Cloning and Characterization of PDK4 on 7q21.3 Encoding a Fourth Pyruvate Dehydrogenase Kinase Isoenzyme in Human*

,

order of glucose homeostasis with complex etiology involving environmental and genetic factors (1). To search for the genetic components, we are conducting a genome-wide scan of DNA microsatellite markers in the Pima Indians of Arizona, who have the highest known prevalence of NIDDM in the world. More than half of this population over the age of 35 years is affected (2,3), and prospective studies show that the onset of the disease in the Pima Indians is predicted by manifestations of impaired insulin action (insulin resistance; Ref. 4). Recently, we detected linkage and association of insulin resistance and NIDDM with a cluster of microsatellite markers at the cytogenetic band q21.3-q22.1 on chromosome 7 (5,6), and one of these markers (D7S479) was also linked with NIDDM in Caucasian families living in Utah. 2 Because these results are consistent with the presence of an NIDDM susceptibility locus at 7q21.3-q22.1, we have begun to search for potential candidate genes using a positional cloning approach.
Several genes have been identified and physically positioned within this region (7)(8)(9), including a novel transcription unit encoding a cDNA with more than 65% identity to the sequences of two isoforms of rat pyruvate dehydrogenase kinase, PDK1 (10) and PDK2 (11). PDKs are serine/threonine protein kinases that selectively inhibit the activity of pyruvate dehydrogenase (PDH), a mitochondrial multienzyme complex that catalyzes the first irreversible step in glucose oxidation (reviewed in Ref. 12). Rat PDK1 and PDK2 belong to a newly recognized family that is structurally related to prokaryotic histidine kinases (10,11). Other members of this family include the rat branched chain ␣-ketoacid dehydrogenase kinase (13), a hypothetical protein (ZK370.5) predicted from a genomic sequence of Caenorhabditis elegans (10,14), and three separate human PDK isoenzymes (PDK1, PDK2, and PDK3) identified most recently (15). All members of this family contain conserved motifs that presumably form the kinase domain (10,11,15). We here identify a new human PDK, and because three other isoforms were already reported (15), we have named this protein PDK4, and refer to the gene as PDK4.
Because PDKs are involved in the regulation of glucose oxidation (which could ultimately affect whole body glucose metabolism), and based on our previous linkage results, the PDK4 gene at 7q21.3 was an excellent candidate for further analyses with respect to insulin resistance and NIDDM in the Pima Indians. Here we report cloning and characterization of PDK4 and show that the corresponding recombinant protein phosphorylates and inactivates PDH. We also performed a comparative molecular analysis of the gene in insulin-sensitive and insulinresistant Pima Indians.

EXPERIMENTAL PROCEDURES
Subjects-The subjects are members of the Gila River Indian Community who have participated in longitudinal studies of NIDDM since 1965 (3,4). Measurements of clinical parameters of insulin action and classification of NIDDM (4), percutaneous biopsies of skeletal muscle (16) and adipose tissue (17), and preparation of genomic DNA samples (18) in this population have been described previously. Subjects selected for comparative studies of the gene and mRNA were not first degree relatives and were divided into an insulin-resistant and insulin-sensitive group according to their insulin-mediated glucose uptake rates determined by the two-step hyperinsulinemic euglycemic clamp (4) using criteria described in Ref. 19.
Polymerase Chain Reactions-Amplifications were performed in GeneAmp 9600 thermal cycler (Perkin-Elmer, Norwalk, CT) as described previously (20). To avoid amplification of nonspecific products, "hot start" PCR was performed by including TaqStart antibody (Clontech, Palo Alto, CA), or by adding the DNA polymerase to each tube after an initial denaturation step (20).
