INTRODUCTION

In accordance with the hypothesis that plastids are of endosymbiotic origin most plastid genes are organized into polycistronic transcription units reminiscent of bacterial operons. Plastid rRNA operons show the typical procaryotic gene order of 16 S, 23 S, and 5 S rDNA. These genes are transcribed as large precursor RNAs that are subsequently processed into the various mature rRNA species (1, 2).

The promoter regions of plastid rrn operons harbor Escherichia coli-like "-10" and "-35" consensus sequences, like most of the plastid transcription units. These E. coli-like consensus sequences serve as promoter structures (3-5) or as regulatory elements (6, 7). However, the interpretation of results on studies of transcriptional regulation in plastids is complicated by the existence of different types of RNA polymerases. One is nuclear encoded (8-11) and the other one is encoded on the plastid genome (12-15). The plastid-encoded enzyme can be considered as "E. coli-like" with respect to its subunit composition (16, 17) and promoter usage (6, 18-20). The composition of the nuclear-encoded RNA polymerase and the promoter structures that are used by this enzyme are not yet clear although several potential transcription start sites for this enzyme have been mapped (5, 21, 22).

In spinach, the rrn operon upstream region contains three different promoter elements (P1, PC, P2), and transcription is thought to be regulated by the transcription factor CDF2 (23). CDF2 acts as a repressor of rRNA transcription by the E. coli-like plastid RNA polymerase and probably as an activator of rRNA transcription by the nuclear-encoded RNA polymerase (6), i.e. rRNA transcription could be regulated exclusively by CDF2. On the other hand, up to now correct initiation at the putative PC start site could not be demonstrated in vitro raising the question whether PC is indeed an initiation site. In addition, two transcription start sites that are different from the three of spinach have recently been reported for the tobacco rrn operon (5). They are differentially used in leaf chloroplasts or in amyloplasts of cultured cells. Considering the high sequence homology of the tobacco and the spinach rrn operon promoter regions, it is very surprising to find differences in rrn expression patterns between two closely related plants. Therefore, we analyzed the rrn expression patterns in both plants in parallel under the same experimental conditions. Also, we analyzed rrn expression of different plastid types in intact plants, because gene expression in cultured cells (in particular in BY2 cells, which are not competent for regeneration) might differ from that in intact tissues of plants.

In general, the rate of ribosome formation (including rRNA synthesis) is adapted to the individual cellular requirement for protein biosynthesis. Correspondingly, overall transcription rates of rrn operons change drastically in accordance with changes of cellular metabolic activities. For instance, in E. coli the rRNA transcription is subject to stringent and growth-rate control (for review see Ref. 24). In higher plants, overall transcription rates of the plastid genome change drastically in a developmental and organ-specific manner (25-27). The transcription rate is about 30 times higher in leaf chloroplasts than in root amyloplasts (27). This difference in gene expression is related to the photosynthetic activities of chloroplasts and correlates with the number of plastid ribosomes. Consequently, rRNA expression should be regulated on the transcriptional level to respond to changes in plastid metabolic activities.

To learn more about the regulation of rrn transcription in different plastid types of intact plants, we have analyzed rrn expression in spinach root and cotyledon tissues by capping and primer extension experiments to determine and compare transcription start sites and transcript quantities. We also analyzed protein/rrn promoter interactions using different protein extracts. Results are interpreted with respect to the presence of multiple RNA polymerases in plastids (for recent review see Ref. 28), and the rrn transcription patterns obtained from plastids of intact plants are compared with transcription patterns observed using cell cultures.

EXPERIMENTAL PROCEDURES

Spinach (Spinacia oleracea L. var. Geant d'hiver) was grown either in vermiculite at 15 and 22 °C (night/day) in a 10-h light/14-h dark cycle or in darkness or kept on moistened filter paper in darkness for 7 days and subsequently illuminated for 2-8 h. After 1 week, cotyledons and roots were collected, washed extensively (three times with sterile water), and homogenized for plastid purification. For nucleic acid analysis, the plant material was immediately frozen in aliquots of 2 g in liquid nitrogen. Tobacco leaves were taken from 45-day-old plants.

