Nuclear Respiratory Factor 1 Activation Sites in Genes Encoding the y-Subunit of ATP S:ynthase, Eukaryotic Initiation Factor 2a, and Tyrosine Aminotrahsferase SPECIFIC INTERACTION OF PURIFIED NRF-1 WITH MULTIPLE TARGET GENES*

originally an activator of the cytochrome c gene and subeequently found to stim- transcription through s-ific sites in other nu- clear genes whose products function in the mitochondria. These include subunits of the cytochrome oxidase and reductase complexes and ‘a component of the mitochondrial DNA replication pachinery. Here we es- tablish that a functional recoeition site for NRF-1 is present in the ATP synthase ytsubunit gene extending the proposed respiratory role of NRF- 1 to complex V. For some experiments heparin-agarose chromatography. The activity eluting col- at M KC1 precipitated ammonium the HA45 fraction (14). An- nealed oligonucleotides were 3’-end-labeled using Klenow enzyme purified by gel electrophoresis. Binding reactions were performed of TM mM Tris, 0.5 mM mM 10% (v/v) glycerol) con- taining between 0.02 4 pg of sonicated thymus DNA (opti-mized for individual fractions), approximately 10 fmol of labeled oligonucleotide, and varying of NRF-1. Competition reactions were performed by adding pmol of unlabeled oligonucleotide to the binding reaction prior of labeled oligonucleotide.

Transcription factor nuclear respiratory factor 1 (NRF-1) was originally identified as an activator of the cytochrome c gene and subeequently found to stimulate transcription through s-ific sites in other nuclear genes whose products function in the mitochondria. These include subunits of the cytochrome oxidase and reductase complexes and 'a component of the mitochondrial DNA replication pachinery. Here we establish that a functional recoeition site for NRF-1 is present in the ATP synthase ytsubunit gene extending the proposed respiratory role of NRF-1 to complex V.
In addition, biologically active NRF-1 sites are found in genes encoding the eukaryotic translation initiation factor 2 a-subunit and tyrosine aminotransferase, both of which participate in the r+-limiting step of their respective pathways of proteia biosynthesis and tyrosine catabolism. The recognition sites from each of these genes form identical cobplexes with NRF-1 as established by competition binbing assays, methylation interference footprinting, and ;UV-induced DNA crosslinking. Cloned oligomers of each NRF-1 binding site also stimulate the activity of 4 truncated cytochrome c promoter in transfected cells. the NRF-1 binding activities for the various target ites copurified approximately 33,000-fold and resic$d in a single protein of 68 kDa. These observations fyrther support a role for NRF-1 in the expression of nyclear respiratory genes and suggest it may help coordlnate respiratory metabolism with other biosynthetia and degradative pathways.
Much of what is known about the control of gene expression in vertebrate organisms comes frop studying highly inducible genes, most of which encode the ppecialized products of terminally differentiated cells. By cqntrast, little is understood of the genetic regulation of metabblic systems required by all animal cells during growth and development. The control of respiratory chain biosynthesis is of particular interest because it requires the concerted modulation of a large number of genes encoded by both nuclear and mitochondrial genetic compartments (1,2). In mammals, mitochondrial genes contribute 13 respiratory subunits along with the ribosomal and transfer RNAs required for mitochondrial translation (1). Nuclear genes encode the majority of the respiratory protein subunits and all of the enzymes and structural proteins required for mitochondrial transcription, translation, and replication. Since the products of both genomes are essential to respiratory function, mechanisms must exist for their coordination in response to cellular energy demands.
We have isolated and characterized the rat (3,4) and human (5) cytochrome c genes with the expectation of uncovering control elements common to other respiratory chain genes. These studies have led to the identification of multiple cisacting elements of potential regulatory significance (6,7). The best characterized of these is the recognition site for NRF-1,' an activator protein which stimulates transcription through specific sites found in several recently isolated nuclear genes encoding mitochondrial functions (7, 8). Two of these genes specify subunits of the cytochrome c reductase (Ref. 9, ubiquinone-binding protein) and oxidase (Ref. 10, subunit VIc) complexes. A third gene encodes the RNA subunit of MRP endonuclease, a ribonucleoprotein enzyme which generates primers for mitochondrial DNA replication (11)(12)(13). The presence of functional NRF-1 sites in these genes is suggestive of a role for NRF-1 in modulating respiratory chain function (14).
