A Transcription Factor Required for Promoter Recognition by Human Mitochondrial RNA Polymerase ACCURATE INITIATION AT THE HEAVY- AND LIGHT-STRAND PROMOTERS DISSECTED AND RECONSTITUTED IN VITRO*

Faithful transcription of human mitochondrial DNA has been reproduced in vitro, using a fraction of mitochondrial proteins capable of accurate initiation at both the heavy- and light-strand promoters. Here we report the initial dissection of this system into two nonfunctional components which, upon mixing, recon-stitute promoter-specific transcriptional capacity in vitro. One of these components copurifies with the major nonspecific RNA polymerase activity, suggesting its identity. The other component lacks significant polymerase activity, but contains a protein or proteins required for accurate initiation at the two individual promoters by isolated mitochondrial RNA polymerase. This factor facilitates specific transcription, but has little or no effect on nonspecific transcription of a synthetic copolymer (poly(dA-dT)), indicating a posi- tive role in proper promoter recognition. The transcription factor markedly stimulates light-strand tran- scription, but only moderately enhances transcription initiation at the heavy-strand promoter, suggesting different or additional factor requirements for heavy-strand transcription. These requirements may reflect the functional differences between heavy- and light-strand transcription in vivo and, in particular, the role of the light-strand promoter in priming of heavy-strand DNA replication. Complete and bovine mtDNA. overall of genomes a of striking compactness mtDNA, little no intragenic spacer throughout length, number, size, and of for and 10 mM MgClZ, 1 mM DTT, 100 pg/ml RNase-free BSA, 400 p~ ATP, 150 pM CTP, 150 pM UTP, 0.2 p~ [~u-~*P]GTP at a specific activity of 410 Ci/mmol (Amersham Corp.), and 16 pg/ml of EcoRI-digested pKB741SP DNA. To detect run-off transcripts initiating at the HSP, a deletion clone (H5'A-60) lacking the LSP was digested with BamHI and EcoRI (to excise the mtDNA insert) and used to direct in vitro transcription at DNA concentrations between 0.25 and 4 pg/ml. After a 30-min incubation at 28 "C, reactions were stopped by the addition of 95 pl of 25 mM Tris-C1, pH 7.35, 12.5 mM EDTA, 100 mM NaC1,4.8% sodium dodecyl sulfate, 700 pg/ml proteinase K (Merck) and incubated 12 min at 37 "C. Following addition of 10 pg of Escherichia coli tRNA as carrier, nucleic acids were ethanol-precipitated in the presence of sodium perchlorate (12). A second ethanol precipitation was performed, and samples were resuspended in 10 pl of 80% formamide, TBE (46 mM Tris borate, pH 8.3, 1 mM EDTA), 0.01% bromphenol blue, 0.01% xylene cyanol, heated at 70 "C for 5 min, and electrophoresed on 7 M urea, 6% polyacrylamide (0.16% bisacrylamide) slab gels in TBE at 800 V until the xylene cyanol had migrated 15 cm. Gels were dried and used to expose preflashed Kodak film at -70 "C with DuPont Cronex Plus intensifying

of replication for the H-strand. Once initiated in the D-loop, transcription proceeds symmetrically around the circular genome, producing polycistronic precursor RNAs encompassing most, if not all, of the genetic information potentially encoded. Maturation of these precursors to mRNAs, rRNAs, and tRNAs requires multiple precise processing events.
The isolation, from human KB cell mitochondria, of a protein fraction capable of initiating transcription accurately on a cloned human D-loop template (5) has made possible in vitro analyses of the sequence-specific protein-nucleic acid interactions required for this event. Through the use of this fraction and a series of D-loop deletion mutants constructed in vitro, the major promoters for both Hand L-strand transcription (HSP and LSP, respectively) were delimited to small regions surrounding the start sites (6, 7). Further analysis by site-directed mutagenesis has identified nucleotides within the promoters critical for their in vitro function (8). While these two promoters are similar to one another, they bear no resemblance to any sequence found near the 5' ends of mouse in vivo D-loop strands (9).
This partially purified RNA polymerase activity has allowed us to define in vitro correlates of in vivo transcriptional events occurring in KB cell mitochondria and to probe their DNA sequence requirements. Here we report the chromatographic resolution of mtRNA polymerase activity and a transcription factor required for accurate initiation at both the HSP and LSP. Following chromatography on phosphocellulose, the fraction containing the peak of RNA polymerase active on a synthetic, nonspecific template shows only very weak selectivity for the HSP or the LSP in run-off assays. When the polymerase is supplemented with a fraction eluted at higher ionic strength, specific initiation is restored. We demonstrate that this fraction contains a protein factor (or factors) that acts as a positive regulator of mitochondrial transcription.

