Molecular cloning of human phosphomevalonate kinase and identification of a consensus peroxisomal targeting sequence.

Two overlapping cDNAs which encode human liver phosphomevalonate kinase (PMKase) were isolated. The human PMKase cDNAs predict a 191-amino acid protein with a molecular weight of 21,862, consistent with previous reports for mammalian PMKase (Mr = 21,000-22,500). Further verification of the clones was obtained by expression of PMKase activity in bacteria using a composite 1024-base pair cDNA clone. Northern blot analysis of several human tissues revealed a doublet of transcripts at approximately 1 kilobase (kb) in heart, liver, skeletal muscle, kidney, and pancreas and lower but detectable transcript levels in brain, placenta, and lung. Analysis of transcripts from human lymphoblasts subcultured in lipid-depleted sera (LDS) and LDS supplemented with lovastatin indicated that PMKase gene expression is subject to regulation by sterol at the level of transcription. Southern blotting indicated that PMKase is a single copy gene covering less than 15 kb in the human genome. The human PMKase amino acid sequence contains a consensus peroxisomal targeting sequence (PTS-1), Ser-Arg-Leu, at the C terminus of the protein. This is the first report of a cholesterol biosynthetic protein which contains a consensus PTS-1, providing further evidence for the concept that early cholesterol and nonsterol isoprenoid biosynthesis may occur in the peroxisome.

The pathway of cholesterol and nonsterol isoprenoid biosynthesis provides a variety of products necessary for growth of mammalian cells, including cholesterol, bile acids, heme A, dolichol, and ubiquinone. This pathway is subject to multilevel regulation at several sites, at both the post-translational and transcriptional level (1). The key regulatory site is 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) 1 reductase, which catalyzes the reaction producing mevalonate. When levels of cholesterol are high, HMG-CoA reductase is subject to downregulation at the transcriptional level, increased turnover of protein, and inactivation of existing enzyme. Conversely, when cholesterol levels are low, transcription of HMG-CoA reductase increases, protein turnover decreases, and the existing protein can be reactivated. In addition, the levels of low density lipoprotein receptors, which are responsible for cellular uptake of cholesterol, are regulated at the level of transcription. Through these mechanisms, mammalian cells have the capacity to precisely regulate cholesterol and nonsterol isoprenoid metabolites, thereby avoiding overaccumulation of potentially toxic metabolites.
Although HMG-CoA reductase is a key regulatory site in the pathway, enzymes and genes further along the pathway are also important in regulation of pathway function. Activities of several enzymes involved in mevalonate disposition, mevalonate kinase (MKase), phosphomevalonate kinase (PMKase), and mevalonate diphosphate decarboxylase (MDDase), are responsive to cholesterol intake in animals (2)(3)(4)(5)(6)(7)(8). The transcription of the genes encoding farnesyl-diphosphate synthase and squalene synthase are also responsive to cholesterol levels, but the mechanisms are different from those involved in the control of transcription for the HMG-CoA reductase and low density lipoprotein-receptor genes (9,10). MKase is subject to feedback inhibition at the protein level by pathway intermediates including geranyl and farnesyl diphosphates (3,4). Preliminary data suggest that MKase responds to cholesterol levels via a sterol regulatory element localized in the promoter, in a fashion similar to that observed for other cholesterol biosynthetic genes (11).
The cholesterol biosynthetic pathway contains a unique series of three sequential ATP-dependent enzymes which convert mevalonate to isopentenyl diphosphate: MKase, PMKase, and MDDase. Since mevalonate is necessary for cholesterol and isoprenoid synthesis, as well as a precursor for isoprenylation of various proteins, the regulation of its disposition in animal cells is of great interest. Although cDNAs encoding MKase have been published, and some regulatory properties revealed, little is known about the regulation and control of the latter two ATP-dependent genes involved in mevalonate disposition, primarily because of a lack of the corresponding mammalian cDNAs. A cDNA encoding PMKase from yeast has been presented, which encodes a protein with a predicted M r ϭ 48,000. However, the purified PMKase from porcine liver has been reported as a monomeric protein of M r ϭ 22,000 (12,13). This divergence from yeast to mammals seems unusual.
