Comparative analysis of plastid transcription profiles of entire plastid chromosomes from tobacco attributed to wild-type and PEP-deficient transcription machineries


*For correspondence (fax ++49-89-1782 274; e-mail
† Present address: Botanisches Institut der Christian-Albrechts-Universität Kiel, Olshausenstrasse 40, D-24098 Kiel, Germany


Transcription of plastid chromosomes in vascular plants is accomplished by at least two RNA polymerases of different phylogenetic origin: the ancestral (endosymbiotic) cyanobacterial-type RNA polymerase (PEP), of which the core is encoded in the organelle chromosome, and an additional phage-type RNA polymerase (NEP) of nuclear origin. Disruption of PEP genes in tobacco leads to off-white phenotypes. A macroarray-based approach of transcription rates and of transcript patterns of the entire plastid chromosome from leaves of wild-type as well as from transplastomic tobacco lacking PEP shows that the plastid chromosome is completely transcribed in both wild-type and PEP-deficient plastids, though into polymerase-specific profiles. Different probe types, run-on transcripts, 5′ or 3′ labelled RNAs, as well as cDNAs, have been used to evaluate the array approach. The findings combined with Northern and Western analyses of a selected number of loci demonstrate further that frequently no correlation exists between transcription rates, transcript levels, transcript patterns, and amounts of corresponding polypeptides. Run-on transcription as well as stationary RNA concentrations may increase, decrease or remain similar between the two experimental materials, independent of the nature of the encoded gene product or of the multisubunit assembly (thylakoid membrane or ribosome). Our findings show (i) that the absence of photosynthesis-related, plastome-encoded polypeptides in PEP-deficient plants is not directly caused by a lack of transcription by PEP, and demonstrate (ii) that the functional integration of PEP and NEP into the genetic system of the plant cell during evolution is substantially more complex than presently supposed.


Chloroplasts are descendants of free-living cyanobacterial-like organisms. During evolution their formerly autarkic gene expression apparatus — together with the corresponding machineries of mitochondria and the nucleus/cytosol — became the integrated, compartmentalised genetic system of the plant cell in which the ‘subgenomes’ are regulated syntonically (Herrmann and Westhoff, 2001; Herrmann, 1997). Due to the eubacterial ancestry of chloroplasts their genetic machinery in many, but not all, respects resembles that of prokaryotes. The genetic potential of plastids is encoded in a reiterated circular chromosome, generally in the range of 120–160 kbp, depending on the organism. Typical plastid chromosomes of vascular plants encode some 120 genes which are usually organised in polycistronic transcription units that are located on both DNA strands (for a review see Sugita and Sugiura, 1996). Contrary to prokaryotes, however, the transcripts of these operons are extensively processed into complex sets of overlapping RNA species, finally often into monocistronic RNAs (Barkan, 1988; Westhoff and Herrmann, 1988). Several instances have been noted where endonucleolytic cleavage of di- or polycistronic transcripts is a prerequisite for efficient translation of mRNAs (Barkan et al., 1994; Hirose and Sugiura, 1997). Essential in higher plant plastids as well are various other transcript maturation processes which are not (i.e. RNA editing) or less commonly (i.e. RNA splicing) associated with prokaryotic systems (Maier et al., 1996; Sugita and Sugiura, 1996). It appears that during organelle evolution distinct RNA processing steps evolved and spread out in the originally cyanobacterial-type genetic system of the ancestral plastid.

In analogy to the evolution of post-transcriptional processing, plastid transcription itself developed into a highly sophisticated process, again representing not merely typical prokaryotic features. In particular, transcription in higher plant plastids is performed by (at least) two different RNA polymerases of different phylogenetic origin. Besides the plastid-encoded RNA polymerase (PEP) which resembles eubacterial enzymes, a nuclear-encoded phage-type RNA polymerase (NEP) is involved in transcribing the plastid chromosome (summarised in Hess and Börner, 1999). The core subunit of the PEP enzyme is specified by the four plastid genes rpoA, rpoB, rpoC1 and rpoC2 which display significant similarity to the three genes encoding the core of the eubacterial RNA polymerase (summarised in Igloi and Kössel, 1992). Sigma-like factors which are essential for transcription initiation by PEP originate in nuclear genes and are post-translationally imported into the organelle (Allison, 2000). Additional PEP-associated polypeptides of nuclear origin with development-dependent compositions have been identified in mustard (Pfannschmidt et al., 2000). Whereas PEP is accepted to trace back to the phylogenetic ancestor of plastids, the NEP locus seems to have evolved by duplication of the gene encoding the mitochondrial RNA polymerase, acquiring a targeting signal for the new organelle (summarised in Hess and Börner, 1999). The ultimate phylogenetic origin of NEP, however, is still unclear. Complicating plastid transcription even more, very recently evidence for a further nuclear-encoded RNA polymerase-type in plastids, that is different from the phage-type NEP enzymes and probably involved in transcription of the rDNA operon has been obtained (Bligny et al., 2000).

The presence of diverse RNA polymerases operating in plastids of higher plants is complemented by the existence of different enzyme-specific promoters and possibly also termination signals preceding and following plastid transcription units, respectively. The PEP enzyme is known to initiate transcription from − 10/− 35 eubacterial-type promoters (Igloi and Kössel, 1992). Analyses of transcription in mutants with disrupted PEP genes or in plastids of non-photosynthetic tissue culture cells led to the identification of NEP-specific promoters that are reminiscent of promoters recognised by mitochondrial and T3/T7 phage RNA polymerases (Hajdukiewicz et al., 1997; Kapoor et al., 1997). Based on analyses of wild-type and PEP-deficient plastids, transcription units have been operationally grouped into three principal classes. Some genes or operons have been suggested to be transcribed by either PEP or NEP, whereas others appeared to be transcribed by both enzymes (Hajdukiewicz et al., 1997). It was proposed that NEP preferentially drives transcription of genes for components of the plastid genetic system, whereas PEP transcribes genes for constituents of the photosynthetic machinery (Hajdukiewicz et al., 1997; Maliga, 1998). Consistently, the accumulation of transcripts of a selected set of photosynthesis-related genes was shown to be reduced dramatically in plants lacking PEP (Allison et al., 1996; Hajdukiewicz et al., 1997). Analysis of run-on transcription activities in PEP-deficient plastids, however, revealed that most segments of the plastid chromosome, independent of the encoded gene class, are transcribed even in the absence of PEP (Krause et al., 2000).

In order to study the still enigmatic functional and phylogenetic implications of multiple transcription machineries in the plastids of higher plants in more detail we established a macroarray-based assay to examine transcription rate and transcript levels of all known genes of the tobacco plastid chromosome. We applied this approach to a comparative study of transcription profiles of leaves from wild-type as well as from PEP-deficient transplastomic mutants (DeSantis-Maciossek et al., 1999) of tobacco. The results presented here indicate a high complexity concerning the functional integration of the different plastid RNA polymerases into the genetic network of the plant cell that is not, as supposed, primarily reflected in gene class-specific transcription but in differential processing, stability and accumulation of resulting transcripts and/or of corresponding polypeptides.


Array design

For array production, fragments representing the entirety of 118 genes and 11 open reading frames identified on the tobacco plastid chromosome (Shinozaki et al., 1986; Wakasugi et al., 1998) were individually PCR-amplified from total plastid DNA. In those cases where small genes contain introns, supplementary PCR products were generated from total tobacco plastid cDNA representing ‘spliced’ genes. This turned out to be crucial, especially for probing the expression of intron-containing tRNA genes, since DNA fragments with intervening sequences did not reliably detect spliced tRNAs under various hybridising conditions.