Frozen tissue (2 g) was ground in liquid nitrogen until a very fine powder was obtained. The powder was poured immediately into an ice-cold mixture of phenol/chloroform/buffer (1:1:2; 10 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, and 1% SDS) for nucleic acid extraction. DNA was removed by two successive LiCl precipitations. RNA was dissolved in sterile water to a final concentration of 2 µg/µl and stored at 20 °C in 15-µg or 1-µg aliquots until usage.

Oligonucleotides were 5-end-labeled with [-³²P]ATP and T4 polynucleotide kinase. The primer (10⁵ cpm, 1-10 ng) was annealed to 1 or 15 µg of the total RNA, and cDNA synthesis was performed using 50 units of Moloney murine leukemia virus reverse transcriptase (Boehringer Mannheim). The reaction products were analyzed on 8% polyacrylamide, 7 M urea gels. Dideoxy-sequencing reactions were performed on double-stranded plasmid DNA using -³⁵S-ATP and the T7 sequencing kit from Pharmacia Biotech Inc. Each primer extension series was performed in parallel using different primer/RNA ratios to verify linearity of signal responses.

For S1 nuclease mapping of capped RNA, 300 ng of single-stranded DNA were hybridized to 15 µg of capped RNA and digested with 100 units of S1 nuclease at 20 °C for 2 h if not otherwise indicated. Capping reactions were performed in 30-µl aliquots containing 15 µg of total RNA, 150 µCi of [-³²P]GTP, and 25 units of guanylyltransferase (Life Technologies, Inc.) in a solution containing 50 mM Tris-HCl, pH 7.9, 1.25 mM MgCl₂, 6 mM KCl, 1 mM dithiothreitol at 37 °C for 30 min.

The 831-bp¹ BglII-PvuII plastid DNA fragment (30) was recloned into Bluescript KS+ after excision by EcoRI from the PHp 34 plasmid. Clones were selected containing the 16 S rRNA 5-end oriented in the opposite direction of the -galactosidase gene. The 831-bp EcoRI fragment was further cleaved by HinPI and inserted into Bluescript KS+ cleaved by EcoRI and AccI.

Plastid DNA was amplified using primers e and f (primer e, 5-AAGATTTGGCTCGGCATG-3; primer f, 5-CCATAGGTACAGCGTTTG-3), and the corresponding fragment was cloned into pCR^TMII (Invitrogen) according to the manufacturer's protocol.

50 ng of primer were annealed to 3 µg of capped RNA in a final volume of 10 µl (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 2 mM MgCl₂, 10 mM dithiothreitol) for 10 min at the melting temperature for the primer. 1 µl of RNase H (5 units) was added, and the reaction was incubated at 37 °C for 20 min. The reaction was stopped by adding 10 µl of loading buffer and was directly run on a denaturing polyacrylamide gel. Control reactions are incubated for the same time without primer or without RNase H.

Purification of root plastids was done according to Deng and Gruissem (27). Isolation of chloroplasts, preparation of protein extracts, and gel shift assays were performed as described (6).

RESULTS

Primer extension assays were optimized to give quantitative results. Fig. 1A shows the analysis of cotyledon and root RNA using a primer that anneals within the mature 16 S rRNA (primer a, 5-GGGCAGGTTCTTACGCGT-3, for primer localization see scheme of Fig. 2A). Radioactivity incorporated into the cDNA that corresponds to mature 16 S rRNA was counted. The obtained values indicated a 77-fold difference in rRNA content in root and cotyledon tissues, which corresponds well to previously reported results (27), and thus confirms that our assay conditions are quantitative. In addition to the mature rRNA, two other RNAs are revealed in cotyledon tissues (Fig. 1A, lane C). The band named Pro2 corresponds in size to the processing intermediate, which was previously characterized in maize and tobacco chloroplasts (3, 29), suggesting that this processing site is universal in plastids of different plant species. The band named PC corresponds to the previously characterized putative transcription start site of the spinach rrn operon (23). Interestingly, PC is not detected in root tissues (lane R). Pro2 appears as a very faint band. The intermediate sized transcript that is marked with an asterisk is probably an artifact since it is not revealed with another primer (compare with Fig. 1B, lane R). The cDNA of about 900 bases indicates the presence of very long precursor transcripts in root tissues. Such a long transcript should include the tRNA(GAC)^Val that is encoded upstream of the rrn operon. Therefore, we named this site PtRNA. The band corresponding to PtRNA is also detectable in cotyledon tissues after longer exposure (not shown).