EXPERIMENTAL PROCEDURES
Initial RNA Polymerase Preparation-A heparin-Sepharose mtRNA polymerase fraction was prepared essentially as described by Walberg and Clayton (5), with the following modifications. Mitochondria isolated from human KB cells in late logarithmic growth phase by the method of Bogenhagen and Clayton (10) were centrifuged in a 1.0-1.5 M sucrose step gradient, collected, and diluted 2fold with mitochondrial storage buffer (20 mM Tris-CI, pH 8.0, 0.5 mM EDTA, 0.25 M sucrose, 15% (v/v) glycerol). Dithiothreitol (DTT) was added to a final concentration of 1 mM, and the mitochondria were centrifuged in a JA-20 rotor for 15 min at 15,000 rpm (18,000 X gav). After the supernatant had been decanted, the pellet was frozen in liquid nitrogen and stored a t -70 "C. Frozen mitochondria from 24 liters of KB cells were thawed on ice, washed once with 100 ml of 20 mM Tris-CI, pH 8.0, 0.2 mM EDTA, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 0.25 M sucrose, 15% glycerol, and repelleted.
Detergent lysis and heparin-Sepharose chromatography were carried out as described (6), except that centrifugation to clear the lysate was a t 45,000 rpm for 60 min in a Type 75Ti rotor (131,000 X gav); MgC12 was omitted from all buffers and the EDTA concentration was 0.1 mM; 0.1% (v/v) Triton X-100 was included throughout the chromatography; and the phenylmethylsulfonyl fluoride concentration was 0.5 mM. Finally, all reported volumes were increased 4-fold. Phosphocellulose Chromatography-After active heparin-Sepharose fractions had been pooled, one-fourth volume (4 ml) was dialyzed against enzyme storage buffer (10 mM Tris-C1, pH 8.0,O.l mM EDTA, 1 mM DTT, 50% glycerol, 0.1% Triton X-100) and stored at -20 "c (with the addition of DTT to a final concentration of 5 mM). This fraction is designated as H S pool. The remaining three-fourths volume (12 ml) was dialyzed against phosphocellulose (PC) column buffer (10 mM Tris-C1, pH 8.0, 0.1 mM EDTA, 1 mM DTT, 7.5% glycerol, 0.1% Triton X-100) containing 100 mM KC1 and loaded a t a flow rate of 20 ml/h onto a 12-ml column of phosphocellulose P-11 (Whatman), which had been precycled according to the supplier's instructions and equilibrated with PC column buffer plus 100 mM KC1. The column was washed with the same buffer (1-2 column volumes) until a constant AzW was recorded, and this step was repeated with column buffer plus 300 mM KCl. The column was eluted with 120 ml of a linear 0.3-0.8 M KC1 gradient in column buffer a t a flow rate of 20 ml/h. Twenty 6-ml fractions were collected and assayed for KC1 concentration by conductivity and for protein concentration by the method of Schaffner and Weissman (11). Onehalf volume of each fraction was then dialyzed against enzyme storage buffer as described above and stored at -20 "C, and the remainder was frozen in liquid Nz and stored at -70 "C. Dialyzed fractions were used in all transcription assays.