Direct cDNA Selection and Isolation of the Full-length cDNA Clone-The direct cDNA selection experiments were conducted as described previously (21). Briefly, four YAC clones (HSC7E1275, HSC7E1301, HSC7E1170, and HSC7E1180) that encompass approximately 3 Mb of DNA and 10 microsatellite markers (7cen-AFMa157tbl-AFMb283xd1-D7S527-D7S1812-D7S821-D7S2539-D7S479-D7S476-D7S491-D7S1 796-7qter) at 7q21.3-q22.1 (7) were used. The YAC DNA was purified away from the endogenous yeast chromosomes by pulsed-field gel electrophoresis, transferred to nylon membrane, and used as substrate for exhaustive hybridization with PCR-amplified cDNA pools made from RNA isolated from 10 tissues. A 1.7-kb cDNA clone (designated FC4 -1#1) was isolated from a frontal cortex cDNA library purchased from Stratagene (La Jolla, CA). Since the 5Ј end of the transcript was not represented in this clone, a two-step PCR protocol utilizing 5Ј-RACE-ready cDNA from skeletal muscle (Clontech) was used. Primary PCR was performed in a 25-l volume with 0.5 ng of the cDNA using antisense primer PDHK-13R (5Ј-CAGCTTTTAACCAATTGCACTG-3Ј) located in exon 2 in combination with the anchor primer provided with the cDNA (Clontech). Because we experienced difficulties in amplification of the 5Ј-UTR and of the first exon using standard PCR conditions (possibly due to the high GC content of this region), dGTP was reduced to 50 M, and 150 M of either 7-deaza-dGTP or dITP were added in separate PCR experiments to facilitate structure-independent amplification (22). Furthermore, AmpliTaq DNA polymerase was not included in the master mix, but was added to each tube after an initial denaturation at 98°C for 3 min. Cycling parameters included 30 cycles of 96°C for 20 s, 57°C for 30 s, and extension for 2 min (at 80°C for the first 5 cycles, and at 72°C for the remaining 25 cycles). A 0.5-l aliquot of the primary PCR was used for a secondary PCR performed under the same conditions with the nested antisense primer PDHK-2R (5Ј-GTAGCTGCTTCATGGA-CAGCG-3Ј) and the same anchor primer. Under both conditions (with 7-deaza-dGTP or dITP), we amplified a specific product as verified by sequencing (see below).
The remaining part of 3Ј-UTR was isolated by 3Ј-RACE using the Marathon cDNA amplification kit (Clontech) in combination with the Expand long template PCR system (Boehringer Mannheim). The sense primer PDHK-7 (5Ј-CCTCAGTTTTCCATCTGTTTTT-3Ј) from the 3Ј end of FC4 -1#1 was used in combination with the adaptor primer AP1 supplied with the Marathon kit.
The 5Ј-and 3Ј-RACE products were subcloned into pCRII vector using the TA cloning kit (Invitrogen, San Diego, CA), and sequenced with the DyeDeoxy terminator cycle sequencing kit or with the Sequenase terminator sequencing kit (both from Perkin Elmer), and run on the ABI automated sequencer model 373A (Perkin Elmer).
YAC and Cosmid Clones-The YAC clones were isolated from a chromosome 7-specific YAC library (23), and the cosmids were obtained from the chromosome 7-specific cosmid library from the Lawrence Livermore National Laboratories.
Determination of the Genomic Organization of PDK4 -Two cosmid clones (61d4, 106d6) were identified by screening of the cosmid library with FC4 -1#1. The exon-intron boundaries were determined by cycle sequencing with various 5Ј 32 P-end-labeled primers designed from the cDNA, using 1 g of cosmid DNA and the Life Technologies, Inc. cycle sequencing kit. The sizes of individual introns were estimated by agarose gel electrophoresis of PCR-amplified products, or determined by direct sequencing of short introns.
The 5Ј-flanking genomic sequence was determined from a ϳ6-kb XbaI-EcoRI fragment subcloned into pBLUESCRIPT SKII (Stratagene). This fragment encompasses the first two exons, plus approximately 4 kb of the 5Ј-flanking region as determined by Southern blotting, PCR, and partial restriction mapping.
Screening for Genomic DNA Variants-Individual exons (with the exception of exon 1, see below) and flanking splice junctions were screened for variants by comparative solid phase sequencing. Each exon was amplified from genomic DNA from 13 insulin-sensitive and 17 insulin-resistant Pima Indians with primers designed to include the coding sequence plus the flanking splice sites in the PCR product. One primer in each set was biotinylated at the 5Ј end (synthesized at Research Genetics, Huntsville, AL), and capture and sequencing of the biotinylated strand was performed with the solid phase Sequenase dye terminator DNA sequencing kit (Perkin Elmer) according to the manufacturer's recommendations, except that the post-priming washes were performed at 37-50°C to minimize nonspecific background. Samples were run on the ABI automated sequencer model 373A (Perkin Elmer). Information on primers used for amplification of individual exons is available upon request.
