A Broad Role for the Zinc Finger Protein ZNF202 in Human Lipid Metabolism*

The ZNF202 gene resides in a chromosomal region linked genetically to low high density lipoprotein cholesterol in Utah families. Here we show that the ZNF202 gene product is a transcriptional repressor that binds to elements found predominantly in genes that participate in lipid metabolism. Among its targets are structural components of lipoprotein particles (apolipoproteins AIV, CIII, and E), enzymes involved in lipid processing (lipoprotein lipase, lecithin cholesteryl ester transferase), and several genes involved in processes related to energy metabolism and vascular disease. Based on the linkage and apparent transcriptional function of ZNF202, we propose that ZNF202 is a candidate susceptibility gene for human dyslipidemia.

The ZNF202 gene resides in a chromosomal region linked genetically to low high density lipoprotein cholesterol in Utah families. Here we show that the ZNF202 gene product is a transcriptional repressor that binds to elements found predominantly in genes that participate in lipid metabolism. Among its targets are structural components of lipoprotein particles (apolipoproteins AIV, CIII, and E), enzymes involved in lipid processing (lipoprotein lipase, lecithin cholesteryl ester transferase), and several genes involved in processes related to energy metabolism and vascular disease. Based on the linkage and apparent transcriptional function of ZNF202, we propose that ZNF202 is a candidate susceptibility gene for human dyslipidemia.
Familial hypoalphalipoproteinemia (HA), 1 the most common form of decreased plasma HDL levels, is an independent risk factor for early coronary disease (1). HDL contains apolipoproteins AI and AII as its major protein components and various amounts of triglycerides, phospholipids, and cholesterol esters.
Its primary function appears to be the "reverse transport" of cholesterol from peripheral tissues to the liver, where it is catabolized into bile acids. A second proposed function for HDL is the uptake of apolipoproteins and free cholesterol from catabolized very low density lipoproteins and chylomicrons.
Phenotypic variation of plasma HDL levels depends on both genetic and environmental factors such as exercise. About 50% of HDL variation is ascribed to genetic influences (2). A number of rare mutations have been described that result in severely depressed HDL levels, among them mutations in the genes encoding apolipoprotein AI, apolipoprotein B, lecithin-cholesterol acyltransferase, and lipoprotein lipase (LPL) (3). Recently, ATP binding cassette transporter-1 was identified as the gene underlying Tangier disease, an extremely rare form of HA (4 -7). Although these rare deficiencies account only for a fraction of individuals with lipid abnormalities, their existence suggests that less severe dyslipidemic phenotypes may be due more subtle disruptions in the expression and function of genes that participate in lipid metabolism.
In an effort to discern other genetic contributors to HA, we performed linkage analysis on large Utah pedigrees that have heritable HA and a family history of early coronary disease. A locus on human chromosome 11q23, clearly distinct from the apoA1/CIII/AIV gene cluster, is linked to HA in many Utah families. 2 One of the genes localized to this region is the zinc finger transcription factor ZNF202 (9). In light of the increasingly recognized role of transcription factors in coordinating the transcription of multiple genes in lipid-related pathways (10,11), we characterized the biochemical function of ZNF202. This analysis reveals that ZNF202 is a transcriptional repressor that binds to the regulatory region of many genes involved in lipid metabolism. ZNF202 thus joins a growing number of transcription factors acting in metabolic coordination.

EXPERIMENTAL PROCEDURES
Cell Culture and Antibodies-HepG2 and HEK293 cells were obtained from ATCC (HB-8065 and CRL-1573, respectively) and grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. A glutathione S-transferase (GST) fusion protein of amino acids 1-199 of ZNF202 was purified from bacterial extracts and used to generate polyclonal antibodies in rabbits. Antisera were affinity-purified on a GST-ZNF202 column.
In Vitro Protein Expression-cDNAs representing the alternatively spliced mRNAs of ZNF202 were cloned into pcDNA3.1 (Invitrogen). Proteins were synthesized and [ 35 S]methionine-labeled in vitro using a coupled transcription/translation system (Promega).
