To learn about promoter specificities of the mitochondrial transcription machinery in Arabidopsis, mitochondrial transcription initiation sites were experimentally determined using a 5’-RACE technique first described by Bensing et al. (1996) (Figure 7), which since has been applied to define primary transcript 5’ termini in different groups of bacteria (Argaman, et al., 2001;Vogel, et al., 2003) and in plastids (Miyagi, et al., 1998). In bacteria as in mitochondria and plastids, primary transcript 5’ ends carry triphosphates while processed transcripts have monophosphates at their 5’ ends. Only the latter are a substrate to RNA ligase, and are in the experimental procedure selectively ligated to an RNA oligonucleotide, to which a forward primer will anneal in a subsequent 5’-RACE step. Primary 5’ termini may be ligated only after removal of a 5’ pyrophosphate through tobacco acid pyrophosphatase (TAP). Consequently, 5’-RACE will yield products from TAP-treated RNA for both primary and processed transcripts, whereas without exposure to TAP, products resulting from primary transcript termini will be significantly reduced or absent. Comparison of 5’-RACE products obtained from TAP-treated and untreated RNA (lanes +T and –T in Figure 8 and Figure 11) would thus identify primary transcripts.

Figure 7: 5’-RACE technique used to distinguish primary from processed transcript 5’ termini.

Transcripts are exposed to TAP to convert 5’ triphosphates to monophosphates (left), or not treated with TAP in a control experiment (right). An RNA linker (green box) is then ligated to the 5’ monophosphate ends, and cDNA (red dashed lines) is synthesized using a primer complementary to the gene of interest (red arrows). Reverse-transcribed ligation products are amplified using a forward primer annealing to the linker sequence and a gene-specific nested reverse primer (small black arrows). RT-PCR products are analyzed by agarose gel electrophoresis, and products derived from primary transcripts (band indicated by an arrow) are identified by comparing TAP-treated and untreated samples as detailed in the text. After Bensing et al. (1996).

The only transcription start site that has so far been experimentally defined in Arabidopsis mitochondria is located upstream of the rrn18 gene and coincides with a conserved nonanucleotide sequence motif (Giese, et al., 1996). Including rrn18 as a control, the genes rrn18, cox2 and atp9 were first investigated for which promoters have been characterized in several dicots (Binder, et al., 1995;Brown, et al., 1991;Giese, et al., 1996;Lizama, et al., 1994). To look for possible tissue-specific variations in promoter utilization, analysis of transcript 5’ ends was performed on RNA isolated from leaves and from flowers of Arabidopsis plants.

Figure 8: 5’-RACE analysis of the mitochondrial rrn18, cox2 and atp9 transcripts.

An rrn18 transcription start site was identified that mapped to position -156 with respect to the mature 18S rRNA 5’ end (Figure 8 and Figure 9). The initiating nucleotide was found to be part of the motif CGTATATAA (initiating nucleotide underlined), which has not yet been described as a promoter motif. The previously determined primary end of this transcript at position -69 (Giese, et al., 1996) appeared to result from processing rather than transcription initiation, as it gave rise to a PCR product that was not enhanced after TAP treatment of transcripts, compared with the control (Figure 8 and Figure 9). In the following, transcriptional starts and their surrounding sequences, which in plant mitochondria encompass the promoter (Caoile and Stern, 1997;Dombrowski, et al., 1999;Rapp, et al., 1993;Rapp and Stern, 1992), will be specified with the letter P (“promoter”), followed by the gene name and position of the initiating nucleotide with respect to the start of the coding sequence or the mature RNA, e.g. Prrn18-156.

Figure 9: 5’ end identification by sequencing across ligation sites of 5’-RACE products.

Chromatograms display the sequences at ligation sites of typical cloned 5’-RACE products derived from transcripts initiated at Prrn18-156 and Prrn18-69 (see Figure 8); RNA linker and transcript portions of sequences are indicated. The mtDNA sequences at Prrn18-156 and Prrn18-69 are displayed below; bent arrows indicate transcription initiation sites.

In the cox2 upstream region, two transcriptional starts were detected by 5’-RACE. Although TAP-treated and non-treated RNAs lead to similar band patterns (Figure 8), extensive sequencing of cloned PCR products revealed that among products of similar lengths, particular 5’ ends were significantly enriched or exclusively present in the TAP-treated sample (Table 9 and Figure 10) and are thus bona fide primary ends. While a nonanucleotide sequence at Pcox2-210 matched the motif found at Prrn18-156 exactly, only limited similarity to any known plant mitochondrial promoter was seen for Pcox2-481.

5’-RACE analysis of atp9 transcripts identified one major and one minor 5’ end, the latter mapping to position -295 within the motif CGTATATAA and the former mapping to position -239 within the sequence CATAAGAGA which, based on sequence comparisons with the experimentally defined atp9 promoter in pea mitochondria, had been predicted to function as a promoter upstream of atp9 and several other genes in Arabidopsis mitochondria (Dombrowski, et al., 1998). However, PCR products resulting from either 5’ end were equally abundant after amplification from TAP-treated and non-treated RNA (Figure 8), and transcripts were found to start with the nucleotides underlined in Figure 10, regardless of the application of TAP (Figure 10). Thus, both 5’ ends were carrying 5’ monophosphates and therefore resulting from processing events, despite the perfect nonanucleotide motifs. Alternatively, mixed populations of primary and processed transcript 5’ ends might have been present, starting with identical nucleotides but carrying either tri- or monophosphates. In order to unambiguously determine whether transcription initiated at positions -239 and -295, the species of atp9 transcript ends mapping to these positions were tested for the presence of in vitro-cappable 5’ termini (see below).

Figure 10: Transcript 5’ termini detected through cloning and sequencing of 5’-RACE products designated Pcox2-210, Pcox2-481, Patp9-239 and Patp9-295 in Figure 8.

Parts of the cox2 and atp9 upstream sequences that surround the four transcription initiation sites are shown; numbers preceding the sequences are the positions of the first displayed nucleotide with respect to the translational start. Numbers written below nucleotide positions indicate frequencies of clones that were found to correspond to transcript 5’ ends mapping to the respective positions (row +T, clone numbers for products of 5’-RACE following TAP treatment; row –T, clone numbers determined without TAP treatment). Only 5’ ends detected more than twice are marked. Numbers behind slashes indicate the numbers of clones that were sequenced in total for each promoter. Nucleotides corresponding to 5’ ends that most likely result from processing events are indicated by open triangles. Transcription initiation sites, which gave rise to TAP-specific 5’-RACE products, are indicated by bent arrows. Processing sites and initiation sites were appointed as detailed in the text. Upstream of Pcox2-210, a small bent arrow marks position -231, which might be a transcription initiation site but did not yield a distinct band in 5’-RACE experiments.

With the aim of identifying additional promoters of the genes rrn18, cox2 and atp9, their 5’ regions were analyzed by 5’-RACE through repeatedly placing reverse primers upstream of identified transcriptional starts, until no further transcript ends could be detected. All three genes were found to possess additional upstream promoters (right panels in Figure 8), of which none matched any known plant mitochondrial promoter sequence (Table 9).

The Arabidopsis mitochondrial genome was screened for additional occurrences of sequence motifs coinciding with experimentally defined transcriptional starts. Of the genes displaying a promoter motif in their upstream regions, rrn26, atp1, atp6-1, atp6-2 and atp8 were selected for an experimental verification of their predicted transcriptional starts. Notably, the atp1 and atp6-1 5’ regions like atp9 displayed the motif CGTATATAA approximately 50 base pairs upstream of the hypothetical CATAAGAGA promoter sequence. The analysis moreover included tRNA-fMet with the predicted promoter motif CGTAAGAGA (Dombrowski, et al., 1998), which had been found to be an element of the atp9 promoter in pea and soybean (Binder, et al., 1995;Brown, et al., 1991). Of those genes not possessing any conserved promoter motif upstream of their coding sequence, rps3 and cox1 were selected for transcript 5’ end mapping.

Figure 11: 5’-RACE analysis of the mitochondrial rrn18, rps3 and atp6-1 transcripts.

Amplified products were separated on agarose gels alongside molecular weight markers; sizes are given in nucleotides (marker lane not displayed). TAP-specific products (lane +T) that correspond to primary transcript 5’ ends are indicated by arrows and labelled with the name of the respective promoter as listed in Table 9. Initiation at Patp6-1-156 and Patp6-1-200 required confirmation through ribonuclease protection analysis of cap-labelled transcripts (compare Figure 12). Control experiments were done by 5’-RACE from RNA mock-treated in TAP buffer without TAP (lanes –T).