Here we describe the purification of NRF-1 to near homogeneity and identify three additional target genes for NRF-1 activation by establishing that a functional recognition site for the factor resides in each. The presence of a NRF-1 site in the gene for the ATP synthase y-subunit (yATP synthase) of complex V (15) lends further credence to a role for NRF-1 in respiratory chain expression. Moreover, NRF-1 sites are also found in genes encoding the eukaryotic initiation factor 2 a-subunit (eIF-2a) (16) and tyrosine aminotransferase (17). Both of these proteins are engaged in the rate-limiting step of their respective pathways of protein synthesis (18) and tyrosine catabolism (19). These observations are indicative of The abbreviations used are: NRF-1, nuclear respiratory factor 1; y-ATP synthase, ATP synthase y-subunit; eIF-2~4 eukaryotic initiation factor 2 a-subunit; RC4, rat somatic cytochrome c gene; ATF, activating transcription factor; COXVIc-2, cytochrome oxidase subunit VIc gene; mMRP, mouse mitochondrial RNA processing; SDS, sodium dodecyl sulfate; BSA, bovine serum albumin; CAT, chloramphenicol acetyltranferase; HEPES, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid.

6999
Target Genes for Respiratory Transcription Factor NRF-1 a more global function for NRF-1 in integrating respiratory activity with other pathways which either consume respiratory energy or generate the required substrates.

EXPERIMENTAL PROCEDURES
Synthetic Oligonucleotides-The following synthetic DNA oligomers were utilized for DNA binding and transfection analysis (TAT indicates tyrosine aminotransferase).
Single-stranded oligonucleotides were synthesized on an Applied Biosystems 4 DNA synthesizer, purified on Applied Biosystems oligonucleotide purification cartridges, and annealed to form doublestranded oligomers as described previously (14). Sequences were verified by the dideoxy sequencing method (20) after cloning the double-stranded oligomers into plasmids pGEM-7Zf(+) or pGEM-4 blue.
Nuclear Extracts and Mobility Shift Assays-Nuclear extracts were prepared from HeLa cells by an established method (21). For some experiments nuclear extracts were partially purified by heparinagarose chromatography. The NRF-1 activity eluting from the column at 0.45 M KC1 and precipitated with 50% ammonium sulfate was referred to as the HA45 fraction as previously described (14). Annealed oligonucleotides were 3'-end-labeled using Klenow enzyme and purified by gel electrophoresis. Binding reactions were performed in 25 pl of TM buffer (25 mM Tris, pH 7.9, 6.25 mM MgCl,, 0.5 mM EDTA, 0.5 mM dithiothreitol, 50 mM KC1, 10% (v/v) glycerol) containing between 0.02 and 4 pg of sonicated calf thymus DNA (optimized for individual fractions), approximately 10 fmol of labeled oligonucleotide, and varying amounts of NRF-1. Competition reactions were performed by adding 2.0 pmol of unlabeled oligonucleotide to the binding reaction prior to the addition of labeled oligonucleotide. After incubation at room temperature for 15 min, the samples were electrophoresed on a 5% polyacrylamide gel (acry1amide:bisacrylamide, 581) in 25 mM Tris, 33.5 mM boric acid, 0.5 mM EDTA (0.5 X TBE) at 10 V/cm for 2.5 h. Gels were then dried and the DNAprotein complexes visualized by autoradiography.
Methylation Interference-Derivatives of the vector pGEM-7Zf(+) containing the cloned oligonucleotides were linearized with either XhoI or MluI and 3'-end-labeled with Klenow enzyme. After a second digestion with either XhoI or M U , the resulting fragments were gelpurified and electroeluted. Each fragment (1.0 X lo6 cpm, approximately 0.5 pmol) was treated with dimethyl sulfate for 5 min at room temperature essentially as described (22). Binding reactions contained 5.0 X 10' cpm methylated fragment and 80 pg of extract fraction HA45. Following electrophoresis, the wet gel was autoradiographed for 2 h at 4 "C, and the free and protein-bound DNA bands were excised and electroeluted. Gel fragments containing the proteinbound bands were soaked 15 min in 0.5 X TBE, 0.2% (w/v) sodium dodecyl sulfate prior to electroelution. The recovered fragments were extracted once with phenol/chloroform, ethanol-precipitated with 10 p g of Escherichia coli tRNA, cleaved at methylated guanosines by treatment with 1 M piperidine, and electrophoresed on a 12% denaturing acrylamide gel.