I n Vitro Transcription Assays-Nonspecific RNA polymerase was assayed wit.h the synthetic alternating copolymer poly(&-dT) as template. Standard reactions contained 10 mM Tris-HC1, pH 8.0, 10 mM MgC12, 1 mM DTT, 100 pg/ml RNase-free BSA, 400 p M ATP, 150 p~ CTP, 150 p~ GTP, 1 p~ unlabeled UTP, 0.2 p~ [w3'P]UTP at a specific activity of 410 Ci/mmol (Amersham Corp.), and 10 pg/ ml poly(dA-dT) (P-L Biochemicals). Incubations were a t 28 "C for 30 min, after which nucleic acids were precipitated in 2 ml of 5% trichloroacetic acid, 50 mM sodium pyrophosphate and filtered through Whatman GF/B paper, and incorporation of radioactivity was assayed by liquid scintillation spectroscopy. Except where noted, 4 p1 of dialyzed enzyme fractions were assayed in 25-pl total volume; thus, the final glycerol and Triton X-100 concentrations were 8 and 0.016%, respectively. Specific activity determinations were performed a t a UTP concentration of 10 p~ (in 50-pl total volume), with all other parameters as above. One unit of poly(dA-dT)-directed RNA polymerase activity is defined as the incorporation of 1 pmol of UMP into cold acid-insoluble material in 30 min under these reaction conditions. LSP-specific transcription was measured in a run-off assay using EcoRI-digested plasmid pKB741SP (see below) as template. Transcription was carried out in 25 pl, at final concentrations (except where noted) of 10 mM Tris-C1, pH 8.0, 10 mM MgClZ, 1 mM DTT, 100 pg/ml RNase-free BSA, 400 p~ ATP, 150 p M CTP, 150 p M UTP, 0.2 p~ [~u -~* P ] G T P a t a specific activity of 410 Ci/mmol (Amersham Corp.), and 16 pg/ml of EcoRI-digested pKB741SP DNA. To detect run-off transcripts initiating at the HSP, a deletion clone (H5'A-60) lacking the LSP was digested with BamHI and EcoRI (to excise the mtDNA insert) and used to direct in vitro transcription a t DNA concentrations between 0.25 and 4 pg/ml. After a 30-min incubation at 28 "C, reactions were stopped by the addition of 95 pl of 25 mM Tris-C1, pH 7.35, 12.5 mM EDTA, 100 mM NaC1,4.8% sodium dodecyl sulfate, 700 pg/ml proteinase K (Merck) and incubated 12 min at 37 "C. Following addition of 10 pg of Escherichia coli tRNA as carrier, nucleic acids were ethanol-precipitated in the presence of sodium perchlorate (12). A second ethanol precipitation was performed, and samples were resuspended in 10 pl of 80% formamide, TBE (46 mM Tris borate, pH 8.3, 1 mM EDTA), 0.01% bromphenol blue, 0.01% xylene cyanol, heated at 70 "C for 5 min, and electrophoresed on 7 M urea, 6% polyacrylamide (0.16% bisacrylamide) slab gels in TBE at 800 V until the xylene cyanol had migrated 15 cm. Gels were dried and used to expose preflashed Kodak XAR-5 film a t -70 "C with DuPont Cronex Lightning Plus intensifying screens.
Construction of Specific mtDNA Template- Fig. 1 shows schematically the three cloned fragments of mtDNA used as templates for in L)itro transcription. inserted into the EcoRI site of the pUCl2 derivative, pSP64 (Promega Biotec). This fragment contains both the H-and L-strand promoters mapped by Chang and Clayton (6). For use as a run-off template, this plasmid was digested with EcoRI, extracted with phenol: chloroform and chloroform, precipitated with ethanol, and resuspended in 10 mM Tris-C1, pH 8.0, 1 mM EDTA.
(either the HSP or the LSP) have been generated by D. D. Chang of Recombinant clones bearing only a single mitochondrial promoter this laboratory by Ba131 deletion (6). The two clones used here as run-off templates were H5'A-60, retaining 60 bp upstream of the Hstrand transcriptional start site (including the HSP), and L5'A-70, with only 70 bp of mtDNA sequence 5' to the start site for L-strand transcription (including the LSP) remaining. These represent extensively deleted plasmids found to sustain accurate initiation a t their respective promoters with maximum efficiency (6).

Isolation of a Nonspecific mtRNA Polymerase
The heparin-Sepharose pool of mitochondrial RNA polymerase activity from human KB cells is capable of initiating transcription accurately at both the HSP and LSP of cloned human mtDNA, contained in linearized restriction fragments (5-8). The synthetic alternating copolymer poly(dA-dT) can also serve as an in vitro template for the HS pool activity. Fig. 2A shows the profile of RNA polymerase activity, assayed with the poly(dA-dT) template, when the HS pool is subjected to chromatography on phosphocellulose. The nonspecific activity is eluted in a sharp peak at approximately 0.45 M KC1. When these fractions are assayed individually for activity on a template containing both the HSP and the LSP of human mtDNA, very little discrete run-off product is detected (Fig.   2B, lanes designated " m t TF). Rather, a diffuse background smear, apparently peaking in the same fraction as the poly(dA-dT) activity (fraction lo), is seen. In fact, in earlier studies, lower resolution purifications resulted in a peak of LSP-specific activity eluting from the column well after the nonspecific polymerase (data not shown).