Table 9 provides a summary of determined transcription start sites and their surrounding sequences. Besides 5’ ends that were unambiguously identified as primary ends in the 5’-RACE, such as those mapping to Prrn26-893 and Prps3-1053, various 5’ termini were detected for which 5’-RACE products were not enhanced following TAP treatment of RNAs but which, as described above for Pcox2-210, nevertheless coincided with genomic sequences exhibiting strong similarity to bona fidepromoters. For these transcripts the pools of 5’ termini cloned from +TAP and from –TAP samples were again compared (fourth and fifth column in Table 9). Mostly, the –TAP pool contained slightly shorter transcripts than the +TAP pool (for examples, compare sizes of +TAP and –TAP 5’-RACE signals obtained for Patp6-1-916/913 and Pcox2-683 in Figure 11 and Figure 8, respectively), and particularly was deprived of longer transcript species that at their 5’ extremities carried A or G nucleotides (see Table 9). These transcripts specific to the +TAP pool are most likely resulting from transcription initiation. Within three promoter regions located upstream of the tRNA-fMet, atp6-1 and atp8 genes, transcription was found to initiate at two different nucleotide positions. From 5’-RACE results it is likely that multiple initiations also occur around Prps3-1133 (data not shown). Multiple promoters were detected for all investigated genes except rrn26, cox1 and orf291. Due to partly identical upstream and coding sequences of cox2 and orf291,the transcriptional start site preceding orf291 was fortuitously found using primers annealing to the cox2 upstream region.

To analyze possible differences in promoter utilization between Arabidopsis leaves and flowers, TAP-specific 5’-RACE signals (lanes +T in Figure 8 and Figure 11) that had been obtained from leaf and from flower RNA for a distinct gene were compared. No primary transcript 5’ end was detected that was exclusively present in leaves or in flowers, indicating that transcription is initiated at identical sites in both tissues. An occasional enhancement of 5’-RACE signals from flower RNA can be attributed to the level of mitochondrial activity being generally higher in flowers than in green tissues (Huang, et al., 1994;Smart, et al., 1994).

As already observed for atp9, 5’-RACE analyses of those atp1, atp6-1, atp6-2 and atp8 transcript 5’ termini mapping to the motifs CATAAGAGA and CGTATATAA did not support transcription initiation at these sequences (Table 9 and Figure 11). The perfect nonanucleotide motifs found at these sites prompted the examination of the corresponding 5’ ends by an independent technique. As a method specifically detecting primary 5’ ends, ribonuclease protection of in vitro-capped transcripts was employedto analyze the respective 5’ termini of the atp9 and atp6-1 mRNAs. This method takes advantage of organellar transcripts being, unlike nuclear mRNAs, not capped at their 5’ ends in vivo. Mitochondrial primary transcripts, which carry 5’ triphosphates, are thus representing guanylyltransferase (capping enzyme) substrates and can be 5’ cap-labelled with the GMP moiety of [³²P]—α-GTP in vitro. Total Arabidopsis RNA was capped and then subjected to ribonuclease protection using RNA probes complementary to the genomic regions containing putative promoters. The rrn26 primary transcript was included as a positive control in the capping study, since its 5’ end had been established by 5’-RACE to map to a promoter that is identical to the sequence surrounding the predicted transcriptional start Patp1-1947, and moreover is highly similar to the hypothetical promoters Patp6-1-156, Patp6-2-148, Patp8-157 andPatp9-239. Additionally, the rrn18 transcript 5’ ends coinciding with positions -69 and -156 were tested for their ability to be capped in vitro.

Initiating nucleotides are underlined; repeatedly observed promoter cores are written bold and the frequent TATATA(A) motif is highlighted. The number of clones that were sequenced for each promoter is given together with the frequency of the respective primary transcript 5’ end as determined from TAP-treated flower RNA, and for selected promoters from flower RNA not exposed to TAP.
^a Consistent with primer extension results in Giese et al. (Giese, et al., 1996).
^b Consistent with previous predictions of Arabidopsis mitochondrial promoters (Dombrowski, et al., 1998).
^c Transcription initiation was found to occur at two different nucleotides in one promoter region; frequencies of transcript 5’ termini are given first for the upstream nucleotide.
n.d., not determined.

Gene

Promoter

Sequence

No. of clones

(+TAP)

No. of clones

(-TAP)

In vitro -cappable

rrn18

Prrn18-156

TAGAATAATACG TATATAATCAGAA

20/23

4/7

+

orf291

Porf291-307

TGGAATAATACG TATATAATCAGAT

7/10

n.d.

n.d.

atp6-1

Patp6-1-200

GCCAATAATACG TATATAAGAAGAG

3/14

n.d.

+

atp9

Patp9-295

CTGGTGCTCTCG TATATAAGAGAAG

8/8

10/11

+

atp1

Patp1-1947

CTGGTGGTATCG TATATAAGAGAGA

8/13

10/15

+

cox2

Pcox2-210

ATGTTGGTTTCG TATATAAGAAGAC

5/39

0/31

+

tRNA-fMet

PtrnM-98 ^b

TTTGAAATATCGTAAGAGAAGAAGG

12/12

n.d.

n.d.

rrn26

Prrn26-893 ^b

CTATCAATTTCATAAGAGAAGAAAG

12/13

0/23

+

atp1

Patp1-1898 ^b

CTATCAATTTCATAAGAGAAGAAAG

13/13

n.d.

+

atp9

Patp9-239 ^b

CTATCAATTTCATAAGAGAAGACGA

21/21

12/13

+

atp6-1

Patp6-1-156 ^b

CTATCAATCTCATAAGAGAAGAAAT

5/14

n.d.

+

atp6-2

Patp6-2-148 ^b

CTATCAATCTCATAAGAGAAGAAAT

7/13

n.d.

+

atp8

Patp8-157 ^b

CTATCAATCTCATAAGAGAAGAAAT

14/22

n.d.

n.d.

rrn18

Prrn18-69 ^a

AGTGGAATTGAATAAGAGAAGAAAG

n.d.

6/8

+

atp8

Patp8-999

ATAAAATTAAATAAAGAGCAAAAAT

9/12

n.d.

n.d.

atp8

Patp8-228/226^c

CATACCATAACA TATATAGAATCGA

1/28, 6/28

0/14, 0/14

n.d.

rrn18

Prrn18-353

TACTTTTCCATCTATATAAAATGAA

10/12

n.d.

n.d.

atp6-1

Patp6-1-916/913 ^c

AGCCCTTTATAT TATATAATAAAGC

2/23, 11/23

0/23, 1/23

n.d.

cox1

Pcox1-355

AATTTATTCAAT TATATAATAATAA

18/23

19/30

n.d.

cox2

Pcox2-481

ATGAATATTCATTAGATAATAGATT

13/43

1/34

n.d.

rps3

Prps3-1133

TAGAAAAAATTATTAGTAATACGTA

6/26

0/15

n.d.

rrn18

Prrn18-424

TCAAATCCTCGG TATATAAAGAGAA

9/10

n.d.

n.d.

cox2

Pcox2-683

GACACGTAAGGTAAAATAAGAATCT

6/12

0/8

n.d.

rps3

Prps3-1053

TTTTTTATTTGGTAGGTAACATCGC

12/14

n.d.

n.d.

atp9

Patp9-487

ATGTCTTATTGGTATGTGATACAAG

13/14

n.d.

n.d.

atp9

Patp9-652

AGAAGATTGAAGTAAGGAGCAGGTT

7/16

0/29

n.d.

atp6-2

Patp6-2-436

TCTTGAATTAAG TATATAGAAAAGA

5/20

n.d.

n.d.

atp6-2

Patp6-2-507

GATAAATTAAGTATAGTAATAAGAA

9/12

n.d.

n.d.

atp8

Patp8-710

ATCGGAGCTGCCAATAAGCTAATCC

4/12

0/13

n.d.

tRNA-fMet

PtrnM-574/573 ^c

CTAATTTATATAAAAAAGACCGGGA

9/18, 9/18

n.d.

n.d.

Table 9: Transcription initiation sites detected by 5’-RACE and in vitro-capping.

Figure 12: Detection of selected rrn18, rrn26, atp6-1 and atp9 primary transcript 5’ ends by ribonuclease protection of cap-labelled RNA.

