Rab Geranylgeranyl Transferase A MULTISUBUNIT ENZYME THAT PRENYLATES GTP-BINDING PROTEINS TERMINATING IN CYS-X-CYS OR CYS-CYS*

Rab proteins are membrane-bound prenylated GTP- binding proteins required for the targeted movement of membrane vesicles from one organelle to another. In the we purified an enzyme that attaches geranylgeranyl residues to Rab proteins that bear the COOH-terminal sequence Cys-X-Cys (such Rab3A) and Cys-Cys (such as RablA). This is designated Rab geranylger-any1 transferase (Rab GG transferase). At high salt concentrations, Rab GG transferase from into two components, designated A and B, both of which are required for activity. We purified Component B to apparent homogeneity and found that it contains two peptides of 60 and 38 kDa. The transferase to p21H-ras-CVLL , is prenylated GG transferase inhibited Zn2+, cation by in of cells at three protein

Rab proteins are membrane-bound prenylated GTPbinding proteins required for the targeted movement of membrane vesicles from one organelle to another. In the current paper we have characterized and purified an enzyme that attaches geranylgeranyl residues to Rab proteins that bear the COOH-terminal sequence Cys-X-Cys (such as Rab3A) and Cys-Cys (such as RablA). This enzyme is designated Rab geranylger-any1 transferase (Rab GG transferase). At high salt concentrations, Rab GG transferase from rat brain cytosol separates into two components, designated A and B, both of which are required for activity. We purified Component B to apparent homogeneity and found that it contains two peptides of 60 and 38 kDa. The purified Rab GG transferase did not attach geranylgeranyl to p21H-ras-CVLL , which is prenylated by a GG transferase of the CAAX type that resembles the CAAX farnesyltransferase. Rab GG transferase was strongly inhibited by Zn2+, a cation that is absolutely required by farnesyltransferase. The Rab GG transferase was also inhibited by NaCl concentrations in excess of 100 mM. Together with previous data, the current findings indicate that mammalian cells possess at least three protein prenyltransferases ( C A M farnesyltransferase, C A M GG transferase, and Rab GG transferase) that are specific for different classes of low molecular weight GTP-binding proteins and other proteins.
Rab proteins are small GTP-binding proteins with molecular masses of 21-25 kDa which are attached extrinsically to membranous organelles in animal and yeast cells (for review see Refs. 1 and 2). In yeast, mutations in one Rab protein, designated YPT1, block the movement of membrane vesicles within the Golgi apparatus. The mammalian counterpart of YPTl is called RablA (3). Another Rab protein, designated Rab3A, is attached to synaptic vesicles in neurons and dissociates reversibly upon calcium-triggered exocytosis (4). These findings have engendered the notion that Rab proteins play some role in the targeted movement of vesicles from one membranous structure to another (1, 2).
Adherence to membranes requires that Rab proteins possess prenyl groups attached by thioether bonds to cysteines Institutes of Health (HL 20948), the Lucille P. Markey Charitable * This research was supported by research grants from the National Trust, and the Perot Family Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. §Recipient of a graduate fellowship from Fundacao Calouste Gulbenkian of Portugal and a Fulbright Scholarship. near COOH termini (for review see Refs. 1 and 2). Most of the known Rab proteins terminate in either of two sequences: Cys-X-Cys (where X is alanine, serine, or glycine) or Cys-Cys. The synaptic vesicle protein Rab3A (also known as smgp25A) contains geranylgeranyl groups on each of the cysteines in its Cys-X-Cys sequence (5,6). RablA contains at least one geranylgeranyl residue attached to its Cys-Cys sequence (6). Prenylation of Rab proteins is the initial step required for membrane attachment (7). Additional protein modifications and binding to specific receptors may also be required (1,2).