Affinity Purification of GST-ZNF202-A GST fusion with ZNF202 zinc fingers 3 through 8 (GST.ZF3.8) was generated by inserting amino acids 474 to 648 of ZNF202 into pGEX-4T-3 (Amersham Pharmacia Biotech). GST.ZF3-8 protein was expressed in BL21 cells and purified essentially as described by the manufacturer. For gel shift assays, protein was eluted from glutathione-Sepharose and dialyzed to remove residual glutathione. Protein concentration was estimated from Coomassie Blue-stained SDS-polyacrylamide gels.
Affinity Selection of Binding Sites-An affinity selection procedure was modified from a method described previously (12). An oligonucleotide designated INVSL2n36 (5Ј-ACC CGA ATT CGG ATC C (N) 36 CG GAA TTC CG-3Ј) was synthesized with a 36-nucleotide random sequence between defined sequences at the 5Ј and 3Ј ends. GST.ZF3.8 was immobilized on GSH-agarose beads (Amersham Pharmacia Biotech) and pre-blocked with 200 ng/ml poly(dIC) for 30 min in 1ϫ GSA buffer (25 mM Hepes pH 7.5, 10 M ZnCl 2 , 10% glycerol, 5 mM dithiothreitol, 5 mM MgCl 2 , 50 mM KCl, 0.05% Triton X-100). This affinity matrix was used to select specific sequences from the pool of random oligomers. Oligonucleotides were added to the GST-ZNF202 agarose beads for 30 min at 22°C in 1ϫ GSA buffer. Unspecifically retained oligomers were * 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.
§ To whom correspondence should be addressed: Myriad Genetics, Inc., 320 Wakara Way, Salt Lake City, UT 84108. Tel.: 801-584-3724; Fax: 801-584-3650; E-mail: swagner@myriad.com. 1 The abbreviations used are: HA, hypoalphalipoproteinemia; HDL, high density lipoprotein; apo, apolipoprotein; GST, glutathione S-transferase; KRAB, krueppel-associated box; LPL, lipoprotein lipase; HNF4 hepatic nuclear factor 4; nt, nucleotides; PAGE, polyacrylamide gel electrophoresis; SEAP, secreted alkaline phosphatase; EMSA, electromobility shift assay. removed by stepwise washes with 3 bed volumes of 1ϫ GSA buffer containing increasing amounts of NaCl (0.1 M and 0.2 M). In the first round, a low and a high stringency-bound fraction were obtained by collecting sequential elutions at increasing salt concentrations. The low salt-bound fraction was obtained by eluting with 0.3 M NaCl. After two more washes at 0.4 M and 0.5 M, a high salt fraction was collected from an elution with 0.6 M NaCl. Both pools were submitted to additional cycles. Subsequent elutions were performed with 0.3 M for the low stringency pool and 0.6 M for the high stringency fraction. From each of the affinity purification rounds, the oligonucleotide pools were amplified for 22-24 cycles with the primers 5Ј-ACC GAA TTC GGA TCC-3Ј and 5Ј-GTG CCT GCT GAA TTC CTA-3Ј in the following PCR protocol: 96°C for 4 s, 62°C for 10 s, and 72°C for 30 s. After the fifth iteration of selection, the pools of selected oligonucleotides were cloned into pBluescript for sequence analysis.
Gel Mobility Shift Assays-Gel mobility shift assays were carried out using purified bacterially expressed GST.ZF3.8 protein or in vitro transcribed and translated ZNF202 (amino acids 304 -648). Gel-purified promoter fragments were end-labeled with polynucleotide kinase and Bioinformatics Analysis-GenBank release 104.0 was screened for accessions whose annotations included the key words "exon1," "promoter," "regulator," or "upstream." More than 7300 human accessions presumed to contain 5Ј upstream regions and annotated promoters were isolated. These accessions were screened using the BLAST algorithm (13) with the binding sequences from the apolipoprotein AIV (apoAIV) (TTGGTGGGGTGGGGGGTGGGGGGTG), apolipoprotein CIII (apoC-III) (GGGTGGGGGCGGGTGGGGGG), and LPL (GGGGGTGGGGAT-GGGGTGCGGGGT) regulatory regions as queries.
