Mammalian Prenylcysteine Carboxyl Methyltransferase Is in the Endoplasmic Reticulum*

Prenylcysteine carboxyl methyltransferase (pcCMT) is the third of three enzymes that posttranslationally modify C-terminal CAAX motifs and thereby target CAAXproteins to the plasma membrane. Here we report the molecular characterization and subcellular localization of the first mammalian (human myeloid) pcCMT. The deduced amino acid sequence of mammalian pcCMT predicts a multiple membrane-spanning protein with homologies to the yeast pcCMT, STE14, and the mammalian band 3 anion transporter. The human gene complemented a ste14 mutant. pcCMT mRNAs were ubiquitously expressed in human tissues. An anti-pcCMT antiserum detected a 33-kDa protein in myeloid cell membranes. Ectopically expressed recombinant pcCMT had enzymatic activity identical to that observed in neutrophil membranes. Mammalian pcCMT was not expressed at the plasma membrane but rather restricted to the endoplasmic reticulum. Thus, the final enzyme in the sequence that modifies CAAX motifs is located in membranes topologically removed from the CAAX protein target membrane.

A number of signaling molecules, including Ras and G proteins, are targeted to the inner leaflet of the plasma membrane by a sequence of posttranslational modifications of a C-terminal CAAX 1 motif that include prenylation, proteolysis, and carboxyl methylation (1). In some cases palmitoylation of an upstream cysteine is also required (2). These modifications render otherwise hydrophilic proteins hydrophobic, promoting association with membranes. The relative contributions of prenylation, proteolysis, and carboxyl methylation to membrane targeting are not well understood. Whereas neutralization of the negative charge on the ␣-carboxyl group by methyl esterification adds to overall hydrophobicity, particularly for farnesylated proteins, this modification contributes little to the affinity of geranylgeranylated proteins for membranes (3). Although processed CAAX proteins can associate with phospholipid vesicles in vitro (4), it is not known whether membrane proteins participate in prenylcysteine membrane association in vivo. The Saccharomyces cerevisiae mating pheromone, a-factor, is a CAAX-processed polypeptide, and both its secretion via the Ste6p transporter (5) and its engagement of the Ste3p G protein-linked receptor (6) are dependent on prenylcysteine carboxyl methylation, suggesting a role for this modification in protein-protein interactions. A cycle of prenylcysteine carboxyl methylation is associated with neutrophil activation (7), and inhibitors of this enzyme block signal transduction in neutrophils (7), macrophages (8), and platelets (9), suggesting that, like bacterial chemotaxis (10), some eukaryotic processes may be regulated by reversible carboxyl methylation.
Because prenylcysteine carboxyl methylation cannot be abolished by mutation of the substrate without also eliminating prenylation, elucidation of the role of carboxyl methylation will require characterization and disruption of the methyltransferase. Until recently, the only prenylcysteine carboxyl methyltransferase (pcCMT) characterized at the molecular level was the STE14 gene product of S. cerevisiae (5). Homologs in Schizosaccharomyces pombe and Xenopus laevis have now been reported (11). Here we report the molecular cloning and preliminary characterization of the first mammalian pcCMT. In addition, using pcCMT tagged with green fluorescent protein (GFP) we show that pcCMT is expressed in the endoplasmic reticulum (ER) but excluded from the plasma membrane that is the target of many CAAX proteins, including Ras.

EXPERIMENTAL PROCEDURES
Molecular Cloning-A text-based search of the expressed sequence tag (EST) data base of the National Center for Biotechnology Information identified a 426-bp partial cDNA from murine placenta (mh77d06.r1) that has amino acid homology to the STE14 gene product of S. cerevisiae. Primers based on this sequence (forward: 5Ј-GCCGG-ACTCAACGCGCTGCTGCTGCTACTCTA-3Ј; reverse: 5Ј-CGTGTACT-CCAGGCTGTGATTCAGGAGGAA-3Ј) were used to amplify by reverse transcriptase-PCR a 218-bp homologous cDNA fragment from HL60 cells. This fragment was then 32 P-labeled and used to screen a unidirectional, size-fractionated HL60 cDNA library constructed in the ZAP system (provided by Dr. Philip Murphy, NIAID). Hybridizations were performed at 56°C overnight. Of ϳ10 5 recombinant plaques screened, a single doubly positive clone was identified and proved positive on secondary screen. This clone contained a 3.6-kb cDNA insert that was sequenced in both directions.