The quantitative difference of the PC transcripts in cotyledon and root tissues was re-analyzed with the same RNA preparations using primer g (5-TTCATAGTTGCATTACT-3). The sequence of primer g is complementary to the rrn precursor region immediately upstream of mature 16 S rRNA. Therefore, the majority of the radiolabeled primer is not trapped by mature rRNA, and the sensitivity of the method is amplified by more than 100 times. Also under this condition, cDNA corresponding to PC initiation is not detectable in root tissues (Fig. 1B, lane R). RNA corresponding to Pro2 cannot be detected since the primer is positioned too close to it. No other intermediate sized RNA is revealed, including that corresponding to the product from the promoter for rrn transcription in cultured tobacco cells (P_tobacco, see Ref. 5).

Pro2-cleaved precursor transcripts are abundant in cotyledon/leaf tissues (Fig. 1A) and are present in newly assembled plastid ribosomes (29). To analyze the fate of the supposed tRNA(GAC)^Val-containing precursor transcript (PtRNA, Fig. 1A) after Pro2 cleavage, we used a primer that is located upstream of processing site Pro2 for primer extension (Fig. 1C; primer c, 5-TCTTTCATTCCAAGGCATAACTTGT-3). Again, PC is observed only in cotyledon tissues but not in root tissues (lanes L, D, and R). In addition, a higher molecular weight cDNA is revealed in roots and cotyledons (Pro1). Fine mapping localizes the 5-end of this transcript 125 nucleotides upstream of the tRNA(GAC)^Val gene (Fig. 1, D and E, for sequence see Ref. 30). The amount of these transcripts does not differ considerably in cotyledon tissues of light-grown or dark-grown plantlets (Fig. 1C, lanes L and D) and in root tissues (lane R).

Next we addressed three questions. 1) Is the rrn operon indeed cotranscribed with the tRNA^Val(GAC) gene, and is PtRNA the transcription start site for this large transcript? 2) If so, is PC a cotyledon/leaf-specific transcription start site or a cotyledon/leaf-specific processing site of the large transcript that starts at PtRNA? 3) Is Pro1 a processing site of the PtRNA-initiated precursor RNA or another transcription initiation start site that produces a transcript that is terminated upstream of Pro2?

To ensure that PtRNA corresponds to cotranscription of the tRNA(GAC)^Val gene with the rrn operon we analyzed root and cotyledon RNA by RT-PCR (Fig. 2B). The cDNA was primed within the sequence of mature 16 S rRNA (primer a), and for amplification a second primer was used that anneals within the tRNA(GAC)^Val gene (primer b, 5-GGTATAACTCAGCGGTAG-3). RNA corresponding to cotranscription of the tRNA(GAC)^Val gene, and the rrn operon is detected in roots and in cotyledons (Fig. 2, lanes 1 and 4). Controls were made omitting reverse transcription prior to PCR amplification (lanes 2 and 5) or omitting primer b during amplification (lanes 3 and 6). The corresponding 417-bp product was cloned and sequenced. The 5 and 3 parts of the sequence are shown in Fig. 2B on the right-hand side.

RT-PCR was also performed with primer pairs b/c and d/c (Fig. 2C; primer d, 5-GGGGTTGATCCGTATCAT-3). Transcripts covering the sequences between primers c and b are more abundant than transcripts extending between primers c and d (Fig. 2C) or a and b (Fig. 2B). This difference could be explained by a processing event that takes place at Pro1. Pro2 has already been determined as a processing site (3, 29).

Having confirmed the existence of transcripts extending from tRNA(GAC)^Val downstream into the rrn precursor region, it is necessary to distinguish between transcription initiation and RNA-processing sites.