In addition to Rab proteins, other small membrane-bound GTP-binding proteins, such as the ~2 1 "~ proteins, are prenylated, as are other extrinsic membrane proteins, such as nuclear lamins and the y-subunit of heterotrimeric G proteins (1, 2). Several protein prenyltransferases catalyze the attachment of either 20-carbon geranylgeranyl or 15-carbon farnesyl groups to cysteine residues in these proteins.
The best characterized prenyltransferase, protein farnesyltransferase, attaches farnesyl to cysteines at the fourth position from the COOH terminus of all known ~2 1 ' "~ proteins as well as several other proteins (8-10). This enzyme recognizes peptides as short as 4 amino acids in length that correspond to the CAAX box consensus sequence, where the A residues are usually (but not always) aliphatic and the COOH-terminal X residue is methionine, serine, glutamine, or cysteine (11). Peptides terminating in leucine are recognized poorly, if at all, by this enzyme (11)(12)(13). Protein farnesyltransferase has been purified to homogeneity from rat brain (8), and cDNAs for its two subunits, designated a and p, have been isolated (14-16). The &subunit contains Zn2+, which is necessary for binding the ~21'"" protein substrate (17). The function of the a-subunit is unknown, but it may participate in binding farnesyl pyrophosphate.
Other protein prenyltransferases attach geranylgeranyl groups to cysteine, and none has yet been purified to homogeneity. One of these enzymes, designated CAAX GG transferase' (12) or GG transferase-I (13), resembles the CAAX farnesyltransferase in prenylating peptides in which cysteine is fourth from the COOH terminus. This GG transferase prefers CAAX sequences that terminate in leucine (12,13,18,19). Substrates include the y-subunit of heterotrimeric G proteins from brain (COOH-terminal sequence CAIL) (20, 21) and raplA/Krevl, a low molecular weight GTP-binding protein whose COOH-terminal sequence is CVLL (12,19,22). Remarkably, this enzyme contains an a-subunit that appears identical to the a-subunit of the farnesyltransferase (12, 13).
14497 Moores et al. (13) recently demonstrated a second GG transferase that attaches geranylgeranyl to Rab proteins that terminate in Cys-Cys. One known substrate is YPTl from Saccharomyces cereuisiae. Other potential substrates that terminate in Cys-Cys include RablA,RablB,and Rab2 (1,2,6). This enzyme differs from the two CAAX prenyltransferases in that it does not recognize short peptides (13). Horiuchi et al. (23) demonstrated a cytosolic GG transferase in bovine brain that prenylates Rab3A, which terminates in Cys-Ala-Cys. The enzyme was detected in crude extracts, but it rapidly lost activity during purification, and no molecular characterization was reported. It is not yet clear whether this GG transferase, which prenylates Cys-X-Cys sequences, is different from the enzyme that recognizes Cys-Cys sequences.
In the current experiments we have purified a Rab GG transferase from rat brain. We found that the loss of activity during purification is attributable to the separation of two components, designated A and B, both of which are required for activity. We have purified the B component (consisting of two peptides of 60 and 38 kDa) to apparent homogeneity and have found that it participates in the attachment of geranylgeranyl to both classes of Rab proteins, namely those that terminate in Cys-X-Cys (Rab3A) and Cys-Cys (RablA). We have designated this enzyme Rab GG transferase.

Expression and Purification of Recombinant Protein Substrates
Recombinant bovine Rab3A was expressed in Escherichia coli and purified as described previously (24). After purification, the column fractions were subjected to electrophoresis on a 12.5% SDS-polyacrylamide gel, stained with Coomassie Brilliant Blue, estimated to be more than 90% pure, divided into multiple aliquots, and stored at -70 "C. Recombinant canine RablA cDNA was kindly provided by Dr. Channing Der (La Jolla Cancer Research Foundation, La Jolla, CA) (Ref. 6). The plasmid was introduced into BL21DE3 E. coli strain, and the expressed protein was purified essentially as described for p21H-"s (12) except that only the DEAE-chromatography step was performed. RablA was estimated to be more than 90% pure by SDSgel electrophoresis, divided into multiple aliquots, and stored at -70 "C. Recombinant human p21H""" and p21H"as~cvLL and purified as described previously (12).