Reconstitution of Specific Transcription
Two possible circumstances could account for this discrepancy between specific and nonspecific activity profiles. One is nonidentity of the two types of activity, indicating the presence of two (or more) RNA polymerases, one intrinsically specific, the other not, in the HS pool (PC load). Alternatively, accurate initiation at the LSP and elongation of the fulllength run-off transcript might depend on two or more dis- -400 possibilities. Fig. 3 displays the RNA species produced by PC fractions 15-20 in the absence (-) or presence (+) of poly(dA-dT)-directed RNA polymerase peak fraction (mt RNAP, assayed alone in the leftmost lane). Marked enhancement of LSP-specific transcription is clearly dependent on the added polymerase, with a peak in fraction 17 (hereafter termed "mitochondrial transcription factor"; eluted at 0.64 M KCl). The complementary mixing experiment, shown in Fig. 2B, demonstrates copurification of nonspecific polymerase with the factor-dependent promoter-directed polymerase activity in PC fraction 10. The data suggest that an intrinsically nonselective mtRNA polymerase is necessary, but not sufficient for LSP-specific transcription.

Characterization of Mitochondrial Transcription Factor
The enhancement effect of mixing transcription factor and polymerase fractions is not simply additive, as the titrations shown in Fig. 4  as described under "Experimental Procedures." The peak of RNA polymerase activity is seen in fraction 10, eluted at 0.45 M KC1. B, phosphocellulose fractions 7-12 (4 pl each) were assayed for LSP-directed mnoff transcription in the absence (-) or presence (+) of mitochondrial transcription factor (mt TF). Transcription factor alone is assayed in the leftmost lane. The template was EcoRI-digested pKB741SP (16 pg/ml), and the length of the expected run-off product initiated a t the LSP is 416 nucleotides (arrowhead). Transcription factor, or a compensating buffer (2 pl), was added to the reaction mixture prior to the addition of RNA polymerase, but no preincubation was required. The final glycerol and Triton X-100 concentrations in these experiments were 12 and 0.024%, respectively. All other conditions were as described under "Experimental Procedures." Size estimates of the products were based on the migration of 32P-labeled HpaIIdigested pBR322 DNA. sociable activities or factors, one being the nonspecific RNA polymerase. In this case, assaying fractions individually could result in an apparent run-off "peak" in the fractions where the activities overlap chromatographically. The experiments depicted in Figs transcription mixture before distribution among individual reactions. Quantitative recovery of this undegraded species from all reactions excludes nuclease inhibition as a significant component of transcription factor action. In Fig. 4B, the factor effect on LSP-specific transcription is quantitated. Fluorograms were scanned densitometrically as described in the legend to Fig. 4, with specific transcription defined as the ratio of the intensity of the band at -416 nucleotides, corresponding to initiation at the LSP and elongation to the end of the template, to that of the internal control band in each lane. At low levels of transcription factor, specific transcription increases linearly with increasing amounts of both factor and polymerase. Saturation of the system is seen at a higher factor concentration, which does not seem to depend on the amount of polymerase activity present. When either component is omitted, a very weak LSP signal is detected, increasing only slightly with increasing factor (Fig. 4A, lanes A-D) or polymerase (lanes E and J). In contrast, a dramatic stimulation is seen when the two fractions are present simultaneously, even at the lowest concentrations, indicating that they supply nonidentical functions essential for accurate initiation.
The transcription factor does not act by general stimulation of mtRNA polymerase activity, as reflected by its lack of a dramatic effect on poly(dA-dT) transcription, presented in Fig. 4C. In the absence of the specific promoter sequence(s), the mtRNA polymerase is virtually indifferent to the presence or absence of mitochondrial transcription factor.
The data presented above suggest that mitochondrial transcription factor acts positively to facilitate proper promoter site selection by mtRNA polymerase. Heating the factorcontaining fraction for 5 min at 100 "C completely abolishes this activity, as seen in Fig. 5, lane D. We conclude that this fraction contains a protein or proteins essential for transcriptional specificity.
Characterization of mtRNA Polymerase Cation Response-Walberg and Clayton (5) characterized the HS pool RNA polymerase activity with respect to salt optima and divalent cation requirements, using denatured calf thymus DNA as template. Fig. 6A (closed circles) shows the response of PC RNA polymerase activity, directed by poly(dA-dT), to KC1 concentration. The drastic inhibition even in low salt concentrations is characteristic of mtRNA polymerases from human (5), Xenopus (13), and yeast (14), when programmed with nonspecific templates. Fig. 6B shows that LSP-specific transcription by the PC enzyme, supplemented with transcription factor, responds very similarly to ionic strength. The open circles in Fig. 6A represent a plot of densitometric scanning data obtained from this gel.
