The Geranylgeranyltransferase I Inhibitor GGTI-298 Induces Hypophosphorylation of Retinoblastoma and Partner Switching of Cyclin-dependent Kinase Inhibitors

The geranylgeranyltransferase I inhibitor GGTI-298 has recently been shown to arrest human tumor cells in the G1 phase of the cell cycle, induce apoptosis, and inhibit tumor growth in nude mice. In the present manuscript, we provide a possible mechanism by which GGTI-298 mediates its tumor growth arrest. Treatment of the human lung carcinoma cell line Calu-1 with GGTI-298 results in inhibition of the phosphorylation of retinoblastoma protein, a critical step for G1/S transition. The kinase activities of two G1/S cyclin-dependent kinases, CDK2 and CDK4, are inhibited in Calu-1 cells treated with GGTI-298. Furthermore, GGTI-298 has little effect on the expression levels of CDK2, CDK4, CDK6, cyclins D1 and E, but decreases the levels of cyclin A. GGTI-298 increases the levels of the cyclin-dependent kinase inhibitors p21 and p15 and had little effect on those of p27 and p16. Most interesting is the ability of GGTI-298 to induce partner switching for several CDK inhibitors. GGTI-298 promotes binding of p21 and p27 to CDK2 while decreasing their binding to CDK6. Reversal of partner switching and G1 block was observed after removal of GGTI-298. Furthermore, GGTI-298 treatment results in an increased binding of p15 to CDK4, which is paralleled with decreased binding to p27. The results demonstrate that the GGTI-298-mediated G1 block in Calu-1 cells involves increased expression and partner switching of CDK inhibitors resulting in inhibition of CDK2 and CDK4, and retinoblastoma protein phosphorylation.

Protein prenylation is an important posttranslational modification that is required for cellular localization and biological function of many proteins (1). These covalent attachments of farnesyl (C 15 ) or geranylgeranyl (C 20 ) to cysteines at the carboxyl-terminal of some proteins are catalyzed by three different enzymes. Farnesyltransferase (FTase) 1 and geranylgeranyltransferase I (GGTase I) modify cysteines of proteins that end in CAAX (C ϭ Cys, A ϭ aliphatic amino acid, X ϭ any amino acid) at their carboxyl-terminal with GGTase I preferring leucine or isoleucine and FTase preferring Met or Ser in the X position (1). GGTase II, on the other hand, geranylgeranylates on two cysteines in proteins that end in XXCC, CCXX, or CXC sequences. Many prenylated proteins are small G-proteins that are integral components of proliferative signal transduction pathways (1). For example, Ras farnesylation is required for its ability to transduce growth signals from receptor tyrosine kinases to transcription factors and the cell cycle machinery that regulates cell division (2). Furthermore, mutated Ras is found in about 30% of all human cancers and is believed to cause malignant transformation by constitutive activation of abnormal growth (3). Farnesylation of oncogenic Ras is required for its cancer-causing activity (4). Similarly, the Rho family of small G-proteins such as RhoA and Rac1 require geranylgeranylation for their biological function. One of their key biological roles is to allow cells to traverse the G 1 phase of the cell cycle and begin DNA synthesis in S phase (5). Recently, RhoA and Rac1 have been shown to be required for malignant transformation by Ras (6,7). Furthermore, constitutively activated RhoA and Rac1 can also lead to oncogenic transformation (6,7). The overwhelming evidence implicating the Ras and Rho family of proteins in aberrant proliferative pathological conditions such as cancer and cardiovascular diseases prompted us and others (reviewed in Ref. 8) to design and synthesize FTase and GGTase I inhibitors.
FTI-277 and GGTI-298 are CAAX peptidomimetics that potently and selectively inhibit FTase and GGTase I, respectively (9, 10). We have found FTI-276 and its methyl ester FTI-277 to be potent inhibitors of oncogenic Ras processing and signaling (9). FTI-276 also potently inhibited the growth in nude mice of human tumors with multiple genetic alterations such as K-Ras mutation and p53 deletion (11,12). FTIs from several other groups have also shown potent antitumor efficacy without toxicity in several animal models (8). Furthermore, several FTIs are presently in phase I clinical trials (13).
