Isoprenoid Modification of G25K (Gp), a Low Molecular Mass GTP-binding Protein Distinct from p2 lras*

a number of low molecular mass GTP-binding post-translational We used two-dimensional immunoblotting to the isoprenoid an antibody a low molecular mass GTP-binding protein originally purified placental,

Cultured murine erythroleukemia (MEL) cells synthesize a number of low molecular mass GTP-binding proteins that undergo post-translational modification by isoprenoids.
We used two-dimensional electrophoresis and immunoblotting to show that a 23-24-kDa protein labeled by the isoprenoid precursor [3H]mevalonate was specifically recognized by an antibody to G25K (Gp), a low molecular mass GTP-binding protein originally purified from placental, platelet, and brain membranes.
Several isoelectric variants of G25K were detected in MEL cells, and all were radiolabeled with [3H]mevalonte.
The G25K-immunoreactive protein did not cross-react with pan-r-as antibody.
Although mature p2 1'8' is known to be localized in the cell membrane, most of the isoprenylated G25K was found in the 100,000 x g supernatant fraction when cells were lysed in buffer without detergent. Blocking isoprenoid synthesis by incubation of MEL cells with lovastatin resulted in a decrease in the concentration of G25K in the particulate fraction and a corresponding increase in immunodetectable protein in the soluble fraction. Lovastatin treatment also produced shifts in the electrophoretic mobilities of the G25K isoforms on twodimensional gels. These observations are consistent with the idea that isoprenylation plays a permissive role in the association of G25K with the cell membrane or other organelles.
However, the high proportion of soluble isoprenylated G25K in MEL cells under normal culture conditions suggests that the role of the isoprenoid modification may be more complex than simply serving as a structural anchor for stable insertion of proteins into the lipid bilayer.
Mevalonic acid (MVA)' serves as a precursor for the biosynthesis of a variety of isoprenoid products including cholesterol, dolichols, the side chain of ubiquinone (coenzyme Q), and the farnesyl moiety of heme a (l-3).
In 1984, Schmidt et ul. (4)  into several categories with respect to their electrophoretic mobilities on SDS gels and their subcellular localizations (7,9). Most recently, progress has been made in characterizing the modifying isoprenoid structures as being either farnesyl or diterpene groups, linked to the proteins via cysteine thioether bonds (10-13).
With only a few exceptions, the identities of the isoprenylated proteins remain to be determined.
Two of the isoprenylated proteins that are localized in the nuclear matrix and which migrate between 66 and 72 kDa have been identified as lamin B (11, 14-16) and prelamin A (15-17). In the latter case, isoprenylation represents a key step in the proteolytic processing of prelamin A to mature lamin A (17). Several recent studies have shown that ~21'~' also undergoes isoprenylation in what appears to be the first step in a series of posttranslational processing events that occur at the C terminus of the protein (18-21). One of the few structural features common to both the lamins and p21ras is the presence of a Cterminal amino acid sequence consisting of cysteine followed by 3 additional amino acids. Since the first 2 amino acids distal to the cysteine generally have aliphatic or hydroxy side chains, this amino acid sequence is frequently referred to as a Cys-A-A-X motif (where A is an aliphatic residue and X is any amino acid). The Cys-A-A-X sequence is found in several fungal mating factors that undergo farnesylation at C-terminal cysteine residues (22-24), and studies using mutational analysis have established that this is indeed the site of isoprenoid modification in mammalian ~21'~' (18) and avian nuclear lamins (16).
Mammalian cells contain many low molecular mass (20-30 kDa) GTP-binding proteins in addition to H-, K-, and N-p21ra', e.g. rho (25-27), yptl/rubl (28, 29), rab 2,3,4 (29, 30), smg21 (rapl) (31, 32), rul (33, 34), and rucl,2 (34, 35). Although these proteins are structurally distinct from each other and from p21ras, their sequences all show some degree of homology with that of p21ra', particularly in the regions contributing to the GTP-binding site. We reported recently that cultured MEL cells and fibroblasts contain several low molecular mass GTP-binding proteins that are not recognized by antibodies to p21ras but nevertheless undergo isoprenylation (36). Although some of these may be novel gene products, it seemed likely that others might be previously identified GTPbinding proteins. Among the low molecular mass GTP-binding proteins described in the literature, one particular protein named G25K (formerly G,,) attracted our interest because of its seemingly ubiquitous distribution in normal tissues and a variety of cultured cell lines (37)(38)(39).
