Evidence for Post-translational Incorporation of a Product of Mevalonic Acid into Swiss 3T3 Cell Proteins*

Previous studies have identified several cellular requirements for mevalonic acid that appear unrelated to cholesterol, dolichol, or ubiquinone. To search for other products of mevalonic acid that might account for these requirements we cultured Swiss 3T3 cells in the presence of mevinolin, an inhibitor of mevalonic acid biosynthesis, then labeled the cells with exogenous radioactive mevalonic acid. Upon analyzing the radio- active material formed, we found that 40-50% of it was not extractable into lipid solvents, and that most of the lipid-insoluble material behaved like protein when treated with sodium dodecyl su1fate:chloro-form:phenol, RNase, or proteinase K. Further analysis by electrophoresis revealed that radioactivity was associated with a few specific proteins that had apparent molecular weights of 13,000-58,000. Control experiments indicated that authentic radioactive (R)-meva- lonic acid was the active precursor. Other lines of evidence suggested that mevalonate was first con- verted to an isoprenoid compound, then covalently incorporated into proteins by way of a cycloheximide-insensitive mechanism. These results suggest that Swiss 3T3 cells possess novel metabolic products of mevalonic acid metabolism that are formed by post-translational incorporation of isoprenoids into specific cell proteins. Compactin and mevinolin competitive of reductase that block the biosynthesis of MVA in cultured cells. Compactin, on DEAE-cellulose-For compe- tition experiments we partially purified radioactive and non-radio-active mevalonate on DEAE-cellulose (11). (RS)-[5-3H]MVA was loaded on a column (0.9 X 30 cm) of acid base washed DEAE-cellulose (COZ- form) in 1 mM ammonium carbonate, pH 9. After washes with 1 mM ammonium carbonate and dilute PBS (1;200 in water), f'H] MVA was eluted with full strength PBS containing penicil- 1in:streptomycin and stored at -75 "C. Non-labeled MVA was spiked with trace amounts of [3H]MVA, then DEAE-purified in the same way. Electrophoresis and Fluorography of Radioactive Cell Material- Samples from equal numbers of cells (approximately 160 pg of pro-tein/gel lane) were dissolved in electrophoresis sample buffer con- taining 5% 2-mercaptoethanol, 2 M urea, and 2% SDS, heated at 100 "C for 10 min, and subjected to electrophoresis in 0.1% SDS, 2 M urea through a 1.5-mm thick 12.5% polyacrylamide slab gel with a 5% stacking gel (12). The gel was stained briefly with Coomassie Brilliant Blue R-250 (13), partially destained, impregnated with EN3HANCE (New England Nuclear), then dried at 60 "C on a gel slab dryer. Preflashed (14) Kodak X-Omat R or AR film was clamped against the gel and exposed at -75 "C. 14C-labeled molecular weight standards (67,000, 43,000, 30,000, 20,000, and 14,000) were chromatographed in adjacent lanes. Estimated molecular weights of the MVA- labeled proteins were obtained from densitometric scans of the fluo-rograms. we established that the use of protease inhibitors is not esential for the demonstration of MVA-labeled proteins. The omission of protease inhibitors during the harvesting did not result in the loss

shape of Swiss 3T3 cells. For each of these three requirements the unidentified active products of MVA metabolism are likely to be quantitatively minor cell constituents. When cholesterol is present in the cell medium, none of the requirements become manifest until sufficiently high concentrations of compactin or mevinolin are added to inhibit residual MVA biosynthesis almost completely. Then, uptake of tiny amounts of exogenous MVA suffices to overcome the effects of the inhibitors.
To search for the active products we studied the metabolic fate of radioactive MVA under culture conditions that were chosen to detect relevant, quantitatively minor products. We cultured large quantities of Swiss 3T3 cells in the presence of cholesterol and amounts of mevinolin sufficient to induce MVA-deficient cell rounding. We then added concentrations of radioactive MVA that we previously found were barely sufficient to prevent (50-100 pM) or reverse (600 pM) the shape change. Upon analyzing the intracellular products that were formed, we obtained evidence for what appears to be a hitherto undescribed pathway of MVA metabolism in mammals (8).

