5′ and 3′ untranslated regions (UTRs) of plastid RNAs act as regulatory elements for post-transcriptional control of gene expression. Polyethylene glycol-mediated plastid transformation with UTR–GUS reporter gene fusions was used to study the function of the psbA, rbcL and rpl32 UTRs in vivo. All gene fusions were expressed from the same promoter, i.e. the promoter of the 16S-rRNA gene, such that variations in RNA and protein levels would be due to the involved UTR elements alone. Transgenic tobacco lines containing different combinations of UTRs showed fivefold variation in the uidA–mRNA level (RNA stability) and approximately 100-fold differences in GUS activity, a measure of translation activity. The rbcL 5′-UTR conferred greater mRNA stability than the psbA 5′-UTR on uidA transcripts. In contrast, the psbA 5′-UTR enhanced translation of GUS to a much greater extent compared to the rbcL 5′-UTR. The psbA 5′-UTR also mediated light-induced activation of translation which was not observed with other constructs. Deletion mutagenesis of an unanalysed terminal sequence element of the psbA 5′-UTR resulted in a twofold drop in uidA-mRNA level and a fourfold decrease in translation efficiency. Exchange of 3′-UTRs results in up to fivefold changes of mRNA levels and does not significantly influence translation efficiency. The mechanical impacts of these results on plastid translation regulation are discussed.
Most plastid genes encode subunits of large functional complexes like photosystems or ribosomes. These complexes consist of nuclear and plastid encoded proteins. Therefore, co-ordinated expression of both nuclear and plastid genes is a basic requirement for the organelle’s biogenesis. The influence of plastid gene expression on nuclear gene expression and vice versa has been well documented in green algae and higher plants. For example, functionally active plastids and chloroplast transcription are necessary for transcription of nuclear encoded plastid genes ( Oelmüller & Mohr 1986; Rapp & Mullet 1991). On the other hand, the plastid gene expression machinery strictly depends on nuclear encoded enzymes and regulatory factors (reviewed in Sugita & Sugiura 1996).
Nuclear encoded genes that encode plastid proteins are regulated primarily at the level of transcription with additional modulation of expression at the level of mRNA stability ( Wanner & Gruissem 1991). Modulation of enzyme activity or modulation of regulatory protein binding may allow plastids to adjust to different light and developmental situations ( Peters & Silverthorne 1995). In plastids, transcriptional regulation of gene expression is less important, although some modulation by light and developmental conditions is found. Changes in transcript levels alone, however, cannot account for the strong changes of protein amounts observed (reviewed in Mayfield et al. 1995 ). Although factors like topology of cpDNA ( Stirdivant et al. 1985 ), cpDNA methylation ( Kobayashi et al. 1990 ), cpDNA phosphorylation ( Tiller & Link 1993), regulation of copy number ( Rapp & Mullet 1991) and differential promoter usage by different RNA polymerases and/or specific regulatory proteins ( Kapoor et al. 1997 ; Sexton et al. 1990 ) are known to influence plastid transcription activity, the expression of many plastid genes is regulated to a large extent post-transcriptionally (for a review on Chlamydomonas see Rochaix 1996; on higher plants see Sugita & Sugiura 1996).
Steady state levels of plastid mRNAs generally appear to diverge from transcription rates as a consequence of different turnover rates ( Klaff & Gruissem 1991; Mullet & Klein 1987; Rapp et al. 1992 ). Turnover rates of plastid transcripts which are far lower than those, for example of bacterial transcripts, vary in a gene specific manner and underlie tissue, light and development dependent fluctuations ( Kim et al. 1993 ). It must be considered that RNA processing and turnover are overlapping events. The 3′-ends of plastid mRNAs are established by processing rather than transcription termination (reviewed in Gruissem & Schuster 1993). The 3′ inverted repeat sequences, a common feature of plastid transcripts, can fold into stem-loop structures and seem to play a role in both RNA 3′-end formation and stabilisation ( Rott et al. 1998 ; Schuster & Gruissem 1991). Several proteins, which either have endonuclease activity themselves or which may induce degradation or stabilisation via protein/protein interactions, are known to bind to the 3′-UTRs. A prominent member of this class of RNA-binding proteins (RNP) is the spinach 28 kd RNP, which shows a phosphorylation modulated RNA affinity and is necessary for 3′ processing and stability ( Lisitsky & Schuster 1995). Sequence elements responsible for regulation of transcript turnover are generally located within the 5′ and 3′ untranscribed regions. A recently published analysis in transplastomic tobacco has demonstrated the capacity of the rbcL leader to stabilise the transcript of a downstream uidA reporter gene transcript ( Shiina et al. 1998 ). However, the coding region may also modify RNA stability via secondary structure or by providing targets for initial endoribonuclease attack followed by exonucleolytic degradation ( Klaff 1995). Biochemical analysis has revealed degradation pathways for transcripts of several genes including psbA and rbcL. Polyadenylation after endonucleolytic cleavage promotes rapid degradation of transcript fragments and thus has a completely different function than in the nucleus ( Kudla et al. 1996 ; Li et al. 1998 ). In addition, mRNA stability is also influenced by translation: ribosome binding seems to facilitate turnover of psbA and rbcL transcripts ( Gruissem & Schuster 1993).
