A Mutation in the Yeast Mitochondrial Core RNA Polymerase, Rpo41, Confers Defects in Both Specificity Factor Interaction and Promoter Utilization*

The yeast mitochondrial RNA polymerase (RNAP) is composed of the core RNAP, Rpo41, and the mitochondrial transcription factor, Mtf1. Both are required for mitochondrial transcription, but how the two proteins interact to create a functional, promoter-selective holoenzyme is still unknown. Rpo41 is similar to the single polypeptide bacteriophage T7RNAP, which does not require additional factors for promoter-selective initiation but whose activity is modulated during infection by association with T7 lysozyme. In this study we used the co-crystal structure of T7RNAP and T7 lysozyme as a model to define a potential Mtf1 interaction surface on Rpo41, making site-directed mutations in Rpo41 at positions predicted to reside at the same location as the T7RNAP/T7 lysozyme interface. We identified Rpo41 mutant E1224A as having reduced interactions with Mtf1 in a two-hybrid assay and a temperature-sensitive petite phenotype in vivo. Although the E1224A mutant has full activity in a non-selective in vitro transcription assay, it is temperature-sensitive for selective transcription from linear DNA templates containing the 14S rRNA, COX2, and tRNAcys mitochondrial promoters. The tRNAcys promoter defect can be rescued by template supercoiling but not by addition of a dinucleotide primer. The fact that mutation of Rpo41 results in selective transcription defects indicates that the core RNAP, like T7RNAP, plays an important role in promoter utilization.

Mitochondria contain a separate genome (mtDNA) 1 that encodes several mRNAs and the rRNAs and tRNAs required for their translation into components of the oxidative phosphorylation system (1)(2)(3). These mitochondrial transcripts are synthesized by an RNA polymerase (RNAP) distinct from those found in the nucleus of the eukaryotic cell (2). In the budding yeast Saccharomyces cerevisiae, two well characterized nuclear genes, RPO41, encoding the mitochondrial core RNAP, and MTF1, encoding a required mitochondrial transcription factor, interact to form the functional mitochondrial RNAP holoenzyme (mtRNAP) (4,5). Homologues of Rpo41 have been identified in many organisms (6), and apparent Mtf1 (sc-mtTFB) homologues required for selective mitochondrial transcription have been described recently in humans and mice (7)(8)(9). Therefore it is clear that more complex eukaryotes also use a multisubunit mtRNAP to express genes from the mtDNA.
The origins of the multi-subunit mtRNAP are still not clear. Rpo41 shares extensive amino acid similarity with RNAP from bacteriophage T7 and T3 (4). However, these phage-encoded single polypeptide RNAPs do not require additional factors for promoter recognition (10). The recent crystal structure determination of Mtf1 has revealed a significant structural similarity with the family of RNA methyltransferases (11). Furthermore, Mtf1 and its functional homologues from multicellular organisms all share amino acid sequence similarity with RNA methyltransferases (7)(8)(9)11). In fact, one of the human Mtf1 homologues has been shown to have methyltransferase activity (7). These recent discoveries leave some open questions about mtRNAP including: why do the T7RNAP-related mtRNAP core enzymes require a separate factor for promoter recognition, and how do a single polypeptide core RNAP and a member of the methyltransferase family interact to create a promoterselective holoenzyme ? We have demonstrated previously (12) that the Rpo41/Mtf1 interaction surface is complex and that essentially the entire Rpo41 and Mtf1 proteins are required for functional interaction. Using point mutations in Mtf1 we have identified at least three regions important for core RNAP interaction (12). Starting with non-interacting mutants of Mtf1 and isolating suppressors of the interaction defects in Rpo41, we identified three regions of Rpo41 that appear to be important for formation of the holoenzyme (13). Although Rpo41 has a 50-kDa N-terminal extension not found in T7RNAP (4), all of the residues in Rpo41 that affect Mtf1 interactions are found in the regions shared with the phage RNAP (13). Because Rpo41 and T7RNAP are very similar in these shared regions, the Mtf1 interaction surface might not be a module unique to the mtRNAP but could be related to an interaction surface present on T7RNAP. T7RNAP is known to interact with the T7-encoded lysozyme (14,15). This interaction, although not required for promoter-selective transcription, does modulate the activity of the T7RNAP and is required for proper expression of different classes of genes during the phage life cycle (15)(16)(17)(18).
