Introduction

Most eukaryotes have maintained mitochondrial DNA (mtDNA) that is theorized to derive from the original bacterial endosymbiont genome acquired early in evolution (Gray et al. 1999). However, the vast majority of genes from the endosymbiont have been either transferred to the nucleus or lost, thus the mitochondrial genome of most eukaryotes only retains genes encoding a small, but essential subset of proteins involved in oxidative phosphorylation. The mtDNA of many species also harbors rRNA and tRNA genes that are necessary for translation of the mtDNA-encoded mRNAs. However, the size, organization, and overall gene content of mtDNA vary dramatically between species (Boore 1999; Burger et al. 2003; Lang et al. 1999). Consequently, the machinery needed for mtDNA replication and expression must also have evolved relatively rapidly to accommodate species-specific changes in mtDNA structure and function.

The core machinery for mitochondrial transcription initiation has been identified in the budding yeast Saccharomyces cerevisiae and in Homo sapiens, which in both organisms comprises a core, single-subunit mitochondrial RNA (mtRNA) polymerase, which is a member of the bacteriophage T7/T3 family (Cermakian et al. 1996; Masters et al. 1987), and transcription factors required for efficient and specific initiation at mtDNA promoters (Shadel 2004). The two-component S. cerevisiae system consists of mtRNA polymerase, encoded by the RPO41 gene (Greenleaf et al. 1986), and the “specificity” factor, sc-mtTFB, encoded by the MTF1 gene (Lisowsky and Michaelis 1989; Schinkel et al. 1987). Together, these two proteins can initiate transcription accurately from a yeast mtDNA promoter in vitro (Jang and Jaehning 1991; Xu and Clayton 1992) and are required for transcription and mtDNA replication in vivo (Greenleaf et al. 1986; Lisowsky and Michaelis 1988). There is also an abundant, high-mobility group-box mtDNA-binding protein in S. cerevisiae, sc-mtTFA/Abf2p (Diffley and Stillman 1991), that stimulates mitochondrial transcription in vitro at low protein-to-mtDNA ratios (Xu and Clayton 1992) but is not essential for transcription in vivo (Diffley and Stillman 1991). Interestingly, cloning of the genes encoding mtTFB orthologues from yeast species closely related to S. cerevisiae revealed that this class of proteins is not highly conserved at the primary structural level, suggesting a high degree of evolutionary flexibility in the amino acid sequence (Carrodeguas et al. 1996).

In H. sapiens, the homologue of sc-mtTFA, h-mtTFA/TFAM, plays a critical role in transcription and was the first mammalian mitochondrial transcription factor identified (Fisher and Clayton 1988; Parisi and Clayton 1991). This contrasts with the function of its yeast orthologue in transcription, which is limited or nonexistent. The additional activities of h-mtTFA have been ascribed to specific domains of the protein not present in the budding yeast orthologue, in particular a 25-amino acid tail that is required for promoter-specific DNA binding and transcriptional activation (Dairaghi et al. 1995).

A second major difference between the yeast and the human mitochondrial transcription systems was revealed from the identification of two human homologues of sc-mtTFB, h-mtTFB1/TFB1M (McCulloch et al. 2002) and h-mtTFB2/TFB2M (Falkenberg et al. 2002). These two proteins each bind directly to the C-terminal tail of h-mtTFA and act as co-activators for initiation of transcription by mtRNA polymerase (McCulloch and Shadel 2003; Shadel 2004). Both h-mtTFB1 and h-mtTFB2 are homologous to a class of site-specific, rRNA adenine methyltransferases (Falkenberg et al. 2002; McCulloch et al. 2002) and h-mtTFB1 has been shown to bind S-adenosylmethionine (SAM) (McCulloch and Shadel 2003) and to possess this enzymatic activity (Seidel-Rogol et al. 2003). Thus, h-mtTFB1 has been postulated to be a dual-function protein, acting as a transcription factor and an rRNA methyltransferase, with its enzymatic activity being separable from its transcription factor function in vitro (McCulloch and Shadel 2003). It is noteworthy that sc-mtTFB, despite only minimal amino acid identity with rRNA methyltransferases, was shown by x-ray crystallography to also be closely related structurally to the bacterial ErmC and KsgA rRNA methyltransferases (O’Farrell et al. 2004; Schubot et al. 2001). Finally, a recent phylogenetic study strongly supports the biochemical and structural arguments that the mtTFB class of transcription factors is indeed descended from an ancestral gene encoding a KsgA-type RNA methyltransferase from the original bacterial endosymbiont (Shutt and Gray 2006).