Translation is an important factor in post-transcriptional regulation since it can provide a tool for the rapid adjustment of protein levels from an existing pool of mRNAs. Although plastid ribosomes are very similar to their bacterial counterparts, in which transcription is coupled with translation (reviewed in Delp & Kössel 1991), plastids have evolved a complex translational control system, probably as a consequence of the integration of plastid and nuclear genomes into a novel, co-ordinated genetic system in the eukaryote formed by endosymbiosis. Nuclear gene products are known to be essential for translation of all, or specific, plastid transcripts (reviewed in Danon 1997). Many of these proteins bind to 5′-UTR sequences, such as the 5′-UTR of psbA in spinach ( Alexander et al. 1998 ). In Chlamydomonas, a psbA specific translation factor, RB47 which is a member of the poly(A) binding protein family has been identified ( Yohn et al. 1998a ; Yohn et al. 1998b ). This factor is essential for psbA translation and in addition mediates light regulation. As there are at least two different mechanisms for the initiation of translation (one using Shine Dalgarno sequences, the other one lacking this consensus box, for a review see Danon 1997), a complex set of interacting proteins not yet fully examined is to be expected.
We have analysed the influence of homologous tobacco psbA, rbcL and rpl32 UTR elements on post-transcriptional regulation. The availability of a tobacco plastid transformation system via PEG treatment of protoplasts ( Golds et al. 1993 ; Koop et al. 1996 ; for protocols see Koop & Kofer 1995) made it possible to follow the expression of a series of UTR–uidA reporter gene fusions in vivo.
UTR elements from psbA and rbcL were chosen for this analysis because these are well-characterised monocistronic genes, and their transcripts undergo limited RNA processing. Moreover, psbA and rbcL RNA stability, RNA binding proteins and translation have been analysed in Chlamydomonas and several higher plant species including barley, spinach and tobacco. No such information is available for the transcripts of the rpl32 ribosomal protein gene. The 3′-UTR of rpl32 has been included in our experiments as it is derived from a gene, which is expected to be expressed constitutively. The main goal of our transformation experiments was to test if and how isolated 5′ and 3′ UTR sequences confer regulation of mRNA stability and translation efficiency to a neutral reporter gene (uidA). Reporter gene fusions with combinations of UTR sequences from different genes were used to study 5′/3′ UTR interactions.
To unambiguously separate the 5′-UTR influence on mRNA levels from promoter-mediated regulation, all leader sequences were precisely isolated according to published transcription maps (for rbcL and psbA see Shinozaki et al. 1986 ; for rpl32 see Vera et al. 1992 ) and placed behind the same promoter sequence (from the tobacco 16S-rRNA gene). In addition, sequences 3′- to the uidA coding region were consistent with in vivo 3′-transcript termini. This contrasts to previous reports where similar plastid reporter gene constructs were investigated ( Staub & Maliga 1993; Staub & Maliga 1994). These authors used the complete 5′-regulatory sequence elements, i.e. the promoters and 5′-UTRs of the genes in question (psbA, 16S rRNA, rps16), which means that regulation of gene expression was a result of both elements rather than of the UTRs alone.
We obtained transplastomic lines harbouring eight different reporter constructs. Their designation and structure is depicted in Fig. 1 and listed in Table 1. Molecular confirmation of transformation and correct insertion of both aadA and uidA cassettes was repeated at several stages of the experiment including F0 and F1 greenhouse plants.
Table 1. Construct names and integration efficiency of GUS:UTR fusion series
Numbers in brackets behind line designation indicate orientation of the GUS insert (see Fig. 1).
After approximately five cycles of regeneration in the presence of spectinomycin.
