Translation Initiation of Cyanobacterial rbcS mRNAs Requires the 38-kDa Ribosomal Protein S1 but Not the Shine-Dalgarno Sequence

Little is known about the biochemical mechanism of translation in cyanobacteria though substantial studies have been made on photosynthesis, nitrogen fixation, circadian rhythm, and genome structure. To analyze the mechanism of cyanobacterial translation, we have developed an in vitro translation system from Synechococcus cells using a psbAI-lacZ fusion mRNA as a model template. This in vitro system supports accurate translation from the authentic initiation site of a variety of Synechococcus mRNAs. In Synechococcus cells, rbcL and rbcS encoding the large and small subunits, respectively, of ribulose-1,5-bisphosphate carboxylase/oxygenase are co-transcribed as a dicistronic mRNA, and the downstream rbcS mRNA possesses two possible initiation codons separated by three nucleotides. Using this in vitro system and mutated mRNAs, we demonstrated that translation starts exclusively from the upstream AUG codon. Although there are Shine-Dalgarno-like sequences in positions similar to those of the functional Shine-Dalgarno elements in Escherichia coli, mutation analysis indicated that these sequences are not required for translation. Assays with deletions within the 5′-untranslated region showed that a pyrimidine-rich sequence in the –46 to –15 region is necessary for efficient translation. Synechococcus cells contain two ribosomal protein S1 homologues of 38 and 33 kDa in size. UV cross-linking and immunoprecipitation experiments suggested that the 38-kDa S1 is involved in efficient translation via associating with the pyrimidine-rich sequence. The present in vitro translation system will be a powerful tool to analyze the basic mechanism of translation in cyanobacteria.

Cyanobacteria are prokaryotes that perform plant-type photosynthesis, and chloroplasts in plants are thought to be derived from an ancestral photosynthetic prokaryote related to cya-nobacteria (1)(2)(3)(4). Cyanobacterial genes are generally transcribed by an Escherichia coli-type RNA polymerase (reviewed in Ref. 5), and the regulation of transcription has been extensively analyzed using several representative cyanobacterial species (e.g. Refs. 6 -8). However, little is known about their posttranscriptional events including RNA processing and translation. The molecular analysis of post-transcriptional processes in cyanobacteria is important to understand the evolutional relationship between cyanobacteria and chloroplasts, as it is known that chloroplast gene expression is regulated mainly at the post-transcriptional level (reviewed in Ref. 9). Several RNA-binding proteins possessing eukaryotic-type RNA-binding motifs (CS-RBD or RRM) (10,11) and their involvement in various physiological phenomena (12,13) have been reported, suggesting that the gene expression of cyanobacteria is also regulated at the RNA level.
Prokaryotic mRNAs generally have the Shine-Dalgarno (SD) 3 sequence in the 5Ј-untranslated region (5Ј-UTR), which participates in translation initiation (14,15). The distance between a functional SD sequence and a start codon is critical with an optimal spacing of 7-9 nucleotides (nt) in E. coli (16,17). Most mRNAs of the cyanobacterium Synechocystis sp. PCC 6803 also possess SD-like sequences in 5Ј-UTRs but their positions vary (18), and only 26% of the genes harbor SD-like sequences at the optimal position (19). In tobacco chloroplasts, about 40% of the protein-coding genes contain no SD-like sequences within 20 nt upstream from start codons and the remaining 60% have SD-like motifs but not always at a conserved position (20). Using an in vitro translation system from isolated chloroplasts, it was suggested that translation initiation of chloroplast psbA, rps2 and atpB mRNAs does not depend on the SD-like sequence and requires trans-acting factors that interact with their 5Ј-UTRs (21)(22)(23). Cyanobacterial rRNAs and ribosomal proteins were assigned mainly based on the DNA sequence similarity with those of E. coli and of chloroplasts (20,24). Recent proteomic analyses revealed that chloroplast ribosomes include several additional proteins not found in E. coli ribosomes (25)(26)(27)(28)(29). On the other hand, few biochemical studies have been made for cyanobacterial ribosomes. Ribosomal proteins from Anabaena variabilis have been analyzed by two-dimensional gel electrophoresis, and 50 proteins were detected (30,31). Ribosomal protein S1 is known to interact with the 5Ј-UTR and regulate translation initiation in E. coli (32)(33)(34). Two proteins related to E. coli S1 were isolated from Synechococcus sp. PCC6301 (35,36). E. coli S1 is the largest ribosomal protein (61 kDa) and contains four repeats of the so-called S1 motif (37), whereas the cyanobacterial proteins are 33 and 38 kDa in size and possess three S1-like motifs. The presence of two S1 homologues has also been predicted from sequenced genomes in several other cyanobacteria, suggesting that these cyanobacteria have unique mechanisms of translation initiation. However, little is known about the mechanism of translation in cyanobacteria due probably to the lack of tools to analyze translation reactions. In this study, we have developed an in vitro translation system from Synechococcus cells and applied this system to investigate the translation mechanism of Synechococcus rbcS mRNA encoding the small subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase.