Northern Blotting-Total RNA (40 g) from HL60 cells grown with or without 1.25% Me 2 SO for 5 days was fractionated on a 1.2% denaturing formaldehyde-agarose gel, transferred to a nylon filter, and hybridized with [ 32 P]dATP-labeled full-length HL60 pcCMT cDNA in Express Hybridization solution (CLONTECH, Palo Alto, CA) at 68°C. A human multiple-tissue Northern blot (CLONTECH) was hybridized with the same probe according to the manufacturer's instructions. The same filters were stripped and rehybridized with a ␤-actin probe.
Antiserum-A peptide corresponding to deduced amino acids 185-201 of HL60 pcCMT (FNHVVQNEKSDTHTLV) was synthesized, linked to keyhole limpet hemocyanin, and used to immunize rabbits. The resulting antiserum was used for immunoblotting of neutrophil light membranes prepared by nitrogen cavitation and discontinuous sucrose density sedimentation (12), for immunoprecipitation of lysates of HL60 cells metabolically labeled with L-[ 35 S]methionine, and for indirect immunofluorescence of COS-1 cells.
Expression of Recombinant pcCMT-The 3.6-kb full-length HL60 pcCMT cDNA and an 852-bp cDNA fragment comprising the ORF were subcloned into the eukaryotic expression vectors pcDM8 and pcDNA3.1 (Invitrogen, Carlsbad, CA). Myc-tagged pcCMT was made by adding a sequence encoding the 10-amino acid epitope of c-Myc recognized by monoclonal antibody 9E10 to the C terminus of pcCMT using patch PCR and subcloning into pcDNA3.1. pcCMT tagged at the C terminus with GFP was generated by subcloning the pcCMT ORF into the expression vector pGFP-N3 (CLONTECH). For pcCMT enzymatic assays, COS-1 cells were grown in 10-cm dishes and transfected with pcCMT with DEAE-dextran as described (13). For fluorescence microscopy, exponentially growing COS-1, CHO, and NIH3T3 cells were plated the day prior to transfection at 50 -60% confluence and transfected with 2 g of DNA using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. Cells were observed at 24 and 48 h after transfection.
Prenylcysteine Carboxyl Methyltransferase Assays-COS-1 cells transfected with HL60 pcCMT or vector alone were harvested at 48 h with 5 mM EDTA and subjected, in parallel with human neutrophils, to nitrogen cavitation and discontinuous sucrose density centrifugation as described (12). Light membranes were assayed for pcCMT activity qualitatively by S-adenosyl-L-[methyl-3 H]methionine labeling (SDSpolyacrylamide gel electrophoresis and fluorography) of Rho GTPases partially purified from neutrophil cytosol and quantitatively by car-boxyl methylation of N-acetyl-S-trans,trans-farnesyl-L-cysteine (AFC) (heptane partition, alkaline hydrolysis, and measurement of vapor phase [ 3 H]methanol) as described (14).
S. cerevisiae Strains, Growth, and ste14 Complementation-Complete (YEPD), synthetic (SD), and synthetic dropout media were prepared as described previously (15), except that dropout medium lacked cysteine. All experiments were performed at 30°C. Yeast transformations were carried out by either the lithium acetate method (16) or the Elble method (17). The ste14 deletion strains used in this study are SM1188 (MATa ⌬ste14 -3::TRP1 trp1 leu2 ura3 his4 can1) and SM1639 (MATa ⌬ste14 -4::URA3 trp1 leu2 his4 can1) (5). Plasmid expression of Ste14p was accomplished by pSM1237 (CEN URA3 STE14). Human pcCMT was expressed in yeast by subcloning the HL60 pcCMT ORF into the BamHI/SalI sites of the yeast expression plasmid pG1 (2 TRP1) that utilizes a GPD promotor and a PGK transcription terminator and polyadenylation signal (18). The resulting construct was designated pG1-hCMT (2 TRP1 hCMT). Patch mating tests were carried out essentially as described previously (15). Briefly, patches of MATa cells grown on selective media were replica plated onto a lawn of the MAT␣ mating tester, SM1068 (lys1), that had been spread on an SD plate. Plates were incubated at 30°C for 3 days. Growth of the prototrophic diploids indicated mating.
