Post-translational modification of low molecular mass GTP-binding proteins by isoprenoid.

Several proteins in mammalian cells are modified post-translationally by the isoprenoid, farnesol. Incubation of cultured cells with [3H]mevalonate, an isoprenoid precursor, results in the labeling of multiple polypeptides, the most prominent of which migrate in the range of 21-26 kDa on sodium dodecyl sulfate-polyacrylamide gels. In Rat-6 fibroblasts transformed by H-ras, one of the farnesylated proteins was identified as p21ras by two-dimensional immunoblotting. However, this protein accounted for only a small proportion of the [3H]mevalonate-derived radioactivity incorporated into 21-26-kDa proteins. Murine erythroleukemia cells, which did not express immunodetectable quantities of p21ras, contained several 21-26-kDa farnesylated proteins distributed in both the cytosolic and particulate fractions. At least eight of these proteins were capable of binding [alpha-32P]GTP on nitrocellulose membranes. Pulse-chase studies showed that the isoprenoid modification did not necessarily result in the translocation of the cytosolic proteins to the cell membrane. A prominent group of carboxyl-methylated proteins in murine erythroleukemia cells overlapped with the 21-26-kDa farnesylated proteins on one-dimensional sodium dodecyl sulfate gels. Methylation of this group of proteins was selectively abolished when cells were treated with lovastatin, an inhibitor of isoprenoid synthesis. Addition of exogenous mevalonate to the lovastatin-treated cells fully restored carboxyl methylation. These studies suggest that the 21-26-kDa farnesylated proteins in mammalian cells are members of a recently discovered family of low molecular mass GTP-binding proteins which, although ras-related, appear to be distinct structurally and possibly functionally from the products of the ras genes. The observed isoprenoid-dependent carboxyl methylation of a group of 21-26-kDa proteins suggests that the low molecular mass GTP-binding proteins may undergo a series of post-translational C-terminal cysteine modifications (i.e. farnesylation, carboxyl methylation) analogous to those recently elucidated for p21ras.

These studies suggest that the 21-26-kDa farnesylated proteins in mammalian cells are members of a recently discovered family of low molecular mass GTP-binding proteins which, although r-as-related, appear to be distinct structurally and possibly functionally from the products of the ras genes. The observed isoprenoid-dependent carboxyl methylation of a group of 21-26-kDa proteins suggests that the low molecular mass GTP-binding proteins may undergo a series of post-translational C-terminal cysteine modifications (i.e. farnesylation, carboxyl methylation) analogous to those recently elucidated for p2 1'"". for some time that depriving cells of MVA leads to arrest of cell proliferation (2-7) and, in some cases, increased expression of differentiated properties (6,7). This occurs even when the cellular requirement for cholesterol is satisfied by exogenous lipoproteins in the culture medium, suggesting that nonsterol isoprenoid derivatives of MVA play a permissive or regulatory role in cell replication (3-7). In recent years, interest in this area has been stimulated by reports that several proteins in mammalian cells are modified covalently by isoprenoid derivatives of . This novel post-translational modification has been discovered in a variety of cell types by tracing the incorporation of radiolabeled MVA into proteins separated on SDS-polyacrylamide gels. In all of the mammalian cell lines studied to date, most of the radioactivity derived from 3H-or '%-labeled MVA is incorporated into a cluster of proteins with molecular masses between 21 and 26 kDa. Smaller amounts of radioactivity are detected in proteins with molecular masses of 17, 45,53,[66][67][68][69][70]9,[12][13][14][15].
The majority of the modified proteins remain to be identified, but progress has been made in two specific cases. Isoprenylated proteins in the range of 66-70 kDa have been localized to the nuclear matrix (12, 15), and two of these proteins have been shown to react with antibodies against lamins A and B (16, 17). One of the low molecular mass proteins has been clearly identified as p21'"". Hancock et al. (18) demonstrated that the H-, K-, and N-~21'"" proteins undergo isoprenoid modification at CYS'~ in what appears to be the first step in a series of post-translational C-terminal processing events, i.e. isoprenylation, proteolytic removal of 3 amino acids distal to the isoprenylation site, carboxyl methylation of the exposed isoprenylated C-terminal cysteine, and palmitoylation of a cysteine residue upstream from the isoprenylated cysteine (U-20). Consistent with the isoprenoid modification of p21'"", Schafer et al. (21) demonstrated that expression of the activity of an oncogenic H-c-ras protein in Xenopus oocytes could be blocked by microinjection of an inhibitor of MVA synthesis. The exact chemical structure of the isoprenoid moiety involved in the modification of mammalian cell proteins remains to be defined. However, our recent chromatographic characterization of the isoprenoid released by sulfonium salt cleavage of low molecular mass proteins from murine erythroleukemia (MEL) cells suggests that the modifying group is farnesol, linked to cysteine via a thioether bond (22). Parallel studies focusing specifically on the isoprenoids released from immunoprecipitated ~21'"" (23) and the nuclear lamins (24) have provided independent evidence for a farnesyl modification of these proteins.