(A) Protected RNA fragments were separated in polyacrylamide gels alongside a molecular weight marker (lane M); sizes are given in nucleotides. Lane C shows total capped RNA prior to ribonuclease protection. Lanes designated rrn18, rrn26, atp6-1 and atp9 show ribonuclease protection results obtained with riboprobes complementary to the rrn18, rrn26, atp6-1 and atp9 upstream regions as detailed in Figure 12B. Specific protected fragments, which correspond to primary transcripts, are indicated by arrows and labelled with the respective promoter name. Asterisks mark fairly strong signals that were considered non-specific, as they were seen with different riboprobes. (B) Diagram of the rrn18, rrn26, atp6-1 and atp9 5’-untranslated regions. Promoters identified in 5’-RACE and capping analyses are indicated by bent arrows; open triangles mark processing sites identified by 5’-RACE. The beginning of the mature rRNA or protein-coding sequence is indicated by a hatched bar. Grey bars are drawn below sequences complementary to the riboprobes that were annealed to cap-labelled primary transcripts in ribonuclease protection assays. The sizes of expected protected RNA fragments are given in nucleotides.

Figure 12A shows the protected cap-labelled RNAs corresponding to the transcription initiation sites Prrn26-893, Prrn18-69 and Prrn18-156, and to the tandem promoters
Patp6-1-156 and Patp6-1-200, and Patp9-239 and Patp9-295. The sizes of protected RNA fragments are in accordance with the expected lengths of transcript 5’ segments annealing to RNA probes, as illustrated in Figure 12B. Most notably, Prrn18-69 as well as Patp6-1-156, Patp6-1-200, Patp9-239 and Patp9-295, which through 5’-RACE could not be confirmed as transcription initiation sites, were found to coincide with in vitro-cappable and thus primary RNA 5’ termini. Capping of the atp9 mRNA mapping to position -295 only yielded a very faint signal, which may be either because of rare utilization of Patp9 -295 as a promoter, or due to rapid in vivo processing of this primary message. In vitro capping moreover verified transcription initiation at Patp1-1898, Patp1-1947 and Patp6-2-148 (data not shown). The analyses of in vitro-cappable atp1, atp6-1, atp6-2 and atp9 mRNAs allow to infer that the transcript 5’ terminus coinciding with Patp8-157 is also derived from transcription initiation.

Table 9 aligns Arabidopsis mitochondrial promoter sequences with respect to experimentally defined transcription start sites and places promoters with similar core sequences in adjacent rows. At positions -7 to -4 with respect to the transcriptional start, the majority of promoters display the previously described core element CRTA (R = A or G) (Fey and Marechal-Drouard, 1999), which here is almost always seen as part of the nonanucleotide motifs CGTATATAA or CATAAGAGA, or the sequences ATTA, AGTA, GGTA or AATA. Only in a few promoters is the distance between core element and start site altered by one base pair.

All primary transcripts characterized in this study originate from transcription initiation at either an A or a G nucleotide (21 and 12 out of 33 start sites, respectively). When comparing nucleotide frequencies within promoters, it appears that A and G start sites favour distinct nucleotides at adjoining positions. For example, while a G as initiating nucleotide is nearly always preceded by an A, initiation at an A essentially requires a T at position -1. Due to these constraints on nucleotide frequencies particularly at positions around transcription initiation sites, promoter sequences were realigned in two subsets. The two alignments of
promoters driving transcription from an A or a G are illustrated in Figure 13 as sequence logos. Among promoters directing initiation at an A, the sequence element TATATA(A) is fairly frequent. Promoters having a G nucleotide at position +1 mostly conform to the consensus CRTAAGAGA that has been suggested previously for dicot mitochondrial promoters (Binder, et al., 1996).

Figure 13: Summary of nucleotide sequences around experimentally defined transcription initiation sites in Arabidopsis mitochondria, as displayed in Table 9.

Two sequence logos are shown that were generated using WebLogo [( , , (Crooks, et al., 2004;Schneider and Stephens, 1990)] from an alignment of 20 promoter sequences activating transcription initiation at an adenine nucleotide (upper sequence logo) and from an alignment of 11 sequences supporting initiation at a guanine nucleotide (lower sequence logo). Position +1 corresponds to the transcriptional start.

To gain an approximate idea of genome-wide promoter distribution in Arabidopsis mitochondria, the mtDNA of this plant was computationally screened for additional occurrences of sequences that are identical to mitochondrial promoter elements determined in the present study. Sequence stretches of various promoters that extended from the initiation site to the core tetranucleotide were used as query (compare Table 9). Potential promoters were detected not only upstream of annotated ORFs but also at sites preceding non-coding sequences of several kilobases and on complementary strands of identified genes. To test whether promoters are active in the Arabidopsis mitochondrial genome that direct the synthesis of presumed non-coding or antisense transcripts, selected motifs emerging from the in silico search were tested for promoter function by 5’-RACE as described in III.1.1.

Figure 14 summarizes the mapping of 5’ termini of antisense transcripts made to different nad genes; sequences at transcription initiation sites are given in Table 10. Intron 3 of nad4 and intron 4 of nad5 harbour sequences reminiscent of the Prps3-1133 promoter on their antisense strands. A nad5 antisense transcripts initiated at this motif (designated Pnad5-AS) was easily amplified from Arabidopsis RNA, whereas the nad4 antisense transcript mapping to a Prps3-1133-like promoter (designated P2nad4-AS) was weak. A far more abundant 5’-RACE product was derived from a nad4 antisense transcript mapping to a Prps3-1053-like sequence on the complementary strand of intron 3 (P1nad4-AS). A prominent signal was also obtained for a nad1 antisense transcript synthesized from a Pcox1-355-like region (P1nad1-AS). An additional nad1 antisense transcript was detected that is initiated from the complementary strand of intron 4 at a sequence comprising a CRTA element (P2nad1-AS). The promoter motif CGTATATAA (compare Table 9) is present in antisense orientation within exon 5 of the nad2 gene and indeed was found to promote transcription (Pnad2-AS). A nad7 antisense RNA mapping to a region on the intron-2 complementary strand that contained typical promoter elements (P2nad7-AS) gave rise to an abundant 5’-RACE product. Thus, several active promoters of the Arabidopsis mtDNA drive the synthesis of antisense transcripts to known genes.

Initiating nucleotides are underlined; typical promoter elements are written bold. The number of clones that were sequenced for each promoter is given together with the frequency of the respective primary transcript 5’ end as determined from TAP-treated RNA.
^a Transcription initiation was found to occur at different nucleotides in one promoter region; frequencies of transcript 5’ termini are given first for the upstream nucleotide.

Promoter

Sequence

No. of clones (+TAP)

P1nad1-AS

GAGAAATACCTTATTATATATATATAA

3,3,4/12 ^a

P2nad1-AS

CAACTAATCTCATAAGTAAACGCCT

3/4

Pnad2-AS

TTTCACTAAGCGTATATAATAAAAT

9/12

P1nad4-AS

TATCATGGTGAAGGTAA G GT A ACGC

4,4/11 ^a

P2nad4-AS

GCCTTTATTAGTAAAGT A AAGCTTT

3/6

Pnad5-AS

TTCTCTATTATATTAGT A AAGGGAA

10/11

Pnad7-AS

ACGTAAGAACTAGTATTGAAAGCTA

4,8/12 ^a

P_38K-nc

TCGATAATATCGTAAGAGAAGAAAA

11/11

P_203K-nc

CCATCTATTTCATAAGAGAATAAAA

11/12

Table 10: Transcription initiation sites associated with antisense and non-coding transcripts.

Of the promoter-like sequences not preceding identified or hypothetical ORFs, a PtrnM-98-like motif (P_38K-nc) which would drive initiation at nucleotide 38411 from the reverse strand and a Prrn26-893-like motif (P_203K-nc) around nucleotide 203267 on the direct strand of the Arabidopsis mtDNA were tested by 5’-RACE for promoter function (Figure 9 and Table 10). Transcripts mapping to the predicted transcriptional starts were amplified for both promoters, indicating that regions of the Arabidopsis mtDNA that lack known genes are indeed expressed from individual promoters.

Figure 14: Synthesis of antisense transcripts to mitochondrial nad genes.