During this procedure we could observe genomic instabilities of transformed lines leading to the loss of either the GUS-cassette alone or both transgenes. This excision seemed to be precise, as Southern blots showed the expected sizes for plastomes with only the aadA gene integrated or wild-type bands ( Fig. 2). This phenomenon may be due to intermolecular recombination events with remaining wild-type plastome copies during the segregation process. Most likely, the 90 bp plastome fragment between the uidA and the aadA cassettes (positions 115265–115355) has served as a homologous sequence for recombination events resulting in the loss of the GUS reporter gene. In the case of line 108–6-(I) (psb::rpl) an intramolecular recombination between the same 90 bp fragment and the 3′ end of rpl32 could also be the reason for instability. The blots ( Fig. 2d, line 108–6-(I)) clearly demonstrate the presence of two different classes of recombinant plastome molecules in the same plant.
Reporter gene elements interfere with transformation efficiency
There was a clear correlation between the UTRs fused to the uidA-gene and the yield of transformants bearing the desired insertions: integration into the plastome of GUS constructs flanked by both 5′ and 3′ UTRs from the same gene was far less efficient than integration of those which were flanked by UTRs from different genes ( Table 1). Note that in this series of experiments transformation efficiency in general was lower than efficiencies reported earlier ( Koop et al. 1996 ). A possible reason could be that the coding regions of the respective genes were replaced by the uidA cassette (psbA by psb::psb, rbcL by rbc::rbc) by homologous recombination during or after transformation. As the removal of either rbcL or psbA coding regions most likely results in pale or white phenotypes, such colonies would not have been selected because they would be indistinguishable from untransformed tissues on spectinomycin plates. Another case of negative influence of a UTR element on transformation efficiencies is given by the reporter gene series containing the 3′UTR of rpl32: from three independent transformation experiments only a single line could be obtained (108–6-(I)). This line also shows the genomic instability mentioned above.
Consequences for transformation vector design
In conclusion, these results demonstrate frequent intra- and intermolecular recombination events due to the presence of the 90 bp plastome sequence between the two marker cassettes. In addition, we assume replacement of the plastid genes when vectors with the 5′ and 3′ UTRs of the same gene were used. These results stress that in vector design for plastome transformation all elements containing unnecessary regions of homology should be strictly avoided. This is particularly true when homologous sequences in the vectors and the insertion site region are in close proximity.
Transcription patterns depend on adjoining sequences
To monitor any influence of the insertion direction of the GUS-cassettes on transgene expression we generated transformants with the same uidA-construct inserted in both orientations. No significant variation of mRNA or protein accumulation could be observed (data not shown). However, there was a vast difference in transcript patterns: with orientation I ( Fig. 3) which was used for all the other constructs a single transcript size (about 2 kbp) was found. In contrast, more than 70% of the uidA transcripts originating from the GUS gene in orientation II were 3.5 kbp, thus overlapping the whole aadA cassette (see Fig. 1).
Supposedly, these differences are due to sequence elements 3′ of the insertion site (starting at position 115 355) which terminate transcription very efficiently. Alternatively, processing might be different in the two orientations.
GUS mRNA and GUS protein accumulation is regulated by UTRs
Site-directed insertion of a reporter gene via homologous recombination excludes variability of gene expression by unknown position effects. Equal reporter gene expression of independently transformed greenhouse plants harbouring the same reporter construct was verified for two lines generated with the same vector. For further analysis, one transgenic line for each uidA construct was used. In cases in which a proportion of the plastome had lost its GUS gene the intensity ratios in Southern blots between ‘transgenic’ and ‘wild-type’ bands ( Fig. 2d) were used to adjust the calculation of mRNA and protein levels. The accumulation of uidA mRNA in the transgenic seedlings was determined by Northern blotting. For comparison with endogenous transcript levels, another Northern blot is shown in which probes for the endogenous rbcL, psbA and 16S transcripts and for uidA have been used ( Fig. 3). As the GUS cassette was under the control of the same promoter in all lines, different steady state levels of mRNA should only depend on the turnover of transcripts and thus reflect mRNA stability differences conferred by the different UTR elements. We also tried to measure mRNA stability directly by treating leaf pieces with the transcription inhibitor actinomycin D and monitoring the decay rates of the uidA transcripts. Although we found a good correlation of the uidA decay rates with the steady state uidA mRNA levels in some lines, these experiments were difficult to interpret (data not shown). The reason is that, in contrast to spinach in which this method is well established ( Klaff & Gruissem 1991), tobacco plastid transcripts have extremely low turnover rates. When measuring uidA transcript degradation over the long time periods necessary to get characteristic decay kinetics (more than 24 h), a general degradation process was observed in the leaf material. It is difficult to clearly separate the UTR-mediated specific uidA decay after actinomycin D treatment from this general degradation.