were expressed Assay for Rab GG Transferase Activity Rab GG transferase activity was determined by measuring the amount of ['H]geranylgeranyl transferred from [3H]GGPP to Rab3A protein. The standard reaction mixture contained the following concentrations of components in a final volume of 50 p1: 50 mM sodium Hepes (pH 7.2), 5 mM MgCI2, 0.3 mM Nonidet P-40,25 p~ EDTA, 1 mM DTT, 0.54 p~ recombinant Rab3A, and 0.5 p~ ['HIGGPP (17,600 dpm/pmol). After incubation for 10 min at 37 "C, the reaction mixtures were processed by trichloroacetic acid precipitation followed by filtration on glass-fiber filters as previously described for assay of protein farnesyltransferase activity (8).

Assays for Protein Farnesyltransferase and CAAX GG Transferase Activities
These enzymatic activities were determined as described previously (8, 12).
Purification of Component B of Rab GG Transferase All steps were carried out at 4 "C. Steps 3-6 were performed with a fast protein liquid chromotography system (Pharmacia LKB Biotechnology Inc.).
Steps 1 and 2: Homogenization and Ammonium Sulfate Fractionation-Brains from 50 male Sprague-Dawley rats (100-150 g) were homogenized as described previously for the purification of protein farnesyltransferase (8). The brain homogenate was centrifuged at 60,000 X g for 70 min. The supernatant fraction was treated with ammonium sulfate, and the material precipitating between 30 and 50% ammonium sulfate was dissolved and dialyzed overnight against buffer containing 20 mM Tris-chloride (pH 7.5), 1 mM DTT, and 20 p M ZnC12 (Buffer 1).
step 3: Zon Exchange Chromatography-Approximately one-third of the dialyzed 30-50% ammonium sulfate fraction (-130 mg of protein) was chromatographed on a Mono Q 10/10 column (Pharmacia) as described in the legend to Fig. 1. The active fractions from three consecutive Mono Q runs were pooled (24 ml) and dialyzed overnight against 6 liters of buffer containing 20 mM sodium Hepes (pH 7.21, 10 mM NaCl, 0.1 mM Nonidet P-40, and 1 mM DTT (Buffer 2).
Step 4: Gel Filtration in IO mM NaC1-The dialyzed Mono Q fraction was concentrated approximately 4-fold in CF25 Centriflo cones (Amicon). One-half of the fraction (-3 ml) was loaded onto a Superdex 200 26/60 column (Pharmacia) that had been equilibrated in Buffer 2. The flow rate was 2 ml/min, and the material eluting between 100 and 200 ml was collected in 4-ml fractions (Fractions 1-25 in Fig. 2). The gel filtration step was then repeated with the other half of the Mono Q fraction.
Step 5: Hydrophobic Interaction Chromatography-The active fractions from both Superdex gel filtration runs in Step 4 were pooled (48 ml) and concentrated approximately &fold in CF25 Centriflo cones. The concentration of ammonium sulfate was adjusted to 0.6 M by the addition of 0.79 g of solid ammonium sulfate, and the material was loaded onto a phenyl-Superose 5/5 column (Pharmacia) that had been equilibrated in buffer containing 50 mM sodium Hepes (pH 7.2), 0.6 M (NH4)*SO4, 0.1 mM Nonidet P-40, and 1 mM DTT (Buffer 3). Chromatography was performed as described in the legend to Fig. 4 with buffer 4 (10 mM sodium Hepes (pH 7.2), 0.1 mM Nonidet P-40, and 1 mM DTT) as the counterbuffer.