The functional significance of having two mtTFB proteins in human mitochondria has not been established. It was reported by Falkenberg et al. (2002) that h-mtTFB2 is able to co-activate transcription to a greater extent in vitro than h-mtTFB1, indicating that h-mtTFB2 may serve as the primary transcription factor. This conclusion is supported by RNAi knock-down studies of Drosophila melanogaster mtTFB1 and mtTFB2 in Schneider cells, which defined a major role for dm-mtTFB2 in transcription and mtDNA copy number regulation (Matsushima et al. 2004) and elucidated a function for dm-mtTFB1 in mitochondrial translation (Matsushima et al. 2005). Together, these results indicate that the functions of mtTFB1 and mtTFB2 in higher eukaryotes may have diverged such that mtTFB1 is acting as the small subunit rRNA methyltransferase, and therefore involved in translation, while mtTFB2 is the primary mitochondrial transcription factor, with no direct role in rRNA methylation or translation. However, whether mtTFB2 has rRNA methyltransferase activity has not been addressed and the possibility of overlapping, but not fully redundant activities of mtTFB1 and mtTFB2 in transcription and rRNA methylation cannot be discounted based on the genetic studies in cultured Drosophila cells (Matsushima et al. 2004, 2005). Furthermore, given the high degree of divergence in mtDNA structure and function between species (Boore 1999; Burger et al. 2003; Gray et al. 1999; Lang et al. 1999), it is necessary to confirm the extent to which mechanisms defined in model organisms correlate with those operating in humans. Therefore, in this study, we have addressed the issue of whether h-mtTFB2 has rRNA methyltransferase activity. In addition, we have performed phylogenetic analyses of mtTFB and its ancestral rRNA methyltransferases that have provided new insight into the evolution and dual-function nature of this interesting class of transcription factors.

Materials and Methods

Plasmids

The plasmids used in this study to express h-mtTFB1 and h-mtTFB2 in E. coli are derivatives of pBluescript II KS+ (Stratagene), which harbors an IPTG-regulated lac promoter. The h-mtTFB1 cDNA was amplified by PCR with primers that introduced an NdeI restriction site (that included the ATG start codon) and a BamH1 restriction site downstream of its natural stop codon. This product was inserted into the plasmid pT7-7 using these enzymes, which placed a consensus ribosome binding site (rbs) upstream of the ATG for optimal bacterial expression. Using an engineered EcoR1 site located upstream of the rbs, an expression cassette was inserted into pBluescript II KS+ as an EcoR1-BamH1 fragment, to create the plasmid pBS-HB1. To construct the h-mtTFB2 expression vector (pBS-HB2), the h-mtTFB1 cDNA in pBS-HB1 was replaced by an analogous h-mtTFB2 PCR product with an NdeI site overlapping its start codon and an SpeI site downstream of its natural stop codon. The h-mtTFB1-G65A and h-mtTFB2-G105SA point mutations were created by a whole-circle PCR mutagenesis. The sc-mtTFB expressing plasmid was essentially the same as the pBS-HB1 and pBS-HB2 plasmids, except a XbaI-HindIII fragment from pGS344 (Shadel and Clayton 1995), a pT7-7 plasmid containing the MTF1 gene engineered with a rbs for bacterial expression, was cloned in pTZ18R instead of pBluescript II KS+.

Bacterial Kasugamycin-Resistance Assays

Drug inhibitory assays were performed using a minor modification of our previously published procedure (Seidel-Rogol et al. 2003) as follows . Briefly, bacterial strains were grown overnight at 37°C in LB medium + 100 μg/ml ampicillin (LB/amp). Cultures were diluted 100-fold in fresh LB/amp and grown to an OD600 of 0.2. The cultures were diluted again to an OD600 of 0.01 in fresh LB/amp + kasugamycin (BIOMOL) at various concentrations (see legend to Fig. 1). Cultures were then grown for 18 h at 37°C and growth was assessed by measuring OD600 the cultures. The bacterial strains used are JM101 (KsgA+ wild-type parent) and the corresponding ksgA mutant (Vila-Sanjurjo et al. 1999), which were kindly provided by Dr. A. Vila-Sanjurgo.