We now demonstrate that proteins identified as G25K (GJ by two-dimensional immunoblotting are modified by isoprenoid in cultured MEL cells. In contrast to the mature form of p21ras, which is localized in the cell membrane, a substantial proportion of isoprenylated G25K is concentrated in the soluble fraction. This finding implies that although isopren- min prior to the initial blocking step. Primary antibodies were diluted to l/l,000 in buffer containing 5% dry milk, 2% Nonidet P-40, and 0.2% SDS, and incubation was performed at room temperature for 2 h. Secondary antibodies (goat '*'I-anti-mouse IgG for anti-p21"' primary antibody, or goat '*'I-anti-rabbit IgG for anti-G25K or anti-Gp primary antibodies) were added to the same buffer at 500,000 dpm/ml, and incubation was carried out for 1 h. Blots were covered with Saran Wrap and exposed to Kodak X-Omat AR film with a Cronex intensifying screen for various periods of time (see tigure legends) at -80 'C. The soluble and particulate fractions were then subjected to SDS-PAGE, transferred to nitrocellulose membranes, and fluorographed to establish the positions of the isoprenylated proteins. The same blots were incubated subsequently with the antibody to G25K or with a monoclonal antibody (ras-11, pan) which recognizes all H-, K-, and N-~21'~' proteins (for additional details, see legend to Fig. 1). In agreement with our previous studies of MEL cells (36), most of the isoprenylated proteins were found in the 21-26 kDa region of the SDS gel and were about equally concentrated in the soluble and particulate fractions (Fig. 1). The soluble fraction contained a protein that reacted strongly with the antibody to GZ5K. The broad band of immunoreactive protein at approximately 23-24 kDa fell within a cluster of proteins labeled heavily with [3H]MVA, prompting further study by two-dimensional electrophoresis (see Fig. 5, described below). The antibody to G25K also gave a weak signal with a protein at the same molecular mass in the particulate fraction.
In a previous study with MEL cells (36), we were unable to detect p21raa by immunoblotting with the pan-r-as antibody, using a horseradish peroxidase-conjugated secondary antibody. However, in the present study, the increased sensitivity of the 1*51-labeled secondary antibody revealed a 26-kDa protein that reacted with the pan-ras antibody in the particulate Each of the fluorographed strips of nitrocellulose was cut in half, and the fluor was washed out. The matching halves of the strips were then immunoblotted with antibodies to G25K or p21ras, using an '*'I-labeled secondary antibody for detection of bound IgG (see "Experimental Procedures"). Exposure time for fluorography of the [3H]MVA-labeled proteins was 5 days. The "'Ilabeled immunoblots were exposed for 4 h. %.s., molecular mass standards.
fraction of the LR-MEL cells. This protein clearly migrated at a higher molecular mass than the protein recognized by anti-G25K; and contrary to G25K, it was not found in the soluble fraction (Fig. 1).
The 26-kDa protein reacting with the pan-ras antibody in the MEL cells appeared to be different from both p21H.r's and G25K with respect to its electrophoretic mobility. Thus, when COS cells were transfected with cDNA coding for ~21".~", transient expression of the protein in cells incubated with [3H]MVA resulted in the appearance of a new MVA-labeled protein with a molecular mass of 21 kDa on Western blots (Fig. 2). This was consistent with the documented isoprenylation of mature p21"' (18-21). The mobility of the 21-kDa ['H]MVA-labeled protein in the transfected COS cells matched that of the greatly increased p21r*' band seen on the immunoblot and was clearly different from that of the 26-kDa protein detected by the pan-ras antibody in the MEL cells (Fig. 2). Confirmation of the Presence of G25K in the Soluble Fractions of Different Cell Lines-The finding that the G25Kimmunoreactive protein detected in LR-MEL cells was concentrated in a soluble fraction generated by lysis of cells in buffer without detergent was surprising since G25K (GJ originally was isolated from placental and brain membranes (37,38). In a previous study of G25K in cultured cells, Polakis et al. (39)  brane fractions or whole cell lysates, these studies did not establish whether a significant proportion of G25K was present in the soluble fraction.
The unexpected subcellular distribution of the immunoreactive protein in the present study led us to consider whether the antipeptide antibody might have been recognizing a soluble protein that shared the G25K pl peptide sequence but was in fact a protein different from the Gp protein originally isolated by Evans et al. (37). To test this possibility we compared the immunoblot patterns obtained with the anti-p1 peptide antibody with those obtained with a polyclonal antiserum raised against whole Gp purified from platelet membranes. As shown in Fig. 3 (panel a), the antibody against the G25K pl peptide gave a strongly positive reaction when tested against authentic placental G,, or platelet G25K. Conversely, the polyclonal antiserum against G,, also reacted with both the placental and platelet proteins. When both antibodies were used in parallel immunoblot assays against soluble and particulate protein fractions from LR-MEL cells, the polyclonal antiserum against Gp gave results that were essentially identical to those obtained with the antipeptide antibody; i.e. a broad immunoreactive band with a mobility corresponding to that of pure G25K, concentrated predominantly in the soluble fraction (Fig. 3, panel b).
