days of incubation with E 2 (Fig. 3a). In contrast to the NHA/RafER cells, however, which became smaller and refractile and SA-␤gal-positive following EGFP⌬Raf-1:ER activation, the tumor cells became elongated with long processes, and SA-␤-gal activity did not increase following EGFP⌬Raf-1:ER activation. The response of these cells was therefore more similar to that of pRb-deficient NHA/E7/RafER cells than that of NHA. Consistent with this idea, all cell lines showed elevated p21 Cip1 expression, ranging from 2-fold increase in U87/RafER cells to 8-fold in U251/RafER and SF126/RafER cells, which began 24 -48 h after 4HT exposure and persisted throughout the experiment (Fig. 3, b-e). As in the NHA/E7/RafER cells, p21 Cip1 induction appeared to be independent of p53, since 1) U251 and SF126 cells lack functional p53 (29,30); 2) U87 cells containing functional p53 (30) showed no increase in p53 protein levels following 4 days of 4HT exposure despite showing p21 Cip1 induction (Fig. 3d); and 3) U87 ϩ E6 cells, which lack functional p53, displayed p21 Cip1 induction and growth arrest similar to that noted in U87 cells (Fig. 3e). p16 Ink4a did not have FIG. 3. Analysis of Raf-induced growth arrest in high grade gliomas. U251/RafER, SF126/RafER, U87/RafER, and U87 ϩ E6/RafER were incubated for 4 days with 4HT to activate Raf and then were harvested and analyzed as described under "Experimental Procedures." a, inhibition of proliferation following Raf activation in U251/Raf301:ER, U251/RafER, SF126/RafER, and U87/RafER was assessed by [methyl-3 H]thymidine incorporation. All values are the mean Ϯ S.D. of at least three experiments. E 2 * values were derived from cells washed extensively to remove E 2 and then grown 4 days in control medium before analysis. b-e, changes in protein expression following Raf activation in U251/RafER cells (b), SF126/RafER cells (c), U87/RafER cells (d), and U87/E6/RafER cells (e) were assessed by Western blot. 96* values were derived from cells washed extensively to remove E 2 and then grown 4 days in control medium before analysis. a role in growth arrest of these glioma cell lines, since all lines had homozygous deletion of p16 Ink4a (21,30). These results suggest that high grade gliomas, which have lost the p16 Ink4aassociated senescence pathway, retain the p21 Cip1 -associated growth arrest pathway present in NHA.
To further distinguish the Raf-activated, p21 Cip1 -associated arrest noted in glioma cells from the p16 Ink4a -associated, senescence-like arrest seen in NHA, further studies were done in U251/RafER and SF126/RafER cells to determine the permanence of the Raf-induced arrest. When U251/RafER cells were incubated for 4 days with E 2 and then incubated for 4 more days in control medium, cells resumed proliferating, as determined by [ 3 H]thymidine incorporation (Fig. 3a). Additionally, the cells reverted back to their original morphology when the E 2 was removed from culture, unlike the NHA/RafER, which maintained their altered morphology following E 2 removal. Finally, when EGFP⌬Raf-1:ER was inactivated in U251/RafER and SF126/RafER cells by the removal of E 2 from the culture, p21 Cip1 expression levels dropped back to the level seen in control cells (Fig. 3, b and c). Therefore, EGFP⌬Raf-1:ER activation in the glioma cell lines resulted in growth arrest that lasted only as long as Raf was activated. This arrest also was associated with increased expression of p21 Cip1 , which returned to normal levels with the inactivation of EGFP⌬Raf-1:ER. These data suggest that the Raf-induced, p21 Cip1 -associated growth arrest in these glioma cell lines is reversible and is dissimilar to the senescence seen in NHA in response to Raf activation.

DISCUSSION
The data presented in this paper demonstrate the presence of two distinct inducible growth arrest pathways in astrocytic cells. These two pathways are qualitatively different in nature, with the choice of which is used dependent on the presence of a functional p16 Ink4a /pRb pathway. In the case of NHA, which have an intact p16 Ink4a /pRb pathway, Raf activation results in an irreversible, senescence-like growth arrest associated with increased p16 Ink4a expression. However, when this pathway is disrupted, as is the case in the NHA/E7/RafER cells and in the high grade glioma cell lines, Raf activation leads to reversible growth arrest associated with increased p21 Cip1 expression.