With the method of primer extension, it is not possible to distinguish whether the revealed RNA results from transcription initiation or from processing (see above and also Ref. 23). Transcription initiation can only be shown by in vitro capping of RNA that contains a 5-triphosphate followed by positioning of the capped 5-end of the RNA by S1 nuclease mapping. Therefore, we analyzed cotyledon and root RNA after in vitro capping. Complementary single-stranded DNA was produced from a cloned DNA fragment that comprises 140 bp of the mature 16 S rRNA gene and 691 bp of the upstream region (831-base probe in Fig. 2A; for sequence see Ref. 30). We observe only one transcript in cotyledons that contains a 5-triphosphate (Fig. 3A, lane 1). This transcript corresponds from its size to initiation at PC. In roots, we could not detect any transcript under the same experimental conditions (lane 2).

To analyze the Pro1-corresponding transcript we cloned a 603-bp DNA fragment that comprises the Pro1 site and the tRNA(GAC)^Val gene but does not extend to the Pro2-processing site (see Fig. 2A). The analysis of capped root (lane 1) and cotyledon (lane 2) RNA is shown in Fig. 3B. The two bands that resist S1 nuclease digestion (lane 2) correspond to the A and G nucleotides located 3 and 4 bases upstream of the primer extension signal (23). This difference of 3 and 4 bases can be explained by the addition of GTP to the analyzed RNA in the capping reaction and to the fact that cDNA (primer extension) and RNA (capping) migrate differently in polyacrylamide gels (31). Lane 3 shows the profile of the capped RNA without hybrid selection.

These results show that PC is a transcription start site of the 16 S rRNA in cotyledon/leaf plastids of spinach plantlets. In root amyloplasts, the 16 S rRNA is cotranscribed with the tRNA(GAC)^Val gene. In cotyledon/leaf tissues, both types of transcription exist (cotranscription with tRNA(GAC)^Val and transcription initiation at PC). The transcript corresponding to Pro1 seems to result from processing since no capping could be demonstrated (Fig. 3B). The start of the primary rrn transcript, which includes the tRNA(GAC)^Val, could not be determined by capping and S1 nuclease mapping because it starts upstream of the DNA fragment that was used for hybridization, i.e. more than 691 bp upstream from the sequence coding for mature 16 S rRNA.

To localize the transcription start site of the tRNA(GAC)^Val gene and to ensure that this transcript in fact extends into the mature 16 S rRNA we had to develop a new method. If the large transcript stops before the mature 16 S rRNA-coding region, S1 nuclease mapping would give wrong results as we cannot determine the 3-end of the hybrid since the labeling by capping is at the 5-end of the transcript. (Usually the 3-end of the single-stranded DNA is labeled.) Therefore, we developed a method that is based on the enzymatic property of RNase H to digest RNA in DNA-RNA hybrids. After the capping reaction, one part of the RNA is annealed to primer a (which anneals within mature 16 S rRNA) and digested with RNase H. The ribosomal precursor RNA should be cleaved at the position of the primer. The product(s) of this reaction is (are) analyzed concomitantly with two control reactions, one which contains only the primer and another one that is treated with RNase H but in the absence of primer. What we are looking for now is the appearance of an additional band within the pattern of capped total RNAs that is not seen in the two control reactions. The size of this band corresponds to the distance of the transcription start site(s) upstream from the position of the primer.

First of all, we analyzed whether capping reactions are reasonably reproducible. This is essential to identify newly appearing bands in total capping patterns. Fig. 4A shows results with total capped RNAs (lanes 1-4). Spinach seeds were germinated in darkness for 7 days (lane 1), and seedlings were subsequently illuminated for 2 (lane 2), 4 (lane 3), or 8 (lane 4) h. Total RNA was isolated from cotyledons and labeled by capping. The most strongly labeled band of about 120 bases was isolated and sequenced. It corresponds to cytoplasmic 5 S RNA (not shown). To avoid the high background of this band the following gels were run long enough to elute it from the gel.