Fluorimetric GUS assays were used to measure GUS protein accumulation. Relative GUS activities observed in plants grown in the greenhouse were higher than in shoot cultures of genetically identical plants. The highest activities were found with psb::psb (line 74–1-(I)), where 120 000 pmol 4-MU (h μg protein)–1 were determined, which is about 100 times higher than the activity of a nuclear tobacco transformants (35S promoter) generated in our laboratory (unpublished data). The ratios between GUS activity and uidA mRNA levels derived from Northern blots ( Fig. 3) were determined and are given as ‘relative translation efficiency’ in Table 2. Whereas the observed mRNA levels varied up to fivefold, differences of more than 100-fold were observed for protein amounts. Comparing lines 74–1-(I) (psb::psb) with 120–9-(II) (rbc::rbc), nearly threefold higher mRNA levels are found in the latter, indicating higher transcript stability mediated by the rbcL UTRs. In plant lines with a combination of UTRs from both genes, the presence of the 5′ UTR of rbcL leads to significantly higher mRNA levels. Thus, the stabilizing effect of this UTR is stronger than that of 3′ UTR of the same gene (see line 120–9-(II) versus 111–5-(I) and 109–9-(I)). This stabilizing influence of the 5′ element of rbcL is enhanced in combination with the 3′ UTR of the same gene. Construct psb::rpl (line 108–6-(I)) yielded the highest steady state mRNA level measured, indicating that the 3′UTR of rpl32 stabilised the transcript. In contrast, the transcript with a synthetic ribosome binding site (consisting of a 26 bp DNA fragment including the ribosome binding consensus sequence GAGG) as the 5′ UTR, i.e. RBS::rbc (line 73–4-(I)), seemed to cause rapid RNA turnover resulting in low mRNA levels. In conclusion, it is clear that both 5′- and 3′-UTRs can modulate RNA stability. Protein accumulation does not necessarily follow the same pattern as mRNA levels. Translation of all transcripts containing the 5′ UTR of psbA (lines 74–1-(I), 109–9-(I) and 108–6-(I)) is highly efficient. Translation efficiency also seems to be influenced by the 3′UTR chosen (compare lines 74–1-(I) and 108–6-(I)), the degree of this influence varies among the different combinations. Lines with transcripts harbouring the 5′ UTR of rbcL produced less than 1% of the relative GUS protein amounts of those with the 5′UTR of psbA, indicating a comparatively low translation efficiency of such transcripts. There was no significant difference between a combination of the 5′ UTR of rbcL either with its own 3′ UTR or that of psbA.
Table 2. Relative levels of steady state mRNA, GUS protein and translation efficiency (GUS protein per uidA mRNA) in leaves of transplastomic Nicotiana tabacum plants
Relative mRNA levels (%)
Relative GUS protein levels (%)
Relative translation efficiency (%)
RNA and protein data originate from at least three independent experiments. Values of line 108–6-(I) were set to 100%. One hundred per cent GUS activity corresponds to 123 000 pmol 4-MU (μg protein h)–1. Experiments were performed with the leaves of greenhouse plants.
33 (± 5)
77.0 (± 10)
39 (± 8)
88.0 (± 10)
100 (± 15)
100.0 (± 12)
92 (± 7)
0.9 (± 0.2)
60 (± 5)
0.6 (± 0.1)
22 (± 5)
0.5 (± 0.1)
Light regulation and tissue specificity are mediated by 5′UTRs
To test tissue specific regulation, mRNA and GUS protein levels of lines 73–4-(I) (RBS::rbc), 120–9-(II) (rbc::rbc), 111–5-(I) (rbc::psb), 108–6-(I) (psb::rpl), 74–1-(I) (psb::psb) and 109–9-(I) (psb::rbc) were determined in photosynthetic (leaves) and non-photosynthetic tissues (roots). Since these tissues are characterized by pronounced differences in general transcription activities it was necessary to discriminate between changes in protein levels caused by differences in transcription and those caused by the UTR elements involved. Steady state levels of 16S-rRNA as determined by using probe ‘16S’ ( Table 3) were about 20 times higher in leaves than in roots (data not shown). The differences in relative uidA transcript levels between the tissues were in the same range for all lines tested. Much larger differences were found in the amounts of GUS protein ( Fig. 4) with all lines containing the 5′ UTR of psbA showing 50 (psb::rpl) to 120 (psb::psb) times higher protein levels in the leaves, whereas 12- to 15-fold differences were found with the other 5′ UTRs analysed. Light regulation was examined by comparing the GUS expression of dark and light grown seedlings of the same transgenic lines, except line 108–6-(I), 14 days after germination. Again, only moderate (two- to threefold, data not shown) differences in uidA mRNA levels contrasted with large differences in the amount of resulting GUS protein ( Fig. 5). Relative GUS activities of light versus dark grown seedlings of all lines carrying the 5′ UTR of psbA exceeded those of lines harbouring GUS fusions with other 5′ UTRs. Moreover, as found in leaf tissues ( Table 2), the absolute values for GUS activity derived from light grown seedlings containing the 5′ UTR of psbA were more than two orders of magnitude higher than those from seedlings with different 5′ UTRs. As with tissue specificity between leaves and roots, the 3′ UTRs chosen also had some, although less pronounced, influence on light regulation. Combining the UTRs from the same gene seemed to lead to a slightly higher light inducibility than using UTR elements from two different genes ( Fig. 5).