Step 6: Gel Filtration in 0.5 M NaCl-The active pool from the phenyl-Superose column (12 ml) was concentrated about 12-fold in Centricon 30 microconcentrators (Amicon) and loaded onto a Superdex 200 16/60 column (Pharmacia) that had been equilibrated in 20 mM sodium Hepes (pH 7.2), 0.5 M NaCI, 0.1 mM Nonidet P-40, and 1 mM DTT (Buffer 5). The flow rate was 0.5 ml/min, and the material eluting between 50 and 90 ml was collected in 1-ml fractions (fractions 1-34 in Fig. 5). The fractions containing Rab GG transferase activity were pooled (usual final volume of 4-5 ml). This material, which is hereafter called purified Component B, was divided into multiple aliquots and stored at -70 "C. The protein content of this material was measured by Method B (see below).
Partial Purification of Component A of Rab GG Transferase Rat brain extract was processed through Steps 1-4 as described above. The sample eluted from Step 4 was adjusted to 0.25 M NaCl, and 48 ml was applied to a 5-ml Red A Dyematrex column (Amicon) that had been equilibrated at 4 "C with buffer containing 20 mM sodium Hepes (pH 7.2), 0.1 mM Nonidet P-40, 1 mM DTT, and 0.25 M NaCl. The flow-through was collected and reloaded onto the column twice. After the last passage, the flow was stopped for 15 min, after which the column was washed with 25 ml of the same buffer. Rab GG transferase Component A was eluted with 30 ml of 2 M NaCl in the same buffer and dialyzed overnight against 6 liters of Buffer 2. The dialyzed fraction was concentrated 10-fold on CF25 Centriflo cones, divided into multiple aliquots, and stored at -70 "C. This material is hereafter called partially purified Component A.
Other Methods SDS-polyacrylamide gel electrophoresis was carried out as described by Laemmli (25), and the gels were calibrated with low range SDS-polyacrylamide gel electrophoresis standards (Bio-Rad). Gel filtration columns were calibrated with a gel filtration calibration kit (Pharmacia) for the Superdex 200 26/60 column and with a gel filtration standard (Bio-Rad) for the Superdex 200 16/60 column. The protein concentration of all samples (except purified Component B) was determined with the Bio-Rad protein assay, a modification of the method of Bradford (26). The protein content of purified Component B was estimated by Coomassie Blue R-250 staining and densitometric scanning of a 10% SDS-polyacrylamide gel in which known amounts (0.5-2 pg) of low molecular weight range Bio-Rad SDS-polyacrylamide gel electrophoresis standard proteins were used as a reference. This procedure is designated protein determination Method B.

RESULTS
For the first step in the purification of Rab GG transferase, we subjected a 30-50% ammonium sulfate fraction of rat brain cytosol to ion exchange chromatography on a Mono Q column. Fractions were assayed for Rab GG transferase activity by trichloroacetic acid precipitation following incubation with [3H]GGPP and recombinant Rab3A isolated from E. coli.
We also assayed the fractions of CAAX GG transferase activity with a chimeric ~2 1~" " " substrate containing CVLL at the COOH terminus and for farnesyltransferase activity with p21H""". Fig. 1 shows that the two GG transferases were clearly separated on the column, with the farnesyltransferase eluting between them.
The Mono Q fractions containing Rab GG transferase activity were pooled and applied to a Superdex 200 gel filtration column in low ionic strength buffer (10 mM NaC1). Fig.  2 shows that under these conditions Rab GG transferase activity eluted at an apparent molecular weight of -330,000, preceding the farnesyltransferase which eluted at an apparent molecular weight of -230,000.