Growth of Cyanobacterial Cells-Synechococcus 7942 or 6301 cells were grown in the BG-11 liquid medium (42) con-taining 5 mM NaHCO 3 at 30°C for 88 h under continuous light (50 E/m 2 /s) with a sterile air of 4 liters/min in a 20-liter polycarbonate vessel. After dark adaptation for 12 h in the above condition except for light, the cells were harvested by continuous centrifugation using a rotor No. 15N (TOMY) at 4,000 ϫ g at 4°C and were stored at Ϫ70°C.
Preparation of S30 Fractions-The cell pellet (ϳ4 g) defrosted at room temperature was suspended by pipetting in 20 ml of EX buffer (30 mM HEPES-KOH, pH 7.7, 10 mM magnesium acetate, 30 mM potassium acetate, 2 mM dithiothreitol and was aliquoted to ϳ10 tubes of a Mini Bead-Beater TM (Biospec Products) with 400 mg/tube quartz sand (Katayama Chemical). After short vortex, the suspensions were frozen in liquid nitrogen and then immediately put into a 37°C water bath for ϳ3 min to allow partial melting. The suspension was agitated by the Mini Bead-Beater TM for 30 s at the maximal speed. The supernatant (cell extract) was obtained by centrifugation at 25,000 ϫ g for 1 min at 2°C, and the precipitate was resuspended in 200 -300 l each of EX buffer followed by centrifugation. The extraction was repeated two to three times to raise the yield of cell extracts. The combined extract (ϳ10 ml) was centrifuged twice at 30,000 ϫ g for 30 min at 2°C. The resulting blue supernatant (ϳ8 ml) was incubated at 30°C for 15 min with 40 M each of 20 non-radioactive amino acids, followed by treatment with micrococcal nuclease (0.26 unit/ml, GE Healthcare) at 30°C for 15 min in the presence of 1 mM CaCl 2 . The nuclease reaction was stopped by adding 300 mM EGTA to the final concentration of 3 mM. The treated extract was dialyzed twice against 4 liters of EX buffer at 4°C for 8 h using Spectra/Pore CE MWCO: 3,500 membrane (Spectrum). The dialyzed fraction (S30, ϳ15 mg/ml protein) was aliquoted (34 l each) and stored at Ϫ70°C. S30 was diluted with EX buffer to adjust its protein concentration to 3 mg/ml before use.
In Vitro Translation Reaction-The standard reaction was carried out at 37°C for 40  . After reaction, proteins were precipitated with 75 l of cold 100% acetone and were rinsed once with cold 50% acetone. The precipitated protein was suspended in the Laemmli sample buffer (Bio-Rad) containing 8 M urea and separated by 0.1% SDS/12.5% PAGE in Tris-glycine or Tris-Tricine running buffer after heat treatment at 95°C for 3 min. Separated products were visualized and quantified by a Bio-imaging Analyzer (BAS2000, Fuji Photo Film Co.).