Purification and Yield-Purification data for mtRNA polymerase are summarized in Table I. The nonspecific assay, with the synthetic poly(dA-dT) template, was carried out with 10 PM UTP present to quantitate RNA polymerase activity in crude lysate, HS pool, and PC fractions. In a separate experiment (not shown), all three fractions were shown to have a linear time course of incorporation for at least 40 min with the poly(dA-dT) template. The crude detergent-high salt lysate, cleared by centrifugation (S-130), supports very low incorporation with the synthetic template, but is extremely active in the run-off assay with EcoRI-digested pKB741SP (Fig. 7, lanes A X ) , allowing the alternative estimates of purification and yield presented in Table 11. Agreement between our two comparisons of HS pool and PC RNA polymerase, using either poly(dA-dT) or the mtDNA LSP, strengthens the argument for identity between the nonspecific and specific polymerase activities. Moreover, this equivalence of the specific to nonspecific activity ratios in HS pool and the PC system strongly suggests that the reconstitution of the LSP-specific activity after PC fractionation is complete, i.e.,

A B C D E 100'
FIG. 5. Heat lability of transcription factor activity. PC fraction 17, containing the peak of mitochondrial transcription factor, or PC fraction 10, containing the mtRNA polymerase, was heated at 100 "C for 5 min and then allowed to cool to room temperature. Runoff transcriptions were then carried out as follows: untreated RNA polymerase only (lune A), untreated RNA transcription factor only (lane B ) , untreated factor + untreated polymerase (lune C), heated factor + untreated polymerase (lune D), and untreated factor + heated polymerase (lune E ) . Final glycerol and Triton X-100 concentrations were 12 and 0.024%, respectively. that no additional specificity factors are present in the HS pool.
The striking discrepancy between activities measured for S-130 in the two assays can be rationalized in two, not mutually exclusive, ways. On the one hand, inhibition of nonspecific transcription due to DNA-binding proteins, to activities, such as ATPases and kinases, which could compete for substrates, and perhaps to nucleic acids might be expected in such a crude fraction. The greater than quantitative recovery of poly(dA-dT) RNA polymerase activity after heparin-Sepharose chromatography probably reflects, at least in part, the removal of such interfering or competing activities. It is noteworthy, however, that contaminating nucleases, active either on the double-stranded DNA template or on RNA products, do not seem to interfere, a t least with run-off activity, as judged by the comparable signal-to-background ratios and internal control recovery at all three stages of purification (Fig. 7). However, nucleases or binding proteins with a preference for single-stranded DNA could skew the results in the direction observed, since the poly(dA-dT) template might be predicted to exist largely in single-stranded form at the low ionic strength of the transcription reaction. On the other hand, such a dramatic difference-in effect, an increase in selectivity on the order of 100-fold relative to HS pool-suggests the presence of specific stimulatory factors in the S-130. These could act either by suppressing nonspecific initiation preferentially ( i e . , inhibiting a competing reaction) [KC11 inhibits both nonspecific poly(dA-dT)-directed incorporation of [32P]UMP into acid-insoluble material by PC fraction 10 mtRNA polymerase (0) and LSP-specific run-off transcription (0) by the PC-fractionated, reconstituted system. Nonspecific activity was measured as described under "Experimental Procedures." Run-off transcription was quantitated by scanning of the fluorogram shown in B, as described in the legend to Fig. 4B. B, LSP-directed run-off activity of the reconstituted PC transcription system. Reactions were catalyzed by 0.6 unit of mtRNA polymerase isolated by phosphocellulose chromatography and supplemented with 2 p1 of mitochondrial transcription factor, at KC1 concentrations of 0, 10, 20, 40, 80, and 160 mM. For valid comparison with the nonspecific assay, these run-off assays were carried out at 150 p~ unlabeled GTP, 1.0 mM unlabeled UTP, and 0.2 p~ [a-32P]UTP at a specific activity of 410 Ci/mmol. An internal control (32P-labeled RNA) was added for densitometry. or positively, by facilitating promoter recognition. In this case, the low yield of LSP-specific activity after heparin-Sepharose chromatography (Table 11) might actually reflect loss of such factors, analogous to the dissociation of transcription factor on phosphocellulose. More recently, we have detected an additional activity (or activities) capable of selectively stim-

A B C D E F G H I -600
Lsp' "w " " -400  Table 11. Final glycerol and Triton X-100 concentrations were 14 and 0.028%, respectively. One unit of LSP-specific activity = the amount of run-off product formation catalyzed by 1 unit of poly(dA-dT)-directed RNA polymerase activity in the experiment shown in Fig. 7 (extrapolated value derived from densitometric data).