Fig. 1.
figure 1

Functional complementation of E. coli KsgA rRNA methlytransferase activity by h-mtTFB2, but not its S. cerevisiae ortholog sc-mtTFB. Results of a kasugamycin-sensitivity assay are shown. Plotted is the culture density (OD600) after growth of cultures in the presence of 300, 500, 700, and 1000 μg/ml kasugamycin. In all experiments two control strains were analyzed, the kasugamycin-resistant ksgA mutant strain with an empty-vector control plasmid (KsgA–) and its isogenic kasugamycin-sensitive KsgA+ JM101 parent (WT). A Results of expressing h-mtTFB1 and the h-mtTFB1-G65A, SAM-binding mutant in the ksgA mutant are shown. Note that h-mtTFB1-expressing cells displayed the same sensitivity to kasugamycin as the parental wild-type strain, thus the plot of their data completely overlaps on the graph. B Results of expressing h-mtTFB2 and the h-mtTFB1-G105A, SAM-binding mutant in the ksgA mutant. C Results of expressing sc-mtTFB in the ksgA mutant are shown. Error bars represent the standard deviation of six replicates.

Full size image

Primer Extension Analysis of Bacterial 16S Adenine Methylation

Bacterial strains used for RNA isolation were grown overnight in LB/amp medium. The next day, cultures were diluted 20-fold in fresh LB/amp and grown for 4 h. Approximately 1 × 109 cells were then harvested by centrifugation and total RNA was obtained using an RNeasy Mini Kit (Qiagen) according to the manufacturer’s guidelines. Total RNA (1 μg) was incubated with 32P- labeled EC16S primer (5′-GATATTTATAACCCCACCGCA-3′) in 10 μl of 1 × reverse transcriptase buffer (50 mM Tris-HCl, 8 mM MgCl2, 30 mM KCl, 1 mM DTT, 1.2 mM dNTPs, pH 8.5) at 55°C for 20 min, and at room temperature for 5 min, and then placed on ice. Annealed samples then were added to 10 μl of extension reaction mix (20 mM Tris-HCl, 10 mM MgCl2, 1.2 mM dNTPs, 0.1 unit/μl RNAsin [Roche], 0.5 unit/μl AMV Reverse Transcriptase [Roche]) and incubated at 42°C for 30 min. Twenty microliters of formamide loading buffer (80% deionized formamide, 10 mM EDTA, 1 mg/ml xylene cyanol, 1 mg/ml bromophenol blue) was added to each reaction and samples were denatured at 95°C for 3 min and then placed on ice. Samples (15 μl) were loaded onto a 10% polyacrylamide/7M urea gel and separated by electrophoresis at 300 V. The gel was then fixed in 10% acetic acid, dried using a vacuum gel dryer, and developed by autoradiography using x-ray film.

Phylogenetic Analysis

Sequences listed in Table 1 were compiled and aligned using ClustalX (version 1.83.1) (Thompson et al. 1997) with default parameters. Alignments were then visually inspected and manually aligned using BioEdit (version 7.0.5) (Hall 2005). Regions of ambiguous sequence, extreme N-terminal and C-terminal sequences, and regions where one sequence caused gaps to form in all other sequences were removed with the final alignment including a total of 251 sites (Supplementary Fig. 1). Gamma distribution values and Rate categories were estimated with Tree-Puzzle (version 5.2) (Schmidt et al. 2002). Sequences were then analyzed using the PHYLIP package (version 3.65) (Felsenstein 2005) as follows. The alignment file was bootstrapped 100 times with SEQBOOT. Trees were generated with PROTPARS using estimated values from Tree-Puzzle and jumbling of sequences ten times. A consensus tree was then generated with CONSENSE and plotted using TreeView (version 1.6.6) (Page 1996).

A mitochondrial transcription factor-only tree was constructed by first aligning mitochondrial transcription factor sequences for mtTFB1, mtTFB2, fungal mtTFB, and E. coli KsgA (excluding protist mtTFBs) with ClustalX as above. Alignments were again manually inspected and edited with the final alignment including a total of 309 sites (Supplementary Fig. 2). A Quartet Puzzling tree was generated with Tree-Puzzle and plotted using TreeView. Neighbor-joining, minimum evolution, maximum parsimony, and UPGMA analyses were conducted with Mega (Version 3.1) (Kumar et al. 2004) using the estimated values from Tree-Puzzle and the JTT substitution matrix (Jones et al. 1992). Bootstrapping values from these analyses were added manually to the Quartet Puzzling tree above.