Quality and amount of PCR products were checked by agarose gel electrophoresis, and each individual amplicon was then adjusted to three different concentrations of 6.25, 25 and 100 ng µ−1 l, respectively. Using a robot device equipped with a 96-pin tool, a total of 134 DNA samples, including pBluescript vector (Stratagene, USA) as a negative control, were arrayed on positively charged nylon membranes according to the pattern shown in Figure 1. To increase the quantity of transferred DNA and to minimise eventual transfer differences, samples were spotted at their destination by 10-fold repetition to the same points onto the membranes. Each sample was finally arrayed in three different quantities (1.25, 5 and 20 ng, respectively) and grouped in duplicates, generating double positives upon probing (see Figure 2a,b).

Figure 1.

Schematic presentation of positions of gene and ORF probes of the entire tobacco plastid chromosome on the array.

Figure 2.

Autoradiograms of nylon filters representing all known genes and ORFs of the tobacco plastid chromosome probed in (a) with run-on transcripts and in (b) with 5′ labelled RNA from wild-type (WT) and PEP-deficient (rpoA) plastids.

Each DNA fragment was arrayed in three different quantities in duplicate using a spotting density of 3 × 3 resulting in the vertical pattern (left/right) 2, 3, 1 and 1, 3, 2, respectively, with 1, 2 and 3 representing pairs of spots with DNA quantities of 1.25, 5 and 20 ng, respectively. DNA fragments were arrayed according to the pattern depicted in Figure 1.

Different controls were performed to check the quality of the arrays obtained. (i) Uniformity and reproducibility of sample loading was evaluated by including an internal control, i.e. by hybridising membranes to end-labelled oligonucleotides complementary to the T7 RNA polymerase promoter sequence that was added to each PCR product as the result of 5′-tagged primers utilised in the amplification reactions (data not shown). (ii) The specificity of sample recognition was monitored by hybridising the arrays to radioactively labelled RNA probes synthesised by transcription of individual PCR products with T7 RNA polymerase. In each case only the expected spots could be detected (data not shown). (iii) In order to estimate the influence of the size of arrayed DNA fragments on signal intensity, radioactively end-labelled plastid RNA was hybridised to psbA PCR products of different lengths. Signal intensities turned out to be only marginally (< 10%) stronger with longer DNA fragments (Figure 3a). Such differences are negligible in a comparative analysis of expression profiles as presented here. (iv) The reproducibility of array hybridisations was verified by scatter plot analysis of data from independent experiments. Correlation coefficients of 0.96–0.97 were typically obtained (Figure 3b).

Figure 3.

(a) To check the influence of fragment length on probe binding three arrayed PCR fragments containing 330, 660 and 967 bp of the psbA coding region were hybridised to end-labelled RNA from PEP-deficient (rpoA) plastids. Note that the signal intensities are only slightly stronger by using longer DNA fragments.

(b) Reproducibility of experiments was verified by Scatter plot analysis. Two different arrays containing the complete set of plastid genes (represented by diamonds) were hybridised with end-labelled wild-type RNA. Signal intensities of single probes (values are given in arbitrary units) are indicated by their location within the 2-dimensional matrix. Correlation between the resulting signals was high with a correlation coefficient of 0.9632 and a curve slope of y = 1.0012×.

(c) Selective labelling of 5′ hydroxyl termini of RNA molecules by PNK was monitored by incubating equimolar amounts of in vitro transcripts (containing 5′-triphosphate termini) and RNA oligonucleotides (containing 5′ hydroxyl termini) in the presence of [γ-32P]ATP and PNK.

Suitability of probe types for array hybridisation

Whereas in array-based approaches, labelled run-on transcripts are the probe type of choice in evaluating transcription rates, transcript levels can be estimated by using either cDNAs or labelled RNAs as probes. Three different kinds of probes were checked to evaluate plastid transcript levels. (i) cDNA probes were generated by reverse transcription either primed with random hexanucleotides or with a mixture of 129 gene-specific oligodesoxynucleotides each complementary to the coding strand of one of the plastid genes or ORFs. Although the use of gene-specific primers in reverse transcription reactions leads to more pronounced signals compared to those derived with randomly primed cDNA, expression of a number of tRNA species could not be demonstrated. This was noted, when primers for reverse transcription span the 3′ genomic sequence of tRNAs. Most probably post-transcriptional modifications of plastid tRNAs by 3′-CCA addition (Gegenheimer, 1995) prevented annealing of primers in the reverse transcription reaction in those cases. Moreover, reverse transcription of a single polycistronic transcript using random primers generally leads to a heterogeneous cDNA population covering the upstream cistrons more abundantly than the downstream ones. This may lead to misinterpretation of expression data when using cDNA as a probe. For production of riboprobes, transcripts were either labelled (ii) at 5′-termini by T4 polynucleotide kinase (PNK) or (iii) at 3′-termini by T4 RNA ligase. PNK specifically catalyses the transfer of the terminal phosphate of ATP to 5′ hydroxyl termini of RNA whereas 5′-triphosphate and 3′ hydroxyl ends remain unaffected. This was experimentally verified by selective labelling of RNA oligonucleotides but not of in vitro transcripts containing hydroxyl or triphosphate groups at their 5′ ends (Figure 3c). The alternative, 3′ labelling of RNAs by T4 RNA ligase, turned out to be less suitable, since it bears the risk of generating artefacts in probing the arrays due to frequent co-ligation of individual transcripts. Taken together, application of RNA probes 5′ end-labelled by PNK proved to be superior in generating reliable results in array-based quantitative analyses of plastid encoded transcripts and thus were used for estimating actual transcript levels. It should be noted, however, that the signals derived reflect only those molecules containing 5′ hydroxyl termini, and thus allow to selectively monitor for the accumulation of processed RNA species, in contrast to the arrays probed with run-on transcripts (see Discussion).

Strand-specific versus non-strand-specific hybridisation

Genes on plastid chromosomes are located on both DNA strands. Therefore, run-through transcription of genes encoded by the opposite DNA strand can lead to antisense RNAs that may contribute to transcript levels if probes include transcripts of both DNA-strands (e.g. run-on transcripts or end-labelled plastid RNA) or if probes are hybridised to double-stranded DNA fragments (e.g. gene-specific amplicons on arrays). The array approach presented here, as nearly all previous work on transcript levels in wild-type and PEP-deficient tobacco plants (Allison et al., 1996; Hajdukiewicz et al., 1997; Krause et al., 2000; Serino and Maliga, 1998) and comparable work with other materials (Inada et al., 1996; Silhavy and Maliga, 1998a, 1998b) is based on double-stranded probes, and hence, inherently does not allow strand-specific transcription to be assessed. To estimate the influence of antisense RNA when hybridising labelled transcripts to arrayed double-stranded DNA fragments, the reverse approach was performed, in which plastid RNA was dotted onto Nylon membranes and hybridised against surplus amounts of radioactively labelled single-stranded antisense RNA (Figure 4a). In general, the results were consistent with those obtained with DNA-based filters. However, differences with respect to wild-type/mutant signal ratios were noted in a number of instances, of which some are illustrated by Figure 4. For atpA, atpI, psbC, psaA, rps14, petB and petD, which all showed decreased transcript levels in PEP-deficient plastids as determined by strand-specific probes, wild-type to mutant signal ratios turned out to be lower with the array approach. The opposite was observed for ycf3, rps18, and 5′ rps12, which exhibited increased transcript levels in PEP-deficient plastids. These differences in wild-type/mutant transcript ratios addicted to the probe-type chosen can be caused by antisense RNA.