When we attempted to purify the Rab GG transferase further by hydrophobic chromatography, the activity was resolved into two components that required mixing to restore activity. In one such experiment, shown in Fig. 3, fractions containing Rab GG transferase activity from a Mono Q column similar to the one in Fig. 1 were applied to a phenyl-Superose column in 0.5 M NH4S04 and eluted stepwise with decreasing concentrations of NH,S04. Although the starting material contained abundant Rab GG transferase activity, virtually no activity was found in any of the five column fractions (panel designated none in Fig. 3). We then performed a systematic mixing experiment in which aliquots of each fraction were added to assays employing each of the in a volume of 10 ml was applied to a Mono Q 10/10 column equilibrated in Buffer 1. The column was run at a flow rate of 1 ml/ min, and 4-ml fractions were collected. The column was washed with 24 ml of Buffer 1 (Fractions 3-9), followed by a 18-ml linear gradient from 0 to 0.15 M NaCl (Fractions 9-14), followed by a second wash with 18 ml of the same buffer containing 0.15 M NaCl (Fractions 14-19). The enzymes were then eluted with a 40-ml linear gradient of 0.15-0.45 M NaCl in the same buffer (Fractions 19-28), after which a step to 1 M NaCl was performed followed by a 20-ml wash at the same concentration of NaCl (Fractions 28-32). An aliquot of each fraction was assayed for Rab GG transferase activity (2 pl) (O), for farnesyltransferase activity (5 pl) (01, and for CAAX GG transferase activity (5 pl) (0). The units of activity represent the total nmol of prenyl groups transferred per 10 min assay calculated for the entire fraction. The dashed lines (---) denote the protein content of each fraction. Step 3) in 10 ml of buffer containing 50 mM sodium Hepes (pH 7.2), 0.5 M ammonium sulfate, 1 mM DTT, and 0.1 mM Nonidet-P40 was loaded onto a phenyl-Superose 5/5 column (Pharmacia) preequilibrated in the same buffer containing 0.5 M ammonium sulfate. The column was run a t a flow rate of 1 ml/min, and the flow-through was collected (Fraction 1). A step to 0.38 M ammonium sulfate was performed with a buffer containing 10 mM sodium Hepes (pH 7.2), 1 mM DTT, and 0.1 mM Nonidet-P40, and Fraction 2 was collected. Three other consecutive steps to 0.28 M ammonium sulfate (Fraction 3), 0.05 M ammonium sulfate (Fraction 4 , and no ammonium sulfate (Fraction 5) were performed. The volume of each fraction (Fractions 1-5) was 10 ml. These fractions and an aliquot (0.5 ml) of the starting material ( S M ) were dialyzed overnight at 4 "C against Buffer 2. An aliquot (10 pl) of each of these dialyzed fractions was then used to measure Rab GG transferase activity in the standard assay incubated for 30 min at 37 "C in the absence or presence of an aliquot (10 pl) of the indicated added fraction. other fractions. We observed a peak of activity that appeared to be maximal when Fractions 2 and 3 were mixed (Fig. 3). The earlier eluting material from the hydrophobic column was designated Component A, and the later eluting material was designated as Component B of the Rab GG transferase.
To follow the activity of Component B during purification, we required a preparation of Component A which was free of Component B so that it could be used to supplement the assays. We found that Component A could be separated quantitatively from Component B by chromatography on a Red A Dyematrex column. Accordingly, we carried a preparation of Rab GG transferase through the Mono Q column and low salt Superdex chromatography steps, which failed to resolve the two components. We then applied the mixture to a Red A Dyematrex column in 0.25 M NaCl. After washing, Component A was eluted with 2 M NaCI. This Component A material had minimal Rab GG transferase activity on its own, but it did restore the activity of Component B (see below).
To continue the purification of Component B, the fractions from the low salt Superdex 200 column of Fig. 2 which contained Rab GG transferase activity were pooled and applied to a phenyl-Superose column in 0.6 M (NH4)&04, which was eluted with a linear gradient containing decreasing ammonium sulfate concentrations (Fig. 4). Fractions were assayed for Rab GG transferase activity either alone (open circles) or in the presence of added Component A (closed circles). In the absence of Component A, the fractions had no transferase activity. In the presence of Component A, we observed a peak of activity emerging at the midpoint of the gradient.