Agarose gel analyses of 5’-RACE products derived from nad antisense transcripts (right; size marker and lane designation as in Figure 8) are displayed together with diagrams illustrating nad gene structures and positions of promoters driving antisense RNA synthesis (left). 5’-RACE signals corresponding to promoters are indicated by arrows beside gels and are specified according to Table 10; asterisks mark RNAs identified by sequencing to represent non-specific products. In nad gene diagrams, orange boxes and horizontal grey lines mark exons [exon assignment as in (Unseld, et al., 1997)] and cis-spliced introns respectively; the absence of horizontal grey lines between exons indicates trans-splicing (compare Figure 2 for nad exon distribution on the Arabidopsis mtDNA). Bent arrows symbolize promoters; gene fragments corresponding to amplified antisense transcripts are represented by bold black lines.

Figure 15: Synthesis of transcripts from intergenic regions of the Arabidopsis mtDNA.

(A) Location of promoters P_38K-nc and P_203K-nc driving transcription of non-coding sequences; mtDNA coordinates (in nt) are indicated. Orange arrows denote functional genes and are labelled accordingly; grey arrows correspond to hypothetical ORFs. Bent arrows symbolize promoters; gene fragments corresponding to amplified transcripts are represented by bold black lines. (B) Agarose gel analysis of 5’-RACE products derived from transcripts initiated at P_38K-nc and P_203K-nc (size marker and lane designation as in Figure 8). Signals corresponding to promoters are indicated by arrows beside gels and are specified according to Table 10.

5’-RACE products displayed in Figure 14 and Figure 15 were obtained using RNA isolated from Arabidopsis flowers; essentially the same signals were obtained for leaf transcripts (data not shown).

The present study aims at dissecting the roles of the two phage-type RNA polymerases present in Arabidopsis mitochondria by characterizing the transcriptional performances of RpoTm and RpoTmp in vitro. Specific transcription initiation at mitochondrial promoters may require complementing recombinant RpoT enzymes with as yet unidentified auxiliary factors (see I.3.3). Therefore, the Arabidopsis genome was screened for sequences encoding candidate cofactors of phage-type RNA polymerases.

BLAST searches using the amino acid sequence not of mtTFB from S. cerevisiae but of a putative Schizosaccharomyces pombe mtTFB homologue as query have previously lead to identifying nuclear genes encoding the transcriptional cofactors mtTFB1 and mtTFB2 of human mitochondria (Falkenberg, et al., 2002;McCulloch, et al., 2002). Hence, the Arabidopsis genome and ESTs were queried with the putative S. pombe mtTFB amino acid sequence (McCulloch, et al., 2002) and the human mtTFB1 and mtTFB2 sequences (Falkenberg, et al., 2002) using the blastp and tblastn algorithms available at the National Centre for Biotechnology Information. Three mtTFB-like dimethyladenosine transferases are predicted to be encoded by the loci At5g66360, At2g47420 and At1g01860 (see I.3.3.1 for details on the structural similarity of yeast and animal mitochondrial transcription factors to rRNA dimethyladenosine transferases), of which the latter corresponds to the previously characterized PFC1 gene coding for a plastidial 16S rRNA dimethylase (Tokuhisa, et al., 1998). BLAST searches of the Arabidopsis genome using any of the three sequences as query did not deliver additional hits. No sequences encoding Arabidopsis mtTFA homologues were identified by screening the database for putative mitochondrial HMG box proteins.

The methyltransferase-like genes at loci At5g66360 and At2g47420 were tentatively designated MetA and MetB (methyltransferase-like), respectively. While MetB is predicted to encode a 353-amino acid polypeptide, available EST data support alternative splicing of the hypothetical mRNA deriving from MetA, which would give rise to two different polypeptides of 352 and 380 amino acids (GenPept accession numbers NP_201437 and NP_975003). However, PCR amplification of the MetA and MetB coding sequences from cDNAs yielded only the longer of the two predicted products for MetA and a fragment of the expected length for MetB (data not shown). Protein sequence comparisons revealed that the presumed optional MetA intron codes for amino acid sequence motifs that are conserved among rRNA dimethylases-like proteins (amino acids 214-241 of the derived MetA polypeptide, see alignment in Annex A). Hence, further sequence analyses were based on the longer deduced MetA polypeptide, and the amplified MetA and MetB cDNAs were used to construct plasmids for MetA and MetB expression in E. coli (see III.2.4).

Both MetA and MetB display ~30% and ~27% amino acid sequence similarity to mtTFB from S. pombe and S. cerevisiae, respectively; ~15% and ~12 % of positions are identical (sequences exclusive of the predicted transit peptides were compared). These values approximately correspond to those obtained from comparisons of human to fungal mtTFB sequences. Similarities of MetA and MetB to each h-mtTFB1 and h-mtTFB2 are ~37%; identities to the human sequences are ~20% for both MetA and MetB.

^a http://www.cbs.dtu.dk/services/TargetP (Emanuelsson, et al., 2000;Nielsen, et al., 1997)
^b http://www.mips.biochem.mpg.de/cgi-bin/proj/medgen/mitofilter (Claros and Vincens, 1996)
(P, probability of mitochondrial targeting)
^c http://www.inra.fr/Internet/Produits/Predotar/
^d http://genoplante-info.infobiogen.fr/predotar/predotar.html (Small, et al., 2004)
^e http://psort.ims.u-tokyo.ac.jp/form.html
^f http://hc.ims.u-tokyo.ac.jp/iPSORT/ (Bannai, et al., 2002)
^g Physicochemical properties were derived using ProtParam (http://expasy.cbr.nrc.ca/tools/protparam.html) from MetA and MetB amino acid sequences lacking the predicted transit peptides.

MetA

MetB

Locus tag

At5g66360

At2g47420

Subcellular localization

TargetP 1.1^a

mitochondrial

not mitochondrial or plastidial

Mitoprot^b

mitochondrial (P=0.59)^b

non-mitochondrial (P=0.25)^b

Predodar 0.5^c

mitochondrial

mitochondrial

Predodar 1.03^d

possibly mitochondrial

not mitochondrial or plastidial

PsortII^e

mitochondrial

nuclear

iPsort^f

mitochondrial

not mitochondrial or plastidial

N-terminal transit peptide

TargetP1.1^a

26 aa

-

Mitoprot^b

27 aa

17 aa

PsortII^e

27 aa

-

Mature protein ^g

Length

354 aa

335 aa

Molecular weight

40 kDa

38 kDa

pI

8.6

8.5

Table 11: Predicted properties of Arabidopsis MetA and MetB.

Several computer algorithms unambiguously predicted MetA to possess an N-terminal transit peptide mediating the import of the protein into mitochondria (Table 11). In contrast, mitochondrial localization of MetB was supported only by the older, less stringent version of Predotar, and according to Psort, MetB might be a nuclear protein (Table 11). Calculated physicochemical properties of the deduced MetA and MetB proteins are given in Table 11.

To experimentally investigate the potential of the MetA and MetB N-termini to function as mitochondrial transit peptides, nucleotide sequences encoding the 64 or 57 N-terminal amino acids of MetA or MetB respectively were fused in-frame to the green fluorescent protein (GFP) coding sequence (see II.8.1). Tobacco protoplasts were transformed with the MetA- and MetB-GFP fusion constructs, and transient expression of the fusion proteins was monitored using fluorescence microscopy. Two plasmids encoding mitochondrial CoxIV-GFP and plastidial RecA-GFP fusion protein were used for reference transformations of tobacco protoplasts.

Figure 16: Transient expression of GFP fusion proteins in tobacco protoplasts.

The MetA and MetB gene fragments encoding putative transit peptides were inserted into plasmid pOL-GFPS65C (Peeters, et al., 2000) to generate vectors driving the expression of MetA-GFP- and MetB-GFP. The control constructs encoding mitochondrial CoxIV-GFP and plastidial RecA-GFP (Peeters, et al., 2000) were kindly provided by I. Small (INRA CNRS, Evry, France). Images were taken by epifluorescence microscopy using a GFP filter (top panels) or a FITC filter set (bottom panels).

Protoplasts expressing MetA-GFP displayed green fluorescence of small structures resembling the fluorescent mitochondria of protoplasts synthesizing CoxIV-GFP (Figure 16), substantiating a mitochondrial localization of MetA. MetB-GFP fluorescence, on the other hand, for the most part enveloped a large round structure, which likely corresponded to the nucleus, and was moreover distributed over the surfaces of cell organelles such as chloroplasts (as inferred from red chlorophyll autofluorescence in Figure 16, bottom panels). GFP distribution in MetB-GFP-expressing protoplasts pointed to a cytoplasmic localization of the fusion protein, indicating that MetB does not possess a mitochondrial (or plastidial) transit peptide. According to import experiments and targeting predictions (Figure 16 and Table 11), MetB may be a cytoplasmic or nuclear protein. By fusing GFP to only the N-terminal portion of MetB, possible nuclear targeting signals may have been removed.