UV Cross-linking and Immunoprecipitation-A 32 P-labeled rbcS mRNA portion (a 120-nt 5Ј-UTR with a 67-nt 5Ј-coding region) was synthesized from the plasmid for rbcS-lacZ fusion mRNA (linearized with PvuI) using the in vitro transcription kit (Stratagene) with [␣-32 P]UTP (Ͼ110 TBq/mmol, Amersham Biosciences) and purified by 4% PAGE containing 8 M urea. The mRNA of 2.5 pmol was incubated for 25 min at 30°C in the 25-l translation mixture as above except 35 S-labeled amino acids were replaced with the corresponding non-radioactive ones, and then the mixture was diluted with 75 l of EX buffer. UV cross-linking of the mRNA and proteins was performed twice on ice using a UV illuminator (254 nm, 900 mJ/cm 2 , FS1500, Funakoshi), and then the mixture was treated with 0.01 unit of RNase A for 30 min at 37°C. Cross-linked proteins were precipitated and rinsed with 80% acetone. 32 P-Labeled proteins were separated by PAGE as above and visualized by the Bioimaging Analyzer. For immunoprecipitation, protein A-Sepharose beads (25 l wet volume each) were mixed with 250 l each of antibodies against Synechococcus 38-kDa S1, EF-Tu, Rbp1, and Rbp2 for 1 h at room temperature and then centrifuged to remove excess antibodies. Anti-S1 and anti-EF-Tu were gifts of M. Sugita, and anti-Rbp1 and anti-Rbp2 were as described (13). The beads were washed seven times with 1 ml each of phosphate-buffered saline, pH 7.4, added to 100 l of the reaction mixture containing UV cross-linked proteins, and the suspension was incubated for 3 h at 4°C. After collecting beads by centrifugation and washing three times with 1 ml each of phosphate-buffered saline, the beads were suspended in 25 l of the Laemmli Sample buffer and subjected to PAGE as above. 32 P-Labeled proteins were visualized by the Bio-imaging Analyzer.

Development of an in Vitro Translation
System-During the course of development and then optimization of an in vitro translation system from the cyanobacterial cells, a psbAI-lacZ fusion mRNA was used as a model template, and translation in vitro was monitored by its 35 S-labeled translation products. The psbAI mRNA of Synechococcus elongates PCC7942 (a strain closely related to Synechococcus sp. PCC6301) encodes the D1 protein (40 kDa) inserted into the photosystem II complex and is known as one of the actively translated transcripts in photosynthetic organisms (39). The fusion to lacZ mRNA makes the translation product soluble, normalizes efficiencies of translation elongation, and eliminates possible effects of the psbAI 3Ј-UTR on translation initiation. The hybrid mRNA template was synthesized with T3 RNA polymerase from a linearized construct including the 5Ј-psbAI and the 3Ј-lacZ gene portions as shown in Fig. 1A.
Synechococcus 6301 or 7942 cells were grown to an early log phase and subjected to dark adaptation for 12 h to reduce endogenous mRNA levels. Cell extracts were preincubated with unlabeled amino acids and then treated with micrococcal nuclease to digest endogenous transcripts. These steps were critical to reduce nonspecific translation in vitro. Extensive dialysis was also required to remove EGTA added to inactivate micrococcal nuclease and to standardize reaction constituents. Using this cell extract (S30) and the model template, a major polypeptide of ϳ18 kDa was detected on PAGE with a relatively low background (Fig. 1B). This size is ϳ3 kDa smaller than the calculated mass of 20.9 kDa. This is probably due to the presence of a long hydrophobic region (21st to 60th amino acids) in the translated products.