Thus, throughout this report, the covalent modification of cellular proteins by isoprenoid derivatives of MVA will be referred to as farnesylation.
We found the recent demonstration of an isoprenoid modification of p21'"" to be of particular interest in light of an 2148 early study in which we observed no obvious differences in the ['%]MVA-labeling of 21-26-kDa proteins in H-ras-transformed fibroblasts compared with nontransformed fibroblasts (12). In the present report, we demonstrate that although p21'"" is indeed modified by isoprenoid in rus-transformed cells, most of the 21-26-kDa MVA-labeled proteins in mammalian cells are immunologically and electrophoretically distinct from p21'"". We also provide the first direct evidence that several of these non-ras farnesylated proteins are capable of binding GTP and that methylation of proteins with the same molecular mass as the farnesylated GTP-binding proteins is blocked by an inhibitor of isoprenoid synthesis. Based on these observations, we suggest that the unidentified 21-26-kDa farnesylated proteins are members of the large family of low molecular mass GTP-binding proteins (25, 26) that may undergo C-terminal modifications similar to those described for ~21'"". Panel A, suspension cultures of MEL cells were initiated at a density of 100,000 cells/ml and grown for 2 days. A total of 4 x lo6 cells was then incubated for 18 h in 2 ml of medium containing 200 pCi/ml [3H]MVA and 25 FM lovastatin. H-ra.s-transformed Rat-6 cells were plated at 12,000 cells/cm* in a 25.cm2 flask and grown for 2 days. The monolayer culture was then incubated for 18 h with ["H]MVA under the same conditions as those used for the MEL cells. The soluble (S) and particulate (P) fractions were obtained from each cell line as described under "Experimental Procedures," and the total protein in each fraction was subjected to SDS-PAGE and fluorography. The gel shown in panel A was exposed for 5 days. The molecular mass marker proteins are shown at the left, and the position of the tracking dye is indicated by the arrow. The amounts of protein in each fraction loaded on the gel were as follows: Rat-6 (S), 55 rg; Rat-6 (P), 200 pg; MEL (S), 110 pg; MEL (P), 280 wg. Panel B, cells were grown as described above except that labeling with ['H]MVA was omitted. Protein from the soluble or particulate fractions from each cell type was subjected to SDS-PAGE, transferred to nitrocellulose, and immunoblotted with ra.s-11 (pan) antibody to p21'"'. Lanes were loaded with 310 rg of soluble protein (S) or 1.5 mg of particulate protein (P), and bound primary antibody was detected with horseradish peroxidase-conjugated goat anti-mouse IgG (see "Experimental Procedures"). Rat-6 cells showed immunologically detectable p21'"", and this was confined to the particulate fraction ( Fig. 1, B and C). This observation suggested that most, if not all, of the low molecular mass farnesylated proteins in the MEL cells were structurally distinct from the ~21'"" proteins. To determine the extent to which the farnesylation of p21"" contributed to the profile of [3H]MVA-labeled 21-26-kDa proteins in the H-ras-transformed Rat-6 cells, the proteins in the particulate fraction were separated by two-dimensional electrophoresis.
Western blots of the [3H]MVA-labeled proteins were subjected to fluorography, then immunoblotted with the ras-11 (pan) antibody. As shown in Fig. 2, the p21'""immunoreactive protein accounted for a relatively small portion of the [3H]MVA labeling seen in the 21-26-kDa region, and it was only one of many farnesylated proteins, most of which were not recognized by the antibody and had slightly higher molecular masses than the ras protein.
