ABSTRACT
We have previously used a homologous in vitro transcription system to define functional elements of the maize mitochondrial atpA promoter. These elements comprise a central domain extending from -7 to +5, relative to the transcription start site, and an upstream domain of 1-3 bp that is purine rich and centered around positions -11 to -12. Within the central domain lies an essential 5 bp core element. These elements are conserved in many mitochondrial promoters, but their functionality has only been tested for atpA. In this study we have introduced mutations into the corresponding elements of two cox3 promoters and show that while the core element is essential for cox3 promoter activity, upstream element mutations have little or no effect. To define the minimal sequence required for in vitro promoter activity a series of short cloned oligonucleotides corresponding to the atpA promoter was used. While some activity was seen with a 14 bp sequence, full activity required 26 bp, suggesting that elements other than the core and upstream region can influence promoter strength. Another series of clones showed that altered spacing between the upstream and core elements of atpA had a significant effect on promoter activity. These results further define important features of the plant mitochondrial transcriptional machinery.
Plant mitochondrial genomes are multipartite and complex, yet contain a relatively small number of genes encoding mainly components of the translational apparatus and the electron transport chain. In the Arabidopsis mitochondrial genome 57 genes are dispersed throughout 367 kb of sequence (1 ). This contrasts strongly with the compact mitochondrial genomes of animals and fungi. Like genome size and content, the transcriptional strategies of plant mitochondria are distinct. Rather than having a single promoter for each strand, like mammals, or a small number of nearly identical promoters, like Saccharomyces cerevisiae, plant mitochondria feature a degenerate promoter sequence which is poorly conserved between monocotyledonous and dicotyledenous plants and which contains only a short consensus sequence, which is itself not universal, especially for rRNA and tRNA genes (2 -7 ; reviewed in 8 ). In addition, many plant mitochondrial genes have multiple promoters (see for example 4 ,9 ), leading to complex mRNA patterns. These aspects of plant mitochondrial transcription raise the question as to how closely the plant transcriptional machinery resembles its counterparts in other organisms and whether unifying features of promoter structure can be found.
Plant mitochondrial promoter structure has been deduced from inspection of sequences surrounding known initiation sites and from in vitro transcription. Sequence comparisons showed that weak consensus motifs could be derived for maize (10 ), Oenothera (11 ), rice (12 ), sorghum (7 ), soybean (13 ) and wheat (14 ), with the most common element being YRTA (Y = C or T; R = A or G) found immediately upstream of the initiation site. To functionally test promoter elements in vitro transcription systems were developed from wheat (15 ), maize (16 ) and pea (17 ) mitochondria. In the case of maize, mutagenesis of the atpA promoter revealed that the YRTA `core' sequence was indeed an essential element and that other sequences both upstream and downstream of the core influenced, but were not absolutely required for, promoter activity in vitro (18 ). These other sequences included a purine-rich motif ~10 bp upstream of the core and other bases surrounding the core in a central domain. The pea extract was shown to transcribe both a mRNA and tRNA promoter, as well as the atp9 promoter of soybean, suggesting that transcription initiation mechanisms can be conserved across species boundaries in spite of low primary sequence homology. In addition, it was shown that the pea atp9 promoter contained an essential element between -7 and -25 relative to the transcription initiation site, upstream of the core sequence. This promoter contains a purine-rich domain which could correspond to the upstream domain of maize atpA.
Apart from the examples of maize atpA and pea atp9, no other mutagenic analyses of plant mitochondrial promoters have been reported. This raises the question of whether conserved sequence elements are indeed functional in each promoter in which they are found. Although the core, central domain and upstream purine-rich domain which had been demonstrated to form parts of the maize mitochondrial atpA promoter are also found in other maize mitochondrial promoters, for example those of cox3 (9 ,10 ), they have not been functionally analyzed. Here we report that the core, but not the upstream elements, of two maize cox3 promoters are required for in vitro transcription initiation. We also use oligonucleotide sequences to define a minimal promoter for maize in vitro transcription.
Mitochondrial protein extracts and in vitro transcription assays were carried out as previously described (18 ), except that the final chromatography step was through a Pharmacia 1 ml Resource-Q column. Templates for in vitro transcription were linearized plasmids or plasmid inserts gel purified using GeneClean II and a GeneSpin column (Bio 101, La Jolla, CA).
Cloning and site-directed mutagenesis were carried out by standard procedures. A plasmid containing the cox3 promoters (9 ) was obtained from Dr R.M.Mulligan (University of California, Irvine, CA). Minimal promoter constructs were made by annealing complementary oligonucleotides with the sequences shown in the figures, repairing them with the Klenow fragment of DNA polymerase and ligating them into pBluescript linearized with EcoRV or SmaI. All constructs were verified by DNA sequencing.
The maize mitochondrial cox3 gene has three upstream promoters, at -316, -355 and -1050 relative to the translation initiation codon (9 ). Sequences in the -316 and -355 promoters that could be aligned with those in maize atpA included the core sequence and upstream purine-rich domains, as shown in Figure 1 and discussed in Mulligan et al. (10 ). To test the importance of these elements for in vitro transcription of cox3, site-directed mutagenesis was carried out. Figure 2 shows that five different mutations in the -355 core element each reduced transcription by 85-100% relative to the wild-type sequence (Fig. 2 B and C, constructs 1-6) and, similarly, four core mutations in the -315 promoter reduced transcription by 85-100% relative to the control (constructs 8-12). Therefore, we conclude that the core element CRTAT is essential for in vitro transcription of multiple maize mitochondrial promoters.