Another possible explanation for the high proportion of G25K in the soluble fraction was that this subcellular distribution was a unique feature of the LR-MEL cells. Therefore, we compared the distribution of G25K between the total soluble and particulate compartments in several additional Isoprenoid Modification of G25K CGJ cell lines (Fig. 4). The salient feature of the results is that a substantial proportion of G25K was present in the soluble fraction in every cell line examined. As in the case of the LR-MEL cells, the protein was predominantly cytosolic in human HL-60 myeloid leukemia cells, simian COS cells, and the parent Friend MEL cell line, which had not been adapted to G2SK  partially purified platelet G25K (9 pg) or placental Gp (20 pg) was subjected to SDS-PAGE and Western blotting as indicated at the top of each lane. lmmunoblotting was then performed with affinitypurified antibody against G25K pl peptide or with polyclonal antiserum against platelet Gr, as indicated at the oorrorn of each punel. After incubation with l*sI-labeled secondary antibody (see "Experimental Procedures") the immunoblots were exposed for 2 h. b, LR-MEL cells from an exponentially growing suspension culture were homogenized and fractionated as described under "Experimental Procedures." Aliquots of soluble (S) or particulate (P) fractions (150 pg of protein/lane) were subjected to SDS-PAGE and transferred to nitrocellulose. Immunoblotting was performed with the anti-G25K peptide antibody or the anti-Gr polyclonal antiserum as indicated below each punel. The blots were exposed for 8 h to visualize the '*'Ilabeled secondary antibody. In normal Rat-6 fibroblasts, the relative amount of immunoreactive protein detected in the particulate fraction was slightly greater than in the soluble fraction whereas in the H-ra.s-transformed Rat-6 fibroblasts the G25K appeared to be about equally distributed between the two compartments. Demonstration of the Isoprenylation of G25K by Two-dimensional Gel Analysis-In view of the overlap of proteins recognized by antibodies to G25K with proteins labeled by [3H] MVA on one-dimensional SDS gels (Fig. l), we performed a two-dimensional immunoblot analysis of MEL cell proteins to establish whether or not G25K was actually one of the isoprenylated proteins. The [3H]MVA-labeled proteins from the soluble and particulate fractions were subjected to twodimensional electrophoresis, fluorography, and immunoblotting with the anti-G25K pl antibody. In anticipation of performing further studies on the effects of blocking isoprenoid synthesis on the electrophoretic mobility and subcellular distribution of G25K (see Fig. 6), we elected to perform these studies with the parent MEL cell line, which had not been adapted to grow in the presence of lovastatin and therefore retained full sensitivity to this inhibitor of MVA synthesis. The results, which are shown in Fig. 5 Cells were harvested from suspension or monolayer cultures that were in the exponential phase of growth. Fractionation of cells into soluble (S) and particulate (P) components was performed as described under "Experimental Procedures." In each case, the total soluble and particulate fractions were loaded in adjacent lanes and subjected to SDS-PAGE and immunoblotting with the anti-G25K pl peptide antibody. The blots shown in paneLs a and b were from two separate experiments. Those in panel a were exposed for 17 h, and those in panel b were exposed for 4 h. The actual protein values for the total soluble and particulate fractions loaded in each case were as follows: o: LR-MEL: P = 330 pg, S = 208 wg; HL-60: P = 264 pg; S = 173 pg. bz LR-MEL P = 780 pg, S = 350 pg; MEL, P = 640 pg, S = 150 pg; COS: P = 740 lg, S = 250 pg; Rat-6 P = 100 pg. S = 40 pg; H-ros-Rat-6 P = 620 pg, S = 120 pg.  to the G25K pl peptide using Y-labeled secondary antibody to localize the bound IgG Uower pun&). Fluorographs were exposed for 7 days (soluble) or 5 days (particulate). Both of the immunoblots were exposed for 18 h. The QWOWS indicate [3H]MVA-labeled proteins that were specifically recognized by the antibodies to G25K. ZEF, isoelectric focusing.