Figure 4.

Autoradiograms of nylon filters onto which (a) 5 µg of wild-type or mutant RNA were dotted and hybridised with 8 µg of radiolabelled in vitro transcripts of atpA, atpI, psbC, psaA, rps14, ycf3, rps18, 5′rps12, petB and ndhJ, respectively.

In (b) double stranded gene-specific DNA fragments were spotted as described for Figure 2 and probed with 5′ end labelled RNA. Ratios of values determined for RNA from wild-type (WT) and PEP-deficient (rpo) plants are given below the gene designation. Note that in (a) membranes with dotted RNA of wild-type and rpo leaf tissue were probed together with strand-specific probes whereas in (b) spotted DNA fragments were probed separately with wild-type and rpo RNA.

In a previous study we have shown that the ratios between sense and antisense RNA can vary widely between individual genes but that the regional amounts of antisense transcripts in both wild-type and mutant leaf tissue were comparable (Krause et al., 2000). The array data are consistent with these findings. The observed shifts in wild-type to mutant transcript ratios are most probably caused by comparable regional fractions of antisense transcripts in both materials which are detected with arrayed double-stranded DNA but not with strand-specific probes. If, in the case of psaA with a wild-type to mutant RNA ratio of approximately 2.58–1.0 an antisense fraction of 0.72 is added, the ratio decreases to 1.92 (3.3–1.72) as measured by the array approach. On the other hand, in the case of rps18, addition of an antisense fraction of 0.6 leads to an increase of the ratio from 0.31 to 0.57 (cf. Figure 4). Thus, quantitative comparison of transcript levels in plastids strongly depends on the probe-type, single-stranded or double-stranded, applied. A full account of the outlined data will be presented elsewhere.

Comparison of expression profiles of wild-type and PEP-deficient plastids

Changes in transcription rate of individual plastid genes do not necessarily lead to comparable changes in steady state transcript levels. To assess both, the activity of the different RNA polymerases as well as the fate of the synthesised transcripts, array-based profiles of transcription rates, as well as of steady-state transcript levels were determined in this study. In the former case, transcription of operons, genes and ORFs of the tobacco plastid chromosome was investigated by hybridising [α-32P]UTP-labelled run-on transcripts from plastid lysates to arrays. Profiles of actual transcript levels, in turn, were studied by probing the arrays with total plastid RNA which had been 5′ end-labelled with PNK in the presence of [γ-32P]ATP. The following sections provide a detailed comparison of transcription rates and transcript steady state profiles of the complete gene set in wild-type and PEP-deficient plastids, taking into account the above outlined technical aspects as well as the proposed classification of transcripts either to be synthesised by PEP, NEP or both enzymes (Hajdukiewicz et al., 1997). This study was complemented by Northern and Western analyses of selected sets of genes in order to trace the fate of transcripts synthesised in wild-type or mutant leaves in more detail.

(i) Genes encoding photosynthesis-related components

Previous reports (Allison et al., 1996; Hajdukiewicz et al., 1997; Serino and Maliga, 1998) have shown that, in the absence of PEP, transcription initiation of tobacco plastid genes and operons is linked to promoters specifically used by NEP. Furthermore, Western analysis verified that plastid-encoded polypeptides involved in photosynthesis do not accumulate in plastids lacking PEP (DeSantis-Maciossek et al., 1999). Our run-on transcription data show that all plastid genes are transcribed in a PEP-deficient background, though into different profiles compared to wild-type plastids (Figures 2a and 5a). This confirms that the absence of plastid-encoded photosynthesis-related gene products in mutants lacking PEP is not primarily based on selective transcription of the corresponding genes by this polymerase (cf. Krause et al., 2000).

Figure 5.

Histograms indicating the relative transcription rates (a) and RNA levels (b) from plastids of wild-type (grey bars) and PEP-deficient (rpoA) leaf tissue (white bars).

Intensities of duplicate spots representing the same gene were averaged and analysed as percentages of the total signal on the array after subtracting background. In those cases, where the probe quantity exceeds the quantity of the corresponding DNA fragments on the filter, values received (designated as ∞) were excluded from the quantitative evaluation of the expression profiles. Genes and ORFs above the baseline are transcribed from left to right, those below from right to left. Data in (a) were normalised by setting data in relation to rrn16 transcription rate in the respective material. In (b) normalisation was done by setting the atpI wild-type/mutant transcript ratio to 0.8 as calculated from Northern filters using double-stranded probes (Allison et al., 1996).

The genes rbcL (encoding the large subunit of Rubisco), psbA, psbB, psbC, psbD, psbE, psbF, psbH, psbJ, psbL, psbM, psbT, psbN, psbZ (ycf9, Swiatek et al., 2001) (encoding subunits of the photosystem II assembly), psaA and psaB (encoding the two core proteins of photosystem I), petB, petD, petL and petN (encoding subunits of the cytochrome b6f complex), ndhG and ndhI (encoding subunits of the NADH dehydrogenase) exhibit relative transcription rates that are significantly lower in mutant than in wild-type plastids, as expected. However, various genes encoding subunits of photosystem I (psaC and psaI), photosystem II (psbI and psbK), of the cytochrome b6f complex (petG) and the NADH dehydrogenase (ndhA, ndhC, ndhD and ndhE) turned out to be transcribed at almost similar rates in both, PEP-deficient mutants and wild-type. Even more astonishingly, various genes, also coding for subunits of thylakoid assemblies, notably atpA, atpB, atpE, atpF, atpH, atpI (specifying constituents of the thylakoid-located ATP synthase), psaJ, petA, ndhB, ndhF, ndhH, ndhJ and ndhK are transcribed at even higher rates in mutant than in wild-type plastids. The same holds for the reading frames ycf3, ycf4 and ycf10 which have been shown to encode proteins with photosynthesis-related functions (Figures 2a and 5a; Boudreau et al., 1997; Rolland et al., 1997; Ruf et al., 1997). It should be noted that individual genes of one operon can show quite different signal intensities (i.e psbE, psbF, psbL and psbJ or rpoB, rpoC1 and rpoC2) with decreasing values from the first to the last gene of the operon. Most probably, this can be attributed to the lower number of longer transcripts generated in run-on assays, possibly also in vivo.

When probing arrays with end-labelled RNAs, in many cases hybridising transcript quantities correlate with findings of run-on analyses (see Figures 2b and 5b). Operons or genes with high transcription rates generally lead to higher transcript levels and vice versa. However, in a number of cases, concerning both photosynthesis and non-photosynthesis-related genes (see below), no correlation between transcription rates and transcript quantities could be found. For instance, psbM and psbT exhibit very low relative transcription rates in mutants (Figures 2a and 5a), but the corresponding end-labelled RNAs were detectable in almost similar quantities in both wild-type and mutant (Figures 2b and 5b). On the other hand, in mutants relative transcription rates of psaI only slightly differ in comparison to wild-type, but the resulting transcript quantities are significantly higher. Surprisingly, petB and petD transcription rate and transcript levels determined by the arrays differ, with intensities for petD being substantially higher. A priori, this is unexpected since petB and petD are co-transcribed as part of the psbB operon and not processed into individual RNAs in tobacco (Monde et al., 2000a). However, in addition to petB/petD-containing processed RNA species including the 1.9 kb psbH/petB/petD and the 1.5 kb petB/petD transcripts a 1.2 kb petH/petB dicistronic RNA accumulates to substantial amounts. No monocistronic petD mRNA is detectable by Northern analysis (Figure 6; Monde et al., 2000a). Thus, the amount of petB containing transcripts is higher than that of petD, as revealed by Northern analysis with strand-specific probes (Figure 6). In the ‘double-stranded’ array approach, signal intensities for petD are influenced by antisense transcripts probably originating from run-through transcription of the rpoA or aadA gene flanking petD on the opposite DNA strand in wild-type and mutants, respectively (data not shown). More striking discrepancies appear when transcription rates are compared with data from Northern analyses. For example, the majority of genes coding for thylakoid proteins, such as psbC, petB and petD, accumulate only very low steady state transcript levels in mutants lacking PEP as judged by Northern analyses (Figure 6 and data not shown), although they are transcribed with appreciable rates, obviously by NEP (Figures 2a and 5a).