In vitro reaction conditions were then optimized using the psbAI-lacZ mRNA (data not shown). Optimal buffer conditions were 10 mM magnesium acetate, 40 mM potassium chlo-ride, and 20 mM ammonium chloride in 60 mM HEPES-KOH, pH 7.7. ATP was essential for translation and its optimal concentration was 2 mM. GTP (the optimum at 0.5 mM) and an ATP-generating system enhanced translation. An RNase inhibitor, leupepcin (thiol proteinase inhibitor), and E. coli tRNAs slightly affected translation, and hence, we did not add them. Translation was hardly detected with the mRNA of 1 pmol or less, but it increased linearly up to 10 pmol and then slowly up to 50 pmol (Fig. 1C). Using 10 pmol of mRNA template, translation increased linearly with S30 amounts (up to 8 mg of protein/ml) (Fig. 1C). Translation was observed after a time lag of ϳ10 min and reached a plateau at 60 min (Fig. 1C). We adopted S30 of 3 mg/ml protein, 10 pmol of mRNA template, and 40-min incubation as our standard in vitro reaction. As expected, translation was completely inhibited with 4 g/ml chloramphenicol (data not shown).
To examine whether the optimized in vitro system translates faithfully cyanobacterial mRNAs, we replaced the authentic initiation codon AUG in the psbAI-lacZ mRNA with UAC. The mutated mRNA lost completely its translation activity (Fig. 1B,  lane 2). No difference was observed when the authentic UGA stop codon was replaced with UAA (Fig. 1B, lane 3). An exchange of the UGA for UGG produced a long polypeptide of 23 kDa as the 3Ј-UTR had no additional in-frame stop codon (Fig. 1B, lane 4). This is consistent with that of the predicted product from the start codon to the 3Ј-end of mRNA (see Fig.  1A), that is, the calculated mass of 25.9 kDa subtracted by 3 kDa due to the hydrophobicity as mentioned above. We also carried out translation reactions in vitro using three additional mRNAs from Synechococcus 6301 rbcS, rpl3, and rpl4 ( Fig. 2A; Refs. 40 and 41). All these mRNAs fused with the lacZ mRNA synthesized translation products of practically expected sizes at varied efficiencies (Fig. 2, B and C). Thus, our in vitro system supports correct translation from various Synechococcus mRNAs.
Translation Initiation Site and SD-like Sequences of rbcS mRNA-The rbcL and rbcS genes encoding the large and small subunits, respectively, of ribulose-1,5-bisphosphate carboxylase/oxygenase constitute an operon in Synechococcus 6301 ( Fig.  3A; Ref. 40), and the rbcS mRNA has two possible initiation codons, both AUGs, separated by three nt, named as "1st AUG" and "2nd AUG" in Fig. 3A. To define the actual translation initiation site, we constructed an rbcS-lacZ fusion mRNA including the entire spacer region (93 nt) between rbcL and rbcS coding regions and the 5Ј-coding sequence (299 nt) of rbcS (see WT (wild-type) in Fig. 4A). An exchange of the 1st AUG for ACG arrested translation completely, while that of the 2nd AUG for ACG still supported translation though reduced activity by half (Fig. 3, A and C). The N-terminal amino acid sequence deduced from the 1st AUG matches that determined from the Synechococcus RbcS (SMKTLPKE---; Ref. 43). This result indicates that the 1st AUG, but not the 2nd AUG, is essential for the rbcS mRNA translation.
In the 5Ј-UTR, there are SD-like sequences (AGGA, AGG, or GGA; Fig. 3A) and additional short sequences (UGA and AUC, Fig. 3B) complementary to the 3Ј-end region of Synechococcus 6301 16 S rRNA (44). To examine the effect of these sequences on rbcS mRNA translation, these sequences (S1, S2, and S3) were mutagenized. As shown in Fig. 3D, none of the mutated mRNAs abolished translation. Mutation of S1 (AGGAU to ACCAU) or S3 (AUC to UAC) enhanced translation and that of S2 (UGA to ACA) retained 80% of the wild-type efficiency. It should be noted that mutation of S1, an SD-like sequence (GGA) at a proper position, resulted in 200 -250% increase in translation. Thus, the interaction between an SD-like sequence and the 3Ј-16 S rRNA is unnecessary for rbcS mRNA translation.