Genomic segments with a high GC content (promoter, 5Ј-UTR, and exon 1) that were difficult to scan by solid phase sequencing with Sequenase were analyzed by SSCP as described previously (20). PCR amplification of exon 1 was improved by including 10% dimethyl sulfoxide and doubling the concentration of AmpliTaq DNA polymerase in the reaction (24). PCR products showing mobility shifts on SSCP gels were sequenced directly, or after subcloning into the pCRII vector with the TA cloning kit (Invitrogen). In our experience, cycle sequencing of the GC-rich products using the DyeDeoxy terminator cycle sequencing kit (Perkin Elmer) and/or the Life Technologies, Inc. kit (with 5Ј 32 Pend-labeled primers) produced satisfactory results. Although no mobility shifts were observed in exon 1, PCR products of this segment amplified from two insulin-resistant and two insulin-sensitive Pima Indians were also cycle-sequenced to exclude the possibility that common variants that may be specific for either group remained undetected by the less sensitive SSCP technique.
Allelic expression of the two-base insertion/deletion polymorphism in the 5Ј-UTR was determined in randomly primed cDNA samples by an initial amplification with primer PDHK-prom (5Ј-GAATCTC-GAACCACTGCATCT) located immediately downstream from the transcription start site, combined with the antisense primers PDHK-12R (5Ј-GCGAGTCTCACAGGCAATTC-3Ј) in the second exon, or PDHK-13R (5Ј-CAGCTTTTAACCAATTGCACTG-3Ј) in the third exon. After 35 PCR cycles, 0.2 pmol of 5Ј 32 P-end-labeled antisense primer PDHK-1R (5Ј-CCTGGGCTGGGGTTTGAG-3Ј) were added to each sample, and one additional cycle (96°C for 20 s, 57°C for 30 s, 72°C for 5 min) was performed to synthesize a complementary strand encompassing the variant site. After denaturation, the singlestranded labeled products were resolved on a 6% sequencing gel to differentiate between the 73-and 75-base-long allelic forms. To assess the possibility of allele-specific differences in expression, the intensity of the labeled RT-PCR products was compared on the same gel with products amplified with PDHK-prom and PDHK-1R directly from genomic DNA from the same subjects including eight heterozygotes, as well as one homozygote for each allele.
Southern and Northern Blot Analyses-DNA samples from CEPH Caucasian subjects were obtained from the NIGMS repository (Coriell Institute, Camden, NJ), and non-human genomic DNAs (for the "zoo" blot) were obtained from Clontech. Southern blots were constructed with 10 g of EcoRI-digested genomic DNA, or with 50 ng of purified cosmid DNA transferred to a nylon membrane, and hybridized as described previously (25). The final post-hybridization wash of human DNA blots was at 65°C in 0.1 ϫ SSC/0.1% SDS, and the "zoo" blot was washed at 55°C in 0.5 ϫ SSC/0.1% SDS.
A human multiple tissue Northern blot containing approximately 2 g of poly(A) ϩ mRNA from heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas was obtained from Clontech. Hybridization with 32 P-labeled FC4 -1#1 insert was performed according to manufacturer's recommendations, and the final wash was at 65°C in 0.1 ϫ SSC/0.1% SDS. To confirm that all lanes contain comparable amounts of intact mRNA, the blot was stripped and rehybridized with a ␤-actin cDNA control probe provided by the manufacturer.
Cloning of PDK4 into the Baculovirus Transfer Vector-The cDNA clone FC4 -1#1 contains a 1748-bp insert with the initiating methionine and stop codon at positions 173 and 1406, respectively. By PCR methodology, a tag encoding for 6 histidine residues was incorporated at the 3Ј end of the open reading frame immediately before the stop codon. This modified cDNA (also lacking 5Ј-and 3Ј-UTR regions) was ligated into pBLUESCRIPT (Stratagene) and the amplified regions were sequenced to confirm that no errors were introduced. The cDNA was then ligated into the baculovirus transfer vector pVL1393 (Invitrogen) for transfection into SF9 cells. This construct is referred to as PDK4-his/pvl.