Table 3. Probes used for Southern and Northern blotting
Mutation of the 5 ′ UTR of psbA results in decreased translation efficiency
Several binding domains for RNA binding proteins (RNABP) have been described for the 5′ UTR of psbA ( Alexander et al. 1998 ; Hirose & Sugiura 1996; Klaff et al. 1997 ). To analyse the influence of a predicted hairpin structure at the 5′-end of 5′-psbA on post-transcriptional regulation and regulatory protein binding, we included a construct with a mutant 5′-UTR of psbA in which 17 nucleotides of the 5′-terminus (nt –85 to nt –69) have been deleted ( Fig. 6).
A comparison between transplastomic lines 74–1-(I) (psb::psb) and 124–1-(I) (psb delta::psb) reveals only a moderate influence of the deletion on mRNA turnover, as transcript levels are only twofold lower in plant line 124–1-(I) ( Fig. 6). In contrast, an eightfold depression of protein accumulation appeared in this line, indicating a fourfold lower translation efficiency of the mutant mRNA.
These experiments demonstrate that the three UTR sequences examined modulate RNA stability approximately fivefold and translation efficiency over 100-fold. Moreover, the psbA 5′-UTR conferred light inducible translation on uidA transcripts. Deletion of the terminal sequences of the psbA 5′-UTR was found to decrease the ability of the UTR to activate translation.
Some limitations of our experimental approach have to be taken into account when interpreting the results achieved. Since the UTR elements were isolated from their original genetic context some regulatory interactions might have escaped our analysis: for example, UTRs might interfere with transcription from the endogenous promoters, although for rbcL this was shown not to be the case ( Shiina et al. 1998 ). UTRs might also interact with the coding regions of the respective genes. Several types of interaction can be envisaged. The coding region might influence secondary structure formation and thus interfere with protein binding. Furthermore, specific endonucleolytic cleavage within the coding region has been shown to initiate the degradation pathway, at least for the psbA transcript ( Klaff 1995). As the UTR-sequences fused to our uidA-constructs are homologous plastid DNA-fragments, which carry target sequences for RNA-binding proteins, possible side-effects caused by the depletion of protein or other factors, if these are limiting, may also be taken into consideration.
UTR elements modulate transcript stability
Analysis of mRNA levels from plants containing various UTR::reporter gene constructs revealed the following order of relative RNA stability: rbc::rbc > rbc::psb > psb::rbc > psb::psb ( Fig. 3, Table 2). The difference between rbc::rbc and psb::psb is about threefold. Because each fusion was placed behind the same promoter, i.e. the 16S-rRNA promoter, the observed differences are most likely only due to characteristic transcript turnover rates mediated by the UTR sequences. The consequences of these findings are: (i) both the 5′ and the 3′ UTRs of rbcL individually confer reduced transcript turnover, i.e. enhanced stability compared to a transcript carrying both psbA UTR elements; and (ii) the effect of the 5′ leader is stronger than that of the 3′ trailer. Turnover of endogenous rbcL and psbA transcripts has been analysed by several groups. Selective modulation of mRNA stability during development was shown, for example for barley ( Kim et al. 1993 ) or spinach ( Klaff & Gruissem 1991). In vivo experiments with Chlamydomonas have demonstrated the influence of rbcL 5′ ( Salvador et al. 1993 ) and 3′ UTRs ( Rott et al. 1998 ) on mRNA accumulation. These data are consistent with our findings confirming the significance of both 5′ and 3′ UTRs on transcript turnover. A recently published study on transplastomic tobacco ( Shiina et al. 1998 ) has proposed an increased mRNA stability conferred by the 5′ UTR of rbcL as the mechanism to compensate for reduced rbcL transcription in the dark. Similar to our experiments, Shiina et al. 1998 measured mRNA levels of a GUS reporter gene with the complete or partly deleted 5′ UTRs of the rbcL gene. The reporter genes were under control of the rbcL promoter and a different 3′ UTR (rps16) was used. Stabilisation of the uidA transcript in the dark was abolished when the proximal part of the 5′ UTR (positions +9 to +122) was deleted, but steady state levels in the light were not significantly touched. Although performing our experiments under different conditions, we found a threefold decrease of mRNA levels in light grown plants when the 5′ UTR of rbcL was reduced to a 26 bp fragment containing the ribosome binding site (RBS::rbc, line 73–4-(I) compared to rbc::rbc, line 120–9-(II)). Obviously, a transcript with only this short 5′ UTR is a target for more