Fluorescence Microscopy-Live cells that had been transfected with pcCMT-GFP and fixed/permeabilized (2% paraformaldeyde/0.2% Triton X-100 at 4°C or methanol at Ϫ20°C) cells stained with 9E10 anti-Myc antibody (Santa Cruz Biotechnology, Santa Cruz, CA), C6 anti-ribophorin I antiserum (provided by Dr. Gert Kriebich, NYU), or the anti-pcCMT antiserum described above, followed by Texas Red-conjugated secondary antisera, were imaged with a Zeiss Axioscope equipped with a Princeton Instruments cooled CCD camera and a KAF 1400 chip or a Molecular Dynamics confocal microscope equipped with an argon laser.

RESULTS AND DISCUSSION
Repeated efforts to characterize mammalian pcCMT by STE14 homology cloning and by enzyme purification and microsequencing were unsuccessful. However, we identified a partial murine cDNA (GenBank accession no. AA022288) in FIG. 1. Deduced amino acid sequence of human pcCMT and homology to related genes. S. cerevisiae STE14 (5) and S. pombe and X. laevis mam4 (11) have prenylcysteine carboxyl methyltransferase activity in vitro. Murine EST mh77d06.r1 and two C. elegans genes were identified by searching the NCBI GenBank. Amino acid identities between human pcCMT and other gene products are indicated by shading. Hydrophobic stretches of human pcCMT that may represent membrane-spanning domains are indicated with solid bars. A region of 36% identity between human pcCMT (amino acids 1-66) and human band 3 anion transporter (amino acids 750 -821) is indicated with a dashed bar. An internal peptide used to raise an anti-pcCMT antiserum is indicated by a dotted line. A potential N-glycosylation site is indicated with an arrowhead.
the EST data base of the National Center for Biotechnology Information that has sequence homology to Ste14p. Using primers based on this sequence, reverse transcription-PCR was employed to amplify from human myeloid HL60 cells a 218-bp homologous sequence. Using this fragment an HL60 cDNA library was screened and a single clone identified that contained a 3.6-kb insert. Sequencing revealed a 3595-bp cDNA (GenBank accession no. AF064084) with an ORF that predicted a 284-amino acid protein that is 26% identical to Ste14p (Fig.  1). Hydropathy analysis revealed six hydrophobic sequences that may represent membrane-spanning domains (Fig. 1). This is consistent with the five or six putative membrane-spanning domains of Ste14p (5), with recent data demonstrating that Ste14p is an integral membrane protein 2 and with our observation that active pcCMT cannot be extracted from membranes with detergents but can be partially reconstituted in phospholipid vesicles (19). The first 65 amino acids of human pcCMT, containing the first two hydrophobic sequences, are 36% identical to amino acids 750 -821 of the human band 3 anion transporter, which represent the 11th and 12th of 14 well characterized membrane-spanning domains, further supporting the hypothesis that human pcCMT is a multiple membrane-spanning protein. Comparison of human pcCMT with all related sequences in the GenBank (Fig. 1) revealed the highest degree of divergence in the N-terminal third of the molecule, suggesting that the catalytic domain is C-terminal. Sequences homologous to the S-adenosylmethionine binding regions described in aspartyl and glutamyl protein carboxyl methyltransferases (20) were not apparent, arguing against an evolutionary link.