The co-crystal structure of T7RNAP and T7 lysozyme has been solved allowing examination of the interaction surface at high resolution (19). Because the T7RNAP residues in contact with T7 lysozyme are conserved among the phage and mtRNAPs, Jeruzalmi and Steitz (19) speculated that the other family members might also be regulated by factors binding to this surface. Although there are no obvious structural similarities between Mtf1 and T7 lysozyme, we hypothesized that Mtf1 may associate with Rpo41 using a surface at least in part similar to that used in the T7RNAP/T7 lysozyme interaction.
To test this hypothesis we used the co-crystal structure of T7RNAP/T7 lysozyme as a model to create site-directed mutations in Rpo41 at specific positions predicted to lie in regions similar to the interaction surface on T7RNAP. One of these mutations (E1224A) did in fact demonstrate reduced interactions with Mtf1. Surprisingly, analysis of the mutant Rpo41 in in vitro transcription reactions revealed that the E1224A mutation also causes promoter utilization defects that are independent of Mtf1 interaction defect. These results indicate that the core mtRNAP itself may, like the promoter recognition competent T7RNAP, have intrinsic promoter recognition and utilization capabilities.
Cell Growth and Genetic Methods-S. cerevisiae (yeast) cells were grown in standard media including YP medium containing 2% glucose (YPD) or 2% each glycerol, ethanol, and lactate (YPGEL) (21). SC medium lacking appropriate amino acids was prepared as described previously (21). 5-Fluoro-orotic acid (5-FOA) was added to SC media lacking Leu (SC-Leu) to a final concentration of 1 g/liter. Yeast cells were transformed by the lithium acetate method (22).
Plasmid Shuffle Constructs and Assays-Recombinant techniques were as described previously (13). Plasmid pJJ1149 (containing a 5.7-kb fragment of RPO41 in Ycplac111 (13) and the site-directed RPO41 mutant PCR products) were digested with HpaI and MscI for the region 1 constructs and were digested with MscI and SnaBI for the region 2 constructs. Fragments of the digested plasmid were separated on a 0.8% agarose gel, and appropriate fragments were isolated using a gel extraction kit (Qiagen). The digested site-directed RPO41 mutant PCR products were purified using a nucleotide removal kit (Qiagen). The purified products were ligated with T4 DNA ligase (Invitrogen) at 16°C overnight. Constructs were confirmed by DNA sequencing (UCHSC Cancer Center DNA Sequencing Core). The RPO41 mutants were tested for function in vivo using a plasmid shuffle technique, in which the LEU2-containing plasmids were transformed into yJJ1095 (his4⌬309, ura3-52, ino1-13, leu2-3, rpo41::KAN, pJJ1148 (RPO41 in Ycplac33 (13)) and plated onto ϪLeu, ϪUra plates. The Leuϩ, Uraϩ transformants were plated onto 5-FOA-Leu to select for the loss of the wild type RPO41, URA3 plasmid. The Leuϩ, UraϪ cells containing the RPO41 mutant plasmids were plated onto YPGEL media to test for growth on a non-fermentable carbon source at 30 and 36°C.