rapid degradation, most likely by 5′-3′-exonucleases ( Drager et al. 1998 ). Therefore, it seems likely that additional elements conferring transcript stability are located between positions +122 and +154. The 26 bp fragment derived from the rbcL gene is known to be sufficient to confer translation initiation. It is the same fragment which was routinely used as the 5′ UTR in the aadA selection marker for plastome transformation experiments in our laboratory ( Kofer et al. 1998a ; Kofer et al. 1998b ; Koop et al. 1996 ). We conclude that the larger portion of the 5′rbcL UTR is an important determinant of transcript stability, whereas the elements necessary for translation are located close to the translation start. This contrasts with the situation found with a deletion mutant of the 5′ UTR of psbA, where a 17 bp element located at the start of the 5′psbA UTR is necessary to confer wild-type level translation efficiency (psb delta::psb, line 124–1-(I) compared to psb::psb, line 74–1-(I), see below). Although the deletion also results in decreased mRNA accumulation (about 60% compared to the wild-type 5′-UTR of psbA), translation efficiency of the reporter gene transcript is reduced by a factor of four. The reduced stability of the mutant mRNA may be an effect of an elimination of the proposed hairpin secondary structure alone, which could impede exonuclease attack, although it seems more likely that a putative binding site of a protein (or the secondary structure involved in binding) was affected by the deletion. The protein might have the dual function to protect the RNA from degradation and stimulate translation at the same time.
In addition to 5′ UTRs, 3′ UTR elements also affect transcript levels, with plants from line 108–6-(I) (psb::rpl) showing the highest uidA transcript levels observed, threefold higher than line 74–1-(I) (psb::psb) which carries the same 5′ UTR. We were not successful in generating transplastomic lines with the other fusion constructs containing the 3′ UTR of rpl32, probably as a consequence of undesired recombination events. The endogenous tobacco rpl32 transcript has several interesting features. It is transcribed from two promoters which are both located inside the coding region of ndhF ( Vera et al. 1992 ; Vera et al. 1996 ). Its extremely long (1101 bp and 1030 bp) 5′ UTRs both originate in the divergently orientated ndhF gene. The 3′-UTR spans sprA ( Vera & Sugiura 1994), a supposedly dispensable RNA gene ( Sugita et al. 1997 ). By comparing different orientations of our GUS fusion cassettes, we found that the sprA gene contains sequence information which appears to act as a very efficient transcription terminator. This is rather unusual for plastid genes with a general tendency for read-through transcription ( Stern & Gruissem 1987). We cannot rule out the possibility that these sequences influence transcript processing and that they may also contribute to transcript stability.
Control of translation efficiency is the major post-transcriptional regulatory mechanism for psbA, but not for rbcL
In general, the turnover of tobacco plastid transcripts appears to be relatively low (P. Klaff, personal communication). We could not find any significant decrease of psbA mRNA levels 30 h after applying actinomycin D at concentrations known to be effective in blocking transcription in other systems. Therefore, transcript turnover, clearly an important means of regulation in other systems, is of limited significance in tobacco plastids and not well suited for rapid adaptation of gene expression. In contrast to the rather modest influence of UTRs on transcript levels we found dramatic differences in GUS protein accumulation. Protein levels from lines with the 5′ UTR of rbcL were 200 times lower than from equivalent constructs with the 5′ UTR of psbA, in spite of the fact that their mRNA levels were 2–3 times higher. The striking difference of regulation between psbA and rbcL expression, both encoding proteins involved in photosynthesis, demonstrates the complex adaptation of control mechanisms for each single gene. Translational control is the predominant mechanism for psbA regulation but not for rbcL. Several sequence elements within the 5′-psbA of tobacco have been described to influence translation ( Hirose & Sugiura 1996), all of them being located in the central region or close to the initiation codon. By a deletion of the terminal 17 nucleotides of the psbA 5′-UTR (psb delta::psb, line 124–1-(I)) we have shown the presence of an as yet unknown sequence element at the transcript start influencing translation efficiency. The rapid turnover of D1 in illuminated plants may have been selected for the very strong, light modulated, translational activation system that is mediated in part by the psbA UTR.