To determine whether the HL60 Ste14p homolog has prenylcysteine carboxyl methyltransferase activity, the cDNA was subcloned into a mammalian expression vector and transiently overexpressed in COS-1 cells. Membranes prepared from these cells were used as a source of recombinant enzyme. Farnesylated Ras GTPases require detergent for extraction from membranes. Therefore geranylgeranylated neutrophil cytosolic Rho GTPases that remain soluble by virtue of association with an accessory protein, guanine nucleotide dissociation inhibitor (GDI), were utilized as endogenous substrates in an in vitro assay (Fig. 2). Untransfected COS-1 cell membranes had little pcCMT activity toward Rho proteins compared with that of membranes derived from human neutrophils (Fig. 2a). Transfection with the HL60 cDNA conferred pcCMT activity toward Rho GTPases on COS-1 membranes, and this activity was blocked by the competitive pcCMT inhibitor, AFC, with an ED 50 (10 M) identical to that for endogenous neutrophil pc-CMT. Carboxyl methylation of Rac2 and RhoA by HL60 pc-CMT-transfected COS-1 cell membranes was enhanced by GTP␥S (Fig. 2b), similar to the activity observed in neutrophil membranes (7). Carboxyl methyltransferase activity was quantitated in membranes of COS-1 cells transfected with HL60 pcCMT using prenylcysteine analogs as defined substrates (14). Untransfected COS-1 cell membranes had 23 Ϯ 10% (n ϭ 5) of the pcCMT specific activity of neutrophil membranes toward AFC. Transfection of COS-1 cells with HL60 pcCMT resulted in an 18 -53-fold increase in specific AFC carboxyl methyltransferase activity (3.0 Ϯ 0.8 to 74.5 Ϯ 6.8 pmol/ mg⅐min, n ϭ 5, p Ͻ 0.0005). Carboxyl methylation of N-acetyl-S-all-trans-geranylgeranyl-L-cysteine (AGGC) was increased in parallel with methylation of AFC (18-versus 20-fold increase, n ϭ 2), consistent with previous studies demonstrating that a single activity carboxyl methylated both farnesylated and geranylgeranylated substrates (21). The Michaelis constants of the recombinant enzyme, K m ϭ 7 M for AFC and 0.6 M for AGGC, were similar to those for the endogenous neutrophil enzyme (19). Thus, the cDNA described above encodes authentic human myeloid pcCMT.
To determine whether human pcCMT could substitute in vivo for the S. cerevisiae pcCMT, Ste14p, we performed a complementation analysis using mating as a biological readout. The HL60 pcCMT cDNA expressed from a plasmid in a ⌬ste14 yeast strain partially restored the mating phenotype (Fig. 3), indicating that a-factor could be carboxyl methylated by human pcCMT. Thus, HL60 pcCMT is a functional human homolog of S. cerevisiae Ste14p.
Expression of the human pcCMT gene was examined by Northern analysis. HL60 cells expressed two pcCMT mRNAs, a 3.6-kb transcript consistent with the isolated cDNA and a 5-kb transcript suggesting a related gene or an alternatively spliced message (Fig. 4a). The level of expression of both transcripts was diminished by granulocytic differentiation of HL60 cells induced by Me 2 SO, consistent with our observation that HL60 membranes have 5-fold greater specific pcCMT activity than  9 -12). In some reactions (lanes 3, 7, and 11) a competitive pcCMT inhibitor, AFC, was included. Reaction products were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. The position of carboxyl-methylated p21s is indicated as is the position of protein phosphatase 2A (PP2A) that is methylated on a C-terminal leucine by a cytosolic enzyme unrelated to pcCMT (33). b, analysis as in A showing GTP dependence (Ϯ10 M GTP␥S) of carboxyl methylation of partially purified human neutrophil Rac2 and RhoA, but not cdc42hs, by COS-1 cell membranes expressing human pcCMT. FIG. 3. Complementation of S. cerevisiae ste14. Strains of S. cerevisiae deficient in mating because of a deletion of ste14 were transformed with pG1 vector alone (a) pSM1237 that drives plasmid expression of STE14 (b), or pG1-hCMT that drives plasmid expression of human pcCMT (c). Transformants were tested by the patch mating assay described under "Experimental Procedures." Mating was scored as prototrophic growth of resulting diploids. membranes of mature, peripheral blood neutrophils. 3 Both transcripts were ubiquitously expressed in human tissues (Fig. 4b).