Figure 6.

Selected Northern filters of RNA from tobacco wild-type (WT) and PEP-deficient mutants (Δrpo). Equal amounts of RNA were electrophoretically separated in agarose/formaldehyde gels and probed with in vitro transcripts specific to the mRNA of atpA, atpI, psbC, psaA, rps14, ycf3, rps18, 5′rps12, petB, petD and ndhJ, respectively. A 25S rRNA control for equal loading is shown below autoradiograms. Signals were quantified with a Fuji phosphorimager and the TINA (raytest) program.

In addition to quantitative differences either in transcription rate or in RNA accumulation, various genes for photosynthesis-related functions also exhibited conspicuous qualitative differences in transcript pattern between wild-type and PEP-deficient mutants. For some genes, such as ndhJ, the signal strength among individual bands of a RNA pattern differs between wild-type and mutant plastids (see Figure 6). This may be caused by operon internal promoters and/or differences in processing kinetics or RNA stability. In other cases, like atpA, atpI, psaA, ycf3, rps14, rps18 or ndhJ (Figure 6), high molecular weight precursors accumulate specifically in mutants lacking PEP (cf. Krause et al., 2000). For instance, whereas wild-type plastids accumulate psaA and rps14 transcripts of approximately 5.2 kb in size, which correspond to the tricistronic psaA/psaB/rps14 message, mutant leaf tissue lack this band but exhibit large transcripts of approximately 7.5 and 9.5 kb. When probing for ycf3, signals of 7.5 and 9.5 kb are detectable again exclusively in mutants (Figure 6). This suggests that in the absence of PEP psaA and psaB messages accumulate as parts of a large polycistronic ycf3/psaA/psaB/rps14 RNA which is consistent with the localisation of a putative NEP promoter upstream of ycf3 in plastids of mustard (Summer et al., 2000). However, PsaA and PsaB proteins are not detectable in mutant plastids (Figure 7).

Figure 7.

Immunoblot analysis of total proteins from tobacco wild-type (WT) and PEP deficient-mutants (rpoA).

Equal amounts were loaded per lane and probed with polyclonal, monospecific antibodies against subunit CP43 of photosystem II (gene: psbC), cytochrome b6 (petB), reaction centre proteins of photosystem I (psaA and psaB), and the nuclear-encoded 33 kDa polypeptide of the oxygen-evolving system (psbO), respectively.

It is known that steady-state RNA concentrations and rates of mRNA translation are often discordant and substantial amounts of thylakoid proteins are frequently translated from minute RNA amounts (e.g. Herrmann et al., 1992). Immunoblot analysis, however, shows no detectable signals of full-size photosystem I core proteins (genes: psaA and psaB), CP43 of photosystem II (psbC) and cytochrome b6 (petB) of the cytochrome b6f complex in the PEP-deficient plastids (Figure 7), although transcripts, even though in reduced amounts, are clearly present (Figure 6).

In summary, immunoblot analyses (Figure 7; plus complementing data of DeSantis-Maciossek et al., 1999) demonstrate that plastid-encoded thylakoid proteins are not or only hardly detectable in plastids lacking an active PEP enzyme, in contrast to those of nuclear origin (e.g. PsbO; Figure 7) which accumulate to comparable levels in both, wild-type and PEP-deficient mutants. Combined with results from array-based expression profiling and from Northern analyses the data show that the lack of plastid-encoded polypeptides for photosynthesis-related functions in PEP-deficient tobacco plants is not directly caused by a lack of transcription of their genes but depends on which one of the distinct RNA polymerases produced their mRNAs.

(ii) Genes encoding components of the genetic apparatus

Since translation takes place in PEP-deficient plastids (Allison et al., 1996; DeSantis-Maciossek et al., 1999), it is not surprising that run-on analyses reveal transcription of all genes coding for components of the organelle translation machinery. Consistent with previous observations (DeSantis-Maciossek et al., 1999; Hajdukiewicz et al., 1997), genes such as rpoB, C1 and C2, trnL(CAA), trnR(UCU), trnS(GCU) and trnV(GAC), rps3, rps7, rps8, rps11, 3′rps12, rps18 and rps19, rpl16, rpl20, rpl22, rpl32 and rpl36 are transcribed at significantly higher rates in PEP-deficient mutants (Figure 5a). On the other hand, comparably to genes for photosynthesis-related components shown above, relative transcription rates for several housekeeping genes in wild-type and mutants are quite similar (i.e. rrn23, rrn16, rrn5 and rrn4.5, trnG(UCC), trnI(CAU), trnL(UAG), trnR(ACG), trnS(GGA), trnT(UGU), trnP(UGG) and trnW(CCA), rps2, rps4, 5′rps12, rps15, and rpl33). Remarkably, various genes [i.e. trnA(UGC), trnC(GCA), trnD(GUC), trnE(UUC), trnF(GAA), trnG(GCC), trnH(GUG), trnK(UUU), trnL(UAA), trnM(CAU), trnfM(CAU), trnN(GUU), trnQ(UUG), trnS(UGA), trnT(GGU), trnV(UAC), trnY(GUA), rps14, rps16, rpl2, rpl23 and ycf14(matK), which is assumed to be involved in splicing of certain group II introns], display higher relative transcription rates in wild-type than in the mutant leaf tissue (see Figures 2a and 5a).

As noted for genes for photosynthesis-related components, the transcript quantities again often do not correspond to transcription rates. For instance, trnS(GGA) shows similar relative transcription rates in mutant and wild-type plastids (Figure 5a), but the resulting transcript quantities hybridising to the trnS(GGA) gene fragment turned out to be higher in wild-type than in mutants (Figure 5b). On the other hand, the rpl2 gene is transcribed at nearly 2-fold higher rates in wild-type plastids (Figure 5a), but in mutants the corresponding transcript quantity was found to be even more than 2-fold that of wild-type plastids (Figure 5b). The amount of Rpl2 polypeptide, however, is significantly higher in wild-type plastids (DeSantis-Maciossek et al., 1999). This again clearly illustrates that, in analogy to the relationship between transcription rates and resulting RNA levels, only limited conclusions can be drawn from transcription rates or RNA concentrations on the accumulation of the corresponding protein.