Reporter Assays-A promoter fragment from the apoAIVgene (Ϫ710 to ϩ10) was cloned into the pSEAP2-enhancer vector, and a fragment of the apoE gene (Ϫ322 to ϩ18) was inserted into pSEAP2-basic (CLON-TECH). 2.0 g of reporter plasmid and the indicated amounts of a ZNF202m1 expression plasmid were cotransfected into HepG2 cells using LipofectAMINEPlus (Life Technologies, Inc.) according to the manufacturer's recommendations. 100 ng of pCMV␤ was added as an internal control. The total amount of DNA was kept constant by supplementing with the parent expression vector pcDNA3.1. Supernatant was collected at 48 h post-transfection, and alkaline phosphatase activity was measured by chemilumiscent detection (Great EscApe chemilumiscence detection kit, CLONTECH) per manufacturer's procedures. Results are expressed as percent of transcriptional activity relative to transcription in the absence of ZNF202m1. Each point represents the average of 3 independent experiments.
Cell Extracts and Western Blotting-HEK293 were transfected with 4 g of expression plasmid with LipofectAMINE as described earlier.
After 48 h, cells were harvested, and total cell extract was prepared by lysis in radioimmunoprecipitation assay on ice. Approximately 20 g of extract was subjected to SDS-PAGE and Western blot with affinitypurified anti-GST-ZNF202 antibodies according to standard procedures.
Northern Blot Analysis-Labeled ZNF202 cDNA probes were prepared using [␣-32 P]dCTP (Amersham Pharmacia Biotech) and a random prime labeling kit (Promega). Two multi-tissue blots (Human and Hu-manII) and a human cancer cell line blot (CLONTECH) were used for hybridization in Quickhyb solution (Stratagene) at 65°C and visualized by autoradiography. The blots were then stripped of radioactivity and reprobed with a 32 P random prime-labeled glycerol-3-phosphate dehydrogenase cDNA probe (CLONTECH) to confirm equal loading.

RESULTS
Structural Features of ZNF202-The ZNF202 gene contains 10 exons spanning 27 kilobases of chromosome 11. It was independently cloned as a candidate gene for predisposition to breast and lung cancer (9). Cloning of ZNF202 cDNAs (Gen-Bank accession numbers AF027219 and AF027218) revealed a number of alternative splices in the 5Ј-non-coding region and two common splice variants in the coding region named ZNF202m1 and ZNF202m3 (Fig. 1A). The full-length m1 form (ZNF202m1) encodes a predicted 648 amino acid protein with several conserved domains. The C terminus encodes eight consecutive C2H2 zinc finger motifs; the first, second, seventh, and eighth C2H2 motifs show significantly higher similarity to C2H2 fingers known to bind DNA than do fingers three through six (14). The central part of the protein is occupied by a Krueppel-associated box (KRAB). KRAB domains, present in one-third of all zinc finger proteins, mediate transcriptional repression through mobilization of a corepressor, for example KAP-1 (15)(16)(17). The N terminus contains a SCAN (or LeR) motif, a suggested protein interaction module, observed in other zinc finger proteins (18). Further analysis of the protein sequence by PESTFIND (19) reveals extensive PEST sequences that suggest a rapid protein turnover, similar to that observed for several other transcription factors. The conceptual translation (GenBank accession AF027218) of the second common splice form, m3 (ZNF202m3), predicted a N-terminal-truncated protein that retains the KRAB domain and the eight C-terminal zinc finger motifs. However, expression of a ZNF202 m3 cDNA revealed that the m3 form encodes a truncated ZNF202 protein of 142 amino acids encompassing only the SCAN domain. Expression of the m1 and m3 gene products was demonstrated by both coupled transcription/translation experiments and transient transfections in HEK293 cells followed by SDS-PAGE and immunoblots (Figs. 1, B and C). If, as has been suggested (18), SCAN domains mediate dimerization, the m1 and m3 gene products may interact with each other and potentially with other SCAN domain-containing proteins.