Most maize mitochondrial promoters contain a purine-rich domain 10-20 bp upstream of the transcription initiation site (see fig. 11 in ref. 10 ). Site-directed mutagenesis showed that for atpA transversions in position -13 or -12 (-9 and -8 relative to the core sequence) caused a 50-70% decrease in transcription activity, thus defining an upstream domain required for optimal activity. To test whether similar sequences in the cox3 promoters also played a role in transcription initiation, GA motifs at positions -9 and -8 relative to the core sequence were modified by site-directed mutagenesis, as shown in Figure 3 . When these templates were used for in vitro transcription, however, only small changes in activity were seen relative to the wild-type control, with residual transcription ranging from 62 to 100%. These results indicate that the mutated nucleotides play only a minor role in cox3 transcription initiation in vitro and that either an upstream domain is not present as defined for atpA or that the upstream domains of cox3 comprise other nearby purines or other sequences.
Since only mutations in the core promoter sequence have pronounced effects on transcription activity, this raises the question of whether the core element itself is sufficient to direct accurate initiation. We had previously (16 ) tested an 11 bp sequence surrounding the atpA initiation site for autonomous initiation activity by cloning it into three different plasmids. However, in only one case was the plasmid transcriptionally active, suggesting that essential sequences were absent. This 11mer extended from 1 bp upstream of the core element to the +6 position, as shown in Figure 5 A. To define the minimal active promoter, a variety of sequences ranging from 14 to 40 bp were cloned in both orientations into two different restriction sites in the pBluescript plasmid and tested for in vitro transcription activity. Figure 5 C shows that all of these sequences were active in both orientations, but that the activities were variable. When corrected for uridine content (the transcripts were labeled with [32P]UTP) and gel loading the 14mer, 22mer and 30mer gave ~25, 40 and 75% of the activity of the 40mer respectively. Since the 14mer spans only the core and upstream domains, we conclude that these elements alone are sufficient to promote accurate transcription initiation in vitro.
In this paper we have shown that the core motif YRTAT found in most plant mitochondrial promoters is functional in the case of two maize cox3 promoters. This extends our previous observations that deletion (16 ) or mutagenesis (18 ) of this sequence strongly reduced or abolished in vitro promoter activity. We conclude that this core sequence is likely to be essential for function in each promoter within which it is found. We have also shown that for cox3 -355, mutation of the initiation position, which is located 4 bp downstream of the core sequence, abolished transcription. Interestingly, the +1 position of the yeast mitochondrial promoter can be mutated to any base with no significant effect on initiation rates (19 ). In contrast, mutagenesis of purine-rich elements in the cox3 promoters corresponding in position and sequence to the enhancing upstream domain of the maize atpA promoter had no discernible effect on transcription. While this result is surprising given the effect of mutations on atpA transcription and the prevalence of purine-rich domains upstream of plant mitochondrial core motifs (10 ), the general A+T richness of mitochondrial non-coding regions could result in fortuitous A-rich domains around promoters.
Attempts to define a minimal functional plant mitochondrial promoter have been difficult. In yeast only seven of the eight conserved bases have constraints (19 ), thus defining a 7 bp minimal promoter, whereas the promoters recognized by the related (20 ) bacteriophage RNA polymerases T3 and T7 are homologous over 23 bp, including 6 bp of transcribed sequence (21 ). In the case of minimal promoters based on the maize mitochondrial atpA gene an 11mer spanning the core sequence and some flanking sequences, representing the most highly conserved sequences based on alignments, was unable to direct transcription initiation in two out of three cloning contexts (16 ), suggesting that it depended on vector sequences in the successful case. In this paper we have shown that a 14mer which includes the core and upstream domains is sufficient to direct initiation, but that longer sequences show significantly higher activities. These results are somewhat difficult to interpret, but may suggest that multiple contacts between the transcription machinery and the template can facilitate initiation and that some of these contacts are lacking in the shorter promoter versions. In the case of T7 promoters so-called initiation and binding domains could be defined, which are clustered in the -9 to +1 region of the promoter (22 ). Immediately upstream of this region are 3 bp that are responsible for discrimination by T3 and T7 RNA polymerases (23 ). Thus, like the maize mitochondrial promoter, T7 promoter elements are distributed over an ~15 bp region and many mutations within this region have only modest effects on promoter activity (see for example 24 ). The quantitative aspects of T7 promoter analysis have been greatly facilitated by the purified polymerase and in vivo assays and clearly such methods will be necessary to conduct a fine analysis of plant mitochondrial promoter function.
The ability of plant mitochondrial polymerases to recognize promoters from multiple species, as seen with our maize extract (W.D.Rapp and D.B.Stern, unpublished results) and in other work (17 ), suggests that there are conserved features of the transcriptional apparatus. Indeed, partial sequences encoding T7/T3/yeast mitochondria-like RNA polymerases have been found in a wide variety of eukaryotes (25 ), and complete sequences for nuclear genes in Chenopodium album (26 ) and Arabidopsis thaliana (27 ) have been published. The mitochondrial localization of the translation products from one of the Arabidopsis genes has been inferred from in vitro import studies, although the presence of the protein in vivo has not yet been verified. Taken together, the data appear to favor a unified mode of transcription initiation in fungal, vertebrate (28 ) and plant mitochondria, with the core polymerase derived from a bacteriophage ancestor. The variability in promoter structure most likely reflects divergence among polymerases, as in the case of T3 and T7 RNA polymerases, where single nucleotide changes in the promoter (23 ) or a single amino acid change in the polymerase (29 ) can determine promoter recognition. Other mechanisms may operate in more primitive organisms. For example, the mitochondrial genome of the protozoan Reclinomonas americana contains four genes that could specify subunits of a typical eubacterial RNA polymerase (30 ).
We thank Shelley Lupold and Bill Rapp for helpful discussions and Mike Mulligan for providing cox3 plasmid DNA. This work was supported by NIH grant GM52560-02 to D.B.S.
REFERENCES