In order to assess the phylogenetic relationships of Arabidopsis MetA, MetB, and of the plastidial methyltransferase Pfc1 to established mitochondrial transcription factors such as yeast and animal mtTFBs and to other rRNA dimethylases such as E. coli KsgA (see I.3.3.1), the MetA, MetB and Pfc1 sequences were compared to available mtTFB sequences and to a set of sequences of characterized and predicted rRNA dimethylases. The latter included all rRNA dimethylase-like ORFs that could be retrieved from the fully sequenced genomes of Arabidopsis, Populus trichocarpa, O. sativa and H. sapiens, as well as additional rRNA dimethylase sequences available from organisms with characterized mtTFBs. Sequence retrieval was done as described in II.9. For plant sequences, a subcellular targeting prediction was performed using the TargetP, Predotar, Mitoprot, Psort and iPsort algorithms (Annex B). Sequences with highest similarity to Arabidopsis MetA were entirely found to have putative mitochondrial transit peptides, and are thus designated MetA in the phylogenetic tree. Sequences best aligning to Arabidopsis MetB were predicted to be neither plastidial nor mitochondrial and are referred to as MetB. The P. trichocarpa sequence designated Pt-Pfc1 was calculated to comprise an N-terminal plastidial transit peptide and show highest similarity to Arabidopsis Pfc1.

Figure 17: Phylogeny of mitochondrial transcription factors and small-subunit rRNA dimethylases.

The phylogeny was reconstructed by Bayesian estimation from conserved amino acid sequence sections indicated in the amino acid sequence alignment in Annex B. The clusters comprising plant MetB proteins and yeast and animal Dim enzymes are shaded grey in the comprehensive phylogram under (A) and are shown in detail under (B). Clades shaded pale yellow comprise mitochondrial or predicted mitochondrial proteins; plastidial components are highlighted green. Prefixes of designations of plant MetA-like, MetB-like and Pfc1-like proteins refer to Arabidopsis thaliana (At), Glycine max. (Gm), Lycopersicon esculentum (Le), Medicago truncatula (Mt), Oryza sativa (Os), Populus trichocarpa (Pt), Zea mays (Zm). Fungal and animal protein names adhere to the nomenclature introduced in I.3.3; prefixes refer to Kluyveromyces lactis (kl) Saccharomyces cerevisiae (sc), Saccharomyces kluyveri (sk), Schizosaccharomyces pombe (sp), Drosophila melanogaster (d), Homo sapiens (h), Mus musculus (m), Rattus norvegicus (r), Xenopus laevis (x). EcKsga and PaeKsgA are 16 S rRNA dimethylases from Escherichia coli and Pseudomonas aeruginosa respectively. Numbers at branching points are posterior branch support values. Branch lengths correspond to the number of inferred amino acid changes per position, as indicated by scale bars.

Amino acid sequences were compared using the Multalin algorithm (Corpet, 1988), and the alignment was refined according to a structure-based alignment generated for fungal mtTFBs and the Bacillus subtilis rRNA dimethylase ErmC’ (Schubot, et al., 2001); see
Annex A for the alignment). Based on the alignment sections that are indicated in Annex A, a Bayesian phylogenetic tree was derived (Figure 17). Essentially the same tree topology was obtained employing maximum likelihood or maximum parsimony analysis (data not shown). Arabidopsis MetB and its plant orthologues appear to be most closely related to a group of rRNA dimethylases containing the yeast nucleolar 18S rRNA dimethylase Dim1.They may therefore represent nuclear or cytoplasmic enzymes, which would be consistent with computational predictions of the subcellular localization of these proteins and with GFP import experiments conducted for Arabidopsis MetB.A sister group to plant, fungal and animal MetB/Dim1-like methyltransferases is formed by predicted mitochondrial rRNA dimethylases of plants including Arabidopsis MetA. Notably, amino acid sequence similarities of Arabidopsis MetA and MetB to yeast Dim1 are 54% and 68% respectively, and considerably exceed similarities to mtTFBs (compare III.2.1). This is contrasted by similarities of sc-mtTFB, h-mtTFB1 and h-mtTFB2 to yeast Dim1 of only 36%, 35% and 28%. Arabidopsis Pfc1 and its poplar orthologue compose a distinct group apart from the Dim1/MetA cluster and from three other well-separated groups formed by animal mtTFB1s, animal mtTFB2s, and the highly diverse fungal mtTFBs. The phylogram shows that plant mitochondrial rRNA dimethylases are decidedly more closely related to nuclear/cytoplasmic enzymes of this type than to fungal and animal mtTFBs. In the absence of any other Arabidopsis mtTFB candidates it was decided to further characterize MetA as a potential cofactor of mitochondrial transcription, and to take along the putative Dim1 orthologue MetB for control experiments.

Yeast mtTFB as well as human mtTFB1 have been described previously to bind to mtDNA sequences in a non-specific manner (McCulloch, et al., 2002;Riemen and Michaelis, 1993). Recombinant MetA was prepared in order to assay the protein for a similar DNA-binding activity, and to moreover test the protein in vitro for a possible function as cofactor of mitochondrial transcription in Arabidopsis. Using the pPROTet.E expression vector (Clontech), MetA was engineered for expression in E. coli as fusion protein carrying an N-terminal hexahistidine tag but lacking the 26 N-terminal amino acids of the methyltransferase, which correspond to the predicted transit peptide and may be expected to not be part of the mature, functional protein. MetB lacking the 17 N-terminal amino acids was in the same way prepared for expression in E. coli so as to have a non-mitochondrial rRNA dimethylase-like protein available for control experiments.

Figure 18: Purification of recombinant MetA and MetB.

MetA and MetB lacking the predicted transitpeptides (Table 11) and fused to N-terminal hexahistidine tags were expressed in E. coli, purified over Ni²⁺-NTA agarose and analyzed by SDS-PAGE followed by Coomassie-blue staining of the gel (left panel). Samples were run alongside a molecular weight marker; sizes are indicated in kDa (marker lane not displayed). Proteins of the expected sizes were found to be enriched, and their identity was confirmed by Western blotting and immunolabelling with an anti-polyhistidine antibody (right panel).

Following expression of the recombinant proteins in the bacterial host, soluble MetA and MetB were enriched from E. coli extracts through a Ni²⁺-NTA agarose purification step (see II.5.3.3). Figure 18 displays an SDS-PAGE analysis of the purified proteins. Two major bands migrated as expected for MetA and MetB, and were confirmed to correspond to the recombinant proteins by immunolabelling with an anti-polyhistidine antibody (Figure 18).

Recombinant MetA was tested for DNA-binding activity in an electrophoretic mobility shift assay as described previously for human mtTFB1 (McCulloch, et al., 2002). Two different double-stranded mtDNA fragments that contained either Patp9-239 or Patp9-295, each representing frequent mitochondrial promoter types in Arabidopsis, were radiolabelled and supplied as target DNA in the binding assay. Binding reactions were subsequently characterized by native PAGE (Figure 19). The addition of MetA to both DNA fragments lead to a mobility shift of the labelled DNA (Figure 19, left panel). This effect was abolished when minor amounts of the non-specific competitor polynucleotide poly(dI-dC) were present in the binding reaction, indicating that the observed DNA binding by MetA is not DNA sequence-specific. Addition of an unlabelled competing mtDNA fragment that did not contain a mitochondrial promoter sequence similarly eliminated the band shift (data not shown). DNA binding by MetB was indistinguishable from the DNA-binding activity of MetA (Figure 19, right panel). To ensure that the observed DNA-protein complex formation was indeed due to MetA or MetB rather than caused by residual factors from the E. coli expression host, E. coli extracts from cells containing the empty expression vector pPROTet.E were subjected to the Ni²⁺-NTA agarose purification procedure and subsequently assayed for DNA binding. No band shifts were observed with these fractions, indicating that the DNA-binding activity of the MetA and MetB preparations is indeed due to the recombinant proteins (data not shown). Not only MetA but also MetB behaved like human mtTFB1 in the DNA-binding assay (compare (McCulloch, et al., 2002). The proteins were further characterized in in vitro transcription studies (see III.4).