Results

Human Mitochondrial Transcription Factor B2, Like Its Paralogue h-mtTFB1, Has rRNA Methyltransferase Activity

Dimethylation of two adenine residues in a 3′-terminal stem-loop of small subunit rRNAs is highly conserved, occurring on bacterial, eukaryotic cytoplasmic, and most mitochondrial rRNAs. In E. coli, the KsgA protein methylates this stem-loop and loss of this activity results in resistance to the translational inhibitory antibiotic kasugamycin (Helser et al. 1972; van Buul and van Knippenberg 1985). We demonstrated previously that h-mtTFB1 is able to functionally complement an E. coli ksgA mutant by methylating this substrate, demonstrating that it has the predicted rRNA methyltransferase activity and implicating it as the enzyme responsible for modification of the homologous stem-loop in the human mitochondrial 12S rRNA subunit (Seidel-Rogol et al. 2003).

Using this same strategy, we determined whether h-mtTFB2 also has this enzymatic activity. As reported previously (Seidel-Rogol et al. 2003), expression of h-mtTFB1 completely restored sensitivity to the antibiotic in a ksgA mutant strain (Fig. 1A). Likewise, expression of wild-type h-mtTFB2 in the same ksgA mutant background resulted in significantly increased sensitivity to kasugamycin compared to the corresponding kasugamycin-resistant control strain (Fig. 1B), indicating that h-mtTFB2 also has rRNA methylation activity. However, h-mtTFB2 only partially restored this phenotype (Fig. 1B), indicating that the activity of h-mtTFB2 was not as robust as h-mtTFB1 in this assay.

Given the apparent lower efficiency of h-mtTFB2 compared to h-mtTFB1, we wanted to confirm that its rRNA methyltransferase activity was co-factor dependent, as we showed previously for h-mtTFB1 (Seidel-Rogol et al. 2003). To examine this, we made a mutation in a conserved residue (G105A) in the predicted SAM-binding site of h-mtTFB2 by site-directed mutagenesis and tested its ability to restore antibiotic sensitivity in the ksgA mutant strain. The analogous mutation in h-mtTFB1 (G65A) was shown previously to inhibit SAM binding (McCulloch and Shadel 2003) and inactivate its enzymatic activity in this assay (Seidel-Rogol et al. 2003) and was used as a control in this experiment (Fig. 1A). As was the case for the h-mtTFB1 G65A mutant, expression of the G105A h-mtTFB2 mutant had no effect on the sensitivity to kasugamycin (i.e., the strain was equally as resistant as the ksgA mutant strain alone; Fig. 1B), indicating that its rRNA methyltransferase activity was likewise SAM-dependent. We note that the final cell densities reached in the current study in the presence of drug differ from those we published previously using this assay (Seidel-Rogol et al. 2003). This difference is due to a modification of the original procedure used in this study that entails beginning the experiment with a greater number of cells (as detailed in Materials and Methods) and not a difference in the doubling times of the strains, which were virtually identical (data not shown).

In S. cerevisiae, sc-mtTFB is the only obvious mitochondrial transcription factor B orthologue. Given that it too has structural similarity to KsgA rRNA methyltransferases (O’Farrell et al. 2004; Schubot et al. 2001), we determined whether it has this enzymatic activity. Expression of sc-mtTFB in a ksgA mutant strain had no effect on kasugamycin sensitivity and, therefore, has no detectable activity in this assay (Fig. 1C).

We next determined if the restoration of sensitivity to kasugamycin by h-mtTFB2 was due to methylation of the E. coli 16S rRNA and not an indirect effect. Dimethylation of the stem-loop has been shown by us (Seidel-Rogol et al. 2003) and others (Lafontaine et al. 1998) to impede extension by reverse transcriptase in vitro, providing a convenient primer extension assay for methylation at this site. Primer extension analysis using a radiolabeled primer that binds to the 3′ terminus of 16S rRNA (Fig. 2) revealed that the conserved stem-loop was indeed methylated in the ksgA– strain in the presence of h-mtTFB1 or h-mtTFB2 (demonstrated by the lack of longer products corresponding to extension of reverse transcriptase beyond the two methylated adenines; Fig. 2, lanes 5 and 7). This methylation-induced block was virtually indistinguishable from that observed using rRNA from the wild type (KsgA+) strain (Fig. 2, lane 4), in which the 16S rRNA is fully methylated. As was the case in the ksgA– strain, which completely lacks methylation (Fig. 2, lane 2), the block in primer extension (i.e., methylation) was significantly reduced in ksgA mutant strains expressing the h-mtTFB1 G65A (Fig. 2, lane 6) or h-mtTFB2 G105A (Fig. 2, lane 8) SAM-binding-site mutants, indicating loss of methyltransferase activity. Finally, strains expressing sc-mtTFB showed little or no methylation activity in this assay (Fig. 2, lane 9). Altogether, these results largely confirm those of the kasugamycin-sensitivity assay (Fig. 1) and demonstrate that h-mtTFB1 and h-mtTFB2 both have SAM-dependent, site-specific rRNA adenine methyltransferase activity, while the budding yeast orthologue, sc-mtTFB, does not.