Transfection of Sf9 Cells-Sf9 cells were maintained in Grace's medium, pH 6.2 (Life Technologies, Inc.), supplemented with 3.3 g/liter TC yeastolate, 3.3 g/liter lactalbumin hydrolysate, 0.35 g/liter NaHCO 3 , 0.1% Pluronic F-68, and 10% fetal bovine serum. Cells were grown as a suspension culture at 27°C at 110 rpm, and passed twice weekly to a concentration of 0.5 ϫ 10 6 cells/ml. Methods for transfection of the cDNA and plaque purification of recombinant virus were essentially as described in Ref. 26. Briefly, 1 g of PDK4-his/pvl and 200 ng of linearized AcRP23lacZ (Pharmingen) were mixed with 5 l of Cellfectin (Life Technologies, Inc.) and transfected into Sf9 cells. Two days later, serial dilutions of the medium were applied to Sf9 cells for plaque assay. Plaques were visualized with an overlay of 0.5% SeaKem-agarose (FMC BioProducts) containing 0.05 mg/ml Neutral Red (Sigma) and 0.25 mg/ml 5-bromo-4-chloro-3-indoyl ␤-D-galactoside (Life Technologies, Inc.). One day later clear plaques were picked and used to infect Sf9 cells. When noticeable infection was present (4 -6 days), the cells were screened for PDK activity by incorporation of 32 P into recombinant E1␣ protein (see "PDK Assay" below). Virus-containing medium from cells that expressed PDK4 was plaque-purified as described in (26).
PDK Assay-Infected cells were lysed in 10 mM imidazole pH 7.2, 5 mM DTT, 0.1 mM EDTA, 10 Ϫ7 M pepstatin, 2 mM phenylmethylsulfonyl fluoride, and 2 g/ml leupeptin with 30 passes in a Thomas homogenizer. The homogenate was centrifuged at 16,000 ϫ g for 20 min, and the supernatant used in the kinase assay. The 50 l assay contained 20 mM Hepes pH 7.5, 50 mM KCl, 2 mM DTT, 0.1 mg/ml bovine serum albumin, 0.5 mM [␥-32 P]ATP (ϳ500 cpm/pmol), 5 mM MgCl 2 , 20 l of supernatant (5-10 g of protein), and 0.85-2.5 g of recombinant E1␣ (a gift from Dr. M. S. Patel, State University of New York, Buffalo). The reaction was assayed for 10 min at room temperature and stopped with 10 l of 5 ϫ SDS buffer. Proteins were resolved by 10% SDS-PAGE and exposed to X-Omat film at Ϫ70°C. In some instances, the band corresponding to E1␣ was excised from the gel and counted in 2 ml of Ecolume scintillation solution (ICN Biomedicals).
Purification of PDK4 -Sf9 cells (5 ϫ 10 7 ) were infected with recombinant PDK4 virus (multiplicity of infection ϳ1) for 3 days. The cells were pelleted and lysed in 8 ml of 60 mM imidazole, pH 7.2, 0.3 M NaCl, 5 mM ␤-mercaptoethanol, 0.2% Triton X-100, 2 mM phenylmethylsulfonyl fluoride, 10 Ϫ7 M pepstatin, and 2 g/ml leupeptin by 30 passes in a Dounce homogenizer. The lysate was centrifuged at 100,000 ϫ g for 10 min, and the resulting supernatant was batch applied with gentle rocking to 0.5 ml of a 50% slurry of Ni-NTA resin (Qiagen) pre-equilibrated in lysis buffer. After 1 h, the resin mixture was applied to a 1-cm diameter column, washed with 1.5 ml each of 1 M NaCl, 0.5% Tween 20, and 15% glycerol, and 0.5 ml of 100 mM imidazole pH 7.2, 0.1% Triton X-100, and eluted with 0.5 ml of 500 mM imidazole, 0.1% Triton X-100. DTT was immediately added to the eluted fraction to a final concentration of 2 mM. The various fractions were assayed for PDK activity as described above. Proteins in each fraction were quantitated using the Bio-Rad protein assay, resolved by SDS-PAGE, and stained with silver.
ATP-dependent Inactivation of PDH Complex-PDK4 was tested for its ability to inhibit PDH complex utilizing the two-step assay as described in Ref. 11 using 22.5 milliunits of kinase-depleted PDH complex (27).
Data Analyses-Sequences were analyzed with programs available in the Genetics Computer Group package version 8 (GCG, Madison, WI), or with the Sequencher 2.1 program (Gene Codes Corp., Ann Arbor, MI). The significance in frequency differences of PDK4 genomic variants was assessed by chi-square analysis of multiple contingency tables.