Selective inhibition of protein geranylgeranylation with GGTI-298 has major consequences on several biological pathways. Pretreatment of fibroblasts with GGTI-298, blocks PDGF-and epidermal growth factor-dependent tyrosine phosphorylation of their corresponding tyrosine kinase receptors (14). In contrast, selective inhibition of protein farnesylation has no effect on receptor tyrosine kinase phosphorylation (14). Furthermore, GGTI-298 inhibits the growth in nude mice of human tumors by a mechanism that is not yet known (12). One possible mechanism may involve GGTI-298-mediated G 1 phase block and apoptosis in cultured human tumor cells (15). The ability of GGTI-298 to inhibit proliferation is not limited to human tumor cells that are of epithelial and fibroblast origin. GGTI-298 also has a major effect on the proliferative pathways of smooth muscle cells (16). For example, GGTI-298 is very effective at inducing G 1 arrest and apoptosis in rat pulmonary artery smooth muscle cells. Furthermore, GGTI-298 enhances the ability of interleukin-1␤ to induce nitric oxide synthase-2 in the same cells (17). This induction of nitric oxide synthase-2 results in a large increase in the production of the nitric oxide radical, which is known to be inhibitory to smooth muscle cell proliferation. GGTI-298 also inhibited the ability of PDGF, interleukin-1␤, and activated Ras to induce superoxide production in smooth muscle cells (18). Taken together, the data indicate that GGTI-298 has antiproliferative effects on fibroblasts, epithelial, and smooth muscle cells, and this cell growth inhibition appears to be mediated through a G 1 phase arrest. Recently, we have shown that in human tumors this G 1 arrest correlated with an induction of a cyclin-dependent kinase inhibitor p21 WAF1 (19). However, whether this induction of p21 WAF1 is responsible for the G 1 arrest is not yet known.
The G 1 to S phase transition of the mammalian cell division cycle is a highly regulated step. The key regulators of G 1 /S progression are a series of kinases that depend on cyclins for activation (for review, see Refs. 20 -22). For example, the Dtype cyclins are made in early G 1 , bind to and activate the G 1 cyclin-dependent kinases CDK4 and CDK6. Activation of these kinases also requires phosphorylation by CDK kinases such as CAK or cyclin H-CDK7 and dephosphorylation by phosphatases such as cdc25 (20). At this stage, the cell has reached a critical checkpoint of its cell cycle called the restriction point (R) at which time the cell checks that all is ready for DNA synthesis. Cyclin D⅐CDK4 and cyclin D⅐CDK6 complexes are now active and phosphorylate the retinoblastoma gene product pRb. Phosphorylated pRb disassociates from the transcription factor E2F, which, once freed from pRb, is able to induce the expression of several genes that prepare the cell for DNA synthesis (20). Among these is cyclin E, which activates CDK2, a kinase required for late G 1 to early S phase transition. The cyclin E⅐CDK2 complex hyperphosphorylates pRb, and the cell proceeds into S phase. At this time cyclin A expression is high, whereas cyclins D and E have been degraded (20). The cyclin A⅐CDK2 complex maintains the phosphorylation of pRb to sustain DNA replication. Finally CDK kinase activities are highly regulated by two families of CDK inhibitors, CKIs such as p21 and p27 and the INKs such as p15 and p16 (20 -22). These inhibitors play a key role in making sure the cells stop at the R point if any DNA damage is detected. This allows the cells to repair the damage before replicating their DNA. Growth factors such as PDGF and epidermal growth factor activate several pathways, some of which have been shown to directly regulate the cell cycle (22). For example, activation of the Ras/Raf/MEK/ERK kinase cascade, which results in increased expression of cyclin D1 that in turn will activate CDK4/CDK6 allowing cells to traverse the R point. PDGF activation of the Ras/RhoA pathway results in the degradation of p27, which also will have the same overall effect of allowing cells to traverse the R point and enter S phase of the cell cycle (23).