The finding that normal human cells contain two distinct growth arrest pathways in response to Raf activation is both novel and consistent with previous published work. The ability of cells to undergo arrest in response to exogenous expression of either p16 Ink4a or p21 Cip1 is widely appreciated (18,(31)(32)(33), and Ras or Raf activation in NHF leads to increased expression of both of these proteins (17,18). In NHF, however, Raf-induced senescence-like growth arrest has been shown to occur in the absence of induced levels of p21 Cip1 (18), a finding reinforced by our observation that in NHA, p16 Ink4a levels, but not levels of p21 Cip1 , increase in response to EGFP⌬Raf-1:ER activation. The novel finding of the present study is that p21 Cip1 , whose potential function in Ras-and Raf-induced arrest in NHF was not closely examined, is associated not with senescence-like growth arrest in NHA but rather with a second distinct arrest pathway that is not associated with markers of senescence and whose presence can be noted only in the absence of the primary p16 Ink4a -associated arrest pathway. The two growth arrest pathways defined in NHA appear to be mutually exclusive (i.e. functional p16 Ink4a /pRb appears to be unnecessary for p21 Cip1associated arrest, and vice versa). In E7-expressing astrocytes, however, levels of pRb were low but not absent, and these cells also contained high levels of p16 Ink4a . The possibility exists, therefore, that the presumed p21 Cip1 -associated response might in fact be simply a muted version of the already described p16 Ink4a -associated arrest. Additionally, since high lev-els of p16 Ink4a have been suggested to displace p21 Cip1 from cyclin D-Cdk4 complexes and to stimulate formation of growthinhibitory p21 Cip1 -cyclin E-Cdk2 complexes (34), the possibility also exists that the presumed p21 Cip1 -associated response might in fact be directly dependent on high levels of p16 Ink4a . The qualitative difference in response to EGFP⌬Raf-1:ER activation in E7-expressing cells (p21 Cip1 -associated arrest in the absence of SA-␤-gal expression or morphologic alterations) versus that in NHA (p16 Ink4a -associated senescence-like arrest), however, argues against a single p16 Ink4a -associated pathway. The presence of a p21 Cip1 -associated arrest pathway in p16 Ink4a null tumor cell lines and the transient as opposed to permanent nature of this arrest further support the contention that two distinct arrest pathways exist in cells of glial origin.
A second important finding of the present work is that even tumors which have dismantled the p16 Ink4a -dependent senescence-like growth arrest pathway retain the secondary, p21 Cip1 -associated growth arrest mechanism. Numerous tumor cell types have been shown to respond to Raf induction by undergoing growth arrest, although in these studies either the tumor cells used were p16 Ink4a /pRb-proficient (35,36) or the arrest did not involve p21 Cip1 (21). We also have noted that activation of EGFP⌬Raf-1:ER induces a SA-␤-gal-associated arrest in p16 Ink4a -proficient tumor cells derived from low grade human gliomas (data not shown). It therefore seems likely that p16 Ink4a /pRb-proficient tumors retain senescence-like arrest pathways similar to those noted in normal human cells following activation of Raf. The arrest noted in the glioma cell lines used in the present study, however, was clearly different from the p16 Ink4a -dependent arrest noted in NHA, since it was reversible, not associated with SA-␤-gal expression, and activated even in tumor cells (U87, U251, and SF126) devoid of p16 Ink4a . Significantly, the p21 Cip1 -associated arrest was also noted in tumor cells lacking functional p53 and p14 ARF (U251, SF126). Recent studies have shown that Ras or Raf activation can induce expression of both p14 ARF and the p53 inhibitor Mdm2 and that the balance of these actions may affect p53 levels and downstream p53-dependent effects (37). The presence of a Raf-induced, p21 Cip1 -associated growth arrest pathway in p14 ARF and p53-deficient cells therefore suggests that the linkage between Raf induction and increased p21 Cip1 expression is independent of not only p16 Ink4a but also of effects of the Ras/Raf/MEK/ERK pathway on the p14 ARF /Mdm2/p53 pathway.
Having established the presence of a secondary p21 Cip1 -associated, Raf-stimulated arrest pathway in NHA, the question arises as to how Raf induces p21 Cip1 expression and how p21 Cip1 induces growth arrest in the absence of pRb/p16 Ink4a . With regard to the latter question, p21 Cip1 has been shown to inhibit the action of E2F in human cervical and bladder carcinoma cell lines that lack functional pRb (38), suggesting that the p21 Cip1 -associated arrest noted in glioma cell lines and, in particular, in the pRb/p16 Ink4a -deficient SF126 line may similarly involve E2F. Alternatively, p21 Cip1 may regulate the activities of cyclin E-Cdk2 and/or cyclin A-Cdk2 complexes (19). With regard to the former question, p53-independent induction of p21 Cip1 has been associated with a variety of transcription factors including Sp1, Sp3, STAT proteins, and E2Fs (39 -41). Raf-induced increases in any of these transcription factors could therefore lead to p53-independent increases in p21 Cip1 levels. These possibilities and, in particular, the interplay between Raf, E2F, and p21 Cip1 , remain to be examined.
Finally, the identification of two distinct growth arrest pathways in astrocytes and astrocytic tumors may be important both to understanding the genesis of tumors and to effectively stopping their growth. Approximately half of grade III anaplas-tic astrocytomas have inactivation of at least one component of the p16 Ink4a /pRb pathway, and homozygous deletion of p16 Ink4a is seen in up to 70% of grade IV glioblastoma multiforme (13). Thus, most glial tumors, by dismantling the p16 Ink4a /pRb pathway, have eliminated the possibility of Ras/Raf-mediated p16 Ink4a -associated permanent growth arrest. Inactivation of p21 Cip1 , however, has rarely been reported in gliomas or in other human tumors (13,19,42). These observations suggest that tumor development may select for cells with Ras/Raf/ MEK/ERK pathway activation high enough to promote growth yet low enough to avoid triggering p21 Cip1 -dependent growth arrest. Indeed, this appears to be the case in experimental human gliomas created by overexpression of mutant Ras, which, despite having very high levels of activated Ras, have only modest levels of ERK1/2 activation. 2 More importantly, these results support the idea that intact arrest pathways lie dormant in tumors (43), including gliomas. Selective activation of these pathways in tumor cells would represent an ideal, noncytotoxic way to prevent the growth and the ultimate fatal consequences of tumors. The definition of at least two distinct growth arrest pathways in cells is a first step toward this goal.