(iii) Heterogenic operons encoding components of both, the photosynthesis and genetic apparatus

Different from plastid operons carrying genes either solely for components of the genetic machinery or for photosynthesis-related functions, rps14 which encodes the protein CS14 of the 30S ribosomal subunit, is transcribed as part of an operon also containing the two photosystem I genes psaA and psaB (see above; Meng et al., 1988). The operon has no obvious NEP promoter and seems to be transcribed preferentially by PEP (Hajdukiewicz et al., 1997) which in wild-type plastids mainly results in the tricistronic psaA/psaB/rps14 message driven by a PEP promoter upstream of psaA (Meng et al., 1988). Our data demonstrate that rps14 is also transcribed in PEP-deficient mutants but at a significantly lower rate compared to wild-type (Figure 5a). However, in contrast to the wild-type pattern no signal for the dominant 5.2 kb psaA/psaB/rps14 co-transcript can be found in PEP-deficient leaf tissue as judged from Northern analyses, and even no monocistronic rps14 mRNA (of approximately 0.5 kb) resulting from a potential NEP promoter is detectable (Figure 6; Hajdukiewicz et al., 1997). In the absence of PEP, rps14 mRNA seems to accumulate only as part of a large polycistronic ycf3/psaA/psaB/rps14 RNA which is consistent with the presence of a putative NEP promoter upstream of ycf3 (Summer et al., 2000). It should be noted that other transcripts with sizes of 2.2 and 4.2 kb not present in the wild-type are detectable in mutant leaf material. Primer extension analysis of the region upstream to rps14 generates a signal specific for mutant RNA at position − 152 with respect to the translational start codon of rps14 (Figure 8). In addition, various other primer extension signals (i.e. at positions 173, 198, 211 and 250) which are more pronounced in RNA from PEP-deficient mutants than in wild-type RNA are also clearly detectable (Figure 8). Since sequences surrounding these signals are identical in both wild-type and mutant, premature termination of cDNA synthesis due to secondary structures in mRNAs can not account for those differences. Together with data from Northern analysis this strongly suggests that transcripts are subject to differential processing not only in a quantitative but also in a qualitative manner depending on the polymerase that produced them. Interestingly, the CTT motif, which has been found to constitute the core of the NEP promoter driving transcription of rrn16 of tobacco (Nt Prrn16–64; Liere and Maliga, 1999; Vera and Sugiura, 1995) is found four nucleotides upstream of the primer extension signal at position − 152, within the coding region of psaB (Figure 8). Individual primer extension signals therefore alternatively may result from (specific or ‘unspecific’) initiations of transcription by NEP.

Figure 8.

Primer extension analysis of rps14 mRNA.

End-labelled primer extension products from wild-type (WT) and PEP-deficient mutants (rpoA) are shown alongside the sequence ladder derived by using the same primer. Numbers refer to the distance in nucleotides from the rps14 translation initiation codon.

Two editing sites have been identified in rps14-coding transcripts of tobacco (Hirose et al., 1998). Post-transcriptional C- to -U conversions at both positions lead to the restoration of codons for leucine residues conserved in rps14 genes of other species. It is generally assumed, and has been experimentally shown for some cases (Bock et al., 1994; Sasaki et al., 2001; Zito et al., 1997) that such conserved amino acid residues restored by editing are of structural or functional importance for the affected polypeptide (Maier et al., 1996). Both sites of rps14 transcripts are edited even in PEP-deficient plastids. Also, 3′ ends of the rps14 message are found to be identical in both wild-type and mutant plastids (data not shown). Thus, expression of a functional CS14 protein (Hirose et al., 1998), which is important in the early stages of 30S assembly (Wimberly et al., 2000), appears to be guaranteed and efficient translation may compensate the only minute amounts of rps14-containing mRNA in the mutants.

(iv) Genes specifying other functions

As illustrated in Figure 5, clpP, encoding a catalytic subunit of the ATP-dependent Clp protease (Maurizi et al., 1990) and ycf2, which has been found to be essential for plastid biogenesis, even though its function is unknown (Drescher et al., 2000), belong to those genes with quite different transcription in PEP-deficient leaf material. In wild-type, clpP and ycf2 are transcribed at significantly lower rates than in mutant plastids (Figure 5a) and the resulting transcript levels determined by the array approach (Figure 5b) are lower, too. Whereas Northern experiments revealed a only very weak signal for ycf2 transcripts in wild-type (data not shown, Hajdukiewicz et al., 1997), clpP mRNA is clearly detectable (data not shown, Serino and Maliga, 1998) consistent with the finding that one of the four clpP promoters is recognised by PEP (Hajdukiewicz et al., 1997). On the other hand, accD, encoding one of the subunits of the prokaryotic-type acetyl-CoA carboxylase involved in lipid biosynthesis (Sasaki et al., 1993) shows high relative transcription rates in both, mutant and wild-type leaf tissue (Figures 2a and 5a) but its mRNA accumulates in larger quantities only in PEP-deficient plastids as revealed by Northern analyses (data not shown, Hajdukiewicz et al., 1997; Krause et al., 2000).

(v) Open reading frames

In addition to defined genes and well-conserved ycfs, the tobacco plastid chromosome harbours 11 annotated ORFs (Shinozaki et al., 1986). Comparative analysis of chromosome regions encoding these ORFs in various plant species indicates that, although transcribed, most of them probably do not encode functional peptides (Schmitz-Linneweber et al., 2001). Whereas some of these ORFs exhibit relative transcription rates with no major difference between wild-type and PEP-deficient mutants (e.g. ORF70A and ORF115; see Figure 5), some of them exhibit higher expression rates in wild-type plastids (i.e. ORF105 and ORF75), others (i.e. ORF31 and 70B) show increased relative transcription rates in plastids lacking the PEP enzyme. As shown in Figure 5, transcription rates and RNA levels of these ORFs often correspond to those of flanking genes.


Transcription in plastids is unique since it is the only ‘prokaryotic system’ known to operate with different kinds of RNA polymerases. In the nucleus the existence of different RNA polymerases has been established for long and the roles of the individual enzymes were shown to operate in the transcription of different gene classes. Whereas the eukaryotic RNA polymerase I transcribes rRNA genes, type II transcribes mRNAs, and RNA polymerase III is involved in the synthesis of generally small RNA molecules, including 5S rRNA and tRNAs (Chambon, 1975). Similarly, the different RNA polymerases acting in plastids were proposed to be involved in differential transcription of photosynthesis-related and housekeeping genes (Hajdukiewicz et al., 1997).

In this study, we have presented an array-based comparative analysis of gene expression profiles of entire plastid chromosomes in leaves of tobacco wild-type and mutants lacking the plastid-encoded RNA polymerase. Whereas transcripts in wild-type may derive from both, PEP and NEP promoters, in mutant plastids transcription initiation is limited to promoters used by NEP and probably the recently detected nuclear-encoded third type of plastid RNA polymerase that does not correspond to the phage-type NEP enzymes of 110 kDa (Bligny et al., 2000). Different probe types and expression levels have been compared in order to estimate the reliability and limitations of the approach. Data were evaluated with regard to (i) differential transcription rates and resulting transcript quantities of individual genes/operons by the wild-type or PEP-deficient transcription apparatus (quantitative point of view) (ii) differential initiation/termination and processing of transcripts either driven by NEP or PEP (qualitative point of view) (iii) differential degradation/stabilisation, and (iv) differential translation of NEP- and PEP-driven transcripts. It appears that no universal relationship between transcription rates, transcripts levels and the amount of translation products exists, and also no differential, gene class-specific transcription by NEP or PEP, different from the nuclear system. Our results demonstrate that the functional integration of NEP into the genetic system of the plant cell during evolution is more complex than commonly supposed.