Purification of Recombinant Mtf1 and Rpo41-Expression and purification of glutathione S-transferase-tagged Mtf1 was as described previously (24) with the following changes. Cells were grown in LBϩcarbenicillin media to an A 600 of 0.5, 0.1 mM isopropyl-1-thio-␤-Dgalactopyranoside was added, and the cells were induced for 8 h, harvested by centrifugation, washed with water, and stored at Ϫ80°C. Cells were lysed in lysis buffer (20 mM Hepes, pH 7.9, 100 mM KCl, 1 mM EDTA, 10% glycerol, 1 mM PMSF) with 0.5 mg/ml of lysozyme at 4°C for 2 h. The lysed cells were incubated at 37°C for 4 min and immediately put into ice. The supernatant was collected after centrifugation (31,000 ϫ g) at 4°C for 1 h. 0.5 mM DTT and glutathione-agarose (Pierce) (1 ml of resin per 0.4 g of cell) equilibrated with lysis buffer were added to the supernatant and incubated at 4°C for 2 h. The mixture was transferred into a 5-ml disposable column (Qiagen), and the resin was washed with 100 volumes of wash buffer (30 mM Hepes, pH 7.9, 100 mM KCl, 1 mM EDTA, 10% glycerol, 5 mM DTT, 1 mM PMSF) with 0.5% Nonidet P-40 and subsequently washed with 100 volumes of wash buffer containing 500 mM KCl. Proteins were eluted with 50 mM Tris-Cl, pH 8.0 25°C, 100 mM KCl, 2 mM EDTA, 10 mM MgCl 2 , 10 mM glutathione, 5% glycerol, 0.1 mM DTT, and 1 mM PMSF. The peak protein-containing fractions were pooled and dialyzed against a buffer containing 30 mM Tris-Cl, pH 8.0, 25°C, 1 mM KCl, 2 mM EDTA, 10 mM MgCl 2 , 5% glycerol, 0.1 mM DTT, 1 mM PMSF. The dialyzed samples were aliquoted, frozen in liquid nitrogen, and stored at Ϫ80°C. (His) 6 tagged forms of wild type and mutant Rpo41 were expressed and purified from bacterial cells as described elsewhere. 2 Transcription Templates and Selective and Non-selective Transcription Assays-Templates containing the 14S rRNA (pJJ83), COX2 (pJJ1109), and tRNAcys (pJJ1110) promoters used in this study and selective transcription assays were as described previously (24). The reactions contained UTP at 100 M, [␣-32 P]UTP (3000 cpm/pmol UTP), 0.8 pmol Mtf1, and 0.8 pmol of wild type or mutant Rpo41. Reactions were stopped, and products were analyzed as described (24). The products of the ϪCTP transcription reactions were analyzed on 7 M urea, 11% polyacrylamide gels. Nonselective transcription assays were performed as described previously (25) except that UTP was present at 100 M, [␣-32 P]UTP (100 cpm/pmol UTP).

Using the T7RNAP/T7 Lysozyme Interaction Surface as a Model for Rpo41/Mtf1
Interaction-The interaction between T7RNAP and T7 lysozyme is not required for promoter recognition, but it does modulate the activity of the RNAP by increasing the K m for the initiating nucleotide and reducing the ability of the RNAP to convert to the elongation mode (16 -18, 26). Although the co-crystal structure of T7RNAP complexed with T7 lysozyme has been solved (see Fig. 1A (18)) the mechanism for these changes in T7RNAP activity is still not clear.
Sousa and co-workers (18,(27)(28)(29) have speculated that binding of T7 lysozyme may alter the position of the C-terminal "foot" of the enzyme that is required for efficient binding of the initiating nucleotide and important for formation of the processive, elongating form of the RNAP. Because the regions of T7RNAP important for lysozyme interaction and the C-terminal foot are also found in Rpo41 (see Fig. 1, A and B (4, 19)), and because Rpo41 complexed with Mtf1 demonstrates a higher K m for the initiating nucleotide than observed in non-selective transcrip-tion by the core mtRNAP (30 -32), we hypothesized that Mtf1 and Rpo41 might interact via a similar interface as the T7RNAP/T7 lysozyme interaction.