* Relative to immediately preceding stage. ulating specific transcription from both the HSP and LSP by fractionation of a cleared lysate on DEAE-Sephacel. A fraction which flows through this resin at 0.1 M KC1 contains no independent polymerase activity (specific or nonspecific), but can stimulate specific transcription in a HS pool or a PC-fractionated, reconstituted system by at least 40-fold. Moreover, promoter-specific stimulation of PC-purified RNA polymerase by this flow-through fraction is dependent upon added transcription factor.' When poly(dA-dT)-directed transcription is used to calculate the RNA polymerase yields (Table I), the heparin-Sepharose step appears to be very efficient. Although major losses early in the purification cannot be ruled out, we have consistently failed to detect RNA polymerase activity (specific or nonspecific) in heparin-Sepharose flow-through fractions (data not shown). Inactivation of the enzyme may occur during heparin-Sepharose fractionation, but seems unlikely in light of the stability to storage at -20 "C of RNA polymerase activity at all stages of purification (at least 3 months for HS and PC, at least 1 month for S-130). Losses due to nonspecific adsorption or dilution, furthermore, would be predicted to be greater during subsequent dialysis and phosphocellulose chromatography steps, where the total protein concentrations are much lower.
The poor yield of RNA polymerase from phosphocellulose chromatography can be attributed almost entirely to losses during the final dialysis against 50% glycerol for storage and subsequent in vitro analysis. A greater than &fold reduction in volume is achieved, without changing the measured protein concentration (compare Table I with Fig. 2 4 ) . Thus, without the addition of carrier protein, roughly 70% of the protein in the PC peak fraction is lost by adsorption to the dialysis tubing. After correcting for this loss, the actual yield is closer to 75%, including PC peak fraction 10 and side fractions 9 and 11.

Transcription Factor Activity at the HSP
At the high DNA concentrations used to assay LSP-specific transcription, little or no HSP-specific run-off product was expected3 or observed, even with S-130 and HS pool enzymes. Moreover, reliable detection of the full-length HSP run-off product at -191 nucleotides is made problematic by LSPdependent products of similar size (see above). However, when compared at low concentrations of templates lacking the LSP, roughly equivalent low. levels of specific H-strand transcription are seen in the PC-fractionated, reconstituted system and in the HS pool. In lanes A-C of Fig. 8A, the HSP-specific run-off activity of the HS pool is assayed at 0.25, 1.0, and 4.0 pg/ml DNA. An identical titration was performed using an equivalent amount of PC-purified RNA polymerase activity (equal units of poly(dA-dT)-directed activity) in the presence (lanes E-G) or absence (lanes I-K) of near-saturating levels of mitochondrial transcription factor. Comparable promoter specificity is evident in the reconstituted system and, as is the case for LSP specificity, HSP activity depends on added transcription factor. Lanes D, H, and L represent identical transcription reactions programmed with 16 pg/ml linearized cloned DNA retaining only the LSP, controls included to ensure equal LSP-specific activity in HS pool and the reconstituted PC systems. At the highest concentration of the HSPbearing template assayed. (4 pg/ml, lane G), numerous additional, factor-dependent or factor-stimulated bands (compare lanes G and K ) appear in the PC-reconstituted transcription system. The significance of these products is obscure, but they probably reflect specific initiation at sites within the vector, perhaps due to homologies to mtDNA promoters, since the most intense signals correspond to species longer than the excised mtDNA insert. Their absence in the reactions cata-R. P. Fisher and D. A. Clayton, unpublished observations. D. D. Chang, personal communication.  A-D) or through the phosphocellulose stage (lanes E-L) were assayed for HSP-specific run-off transcription at 0.25 pg/ml (lanes A , E, and I), 1.0 pg/ml (lanes B, F, and J), or 4 pg/ml (lanes C, G, and K ) of a template lacking the LSP (H5' A-60) or at 16 pg/ ml of a template lacking the HSP (L5'A-70) (lanes D, H , and L). Expected run-off products from the HSP and LSP are -191 nucleotides (double arrowhead) and -416 nucleotides (closed arrowhead) long, respectively. In lanes E-H, promoter selectivity was reconstituted by the addition of 4 pl of PC-purified mitochondrial transcription factor; lanes I-L contain no added factor. Final glycerol and Triton X-100 concentrations were 16 and 0.32%, respectively. B, copurification of HSP-and LSP-specific transcription factor activity on phosphocellulose. PC fractions 15-20 (4 p1 each) were assayed for run-off transcription initiating at the HSP (double arrowhead) in the absence (-) or presence (+) of 1.2 units of mtRNA polymerase (mt RNAP) (PC fraction lo), assayed alone in the leftmost lane. Enhancement of HSP-directed transcription is maximal in fraction 17, corresponding to the peak of LSP-specific enhancement seen in Fig. 3. Reactions were carried out as for A a t 0.25 pg/ml template bearing only the HSP.
lyzed by the HS pool may be a function of different proportions of factor to polymerase and/or factor to DNA in the two systems.