Fig. 2.
figure 2

SAM-dependent methylation of an evolutionarily conserved stem-loop in the E. coli 16S rRNA by h-mtTFB2, but not its S. cerevisiae orthologue, sc-mtTFB. An autoradiogram of primer-extension products from the E. coli 16S rRNA is shown. The sequence of the 3′-terminal stem-loop of E. coli 16S rRNA is shown to the right, with the bent arrow representing the radiolabeled primer used in the assay and dimethylation of the two adenines in the loop shown. Reverse transcription from this primer is blocked by either dimethylation event, demonstrated by the lack of longer species that extend past the two adenine residues (i.e., that migrate slower than the primer alone in the gel, indicated by the arrow). Lane 1, radiolabeled primer alone; lane 2, ksgA – strain plus empty pBluescript KS+ (16S rRNA not methylated); lane 3, ksgA – strain plus empty pBluescript KS+ (control reaction lacking dATP, dCTP to mimic methylation block); lane 4, isogenic wild-type (KsgA+) strain plus empty pBluescript KS+ (16S rRNA fully methylated); lane 5, ksgA – strain + h-mtTFB1 plasmid; lane 6, ksgA – strain + h-mtTFB1-G65A plasmid; lane 7, ksgA – strain + h-mtTFB2 plasmid; lane 8, ksgA – strain + h-mtTFB2-G105A plasmid; lane 9, ksgA – strain + sc-mtTFB plasmid.

Full size image

The Existence of Two Mitochondrial Transcription Factors Is the Result of a Duplication Event That Occurred Prior to the Divergence of Fungi and Metazoans

Using a refined sequence alignment (Supplementary Fig. 1) of representative sequences for prokaryotic, eukaryotic, and archaeal dimethyltransferases, as well as known and predicted mtTFB orthologues (Table 1), we performed parsimony analysis using the PHYLIP package (Felsenstein 2005) to determine a consensus phylogenetic tree for this family of proteins (Fig. 3). Consistent with the findings of Shutt and Gray (2006), we find that the bacterial and eukaryotic cytoplasmic KsgA orthologues form distinct clades that diverge from a common ancestral point and that archaeal orthologues diverge early from the eukaryotic lineage. Likewise, we found that the mtTFB1 and mtTFB2 groups also form distinct lineages. However, perhaps unexpectedly, the mtTFB1 and mtTFB2 proteins do not form a single clade that is distinct from the fungal mtTFBs. Instead, the divergence between mtTFB1 and mtTFB2 occurs such that the fungal mtTFBs and mtTFB2 form a clade (bootstrapping value of 100) that is separated from the mtTFB1 clade, consistent with a much earlier gene duplication event than proposed previously (Rantanen et al. 2003). Additionally, divergence of the fungal mtTFBs from the mtTFB2 group to form a separate clade is highly supported (bootstrapping value of 91) (Fig. 3).

Table 1. Accession numbers of amino acid sequences used in this study
Full size table
Fig. 3.
figure 3

Phylogenetic analysis of the B family of mitochondrial transcription factors and rRNA adenine dimethyltransferases from bacteria, archaea, and eukaryotes. Shown is a consensus parsimony tree generated from PROTPARS using a gamma model with α = 2.04 and eight rate categories as estimated by Tree-Puzzle. Full genus and species names are listed at the ends of the branches they represent, as is a designation of the type of protein: rRNA dimethyltransferase (DMT), prokaryotic rRNA dimethyltransferase (KsgA), mitochondrial transcription factor B1 (mtTFB1) orthologue, mitochondrial transcription factor B2 orthologue (mtTFB2), and a separate designation for organisms that contain only a single mitochondrial transcription factor B orthologue (mtTFB). Clades are defined by brackets to right, with asterisks denoting protist mtTFB homologues. Bootstrap values from 100 replicates are listed at each node. Arrows signify nodes discussed in text. The scale bar represents 100 substitutions per sequence.