GTP Binding to 21-26-kDa Farnesylated Proteins-A number of GTP-binding proteins with molecular masses in the Zl-26-kDa range are not recognized by the standard pan-ra.s antibodies (37)(38)(39)(40). To determine whether any of the nonimmunoreactive farnesylated proteins in the MEL cells might be members of this family of proteins, we took advantage of the fact that the low molecular mass GTP-binding proteins retain their capacity to bind guanine nucleotides after being transferred from SDS gels to nitrocellulose membranes (35,38). As shown in Fig This necessitated longer exposures to visualize the bound GTP and raised the possibility that the number of GTPbinding proteins might be underestimated. Nevertheless, these disadvantages were outweighed by the ability to match [3H]MVA-labeled proteins and [32P]GTP-binding proteins precisely on the same blot. Using this approach, it appeared that at least eight of the farnesylated proteins in the MEL cell-soluble fraction and two of the proteins in the particulate fraction were capable of binding GTP (Fig. 3, C and E). The two prominent farnesylated GTP-binding proteins in the particulate fraction (Fig. 3E) closely matched two of the farnesylated GTP-binding proteins in the soluble fraction ( Fig. 3C) with respect to their electrophoretic mobility, suggesting that they might be the same or closely related proteins. Rat-6 cells were plated in a 75-cm2 flask at a density of 13,000 cells/cm' and grown for 24 h. Medium containing 200 &i/ml [3H]MVA and 25 pM lovastatin was then added, and the incubation was continued for 16 h. The particulate protein (150 pg) was separated by two-dimensional electrophoresis and transferred to nitrocellulose (see "Experimental Procedures"). Panel A shows the fluorograph of the blot after a lo-day exposure, with the arrow indicating the [3H]MVA-labeled protein overlapping with the ~21"" protein detected by immunoblotting with the ras-11 (pan) antibody (panel B). The radiolabeled (panel A) or prestained (panel B) molecular mass standards shown on the right edge of the gel were run in the second dimension. The acidic end of the isoelectric focusing (ZEF) gel was oriented toward the left.

Distribution of Farnesylated Proteins between Soluble and
Particulate Fractions in MEL Cells-Farnesylated ~21"" was detected only in the particulate fraction of transformed Rat-6 cells (see Fig. l), presumably because the mature palmitoylated form of the protein is localized in the cell membrane (18,19,41). In contrast, the farnesylated 21-26-kDa proteins in MEL cells were equally distributed between the soluble and particulate fractions. To test the possibility that the soluble [3H]MVA-labeled proteins were intermediates in a series of processing steps which would ultimately lead to their insertion into the membrane, we performed a pulse-chase study depicted in Fig. 4. Cells were incubated with [3H]MVA in the presence of lovastatin to allow maximum labeling of farnesylated proteins. The labeled MVA was then removed from the medium, and the cells were incubated for 24 h without lovastatin, allowing the endogenous biosynthetic pathway to generate MVA for continued cell growth. When the electrophoretic profiles of the labeled proteins in the soluble and particulate fractions were examined at intervals after removing [3H]MVA from the medium, there was no evidence that the farnesylated proteins present in the cytosol at the end of the pulse-labeling period were translocated to the membrane compartment during the chase period. Instead, there was little change in the distribution of radiolabeled proteins between the soluble and particulate fractions. By 24 h we observed a uniform diminution of radioactivity in the proteins in both subcellular compartments as cell growth proceeded in the absence of labeled MVA. In view of our previous finding that the farnesylated proteins in MEL cells are relatively stable (14), we attribute the decrease in radioactivity to dilution of the labeled farnesylated proteins as cell    growth and protein synthesis continued in the absence of labeled MVA.
Isoprenoid-dependent Methylation of 21-26-kDa Proteins-As mentioned earlier, newly synthesized ~21'"" (pro-p21) is rapidly farnesylated at Cye?, triggering proteolytic removal of three C-terminal amino acids and carboxyl methylation of the exposed C-terminal cysteine (18,19). With this sequence of events in mind, we considered the possibility that the nonras GTP-binding proteins also might undergo carboxyl methylation in conjunction with their farnesylation. The gel slice profiles in Fig. 5 show that a prominent group of methylated proteins overlapped with the 21-26-kDa ["HIMVA-labeled proteins in MEL cells (compare panels A and B). This in itself was not particularly informative, since the methylation assay detects O-methyl derivatives of aspartate and glutamate (36) as well as the newly discovered cysteine carboxyl methylation (20,42,43). However, upon addition of lovastatin, an inhibitor of MVA synthesis which blocks the formation of cellular isoprenoids, the methylation of the 21-26-kDa proteins was selectively diminished, whereas methylation of other proteins in the gels was not noticeably affected (Fig. 5C). To confirm that the decreased methylation of these proteins was due specifically to MVA deprivation, we added exogenous MVA to the cultures along with lovastatin and were able to restore methylation of the 21-26-kDa proteins (Fig. 50). Because the proteins were separated by one-dimensional SDS-PAGE, it is not possible to state definitively that the methylated proteins in the 21-26-kDa region were the same proteins that were labeled with ["HIMVA. However, the sensitivity of the methylation of this specific group of proteins to lovastatin strongly supports the notion that their methylation was dependent on prior or concurrent isoprenoid modification.