Northern blots containing a wide range of tissues and several cell lines were probed with ZNF202 cNDA to determine its expression pattern. Maximal expression was detected in heart, lung, liver, and testis (Fig. 2).
ZNF202 Binds Specific Promoter Elements-The protein motifs observed in the ZNF202 m1 gene product imply that it functions as a DNA-binding protein and as a transcriptional regulator. Fluorescent microscopy of HEK293 cells expressing green fluorescent protein-ZNF202m1 fusions reveals a strong nuclear fluorescence (data not shown), as expected for a transcription factor. To define a DNA binding site for ZNF202m1, we employed a GST fusion protein containing the last six zinc fingers of ZNF202 m1(GST.ZF3-8) to select a consensus binding site from a complex mixture of random oligonucleotides (12). After five rounds of selection and amplification, the bound fraction was cloned and sequenced. Analysis of 55 clones revealed a simple consensus binding motif (5Ј-GGGGT-3Ј), as shown in Fig. 3A. The rules of Choo and Klug (20) predict essentially the same binding specificity for these zinc fingers. Direct repeats of the selected motif were found in the apoAIV and LPL promoters and in the apoCIII enhancer (Fig. 3B); this enhancer affects the expression of the entire apoA1/CIII/AIV gene cluster. To confirm that GST.ZF3-8 binds the predicted elements, we performed electromobility shift assays with promoter fragments of apoAIV and apoCIII. As shown in Fig. 4, A  and B, promoter fragments that include the presumed ZNF202 site were bound by GST.ZF3-8. Subdivision of the apoCIII and apoAIV probes into smaller fragments revealed that fragments containing the consensus repeat were bound by GST.ZF3-8, whereas adjacent fragments lacking the consensus site were not bound. Also, several non-related DNA fragments derived from the promoters of heat shock protein 90 and glycerol-3phosphate hydrogenase were negative for binding by GST.ZF3-8. Methylation interference experiments indicated that GST.ZF3-8 binds to the predicted sequence elements (data not shown). To demonstrate binding specificity, we performed a competition assay. Fig. 4C shows binding of ZNF202 to a fragment of the ␤3-adrenergic receptor promoter containing a predicted ZNF202 binding site. The addition of a 50ϫ excess of the cold ␤3-adrenergic receptor promoter fragment effectively competed for binding to the labeled ␤3-adrenergic receptor probe. A negative control fragment of the neurofibromin promoter did not compete under identical conditions. Protein dilution experiments suggested the binding of GST.ZF3-8 to the apoAIV DNA fragment has an apparent K d of approximately 10 nM, an upper limit value that assumes that all purified GST.ZF3-8 is in an active conformation. This binding constant is similar to the apparent K d described for other zinc finger proteins (21). We thus conclude that binding of GST.ZF3-8 to these promoter fragments is sequence-specific.
To identify other possible target promoters, we generated a custom unbiased data base of known human promoters from GenBank entries through a keyword search. These sequences were searched with the consensus binding motifs found in the apoAIV and LPL promoters or in the apoCIII enhancer, which had been shown to bind ZNF202 zinc fingers in vitro. Remarkably, the motif appeared in only a small number of promoters (Table I). Moreover, the majority of the genes identified are known to be involved in lipid metabolism or to be associated with metabolic disorders. We amplified most of these putative elements from genomic DNA and used them as probes in gel shift experiments to confirm ZNF202 binding. These results are summarized in Table I, and an alignment of the binding sites is displayed in Fig. 5. Notably, several of our binding elements coincide with sites previously mapped as DNA-protein interaction sites (11). For example, the ZNF202 site in the apoCIII promoter overlaps with an element at Ϫ611 to Ϫ592 Ϫ103 to ϩ120 relative to the start of transcription) or a control fragment from the promoter of neurofibromin (NF1*, Ϫ361 to Ϫ194). Binding assays were run in the absence (lanes 1, 2, 5, and 6) or presence of 50ϫ cold specific competitor (␤-3 adrenergic receptor (␤3AR), lanes 3 and 7) or 50ϫ nonspecific competitor (NF1*, lanes 4 and 8).