In Fig. 1A is a model of the T7RNAP/T7 lysozyme co-crystal structure as solved by Jeruzalmi and Steitz (19). Region 1 and region 2 denote residues that form the contact surface between the two proteins. In Fig. 1B these two regions are shown in comparison with the related amino acids from core mtRNAP from budding (Sc) and fission (Sp) yeast, Neurospora (Nc), and FIG. 1. Using T7RNAP and T7 lysozyme co-crystal structure to predict a Rpo41/Mtf1 interaction surface. A, T7RNAP/T7 lysozyme co-crystal structure (19). Region 1 and Region 2 refer to amino acids of T7RNAP involved in direct contacts with T7 lysozyme. B, alignment of T7RNAP with ScRpo41 and comparison of amino acid sequences of regions 1 and 2 of the phage and selected mitochondrial core RNAPs. On the linear schematics of ScRpo41 and T7RNAP region 1 and region 2 are highlighted as darker boxes. The known secondary structures of the T7RNAP corresponding to regions 1 and 2 are shown below the linear schematics with boxes and an arrow denoting ␣-helices and ␤-sheets, respectively (19). Sequence alignments were performed with ClustalW (clustalw.genome.ad.jp/, (66)) comparing regions 1 and 2 of T7RNAP (amino acids 299 -335 and 829 -856, NCBI accession number 4RNP_A) and mitochondrial core RNAPs from S. cerevisiae (ScRpo41; amino acids 705-741 and 1206 -1232, NCBI accession number NP_116617), S. pombe (SpRpo41; amino acids 547-583 and 1044 -1070, NCBI accession number NP_594459), N. crassa (NcRpo41; amino acids 660 -696 and 1196 -1222, NCBI accession number AAA33587), and Homo sapiens (HsRpo41; amino acids 679 -714 and 1168 -1194, NCBI accession number NP_005026). The asterisks above the amino acid sequence indicate residues important for T7 lysozyme association based on the co-crystal structure. The filled square above the amino acid sequence denotes T7RNAP Ser-856, mutation of which to proline reduces association with T7 lysozyme (16). Phage consensus indicates that all amino acids at this position from each protein are similar (ϩ) or identical (amino acid). Fungal consensus indicates that all amino acids from the fungal (Sc, Sp, and Nc) mtRNAPs are similar (ϩ) or identical (amino acid). Positions on the ScRpo41 sequence indicated by arrows were changed to alanine in this study. humans (Hs). Based on the co-crystal structure, T7RNAP amino acids Asp-310 and Tyr-312 in region 1 and Asp-851 and Glu-855 in region 2 (marked with asterisks) should make direct contact with T7 lysozyme. Consistent with these regions being critical for interaction, Moffatt and Studier (16) demonstrated that an S856P mutation in region 2 (marked with the filled square; see Fig. 1B) decreases affinity for T7 lysozyme. The aligned amino acids in Fig. 1B do not demonstrate extensive identity; however, some positions are conserved between the phage and core mtRNAPs (phage consensus), and other positions are specifically conserved within the family of fungal core mtRNAPs (S. cerevisiae, Schizosaccharomyces pombe, and Neurospora crassa; fungal consensus). Based primarily on the fungal consensus patterns we chose positions (marked by the arrows) to mutate to alanine (see "Experimental Procedures"). These site-directed mutations were then tested for in vivo and in vitro function as described in the following sections.
Rpo41 Mutations E705A and E1224A Are ts for in Vivo Function-Mutations that disrupt the Rpo41/Mtf1 interaction result in the inability to utilize non-fermentable carbon sources (petite phenotype (12,13)). Because removal of either mtRNAP subunit ultimately results in loss of mtDNA (33,34), we used the plasmid shuffle technique, replacing the wild type copy of Rpo41 with the mutant versions (12), to screen the site-directed mutations for those that conferred a petite phenotype (glycerol minus). As summarized in Table I, we found two mutations, E705A from region 1 and E1224A from region 2, that were defective for growth on glycerol at 30°C, and more severely defective at 36°C, when compared with wild type Rpo41.
Isolation of Recombinant Forms of Rpo41 and Measurement of Non-selective Transcription Activity-To ask whether the defects in the Rpo41 E705A and E1224A mutants were simply because of misfolding or loss of catalytic activity we made N-terminal histidine-tagged constructs and then expressed and purified the recombinant proteins from bacterial cells. 2 Both mutant proteins were soluble and could be expressed and purified by the same protocols as wild type Rpo41. The highly purified proteins are shown in Fig. 2A. We used these purified proteins in non-selective transcription reactions using poly-[d(AT)] as the template (see "Experimental Procedures"). Although the E705A mutant was catalytically active, its specific activity (units/mg protein) was significantly lower (6-to 8-fold) than either wild type Rpo41 or the E1224A mutant. These measurements were made at 30°C; similar results were obtained in 38°C reactions (not shown). Although the E705A mutation is not ts for non-selective transcription, this substantial reduction in catalytic activity is the probable cause of the in vivo ts petite phenotype. We therefore focused further studies on the E1224A mutant.