The next step took advantage of the fact that Component B had been separated from Component A, and therefore Component B should have a lower apparent molecular weight on gel filtration. Accordingly, the fractions from the phenyl-Superose column with Component B activity were pooled and reapplied to a Superdex 200 gel filtration column, this time in the presence of high salt (0.5 M NaC1). The peak of protein eluted ahead of the 158-kDa marker (Fig. 5). The activity of Rab GG transferase Component B, assayed in the presence of added Component A, was well separated from the protein peak, eluting at a position corresponding roughly to an apparent molecular weight of 90,000. Table I summarizes Fig. 2 were pooled and processed as described under "Experimental Procedures" and applied in a volume of 10 ml (0.6 mg of protein) to a phenyl-Superose 5/5 column equilibrated in Buffer 3 containing 0.6 M ammonium sulfate (Fractions 1-4). The column was run at a flow rate of 0.4 ml/min, and 3-ml fractions were collected. The column was washed with 11 ml of Buffer 3 (Fractions 5-7), after which a 20ml linear gradient was started using Buffer 4 from 0 to 60% (0.6-0.24 M ammonium sulfate) (Fractions 8-15). A step to 100% Buffer 4 (no ammonium sulfate) was then performed followed by a 10-ml wash with the same buffer (Fractions 16-18). An aliquot of each fraction (1 p l ) was assayed for Rab GG transferase in the presence (0) or absence (0) of 2.6 pg of partially purified Component A (see "Experimental Procedures"). The units of activity represent the total nmol of prenyl groups transferred per 10-min assay calculated for the entire fraction. A blank value determined in assays containing Component A in the absence of any column fraction (1.35 nmol/fraction) was subtracted from the plotted data. The blank value in assays containing column fractions in the absence of Component A was negligible (0.072 nmol/fraction). The dashed lines (---) denote the protein content of each fraction. excess Component A. The results suggested that Component B had been purified approximately 17,000-fold relative to its activity in crude cytosol, with an overall yield of approximately 17%. With this procedure, we were able to isolate -16 pg of purified protein from 1.6 g of rat brain cytosol protein.
These figures should be considered as gross approximations since the assays for Component B activity in crude cytosol are not accurate and since the protein concentration of the final purified material was not measured chemically but was estimated from the intensity of the bands observed on Coomassie-stained SDS-polyacrylamide electrophoresis gels (see below).
SDS-polyacrylamide gel electrophoresis of the fractions that emerged from the final high salt Superdex 200 column revealed two peptides with molecular masses of 60 and 38 kDa that co-eluted with the Rab GG transferase Component B activity (Fig. 6, Panel A ) . When the pooled active fractions from this Superdex 200 column were subjected to further analytical chromatography on Mono Q, both peptides again eluted coincident with Component B activity, and no other bands were visible in the peak fraction (Fig. 6, Panel B ) . In multiple preparations of enzyme purified by the above procedure, we always observed both the 60-and 38-kDa proteins in an approximate 1:l ratio as indicated by Coomassie Blue staining of the gels. Moreover, the molecular weight of the final Component B preparation, as estimated by gel filtration in high salt, was -90,000 (Fig. 5). We therefore believe that Component B may be a 1:l heterodimer of 60-and 38-kDa subunits. Fig. 7 shows that purified Component B required Component A for Rab GG transferase activity. The partially purified preparation of Component A had slight Rab GG transferase activity on its own, but this was stimulated 20-fold with the addition of 64 ng of Component B (Fig. 7A). Similarly, Component B had little detectable activity unless supplemented with Component A (Fig. 7B). The two components appeared to saturate each other when present in a ratio of 64 ng of Component B to 2.6 pg of Component A. The inequality of this ratio reflects the impurity of the Component A preparation.