MATERIALS AND METHODS
Except as noted, all radioisotopes were obtained from New England Nuclear and all reagents and enzymes from Sigma. (R)-[l-"C]Mevalonolactone (10 mCi/mmol) was a generous gift from Dr. J. Watson (University of California, San Francisco). Mevinolin was the kind gift of Dr. A. Alberts (Merck, Sharp & Dohme Research Laboratories, Rahway, NJ) and was converted to its Na+ salt before use (2). PBS without Ca2+ and M e was prepared by the method of Dulbecco (9).
Tissue culture reagents were from Gibco Laboratories (Grand Island, NY).
Cell Culture-Mass cultures of Swiss 3T3 cells (2) were obtained by inoculating approximately 2 X lo6 cells into 850 cm2 roller bottles (Corning Scientific Products, Corning, NY) in 100 ml of Dulbecco's modified Eagle's medium containing 10% calf serum, non-essential amino acids, sodium pyruvate, and penicil1in:streptomycin. The bottles were gassed with 5% COs/air, sealed, and maintained a t 37 "C on a roller bottle apparatus (Wheaton Scientific, Millville, NJ) until the cultures reached confluence. The complete culture medium contained approximately 60 pg/ml of total cholesterol.
The preceding delipidation procedure appears to be complete with respect to MVA-labeled lipids because further extractions with either the same solvents or with CHCl3:methanol:water (10103) failed to release significant amounts of MVA-derived radioactivity. At least 95% of cellular cholesterol was extracted by this procedure (4). The labeled lipids have not all been identified but approximately 50% of the acetone-soluble material was in a non-saponifiable lipid that comigrated with cholesterol during thin-layer chromatography on Silica Gel H in heptane:ether:methanol(909:15) (4) while 10-30% behaved as MVA or more polar lipids (not shown).
In other experiments we incubated confluent cultures with 30 p~ mevinolin in fresh medium for 24 h, then scraped the cells directly into the medium. Aliquots of cells were spun down, resuspended in a small volume of the same mevinolin-containing medium, and incubated with t3H]MVA (200 pCi/ml) for 3 h at 37 "C in a 5% COn atmosphere with occasional agitation. The labeled cells were washed twice with PBS, incubated with protease inhibitors, extracted three times with 2 ml of acetone, then dissolved in electrophoresis sample buffer.
Partial Purification of f H ] M V A on DEAE-cellulose-For competition experiments we partially purified radioactive and non-radioactive mevalonate on DEAE-cellulose (11). (RS)-[5-3H]MVA was loaded on a column (0.9 X 30 cm) of acid base washed DEAE-cellulose (COZ-form) in 1 mM ammonium carbonate, pH 9. After washes with 1 mM ammonium carbonate and dilute PBS (1;200 in water), f' H] MVA was eluted with full strength PBS containing penicil-1in:streptomycin and stored at -75 "C. Non-labeled MVA was spiked with trace amounts of [3H]MVA, then DEAE-purified in the same way.
Electrophoresis and Fluorography of Radioactive Cell Material-Samples from equal numbers of cells (approximately 160 pg of protein/gel lane) were dissolved in electrophoresis sample buffer containing 5% 2-mercaptoethanol, 2 M urea, and 2% SDS, heated at 100 "C for 10 min, and subjected to electrophoresis in 0.1% SDS, 2 M urea through a 1.5-mm thick 12.5% polyacrylamide slab gel with a 5% stacking gel (12). The gel was stained briefly with Coomassie Brilliant Blue R-250 (13), partially destained, impregnated with EN3HANCE (New England Nuclear), then dried at 60 "C on a gel slab dryer. Preflashed (14) Kodak X-Omat R or AR film was clamped against the gel and exposed at -75 "C. 14C-labeled molecular weight standards (67,000, 43,000, 30,000, 20,000, and 14,000) were chromatographed in adjacent lanes. Estimated molecular weights of the MVAlabeled proteins were obtained from densitometric scans of the fluorograms.
During the course of this investigation we established that the use of protease inhibitors is not esential for the demonstration of MVAlabeled proteins. The omission of protease inhibitors during the harvesting procedure did not result in the loss of any MVA-labeled bands and resulted in only minor changes in their apparent molecular weights: the 45,000, 29,000, 24,000, and 23,000 bands were altered by less than 5% (not shown). Furthermore, all changes not involving the 45,000-Da band were due to an effect of N-ethylmaleimide which was apparently unrelated to inhibition of proteolysis (not shown). Similarly, cells harvested by direct solubilization in hot 0.1 M leupeptin in 1% SDS yielded proteins of the same apparent molecular weights as cells harvested without protease inhibitors (not shown). It might be argued, therefore, that protease inhibitors can be omitted routinely from the cell harvesting procedure.