The G 1 arrest brought about by GGTI-298 inhibition of protein geranylgeranylation could be due to effects on several important steps in G 1 /S such as the inhibition of expression of cyclins D and/or E, CDK2, CDK4 and/or CDK6, or increased expression of CKIs and/or INKs. The work described in this study suggests a possible mechanism for GGTI-298 growth arrest involving increased expression and partner switching of CDK inhibitors, resulting in inhibition of CDK2 and CDK4, and pRb phosphorylation.
Flow Cytometry Analysis-Cells were treated and harvested as described for Western blotting, and nuclei were stained with propidium iodide. DNA content was analyzed by fluorescence-activated cell sorter as described previously (19).
Cyclin-dependent Kinase Assay-To measure the activity of CDK2, histone H1 was used as the substrate; for CDK4 and CDK6, GST-Rb (C-terminal fragment of pRb) was used as a substrate. The CDK immunoprecipitates were resuspended in 10 l of 50 mM Hepes (pH 7.4) containing 10 mM MgCl 2 , 5 mM MnCl 2 , 1 mM dithiothreitol, 10 Ci [␥-32 P]ATP and 100 g/ml histone H1 (BM) or 20 g/ml GST-Rb, and then incubated for 30 min at 30°C with occasional mixing. The reaction was terminated with an equal volume of 2ϫ loading buffer (93.75 mM Tris, pH 6.8, 15% glycerol, 3% SDS, 7.5% ␤-mercaptoethanol). The sample was fractionated by SDS-PAGE, and phosphorylated proteins were visualized by autoradiography.

GGTI-298 Induces Accumulation of Hypophosphorylated pRb in the Human Lung Carcinoma
Calu-1-GGTI-298 was previously shown to inhibit the growth in nude mice and to induce G 1 block of human tumor cells (12,15). We sought to understand how GGTI-298 prevents Calu-1 cells from traversing G 1 and entering S phase of the cell cycle. We first determined the ability of GGTI-298 to affect phosphorylation of pRb, one of the key events required for G 1 /S transition. Calu-1 cells were treated for 48 h with GGTI-298 (15 M), and the cell lysates were immunoblotted with an anti-pRb antibody that recognizes both hypo-and hyperphosphorylated forms of pRb as described under "Experimental Procedures." Other cell lysate aliquots were immunoblotted with antibodies to either Rap1A or RhoA, small G-proteins that are exclusively geranylgeranylated. Finally, a set of cells were analyzed by flow cytometry to determine the proportion of cells at different phases of the cell cycle. Fig. 1A shows that Calu-1 cells treated with the vehicle contained predominantly hyperphosphorylated pRb. Treatment with GGTI-298 resulted in hypophosphorylation of pRb (Fig. 1A). Hypophosphorylation of pRb correlated with inhibition of the geranylgeranylation of the GGTase I substrates, Rap1A and RhoA, and increased the proportion of Calu-1 cells in the G 1 phase of the cell cycle (Fig. 1A). Because processed and unprocessed RhoA migrated closely (Fig. 1A), we confirmed the effects of GGTI-298 on the processing of RhoA by isolating membranes and cytosolic fractions and showing that GGTI-298 decreases the membrane levels of RhoA while inducing accumulation of RhoA in the cytosol (Fig. 1B).