We used the mixture of purified Component B and partially purified Component A to characterize the enzyme reaction   Fig. 5 were subjected to electrophoresis (30 mA, 45 min, 24 "C) on a 10% SDS-polyacrylamide minigel, and the protein bands were detected by silver staining. Panel B, the pooled active fraction obtained after Superdex 200 chromatography in 0.5 M NaCl (-16 pg of protein in a volume of 5 ml) was diluted 4-fold to adjust the NaCl concentration to 0.125 M and then applied to a Mono Q 5/5 column (Pharmacia) equilibrated in buffer containing 20 mM sodium Hepes (pH 7.2), 0.1 mM Nonidet P-40, and 1 mM DTT (Fractions 1-10). The column was run at a flow rate of 1 ml/min, and 2-ml fractions were collected. The column was washed with 5 ml of the same buffer (Fractions 11-13), followed by 5 ml of the same buffer containing 0.23 M NaCl (Fractions 13-15). Component B of Rab GG transferase was eluted with a 24-ml gradient from 0.23 to 0.35 M NaCl in the same buffer (Fractions 16-27). An aliquot (2 pl) of the fractions eluting between approximately 0.28 and 0.30 M NaCl was assayed (Fractions 21-24), and another aliquot (25 p1) of the same fractions was subjected to electrophoresis (30 mA, 45 min, 24 "C) on a 10% SDS-gel electrophoresis minigel, and the protein bands were detected by silver staining. For each fraction, the activity of Rab GG transferase Component B is shown at the top. The gels were calibrated with the indicated protein molecular weight standards. Arrows denote the presence of the 60-and 38-kDa proteins that correlate with the activity of Component B. Step 3) and the standard reaction components in the presence (+) or absence (-) of 5 mM MgCl,. The gel was treated with Entensify solution (Du Pont), dried, and exposed to XAR film for 2 days at -70 "C. The migration of "C-methylated markers loaded onto an adjacent lane is shown. kinetically (Fig. 8). The concentrations of GGPP (Panel A ) and Rab3A (Panel B ) giving half-maximal reaction velocity were both approximately 0.2 pM. The ['HJgeranylgeranyl from ['HIGGPP was indeed transferred to the Rab3A protein, as verified by SDS-polyacrylamide gel electrophoresis and autoradiography (inset to Fig. 8). The transfer of ['Hlgeranylgeranyl from ["HIGGPP was competitively inhibited by unlabeled GGPP, with 50% inhibition occurring at -0.2 pM (Fig.  9). Farnesyl pyrophosphate at concentrations as high as 10 p~ failed to inhibit (Fig. 9).
The purified Rab GG transferase showed an absolute requirement for MgC12 with a half-maximal concentration of about 0.5 mM (Fig. 1OA). In the presence of saturating concentrations of MgCI2, the addition of ZnCln was inhibitory with 100% inhibition achieved at 0.2 mM (Fig. 1023). The enzyme activity was very sensitive to the salt concentration in the assay (Fig. 11). The activity was inhibited by 50% at 100 mM concentrations of NaC1, NaOAc, or KCl, and was abolished at 300 mM concentrations of these salts.
In addition to RabsA, which terminates in Cys-Ala-Cys, the purified Rab GG transferase prenylated RablA, which terminates in Cys-Cys (Fig. 12). The concentrations of the two protein substrates giving half-maximal reaction velocities Incubations were carried out in duplicate for 10 min at 37 "C, and trichloroacetic acid-precipitable radioactivity was measured as described under "Experimental Procedures." A plus the indicated amounts of purified Component B. For both panels, duplicate samples were incubated for 10 min at 37 "C, and trichloroacetic acid-precipitable radioactivity was measured as described under "Experimental Procedures." were similar, but the maximal velocity was about 2-fold higher for the Rab3A substrate. The Rab GG transferase did not attach geranylgeranyl residues to a mutant p21H""s that contained CVLL at the COOH terminus despite the fact that this protein is a good substrate for the CAAX GG transferase ( Fig.   1 and Refs. 12 and 19).
To confirm that the prenylation of RablA required Component A as well as Component B, we compared the effects of the two components of the prenylation of the two Rab proteins in the same experiment (Fig. 13). Prenylation of both proteins required Component A and B, suggesting that a single enzyme prenylates both Rab proteins.