FIG. 4. ZNF202 binds the apoCIII enhancer and apoAIV promoter.
A, gel shift analysis was performed as described under "Experimental Procedures," with 32 P-end-labeled fragments (S0, S1, S2, S3) of the apoCIII enhancer and bacterially expressed GST.ZF3.8. Positions relative to the transcription start site are given for each fragment. The ZNF202 binding site depicted in Fig. 3B is denoted by GnT. The barred regions labeled I, H, and F indicate footprinted regions as described in Kardassis et al. (11) that contain elements related to GnT. Fragments marked ϩ are shifted by GST.ZF3-8. B, as in A, except that fragments of the apoAIV promoter were used in the gel shift analysis. C, sequencespecific binding of ZNF202. Gel shifts were performed using in vitro translated ZNF202m1 (amino acids 304 -648) and a fragment of the ␤3-adrenergic receptor promoter containing a ZNF202 site (␤3AR), The informatics search for ZNF202 binding sites in promoters was performed as follows. Twenty-one promoters were identified with homologous sequences in their promoters; 17 of these were tested in EMSAs and were confirmed to bind GST.ZF3-8. Three additional promoters (apoE, lecithin-cholesterol acyltransferase (LCAT), and HNF4␣) were identified in less stringent searches and were confirmed in a similar manner, thus yielding the 20 EMSA-positive promoters shown in the table. Additionally, several control promoter fragments lacking sequences similar to the consensus site (heat shock protein 90, glycerol-3-phosphate dehydrogenase, and NF1*) were employed in EMSAs and were found not to bind ZNF202m1 derivatives. Results from transient transfection assays in HepG2 cells are given in the column labeled "Repression Function." The ranges of the DNA fragments used are given under "Experimental Procedures." PLTP, phospholipid transfer protein; HTGL, hepatic triglyceride lipase; VEGF, vascular endothelial growth factor; IA-1, insulinoma-associated gene 1; ␤3AR, ␤3:adrenergic receptor; CRABP2, cellular retinoic acid-binding protein type II; CALRT1, calretinin; PNMTA, phosphatidyl ethanol amine N-methyl transferase; PLP, phospholipid protein. relative to the transcription initiation site (Fig. 4a); this element is thought to bind Sp1 and an unidentified factor CIIIJ1 from rat liver (11), and the rat ortholog to ZNF202 may be that factor. ZNF202 Transcriptional Activity-Having established that ZNF202m1 zinc fingers bind DNA in a sequence-specific manner, we next analyzed the effect of full-length ZNF202m1 on transcription from target promoters. The presence of a KRAB domain suggested that the ZNF202m1 gene product acts as a transcriptional repressor (15)(16)(17). To test this hypothesis, we used several reporter constructs containing portions of the apoE and apoAIV genes; these fragments all had been shown to bind ZNF202 in gel shift experiments. The reporter constructs were introduced by transfection into HepG2 human hepatoma cells with varying amounts of a ZNF202m1 expression plasmid. Reporter gene expression was measured in two independent systems by monitoring either secreted alkaline phosphatase (SEAP) or luciferase activity. As shown in Fig. 6, co-expression of ZNF202 represses transcription from the apoAIV and apoE reporter constructs 5-10-fold relative to the transcriptional activity in the absence of ZNF202. This repression is dose-dependent. No effect was observed for a control plasmid lacking a ZNF202 binding site. Identical results were obtained using a luciferase reporter (data not shown). Additional fragments derived from the promoters of phospholipid transfer protein and hepatic nuclear factor-4 (HNF4) and the apoCIII enhancer that had been shown to bind GST.ZF3-8 in vitro were also used in reporter assays. Transcriptional repression in the presence of ZNF202 was similar to the one observed for the apoE and apoAIV promoters (Table I). DISCUSSION Our results establish that ZNF202 is a transcriptional repressor that can bind in vitro to the regulatory regions of a large number of genes related to lipid processing. The target