By hybridising run-on transcripts from wild-type and PEP-deficient plastids to filter arrays carrying all plastid encoded genes (Figures 2a and 5a) we could show that NEP, as well as probably PEP, alone is able to transcribe the entire genetic information of the tobacco plastid chromosome. The transcription rates of individual genes or operons however, often do not correlate with steady state transcript levels as determined by hybridising end-labelled RNA to arrays (Figures 2b and 5b) and by Northern analysis (Figure 6), and only a distinct subset of the corresponding proteins accumulates in PEP-deficient plastids (Figure 7 and DeSantis-Maciossek et al., 1999). In fact, independent of the function of the encoded gene product or of the multisubunit assembly (i.e. photosynthesis-related structure or ribosome), all three patterns, increase, decrease and virtually no change of RNA levels, have been observed comparing transcript/transcription profiles of wild-type and PEP-deficient plastids. Taken together, these findings exclude that selective accumulation of non-photosynthesis related gene products in plants lacking PEP is based primarily on the selective transcription of the corresponding genes (cf. also Krause et al., 2000). This raises the question upon the cause and physiological relevance for the observed differences in the expression pattern.

The evaluation of array-based expression profiles in plastids is particularly challenging in comparison to nuclear expression profiles (e.g. DeRisi et al., 1997; Lashkari et al., 1997; Wodicka et al., 1998) because plastid genes are organised in polycistronic transcription units that are subsequently processed into complex sets of overlapping transcripts which may differ in stability and hence in relative stationary concentrations. Furthermore, transcription of individual genes of a given operon can be driven by several, even intracistronic, promoters (Hajdukiewicz et al., 1997; Kapoor et al., 1994; Orozco et al., 1990) which adds to the complexity of transcript patterns.

Whereas the transcription rates of single plastid genes can be estimated by hybridising labelled run-on transcripts to the arrays, transcript steady-state levels in principle can be assessed by either using labelled RNA or cDNA derived by reverse transcription of plastid RNA as probes. Our results show that cDNA probes are not recommendable for the array-based transcription profiling of plastid genomes, since these are transcribed in polycistronic units and reverse transcription can lead to a heterogeneous cDNA population covering the upstream cistrons more abundantly than the downstream ones. This can lead to incorrect expression level estimations. Concerning the use of RNA probes, application of transcript probes 5′ end-labelled by PNK turned out as most suitable, since 3′ end labelling using T4 RNA ligase can lead to co-ligation of independent transcripts (data not shown) and therefore may produce artefacts in the determination of actual transcript levels. PNK on the other hand selectively labels 5′ hydroxyl termini and therefore excludes primary transcripts from probing. The fraction of primary transcripts in plastids, however, is generally low, in Chlamydomonas not even detectable by capping experiments (Monde et al., 2000b). The finding that transcript levels of the first cistron of a polycistronic operon are not generally under-represented (Figure 5b) clearly shows that only a minor fraction of transcripts, of both wild-type and mutant plastids, contain triphosphate groups at their 5′ ends. Even those transcripts which are known to undergo only limited processing in tobacco plastids (e.g. Eibl et al., 1999) are significantly labelled by PNK, which suggests that their 5′ ends are present in dephosphorylated form. Thus the error introduced by using PNK-labelled probes is negligible. On the other hand, although PNK labels only a subpopulation of RNAs the transcript quantities estimated by the array approach, especially in PEP-deficient mutants, often exceed those determined by Northern experiments (e.g. petB, petD, psbC or rps14, Figures 5b and 6).

These discrepancies can be explained in several ways. (i) Since the arrayed double-stranded DNA fragments hybridise to both, sense and antisense transcripts, a fraction of the signals measured may trace back to antisense RNA. The influence of antisense transcription on array-based data has been checked by the reciprocal experiment in which filters with dotted RNA were individually probed with gene-specific in vitro transcripts (Figure 4). It turned out that wild-type/mutant signal ratios for transcripts less abundant in mutant leaf tissue were higher using strand-specific probes (Figure 4a). The opposite was observed for transcripts more abundant in mutants. These generally exhibited lower wild-type/mutant transcript ratios when using strand-specific probes. This can be caused by fractions of antisense RNA, with regional similarity in both mutant and wild-type, which hybridise to double-stranded amplicons but are not detected with strand-specific probes. (ii) It is conceivable that a significant fraction of end-labelled RNA molecules which hybridise to a distinct gene probe constitute run-through transcripts initiating at upstream genes (or downstream genes on the opposite DNA strand, as in the case of petD) as well as to transcripts initiating within the coding region (sense and/or antisense strand) of the tested gene. For instance, a promoter specifically used by NEP has been identified upstream of clpP on the antisense strand within the psbB coding region (Hajdukiewicz et al., 1997). (iii) Also, ‘unspecifically’ initiated transcripts may account for the observed discrepancies. Promoter core sequences recognised by NEP exhibit only a weak consensus motif (Weihe and Börner, 1999). Using an in vitro transcription assay, directed mutagenesis of a 15 nucleotide segment, containing the PrpoB− 354 NEP promoter, revealed a major decrease in transcription activity (< 30%) only if the core CRT sequence had been changed whereas other positions seem to be less important (Liere and Maliga, 1999). Therefore, NEP transcription may initiate at various sites spread over the plastid chromosome leading to a heterogeneous population of perhaps mostly abortive transcripts. This may find support in numerous primer extension signals not only found in intergenic but also in coding regions in PEP-deficient leaf tissue (Figure 8 and data not shown). However, at least in cases where signals appeared from both, wild-type and mutant RNA, such bands may also be caused by processed RNA or premature termination of cDNA synthesis due to RNA secondary structures. Moreover, both sense and antisense RNA fragments resulting from ‘unspecific’ transcription initiation are able to hybridise to the respective gene-specific PCR products on the array. (iv) Finally, small transcript fragments (i.e. degradation products of unspecifically synthesised or not well stabilised RNAs), which during RNA gel electrophoresis are separated from distinct bands prior to blotting and appear as background in gels, may escape detection in Northern filters but could contribute significantly to the increased transcript quantities measured by array hybridisation.

The steady state level of a mRNA is dependent on both its kinetics of synthesis as well as of degradation. The interplay between transcriptional regulation and control of RNA stability is known to be quite complex in plastids (reviewed in Monde et al., 2000b). RNA stabilities can vary during plastid development and their regulation during chloroplast biogenesis in response to photosynthetic activity has been suggested (Monde et al., 2000b). Interestingly, the redox state may influence RNA degradation rates in chloroplasts of Chlamydomonas reinhardtii (Salvador and Klein, 1999). Because of the different redox conditions in wild-type and PEP-deficient leaf tissues (DeSantis-Maciossek et al., 1999), RNA stability may considerably differ between transcripts of one and the same gene in the two materials. Stability of single transcripts possibly could also differ in dependence of its synthesis by either NEP or PEP. It is known that stabilisation/degradation of plastid transcripts is often modulated by cis elements located predominantly in the 5′- and 3′- UTRs (Hayes et al., 1999; Schuster et al., 1999). In addition, transcript degradation can also be initiated by endonucleolytic cleavages within the coding region of a mRNA (Klaff, 1995). Since the PEP and NEP enzymes operate with different promoters, which may be separated by hundreds of base pairs, at least 5′ ends of the primary transcripts produced by the two polymerases do not coincide. This could influence transcript stability. It is conceivable that (hitherto unknown) polymerase-specific transcription termination signals account for differences in transcript accumulation as well, since the 3′ terminal sequence of a plastid transcript, often folding into secondary structure, is critical for its stability (Hayes et al., 1999). Run-on transcription in plastids treated with tagetitoxin, a potent inhibitor of PEP, was shown to be linear for over 1 h in the presence of heparin (Kim et al., 1993). This suggests that at least in vitro NEP does not leave the template which may be the cause for the overall transcription of the plastid chromosome observed in PEP-deficient plants and, implicitly, 3′ transcript termini different from those of PEP-driven RNAs.