Leader/trailer interaction may be involved in the regulation of plastid translation
Relative translation efficiency was reduced to 50% in the psb::rpl versus the psb::psb (108–6-(I) versus 74–1-(I)) combination ( Table 2). Since 3′-UTRs cannot be expected to modulate translation efficiencies on their own, this indicates interaction between both RNA ends. Exchanges of the 3′ UTRs of psbA and rbcL (psb::psb versus psb::rbc, and rbc::rbc versus rbc::psb, respectively) on the other hand did not influence translation efficiency significantly. As the proposed interaction of mRNA ends most likely involves RNA binding proteins, one might conclude that the same or similar protein(s) recognise the UTRs of psbA and rbcL, both genes coding for proteins involved in photosynthesis, while at least the 3′ UTR of rpl32, a protein of the genetic machinery, interacts with different components. Indeed, several proteins have been described that bind both psbA and rbcL UTRs ( Stern et al. 1989 ).
Another indication for possible 5′/3′ UTR interactions derives from the results of our 5′-psbA deletion mutant when compared to experiments performed with in vitro translated psbA-lacZ fusion transcripts ( Hirose & Sugiura 1996). No reduction of the in vitro translation efficiency appeared when a 5′-terminal 37 bp fragment including the complete hairpin secondary structure of 5′-psbA was deleted. In contrast, several internal mutations or deletions of the 5′-psbA sequence had a strong influence on translation efficiency. Basic differences between the plastid in vitro translation system and our in vivo system might be a possible reason for the contrasting results. In an in vitro system, for example, any regulatory effect caused by the supposedly protein-mediated targeting of the transcripts to membrane-bound ribosomes ( Yohn et al. 1998a ) is absent. However, another good explanation would be that the sequence which has been deleted in our psb delta::psb construct (line 124–1-(I)) is involved in leader/trailer interactions. This putative interaction would not have been detected in the in vitro translation experiment as the psbA:lacZ transcript used by Hirose & Sugiura (1996) did not contain any plastid 3′-sequence. If this assumption is true then the reduced translation efficiency of the psb delta::psb reporter transcript (line 124–1-(I) versus line 74–1-(I)) could be at least partly due to co-operative effects between leader and trailer.
Leader/trailer interactions mediated by RNA binding proteins are known to play an important role in stabilisation and translation of nuclear transcripts ( Piron et al. 1998 ; Wells et al. 1998 ; reviewed in Gallie 1998). These authors describe that communication of the 5′-cap structure and the polyadenylated trailer results in formation of circular mRNA structures which influence translation. Initiation factors and poly(A) binding proteins (PABP) have been identified to be components of these complexes. Although translatable plastid transcripts are neither capped nor polyadenylated, polyadenylation, as in bacteria, rather seems to be a marker for degradation ( Kudla et al. 1996 ; Lisitsky et al. 1996 ) and some of the proteins which bind to plastid UTRs resemble such nuclear PABP ( Yohn et al. 1998a ). A well-characterised example for a nuclear-encoded plastid RNA binding protein of the PABP family is the Chlamydomonas protein RB47. This protein binds specifically to the 5′ UTR of psbA and is essential for psbA mRNA translation ( Yohn et al. 1998a ; Yohn et al. 1998b ). The presence of PABP, which has been described to mediate mRNA circularisation in other genetic compartments, together with our results let us assume that interaction between both transcript ends is a prerequisite for maximum but not for basic level translation. Protein dependent circularisation of mRNA ends may also be involved in the targeting of transcripts to membrane-bound ribosomes. Proteins which show phosphorylation or redox dependent RNA binding affinity like the Chlamydomonas RB47 may function as a component of the signal transduction chain. It is highly possible that RNA binding proteins interacting with cap-structures or poly (A) tails of nuclear transcripts have evolved changed binding domains to fulfil similar functions in the plastid.
Construction of uidA-fusion fragments and transformation vectors
Each UTR element was amplified by PCR using tobacco plastid DNA as template, and desired restriction sites were introduced at both ends ( Table 4). Correct amplification was confirmed by sequencing. The UTR fragments were then fused to the NcoI/EcoRI fragment of plasmid pRAJ275 (gift from Mike Bevan, Nottingham University, UK), which comprises of the uidA coding region, and the fusion products were placed behind the 95 bp plastid 16S-rDNA promoter fragment ( Koop et al. 1996 ). The cassettes were inserted into the HincII site (plastome position 115 355; Shinozaki et al. 1986 ) of vector pFaadAI ΔSN by blunt end ligation. pFaadAI ΔSN is a modified version of the previously described transformation vector pFaadAI ( Koop et al. 1996 ) from which the SmaI/NdeI fragment has been deleted and from which the HincII site in the polylinker region has been removed by subcloning of the vector insert. pFaadAI ΔSN contains an aadA resistance marker under control of the 16S-rDNA promoter, a 25 bp ribosome binding site, synthesised to match the corresponding sequence of the tobacco rbcL gene, and the 3′-end (450 bp) of the Chlamydomonas reinhardtii rbcL gene. All cloning procedures were carried out using standard methods described by Sambrook et al. (1989) .