DISCUSSION
Since Schmidt et al. (8) first described the covalent modification of 3T3 cell proteins by unidentified isoprenoids derived from MVA, reports from several laboratories have established that this novel type of post-translational modification is ubiquitous in mammalian cells (9-15). Although it has been known for some time that yeast peptide-mating factors undergo isoprenoid modification and post-translational processing at their C termini (i.e. farnesylation of cysteine, proteolytic removal of 3 distal amino acids, and carboxyl methylation of the farnesylated cysteine) (44)(45)(46)(47)(48), this fact initially received little attention because of the apparent lack of structural or functional relationships between these peptides and the isoprenylated proteins in mammalian cells. However, recent reports that ~21'"" undergoes isoprenoid modification at Cysla6 (18, 23), C-terminal proteolytic processing (18,19), and carboxyl methylation (18, 20) have suggested that a homologous C-terminal amino acid motif in the yeast-mating peptides and ~21'"" (i.e. Cys-A-A-X where A is an amino acid with an aliphatic side chain and X is any amino acid), may represent a consensus sequence for farnesylation. Two additional observations tend to support this hypothesis. The first is the finding that the modifying isoprenoid released from mammalian proteins by thioether cleavage does indeed behave as farnesol or its rearrangement products in several chromatography systems (22)(23)(24), and the second is the observation that the nuclear lamins, which undergo isoprenoid modification (16,17,24) and carboxyl methylation (49,50), apparently share the Cys-A-A-X C-terminal motif (51)(52)(53) into 21-26-kDa proteins to an extent comparable to that seen in H-ras-transformed fibroblasts. In both the MEL cells and the rus-transformed fibroblasts, most of the radioactivity derived from [3H]MVA was incorporated into nonimmunoreactive proteins with slightly higher molecular masses (23-26 kDa) than the immunoreactive ~21'"". Several of these farnesylated proteins in the MEL cells were able to bind GTP when transferred to nitrocellulose membranes, a feature that is characteristic of a growing number of non-ras guanine nucleotide-binding proteins identified recently in mammalian cells. Although the physiological functions of these proteins have yet to be defined, they generally exhibit some degree of homology with ~21"" and have hence been termed "ras-related" (25,26). Using the information available in the literature, it is possible to separate these proteins into three general groups based on their C-terminal amino acid sequences and their sizes. The first group consists of 21-kDa proteins such as rho (40,54,55), smg-p21(56), ral(57), and rap1 (58), which have a relatively high degree of homology to the H-, K-, and N-r-as proteins and share the C-terminal Cys-A-A-X motif. The second group consists of 23-24-kDa proteins such as yptl (59), rabl, and rab2 (60), which are more closely related to the yeast YPTl and SEC4 proteins than to p21'"", and have C-terminal sequences ending with 2 consecutive cysteine residues. The third group includes the smg25 proteins (61) and the rab3 and rab4 proteins (60, 62), which have molecular masses of approximately 25 kDa and a C-terminal motif of Cys-X-Cys. In addition to the proteins for which sequence data are available, a number of poorly characterized 24-26-kDa guanine nucleotide-binding proteins have been isolated from diverse tissues such as human placenta and platelets (63), bovine brain (64), human leukemia cells (65), and 3T3 flbroblast membranes (39). Based on their molecular masses, most of the farnesylated GTP-binding proteins in the MEL cells correspond more closely to proteins such as yptl, rab, and smg25 than to the ros-like PI-kDa proteins rho, ral, and smg21. Thus, our findings raise the possibility that classes of GTP binding proteins with C-terminal sequences ending in Cys-Cys or Cys-X-Cys, instead of Cys-A-A-X may also undergo post-translational modification by isoprenoid.
Studies of mammalian and yeast ras proteins have suggested that they are methylated by a novel carboxyl methyltransferase activity that is specific for C-terminal cysteine (20,42). The observation that a Cys + SerZs6 point mutation prevents carboxyl methylation of ~21'"" (18) is consistent with a model in which methylation of CyslE6 is contingent upon its isoprenoid modification and the subsequent proteolytic removal of 3 distal amino acids. Although it is not known whether other low molecular mass GTP-binding proteins undergo C-terminal cysteine carboxyl methylation, two recent observations suggest that this may occur. First is the demonstration of this type of methylation in a group of 23-29-kDa proteins from retinal rod outer segment membranes (43), and second is the finding that in vitro methylation of non-ra.s 20-23-kDa macrophage membrane proteins is stimulated by GTP (66). In the present report we provide the first direct evidence that carboxyl methylation of 21-26-kDa proteins in cultured mammalian cells is selectively abolished by an inhibitor of isoprenoid synthesis and restored by provision of the isoprenoid precursor, MVA. Thus, it is reasonable to speculate that the non-ras low molecular mass GTP-binding proteins undergo C-terminal modification similar to those described for p21"" and that farnesylation of cysteine residues at or near the C-terminal is a prerequisite for their methylation.