UV-induced DNA Cross-linking-Labeled DNA was prepared by the method of Chodosh et al. (23). The synthetic oligonucleotide tyrosine aminotransferase -1090/-1069 was cloned into M13 phage and served as the template for the incorporation of 5-bromo-2'deoxyuridine triphophosphate (Pharmacia LKB Biotechnology Inc.) and [cY-~'P]~CTP. The DNA was digested with BamHI and HindIII and the 32-nucleotide probe gel-purified. Binding reactions were performed in 30-50 p1 using the conditions described for mobility shift assays except for the inclusion of 10 pg of BSA, 0.01-25 pg of sonicated calf thymus DNA, and 0.3-25 pg of denatured calf thymus DNA, each optimized for individual fractions. Approximately 0.5 units of NRF-1 binding activity (one unit is defined as the activity required to shift 5 fmol of labeled oligonucleotide tyrosine aminotransferase -1090/-1069 under standard mobility shift conditions) for fractions prior to the affinity steps and 1.0 unit of activity for affinity fractions were used. Following incubation at room temperature for 20 min, the reaction mixture was irradiated at 254 nm under a UV lamp for 30 min at a distance of 2.5 cm. After addition of 50 pl of 2 X SDS sample buffer, mixtures were subjected to electrophoresis on a 10% polyacrylamide SDS gel and the radiolabeled DNA-protein complexes visualized by autoradiography of the dried gel. For competition experiments unlabeled competitor oligonucleotides were added at a 200-fold molar excess and incubated for 10 min prior to the addition of labeled probe. Cell Culture and DNA Transfection-To assay the activity of transfected promoters containing the various NRF-1 sites, annealed oligomers were cloned as described previously (7,14) into the polylinker of expression vector pRC4CATBA/-66BA, which contains the rat cytochrome c/CAT fusion gene deleted of cytochrome c sequences upstream from position -66. In these constructs, the cloned oligomers are positioned 95 base pairs upstream from the cytochrome c transcription start site. The plasmid p4xmMRP -311/-292, a derivative of pGEM-7Zf(+), was constructed by converting the HindIII site flanking the cloned oligonucleotide to a BglII site, followed by two rounds of dimerization using the flanking BamHI and BglII sites as described previously (14).
COS-1 cells were grown and transfected by the Capol method as described (6). Forty-eight hours following transfection, cells from triplicate plates were harvested into 3 ml of phosphate-buffered saline. One-half of these pooled cells were used for preparation of cell lysates for chloramphenicol acetyltranvferase assays, and one-half were used for the preparation of low molecular weight DNA by the Hirt method (24) to normalize for transfection efficiency. Promoter activity values were the average of between two and five separate transfections of 3 plates each.
Purification of NRF-1-Nuclear extracts (21) were prepared from 12 liters of HeLa cells grown in spinner culture to a cell density of approximately lo6 cells/ml. Crude nuclear extract was applied to a heparin-agarose column (1.5 X 18 cm) equilibrated with buffer D (20 D M HEPES, pH 7.9, 20% glycerol, 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol, 0.1 M KCl). After consecutively washing the column with three bed volumes of buffer D and buffer D containing 0.25 M KCl, NRF-1 was eluted with buffer D containing 0.45 M KC1. This 0.45 M fraction was diluted to 0.1 M KC1 by the addition of buffer D without KCl. The diluted fraction was adjusted to 0.1% Nonidet P-40 and 10 p~ ZnCh and applied to a DNA-cellulose column (1.6 X 10 cm) equilibrated with buffer E (buffer D with 0.1% Nonidet P-40 and 10 p~ ZnCl,). After washing the column with three bed volumes of buffer E, the protein was eluted with a linear gradient of buffer E containing 0.1-1 M KCI. Fractions containing NRF-1 activity were pooled and diluted to 0.1 M KC1 with buffer E without KC1 and applied to a 3-ml NRF-1 affinity column equilibrated with the same buffer. The affinity column was prepared by annealing the COX-VIc +20/+46 oligonucleotides (5'-GATCCGGCTCCTGCGCATGCGTGCTGCTA-3' and 5"GATCT-AGCAGCACGCATGCGCAGGAGCCG-3'1, ligating the doublestranded oligomers end-to-end with T4 DNA ligase, and coupling the ligated material to CNBr activated Sepharose (Pharmacia) by the method of Kadonaga and Tjian (25). Proteins were eluted with a 15ml linear gradient of buffer E containing 0.1-1 M KCl. Fractions containing NRF-1 activity were pooled and diluted to 0.2 M KC1 with buffer E without KCl. The diluted fraction was adjusted to 160 rg/ ml sonicated calf thymus DNA, 50 pg/ml denatured sonicated calf thymus DNA, and 400 pg/ml insulin, incubated on ice for 10 min, and reapplied to the same 3-ml affinity column equilibrated with buffer E containing 0.2 M KCl. Proteins were eluted as described for the first pass, the active fractions pooled, and subjected to a final passage through the affinity column. Fractions containing NRF-1 activity were pooled and stored in small aliquots at -80 "C.