To characterize endogenous pcCMT, a polyclonal antiserum was raised against an internal HL60 pcCMT peptide (amino acids 185-201, Fig. 1). Immunoblots of neutrophil and pcCMTtransfected COS-1 cell membranes using this antiserum revealed a 33-kDa protein (Fig. 4c) corresponding to the predicted size of the protein encoded by the HL60 pcCMT cDNA, confirming that this cDNA, which lacked a termination codon 5Ј of the ORF is, in fact, full-length. The same antiserum immunoprecipitated a 33-kDa protein from HL60 cells (Fig. 4c). Although the deduced amino acid sequence of pcCMT reveals a potential N-glycosylation site (Fig. 1), these data argue against glycosylation.
Ras is constitutively carboxyl methylated (22) and expressed at the plasma membrane (23), a localization required for its biological activity (24). Like all CAAX proteins, Ras lacks a signal peptide, is synthesized in the cytosol, and modified immediately posttranslationally by a cytosolic prenyltransferase (25). Because prenylcysteine carboxyl methylation is catalyzed by an intrinsic membrane protein and represents the last of the three posttranslational modifications of the CAAX cysteine that enhance the affinity of Ras for membranes, the simplest model of plasma membrane targeting predicts pcCMT expression in the target membrane. However, activities associated with the further processing of prenylated proteins, including pcCMT (26), S-isoprenyl-CAAX high affinity binding (27), Sisoprenyl-CAAX protease (28,29), and palmitoyltransferase (30) activities, have all been reported in microsomal fractions. Furthermore, one of the two S-isoprenyl-CAAX proteases recently identified in yeast has a putative ER retention sequence (31,32), and a double deletion of these genes led to mislocalization of yeast Ras2p to internal membranes and cytosol (31). Nevertheless, none of these studies excluded expression of prenylcysteine-modifying activities from plasma membranes. Indeed, we have reported pcCMT activity in neutrophil subcellular fractions enriched for surface membrane (19).
To determine the subcellular localization of pcCMT we constructed a recombinant pcCMT tagged at the C terminus with GFP. In CHO (Fig. 5, a-c), COS-1 (Fig. 5, d-e), and NIH3T3 (not shown) cells, CMT-GFP was visualized in the ER, Golgi, and nuclear membranes but not in the plasma membrane. This assay afforded a sensitivity and resolution that allowed visu-alization of individual ER canaliculi extending to the periphery of the cell (Fig. 5, c and d) and was therefore definitive in exlcuding expression in the plasma membrane. ER expression was confirmed by colocalization with ribophorin I (Fig. 5f), a component of the glycosyl transferase complex restricted to ER. Because the 27-kDa GFP added to the C terminus of pcCMT might alter its native subcellular localization, the CMT-GFP localization was confirmed with a pcCMT tagged at the C terminus with a 10-amino acid Myc epitope (Fig. 5g). Finally, an anti-peptide antiserum reactive with pcCMT was used to 3 2) with preimmune (p) or immune (i) serum from a rabbit immunized with a pcCMT peptide. Membranes from human neutrophils (PMN, lanes 3 and 4) and from COS-1 cells transfected with pcCMT (CMT, lanes 5 and 6) or vector alone (VEC, lane 7) were immunoblotted with the same sera. IP, immunoprecipitate; IB, immunoblot. confirm these data by localizing endogenous pcCMT to ER and nuclear membranes in COS-1 cells (Fig. 5, h-i). The absence of Golgi staining for endogenous pcCMT suggests that the Golgi localization of ectopically expressed pcCMT may result from gene overexpression.
These data demonstrate that mammalian pcCMT is an intrinsic membrane protein localized to a compartment topologically removed from the plasma membrane. Similar observations have recently been made in yeast. 4 It is curious that proteins such as Ras, synthesized on free ribosomes, prenylated in the cytosol, and destined for the cytoplasmic leaflet of the plasma membrane, are diverted to the ER for processing. Moreover, the ER restriction of pcCMT suggests that an uncharacterized transport pathway must mediate the translocation of fully processed GTPases from internal membranes to the cell surface. Such a pathway could utilize the cytoplasmic surface of secretory vesicles, cytosolic accessory molecules analogous to GDI, or a novel transport system. Investigation of this aspect of Ras biology may open new approaches to the development of pharmacologic inhibitors of oncogenic Ras.