RNase H mapping of capped cotyledon RNA using primer a produces two different supplementary bands of about 250 and 900 bases (marked by arrows in Fig. 4B, lane 2). The 250-base RNA corresponds to PC. The 900-base transcript confirms the existence of a very long rrn precursor transcript that comprises the tRNA(GAC)^Val and corresponds well in size to the large transcript revealed by primer extension using the same primer (see Fig. 1). The asterisk (Fig. 4B) indicates the position predicted for the second rrn promoter that was observed in tobacco plastids (5, 21).

For fine mapping of the 900-base RNA, we hybrid-selected capped RNA using complementary single-stranded DNA with the 5-end localized about 200 nucleotides downstream of PtRNA. The S1 nuclease-protected RNA is shown in Fig. 4C. It localizes the transcription start site (PtRNA) 457 bp upstream of the tRNA(GAC)^Val gene. This coincides well with the 870-base RNA found by cleavage with primer a in the RNase H mapping experiment. PtRNA is not preceded by E. coli-like consensus sequences (Fig. 4D) suggesting transcription by the nuclear-encoded plastid RNA polymerase. If we compare PtRNA with the main rrn transcription start site in rpoB-depleted tobacco plastids (21) we notice that 7 of 8 bases are identical (Fig. 4D, bold letters).

A comparison of spinach and tobacco rrn promoter regions is shown in Fig. 5. Differences in base composition are marked by open boxes, transcription start sites are indicated by bent arrows. The fact that PC is obviously not used in spinach root tissues and that transcription start sites in this region are different in spinach (PC) and tobacco (P1 and P2) plastids prompted us to analyze and compare DNA/protein interactions by gel retardation. The rrn promoter fragment (WT) and the promoter mutation (M1) on which CDF2 does not bind (6) were analyzed in the presence of plastid extracts obtained from spinach cotyledon (Fig. 6B, lanes 1-6) or root tissues (Fig. 6B, lanes 7-12). Taking into account that rrn transcription is 50 to 80 times lower in root than in leaf tissues we used much higher protein concentrations to analyze root extracts compared with leaf extracts. The formation of the CDF2-promoter complex is detectable using leaf extracts (lanes 1-6). No protein/promoter interaction is detected with root plastid extracts even if the protein concentration was scaled up to 52 times that of cotyledon extracts (lanes 9 and 12).

Fig. 6C shows DNA/protein interactions among the spinach rrn promoter DNA fragment, three mutated fragments (see Fig. 6A), and tobacco leaf plastid extract. The complex that is formed with the wild-type promoter fragment (lanes 1-3) is highly sensitive to M2 and M3, i.e. to modifications in the "-35" and "-10" sequence elements of the E. coli-like spinach P2 promoter. Using M1, the mutation of the CDF2-binding site, DNA/protein interactions are only moderately reduced (lanes 11 and 12) in contrast to results obtained with spinach plastid extracts (compare with Fig. 6B). This experiment suggests that the differences in spinach and tobacco chloroplast rrn initiation (Fig. 5) are due to differences in protein/promoter interactions in the two plant species.

If we compare root and leaf RNA from spinach and tobacco directly by primer extension analysis using primer c (Fig. 7A), we find that cotranscription of the rrn operon with the tRNA(GAC)^Val gene seems also to exist in tobacco plastids (lane C Spinach and lanes C and R Tobacco, Pro1). The differential localization of the transcription start sites (PC in spinach, P1 in tobacco) is evident (compare lane C Spinach to lane C Tobacco). Analysis of rRNA from tobacco cell cultures (BY2) shows the two reported transcription start sites P1 and P2 (Fig. 7B, lane 2). In contrast, in plastids of spinach cell cultures we determine only transcripts corresponding to PC (lane 1).