Selective Enzymatic Digestion 'of Radioactive Macromolecules-Samples from 1/10 of a roller bottle were incubated with or without 20 pg of proteinase K (Beckman) in 220 pl of 1% SDS, 50 mM EDTA, 0.1 M Tris, pH 8 (pretreated with diethyl pyrocarbonate), for 1 h at 45 "C. Remaining macromolecules were precipitated with 2 ml of cold 95% EtOH, then dissolved in electrophoresis sample buffer.
Similar samples were suspended in 200 pl of 50 mM Tris, pH 7.4 (pretreated with diethyl pyrocarbonate), sonicated briefly in a sonicating water bath, then incubated with or without 20 pg of RNase A (Calbiochem-Behring) in the same buffer (previously boiled for 10 Proteins by Mevalonic Acid min) at 37 "C for 60 min. Remaining macromolecules were precipitated with 2 ml of cold 95% EtOH and redissolved in electrophoresis sample buffer.
Extensive Proteolysis of MVA-labeled Proteins and Subsequent Purification-MVA-labeled proteins from the organic phase of the SDS:CHCb:phenol extraction were suspended in 50 mM Tris acetate, 3 mM calcium acetate, pH 7.8, containing 0.3 mg/ml of Pronase for 3 days at 37 "C (15). After centrifugation the supernatant was discarded, and the insoluble material dissolved in 50 mM Tris acetate, 10 mM EDTA, 0.5% SDS, pH 8. Proteinase K (0.1 mg/ml) was added at 0 and 3 h and the digestion continued at 45 "C for 24 h (16). After the pH was adjusted to 5-5.5, carboxypeptidase Y (200 pg/ml) was added at 0 and 3 h, and the incubation continued for 24 h at 37 "C (17).
The resulting hydrolysate was applied directly to a column (1.0 X 10 cm) of Bio-Rad AG 1-X2 (formate) in 1 M ammonium formate, pH 6.6. After washes with 1 M ammonium formate and water the radioactivity was eluted with a nonlinear gradient from water to 88% MVA (50 pCi/ml), then analyzed the radioactive intracellular products that had been formed. We found typically ( Table I) that about half of the cell-associated radioactivity could not be extracted with lipid solvents. By three criteria, most of this non-lipid radioactivity seemed to be associated with protein rather than RNA (19). First, when we employed an SDS: CHC13:phenol extraction technique to separate nucleic acids from proteins, we found ( Table I) that about 80% of the nonlipid radioactivity was soluble in the protein-containing SDS:CHC13:phenol phase, whereas only 16% of the radioac- Initial cell-associated radioactivity for MVA-, uridine-, and leucine-labeled cells was 846,930,385,570, and 894,400 cpm, respectively.
Numbers refer to per cent recovery of non-lipid radioactivity in the two phases of the phenol extraction.
108.1 96.0 118.8 tivity was soluble in the RNA-containing aqueous phase. This distribution of radioactivity was similar to that obtained when corresponding non-lipid material from ['Hlleucine-labeled control cells was fractionated by the same technique, but differed dramatically from that obtained using comparable ["Hluridine-labeled material (Table I).
Second, SDS-gel electrophoresis revealed that the MVAderived non-lipid radioactivity was associated with discrete macromolecules (Fig. 1). Most labeled macromolecules were present primarily or exclusively in the organic phase, although prominent bands of less than 14 kDa were found only in the aqueous phase. Eighty per cent of the organic phase label, as demonstrated by gel slicing and counting, was in macromolecules that had apparent M,, = 20,000-30,000. Individual leucine-labeled bands from control cells also were found primarily in the organic phase, but most of them were macromolecules larger than 30 kDa.
Third, MVA-labeled macromolecules were degraded under conditions of selective proteolysis.