Copurification of HSP-and LSP-specific Transcription Factor Activity
In the experiment shown in Fig. 8B, consecutive PC fractions 15-20 were assayed in the absence (-) or presence (+) of mtRNA polymerase (assayed alone in the leftmost lane) for run-off transcription initiating at the HSP. Specific transcription is dependent on added polymerase and peaks in fraction 17, together with LSP-specific activity (see Fig. 3). This copurification is suggestive of a single factor conferring specificity for both the Hand L-strand promoters of mtDNA. A second peak of HSP enhancement is seen in fraction 20, in contrast to the situation for the LSP (Fig. 3), where only a single peak is apparent. This raises the possibility that a second, HSP-specific transcription factor is eluted from phosphocellulose at still higher ionic strength.

Copurification of HSP-and LSP-specific, Factor-dependent mtRNA Polymerase
We were also able to demonstrate that a single RNA polymerase is probably responsible for both H-and L-strand transcription. A phosphocellulose column profile of HSP-specific, factor-dependent mtRNA polymerase activity is shown in Fig.  9. PC fractions 7-12 were assayed for run-off transcription both in the absence (-) and presence (+) of mitochondrial transcription factor (mt TF, assayed alone in the leftmost lane). These assays were supplemented with 2 p1 each of the DEAE flow-through fraction, mentioned above, in order to increase the HSP-specific signal. One consequence is the elevated level of "background run-off activity in the reaction containing transcription factor alone. This probably reflects minor contamination of this fraction with mtRNA polymerase, despite the good resolution of the two activities on phosphocellulose, and should not interfere with interpretation of the results. It is clear that the HSP-driven activity is peaking in PC fraction 10, exactly coincident with the peak of nonspecific RNA polymerase (poly(dA-dT)-directed) and with the peak of factor-dependent RNA polymerase activity selective for the LSP (see Fig. 2).

DISCUSSION
We have dissociated and reconstituted in vitro a promoterselective transcriptional activity of human mitochondria. At present, the simplest (i.e., two-component) models for this complementation seem quite plausible. These would entail a mtRNA polymerase lacking intrinsic promoter specificity, and a single accessory factor, conferring selectivity either by binding to the "core" polymerase or by binding to the DNA template in a sequence-specific manner. The paradigms for the first type of agent are bacterial u factors (15) and perhaps the regulatory subunits of eukaryotic nuclear RNA polymerases (16). Well-characterized examples of the latter mode of action include the nuclear transcription factors: transcription factor IIIA, which binds intragenic control sequences in 5 S ribosomal RNA genes, leading to their specific transcription by RNA polymerase I11 (17, 18), and Spl, which recognizes and binds sequences upstream of SV40 genes transcribed by RNA polymerase I1 (19).
There is indirect evidence which leads us to anticipate a greater complexity. Comparisons of a crude S-130, heparin-Sepharose pool, and phosphocellulose-fractionated reconstituted transcription systems were made using both poly(dA-dT) and a linearized, cloned D-loop template (see "Results"). While the two more extensively purified systems (HS and PC) show nearly identical specific-to-nonspecific activity ratios, i.e., equal selectivity, the crude extract (S-130) displays a 100-fold higher selectivity for the LSP. We believe that these data are best interpreted as indicating the removal from the extract, by heparin-Sepharose chromatography, of additional factors capable of stimulating promoter-directed transcription selectively.
In the promoter mapping of Chang and Clayton (6) and Hixson and Clayton (8), a regulatory role was inferred for sequences upstream of the minimal promoters; their deletion or alteration decreased the efficiency, but not the accuracy, of initiation. It remains to be seen whether the transcription factor we have isolated acts at these upstream sequences or on the minimal promoter itself. It seems likely, though, based on quantitative comparisons of promoter-specific transcription at the three stages of purification, that the protein requirements elucidated here are indeed minimal ones.