Recovery of NRF-1 Binding Activity from SDS-Polyacrylamide Gels-Approximately 2 pg of second pass affinity material was concentrated by ultrafiltration with a Centricon-10 concentrator (Amicon) and fractionated by electrophoresis on a 10% SDS-polyacryl-amide gel. The region of the gel lane between 76 and 44 kDa was excised and cut into four slices. Proteins were eluted and renatured by the method of Briggs et al. (26) with slight modification. Briefly, gel slices were crushed with a siliconized glass pestle in siliconized 1.5-ml Eppendorf tubes. After addition of 1 ml of elution buffer containing 50 mM Tris-HC1 (pH 7.9), 150 mM NaCl, 0.1 mM EDTA, 5 mM dithiothreitol, 0.1% SDS, and 10 pg/ml BSA, each tube was rocked gently overnight at room temperature. The gel residue was removed by centrifugation and the protein precipitated from the supernatant by addition of 100 pg of BSA and five volumes of acetone. The resulting pellet was rinsed with 1 ml of 80% acetone, air-dried, and resupended in 50 pl denaturation buffer containing 50 mM Tris-HCl (pH 7.9), 12.5 mM MgCl,, 20% glycerol, 1 mM EDTA, 1 mM dithiothreitol, 0.1% Nonidet P-40, 1 pM ZnCl,, 50 mM KC1, 6 M guanidine hydrochloride. After 20 min at room temperature, guanidine was removed by passage through a (2-25 Sephadex (Sigma) column equilibrated with denaturation buffer without guanidine hydrochloride. NRF-1 activity was assayed by mobility shift as described above with the exception that binding reactions contained 3 pg of BSA and calf thymus DNA was omitted.

Seven Nuclear Genes Contain NRF-1 Recognition Sites-
NRF-1 was initially defined by its ability to recognize a specific sequence contained within four recently isolated nuclear genes whose products were either directly or indirectly involved in mitochondrial respiratory function (7, 14). In two of these genes (cytochrome c and MRP RNA) the sites were found to be functionally conserved between rodent and human promoters. Fig. 1 shows a perfect match to the NRF-1 consensus at the 5' end of a recently published cDNA encoding y-ATP synthase (15). In addition, strong similarities to the other NRF-1 sites were observed in genes encoding eIF-2a and tyrosine aminotransferase (Fig. 1). Both genes have sequences within their 5"flanking DNA which differ by only a single nucleotide from the NRF-1 consensus derived from a compilation of functionally defined sites.

NRF-1 Complexes Formed with the Various Target Sequences Are
Indistinguishable-If specific recognition of each gene occurs through the same molecular species, the various sites should display similar patterns of nucleotide contacts. Methylation interference was performed to identify the major groove guanine nucleotides involved in NRF-1 binding to y-ATP synthase +1/+20, eIF-2a -42/-21 and tyrosine aminotransferase -1090/-1069. As shown in Fig. 3A, methylation of similar nucleotides within the core recognition sequence from each gene either completely or partially interfered with the formation of a specific DNA-protein complex. All of the nucleotide contacts within the three new sequences are found within the consensus of contact nucleotides derived from a comparison of the four previously characterized NRF-1 binding sites (14) and coincide with the directly repeated TGCGCA motifs (Fig. 3B).