In three experiments, MVA-labeled macromolecules from both the aqueous and organic phases were completely hydrolyzed by proteinase K in 1% SDS, but not by boiled RNase (not shown). The sole exception was the broad MVA-labeled band of approximately 14 kDa that appeared to be completely resistant to proteinase K. Control digestions of leucine-and uridine-labeled material showed that the digestion conditions were completely selective for protein and RNA. Taken together, these results strongly suggest that most of MVA-derived non-lipid radioactivity is tightly linked to protein.
A minor fraction of the MVA-labeled, non-lipid material, however, probably consisted of isopentenylated tRNA. We found previously (20) that some MVA-labeled, aqueous phase material co-migrates with tRNA in an electrophoretic systev Other lanes contain "C-labeled molecular weight markers. In this experiment, the MVA-labeled non-lipid material contained approximately 40,000 cpm/mg of protein.

Proteins by Mevalonic
Acid 10177 (21) for RNA. Furthermore, in the present investigation we hydrolyzed aqueous phase material in 0.3 N KOH for 18 h a t 37 "C and found that 19% of the label was converted to a perchloric acid-soluble form that co-chromatogrhphed with AMP on Sephadex G-25 (not shown). Only 6% of the organic phase label was released in perchloric acid-soluble form and less than 2% of it co-chromatographed with AMP; the remaining 4% was in unidentified higher molecular weight material.
Evidence That the Radioactivity Incorporated into 3T3 Cell Protein Is Derived from Authentic Labeled (R)-MVA-AIthough the experiments described above provided strong evidence that much of the cell-associated radioactivity in our experiments was present in protein, it remained to be demonstrated that this radioactivity was actually derived from radioactive MVA. These results strongly suggest that authentic radioactive (R)-MVA is indeed the source of the labeled moiety that becomes incorporated into 3T3 cell proteins.
Using a modified labeling system we obtained evidence suggesting that unlabeled MVA can compete with radioactive MVA for incorporation into cell proteins whether the unlabeled MVA is derived from endogenous or exogenous sources. Thus, when we incubated cells with or without mevinolin for 24 h, then labeled them with radioactive MVA and examined the proteins by SDS gel electrophoresis, we found (Fig. 2) detectable radioactivity in macromolecules only in the cells that had been pretreated with mevinolin. In the same experiment we also added increasing concentrations of unlabeled MVA to the culture medium of mevinolin-pretreated cells a t the same time that we added radioactive MVA, and found that concentrations of unlabeled MVA greater than 100 p~ competitively prevented the incorporation of radioactivity into cell protein. Below 100 p~ unlabeled MVA, the amount of protein-bound radioactivity was unaffected, presumably because insufficient MVA was taken up by the cells during the relatively short labeling period to fill all potential protein acceptor sites.
Evidence That Radioactivity from MVA Is Incorporated into Proteins by a Post-translational Mechanism-We tested the abilities of cycloheximide (12 pg/ml) and chloramphenicol (60 or 300 pg/ml) to prevent mevinolin-treated cells from incorporating radioactive MVA into proteins. Neither inhibitor affected the incorporation of labeled MVA into cell proteins (Fig. 3) even though cycloheximide dramatically reduced the incorporation of labeled leucine into the proteins of control cells (not shown). Similar results were obtained both times the experiment was repeated. Cycloheximide had no effect on the conversion of labeled MVA into lipids which co-migrated with cholesterol, cholesteryl oleate, dolichol, or ubiquinone during thin-layer chromatography in heptane:ether:methanol (90:9:15) (not shown). Thus, it appears that the labeling of cell proteins by MVA occurs by a post-translational mechanism rather than a pre-or co-translational one.
Partial Purification of Proteolytic Fragments of MVA-labeled Proteins-We have begun to purify proteolytic fragments of MVA-labeled proteins in order to determine the structure of the MVA-derived moiety and its link to protein.