Cloning and Characterization of a Full-length cDNA-
The initial PDK4 cDNA clone (E1301cd24) was isolated by the technique of direct selection (28,29) from YAC clone HSC7E1301 (7), which encompasses approximately 1 Mb of DNA in the 7q21.3 region. The gene was positioned between PON and D7S527 immediately adjacent to the microsatellite marker AFMb283xd1. The order of DNA markers and genes in this region has been described previously (7,9). The E1301cd24 fragment (0.4 kb) was used to screen a frontal cortex cDNA library, and a 1.7-kb clone (FC4 -1#1) was isolated. The se- Data base searches detected a significant similarity of the predicted protein with rat PDK1 (65% identity/81% similarity allowing for conservative substitutions), rat PDK2 (65%/82%), and with the hypothetical protein ZK370.5 from C. elegans (48%/65%). The deduced protein described here, PDK4, has 62%-66% identity/79%-82% similarity with the three human PDK isoforms, and we performed multiple alignment of all four proteins to determine shared regions of homologies. As shown in Fig. 2, PDK4 contains all motifs characteristic for the putative kinase domain of the eukaryotic mitochondrial kinase family as defined previously (11,15), including subdomain II characterized by an invariant asparagine (sequence KNXMRAT), subdomain IV defined by an invariant tyrosine (Tyr 309 in PDK4), and the glycine-rich subdomains III (DXGGG) and V (GFGYGLP). Subdomain I, characterized by an invariant histidine (11,15), is represented by His 127 in PDK4. Additional conserved patterns in Fig. 2 include a segment of 29 amino acids with 78% identity/86% similarity between all four human proteins (residues 31-67 in PDK4), and several shorter motifs (e.g. PDK4 residues 130 -137, 156 -176, 343-352, and 357-368) that are found at comparable positions in other mitochondrial protein kinases (15).
To determine the tissue distribution of PDK4 mRNA, clone FC4 -1#1 was used to probe a Northern blot containing poly(A) ϩ RNA from eight human tissues. As shown in Fig. 3, a single band approximately 4 kb long was detected in all lanes, and the strongest signal was observed in heart and skeletal muscle. Because the 1.7-kb clone did not represent the fulllength transcript, we isolated the corresponding ends by 5Јand 3Ј-RACE PCR. All four sequenced 5Ј-RACE clones (see "Experimental Procedures") begin at the same adenosine residue, that we have designated as the putative transcription start site ϩ1 in Fig. 1 and Fig. 5. By sequencing of a 3Ј-RACE product, we found that the poly(A) tail is preceded by the typical polyadenylation signal AATAAA (Fig. 1).
To determine if other species have genomic sequences homologous to PDK4, the FC4 -1#1 probe was hybridized to a Southern blot of EcoRI-digested genomic DNA from human, rhesus monkey, dog, pig, cow, mouse, rat, and chicken, and washed under moderately stringent conditions (see "Experimental Procedures"). Four bands (ϳ9.3, ϳ7.5, ϳ4.2, and ϳ1 kb) were detected in human DNA, and three to five distinct bands were seen in the remaining species (data not shown).
Exon-Intron Structure and Promoter Sequence of PDK4 -The exon-intron organization was determined by direct sequencing and PCR analysis with exon-specific primers of cosmid clones 61d4 or 106d6 containing the entire locus. PDK4 consists of 11 exons, and the sizes of individual introns range between 0.2 and 2.5 kb (Fig. 4). The entire locus spans approximately 16 kb, and the sequences of all exon-intron junctions are in agreement with the consensus GT-AG motifs for splice donor and acceptor sites in eukaryotic genes (30). The ZK370.5 gene from C. elegans, consisting of seven exons and six introns (14), is the only other member of this kinase family for which the genomic structure is known. When the protein and cDNA sequences of ZK370.5 and PDK4 were aligned, four of the six exon junctions in the cDNA from the C. elegans gene (corre- sponding to intron number 1, 2, 4, and 5) are located at analogous sites in PDK4 (intron 3, 4, 7, and 8), consistent with a conservation of their genomic structures (data not shown).