Table 4. Plastome positions (a) of UTR sequences and primers used for cloning
Shoot cultures of Nicotiana tabacum L. cv. Petit Havanna were grown in vitro from seedlings as described by Koop et al. (1996) and protoplasts were prepared from leaves of 3-week-old cultures.
For the analysis of light regulation, sterile seedlings were grown for 2 weeks on a sugar free agar (0,8%) containing MS salts ( Murashige & Skoog 1962). Seedlings were either kept in the dark or in the light (16 h light, 0.5–1 W m–2, Osram L85 W/25 Universal-White fluorescent lamps) at 25°C.
Preparation of protoplasts, PEG-mediated transformation, culture of protoplasts in thin alginate layers ( Dovzhenko et al. 1998 ) and selection and shoot regeneration from transgenic calli on RMOP medium ( Svab et al. 1990 ) were carried out as reported previously ( Koop et al. 1996 ). After 4–6 cycles of repetitive shoot regeneration on spectinomycin and streptomycin (500 mg l–1 each) containing RMOP medium, shoots were transferred to B5 medium ( Gamborg et al. 1968 ) with 500 mg l–1 spectinomycin for root formation. Plantlets were transferred to soil, grown to maturity in the greenhouse, and seedlings of selfed plants were either germinated on soil or sterilised and germinated on sugar-free agar medium containing MS salts for analysis of gene expression.
Analysis of transgenic lines by PCR, Southern and Northern hybridisation
Total DNA was extracted from 100 to 150 mg of in vitro or greenhouse plant material using the ‘DNeasy plant DNA isolation kit’ from Qiagen (Hilden, Germany). About 5 μg of DNA were restricted, electrophoresed on 1% agarose gels and transferred to Hybond N+ (Amersham, Braunschweig, Germany) membranes using standard procedures ( Sambrook et al. 1989 ). DNA probe fragments ( Table 3) were random prime labelled with α-32P using Klenow fragment and hybridised to the membranes in 250 m m sodium phosphate, 7% SDS at 65°C. Washing of membranes was carried out with 0.5–0.2 × SSC, 0.1% SDS at the same temperature. Signals were detected using a phosphoimager (Fujifilm, model BAS 1500).
PCR analysis of transgenic plant material was undertaken with DNA isolated by the method described above using Taq-polymerase from Qiagen (Hilden, Germany).
Total RNA was isolated from 20 to 50 mg plant material (leaves or roots) using 400–1000 μl of Trizol reagent (Gibco BRL, Eggenstein, Germany), separated on 1.2% formaldehyde/agarose gels and transferred to nylon membranes. Random prime α32P-labelled DNA probes were hybridised to the membranes as described for Southern blotting. The uidA and aadA probes were used simultaneously in the same reaction. Washing of membranes was done with 0.1 × SSC, 0.1% SDS at 65°C. After hybridisation with the uidA and aadA specific probes the membranes were stripped and re-hybridised with a 16S-rDNA probe to confirm equal loading of the gel slots. Signal strength was determined using the phosphoimager. The aadA signal was used as an internal standard to compensate for variations of reporter gene copy numbers.
A fluorimetric GUS assay using 1 m m 4-methyl-umbelliferyl-β- d-glucuronid (4-MUG) from Sigma (Deisenhofen) as a substrate for the β-glucuronidase enzyme was carried out from 20 to 100 mg of plant material as described by Jefferson (1988).
Western blot analysis
Approximately 20 mg of total protein was separated on an SDS-PAGE gel containing 12% acrylamide and either stained with Coomassie blue or electroblotted to Hybond-C nitrocellulose membranes (Amersham, Braunschweig, Germany). Membranes were treated with polyclonal antisera from rabbit specific for GUS (5 Prime → 3 Prime, Inc., Bolder, USA). Mouse monoclonal anti-rabbit IgG alkaline phosphatase conjugate (Sigma, St Louis, USA) was used as a secondary antibody to detect the GUS-signal.
This work was funded by Deutsche Forschungsgemeinschaft (Ko 632–13/1,2). The expert technical assistance by Mrs Petra Winterholler and Mrs Hille Rädler is gratefully acknowledged. Dr Waltraud Kofer assisted in maintaining the transplastomic lines, contributed valuable discussions and helped with some of the experiments. We thank Mike Bevan (Norwich University, UK) for providing us with the pRAJ-275 plasmid.