UV-induced DNA-protein cross-linking was performed to investigate the size and complexity of the proteins responsible for the NRF-1 binding activity present in crude nuclear extracts. Bromodeoxyuridine was incorporated in place of thymidine into the NRF-1 site from the tyrosine aminotransferase gene that was also radiolabeled with [(u-~*P]~CTP. This site was chosen because it contains an additional AT base pair for bromodeoxyuridine incorporation which would facilitate UV cross-linking efficiency. Two labeled DNA-protein complexes, represented by a major band at about 90 kDa and a less intense band at about 110 kDa, were detected following separation of UV-irradiated complexes formed with fraction HA45 on denaturing acrylamide gels (Fig. 4, lane 1 (14) are compared with the newly identified NRF-1 sequences found in the -y-ATP synthase (-/-ATPs) (15), eIF-2a (16) [1][2][3][4][5][6] y-ATP synthase +1/+20 (lanes [7][8][9], eIF-20 -42/ -21 (lanes [10][11][12], tyrosine aminotransferase -1090/-1069 (lanes 13-15), and RC4 -281/-256 (lanes [15][16][17][18][19][20] were incubated with 25 I.cg of HeLa nuclear extract and assayed for the formation of the NRF-1 shifted complex by gel electrophoresis. Competition binding reactions contained a 100-fold molar excess of the unlabeled competitor oligonucleotide indicated above each lane. None refers to no competitor. y-ATPS, y-ATP synthase; TAT, tyrosine aminotransferase. major and minor species, respectively. Neither complex was formed when cross-linking was performed in the presence of an excess of unlabeled NRF-1 oligomers from the functional recognition sites of each of the seven target genes (lanes 2-8). However, both complexes were formed in the presence of an excess of an unrelated RC4 -281/-256 ATF oligomer (lane 1 1 ) and an hCCl -454/-431 oligomer (lane 10). The latter was derived from a region of the human cytochrome c1 gene which closely resembles the NRF-1 consensus sequence yet has been found to have no NRF-1 binding or transcriptional activity (14). The hCC1 -454/-431UP oligomer has two nucleotide differences with hCCl -454/-431 and has been shown to be active in both NRF-1 binding and transcriptional activity (14). An excess of unlabeled hCCl -454/ -431UP abolished the formation of both labeled complexes (lane 9 ) , further confirming their specificity.

NRF-1 Recognition Sites from y A T P S y n t h e , eIF-2a, and Tyrosine Aminotransferase Mediate Transcriptional Activation in Trunsfected
Cells-The previously identified NRF-1 sequences from respiratory genes were all able to stimulate transcription when cloned upstream in an RC4 promoter/ CAT fusion gene (14). The expression vector RC4CATBA.I -66B4 utilized for transient transfections, contains the RC4 promoter deleted of sequences upstream from -66 nucleotides relative to the transcription start site (7,14). In the absence of cloned oligomer this vector has only a low level of basal promoter activity. When vectors containing NRF-1 sites were transfected into COS-1 cells using pGEM-4 blue as a nonspecific carrier DNA, the level of promoter stimulation directed by 7-ATP synthase +1/+20, eIF-2a -42/-21 and tyrosine aminotransferase -1090/-1069 was similar to that observed with control vectors containing RC4 -172/-147 and COX-VIc-2 +20/+46 (Table I). To further confirm that the transcriptional activation observed with the various cloned oligomers was NRF-1-dependent, a high copy (300-fold molar

A T C C~A O O T T C C O C A T O C O C~O T~~A A A~ TAGGCTCCAAOOCOTACOCGCCACCTTTC
.. ..    TAT -1090/-1069 28.  13.8 Protein concentration was determined by the Bradford assay in fractions prior to the first affinity step. For second and third pass affinity fractions, protein concentrations were estimated by densitometric scanning of silver-stained 10% SDS-polyacrylamide gel lanes using known amounts of BSA as a standard.
'One unit is defined as the activity required to shift 5 fmol of labeled oligonucleotide tyrosine aminotransferase -logo/-1069 under standard mobility shift conditions. PI, first pass affinity fraction; PII, second pass affinity fraction; PIII, third pass affinity fraction. excess) of NRF-1 sites was introduced in trans by using an excess of carrier plasmid containing four tandem copies of the mMRP RNA -311/-292 NRF-1 oligomer (p4xmMRP -3111 -292). Cotransfection with this plasmid resulted in a substantial reduction in the CAT activity directed by each of the cloned NRF-1 oligomers (Table I). The reduction of promoter activity was the result of specific trans-competition because the stimulation of CAT activity by RC4 -187/-161, a functionally independent cytochrome c promoter element distinct from the NRF-1 site (7), was unaffected by the excess of cotransfected NRF-1 sites. These results demonstrate that the promoter activation mediated by the three newly identified NRF-1 sites occurs through a specific interaction of NRF-1 with its promoter recognition sequence.