In order to monitor our purification we first prepared delipidated, phenol-extracted proteins that had been labeled either with [2-14C]MVA (as in Table I) or with 3H-amino acids (by growing cells for 4 days in medium that contained either labeled phenylalanine, tyrosine, tryptophan, cysteine or a labeled algal protein hydrolysate). After separately incubating these proteins with Pronase for 3 days, we found in eight experiments that 96-9996 of the 3H-labeled material became water-soluble whereas 88-92% of I4C-labeled material remained insoluble. By centrifuging the hydrolysate and decanting the supernatant we thus effected an approximately 30-fold enrichment of the MVA-labeled material. We then dissolved the insoluble residues in SDS, pooled them, and further incubated them with proteinase K and carboxypeptidase Y. Although these digestions had little effect on MVAlabeled material, 3H-labeled material was degraded further (not shown). By passing the final hydrolysate over an AG 1x 2 (formate) column in 1 M ammonium formate, we separated water-soluble fragments (which washed through the column) from SDS and MVA-labeled material (which bound). We then eluted MVA-labeled products (substantially free of SDS) with formic acid: EtOH (1:4). In seven experiments, this step caused a further 2.2-15-fold purification. Analytic chromatography of the MVA-labeled eluate over Sephadex LH-20 in formic acid:EtOH (1:4) revealed that it contained at least two major components with apparent sizes of 1000 and 500 Da (Fig. 44).
By preparative LH-20 chromatography we separated these components from each other (Fig. 4B)  achieved an additional 3-fold purification. The MVA-labeled material in the resulting pools showed an overall enrichment of 400-750-fold, with recoveries of up to 70%. Nevertheless, it still appeared to contain substantial amounts of other fragments that were prelabeled by the amino acids (not shown).
Both the material of 1000 Da and that of 500 Da were extremely hydrophobic and even insoluble in 6 M guanidine hydrochloride. They also were almost insoluble in hexane but appeared to be very soluble in acidic polar organic solvents. Upon being chromatographed on thin-layer plates of silicic acid in solvent systems for neutral lipids (4,22), they remained at the origin, unlike cholesterol, dolichol, and ubiquinone. On the other hand, they migrated near the solvent front in a phospholipid solvent system (23). These properties, and the results of our other work, are consistent with the possibility that the MVA-labeled, proteolytic fragments are small, hydrophobic peptides that contain polyisoprenoid side chains.

DISCUSSION
In this investigation we studied the metabolic fate of radioactive MVA in mevinolin-treated cultures of 3T3 cells and obtained results strongly suggesting that a product of authentic (R)-MVA was converted to a protein-bound form. Since no protein-bound products of MVA in mammalian cells have been described previously, several important questions arise. Could the labeled proteins be experimental artifacts? In what form is the radioactivity bound to protein? Are the modified proteins likely to be biologically important?
We considered two potential artifacts: 1) spurious labeling of proteins by a radiolabeled contaminant, and 2) protein ''labeling'' as a result of nonspecific complex formation between proteins and labeled lipids. Our experiments provided strong evidence that authentic (R)-MVA serves as the active precursor (see "Results"). Moreover, they made it seem very unlikely that the MVA-labeled proteins represent noncovalent complexes of unlabeled proteins with either MVA or labeled isoprenoids (including isopentenylated tRNA). Thus, the radioactive products released by extensive proteolytic digestion clearly differed from MVA during chromatography on Sephadex LH-20, and also behaved differently from cho-lesterol, dolichol, ubiquinone, and cholesteryl oleate during thin-layer chromatography. Furthermore, we were not able to separate the radioactive moiety from proteins using any of the following methods: extraction with organic solvents, extraction with SDS:CHCls:phenol, SDS-gel electrophoresis, or chromatography in formic acid:EtOH (1:4). We were also unable to liberate the radioactivity using extractions for farnesyl-containing heme a (24)(25)(26) or solvents that disrupt tight, noncovalent binding of polyphosphorylated lipids to proteins (27). These results strongly suggest, therefore, that the MVAderived moiety is covalently linked to protein. Since, to the best of our knowledge, this type of linkage has never been described in mammalian cells, it seems likely that MVAlabeled proteins represent novel products of MVA metabolism, either known products linked in novel ways or entirely novel metabolites of MVA.
Radioactivity in MVA-labeled proteins might have reflected the presence of MVA itself, a product of a degradative pathway such as the transmethylglutaconate shunt, or an isoprenoid product of MVA. One type of experiment strongly suggests, however, that radiolabeled MVA was not involved. When we incubated mevinolin-treated 3T3 cells with either [1-I4C]-or [2-14C]MVA for 24 h then measured radioactivity in proteins by electrophoresis and fluorography, we found that l-14C was incorporated into proteins at least 91-fold less efficiently than 2-'*C.' Because C1 is selectively lost from MVA during the formation of isopentenyl pyrophosphate, we postulate the MVA-labeled proteins are actually labeled by either isopentenyl pyrophosphate or one of its metabolic products.