DISCUSSION

Regulation of transcription in plastids is not yet well understood, and results on the expression of higher plant rDNA transcription seem to be contradictory. (i) The spinach plastid 16 S rDNA upstream region contains two tandem E. coli-like promoters, P1 and P2, which are efficiently used in vitro by E. coli RNA polymerase (Ref. 32 and Fig. 5). However, in vivo, we could not find products that correspond to these two initiation sites in photosynthetically active organs (7, 23). On the other hand, tobacco plastid rDNA transcription starts mainly at an E. coli-like promoter (P1), which corresponds to the spinach P2 promoter (Refs. 3 and 5 and Fig. 5). (ii) The higher plant plastid genome is transcribed by at least two different types of RNA polymerase (for recent review see Ref. 28). In spinach, plastid rDNA is probably transcribed by the nuclear-encoded enzyme as suggested by in vitro transcription studies (6) and the results presented here. On the other hand, transcription at the P1 promoter of tobacco plastid rDNA is likely to be initiated by the E. coli-like plastid-encoded RNA polymerase (21). (iii) A second minor transcription start site that is not preceded by E. coli-like promoter elements has been demonstrated recently in tobacco plastids. The steady-state level of the corresponding transcript is higher in BY2 plastids than in leaf chloroplasts (5). When the plastid rpoB gene is deleted by plastid transformation only this non-E. coli-like promoter is active suggesting that the minor transcription start site is used by the nuclear-encoded plastid RNA polymerase (21). In spinach leaf chloroplasts, this start site has not been reported so far.

These contradictions prompted us to analyze rrn expression in spinach plastids. We found that transcription of the rrn operon in intact spinach plants is regulated by usage of two promoters. In photosynthetic and non-photosynthetic organs (cotyledons and roots), the rrn operon is cotranscribed with the preceding tRNA(GAC)^Val gene (Figs. 1 and 2). The transcription start site of the tRNA(GAC)^Val-including precursor RNA (PtRNA) does not employ E. coli-like consensus sequences (Fig. 4). Instead, there is some upstream sequence homology to the recently reported rrn promoter that is active in rpoB-depleted plants (Ref. 21, P2 in Fig. 5, and see also Fig. 4C). This suggests that PtRNA is transcribed by the nuclear-encoded RNA polymerase. To compare our results with tobacco we performed primer extension analysis also with tobacco leaf and root RNA. The results indicate that cotranscription of the rrn operon with tRNA(GAC)^Val exists also in tobacco (Fig. 7A).

The enhancement of rrn transcription in photosynthetically active cotyledon/leaf tissues is due to the organ-specific activation of the PC promoter in spinach (Figs. 1 and 3). This transcription start site is different from the main transcription start site of tobacco leaf plastids (Fig. 7A, compare also spinach PC and tobacco P1 in Fig. 5). Also, the analysis of RNA isolated from spinach cell cultures in parallel with tobacco BY2 cells shows differences in rrn expression patterns. The tobacco P2 transcript is only detectable in tobacco BY2 cells but not in spinach cell cultures (Fig. 7B). The reasons for these differences remain unclear. It seems, however, that the spinach PC and the tobacco P1 transcription start sites are recognized by different RNA polymerases and/or transcription factors (6, 21) (Fig. 6). The evolutionary transfer of genes coding for components of the plastid transcriptional machinery to the nucleus and the existence of multiple RNA polymerases in plastids provide a large background for species-specific rearrangements of the plastid transcriptional apparatus. Nevertheless, the difference in transcription initiation between tobacco and spinach remains surprising. Full understanding of these differences will require the purification of transcriptional components and in vitro reconstitution of initiation events in both systems.

The tRNA(GAC)^Val-including precursor RNA is probably processed at the Pro1 site. This can be concluded from the negative capping result (Figs. 4 and 6) and from the quantitative differences of RT-PCR products which start upstream or downstream from Pro1 (Fig. 2C). The hypothetical secondary structure of the Pro1-processing site, as shown in Fig. 1E, locates the cleavage site within the loop of a hairpin structure. It was demonstrated that chloroplasts contain an endonuclease activity that cleaves within loop structures of inverted repeats (33). An enzyme of this type might be implicated in the processing mechanism occurring at Pro1.

Altogether, our results show species-specific assembly of the plastid transcriptional complexes and the existence of novel mechanisms of gene expression in plastids, i.e. the organ-specific usage of different promoter structures. Fig. 8 summarizes the results obtained for spinach plastid rrn transcription.