Electrophoretic Mobility of G25K-Recent studies of ~21'~' (punel a), pretreatment of MEL cells for 24 h with 25 pM have shown that treatment of cultured cells with inhibitors lovastatin resulted in a decrease in the relative amount of of isoprenoid biosynthesis prevents the isoprenylation and immunodetectable G25K in the particulate fraction and an post-translational proteolytic processing of the protein at its increase in the relative amount of immunodetectable G25K C terminus (i.e. conversion of pro-p21"' to c-~21~") and the concentrated in the soluble fraction. The broadening of the subsequent palmitoylation and membrane localization of the band of immunoreactive protein in the soluble fraction of protein (conversion of c-p21r'* to m-p21ras) (18, 21). The MEL cells acutely exposed to lovastatin suggested that forms inhibition of p21ras processing is manifested by the accumu-of G25K with slightly altered electrophoretic mobilities might lation of pro-p21ras, which can be distinguished from c-~21~" have accumulated when MVA synthesis was blocked. Howand m-p21r"' by virtue of its higher apparent molecular mass ever, even when gradient gels were used, the resolution afon one-dimensional SDS gels (18,19,21). In view of these forded by one-dimensional SDS-PAGE was insufficient to findings, we conducted studies to determine what effect block-delineate clearly a pro-form of G25K analogous to pro-p21ras. ing MVA synthesis might have on the subcellular distribution To explore this possibility further, two-dimensional electroand electrophoretic mobility of G25K. As shown in Fig. 6 phoresis was used to resolve the isoforms of G25K (Fig. 6 punel b). Under these conditions, shifts in both the p1 values and the apparent molecular masses of the G25K isoforms were clearly discernible after lovastatin preincubation.

DISCUSSION
The present study demonstrates that in cultured MEL cells a protein that is immunologically indistinguishable from G25K undergoes post-translational modification by isoprenoid groups derived from [3H]MVA. G25K (GJ is a relatively abundant GTP-binding protein in platelets (34), brain (38), and placenta (37), and it was one of the first low molecular mass GTP-binding proteins to be purified from mammalian tissues. Although its function remains to be defined, the studies of Polakis et ul. (39) coupled with those in this report imply that G25K is a ubiquitous protein found in a variety of normal and transformed cell lines. G25K is customarily viewed as being a member of the rapidly expanding family of rus-related proteins. However, the available amino acid sequence information derived from four peptide fragments indicates that compared with other low molecular mass GTPbinding proteins, G25K has relatively little homology with p21r'* (39). Variation is observed even in the highly conserved Asn-Lys-X-Asp element (where X is an amino acid), which has been proposed to contribute to the guanine nucleotide binding site in the t-as-related proteins (52, 53). Therefore, it was not surprising that antibodies generated against the unique pl peptide sequence of G25K (39) showed no crossreactivity with proteins recognized by the pan-ro.s antibody in the present study.
As mentioned previously, reports that isoprenoid modification of p21raS, prelamin A, lamin B, and fungal mating factors occurs at a cysteine within a C-terminal Cys-A-A-X sequence have led to speculation that this sequence may represent a consensus motif for isoprenylation. Therefore, a major question is whether the C-terminal sequence of G25K is consistent with this pattern. The full-length cDNA sequence for G25K is not yet available. However, preliminary information indicates that there may be two different forms of G25K, one with a C-terminal sequence fitting the pattern Cys-A-A-X' and the other with a variation of this motif, i.e. Cys-Cys-A-X." Although the electrophoretic variants of G25K seen in the present study could have been due to posttranslational modifications of a single protein, it is also conceivable that two or more slightly different G25K gene products are expressed in MEL cells. If the latter is true, then our finding that all of the electrophoretic variants of G25K in MEL cells were isoprenylated (see Fig. 5) raises the possibility that proteins with a Cys-Cys-A-X motif, as well as the usual Cys-A-A-X, may undergo isoprenylation.
Preincubation of MEL cells with an inhibitor of isoprenoid synthesis resulted in changes in the isoelectric points and molecular masses of the G25K-immunoreactive proteins on two-dimensional gels (see Fig. 6). This observation provides some support for the notion that as in the case of p21raS, posttranslational processing of G25K is dependent on isoprenoid availability. However, it remains to be determined whether the shifts in electrophoretic mobility of G25K are due solely to accumulation of nonisoprenylated forms of the protein or whether blocking isoprenoid synthesis prevents additional steps such as proteolytic cleavage or carboxylmethylation at the C terminus.