Full size image

To investigate these observations in more detail and eliminate the possibility of other functionally distinct sequences artificially producing this branching order, a second round of phylogenetic analysis was performed using only the mitochondrial transcription factor B sequences and E. coli KsgA as an outgroup. The protist mtTFB homologues were excluded from this alignment in order to reduce the possibility of lateral gene transfer events complicating tree construction. Sequences were again aligned using Clustal and manually inspected. The final refined alignment (Supplementary Fig. 2) was then subjected to phylogenetic analysis. Trees generated by five methods showed very similar branching patterns (Fig. 4). In this phylogenetic analysis three distinct clades are observed: mtTFB1, mtTFB2, and fungal mtTFB. However, where previously the branching order of these clades had not been resolved (Shutt and Gray 2006), we again observed mtTFB2 and fungal mtTFB sequences forming a clade distinct from mtTFB1. This branching order had very high support values by all methods (Fig. 4).

Fig. 4.
figure 4

Phylogenetic analysis of the mtTFB family of mitochondrial transcription factors. Shown is a quartet puzzling tree generated by Tree-Puzzle. Full genus and species names are listed at the ends of the branches they represent as is a designation of the type of protein: prokaryotic rRNA dimethyltransferase (KsgA), mitochondrial transcription factor B1 (mtTFB1), mitochondrial transcription factor B2 (mtTFB2), and a separate designation for fungal mitochondrial transcription factor B (mtTFB). Quartet puzzling (QP) values from Tree-Puzzle and bootstrap values of neighbor-joining (NJ), maximum evolution (ME), maximum parsimony (MP), and UPGMA analyses (gamma model, α = 3.18) from the Mega 3.1 software suite are listed in the order indicated at the four key branch points of the tree (arrows). The scale bar represents 0.1 substitutions per sequence.

Full size image

Discussion

An important recent advance in our understanding of mitochondrial gene expression came with the discovery that humans have two paralogues of the well-characterized S. cerevisiae mitochondrial transcription factor B, h-mtTFB1 and h-mtTFB2. An additional unexpected finding was that this class of transcription factor is related to a family of site-specific rRNA methyltransferases at both the primary and the tertiary structural level (Falkenberg et al. 2002; McCulloch et al. 2002; O’Farrell et al. 2004; Schubot et al. 2001). In fact, h-mtTFB1 was shown to possess this enzymatic activity, strongly suggesting it is a dual-function protein involved not only in transcription, but also in some aspect of translation via its ability to methylate the mitochondrial 12S rRNA (Seidel-Rogol et al. 2003). However, whether h-mtTFB2 also possesses rRNA methyltransferase activity had not been addressed and the precise contribution of each factor to transcriptional activation and rRNA methylation has yet to be established. As we will discuss, the results of this study demonstrate that, like h-mtTFB1 (Seidel-Rogol et al. 2003), h-mtTFB2 possesses rRNA methyltransferase activity, which has novel implications regarding how these factors coordinately contribute to the regulation of mitochondrial gene expression in humans. In addition, our phylogenetic analysis leads us to conclude that, during the course of evolution, there has been differential selection of transcription and enzymatic activities of mtTFB orthologues within and between species in order to coevolve the functions of these factors with changes in mtDNA structure and function in various organisms.

The KsgA class of rRNA methyltransferases dimethylates two adjacent adenine residues in a stem-loop structure located at the extreme 3′-end of the small subunit rRNA (Fig 2) (Helser et al. 1972; van Buul and van Knippenberg 1985). This stem-loop is highly conserved from bacteria to vertebrates, and is methylated in this fashion in both cytoplasmic and mitochondrial ribosomes in most eukaryotes. Lack of methylation at this site in bacteria results in resistance to the antibiotic kasugamycin, which normally inhibits bacterial translation (Helser et al. 1972; van Buul and van Knippenberg 1985). We reported previously that h-mtTFB1 functionally complements a ksgA mutation in E. coli by restoring methylation at the heterologous stem-loop, demonstrating that h-mtTFB1 has rRNA methyltransferase activity (Seidel-Rogol et al. 2003). Utilizing this same strategy, we show here that h-mtTFB2 has rRNA methyltransferase activity as evidenced by its ability to restore sensitivity to kasugamycin in an E. coli ksgA mutant (Fig. 1B) and methylate the 16S bacterial rRNA (Fig. 2). However, its activity is qualitatively lower than that of h-mtTFB1, which was analyzed in parallel (Fig. 1A).