In order to undertake analysis of the regulation of mamma-* This work was supported by Grant-in-aid 94010450 from the American Heart Association (to K. M. G.) and Ho 966-4/1 from the Deutsche Forschungsgemeinschaft (to G. F. H.). 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.
The lian PMKase, a cDNA encoding this protein was isolated and the discrepancies in protein structure were addressed. In the present report, the molecular cloning of human PMKase is described. Our data indicate that human PMKase is subject to regulation by sterol at the level of transcription. The human gene contains a consensus peroxisomal targeting motif (Ser-Arg-Leu) (PTS-1) at the C terminus of the protein, the first identified in a cholesterol biosynthetic enzyme. This finding provides additional evidence that early cholesterol biosynthesis may occur in the peroxisome.

EXPERIMENTAL PROCEDURES
Protein Purification and Sequencing-Porcine liver PMKase was purified as described previously (13). N-terminal and tryptic peptide amino acid sequence was determined as described (14).
Amplification of Porcine Phosphomevalonate Kinase cDNA-The mixed oligonucleotides primed amplification of cDNA procedure was used to generate a DNA probe for library screening. First strand cDNA was synthesized with murine Moloney leukemia virus reverse transcriptase using a downstream antisense primer (GGCTCGAGGGNAC-NGTYTGNGTNA) according to Kawasaki (15). Thirty cycles of PCR were performed on the cDNA using the upstream primer GGGAAT-TCGGNAARAGRAARGAYGG and the above downstream primer with annealing for 45 s at 42°C, extension for 2 min at 72°C, and denaturing for 45 s at 94°C. A second round of PCR was carried out using 1 l of the first PCR as template. The same upstream primer was used, but a nested downstream primer with the sequence GGCTCGAGGTYTCYT-GRAACCAYTG replaced the previous downstream primer. The same temperature profile and cycle number were employed. The PCR product of 317 bp was gel-purified with Qiaex beads (Qiagen) and subcloned into the pBluescript II SK ϩ vector (Stratagene) using the EcoRI and XhoI restriction sites engineered onto the ends of the primers. Five isolates of the clone were sequenced and proven to be identical. The clone contained an open reading frame which encoded all of the sequenced porcine tryptic peptides.
Isolation of Human Phosphomevalonate Kinase cDNA-The 317-bp porcine liver PMKase fragment was used to screen Ͼ5 ϫ 10 5 plaques from a human liver ZAP cDNA library (Stratagene) using standard procedures. One 827-bp cDNA clone was isolated and sequenced. The human clone was 94% identical with the 317-bp porcine clone in the aligned region and contained a poly(A) tract at its 3Ј end; however, the human clone did not contain the complete N-terminal coding region. The remaining 5Ј end of the human cDNA was amplified from a Super-Script human liver cDNA library (Life Technologies, Inc.) using a set of nested, nondegenerate PMKase-specific antisense primers and an anchored sense primer in the library vector. The first amplification was carried out with the anchored primer (TAATACGACTCACTATAGGG) and the most downstream gene specific primer (TCCGAAAGGCCTC-CTTGTAG) using the amplification profile described above with the exception of an annealing temperature of 55°C and an elongation time of 1 min. A second amplification was performed with the same anchored sense primer and the nested antisense primer TCTGGAAGTTCAAGC-CATGC, using 1 l of the first amplification reaction as template. The same temperature profile and cycle number were used. The amplified product (341 bp) was treated with the Klenow fragment of T4 DNA polymerase to create blunt ends and digested at the library vector XhoI site. The resulting 255-bp fragment was cloned into the XhoI and SmaI sites of pBluescript II SK ϩ and sequenced.
DNA and Protein Sequence Analysis-DNA sequencing was determined by the dideoxynucleotide chain termination method (16). Doublestranded DNA sequencing was performed with the Sequenase-modified T7 DNA polymerase (U. S. Biochemical Corp.) following the manufacturer's protocol. DNA and protein sequences were aligned and compared using the MacDNAsis computer program (Hitachi). The EMBL/ GenBank TM data bases were searched for homologies to the DNA and protein sequences of PMKase.