Rpo41/Mtf1 Interactions Are Reduced by the E1224A Mutation-We used yeast two-hybrid constructs to investigate whether the in vivo ts petite phenotype of the E1224A mutant was because of interaction defects with Mtf1 (see "Experimental Procedures" (12,13)). The in vivo abundance of the mutant E1224A VP16-Rpo41 protein was equal to that of the wild type VP16-Rpo41 protein (data not shown). As shown in Fig. 3, interactions between Mtf1 and Rpo41 E1224A were reduced nearly 2-fold relative to interactions with wild type Rpo41 at both 30 and 38°C. However, because the interaction defect does not demonstrate sensitivity to elevated temperature, it does not entirely explain the original observation that the E1224A mutation causes an in vivo ts petite phenotype (Table I). FIG. 3. The E1224A Rpo41 mutant reduces interaction with Mtf1. Yeast two-hybrid assays were performed on extracts from cultures containing lexA-MTF1 and VP16-RPO41 constructs grown at 30°C (white bar) and 38°C (gray bar) as described by Cliften et al. (12). Measurement of ␤-galactosidase activity is described under "Experimental Procedures." TABLE I Assessment of RPO41 mutant in vivo function using a plasmid shuffle assay Plasmid-borne copies of the indicated RPO41 mutants were used to replace a wild type copy at RPO41 as described in "Experimental Procedures." Strains bearing the mutated genes were tested for their ability to grow on a nonfermentable carbon source (YPGEL) at 30°C and 36°C. ϩϩϩ indicates growth equivalent to wild type. ϩϩ or ϩ indicate growth slightly or significantly less than wild type. Ϫ indicates no growth.

FIG. 2. Purification of the (His) 6 -tagged wild type and mutant core RNAPs and measurement of non-selective transcription activity.
A, (His) 6 -tagged wild type and mutant core RNAPs were purified and analyzed by gel electrophoresis and staining with Coomassie as described. 2 B, specific activities (units/mg) in the non-selective transcription assay for each purified protein are listed.

Rpo41 E1224A Is ts for Selective Transcription-To investi-
gate further the defect in the Rpo41 E1224A mutant we performed in vitro selective run-off transcription reactions using purified recombinant proteins (see "Experimental Procedures"). We used linear templates containing the three different yeast mitochondrial promoters shown in Fig. 4A: the consensus 14S rRNA promoter (35), the variant COX2 promoter (36), and the weak tRNAcys promoter (37). We have shown previously (12,13,24) that mutations in Mtf1 or Rpo41 that significantly reduce subunit interactions abolish the ability of the mtRNAP to carry out selective transcription on these templates. In contrast, we found that mtRNAP reconstituted with the Rpo41 E1224A mutation was nearly as active as wild type mtRNAP for promoter-selective transcription at 30°C (Fig. 4,  B and C). Apparently the interaction defect is not sufficient to impair the ability of the reconstituted mtRNAP to recognize and utilize promoter-containing templates. However, the Rpo41 E1224A mutant showed a significant reduction in transcription activity on all three promoters at 38°C relative to the wild type mtRNAP (Fig. 4, B and C). Although we have not tested the E1224A mutation on all mitochondrial promoters, failure to transcribe even one mitochondrial encoded transcript will result in a petite phenotype, because every mitochondrially encoded gene product is essential for functional respiration (38). Therefore, the fact that Rpo41 E1224A is ts for promoterselective transcription is the most likely explanation for its in vivo petite phenotype. Because we do not observe a ts defect for interaction with Mtf1 (Fig. 3), this defect in promoter utilization must be an intrinsic feature of the Rpo41 E1224A mutation.