Sequences Necessary for rbcS mRNA Translation-As described above, translation of the Synechococcus rbcS mRNA requires no SD-like elements, and therefore an additional mechanism should operate to recognize the genuine start site. To identify sequences necessary for the translation initiation of rbcS mRNAs, we first constructed a series of deleted mRNAs based on the rbcS-lacZ fusion mRNA (Fig. 4A) and assayed their translation. Deletion from the 5Ј-end (Ϫ120) to Ϫ85 did not significantly affect translation; however, surprisingly a 4-fold increase was observed by deletion up to Ϫ47 (Fig. 4B). This result suggests that the original mRNA construct does not possess a maximal activity and that the Ϫ84 to Ϫ47 region includes a negative element to regulate rbcS mRNA translation.
As a long pyrimidine-rich sequence lies in the Ϫ46 to Ϫ15 region (Fig. 4A), mRNAs with further deletion toward the initiation site were used to examine the role of this sequence. The translation efficiency of these mRNAs was dramatically reduced compared with ⌬Ϫ66 and ⌬Ϫ47 mRNAs, and deletion to Ϫ15 (removal of the entire pyrimidine-rich sequence) virtually abolished translation (Fig. 4B, ⌬Ϫ15). We then examined mRNAs with internal deletions. As shown in Fig. 4C, deletion of the whole pyrimidine-rich sequence (⌬Ϫ46/Ϫ15) caused a drastic reduction (ϳ10% of the wild-type), and translation decreased to ϳ20% by deleting half of the pyrimidine-rich sequence (⌬Ϫ27/Ϫ15). mRNA quantity was monitored during incubation using WT, ⌬Ϫ66, and ⌬Ϫ15 mRNAs. Despite the presence or absence of the pyrimidine-rich region, these mRNAs decayed in a similar manner (ϳ70% for 40 min), suggesting that the observed difference is not due to the difference in mRNA stability (supplemental Fig. S1). These results confirmed that the pyrimidine-rich region is an important element for efficient translation of rbcS mRNAs. However, this region was not essential for translation because its complete removal still retained translation (ϳ10% of the wild-type). Translation  DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 was also arrested by deleting the immediately upstream sequence from the start site (⌬Ϫ14/Ϫ1), suggesting that an additional sequence(s) is required for translation. We next examined whether the enhancer activity requires a specific sequence or a high pyrimidine content in the pyrimidine-rich region. The Ϫ27 to Ϫ15 region of rbcS mRNAs was replaced with random pyrimidine-rich sequences or random purinerich sequences. Replaced constructs were cloned and 14 mRNAs derived from randomly selected clones were analyzed. As listed in Table 1, all the mRNAs with different purine-rich sequences showed lower translation efficiency than the original mRNA (42 to 5% of the wild-type). On the other hand, pyrimidine-rich constructs exhibited generally the opposite effect; 9 out of the 14 mRNAs showed higher efficiency than the wildtype mRNA, and the remaining five mRNAs showed a moderate reduction (Y02, Y15A, Y13A, 09A, and Y10A). This result indicates that the pyrimidine-rich sequence in the 5Ј-UTR involves an important role for translation and that its enhancing capacity requires high pyrimidine contents but not unique sequences.