Expression of Human Phosphomevalonate Kinase cDNA in Bacteria-A composite human liver PMKase cDNA clone was constructed in the glutathione S-transferase fusion vector pGEX-4T-1 (Pharmacia Biotech Inc.) for the purpose of protein expression. The 255-bp amplicon described above which contained 35 bp of 5Ј-untranslated region and 220 bp of coding region was excised from the pBluescript plasmid with NotI and ligated into the NotI site of the pGEX vector. The 5Ј-untranslated region, which contained two in-frame stop codons, was excised using NcoI and SmaI restriction endonucleases. A StuI/XbaI restriction fragment from the 827-bp human liver PMKase cDNA, which contained the remainder of the coding region and 116 bp of 3Ј-untranslated region, was then ligated into the construct. The resulting plasmid was transformed into Escherichia coli strain DH5␣ which were grown and harvested as described previously (14). Cell pellets were resuspended in 0.1 M MOPS/KOH buffer (pH 7.0) supplemented with 0.1 mM dithiothreitol and sonicated on ice for 3 min in 1-min intervals with 1 min between each interval. Cell debris was removed by centrifugation at 12,000 ϫ g for 10 min at 4°C. Various amounts of the cleared sonicates were assayed for PMKase activity using a radiometric assay which was a modification of that employed for MKase activity determination (17). The substrate was R-[2-14 C]mevalonate 5-phosphate, specific activity 45.5 mCi/mmol (DuPont NEN). Each assay contained 0.015 Ci of substrate, final assay concentration 4 M. Incubation was carried out at 37°C for 10 min. For additional verification, PMKase/glutathione Stransferase fusion protein was isolated from cleared sonicates with glutathione-Sepharose 4B as described by the manufacturer. Purified protein obtained using the fusion and vector only constructs was analyzed by SDS-polyacrylamide gel electrophoresis and exhibited protein bands of the expected molecular weight (M r ϭ 48,000 for the fusion and 26,000 for vector only). The purified fusion protein was assayed spectrophotometrically as described previously (3). Mevalonate 5-phosphate was produced in situ in this assay system using recombinant human MKase (18) which was produced and purified using conditions identical to those described for PMKase.
Northern and Southern Blot Analysis-Probing of human multiple tissue Northern blots (Clontech) was performed using the human liver PMKase 827-bp cDNA as probe (14). For RNA regulatory studies in cultured lymphoblasts, cells were grown under different conditions for 48 h. Total RNA was isolated from human lymphoblasts by standard procedures and separated on a 1% agarose gel containing 2.2 M formaldehyde. The gel was blotted and probed with the human liver PMKase cDNA. Blots were stripped of the PMKase probe and reprobed with a human ␤-actin probe for quantitation purposes. RNA levels were quantitated by scanning with an IS-1000 Digital Imaging System (Alpha Innotech Corp.) and comparing ratios of PMKase:␤-actin transcript levels. Human genomic Southern blots were prepared as described previously and probed with the human liver PMKase cDNA (14).

Isolation and Characterization of Liver
PMKase cDNAs-Amino acid sequence from purified porcine liver PMKase provided information for the design of oligonucleotide primers for reverse transcription and PCR amplification of the porcine PMKase gene as described under "Experimental Procedures." A 317-bp amplicon was cloned and shown to encode the amino acid sequence predicted from 5 tryptic peptides of porcine PMKase. Screening of a human liver library with the porcine PMKase clone yielded a single 827-bp human liver cDNA. Sequence analysis of the human clone revealed 94% identity with the porcine clone in the overlapping region. The human clone contained a poly(A) tract at the 3Ј end but did not contain the complete 5Ј end coding region. The remaining 5Ј end of the human cDNA was amplified from a second human liver cDNA library using two downstream nested PMKase specific primers and an upstream primer anchored in the library vector sequence. An additional 193 bp of human PMKase sequence was obtained which contained 35 bp of 5Ј untranslated region as well as the N-terminal region of the protein (Fig. 1).
The open reading frame of the human liver cDNA predicts a protein of 192 amino acids with a molecular weight of 21,862 consistent with previous reports for mammalian PMKase (M r ϭ 21,000 -22,500) (12,13). Comparison of the predicted amino acid sequence from the cDNA with that obtained from amino acid sequencing indicates that a single methionine is removed from the N terminus upon maturation of the protein. A consensus peroxisomal targeting sequence (PTS-1), Ser-Arg-Leu, resides at the C terminus of the protein.