Purification of NRF-1 to Near Homogeneity and Its Specific Recognition of Target Genes-Definitive proof that the same NRF-1 polypeptide is responsible for specific recognition of the various target genes requires the purification of NRF-1. To this end, NRF-1 was purified approximately 33,000-fold from a HeLa cell nuclear extract (Table 11, Fig. 5A). Following heparin-agarose and DNA-cellulose chromatography, the partially purified material was subjected to three passages through a specific DNA affinity column constructed by coupling a concatenated oligomer from the COXVIc-2 NRF-1 recognition site to cyanogen bromide-activated Sepharose  lanes 4-6), ?-ATP synthase +1/+20 (lanes 7-9), tyrosine aminotransferase -1090/-1069 (lanes [10][11][12], and RC4 -281/-256 (lanes [13][14][15] were incubated with approximately 2.5 ng of third pass affinity material and assayed for the formation of NRF-1-DNA complex by mobility-shift assay. Competition binding reactions contained a 200-fold molar excess of unlabeled RC4 -172/-147 (specific) or RC4 -281/-256 (nonspecific) competitor oligonucleotide as indicated above each lane. None indicates no competitor added. r-ATPS, ?-ATP synthase; TAT, tyrosine aminotransferase. (25). Although the DNA-cellulose step achieved little in the way of purification, it appears to have removed contaminants which reduced the efficiency of the subsequent DNA affinity steps. After three passes of the combined NRF-1-containing DNA-cellulose fractions through this affinity resin, the material eluting in gradient fractions between 1.0 and 1.2 M KC1 displayed a single major protein band at about 68 kDa on a denaturing acrylamide gel and a less intense band at about 55 kDa (Fig. 5A, lane 7). This fraction had all of the recovered NRF-1 binding activity and formed complexes of identical electrophoretic mobility with the NRF-1 recognition sites from RC4 -172/-147 (Fig. 5 4 (lanes 3,6,9, and  12). Moreover, no specific complex formation was observed with the latter as a labeled probe (lanes [13][14][15]. These results demonstrate that the NRF-1 activities responsible for specific recognition of the various target genes can be copurified approximately 33,000-fold and likely reside in one or both of the major protein bands visualized by silver staining. If the same proteins are responsible for specific binding to the various target genes the same DNA-protein complexes should be detectable by UV-induced DNA cross-linking using fractions from throughout the purification scheme. As shown in Fig. 6A, the same major DNA-protein complexes of 90 and 110 kDa detected using crude extracts were also found using NRF-1-containing fractions from each of the chromatographic steps including the most highly purified affinity fractions (Fig. 6A, lane 6). Like the complexes formed using the HA45 fraction (Fig. 4) those formed using affinity-purified fractions (Fig. 6B, lane 1 ) were specifically competed away by a n excess of unlabeled NRF-1 oligomers from RC4 -172/ -147 (lane 2), eIF-2a -42/-21 (lane 3 ) , ?-ATP synthase +1/ +20 (lane 4 ) , and tyrosine aminotransferase -1090/-1069 (lane 5) but not by an excess of RC4 -281/-256 ATF oligomer (lane 6). When corrected for the labeled oligomer, the 90-kDa complex corresponds to a protein of about 70 kDa. This agrees well with the major polypeptide of 68 kDa detected in the most highly purified affinity fraction, strongly suggesting that it is the NRF-1 protein responsible for recognizing the various target genes (Fig. 6A, lane 6).
The presence of a 110-kDa cross-linked complex using the most highly purified affinity fraction was unexpected. This complex implied the existence of a 90-kDa NRF-1 species assuming the complex resulted from an equimolar ratio of protein and labeled oligomer. No protein of 90 kDa was visible in the affinity-purified NRF-1 fraction, nor could a NRF-1 binding activity be detected in this region of the gel lane upon renaturation (see below). Moreover, when the UV-crosslinked material was subjected to nuclease digestion, only a single specific band corresponding to a complex of about 76 kDa was detected upon denaturing gel electrophoresis (Fig.  6C, lanes 2-4). Allowing for the 17 nucleotides known to be protected from nuclease digestion by NRF-1 binding (7), this band corresponded to a protein of about 66 kDa. Once again this agrees well with the major 68-kDa band detected upon denaturing gel electrophoresis of the most highly purified NRF-1 fraction. The elimination of the larger specific complex upon nuclease digestion indicates that it is comprised of additional nucleic acid rather than a larger polypeptide. One possibility is that the 110-kDa cross-linked complex results from the binding of NRF-1 to dimerized probe which would account for the extra 20 kDa of nuclease-sensitive material. Nevertheless, the only specific, nuclease-resistant complex corresponds to a polypeptide whose estimated mass agrees with that of the major affinity-purified protein.