Effects of GGTI-298 on the Expression of Various Cell Cycle Components-
The mechanism by which GGTI-298 results in inhibition of the activities of CDKs could involve inhibition of the expression of the G 1 kinases themselves or their cyclins A, D, and E, or because of an increase in the levels of cyclin-dependent kinase inhibitors. To further investigate these possibilities, Calu-1 cells were treated with GGTI-298 for 48 h, and the lysates immunoblotted with antibodies to CDK2, CDK4, CDK6, cyclins D1, E, and A, p21, p27, p15, and p16 as described under "Experimental Procedures." Fig. 3 shows that controldividing Calu-1 cells expressed all cell cycle components evaluated except p21, which was barely detectable. Treatment of Calu-1 cells with GGTI-298 did not alter the levels of CDK2 and CDK4 or those of cyclins E and D1, indicating that the inhibition of CDK2 and CDK4 activities is not because of a decrease in the levels of cyclins E and D1 or the kinases (Fig. 3). However, the amount of cyclin A was decreased by 20%, and this could contribute to the block of CDK2 kinase activity. The levels of CDK6 decreased by 15%, and this could account for the decrease in its kinase activity (Fig. 2). Furthermore GGTI-298 greatly increased (7-fold) the levels of p21, and to a lesser extent (2-fold) increased the levels of p15 and had little effect on the levels of p16 and p27 (Fig. 3).

GGTI-298 Induces Partner Switching of CDK Inhibitors-
The above data suggest that the inhibition of CDK2 and CDK4 and the subsequent hypophosphorylation of pRb are caused by increased protein levels of some of the CDK inhibitors. Therefore we investigated whether association of the CDKs with their inhibitors was also affected. Calu-1 cells were treated with vehicle (control) or GGTI-298 for 48 h, the lysates were immunoprecipitated with various CDKs, the immunoprecipitates were separated by SDS-PAGE and immunoblotted by various CKIs. Fig. 4 shows that in control cells, p21 and p15 were associated with primarily CDK6, whereas p27 was primarily associated with CDK4 and CDK6. Treatment with GGTI-298 increased (7-fold) association of p21 with CDK2 while decreasing (70%) its association with CDK6. Similarly, GGTI-298 treatment decreased binding of p27 to CDK6 (85%) and CDK4 (22%) and increased binding to CDK2 (2-fold) (Fig.  4). Furthermore, GGTI-298 treatment increased (2-fold) association of p15 to CDK4 with no effects on association with CDK6 (Fig. 4). Finally, GGTI-298 had no effect on the association of p16 with CDK4 or CDK6 (data not shown). Thus, GGTI-298 treatment resulted in a partner switch for p21 and p27 from CDK6 to CDK2. Furthermore, GGTI-298 treatment also induced an increased association of CDK4 with p15 while decreasing its association with p27.
We next determined whether the GGTI-298-induced partner switching is unique to Calu-1 cells. To this end, we treated three other human tumor cell lines; A-549 lung carcinoma, T-24 bladder carcinoma, and A-253 head and neck squamous cell carcinoma with GGTI-298, immunoprecipitated the lysates with CDK2 and CDK6 antibodies, and blotted with p27 antibody as described under "Experimental Procedures." Fig. 5 shows that in all three cell lines treatment with GGTI-298 increased the levels of p27 associated with CDK2 while it decreased the levels associated with CDK6.
Removal of GGTI-298 Results in Reversal of Partner Switching-To determine whether removal of GGTI-298 would result in reversal of G 1 block and partner switching, we treated Calu-1 cells with GGTI-298 for 48 h, fresh medium lacking GGTI-298 was then added, and the cells collected after 0, 4, 24, and 48 h post-GGTI-298 removal. The cells were then harvested and processed for p21 waf induction, immunoprecipitation/blotting and flow cytometry as described under "Experimental Procedures." Fig. 6 shows that, consistent with Fig. 4, GGTI-298 induced p21 waf , partner switching of p27 from CDK6 to CDK2 and G 1 block (compare control versus GGTI-298 lanes of Fig. 6). GGTI-298 had little effect on the association of p27 with CDK4. Removal of GGTI-298 resulted in a time-dependent reversal of p21 waf induction, partner switching, and G 1 block. Little change was observed 4 h after removal. However, significant reversal occurred at 48 h (Fig. 6). DISCUSSION Although there are more geranylgeranylated proteins than farnesylated proteins, more efforts have been spent on designing, synthesizing, and biologically characterizing FTase rather than GGTase I inhibitors (8). The intense search for FTase inhibitors was prompted some years ago by the realization that farnesylation is required for the cancer-causing activity of the important oncoprotein Ras (8). Recently, however, more attention has been directed toward understanding the effects of inhibiting (by GGTase I inhibitors) the function of geranylgeranylated proteins. This is primarily due to the discovery that some geranylgeranylated proteins such as those from the Rho family are essential for normal and aberrant proliferation in several cell types (1,5). In addition to their ability to inhibit human tumor growth, GGTase I inhibitors may also have great therapeutic potential in cardiovascular diseases. For example, GGTI-298 blocks the ability of smooth muscle cells to proliferate by inducing a G 1 block and apoptosis (16). This may be related to the inhibition by GGTI-298 of PDGF and Ras induction of superoxide formation (18) suggesting that a geranylgeranylated protein downstream of Ras is critical to events regulating cell division. This is consistent with a recent report, which shows that the geranylgeranylated protein Rac1 mediates superoxide formation and transformation by Ras (24). In this study, we demonstrate that GGTI-298 treatment of the human lung carcinoma Calu-1 cells results in a large increase of the CDK inhibitor p21 and a modest increase of p15 with little effect on p27 and p16. In nontreated dividing cells, CDK2 and CDK4 bound mainly p27 and their kinases were active, whereas CDK6 bound all inhibitors and was also active. Upon treatment with GGTI-298, p21 and p27 switched partners from CDK6 to CDK2, whereas p15 became bound to CDK4. The effects of GGTI-298 on the observed partner switching was reversible. We found that removal of GGTI-298 resulted in reversal of the G 1 phase block, which was paralleled by a reversal of partner switching of p27 from CDK2 to CDK6.
The fact that in dividing cells, CDKs are bound to some inhibitors is not surprising, because low levels of such inhibitors may be required to stabilize cyclins with the corresponding CDKs (25). This is also consistent with the suggestion that some CDKs may serve as cellular reservoirs for low levels of inhibitors (26). Furthermore, although there was little increase in total cellular levels of p27 after GGTI-298 treatment, the amount of p27 bound to CDK2 was increased. This suggests that p27 that was released from CDK6, and to a lesser degree from CDK4, bound to CDK2. Taken altogether, the data suggests that GGTI-298 induced partner switching for p21 and p27 from CDK6 to CDK2. This partner switching was not unique to Calu-1 cells and was found to occur in three other human tumor cell lines, A-549, T-24, and A-253. These results are similar to those obtained from cells that were arrested in late G 1 following treatment with transforming growth factor ␤ (27). In these cells, transforming growth factor ␤ induced the expression of p15, which bound CDK4 and CDK6 and displaced p27 from cyclin D⅐CDK4 and cyclin D⅐CDK6 complexes (27). The free p27 bound the cyclin E⅐CDK2 complex and inhibited its activity, which resulted in preventing cells from entering S phase (27).
The fact that inhibition of protein geranylgeranylation results in G 1 arrest suggests that some geranylgeranylated proteins are required for cell progression from G 1 to S. The most likely candidates are members of the Rho family of proteins such as RhoA and Rac1. In this study, we showed that GGTI-298 inhibits the processing of RhoA, and recently we have demonstrated that RhoA represses p21 waf transcription and suggested that GGTI-298 induction of p21 waf is mediated by inhibition of RhoA geranylgeranylation (29). Furthermore, it has been shown that both RhoA and Rac1, when microinjected into G 1 -arrested cells, promote progression to S phase (5). Furthermore, dominant-negative Rac1 and RhoA reverse oncogenic Ras transformation (6,7). Rac1 may regulate transcription by activating the Jun kinase pathway that in turn will affect genes that are regulated by AP-1. Furthermore, very recently Tapon et al. (28) discovered a novel effector of Rac1, POSH, a kinase that appears to be required for activation of Jun kinase by Rac1. It would be of great interest to determine whether activated Jun kinase or POSH would rescue the cells from the GGTI-298 mediated G 1 block. These studies, which are in progress, will further enhance our understanding of how inhibition of protein geranylgeranylation results in G 1 arrest.