Expression of Human PMKase cDNA in Bacteria-A composite clone was constructed which contained the complete coding region of human liver PMKase in-frame with the glutathione S-transferase (GST) gene. The construct was transformed into bacteria which were grown and induced as described under "Experimental Procedures." Lysates of the fusion clone and a control clone were assayed for PMKase activity by a radiometric assay using R-[2-14 C]mevalonate 5-phosphate as substrate. While the control clone (vector only) had no detectable PMKase activity, the fusion clone quantitatively converted the radioactive monophosphate to diphosphate (Table I). This activity was completely dependent upon exogenous ATP. For additional enzymatic verification, the PMKase/GST fusion protein, as well as GST alone, was isolated from the bacterial lysates via glutathione-Sepharose 4B affinity chromatography. The purified proteins were analyzed by gel electrophoresis and were shown to be of the expected molecular weight (data not shown). The proteins were assayed for PMKase activity by a spectrophotometric assay which again demonstrated substantial PMKase activity with the fusion protein and only background levels of activity with the GST protein (Table I).
Northern Blot Analysis of Human PMKase-Northern blot analysis of several human tissues revealed a broad hybridizing region at approximately 1 kb (Fig. 2A). The blot revealed differential tissue expression of the PMKase transcripts with the highest levels present in heart and skeletal muscle and slightly lower levels in liver, kidney, and pancreas. A low but detectable transcript level is present in brain, lung, and placenta. Upon closer inspection of the Northern blot data in underexposed gels and tissues with lower transcript levels, there appears to actually be a doublet of transcripts.
Regulation of PMKase Transcription and Genomic Southern Blot Analysis-To assess the possibility that PMKase is regulated at the level of transcription as are other cholesterol biosynthetic enzymes, we examined the level of PMKase mRNA in lymphoblasts grown under various conditions. Northern blot analysis of RNA from lymphoblasts subcultured in untreated fetal calf sera, lipid-depleted sera (LDS), and LDS supplemented with 1 M lovastatin is presented in Fig. 2B. Quantitation was performed by scanning of the autoradiographs and comparison of PMKase transcript levels with ␤-actin transcript levels. For cells subcultured in untreated fetal calf sera, the normalized ratio of PMKase transcript to ␤-actin was set at 1 (n ϭ 5 cell lines). When cultured in LDS, the normalized ratio increased slightly to 1.2 Ϯ 0.1 (n ϭ 5; mean Ϯ S.E.). The ratio increased to 2.3 Ϯ 0.6 (n ϭ 5) for cells grown in LDS supplemented with 1 M lovastatin.
Genomic Southern blot analysis of the PMKase gene revealed a single hybridizing band when any of three different restriction endonucleases were used for genomic DNA digestion (Fig. 3). Size estimations indicated that PMKase is a single copy gene, occupying slightly less than 15 kb in the human genome.

DISCUSSION
This report describes the first mammalian cDNA encoding PMKase, one of three ATP-dependent enzymes which convert mevalonate to isopentenyl diphosphate. A genomic DNA clone encoding PMKase (ERG8) from yeast has been described (19).

FIG. 1. Nucleotide and deduced amino acid sequences of human (H) and porcine (P) liver PMKase.
Dots in the porcine sequences represent bases or amino acid residues identical to human sequences. Underlined amino acid residues were determined by amino acid sequencing of porcine liver PMKase. The first 15 and last 9 amino acid residues of porcine PMKase are known only from amino acid sequencing performed in our laboratories. The consensus peroxisomal targeting sequence (PTS-1) is boxed, and the consensus polyadenylation signal (AATAAA) is underlined. a Activities estimated with the radiometric method are an underestimate because of limiting substrate concentration (all substrate was converted to product). The radiometric assay was used with crude bacterial extracts.