The pulse-chase experiment described in this report (Fig.  4)  For example, the proteins found in the soluble and particulate fractions may be structurally unique so that in some cases farnesylation confers sufficient hydrophobicity for stable membrane localization, whereas in other cases it does not. Alternatively, the soluble and particulate proteins may be closely related or identical translation products that are in a state of dynamic equilibrium between the two subcellular compartments, possibly determined by an additional post-translational modification such as fatty acylation. Although further studies are needed to distinguish between these models, the present studies would seem to rule out the concept that farnesylation necessarily targets all low molecular mass GTP-binding proteins to the cell membrane. The biochemical basis for the well documented arrest of cell cycling which occurs when cells are deprived of MVA remains to be firmly established. The fact that growth arrest cannot be reversed by providing exogenous sterols (3-7) has led to a search for nonsterol isoprenoids that may be required for cell replication.
Initially it was thought that isopentenyladenine, a precursor for isopentenyl tRNA, might be a key isoprenoid involved in growth regulation (67), but subsequent studies failed to support this hypothesis (9,12,68,69). The intracellular supply of ubiquinone (coenzyme Q), an isoprenoid carrier of reducing equivalents in the mitochondrial electron transport chain (70), also does not appear to be critical, since cells can tolerate a 50% reduction in the mitochondrial ubiquinone pool with little effect on respiratory function (71). Moreover, supplementing cells with ubiquinone does not reverse the arrest of growth induced by inhibiting MVA synthesis (12), despite the fact that exogenous ubiquinone is readily taken up by mitochondria in cultured cells (72). Similar studies have shown that supplementing cells with dolichol cannot reverse the arrest of cell cycling in MVAdepleted cells (12, 73). However, because of uncertainties regarding the extent to which free dolichol may be converted to the phosphate ester, the depletion of intracellular dolichyl phosphate and consequential perturbation of N-linked protein glycosylation still must be considered as a viable mechanism that could lead to growth inhibition.
The discovery that multiple proteins in mammalian cells undergo covalent post-translational modification by MVAderived farnesyl groups has placed the MVA requirement for cell cycling in an entirely new perspective.
Soon after the discovery of protein isoprenylation, it became clear that this type of modification could provide an explanation for the fact that proliferating cells require small amounts of MVA for cell growth, even when their cholesterol requirement is met. For example, in a Chinese hamster ovary cell line (mev-1), which is auxotrophic for MVA, cell cycle arrest induced by MVA deprivation was closely correlated with a decline in the concentration of proteins radiolabeled with MVA (9). Similar findings were observed in neuroblastoma cells induced to enter a quiescent differentiated state by treatment with an inhibitor of MVA synthesis (12). The potential connection between protein isoprenylation and cell growth was strengthened by our recent finding that this modification is tightly coupled to MVA synthesis and isoprenoid supply, i.e. if MVA synthesis is blocked, the pool of isoprenoid (farnesyl) groups available for protein modification is immediately depleted, leading to a rapid accumulation of nonmodified proteins (14). The nuclear lamin proteins are thought to play important roles in the organization of the nuclear envelope (74), and it is therefore conceivable that blocking their isoprenoid modification (16, 17) could disrupt their normal interactions with components of the nuclear membrane and scaffold. At the same time, prevention of the farnesylation of ras proteins or other GTP-binding proteins could impinge on cellular responses to growth factors, signal transduction, or intracellular transport functions in which these proteins have been postulated to play a role (25,26,(75)(76)(77). Thus, decreased MVA synthesis may ultimately affect cell cycling by interfering with a multitude of protein-mediated events that are presently difficult to differentiate from one another. With this in mind, it would seem that the most productive approach to understanding the complex interrelationships between isoprenoid synthesis and cell growth must involve the delineation of the structures and functions of the individual farnesylated proteins and the determination of how the modification of these proteins affects their normal activity. The findings reported herein represent a starting point for such studies insofar as they demonstrate that post-translational modification by isoprenoid has broad relevance for a class of low molecular mass GTP-binding proteins distinct from ~21'"".