To investigate the promoter region, we have sequenced approximately 4 kb of the 5Ј region preceding the start of the coding sequence. Although the gene does not have a typical TATA box, a TATA-like motif (AAATAAAA) was found beginning at position Ϫ26 from the putative transcription start site. Other predicted transcriptional regulatory elements (31) include an inverted CCAAT box at Ϫ114, four putative SP1binding sites (GGGCGG) at Ϫ16, Ϫ143, Ϫ338, and Ϫ561, and an AP2 site (GGCAGCCC) at Ϫ631. As shown in Fig. 5, all of these elements are clustered within a ϳ640-bp segment immediately upstream from the transcription initiation site.
Screening for DNA Variants in PDK4 -To search for large variants or rearrangements in the locus, Southern blots of EcoRI-digested genomic DNA from 52 Pima Indians (23 insulin-sensitive and 29 insulin-resistant), and from a Caucasian subject were hybridized with the FC4 -1#1 clone. Using stringent washing conditions, four invariant bands were observed in all subjects, consistent with the absence of gross structural alterations of the gene in the Pima Indians (data not shown).
PCR-amplified genomic segments from 30 subjects were screened for small (e.g. single-base) variants as described under "Experimental Procedures." The only difference detected in the coding sequences was a silent substitution (GCA 3 GCC) at the Ala 341 codon in exon 10, with a nearly identical frequency in the insulin-resistant and insulin-sensitive groups (0.68 and 0.69, respectively; p ϭ 0.88).
The promoter and 5Ј-UTR were analyzed in three segments to include the transcription start site, and the recognized putative transcription factor binding sites described above. As shown in Fig. 5, we detected two variants in the 5Ј-UTR, including a two-base (C/CCC) polymorphism (presumably an insertion or deletion) at position ϩ38 (referred to as C ϩ38 and CCC ϩ38 , respectively). The frequency of the CCC ϩ38 allele was 0.28 in the insulin-resistant group, and 0.31 in the insulinsensitive group (p ϭ 0.64). The CCC ϩ38 variant was also found at a comparable frequency (0.28) in 9 CEPH Caucasian subjects. A less common A 3 G substitution was detected at position ϩ15 with a frequency 0.11 in the insulin-resistant, and 0.06 in the insulin-sensitive group (p ϭ 0.2).
Two single-base substitutions were also detected in the promoter, including a T 3 C substitution at Ϫ153 located 5 bp upstream from the second SP1 site, and a C 3 T substitution at Ϫ208 (Fig. 5). Both variants create a RFLP at an MspI site (recognition sequence ϭ CCGG), including a loss of the restriction site at Ϫ208 (CCGG 3 CTGG; designated site 1), and a gain of the restriction site at Ϫ153 (CTGG 3 CCGG; designated site 2). Because of their close proximity, both variants were typed simultaneously on a single PCR product (see "Experimental Procedures"). We observed three combinations (haplotypes) of the variants at sites 1 and 2 in the Pima Indi- ans: haplotype A containing MspI site 1 and lacking site 2 (identical with the cloned genomic sequence in Fig. 5), haplotype B containing both MspI sites, and haplotype C lacking both MspI sites. The respective frequencies of haplotypes A, B, and C were 0.60/0.31/0.09 in the insulin-resistant subjects, and 0.59/0.36/0.05 in the insulin-sensitive subjects, and the differences between both groups were not statistically significant (p ϭ 0.47).
Quantitative Analysis of PDK4 mRNA Level in Skeletal Muscle and Adipose Tissue-To test if any of the variants in the promoter or 5Ј-UTR were affecting PDK4 mRNA level, we used quantitative RT-PCR analysis of cDNA samples prepared from skeletal muscle biopsies from 36 non-diabetic Pima Indians, and from subcutaneous fat biopsies available from 22 of these individuals. No correlation was found between any of these polymorphisms and differences in mRNA levels in either tis-sue. To test whether the structural difference between the C ϩ38 and CCC ϩ38 variants could influence the relative level of either allelic form of the transcript (e.g. by affecting their transcription rates differently), we compared both forms by a two-step RT-PCR (described under "Experimental Procedures") using skeletal muscle cDNAs from eight C ϩ38 /CCC ϩ38 heterozygous subjects. Direct comparison between both RT-PCR products in each heterozygote, and comparisons with the patterns obtained by amplification of this region from genomic DNA from the same individuals did not indicate significant differences in the relative level of either allelic form of the transcript.