To further establish that the 68-kDa protein is the active species, slices spanning the range between 76 and 44 kDa were excised from a denaturing acrylamide gel (indicated in Fig. 5A), and the proteins eluted, renatured, and assayed for NRF-1 binding activity. Of the four slices taken from the region of the gel lane containing the protein bands visible by silver staining, only slice 1, corresponding to the 68-kDa protein, displayed specific NRF-1 binding activity (Fig. 7A,  lanes 3 and 4 ) . The mass of this protein is in good agreement with the major 70-kDa cross-linked polypeptide.
Further slices in the direction of higher molecular mass failed to reveal an activity that would account for the larger UV-cross-linked complex (not shown). The DNA-protein complex formed with the renatured 68-kDa protein had a mobility identical to that formed with affinity purified NRF-1 (compare lanes 1 and 3 ) .   Fig. 5A were excised from a 10% SDS-acrylamide gel lane containing affinitypurified NRF-1. Following elution and renaturation the protein recovered from each gel slice was assayed by mobility shift using either specific (RC4 -172/-147) or nonspecific (RC4 -281/-256) endlabeled oligonucleotides as indicated above each lane. For comparison lanes I and 2 depict complexes formed with the second pass affinity fraction (PZZ). B, renatured material from gel slice 1 was incubated with end-labeled oligonucleotide RC4 -172/-147 in the absence of competitor ( l a n e 2) or in the presence of a 200-fold molar excess of the unlabeled competitor oligonucleotides indicated above each lane (lanes 3-7). For comparison lane I shows the NRF-1-DNA complex formed using the second pass ( H Z ) affinity fraction. y-ATPS, yATP synthase; TAT, tyrosine aminotransferase.
Addition of renatured material from gel slices 2 through 4 to binding reactions had no effect on either the mobility or intensity of the complex formed using slice 1 protein (not shown), indicating that the 55-kDa protein and other less intense protein bands are contaminants and not subunits of NRF-1. Competition binding assays demonstrated that the complex formed with the renatured material from slice 1 (Fig.  7B, lane 2) was displaced using an excess of unlabeled NRF-1 oligomers from RC4 -172/-147 ( l a n e 3), eIF-2a -42/-21 ( l a n e 4 ) , ?-ATP synthase +1/+20 ( l a n e 5 ) , and tyrosine aminotransferase -1090/-1069 ( l a n e 6) but not by an excess of RC4 -281/-256 ATF oligomer ( l a n e 7). These experiments establish that a NRF-1 protein of about 68 kDa is sufficient to account for the specific recognition of the NRF-1 sites present in each of the target genes.

DISCUSSION
The limited coding capacity of mammalian mitochondrial DNA necessitates a major contribution by the nuclear genome to the control of mitochondrial function. We have investigated the structure and expression of mammalian cytochrome c genes with the aim of identifying control elements common to nucleus-encoded respiratory genes. As a result of this approach, NRF-1 was discovered as a trans-activator of cytochrome c gene expression (7) and its recognition site found in other nuclear genes encoding oxidase and reductase subunits and the MRP RNA (14). We now establish that functional NRF-1 sites reside in genes encoding the ATP synthase y-subunit, the eukaryotic translation initiation factor 2 asubunit, and tyrosine aminotransferase. Moreover, we have purified NRF-1 binding activity to near homogeneity and demonstrated that a single NRF-1 polypeptide of 68 kDa can account for the specific recognition of all seven target genes.