Rpo41 E1224A Has a ts Defect in Promoter Opening, but Not First Bond Formation-The ts promoter utilization defect of the E1224A mutant could be because of defects in DNA binding, open complex formation, nucleotide binding, first bond formation, promoter escape, or a combination of these transcriptional stages (39). To test whether the defect affected first bond formation we added the dinucleotide primer AU that represents the ϩ1 and ϩ2 position of the transcript from the tRNAcys template to the transcription reactions (40). As shown in Fig. 5A, we found that, although overall levels of transcription were increased on this weak promoter by addition of the primer (not shown; see Ref. 24), the ts defect of Rpo41 E1224A was not rescued (compare with Fig. 4C, lower panel).
We next asked whether the E1224A mutant was defective in promoter melting. Because use of a supercoiled template can FIG. 4. The E1224A interaction-defective Rpo41 mutant has temperature-sensitive defects in promoter-selective transcription. A, the promoter sequences used in this study, with ϩ1 indicating the transcription start site. The number in parentheses is the length of the run-off transcript. Boxed nucleotides indicate differences from the consensus. The circled ϩ2T in the tRNAcys promoter causes it to be weak compared with the ϩ2 purines in the other promoters (67). B, the 373-nucleotide in vitro transcript from run-off reactions at 30 and 38°C containing wild type or E1224A core mtRNAPs, Mtf1, and linear 14 S rRNA template was analyzed by gel electrophoresis as described under "Experimental Procedures." C, comparison of run-off transcript abundance from wild type and the E1224A Rpo41 mutant on the three linear templates (14S rRNA, COX2, and tRNAcys). Signal intensities were normalized to the value of wild type at 30°C for each template. The bars correspond to the average of two reactions, and the lines indicate the range. The white and gray bars indicate reactions at 30 and 38°C, respectively.
FIG. 5. Addition of a dinucleotide primer does not rescue the ability of E1224A Rpo41 to transcribe at 38°C, but supercoiling the tRNAcys template does restore activity. A, upper panel, the 105-nucleotide in vitro transcript from run-off reactions at 30 and 38°C containing wild type or E1224A core RNAPs, Mtf1, linear tRNAcys template, and 50 M dinucleotide (AU, ϩ1ϩ2) was analyzed by gel electrophoresis as described under "Experimental Procedures." Lower panels, transcript abundance from the linear tRNAcys template plus dinucleotide normalized to the average value of wild type at 30°C is presented. B, upper panel, the 38-nucleotide ϪCTP in vitro transcript from reactions performed at 30 and 38°C containing wild type or E1224A core RNAPs, Mtf1, supercoiled tRNAcys template, and all nucleotides except CTP was analyzed by gel electrophoresis as described under "Experimental Procedures." Lower panel, transcript abundance normalized to the average value of wild type at 30°C. correct promoter melting defects in many RNAPs (41,42), we used a supercoiled form of the tRNAcys promoter as a template in the in vitro selective transcription assays in the absence of CTP (24). Termination at the first C residue in the transcript gives a 38-nucleotide ϪCTP transcript as shown in Fig. 5B  (upper panel). Unlike the results using linear templates (Fig.  4C), both wild type and the E1224A mutant had similar activity at 30 and 38°C on the supercoiled template (Fig. 5B). Supercoiling therefore increases the activity of the Rpo41 E1224A mutant nearly 3-fold on the tRNAcys promoter. These results suggest that the major defect in the E1224A mutant is at the level of forming or stabilizing an open promoter complex. DISCUSSION We have identified a mutation in the yeast core mtRNAP (Rpo41) that affects interactions with the specificity factor (Mtf1), as well as promoter-selective transcription. Neither of the mtRNAP subunits has been observed to make promoterspecific contacts on its own. Instead, much like the multisubunit bacterial RNAPs (43), Rpo41 must associate with Mtf1 to create a holoenzyme that can recognize and initiate from a simple promoter (consensus ATATAAGTA (35)). Also like the bacterial RNAPs, the contacts that create the promoter competent form are transient; shortly after initiation Mtf1 is released from the elongating Rpo41 (25), like sigma factor is released from the bacterial core RNAP (43)(44)(45). Despite these functional similarities, the T7RNAP-like, single polypeptide core mtRNAP is very different from the multi-subunit bacterial core RNAP, and Mtf1 shares only limited amino acid sequence (5) and no obvious structural similarity with sigma factor (11,45). We therefore looked to the well characterized association between T7RNAP andT7 lysozyme as a potentially useful model for understanding interaction between Rpo41 and Mtf1.