In addition to establishing that G25K is isoprenylated, our results provide the first indication that G25K is normally localized in both the soluble and particulate fractions of several cultured cell lines. Of particular note was the observation that in MEL cells most of the G25K partitioned in the soluble fraction, despite the fact that the protein was isoprenylated. This finding argues against the idea that isoprenylation necessarily results in the stable insertion of proteins into the membrane lipid bilayer and points to a more complex and subtle role for this structural modification. The recent studies with p21raS have established that isoprenylation of CyslB6 is sufficient to promote interaction of the protein with the cell membrane and expression of transforming activity (18,19). However, the presence of an isoprenylated intermediate form of p21ras (c-p21rn*) in the soluble fraction of COS cells led Hancock et ul. (18) to conclude that although the first steps in p21raS processing (i.e. isoprenylation, proteolytic cleavage, and carboxylmethylation) allow some degree of membrane association, this association is not as strong or stable as that which is attained after the protein is palmitoylated. Additional insight into the function of isoprenylation comes from studies of the nuclear lamins. For example, although it is known that prelamin A is isoprenylated (15-17) and that mutations in the Cys-A-A-X motif abolish translocation of the protein to the nuclear envelope (54), forms of prelamin A with point mutations in the a-helical domain remain dispersed in the nucleoplasm despite the fact that they have a normal Cys-A-A-X C-terminal sequence (54). Based on these studies, it has been suggested that isoprenylation alone confers only a weak affinity for membranes and that other structural features or protein-protein interactions are also required for assembly of lamin A into the nuclear envelope (54). A similar situation may exist in the case of lamin B, in which isoprenylation at a C-terminal cysteine (11, 14-16) and interaction with a specific receptor (55) may both contribute to the stable asso-Isoprenoid Modification of G25K (GJ ciation of this protein with the nuclear envelope. Consistent with the foregoing concepts, our studies suggest that although isoprenylation has some influence on the localization of G25K to the particulate fraction, it is not the exclusive determinant of the proportion of G25K associated with the membrane or other organelles. One possibility is that isoprenylation of G25K constitutes a permissive event for subsequent post-translational modifications such as fatty acylation (56) or carboxylmethylation (57, 58), which confer additional hydrophobicity and thereby increase the affinity of the protein for the cell membrane.
This could explain why blocking isoprenoid synthesis with lovastatin resulted in a decreased concentration of G25K in the particulate fraction of MEL cells (see Fig. 6). Alternatively, weak membrane interactions promoted by isoprenylation may serve to bring G25K into proximity with a membrane receptor protein in which other factors (e.g. GTP binding and hydrolysis, phosphorylation) ultimately control the release of the protein into the soluble fraction. Several lines of evidence suggest that the latter possibility merits serious consideration. For instance, Nagata et ul. (59) found that a membrane-bound small GTPbinding protein in platelets could be phosphorylated by cyclic AMP-dependent protein kinase whereas a similar cytoplasmic GTP-binding protein was incapable of being phosphorylated by the same enzyme unless it was first treated with phosphatase. Hart et ul. (60) reported recently that purified platelet G25K undergoes epidermal growth factor stimulated tyrosine phosphorylation when reconstituted into phospholipid vesicles with purified epidermal growth factor receptor and speculated that phosphorylation might serve to stimulate release of the protein from membranes in uiuo. In this regard it is noteworthy that several of the 20-30-kDa isoprenylated proteins have been shown previously to be phosphorylated in murine lymphoma cells (9). It is not yet known if any of these proteins corresponded to G25K, but it is conceivable that phosphorylation could account for the multiple isoforms of G25K observed in the present study. lated proteins of MEL cells by sulfonium salt cleavage, we found material with chromatographic characteristics of farnesol, nerolidol, and additional unidentified rearrangement products (10). More recent studies of isoprenylated proteins from Chinese hamster ovary cells (12) and HeLa cells (13), employing gas chromatography-mass spectrometry to characterize products released by Raney-Nickel-catalyzed desulfurization, have demonstrated the existence of a second class of CZO (geranylgeranyl) isoprenoids. Thus, it is possible that C& isoprenoid alcohols might have accounted for some of the unidentified material released from MEL cell proteins with methyl iodide cleavage (10). In view of the apparent heterogeneity of the modifying isoprenoid groups, it will be necessary to determine whether or not specific proteins such as G25K can be modified interchangeably by farnesyl or geranylgeranyl groups, depending on substrate availability or prenyltransferase activity in a given tissue. An alternative possibility is that variations in the C-terminal motifs may dictate which proteins will act as substrates for particular prenyltransferases. Clarification of these issues should be forthcoming as additional isoprenylated proteins are identified and their modifying groups are characterized.