Translation Initiation of Cyanobacterial rbcS mRNAs
The 38-kDa Ribosomal Protein S1 Interacts with the Pyrimidine-rich Sequence-Our in vitro experiments indicated that translation of the rbcS mRNA starts exclusively from the 1st AUG, but not from the 2nd AUG, and requires the pyrimidine-rich region (Ϫ46 to Ϫ15) but not SD-like sequences. It is therefore expected that a trans-acting factor(s) is involved in recognizing the correct initiation site. To examine the existence of a protein factor(s) functionally interacting with the pyrimidine-rich region of the rbcS mRNA, we carried out competition assays with 2.5 pmol of the mRNA template ( 1 ⁄ 4 of the standard condition) to detect competition sharply. As mentioned above, not only the Ϫ46 to Ϫ15 region but also a variety of pyrimidine-rich sequences enhanced translation, and hence we used a mixture of 30-mer random deoxypyrimidine sequences (30Y) or of 30-mer random deoxypurine sequences (30R) as competitors. Translation was reduced by adding 2.5 pmol of 30Y (an equivalent amount of the mRNA template) and arrested completely by 25 pmol or more (10-fold excess or more) of 30Y (Fig. 5). On the other hand, 2.5 pmol of 30R did not affect translation, and 25 pmol of 30R decreased translation, and no translation was detected by 100-fold excess 30R. These results suggest that a depletable factor(s) interacts with the pyrimidine-rich region of rbcS mRNAs for efficient translation.
UV cross-linking experiments were then performed to detect a possible protein factor(s) interacting with the rbcS mRNA. The 32 P-labeled rbcS mRNA fragments consisting of the 5Ј-UTR of 90 nt and its 5Ј-coding region of 66 nt (no lacZ mRNA) was incubated in the standard translation mixture. After RNase A treatment of UV cross-linked samples, a clear protein band of 38 kDa (p38) was detected on PAGE (Fig. 6A). The translation of rbcS mRNAs is strictly dependent on Mg 2ϩ (the optimum at 10 mM) as described, and no translation was detected at 2 mM Mg 2ϩ (data not shown). The p38 band was extremely faint by UV cross-linking with 2 mM Mg 2ϩ , indicating that p38 binding occurs practically in translating conditions. UV cross-linking assays were also carried out using external deletion mutants of the rbcS mRNA (Fig. 6B). Deletion Arrows indicate mutated nucleotides with site names (S1, S2, and S3). B, the 3Ј-end sequence in Synechococcus 6301 16 S rRNA and its complementary sequence (gray characters). Underlined sequences are present in the 5Ј-UTR (shown in A). C, identification of the bona fide translation start codon from two AUG codons. Translation activities were calculated from band intensities and shown above as a bar graph (the wild-type mRNA (WT) as 100%, n ϭ 3). Mutated nucleotides are indicated in A. D, effect of potential ribosome-binding sites on translation. Translation activities were shown as described for C.
(⌬Ϫ47) up to the pyrimidine-rich region (Ϫ46 to Ϫ15) retained p38 interaction at a level comparable with the wild-type mRNA. Further deletion into the pyrimidine-rich region (⌬Ϫ28 and ⌬Ϫ21) reduced cross-linked signals ϳ60% and ϳ10% of the wild-type, respectively. A residual signal was still detected by deleting the entire pyrimidine-rich region (⌬Ϫ15) and the whole 5Ј-UTR (⌬Ϫ1), probably due to nonspecific binding of p38. A similar result was obtained using internal deletion mRNAs (⌬Ϫ24/Ϫ15 and ⌬Ϫ46/Ϫ15 in Fig.  6B). These results indicate that p38 preferentially binds to the pyrimidine-rich region of the rbcS mRNA.
As can be seen in Figs. 4 and 6, the level of translation correlates roughly with that of p38 cross-linking using a set of deleted mRNAs.
No additional protein bands could be detected by the UV cross-linking assay, suggesting that p38 is a major trans-acting factor for efficient translation of rbcS mRNAs. Previously, we reported that the ribosomal protein S1 from Synechococcus 6301 is 38 kDa in size (35). To examine whether p38 is S1, we carried out immunoprecipitation using the antibody against the cyanobacterial S1. The p38 protein crosslinked with the 32 P-labeled rbcS mRNA was digested with RNase A, and then it was applied to immunoprecipitation followed by PAGE. As shown in Fig. 6C, p38 cross-reacted exclusively with the anti-S1, indicating that p38 is the 38-kDa ribosomal protein S1.