We used both the T7RNAP/T7 lysozyme co-crystal structure and a comparison of amino acid sequences from the family of core mtRNAPs to select several residues for site-directed mutagenesis. These regions are highly conserved in the bacteriophage RNAPs but demonstrate less sequence conservation in the core mtRNAPs (see Fig. 1 (46)). This may reflect significant changes in the interaction partners between the phage and mtRNAPs and among the different mtRNAPs. If these residues did in fact define an interaction surface, then changes would lead to disruption of the Rpo41/Mtf1 interactions and loss of function inside the mitochondrion. Although only one of the nine sites selected resulted in this predicted outcome, the analysis of this mutation, E1224A, has provided new insight into how the mtRNAP may utilize promoters. The E1224A Rpo41 mutation confers a ts petite phenotype when introduced into yeast. In addition, the mutation results in a significant decrease in interaction with Mtf1, but the interaction defect is not ts and therefore does not correlate with the in vivo phenotype. Expecting to see this correlation is justified by our previous identification of point mutations in Mtf1 that are ts for both Rpo41 interaction and petite phenotype (12). In addition, the interaction reduction caused by the E1224A mutation is not as great as that observed with other non-interacting mutants of Mtf1 and Rpo41 (12,13). Instead of the interaction defect explaining the in vivo ts phenotype, we observed that promoter-selective transcription by the E1224A mutant was very sensitive to elevated temperature, even though the mutant enzyme was temperature-resistant in a non-selective assay. In combination with our observation that the transcription defect on one of the mitochondrial promoters can be corrected by supercoiling the DNA template, these data provide the first support for the idea, raised by several investigators in the field (11,31,40,47,48), that Rpo41 itself may be making critical promoter-specific contacts during initiation.
Because the E1224A Rpo41 mutation reduces interaction with Mtf1 without altering the catalytic function of the protein, it therefore defines an additional element of the Mtf1/Rpo41 interaction surface. Using a threaded structural model of Rpo41 including just the regions conserved with T7RNAP (49 -51), we have tentatively predicted the positions of amino acids known to be involved in Mtf1 interactions (Fig. 6). We previously identified Rpo41 residues critical for Mtf1 interaction in regions corresponding to the intercalating ␤-hairpin (Ala-631 and Ala-633) and the specificity loop (Lys-1127 and Glu-1124) of T7RNAP (shown in yellow in Fig. 6 (13)). When the positions of these residues are compared with that of the E1224A mutation identified in this work (shown in blue in Fig. 6), a potential contact surface can be visualized on one face of this speculative threaded model of Rpo41.