The observed rRNA methyltransferase activity was eliminated in both h-mtTFB1 and h-mtTFB2 by a point mutation in a conserved residue required for binding SAM (McCulloch et al. 2002) (Figs. 1 and 2 and data not shown), demonstrating that the activity of each enzyme is cofactor dependent. However, unlike what would be predicted based on the drug-sensitivity assays (Fig. 1), a lower amount of methylation by h-mtTFB2 compared to h-mtTFB1 in the primer extension assay was not observed. This is most likely explained by primer extension being blocked by partially methylated rRNA (i.e., either single adenine dimethylation event), and sensitivity to kasugamycin requiring dimethylation of both of the target adenines in the 16S stem-loop substrate (Fig. 2). This would result in the primer extension assay overestimating the amount of dual adenine dimethylation and hence an inability to read out reduced amounts of enzyme activity by h-mtTFB2 effectively. We also note that the differences observed in these assays cannot be attributed to large differences in expression levels of in E. coli or to differences in total RNA used in these experiments (data not shown). We therefore conclude from these results that both h-mtTFB1 and h-mtTFB2 have rRNA methyltransferase activity, but that h-mtTFB1 is likely a more active enzyme, at least on this heterologous substrate. However, more rigorous biochemical analysis is necessary to address precisely how the rRNA methyltransferase activities differ between h-mtTFB1 and mtTFB2.

The potential regulatory significance of h-mtTFB2 having rRNA methyltransferase activity is worthy of discussion. In particular, this result leaves open the interesting possibility that h-mtTFB1 and h-mtTFB2 have partially overlapping functions in transcription and rRNA methylation, as opposed to separate and nonoverlapping functions (i.e., h-mtTFB1 is the sole rRNA methyltransferase and h-mtTFB2 is the sole transcription factor) as suggested by others (Matsushima et al. 2004, 2005). Having two mtTFB factors, each with two activities (but with different absolute magnitudes), might provide a relatively simple mechanism to differentially regulate mitochondrial gene expression based on the relative abundance (and/or activity) of each in different tissues (Fig. 5). Therefore, it remains quite possible that the amount of mitochondrial transcription in different cell types may actually be modulated by the relative levels (and/or activities) of h-mtTFB1 and h-mtTFB2 present (Fig. 5). A similar regulatory scenario can be envisioned for modulation of translation via the rRNA methyltransferase activities of h-mtTFB1 and h-mtTFB2. This model is supported further by the fact that the relative expression of these factors can fluctuate under different growth conditions in cultured cells (Gleyzer et al. 2005) and varies dramatically in different human and mouse tissues (Falkenberg et al. 2002; Rantanen et al. 2003). Finally, we emphasize, as did Kaguni and colleagues (Matsushima et al. 2004, 2005), that the knock-down studies in Drosophila Schneider cells do not rule out the possibility of partially overlapping, but nonredundant transcription and rRNA methylation activities of mtTFB1 and mtTFB2.

Fig. 5.
figure 5

A putative mtDNA regulatory scheme based on h-mtTFB1 and h-mtTFB2 having partially overlapping, but nonidentical transcription and methylation activities. Shown are four regulatory scenarios (A–D), with hypothetical effects on transcription from the light-strand promoter (LSP) or heavy-strand promoter (HSP), methylation of the mitochondrial small subunit rRNA stem-loop (12S methylation), and LSP-dependent mtDNA replication (replication) indicated. One plus (+) denotes moderate or basal activity, while two pluses (++) indicate higher activity of the indicated process. The two orthologues of mtTFB are depicted as gray and white shapes, which can represent either h-mtTFB1 or h-mtTFB2. A situation where one paralogue of h-mtTFB predominates significantly over the other (e.g., due to its higher relative abundance) is shown in scenarios A and B. In A predominant activity of one form (white) is shown, resulting in higher levels of LSP transcription and replication. In B, a situation where the other paralogue (gray) predominates, HSP transcription and 12S methylation are preferentially increased. In scenarios C and D, both forms are shown contributing significantly, but differentially (based on their relative abundance), such that basal (C) or induced (D) levels of all processes are attained. Depending on the precise relative abundance of h-mtTFB1 and h-mtTFB2, a variety of outputs can be imagined that globally regulate the system according to changing cellular needs or in different tissues. A similar regulatory scenario could also pertain to differences in activity (as opposed to abundance) and the activities of h-mtTFB1 and h-mtTFB2 in transcription are almost certainly also a function of their relative levels compared to those of h-mtTFA and mtRNA polymerase.