DISCUSSION
Our in vitro system supported faithful translation initiation from authentic initiation sites and accurate translation termination at stop codons (see Fig. 1). This system would be useful to define genuine protein-coding regions when multi- The pyrimidine-rich region is shaded. B, translation of 5Ј-deleted mRNAs. Translation activities were as described for Fig. 3C. C, translation of internal deletion mutants. Translation activities were as described for B.  DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 ple possible start codons exist. A stop codon-less psbAI-lacZ mRNA (no in-frame stop codon till the 3Ј-end of mRNA) produced a larger translation product but its quantity was less than the wild-type product. This may be due to stalling ribosomes at the 3Ј-end of mRNAs followed by trans-translation, as a tmRNA homologue has been identified as 10Sa RNA (45). We showed that the pyrimidine-rich sequence (Ϫ45 to Ϫ15) in the 5Ј-UTR is required for efficient translation of the mRNA and that the SD-like sequence is not necessary for its translation initiation. Furthermore, mutation of the SD-like sequence (S1, from AGGAU to ACCAU, Ϫ14 to Ϫ10) significantly enhanced translation. A possible explanation is that the mutation generated a longer pyrimidine-rich sequence (the Ϫ46 to Ϫ6 region; pyrimidine:purine ϭ 32:8) than that of the wild-type mRNA (the Ϫ46 to Ϫ15 region; pyrimidine:purine ϭ 24:6). The region between the start codon and the pyrimidine-rich sequence (Ϫ14 to Ϫ1) is also important for translation probably not in a sequence-specific manner but as a distance because base changes did not affect translation but its deletion inhibited translation. Our UV cross-linking and immunoprecipitation experiments showed that the 38-kDa ribosomal protein S1 interacts with the pyrimidene-rich sequence under translation conditions. Pyrimidine-rich sequences are found in the 5Ј-UTR of other mRNAs from Synechococcus 7942 and 6301. For example, psbAI (39), nrtA (46), zwf (47), and sigA (Kyoto Encyclopedia of Genes and Genomes data base) contain over 75% pyrimidine residues in the Ϫ10 to Ϫ30 region. The 38 kDa-S1 protein is possibly involved in efficient translation of these mRNAs because E. coli S1 has no strict sequence specificity (33).

Translation Initiation of Cyanobacterial rbcS mRNAs
Previously, we reported that the S1 protein is present not only in ribosomal fractions but also in soluble fractions in Synechococcus 6301 (36). Based on the model proposed for efficient translation of rpsA mRNAs encoding S1 in E. coli (48), free S1 may also be inhibitory for translation in Synechococcus. E. coli S1 possesses a long N-terminal domain that is responsible for protein-protein interactions, whereas Synechococcus S1 has no such domain and consists mainly of three S1 motifs (four in E. coli S1). These two S1 proteins exhibit different nucleic acid binding specificities in vitro; E. coli S1 binds poly(C), poly(U), and poly(A), while Synechococcus S1 binds poly(G), poly(U), and poly(A) (36). The S1-binding target in 5Ј-UTRs is AU-rich in E. coli (49) but CU-rich in Synechococcus.
Synechococcus 6301 cells possess an additional S1 homologue of 33 kDa in size. It is not associated with ribosomes, and its nucleic acid binding specificity is distinct from that of the 38-kDa S1 protein: high affinity for both single-and doublestranded DNAs as well as poly(G) and poly(A) (36). This 33-kDa protein may participate in translation initiation of a different type of cyanobacterial mRNAs. As a summary, our in vitro translation system will provide powerful approach to analyze translation reactions in cyanobacteria and to disclose novel translation mechanisms unique to cyanobacteria. Comparison of the translation mechanism elucidated by using our in vitro systems from cyanobacteria and tobacco chloroplasts will provide basic data to understand an interesting evolutional relationship.