b As an NADH consumption spectrophotometric assay, crude extracts could not be used. In these studies, the proteins from bacterial extracts were isolated on GST-Sepharose affinity columns. c In the coupled spectrophotometric assay, recombinant human mevalonate kinase, expressed in bacteria and purified in a manner similar to that described in the current report, was added to the assay system to generate mevalonate 5-phosphate in situ. All NADH consumption due to recombinant mevalonate kinase (the result of ADP production) was subtracted in final calculation of enzyme specific activities. However, the DNA and predicted amino acid sequence of the yeast ERG8 gene has no significant homology to the cDNA which we isolated. Further, the deduced amino acid sequence encoded by the human liver PMKase cDNA does not contain any part of three conserved regions identified in mammalian MKase, bacterial and yeast galactokinase, and yeast PMKase (18). One of these conserved regions is believed to encode part of a conserved ATP binding sequence. Because of this apparent divergence from yeast to man, we sought further verification of the cDNA isolated through expression analysis in bacteria.
Several lines of evidence strongly support the identity of the isolated cDNA as one which encodes PMKase. The deduced M r of the protein encoded by the cDNA is 21,862, which is in good agreement with the M r of the purified porcine liver PMKase previously presented (12,13). The cDNA encoding PMKase was fused to a pGEX expression vector system with a bacterial host. With this system, the expression plasmid produces a hybrid protein of the cDNA insert fused to the C terminus of the glutathione S-transferase protein. The fusion protein provides a convenient system from which to isolate the hybrid protein using glutathione-Sepharose affinity chromatography. Two independent assay systems were employed to assess the identity of the enzyme activity encoded by the cDNA.
In the first assay system, the conversion of radiolabeled mevalonate 5-phosphate to mevalonate 5-diphosphate was monitored in clarified bacterial sonicates, with product separation achieved through thin layer chromatography and quantitation by liquid scintillation counting. A second assay system was spectrophotometric, based upon NADH consumption linked to the production of ADP in the PMKase reaction (3). Recombinant human MKase, expressed in bacteria in a fashion similar to that herein described for PMKase (results to be reported elsewhere) was included to produce mevalonate 5-phosphate in situ as substrate for the PMKase reaction (Table I). For the second assay system, it was necessary to isolate the PMKase fusion protein from the bacterial extract. Analysis of the glutathione S-transferase/PMKase fusion protein, isolated from bacterial sonicates using glutathione-Sepharose chromatography, by SDS-polyacrylamide gel electrophoresis, revealed a protein of M r approximately 48,000. The glutathione S-transferase protein has an estimated M r of 26,000, suggest-ing that the cDNA-derived protein had the expected M r of 22,000 (data not shown), in agreement with the deduced amino acid sequence (Fig. 1). Our protein and enzymatic data, therefore, supported the identity of the cDNA as that encoding human liver PMKase.
The human PMKase cDNA isolated in the current study does not contain the putative ATP binding amino acid sequence which has been detected in other kinases (18). This sequence, GXGXXGX 15-21 AXK (where X represents any amino acid), is detected in MKase but not PMKase. The recently cloned cDNA encoding mevalonate diphosphate decarboxylase (MDDase) from human and yeast also lacks this amino acid sequence (20), yet both PMKase and MDDase utilize ATP as cosubstrate. Although this putative ATP binding site sequence is not mandatory, it may provide insight into the regulation of the three ATP-dependent mevalonate catabolic enzymes. Geranyl and farnesyl diphosphates are potent competitive inhibitors of MKase at the ATP binding site. PMKase is not subject to similar competitive inhibition (4). The amino acid sequence of the MKase ATP binding pocket may enable selective inhibition of MKase by various pathway intermediates, which does not occur in PMKase and MDDase.
Our results provide the first evidence for regulation of mammalian PMKase gene expression at the level of transcription in response to sterol. Earlier studies of PMKase activity in the liver of animals fed a high cholesterol diet or diets containing cholesterol sequestering reagents, such as cholestyramine, were extended in the present study by demonstrating that PMKase messenger RNA increases upon removal of sterol from cultured cell growth medium and further increases when the medium is supplemented with the HMG-CoA reductase inhibitor lovastatin (Fig. 2B). It will be of interest to determine if a larger induction of transcription than reported here might be detected in the livers of animals maintained on different dietary regimens, supplemented with cholesterol or cholesterollowering agents. The data suggest that PMKase is subject to transcriptional regulation in a fashion similar to other cholesterol biosynthetic genes in terms of sterol responsiveness, consistent with the concept of coordinate pathway regulation.