One of the more interesting findings of the study is that transcripts differ not only quantitatively, but in some cases (e.g. psaA, rps14, ycf3 see Figure 6) also qualitatively in wild-type and PEP-deficient plastids (Figure 6). This may be caused by differential processing of transcripts driven by either PEP or NEP. In mammalian cells it has been shown that splicing, processing of 3′ ends and termination of transcription all depend on the presence of a carboxy-terminal domain in RNA polymerase II (McCracken et al., 1997). Specific interaction of processing factors with one of the different plastid RNA polymerases could be the cause for the observed differences in transcript patterns as well. Such factors acting at the interface between transcription and RNA processing may be represented by PEP associated nuclear-encoded polypeptides recently identified in mustard (Pfannschmidt et al., 2000). The fate of a transcript may therefore already be determined by its synthesis by one of the different RNA polymerases.

It is known that substantial amounts of thylakoid proteins are often translated from minute amounts of RNA (e.g. Herrmann et al., 1992). However, although corresponding transcripts accumulate to detectable amounts even in PEP-deficient plastids, a number of proteins with photosynthesis related functions are not detectable in mutant leaf tissue (e.g. PsaA and PsaB photosystem I core proteins; Figure 7). This may be ascribed to the fact that the message can not be translated in a certain context (e.g. from the polycistronic RNA in the case of psaA and psaB) as reported for maize petD (Barkan et al., 1994) and tobacco ndhD (Hirose and Sugiura, 1997) mRNAs. Translation of individual plastid transcripts has also been described to be under control by epistasy in Chlamydomonas (Wollman et al., 1995). It is therefore conceivable that the absence of thylakoid membrane complexes in PEP-deficient tobacco plants prevents translation of certain mRNAs by epistatic effects, too. On the other hand, peptides (and other components of the photosynthetic apparatus) may escape detection if they are made but rapidly degraded, because the thylakoid components are not assembled in mutants. Then, however, it remains puzzling why nuclear-encoded thylakoid components clearly accumulate close to wild-type levels in the mutant plastids in the absence of their plastid-encoded partners (Figure 7; DeSantis-Maciossek et al., 1999).

Nearly all data on NEP-specific transcription patterns were derived from transplastomic tobacco (e.g. Hajdukiewicz et al., 1997; Krause et al., 2000) and naturally arisen mutants (e.g. Hübschmann and Börner, 1998; Silhavy and Maliga, 1998a) lacking the PEP enzyme or ribosome-deficient plastids of heat-bleached plants (e.g. Falk et al., 1993). The PEP-deficient tobacco plants, as in this work, were generally cultivated in tissue culture and varying quantities of supplements like amino acids, hormones and vitamins may influence gene expression. These ‘artificial’ systems may be useful to dissect the activity of different RNA polymerases, but it is difficult to envisage presently (with the data available) that at any developmental stage NEP acts without a contemporary activity of the PEP enzyme. Notably, since transcription of PEP-coding genes appears to be driven by NEP, which results in relatively high transcription rates and transcript accumulation of the plastid-encoded rpo genes in PEP-deficient mutants, one would expect an increased expression of the PEP genes as soon as NEP is active. A major task of the NEP polymerase could be to ensure an increased expression of the PEP enzyme at certain stages during plastid ontogeny and to allow pronounced expression of non-photosynthesis-related genes (e.g. accD, clpP and ycf2) in certain cell/tissue types. In fact, expression of PEP is significantly increased at an early stage of chloroplast development (Baumgartner et al., 1993; Inada et al., 1996), the period where NEP activity was proposed to be highest (Hajdukiewicz et al., 1997). It is inherently difficult to clarify the role of components of a syntonically regulated system by removing one of the ‘team mates’. Recently published data on pale green Arabidopsis mutants defective in the expression of one of the six nuclear-encoded sigma factors, namely SIG2 (Kanamaru et al., 2001; Shirano et al., 2000), are in line with this assumption. Sigma factors interact with the PEP core for regulation (Allison, 2000). In the absence of SIG2, transcripts for several tRNAs are drastically reduced (Kanamaru et al., 2001), whereas transcripts for photosynthesis-related functions accumulate to wild-type levels. Thus, implying a SIG2 activation of PEP, also expression of various genes for the genetic machinery is strongly dependent on PEP. The finding that all tested ‘NEP transcripts’ (e.g. clpP and accD) were increased in the absence of SIG2 complicates the situation further. It suggests a direct or indirect influence of SIG2 on NEP activity, too. However, the amounts of ClpP and AccD polypeptide were not increased in spite of several-fold accumulation of their mRNAs. This may be caused by a general decrease in translational activity in the SIG2 mutants which also seems to be reflected in the reduced amounts of photosynthesis-related polypeptides produced from transcripts present in wild-type levels (Kanamaru et al., 2001). These data can not be reconciled with a simple interactive regulation of NEP and PEP.

In summary, accumulating data on the various RNA polymerases involved in the transcription of plastid chromosomes clearly point out the complexity of mechanisms controlling and implementing plastid gene expression. The plastid is an essential genetic element of the plant cell that traces back to an endosymbiotic event. Data on the expression of plastid chromosomes and its evolution add new insight to the entire genetic set-up of the plant cell. Our current knowledge might well be only scratching the surface of what awaits. To assess the role of the different plastid RNA polymerases it is mandatory to investigate their activities in biological relevant situations, i.e. distinct tissues, cells and developmental stages. Without doubt, this is a exceedingly challenging venture and will continue to be a lively field for the future.

Experimental procedures

Plant material

Tobacco plants (Nicotiana tabacum, cv. Havanna) and transplastomic mutants in which the plastid-encoded RNA polymerase was inactivated by disruption of rpoA, encoding the α-subunit of the enzyme (DeSantis-Maciossek et al., 1999), were chosen for this study. Wild-type and mutant plants were grown at 24°C with 8/16 h dark light−1 cycles at 100 µE m−2 sec−1 standard light (Osram L85W/25 Universal White fluorescent lamps). Wild-type tobacco plants were grown from seeds on B5 medium (Gamborg et al., 1968) containing vitamins (100 mg l−1 myo-inositol, 10 mg l−1 thiamine HCl, 1 mg l−1 pyridoxine HCl, 1 mg l−1 nicotinic acid) and 0.2% sucrose. Transplastomic mutants were regenerated from tissue culture on RMOP medium (Svab et al., 1990) and shoots developing roots were grown on VBW medium (Aviv and Galun, 1985) containing vitamins (100 mg l−1 myo-inositol, 0.01 mg l−1 thiamine HCl, 0.05 mg l−1 pyridoxine HCl, 0.05 mg l−1 nicotinic acid and 0.02 mg ml−1 glycine), hormones (0.2 mg l−1 kinetin and 2 mg l−1 indole-3-acetic acid), amino acids (0.05% casein hydrolysate; ICN Biochemicals, USA) and 0.3% sucrose. Fully expanded leaves of the same size of both mutant and wild-type plants grown on VBW medium (mutants) or B5 medium (wild-type) for additional 6 weeks after shoot regeneration were used for analyses.