Copurification of HSP-and LSP-specific mtRNA polymerase activities provides the most compelling evidence thus far that, at least in vitro, a single mtRNA polymerase is sufficient for both H-and L-strand transcription. Consistent with this notion are the similar responses of HSP-and LSPspecific run-off transcription, both in the HS pool and in similar preparations to variations of ionic strength (data not shown) and of temperature (Ref. 20; data not shown). These systems, however, contain both mtRNA polymerase and mitochondrial transcription factor, and possibly other activities which could influence these responses. At least two components of the transcription machinery are engaged by both promoters; for a given property to be properly ascribed to, say, RNA polymerase rather than transcription factors, would require that the complex series of reactions leading to a runoff product be broken down into individual steps (e.g., promoter recognition, initiation, elongation). We have taken an alternative approach, using run-off transcription assays, to monitor the separation of the system into individual components. Although we cannot rule out completely the existence of exactly copurifying transcription factors and/or RNA polymerases with distinct specificities for the HSP and the LSP, the data presented here support the conclusion that at least these two basic elements of the transcriptional machinery are common to both promoters.
Considerable variability in HSP-specific activity among different HS pool preparations has been noted (5-7, 20), whereas LSP-specific activity appears to be more constant; the heparin-Sepharose pool comprising the input of the present phosphocellulose column is relatively deficient in HSPdirected run-off transcription. This variability implies that additional factors will prove necessary for efficient H-strand transcription. The two promoters, while similar in sequence, are not identical and exhibit markedly different sensitivity to base substitution (8), further strengthening the case for different protein requirements. It is worth noting, however, that the cleared lysate shows a promoter preference (i.e., LSP over HSP) comparable to that seen after further purification. This observation argues that additional HSP-specific factors, if they exist, may be already depleted in this crude extract. Depletion or deficiency of a putative HSP-specific factor (or factors) not required for initiation at the LSP could explain the different responses to increasing DNA concentration seen in run-off assays with LSP-and HSP-bearing templates. Hstrand transcription plateaus at DNA concentrations below 1 pg/ml (Fig. 8A), whereas LSP-directed transcription increases approximately linearly over a wide range of DNA concentrations (data not shown). Alternatively, or additionally, mtDNA topology may influence HSP activity preferentially. Wu and Dawid (13) noted changes in strand selectivity of purified Xenopus laevis mtRNA polymerase upon denaturation of the mtDNA template. In general, the run-off assay using linearized templates seems to accentuate the LSP-HSP disparity in the unfractionated HS pool; more congruent levels are seen in assays employing supercoiled templates (6). Unfortunately, a direct test of the effect of supercoiling on transcription must await further purification steps to remove a contaminating topoisomerase activity (data not shown).
That the same combination of proteins can support aggressive transcription of the L-strand and only weak activity on the H-strand of linear mtDNA may have important implications for the regulatory function of the D-loop. Multiple control elements relevant both to transcription of the Lstrand and to replication of the H-strand are positioned downstream of the LSP (21). A mechanism has been inferred from in vivo mapping studies whereby initiation by RNA polymerase at the LSP can give rise either to transcription of the L-strand in its entirety or to processed molecules priming H-strand DNA synthesis, which in turn can lead either to precisely truncated D-loop DNA strands or to a full round of DNA replication (22). Such numerous and complex pathways do not seem to exist for transcripts initiated at the HSP, the primary or sole function of which appears to be the production of rRNAs, tRNAs, and mRNAs for mitochondrial translation. Thus, maintenance of a steady state in vivo may require more stringent regulation at the level of initiation. Indeed, the LSP may itself be an important component of HSP regulation. It could act directly, by competing for factors and RNA polymerase molecules or through a topological intermediate, ie., the generation of the D-loop structure in mtDNA. Competition between the two promoters has been demonstrated in uitrowhen either promoter is deleted, initiation at the remaining promoter is enhanced-and effects of overall template topology have been noted (6).
Although the number of identified mitochondrial promoter sequences is at present limited, it is clearly apparent that these control elements are not conserved. The human HSP and LSP sequences (6) do not appear in the mouse mtDNA genome (I), and promoter mapping data for mouse mtDNA indicate that the HSP and LSP are conserved in position, but not primary sequence, relative to human mtDNA.4 Furthermore, recently identified yeast mitochondrial promoters (23) show no homology to the mammalian promoters, although the relative position of promoter to transcriptional start site D. D. Chang and D. A. Clayton, unpublished observations. is similar. Such lack of conservation among these control elements implies a stringent species specificity in the transcriptional apparatus. Mitochondrial RNA polymerases from human cells (5), Neurospora (24), Saccharomyces (14), and Xenopus (13) have been characterized and display striking similarities with respect to size, ionic strength optima, divalent cation requirements, and template preferences. The existence of dissociable transcription specificity factors could resolve the seeming paradox of highly conserved RNA polymerases recognizing profoundly divergent promoter sequences.