genes of ZNF202 can be divided into three distinct classes. The first group contains genes that encode structural components of lipoprotein particles: apoAIV, apoCIII, and apoE. Additionally, through its action on the apoCIII enhancer, ZNF202 also affects transcription of the apoA1 gene (11). A second class comprises enzymes involved in lipid processing: LPL, lecithin-cholesterol acyltransferase, and hepatic triglyceride lipase. Members of these first two groups are essential for maintaining lipid and cholesterol homeostasis. The overall effect of ZNF202 on lipid metabolism is difficult to predict because several of the genes products are thought to have counterbalancing functions. For example, in our hands ZNF202 represses transcription of both apoCIII and LPL reporter gene constructs, yet under some conditions these two genes have opposite effects on triglyceride hydrolysis because apoCIII inhibits LPL (22). The third group of target genes includes several genes already suspected of contributing broadly to metabolic processes: HNF4␣, insulinoma-associated gene 1, ␤-3 adrenergic receptor, and vascular endothelial growth factor. In particular, ␤-3 adrenergic receptor variants have been associated with body mass index (23,24), insulin resistance (24), and plasma lipid levels (25).
Several reports have provided genetic evidence that the chromosomal region containing the ZNF202 gene is linked to changes in lipid metabolism. Linkage analysis in Utah pedigrees identified a locus on chromosome 11q23 that predisposes to hypoalphalipoproteinemia. 2 A genome scan in Pima Indians found evidence for a 11q23 locus that influences body mass index (26). Similar results were obtained from a study in Massachusetts residents (27). In addition, a gender-specific obesity mouse quantitative trait locus that localizes to the syntenic region on mouse chromosome 9 was recently described (28). The biochemical evidence presented here raises the possibility that ZNF202 may underlay these genetic linkages to obesity and hypoalphalipoproteinemia. Several actions of ZNF202 may contribute to the clinical associations between dyslipidemia, diabetes, and obesity. For example, ZNF202 represses expression of HNF4␣, an important activator of liver-specific genes. It FIG. 6. Repression of apoAIV and apoE promoters by ZNF202 m1 in HepG2 cells. Transient cotransfection experiments in which SEAP reporter plasmids with either apoAIV(Ϫ710 to ϩ10) in pSEAP2enhancer or apoE(Ϫ322 to ϩ18) in pSEAP2-basic were introduced into HepG2 cells with indicated amounts of ZNF202m1 expression plasmid and a fixed amount of pCMV␤ control plasmid. The phosphatase activity in the supernatant and the galactosidase activity in the lysates were measured in luminometric assays; phosphatase activity was divided by galactosidase activity to adjust for possible variations in transfection efficiency. The apoAIV data represent an average of two plates, and the apoE data, an average of three plates with S.E. shown. also represses the expression of several other transcription factors and signaling molecules known to participate in metabolic processes (Table I).
Expression of structural components of lipoproteins and of the enzymes that act on them likely is coordinated, and shared regulatory elements afford one mechanism by which this coordination may occur. Specifically, the liver X receptor ␣ transcription factor binds to DR-4 sites in the promoters of genes involved in cholesterol catabolism such as cholesterol 7␣-hydroxylase (8). Similarly, farnesoid X receptor orchestrates bile acid synthesis and recovery through opposing actions on the transcription of the ilial bile acid protein gene and of cholesterol 7␣-hydroxylase (10). Finally, HNF4␣ enhances transcription of many genes related to lipid metabolism, including the genes for apoAI, apoB, apoCIII, apoAIV, and acyl-CoA dehydrogenase (11). The ZNF202 binding sites elucidated in this study constitute yet another example of a single transcription factor influencing multiple genes in a specific metabolic pathway.