Full size image

We previously proposed that h-mtTFB1 is responsible for methylating the human mitochondrial 12S rRNA, based on its ability to methylate the homologous bacterial 16S substrate (Seidel-Rogol et al. 2003). Given that h-mtTFB2 has this same specificity, albeit with apparent lower activity, we postulate that both h-mtTFB1 and h-mtTFB2 can serve to methylate the mitochondrial 12S rRNA. Interestingly, the small subunit mitochondrial rRNA is reportedly not methylated at this stem-loop in S. cerevisiae (Klootwijk et al. 1975). This, coupled with our demonstration that sc-mtTFB does not possess this activity (Figs. 1C and 2), provides additional support that it is one or both of the mtTFB factors in mitochondria that methylate this conserved site in mitochondrial small subunit rRNAs in most species.

Recent phylogenetic analysis of mitochondrial transcription factor B orthologues revealed that these proteins are very likely descendents of the KsgA-like dimethyltransferase of the original mitochondrial endosymbiont (Shutt and Gray 2006). The fact that there are two mtTFB genes in many species (including most metazoans), and that they form distinct clades (Fig. 3), indicates a complex evolution that ultimately resulted in a gene duplication event such that two genes encoding mtTFB were located in the nucleus. Rantanen et al. (2002) concluded that this was the result of a relatively late gene duplication event in the metazoan lineage. However, our phylogenetic analysis (Figs. 3 and 4) strongly suggests a different mechanism. Specifically, our results indicate that a gene duplication event likely occurred much earlier in eukaryotic evolution, prior to the divergence of fungi and metazoans (Fig. 4). Based on this new model, the presence of a single orthologue of mtTFB in fungi and in other eukaryotes (e.g., C. elegans) likely represents the loss of one of the orthologues due to differential selection for the rRNA methyltransferase and/or transcription factor functions during evolution. In the case of S. cerevisiae, this apparently involved reduced selective pressure for methylation at the conserved rRNA stem-loop and eventual loss of rRNA methyltransferase activity of its mtTFB2 orthologue (Figs. 1B and 2), as well as, an apparent loss of the gene encoding its h-mtTFB1 paralogue. For the nematodes analyzed here, the lone mtTFB homologue appears to be very closely related to the mtTFB1 family. This would suggest either that the mtTFB2 homologue has been lost or that it has evolved to such an extent that it cannot be recognized by sequence alone. This presents an interesting situation for mitochondrial transcription in these worms. If indeed there is only one mtTFB homologue and it is more closely related to mtTFB1, then based on the relative activities demonstrated here, one would postulate that mitochondrial transcription in these species might involve additional transcription factors or a unique dependence on mtTFB1 to achieve a species-specific transcriptional output. Additionally, structural comparison of E. coli KsgA and S. cerevisiae mtTFB reveal striking overall similarity. This may suggest that while sc-mtTFB is most likely not an active RNA methyltransferase (Figs. 1 and 2), the general structure of this class of enzymes is essential for their transcription factor activity. The divergence of the mtTFB family into two groups with fungal and mtTFB2 factors forming one clade and mtTFB1 factors forming the other (Figs. 3 and 4) is consistent with the mtTFB2 clade representing the more active transcription factors, which corroborates the published in vivo and in vitro studies to date (Falkenberg et al. 2002; Matsushima et al. 2004, 2005). However, preservation of two paralogues in most metazoans, in contrast to fungi and protists, suggests that maintenance of two activities in both proteins is important to regulate overall mitochondrial gene expression.

In conclusion, this study provides important new insight into the function and evolution of the mitochondrial transcription machinery. First, it demonstrates that h-mtTFB1 and h-mtTFB2 both have rRNA methyltransferase activity, which has significant regulatory implications with regard to how these factors work together to control transcription and translation in human mitochondria (Fig. 5). Second, the presence of two genes of mtTFB in the metazoan lineage is apparently the result of a gene duplication event that occurred early in eukaryotic evolution. Thus, the different relative transcription and rRNA methylation activities of h-mtFB1 and h-mtTFB2 (and the complete loss of one mtTFB paralogue in certain organisms) are likely the result of differential selective pressure on these activities to meet species-specific needs. Clearly, much remains to be learned about the precise mechanism of mitochondrial gene expression and how this unique system evolved to create the wide variations observed in different organisms.