As diagrammatically summarized in Fig. 8, these findings suggest a model where NRF-1 may help coordinate nuclear and mitochondrial expression of the respiratory chain. NRF-1 may respond to regulatory signals which act on the nucleus and relay the information to the mitochondria through its transcriptional stimulation of genes involved in both respiration and mitochondrial DNA replication. A prediction of this model is that other nuclear genes required for mitochondrial DNA replication and transcription will also be targets for NRF-1 activation. The present finding of a functional NRF-1 activation site in ?-ATP synthase gene further supports the role of NRF-1 in respiratory chain expression. While this work was in progress the sequence of the mouse gene encoding cytochrome c oxidase subunit Vb was reported (27).
Inspection of the 5"flanking sequence of this gene reveals a near perfect match to the NRF-1 consensus binding site (-92 CGCACATGCGCA -103) located upstream from the site of transcription initiation. It should be noted that at least three other respiratory genes encoding cytochrome c1 (28), cytochrome oxidase subunit IV (29), and the ATP synthase psubunit (30) appear to be devoid of NRF-1 recognition sites. However, it is intriguing that two of these genes encoding cytochrome oxidase subunit IV and the ATP synthase psubunit have binding sites for a second factor designated as NRF-2 (31), suggesting that subsets of respiratory chain genes may be differentially regulated through the utilization of multiple NRF factors. Alternatively, the present finding of a NRF-1 site in an upstream enhancer of the tyrosine aminotransferase gene raises the possibility that NRF-1 activation sites have not yet been detected in some respiratory chain genes because they reside in unknown upstream or downstream enhancer elements.
The demonstration of functional NRF-1 activation sites in both eIF-2a and tyrosine aminotransferase genes indicates that transcriptional control by NRF-1 is not exclusive to genes directly involved in mitochondrial function. This finding supports a more integrative role for NRF-1 in metabolic regulation. In both cases the NRF-1 sequences coincide with regions of the promoter identified by their hypersensitivity to DNase I digestion (17, 32). The NRF-1 site in the tyrosine aminotransferase gene corresponds to a footprint obtained in vivo using dimethyl sulfate (32). Although the tyrosine aminotransferase gene is known to be highly regulated, this and Nucleus Cytoplasm Mitochondria FIG. 8. Summary of the interactions between NRF-1 and seven target genes. Arrows depict the potential activation of the indicated target genes by NRF-1 and the site of action of the various gene products on mitochondrial function. ETC, the electron transport chain and oxidative phosphorylation system of the mitochondrial inner membrane. adjacent footprints occur in the DNase I hypersensitive region of an upstream enhancer for which no regulatory function has yet been assigned. The NRF-1 site in the eIF-2a gene corresponds to a palindromic sequence which binds a protein in the range of 66-68 kDa (32). This protein appears to stimulate transcriptioil through two nonidentical binding sites in the eIF-Pa promoter. Although only one of the sites is tested here, the other (-54 TGCGCATGCGAG -43), which is thought to be a lower affinity site, also contains a single mismatch to the NRF-1 consensus and is almost certainly a recognition site for NRF-1. Although it is likely that the same 68-kDa protein is responsible for the recognition of these as well as the other NRF-1 sites, the possibility that structurally distinct proteins are involved in the recognition of the various target genes cannot be excluded. Families of transcriptional activators whose members are similar in size and bind identical target sequences have been described (33).
If NRF-1 participates in respiratory chain expression what is the significance of finding NRF-1 activation sites in nonrespiratory genes? A feature common to both eIF-Pa and tyrosine aminotransferase is that they are both involved in the rate-limiting steps of their respective pathways. Translation factor eIF-2 catalyzes the GTP-dependent interaction of the 40 S ribosomal subunit with the initiator tRNA and is known to be an important site for regulation of translation initiation (18). Because protein synthesis is an energy-expensive process it may be useful, under certain physiological conditions, to coordinate the synthesis of initiation factors with respiratory chain proteins.
Tyrosine aminotransferase is a highly regulated, liver-specific enzyme that initiates the degradation of tyrosine ultimately leading to the production of fumarate and acetoacetate (reviewed in Ref. 19). Fumarate enters the citric acid cycle where it gives rise to reducing equivalents which can feed directly into the respiratory chain. Acetoacetate, while produced in the liver, is transported to other tissues such as heart muscle, where it serves as a preferred substrate for oxidative energy production (34). NRF-1 may provide a regulatory connection between tyrosine catabolism and mitochondrial respiration for the generation of respiratory fuels. It remains to be determined whether NRF-1 activation sites will appear in other genes encoding proteins involved in the rate-limiting steps of other key biosynthetic and degradative pathways.