Figure 19: Gel mobility shift competition assay showing non-specific DNA binding by MetA and MetB.

(A) Linear diagram of the atp9 upstream region. Grey lines mark the positions of the two mtDNA fragments used as probes in the gel mobility shift assay; parts of the comprised promoter sequences are indicated. Positions of transcriptional starts and processing sites are given as negative distances (in base pairs) from the translational start equalling position +1, which is highlighted by a filled circle; other symbols as in Figure 12. (B) PAGE analysis of DNA binding by MetA and MetB. Binding reactions were set up with 50 or 100 ng of recombinant MetA (left panel) or MetB (right panel) and with 5’ end-labelled mtDNA fragments containing the Patp9-295 or the Patp9-239 promoter as denoted above the autoradiographs. The addition of increasing amounts (5 and 10 ng) of the non-specific competitor poly(dI-dC) to binding reactions is indicated below. Signals corresponding to DNA-protein complexes (C) and unbound DNA (F) are marked.

Chapter III.1 describes the mapping of transcription initiation sites in the mitochondrial genome of Arabidopsis. The knowledge of mtDNA sequences that are recognized as promoters by the transcription machinery in vivo enabled an in vitro system initiating transcription at Arabidopsis mitochondrial promoters to be set up. To study transcription from mitochondrial promoters in vitro, DNA templates were constructed by inserting promoter regions of the Arabidopsis mtDNA into pKL23 (Liere and Maliga, 1999) upstream of the two bacterial ρ-independent terminator sequences hisa and thra(Barnes and Tuley, 1983;Gardner, 1982) that are present in pKL23 (Figure 23A, Figure 25A, and Figure 27). RNA synthesis driven by a plant organellar phage-type RNA polymerase has been described earlier to efficiently stop at hisa and thra(Liere and Maliga, 1999). When providing a circular pKL23 derivative as template in run-off experiments, transcription initiated at the introduced promoters should thus be terminated at hisa and/or thra and/or at downstream cleavage sites on an EcoRI- or XhoI-linearized plasmid (compare Figure 23A, Figure 25A, and Figure 27), thereby generating RNA products of distinct lengths.

Recombinant Trx-(His)₆-tagged RpoTm and RpoTmp were assayed for transcription initiation at several mitochondrial promoters in the presence or absence of the mitochondrial mtTFB homologue MetA. In order to discern non-specific effects of MetA addition, control reactions were set up with the presumably non-mitochondrial MetA homologue MetB. The experimental design essentially followed that described by Falkenberg et al. (2004) for a human mitochondrial in vitro transcription system reconstituted from individual recombinant components (see Materials and Methods). Recombinant proteins were prepared for application in in vitro transcription assays as described in sections III.3 (RpoT) and III.2.4 (Met).

Plasmid pKL23-atp6-1-A containing the promoters Patp6-1-156 and Patp6-1-200 in tandem was selected as DNA template for initial in vitro transcription experiments (Figure 23A; for promoter sequences, see Table 9). Both promoters give rise to fairly abundant in vitro-cappable 5’ ends (compare Figure 12) and were thus considered efficient in vivo promoters. Moreover, they differ in their architecture, thus making pKL23-atp6-1-A an ideal template for preliminary studies investigating the experimental conditions for correct promoter utilization in vitro.

core RNA polymerase. Unlike transcriptionally active preparations from plant mitochondria (Binder, et al., 1995;Hanic-Joyce and Gray, 1991;Rapp and Stern, 1992), RpoTm did not specifically transcribe linear DNA from Patp6-1-156 or Patp6-1-200 (data not shown, compare Figure 28). Based on a study reporting that a phage-type RNA polymerase activity isolated from tobacco plastids was dependent on supercoiled DNA as a template in vitro (Liere and Maliga, 1999), and on previous observations that recombinant RpoTm and RpoTmp were considerably more active in the transcription of supercoiled compared to linear DNA (Kühn, 2001), the experiments were repeated using a circular, negatively supercoiled plasmid template.

Transcription of pKL23-atp6-1-A by RpoTm produced three major discrete RNA products of apparent lengths of approximately 300 and 370 nucleotides (Figure 23B; major transcripts are indicated by black arrows). Following their 5’ end characterization, these products could be attributed entirely to transcription initiation at Patp6-1-200 (see below). While the upper band resulted from transcription termination at thra, termination at hisa produced an RNA migrating as a double band. Since for all plasmid templates used in subsequent experiments, transcripts ending at hisa appeared as double bands, the latter presumably resulted from secondary structure formation at the transcript 3’ end or termination at two different adjacent sites at hisa. The high-molecular-weight signals that are visible at the top of the autoradiograph may be attributed to transcripts initiated non-specifically on the plasmid template at sequences other than the atp6-1 promoters and to RNAs not terminated at hisa and thra. Non-specific high-molecular-weight transcripts were likewise seen in in vitro transcription studies with mitochondrial extracts (Binder, et al., 1995;Hanic-Joyce and Gray, 1991;Rapp and Stern, 1992). The addition of equimolar amounts of MetA to RpoTm appeared to have no effect on RpoTm-driven transcription. No transcripts were made in reactions containing MetA or MetB but not RpoTm. It is thus unlikely that the transcripts seen in Figure 23B are due to an RNA polymerase activity from E. coli exploited as RpoTm expression host, as RpoTm, MetA, MetB preparations were made following essentially the same protocol and may be expected to contain similar residual E. coli contaminants. Subsequent in vitro experiments revealing non-identical transcription activities of RpoTm and RpoTmp (see below) supported that the observed in vitro RNA synthesis was indeed due to the recombinant enzymes.

Figure 23: In vitro run-off transcription of pKL23-atp6-1-A by RpoTm initiates at Patp6-1-200.

(A) pKL23-atp6-1-A was constructed by inserting a 300-bp fragment of the Arabidopsis mtDNA containing the promoters Patp6-1-156 and Patp6-1-200 (compare Figure 31) into pKL23 via the SacI and PstI restriction sites. The positions of a T7 promoter (P_T7) represented by a dark grey bar and of the bacterial attenuators hisa and thra symbolized by red bars are indicated. Red open circles denote sites of transcription termination within hisa and thra; bent arrows mark start points of transcription at Patp6-1-156, Patp6-1-200, and P_T7. Only the plasmid region between P_T7 and the XhoI cleavage site, including the mtDNA insert, is drawn to scale. Run-off products expected from initiation at Patp6-1-156 and Patp6-1-200 and termination at hisa and thra are indicated by horizontal black arrows labelled with the respective RNA length. Red arrows and numbers mark the positions of primers P2hisa (2) and P3hisa (3) employed for transcript 5’ end mapping. (B) RpoTm was assayed for promoter-specific transcription of supercoiled pKL23-atp6-1-A in the presence or absence of MetA or MetB as described in section II.7.2. [-³²P]-UTP-labelled RNA products were electrophoresed in a 5% sequencing gel alongside an RNA size standard; sizes are given in nucleotides. In vitro transcription reactions were supplemented with recombinant proteins (400 fmol each) as indicated above individual lanes. Major discrete RNA products are indicated by black arrows and, after transcript 5’ end mapping (Figure 24), were attributed to transcription initiation at Patp6-1-200 followed by termination at hisa (signals labelled P-200-hisa) and thra (signal P-200-thra) as detailed in the text. Minor signals indicated by grey arrows labelled with asterisks may be due to differently migrating major products.

From a comparison of expected and apparent transcript lengths (Figure 23), the discrete in vitro transcription products indicated in Figure 23B could be assigned neither to transcription initiation at Patp6-1-156 nor to initiation at Patp6-1-200. Discrepancies between expected and apparent sizes were observed for nearly all in vitro-synthesized transcripts and may be due to an altered migration behaviour of these products compared to the RNA size marker. In order to identify the 5’ termini of in vitro-synthesized transcripts, 5’-RACE was performed on RNA products made by RpoTm essentially as described in section III.1.1 (Figure 24, lane +L). As a control, in vitro transcription products not ligated to the RNA oligonucleotide were subjected to RT-PCR (Figure 24, lane -L), thereby allowing to distinguish RNA-derived PCR products from signals resulting from non-specific amplification of of sequences from the carried-over DNA template.

Figure 24: 5’-RACE analysis of pKL23-atp6-1-A-derived transcripts synthesized in vitro by RpoTm.