Expression, Purification, and Characterization of Recombinant PDK4 -Histidine-tagged PDK4 expressed in Sf9 cells was purified by nickel affinity chromatography using recombinant E1␣ as substrate. Fig. 6A shows the results of PDK assay of a typical purification of the recombinant protein. PDK activity was present in the supernatant but not in the flow-through from the Ni-NTA resin (lanes 1 and 2), indicating that all of the histidine-tagged PDK4 bound to the resin. No activity was present in any of the washes, except for a small amount that was usually seen in the 100 mM imidazole wash. Essentially all of the activity eluted in the 500 mM imidazole fraction (lanes 7-9), as further washing with 1 M imidazole eluted very little activity (data not shown). Fig. 6B shows a silver stained gel of the purified protein. As can be seen in lane 4, PDK4 was purified to near homogeneity, migrating as a single band of 47.5 kDa, consistent with its predicted size (ϳ46.5 kDa encoded by the cDNA, plus 0.93 kDa from the histidine tag). There were two minor contaminating bands at ϳ58.5 and 30 kDa, and we estimate that PDK4 is at least 80% pure based on the silver staining pattern. The specific activity of PDK4 toward 22.5 milliunits of PDH was 8.4 nmol/min⅐mg. The purified PDK4 was also able to inhibit the activity of PDH complex. Fig. 7 shows a time course of residual PDH activity after incubation with 1 g of PDK4. The PDH activity decreased with time of incubation with PDK4, and a complete inhibition was reached at about 2 min. Similar inhibition of PDH activity was observed using 50 ng of PDK4, although the rate and maximal inhibition were less (data not shown). There Kinase-depleted PDH was incubated with 1 g of PDK4 for 0.5, 1, 2, and 4 min, and the residual PDH activity measured as described under "Experimental Procedures." Values on the left axis are expressed as % maximal activity, which is defined as the amount of PDH activity measured in the absence of PDK4. The right axis shows the amount of pmol of 32 P incorporated into the E1␣ subunit of the PDH complex over the same time course. was no inhibition of PDH activity if ATP was omitted from the reaction (data not shown). To determine if PDK4 phosphorylated the E1␣ subunit of the complex, the phosphorylation part of the two-step assay was repeated using [␥-32 P]ATP. The radioactive band migrating at ϳ46.5 kDa corresponding to the E1␣ subunit was excised and counted. The results, plotted on the right axis of Fig. 7, indicate a time-dependent increase in the phosphorylation of the E1␣ subunit that closely coincides with the time course of the inhibition of PDH activity. DISCUSSION Pyruvate dehydrogenase kinase is important for the regulation of mitochondrial PDH activity which plays a key role in the oxidative metabolism of glucose. Recent studies have described multiple, genetically distinct mammalian PDK isoenzymes, including two forms in rat (10,11), and three in human (15). Here we describe a fourth human isoform (PDK4) encoded by a gene located at 7q21.3. Our evidence that this protein is an authentic PDK includes: 1) a high degree of identity with all known mammalian PDK isoforms, 2) the ability to phosphorylate the E1␣ subunit of PDH, and 3) the ATP-dependent inactivation of the PDH complex. The PDK4 transcript was present in all tissues examined, and the highest level was detected in heart and skeletal muscle, as was also found for other PDK isoforms (15). Assuming that the protein level corresponds to mRNA, we expect that the amount of PDK4 varies considerably in different tissues. Development of PDK4-specific antibodies will help to address this issue, and also to determine what proportion of the total PDK protein is contributed by this isoform.
Our Southern blot data indicate that the PDK4 cDNA probe does not cross-hybridize with other members of the PDK family in human, and we predict that the simple distinct hybridization patterns detected on the "zoo" blot correspond to conserved homologues of PDK4 in other species. In addition, the similarity of the exon-intron organization between PDK4 and the ZK370.5 gene in C. elegans indicates an evolutionary conservation of their genomic structure.
The PDK4 gene was identified by positional cloning in a region on 7q that is linked with insulin-resistance and NIDDM in the Pima Indians. Based on comparative analyses of the genomic sequences and of the transcript, we conclude that alterations in this gene are unlikely the underlying basis for the linkage of the 7q21.3-q22.1 region with insulin resistance and NIDDM in the Pima Indians. Therefore, other genes in this chromosomal interval need to be investigated as potential candidates. Identification of a fourth PDK isoenzyme in human, and knowledge of the organization of the gene and of its promoter sequence provide new information that should facilitate molecular genetic studies of other members of this mitochondrial kinase family, and of their function in the regulation of glucose metabolism.