Isolation of plastids

Leaf tissue was homogenised in extraction buffer (0.33 m sorbitol, 25 mm HEPES, 25 mm MES, 4 mm Na-ascorbate, 1.2 mm MnCl2, 0.8 mm MgCl2, 4 mm EDTA, 1 mm KH2PO4, 4 mm DTT, 0.2% (w/v) BSA, 0.1% (w/v) PVP-10, pH 6.8). Homogenates were filtered through Miracloth (Calbiochem, La Jolla, CA, USA), and the filtrate was centrifuged at 4000 g for 2 min. The pellet was resuspended in buffer containing 0.33 m sorbitol and 50 mm HEPES/KOH, pH 7.6, and the plastids were fractionated on discontinuous 30%/80% (mutant leaf material) or 40%/80% (wild-type leaf material) Percoll gradients centrifuged at 15 700 g for 20 min at 4°C. The band at the cushion interphase containing intact plastids was recovered. The plastids were washed and finally resuspended in 0.33 m sorbitol, 50 mm HEPES/KOH, pH 7.6. Plastid numbers were determined microscopically, and the suspensions adjusted to 2 × 106 plastids per µl.

Run-on transcription assays

Run-on transcription assays with 2 × 107 lysed plastids were carried out in a 100-µl volume in the presence of heparin as described in Klein and Mullet (1990). Unincorporated nucleotides were removed from the reaction assay by MicroSpin S-200 HR Columns (Amersham Pharmacia Biotech, Freiburg, Germany) following the manufacturer's instructions. Incorporation of α-32P-UTP into elongating transcripts was determined as described by Hallick et al. (1976) with aliquots spotted onto DE81 filters (Whatman, Maidstone, UK).

Isolation of plastid RNA and synthesis of cDNA

RNA was isolated from intact plastids using TRIzol reagent (Gibco/BRL, USA) according to the manufacturer's protocol. The RNA preparations were treated with RNase-free DNase I, extracted with phenol-chloroform and precipitated with 2 volumes of ethanol in the presence of 0.3 m NaOAc, pH 4.8. Aliquots of 3 µg plastid RNA were reverse transcribed in the presence of α-32P-dCTP with Moloney murine leukemia virus RNase H-free reverse transcriptase (SuperscriptTM, Gibco/BRL, USA). The reactions were either primed with random hexanucleotides or with a mixture of 129 oligodesoxynucleotides (8 pmol each), each complementary to the coding strand of one of the plastid encoded genes as well as of several open reading frames. The reaction was terminated by addition of 0.2 volumes of 0.5 m EDTA, pH 8.0. To remove RNA, the product mixture was ethanol precipitated, and the pellet was incubated in 0.3 m NaOH, 5 mm EDTA, pH 8.0, for 30 min at 65°C. The sample was then neutralised by addition of 1.2 volumes of 1 m Tris/HCl, pH 7.5, and unincorporated nucleotides were removed by column chromatography as described above. Oligonucleotide sequences can be obtained on request.

End-labelling of plastid transcripts

5′ ends of DNase I-treated RNA (6 µg) were labelled using 5 units T4 polynucleotide kinase (PNK) in the presence of 30 µCi of [γ-32P]ATP in a final volume of 20 µl according to the supplier's protocol (New England Biolabs, USA). The reaction was stopped by heat inactivation at 65°C for 20 min and remaining nucleotides were removed by the use of columns as described for run-on transcription assays (see above).

PCR amplification of plastid-encoded genes

PCR amplification of each individual gene as well as of open reading frames of the tobacco plastid chromosome was performed using gene/ORF-specific primer pairs on a MultiPROBE II robot (Packard, USA). In cases where genes are interrupted by introns, the respective spliced molecules were amplified using total plastid cDNA as template. Those oligodesoxynucleotides priming the antisense strand synthesis were designed to contain a 5′ extension with the T7 RNA polymerase promoter sequence. Thus, each amplicon is flanked by an identical 30-nucleotide sequence. This allows checking the uniformity of DNA quantities loaded on the nylon filters with a T7 promoter-specific probe. In addition, these amplicons provided templates for the synthesis of strand-specific RNA probes for each plastid-encoded transcript.

RNA gel blot analysis

RNA was electrophoresed in 1.2% agarose formaldehyde gels and transferred onto Hybond N+ membrane (Amersham Pharmacia Biotech) by capillary blotting. The transferred RNAs were hybridised in ,Church buffer (0.25 m Na2HPO4, 7% SDS) at 65°C with strand-specific RNA probes generated by transcription of gene-specific PCR products (see above) in the presence of [α32P]UTP using T7 RNA polymerase.

Nylon filter array preparation

Amplicons were arrayed on 7.8 × 11.9 cm positively charged nylon filters (Hybond-N+, Amersham Pharmacia Biotech) using a 96-pin tool (0.4 mm pins) with a BioGrid spotting device (BioRobotics, UK). Assuming that a pin of 0.4 mm diameter transfers 20 nl of liquid per spotting event three different dilutions of each amplicon (6.25, 25 and 100 ng µl−1, respectively) were prepared. Each sample was spotted 10-fold to the same position in order to obtain final DNA quantities of 1.25, 5 or 20 ng per sample on different spots. Each sample was arrayed in duplicate using a spotting density of 3 × 3. As a negative control pBluescript vector DNA was spotted onto the filters. Prior to hybridisation the arrayed DNA was UV-crosslinked to the filter.

Array hybridisation

Arrays were pre-hybridised for 2 h at 65°C in Church buffer (0.25 m Na2HPO4, 7% SDS). Hybridisation of 32P-labelled run-on transcripts, 32P-end-labelled RNAs and 32P-labelled cDNAs, respectively, was performed under the same conditions over night. Arrays were then successively washed for 10 min at 65°C in 2 × SSC, 0.1% SDS, 10 min at 65°C in 1 × SSC, 0.1% SDS, 10 min at 65°C in 0.5 × SSC, 0.1% SDS followed by a final wash at 65°C in 0.2 × SSC, 0.1% SDS for 10 min.

Image analysis

Radioactive signals were detected using a Fujifilm BAS 1500 Phosphorimager (Fuji, Japan). Images were directly imported into the TINA (raytest, Germany) or ArrayVision software (Imaging Research Inc., Canada) and analysed. In the case of arrays intensities of duplicate spots representing the same gene were averaged. Background correction was done against the negative control (pBluescript).

Immunoblot analysis

Total leaf proteins separated in SDS polyacrylamide gels were electrophoretically transferred onto nitrocellulose filters (Schleicher und Schuell, Dassel, Germany). The filters were probed with specific antisera and developed with goat antirabbit IgG serum conjugated with horseradish peroxidase. The peroxidase activity was detected by reaction with luminol and H2O2 as described by the supplier (Amersham Pharmacia Biotech).

Primer extension analysis

Primer extension reactions were carried out on 50 µg of total leaf RNA with SuperscriptTM reverse transcriptase (Gibco/BRL, USA) and a fluorochrome labelled primer (5′-IRD700-CCCTCTGAA TCAAACTTTTCCTTGCC-3′; MWG Biotech, Ebersberg, Germany) annealing to the region between + 3 and + 21 relative to the first nucleotide of the rps14 coding sequence. DNA sequences were generated by using the same primer and the Thermosequenase kit (Amersham Pharmacia Biotech). Products were analysed with the LI-COR 4200IR2 two-laser system (MWG Biotech, Ebersberg, Germany).


We thank Christian Schmitz-Linneweber for helpful comments and Dr Werner Gunja (Imaging Research Inc., Canada) for assistance in array evaluation. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 184, SFB-TR1) and the Fonds der Chemischen Industrie.