5’-RACE was performed on RNA linker-ligated transcripts (lane +L) and, as a control, on non-ligated transcripts (-L). PCR products were separated on an agarose gel alongside a molecular weight marker; sizes are given in base pairs (marker lane not displayed). The signal that corresponds to transcript 5’ ends mapping to Patp6-1-200 is indicated. The chromatogram below displays the sequence at the ligation site of a typical cloned 5’-RACE product; RNA linker and transcript portions of the sequence are indicated. The mtDNA sequence at Patp6-1-200 is displayed below; the bent arrow indicates the in vivo transcription initiation site.

Electrophoresis of 5’-RACE products showed a single band, indicating that the multiple transcripts synthesized from pKL23-atp6-1-A by RpoTm are due to transcription initiation at only one of the two promoters (Figure 24). Sequencing of the cloned PCR product revealed that the transcripts had a 5’ end identical to that of transcripts initiated at Patp6-1-200 in vivo, allowing to attribute the in vitro-synthesized RNAs to correct transcription initiation at this promoter (Figure 24). On the other hand, in vitro transcription and 5’-RACE experiments together provided no evidence for transcripts initiated specifically at Patp6-1-156. The minor signals indicated by grey arrows and marked with asterisks in Figure 23B were assumed to correspond to differently migrating major transcripts rather than additional defined RNA 5’ ends, as no discrete 5’-RACE products other than those resulting from initiation at Patp6-1-200 were detected.

Additional in vitro transcription experiments investigated the initiation by RpoTm at the promoters Prrn26-893 and PtrnM-98. While the sequence around the transcription initiation site of Prrn26-893 is nearly identical to that of Patp6-1-156, PtrnM-98 displays elements of both Patp6-1-156 and Patp6-1-200 (see Table 9). Supercoiled pKL23-rrn26 and pKL23-trnM gave rise to discrete RNA products (Figure 25), whereas no specific transcripts were seen with the linearized plasmids (data not shown). The transcript 5’ termini were found for both templates to accurately reflect those of transcripts generated in vivo from Prrn26-893 and PtrnM-98, respectively (Figure 26). Initiation at Prrn26-893 and PtrnM-98 was, however, considerably less efficient than at Patp6-1-200. As observed with pKL23-atp6-1-A, MetA had no effect on RpoTm-driven transcription from pKL23-rrn26 and pKL23-trnM.

Figure 25: In vitro transcription from PtrnM-98 and Prrn26-893 by RpoTm.

(A) pKL23-trnM and pKL23-rrn26 were constructed by inserting fragments of the Arabidopsis mtDNA containing the promoter PtrnM-98 or Prrn26-893 (compare Figure 31) into pKL23 via the SacI and PstI restriction sites. Symbols and illustration of expected run-off products as in Figure 23A. (B) RpoTm was assayed for promoter-specific transcription of supercoiled pKL23-trnM or pKL23-rrn26 in the presence or absence of MetA or MetB as indicated, and RNA products were analyzed as described above for pKL23-atp6-1-A. Major discrete RNA products are indicated by arrows and, after transcript 5’ end mapping (Figure 26), were attributed to transcription initiation at PtrnM-98 followed by termination at hisa (signals labelled P-98-hisa) and thra (signal P-98-thra), or initiation at Prrn26-893 followed by termination at hisa (P-893-hisa) and thra (P-893-thra).

The described in vitro assays showed RpoTm to specifically initiate transcription at three different mitochondrial promoters, Patp6-1-200, Prrn26-893 and PtrnM-98, from supercoiled DNA without the aid of auxiliary factors. The preference of promoters over random start sites varied between different promoter sequences.

Figure 26: 5’-RACE analysis of RNAs transcribed from PtrnM-98 and Prrn26-893 by RpoTm in vitro.

(A) Analysis of pKL23-trnM-derived RNAs. 5’-RACE was performed on RNA linker-ligated transcripts (lane +L) and, as a control, on non-ligated transcripts (-L). PCR products were separated on an agarose gel alongside a molecular weight marker (left); sizes are given in base pairs (marker lane not displayed). The signal corresponding to transcript 5’ ends mapping to PtrnM-98 is indicated. The chromatogram to the right displays the sequence at the ligation site of a typical cloned 5’-RACE product; RNA linker and transcript portions of the sequence are indicated. The mtDNA sequence at PtrnM-98 is displayed below; the bent arrow indicates the in vivo transcription initiation site. (B) Analysis of pKL23-rrn26-derived RNAs as described under (A). The 5’-RACE signal corresponding to transcript 5’ ends mapping to Prrn26-893 is indicated (left); the sequence at Prrn26-893 is displayed below the chromatogram of a typical cloned 5’-RACE product (right).

Arabidopsis mitochondria possess two phage-type RNA polymerases (Hedtke, et al., 1997;Hedtke, et al., 2000); yet no data have been provided so far that would define possible functional differences between RpoTm and RpoTmp. A series of experiments therefore compared the abilities of RpoTm and RpoTmp to initiate transcription in vitro at diverse mitochondrial promoter sequences. Figure 27 illustrates the design of plasmid templates that were constructed for this study and supplied as supercoiled or linearized templates (for promoter sequences, see Table 9).

Figure 27: Maps of in vitro transcription templates.

pKL23 derivatives were constructed by inserting Arabidopsis mtDNA fragments containing promoter sequences of the atp6-1, atp6-2, atp8, atp9, rrn18, and cox2 genes (compare Figure 31) into pKL23 via the SacI and PstI restriction sites. RNA molecules expected from transcription initiation at the introduced promoters followed by termination at hisa or thra or the restriction site used for plasmid linearization are exemplified for pKL23-atp6-1-A. Symbols and illustration of run-off products as in Figure 23A. The following RNAs may be expected from transcription of linearized or supercoiled pKL23 derivatives:

Template

Promoter

Cleavage site

Expected transcript length due to termination at

cleavage site

hisa

thra

pKL23-atp6-1-A

Patp6-1-156

EcoRI

130 nt

279 nt

349 nt

Patp6-1-200

174 nt

323 nt

393 nt

pKL23-atp8

Patp8-157

XhoI

309 nt

206 nt

276 nt

Patp8-228/226

380 nt

277 nt

347 nt

pKL23-rrn18

Prrn18-69

XhoI

295 nt

192 nt

262 nt

Prrn18-156

382 nt

279 nt

349 nt

pKL23-cox2

Pcox2-210

XhoI

290 nt

187 nt

257 nt

Pcox2-481

561 nt

458 nt

528 nt

pKL23-atp6-1-B

Patp6-1-916/913

-

-

369/366 nt

439/436 nt

pKL23-atp6-2

Patp6-2-436

-

-

350 nt

420 nt

Patp6-2-507

-

421 nt

491 nt

pKL23-atp9

Patp9-487

-

-

233 nt

307 nt

Patp9-652

-

398 nt

468 nt

In vitro transcription of pKL23-atp6-1-A confirmed initiation at Patp6-1-200 but not
Patp6-1-156 on supercoiled but not EcoRI-cleaved DNA by RpoTm (Figure 28). While non-specific RpoTmp-driven transcription of pKL23-atp6-1-A far exceeded RpoTm activity, no specifically initiated RNAs were seen in reactions with RpoTmp. Neither did RpoTmp recognize Patp6-1-156 or Patp6-1-200 on the linearized plasmid (data not shown). The presence of MetA appeared to have no effect on the activity or specificity of RpoTmp.

Figure 28: Run-off transcription from the atp6-1, atp8, rrn18 and cox2 upstream regions by RpoTm and RpoTmp.

RpoTm and RpoTmp were assayed for promoter-specific transcription of pKL23-atp6-1-A, pKL23-atp8, pKL23-rrn18 and pKL23-cox2. In vitro transcription reactions were run with RpoTm or RpoTmp and supercoiled (ccc) or linear (lin) DNA in the presence or absence of MetA (A) or MetB (B) as indicated above lanes, and RNA products were analyzed as described above. Transcripts made by RpoTmp were diluted 1:5 prior to PAGE due to RpoTmp showing a markedly higher activity than RpoTm. Discrete RNA products are specified as in Figure 23. Signals denoted P*-345-hisa and P*-345-thra are due to non-specific initiation on pKL23-cox2 (see text). Different background signal intensities observed for the same enzyme and template conformation are due to inherent experimental variation.