DISCUSSION
The current results indicate that Rab GG transferase is a complex enzyme with at least two components, designated A and B. Purified preparations of Component B contain two peptides of approximately 38 and 60 kDa. We have been unable to resolve these two peptides with retention of Component B activity, and we therefore believe provisionally that both are subunits of this component. Component A was purified only partially, and its subunit composition is unknown.
Component B has an apparent molecular weight of approximately 90,000 on gel filtration in high salt (Fig. 5), raising the possibility that it is a 1:l heterodimer of the 60-and 38-kDa subunits. When the intact Rab GG transferase was subjected to gel filtration on low salt prior to resolution of the two components, all of the activity emerged at a position corresponding to a molecular weight of -330,000 (Fig. 2). The peak fractions possessed both Components A and B. Assay of other column fractions in the presence of added Components A or B failed to generate additional activity, suggesting that the 330-kDa peak fraction contained essentially all of the Components A and B that were present in the crude extract. These data suggest that the native GG transferase is a macromolecular complex of about 330 kDa that contains Components A and B.
The distinction between Components A and B was first appreciated when we lost nearly all Rab GG transferase activity during hydrophobic chromatography and were able to restore activity by mixing fractions from the column (Fig.  3). Subsequent experiments showed that the separation of the two components was not the result of exposure to the hydrophobic resin but rather that it occurred whenever the salt concentration was raised above about 200 mM. We took advantage of this observation to separate the two components by absorption to a Red A Dyematrex column followed by elution with 2 M NaCl, which eluted Component A essentially free of Component B.
The activity of the purified enzyme depended on Component A as well as Component B. Moreover, the activity was sensitive to the ionic strength of the assay mixture, with 50% inhibition at a salt concentration of 100 mM. This is close to the salt concentration necessary to separate Components A and B and raises the possibility that the two components must be physically associated for geranylgeranyl transfer to occur.
We were surprised to find that the purified Rab GG transferase used RablA as well as Rab3A as an acceptor of geranylgeranyl groups. Transfer of geranylgeranyl to RablA required Component A as well as Component B (Fig. 13), and the reaction was inhibited by Zn2+ and sodium chloride in the same fashion as was transfer to Rab3A (data not shown). Inasmuch as our preparation of Component B appears to be pure, these data strongly suggest that a single Component B participates in the prenylation of both proteins. In view of the impurity of the Component A preparations, we cannot be certain that the same Component A participates in both reactions. However, the similarity in the requirement for Component A and the similarity in the required reaction conditions suggest that a single Component A is involved in both reactions. For this reason, we have designated the enzyme as Rab GG transferase.
The mechanism by which the Rab GG transferase recognizes two different COOH-terminal sequences is not yet known. In experiments not shown, we found that the enzyme was not inhibited by peptides as long as 13 residues corresponding to the COOH-terminal sequence of Rab3A. Thus, the enzyme must recognize additional sequences that are remote from the COOH terminus. It is possible that the enzyme recognizes sequences that are similar in the RablA and Rab3A proteins and may be common to all members of the Rab family. In this respect the Rab GG transferase differs markedly from the two CAAX prenyltransferases whose sub-strate recognition depends on the COOH-terminal 4 amino acids. Indeed, the Rab GG transferase did not transfer geranylgeranyl residues to a mutant p21H"'" in which the COOHterminal sequence was changed to CVLL (Fig. 12). The latter protein is a good substrate for the CAAX GG transferase (12,19).
In vivo Rab3A has been shown to contain geranylgeranyl groups on both cysteines of the Cys-X-Cys sequence ( 5 ) . We do not know whether the enzyme can prenylate both cysteines in the COOH-terminal Cys-Cys sequence of RablA. The experiments of Figs. 12 and 13 suggest that the enzyme incorporated twice as much [3H]geranylgeranyl into Rab3A as into RablA. However, the kinetics of the geranylgeranyl transfer to both substrates appears to be complex, and more studies with mutant Rab3A and RablA substrates that contain only one of the 2 cysteine residues will be necessary to define the stoichiometry of the reaction.