Our results make it seem unlikely that labeled isoprenoid pyrophosphates are first degraded to smaller labeled intermediates, then incorporated into proteins. Most importantly, 5-3H-, 3-14C-, and 2-14C-labeled MVA were each incorporated into the same proteins and [2-'4C]-and [5-3H]MVA, when added to cells concomitantly, were both incorporated into proteins in the same ratio as into lipids. This strongly suggests that the bonds between the labeled atoms were not cleaved during the conversion of labeled MVA to labeled proteins. It also specifically argues against a role of the transmethylglutaconate shunt (28) because the shunt converts [5-3H]-and [2-I4C]MVA to labeled acetyl-coA and acetoacetate, respectively. In addition, the conversion of labeled MVA to long chain fatty acids by the shunt is not a quantitatively important pathway under our culture conditions. Less than 2.5% of the radioactivity in the lipid or protein extracts of MVAlabeled cells was in a saponifiable form that co-migrated with fatty acids in a neutral lipid chromatography ~y s t e m .~ Taken together, these results strongly suggest that the fragment of MVA that is incorporated into proteins is not an end product of the shunt pathway or any other degradative pathway. On the other hand, they do not rule out the possibility that labeled proteins are formed from an early intermediate in the shunt (e.g. dimethyl acrylic acid).
A final possibility is that the MVA-labeled proteins contain isoprenoid products of MVA. At least three mechanisms can be envisioned whereby an isoprenoid compound might become covalently bound to protein. In one, a cysteine residue might condense across the double bond in isoprenoids, as may occur in felinine (29). In a second, isoprenoid fatty acids such as have been found in bovine retina (30) might be linked to proteins through amide or ester bonds, although the latter

results.
Labeling of Swiss 3T3 Cell Proteins by Mevalonic Acid possibility seems unlikely because the radioactivity is not released by treatment with 0.5 M hydroxylamine, pH 11, or 0.1 N NaOH even though ethyl acetate is hydrolyzed.' In a third mechanism, the isoprenoid precursor might be converted to a reactive intermediate with carbonium ion characteristics which then couples to electron-rich sites in proteins via its C, carbon atom. The latter mechanism is thought to underlie the polymerization of isoprenoids (reviewed in Ref. 31), and may account for the biosynthesis of isopentenyl adenosine (32), dimethylallyl tryptophan (33), heme a (34), and yeast peptidal sex hormones (35). In order to choose among these mechanisms, we shall need to determine the structure of the MVAderived moiety and its linkage to protein from fragments of MVA-labeled proteins. We have made progress toward this goal and have purified labeled proteolytic fragments of 3T3 cell proteins by approximately 400-750-fold. This approach has been limited, however, by the small amount of MVAlabeled protein contained within 3T3 cells (see below) and we are currently investigating alternative starting materials that contain more modified protein.
Three arguments raise the possibility that MVA-modified proteins are related to cellular requirements for MVA such as DNA synthesis, shape control, and HMG-CoA reductase regulation. First, appropriate amounts of MVA-labeled protein are present within 3T3 cells. Thus, when we cultured 3T3 cells for four population doublings in the presence of 32 pt M compactin and 92 PM radiolabeled MVA, we calculated the total protein-bound label to be equivalent to 265 x 10"' mol of MVA per cell (approximately 265 pmol of MVA/mg of cell protein). This estimate is quite similar to our previous finding (4) that the uptake of as little as 250 X 10"' mol of MVA/ cell is sufficient to reverse the cellular shape change due to MVA deficiency. Second, a major fraction (26-46% in five experiments) of the MVA that partially prevents cell rounding in the roller bottle culture system is converted ultimately to protein-bound form. Finally, unpublished results in our PDGF-stimulated cell culture system (2,4) indicate that MVA-labeled proteins are major metabolic products of MVA that turn over very slowly but nonetheless become deficient at the same time that mevinolin-treated cells change shape and are prevented from synthesizing DNA. Taken together, these results support the existence of MVA-labeled proteins as important novel metabolic products of MVA. Future studies into the identity of the individual labeled proteins, their function within cells, and their relation to cellular requirements for MVA will be of great interest.