The recent structure of the elongating form of T7RNAP (52, 53) allows us to make some predictions about the fate of this binding interface. Initial DNA contacts are made by amino acids in the T7RNAP specificity loop, which makes sequence specific contacts with the promoter (54 -58), and the ␤-hairpin, which makes non-sequence specific contacts that stabilize the open promoter complex (56 -58, 60). Coincident with the synthesis of a short transcript and the formation of a DNA/RNA hybrid, T7RNAP undergoes extraordinary conformational changes resulting in the reorganization of these DNA binding elements and the creation of the processive and nonspecific contacts of elongation (52,53,61). In particular, the specificity loop moves to create part of the RNA exit tunnel and now makes contact with the nascent transcript (52, 53, 62). In  (58)) as the starting structure to create the threaded model coordinates. The image was generated using the Swiss PDB Viewer (www.expasy.org/spdbv/) (68) and POV-Ray (www.povray.org/). We independently used 3D-PSSM at www.sbg.bio.ic.ac.uk/ϳ3dpssm (69,70). This program chose the same T7RNAP structure as a starting point, and the final Rpo41 coordinates generated were essentially the same. The PSSM e-value comparing the T7RNAP/promoter DNA complex and the threaded model of Rpo41 is 6.39 ϫ 10 Ϫ5 (data not shown). addition, the intercalating ␤-hairpin moves away from the DNA, becomes disorganized, and does not play a role in elongation (52,53,59). It is interesting that the T7 lysozyme interaction surface does not change significantly during this transition (52,53), explaining the ability of lysozyme to interact with both the initiating and elongating forms of T7RNAP (17,27). If similar changes occur as Rpo41 converts to its elongation mode, the contact surface modeled in Fig. 6 would change beyond recognition, clearly explaining the dissociation of Mtf1 observed shortly after initiation (25).
Although the Rpo41 E1224A mutation does affect Mtf1 interaction, its major defect is in promoter utilization at 38°C, a defect that can be corrected by supercoiling the promoter. The fact that supercoiling but not addition of the initiating dinucleotide can correct the E1224A mutant indicates that the defect is probably in creating or stabilizing the melted open form of the promoter, not in first bond formation. The mutation is located near the C-terminal "extended foot" region of the RNAP (19), immediately adjacent to a large insertion uniquely found in some of the fungal Rpo41s (6). The C terminus of T7RNAP includes highly conserved aromatic residues that are essential for promoter-specific function (29). Mutation or deletion of these residues results in an increased K m for the initiating nucleotide (GTP) and an increase in abortive transcription (28,29). These alterations in T7RNAP activity are similar to those seen in the presence of T7 lysozyme (18,26), leading Villemain and Sousa (18) to speculate that the effects of lysozyme may be because of alterations in the structure of the C terminus and its relationship to the active site. Although effects on promoter melting have not been reported for T7RNAP foot mutants, perhaps the effect of the Rpo41 E1224A mutation is an alteration of this region, which then destabilizes the formation of the initiation complex. This idea may be supported by the properties of the E705A mutation (shown in green in Fig.  6), in a region similar to the lysozyme interaction domain, which results in decreased overall catalytic activity even though it is relatively distant from the active site.
Rpo41 is predicted to have regions that correspond to the specificity loop and intercalating ␤-hairpin used by T7RNAP for promoter recognition and melting (Fig. 6). Why does it need a separate factor for promoter-specific initiation? The fact that the E1224A mutation has lost this activity at high temperature supports the idea that the core RNAP itself is critical for this process. However, we have previously described point mutations in Mtf1 that are not defective for Rpo41 interaction but lose the ability to transcribe non-consensus promoters (24). Therefore, Mtf1 is also playing an important role in promoter utilization. Because the points of interaction on the Rpo41 surface are all positioned adjacent to residues important for DNA interactions and initiation, Mtf1 may be inducing conformational changes in Rpo41 that allow the core mtRNAP to assume an initiation-competent form. Alternatively, Mtf1, like sigma factor, may be directly providing contacts critical for initiation but not required during elongation. In this regard it is interesting to consider selection of the initiating nucleotide. In the bacterial RNAP sigma factor provides unique contacts for the initiating nucleotide (43,63). In T7RNAP, specificity for the initiating GTP is provided by Hoogsteen contacts predicted to be made by Arg-425 and Arg-632 (64). It is intriguing that although both these residues are highly conserved in the phage RNAPs, all of the mtRNAP have substitutions at the position of T7RNAP Arg-632 (6). Perhaps, like the association of T7 lysozyme with T7RNAP, and sigma factor with the bacterial core, Mtf1 association with Rpo41 alters the affinity of the core RNAP for the initiating nucleotide. Because the presence of the initiating nucleotide can also help to stabilize an open promoter complex (65), this could explain why mutations in both mtRNAP subunits can be corrected by supercoiling the DNA. These predictions can be tested by additional biochemical experiments and confirmed with structural information about the mtRNAP.