Exclusively non-specific transcription of linear templates was also observed in all subsequent in vitro transcription experiments for both RpoTm and RpoTmp (data not shown for RpoTmp). Modified experimental conditions, such as altered concentrations of monovalent ions or the catalytic Mg²⁺, did not enable RpoTm to initiate RNA synthesis at promoters located on linear templates or at promoters not recognized under the conditions of the standard in vitro transcription protocoll (see II.7.2; linear and supercoiled templates were purified using the same purification system). Neither RpoTm nor RpoTmp was found to be influenced in its activity by the addition of different amounts of MetA (only experiments in which equimolar amounts of RpoTm/RpoTmp and MetA were used are displayed).

Of the two mitochondrial promoters residing on pKL23-atp8, only Patp8-228/226 significantly supported specific transcription initiation by RpoTm while Patp8-157, which displays a sequence identical to Patp6-1-156 around the in vivo transcription start site (see Table 9), was apparently not recognized by RpoTm as a promoter (Figure 28). Enhanced transcription of pKL23-atp8 by RpoTm in the presence of MetA was considered irrelevant as a similar increase in RpoTm activity was seen following MetB addition. 5’-RACE performed on pKL23-atp8-derived RNAs confirmed the major defined transcripts to map to one out of two adenines in Patp8-228/226 used as initiating nucleotides by the mitochondrial transcription machinery in vivo (compare Table 9 and Table 12). Extensive sequencing of cloned 5’-RACE products identified minor discrete transcript 5’ ends mapping to Patp8-157 (Figure 29). However, these transcripts were not distinguishable from the background of non-specific products in the autoradiograph (Figure 28).

RpoTm-driven transcription of pKL23-rrn18 was found to efficiently and accurately initiate at Prrn18-156 but not Prrn18-69 in vitro (Figure 28 and Figure 29). The minor signals indicated by grey arrows and marked with asterisks in Figure 28 were assumed to correspond to differently migrating major transcripts rather than additional defined RNA 5’ ends, as they did not give rise to discrete 5’-RACE products (Figure 29). Minor signals at similar distances from major bands were likewise seen for the template pKL23-atp6-1 (Figure 23). Transcripts with altered migration behaviour may have retained unmelted secondary structures in the denaturing polyacrylamide gel.

Transcription of pKL23-cox2, which harbours two mitochondrial promoters, yielded discrete RNA products mapping to three different initiation sites (Figure 28). Two of these sites were through 5’-RACE confirmed to correspond to the in vivo initiating nucleotides of Pcox2-210 and Pcox2-481 (Figure 29). The Pcox2-481 sequence additionally gave rise to a transcript 5’ end not observed in vivo (Table 12). Erratic initiation by RpoTm (signals indicated by grey arrows in Figure 28) occurred at a sequence that resembles Arabidopsis mitochondrial promoters but had not been detected to function as promoter in vivo. The 5’-terminal nucleotide of the non-specific transcript was determined to correspond to position -345 upstream of the beginning of the cox2 coding sequence on the Arabidopsis mtDNA, and the initiation site was denoted P*cox2-345 (Figure 28 and Figure 29, and Table 12). Of the plasmids pKL23-atp8, pKL23-rrn18 and pKL23-cox2, none supported specific transcription initiation by RpoTmp (Figure 28). Neither didpKL23-rrn26 or pKL23-trnM stimulate specific transcription by RpoTmp (data not shown).

In order to rule out the possibility that the failure of RpoTm to efficiently initiate at Patp6-1-156, Patp9-239,and Prrn18-69 was due to the presence of a second, favoured upstream promoter on the plasmid template (compare Figure 27), additional templates were designed containing the Patp6-1-156, Patp9-239,or Prrn18-69 promoter region but lacking the upstream Patp6-1-200, Patp9-295, or Prrn18-156 sequences, respectively. However, the removal of the latter promoters from in vitro transcription templates did not enhance initiation at Patp6-1-156, Patp9-239,or Prrn18-69 (data not shown).

Figure 29: 5’-RACE analysis of in vitro-synthesized transcripts indicated in Figure 28.

5’ end mapping was performed as described (see Figure 24) on RNAs synthesized by RpoTm from supercoiled pKL23-atp8, pKL23-rrn18, and pKL23-cox2 (chromatograms not displayed). Signals are labelled with the corresponding promoter name or initiation site.

In the described in vitro assays, specific utilization of mitochondrial promoters as transcription start sites by RpoTmdepended on a supercoiled conformation of DNA templates. RpoTmp was unable to specifically initiate transcription, regardless of the template structure. Promoter-specific transcription initiation by the two RNA polymerases was apparently not stimulated by the mtTFB-like protein MetA.

In a study by Binder et al. (1995), a transcriptional activity prepared from pea mitochondria specifically and exclusively initiated transcription at promoters essentially conforming to the CRTAAGAGA nonanucleotide consensus derived previously for dicot mitochondrial promoters (Binder, et al., 1995). Recognition of a deviating mitochondrial promoter by this in vitro transcription system was later described by a single report (Kuhn and Binder, 2002). Additional promoters of the Arabidopsis mtDNA that do not possess a CRTA core element were thus tested for their ability to direct transcription by RpoTm or RpoTmp in vitro.

Figure 30: In vitro transcription initiation at promoters not displaying a CRTA sequence element.

(A) RpoTm and RpoTmp were assayed for promoter-specific transcription of supercoiled pKL23-atp6-1-B, pKL23-atp6-2, and pKL23-atp9. In vitro transcription reactions were run with RpoTm or RpoTmp in the presence of MetA (A) or MetB (B) as indicated, and RNA products were analyzed as described above. Transcripts made by RpoTmp were diluted 1:5 prior to PAGE. Discrete RNA products are specified as in Figure 23. (B) 5’ end mapping was performed as described on RNAs synthesized by RpoTm from pKL23-atp6-1-B, pKL23-atp6-2, and pKL23-atp9 (chromatograms not displayed). Signals are labelled with the corresponding promoter name; asterisks denote non-specific products amplified from RNA linker-ligated transcripts (lane +L). Different background signal intensities observed for the same enzyme and template conformation are due to inherent experimental variation.

While no specific RNAs were synthesized from supercoiled plasmids containing the Prps3-1053, Prps3-1133, and PtrnM-574/573 sequences (data not shown), defined transcripts were made from Patp6-1-916/913, Patp6-2-436, Patp6-2-507, and Patp9-487 by RpoTm (Figure 30; for promoter sequences, see Table 9). 5’-RACE showed that transcription initiated in vitro at Patp6-1-916/913 at the two adenine nucleotides defined previously as in vivo transcriptional starts (Figure 30 and Table 12). The RNA 5’ termini identified forPatp6-2-436 and Patp6-2-507 did not exactly correspond to the in vivo primary 5’ ends but mapped to positions one and three nucleotides respectively downstream of the correct start sites (Table 12). Minor discrete RNAs equalling in size those obtained with RpoTm were generated by RpoTmp from pKL23-atp6-1-B and pKL23- atp6-2. Precise mapping of the 5’-termini of these products was not attempted as they were hardly visible within the background of far more abundant non-specific transcripts, thus rendering the identification of defined RNA 5’ termini extremely difficult. The preference of Patp9-487 over random initiation sites by RpoTm was very weak. While RpoTm did apparently not recognize Patp9-652 as a promoter, additional defined transcripts synthesized from pKL23-atp9 were seen which mapped to non-specific sites. No specific transcripts were made by RpoTmp from pKL23-atp9.

In vitro transcription experiments demonstrated RpoTm to recognize diverse promoter sequences, though not all promoters tested, on supercoiled templates without the aid of transcriptional cofactors. In contrast, RpoTmp did not significantly prefer mitochondrial promoters over random initiation sites.

Two mitochondrial proteins from Arabidopsis encoded at loci At1g80270 and At1g15480 (mitochondrial targeting confirmed by B. Hedtke, HU Berlin), which are homologous to the wheat PPR protein p63 implicated in mitochondrial transcription (Ikeda and Gray, 1999); see I.3.3.3), were additionally tested in in vitro transcription assays for their potential to modulate the transcriptional performances of RpoTm and RpoTmp. Unlike reported for wheat p63, the Arabidopsis proteins did not stimulate transcription initiation at mitochondrial promoters by RpoTm or RpoTmp (in vitro transcription experiments were set up both with and without MetA; data not shown). A transcriptional role of Arabidopsis p63-like proteins was thus not confirmed.