It appears that many steps of cholesterol and nonsterol isoprenoid biosynthesis may be localized to the peroxisome, a mammalian organelle involved predominantly with fatty acid ␤-oxidation, plasmalogen biosynthesis, and respiration. Krisans et al. (21) have provided several elegant cell fractionation studies which indicate that the steps from acetoacetyl-CoA thiolase through farnesyl-diphosphate synthase occur within the peroxisome. These studies have relied upon both enzymatic analysis and immunoblotting with specific antisera made against the respective proteins. Recent work from this group provides strong evidence that MKase, PMKase, and MD-Dase are found in the peroxisome (22). The sterol carrier protein SCP-2 has been at least partially shown to associate with the peroxisome, and additional data reveal that dolichol synthesis from farnesyl diphosphate may occur in the peroxisome (23)(24)(25). Although the evidence that these enzymatic activities are peroxisomally associated is substantial, until the present report there has been no evidence to indicate the exact mechanisms by which these proteins may be imported into the peroxisome.
Subramani and co-workers have presented evidence concerning the mechanisms of protein import into peroxisomes (for a review, see Ref. 26). These investigators, and others, have defined peroxisomal targeting sequences at the C terminus (PTS-1; Ser-Lys-Leu and variants) and N terminus (PTS-2, (Arg/Lys)-(Leu/Val/Ile)-(X) 5 -(His/Gln)-(Leu/Ala)) of some peroxisomally localized proteins. Although these sequences are by themselves sufficient to facilitate peroxisome import, they are not mandatory elements, and other undefined targeting sequences and import mechanisms do exist. Of the cholesterol and nonsterol isoprenoid biosynthetic enzymes which have been reported to have peroxisomal locations, none have been shown to have PTS-1 or PTS-2 sequences until now. The PMKase sequence reported here is the first enzyme of the pathway shown to contain a consensus PTS-1 element, Ser-Arg-Leu, at the C terminus. Analysis of expressed sequence tags (ESTs) using the Blastn program revealed homology of our PMKase cDNA with several human, mouse, and rat ESTs. One mouse EST encoded the complete PMKase open reading frame and revealed the consensus PTS-1, Ala-Arg-Leu, at the C terminus of the deduced amino acid sequence, showing evolutionary conservation of the PTS-1 element in PMKase. Sequence analysis of the MKase polypeptide sequence reveals a consensus PTS-2 motif of Lys-Val-(X) 5 -His-Ala within the first 40 amino acids of the mature N terminus, making MKase potentially only the third protein known with a PTS-2 sequence, in addition to peroxisomal thiolase from various species and malate dehydrogenase from watermelon (27). Using computer analyses, the polypeptide sequences of other cloned cholesterol biosynthetic genes were screened, and it was found that human isopentenyl-diphosphate isomerase has a close match to the consensus PTS-2 motif with the sequence His-Leu-(X) 5 -Gln-Leu. Of the four enzymes which convert mevalonate to isopentenyl diphosphate, only MDDase appears to lack a consensus PTS-1 or PTS-2 element, although convincing evidence indicates that this protein is located in the peroxisome (22). Although these targeting sequences support the concept of a peroxisomal location for these enzymes, to our knowledge none of the sequences in these proteins have been experimentally mutagenized to verify their role in peroxisomal import.
A peroxisomal location of the cholesterol biosynthetic enzymes may provide further insights into the pathophysiology of the peroxisomal biogenesis disorders (28). It has long been known that patients with these disorders, including Zellweger syndrome, neonatal adrenoleukodystrophy, infantile Refsum disease, hyperpipecolic acidemia, and rhizomelic chondrodysplasia punctata, have decreased circulating levels of cholesterol. Cultured cells derived from these patients also invariably demonstrate peroxisome ghosts, in addition to decreased activities of several cholesterol biosynthetic activities (21,28). Identification of putative peroxisome targeting motifs in MKase, PMKase, and isopentenyl-diphosphate isomerase is consistent with decreased levels of circulating cholesterol in patients with peroxisomal biogenesis disorders. The isolation and character-ization of a cDNA encoding human PMKase is an important first step in understanding the regulation of PMKase gene expression and its role in the regulation of cholesterol and nonsterol isoprenoid biosynthesis.