Analysis of the Translational Initiation Region on the Euglena gracilis Chloroplast Ribulose-bisphosphate Carboxylase/Oxygenase (rbcL) Messenger RNA*

The chloroplast mR.NAs from Eugienagracilis fall into two classes. One group of mRNh from this organelle contains a Shine-Dalgarno sequence 5‘ to the start codon, while the other group of mRNAs does not have a conserved sequence signal in the 5’-untranslated region. “0 investigate the start signals for E. gracilie chloroplast mRNAs that do not carry a Shine-Dalgarno sequence, 30 S initiation complex formation has been studied using a series of transcripts carrying the wild-type translational start site of ribulose-bisphosphate carboxylase/ oxygenase (r&cL) or mutated derivatives of this site. Mutation of the start codon of the rbcL gene indicates that the chloroplast 30 S subunit is recognizing only the correct AUG codon. The analysis of the messages from a series of deletion mutants shows that a minimum of A20 residues 5’ to the AUG codon is required for activity. Maximal activity requires the full 55-base leader se- quence. Surprisingly, a transcript carrying the inverse complement of 48 bases in the leadex is A6Wo as active as the wild-type message in promoting initiation complex formation. Introduction of a Shine-Dalgarno sequence in the 5‘-leader increases the activity of the mRNA only Al.42-fold. The presence of an oligodeoxynucleotide

of the small subunit rRNA (16 S rRNA) during initiation complex formation. This rRNA-mRNA interaction facilitates the correct selection of the start codon by the 30 S subunit (6)(7)(8). In contrast, in the eukaryotic cytoplasmic system, 40 S ribosomal subunits carrying the initiator tRNA interact with the cap structure on the mRNA and, in the presence of appropriate initiation factors, scan the mRNA until it reaches the first AUG codon from the 5'-end, where initiation generally occurs (9). In higher plants, statistical analysis of putative start sites on chloroplast mRNAs indicates that some of these messages contain Shine-Dalgarno sequences 5' to the start codon ( 10). Other mRNAs do not appear to have an equivalent sequence. The signals used to designate a certain AUG codon as a start codon in this latter class of mRNAs remain to be clarified.
The analysis of putative translational start sites based on DNA sequence information from the chloroplasts of E. gracilis has been complicated by the presence of Group I11 introns, which are difficult to predict based on sequence information alone (11,12). However, it appears that less than half of the chloroplast mRNAs in this organism have traditional Shine-Dalgarno sequences (13). Previously, we proposed that the translational machinery in the chloroplas~ of E. gracilis has two potential mechanisms available for the recognition of translational start sites (13). In one class of sites, a sequence equivalent to the Shine-Dalgano sequence is present and can hydrogen-bond to the CUCCC sequence found at the 3'-end of the small subunit rRNA in E. gracilis chloroplasts (14). In the second class of mRNAs, no Shine-Dalgarno sequence is present.
In these mRNAs, the A/U content of the region just 5' to the AUG start codon is very high (generally 240%). This portion of the mRNA may therefore have very little stable secondary structure. In this class of mRNAs, the AUG start codon may be selected because it resides in a region of the mRNA that has little secondary structure, making it readily available for interaction with the 30 S ribosomal subunit (13).
The investigation into the nucleotide sequence signals in E. gracilis chloroplast mRNAs has been facilitated by the development of an in vitro system that measures the formation of initiation complexes (13). In this paper, we have examined the role of primary sequence information in initiation complex formation with the chloroplast mRNA encoding the large subunit of ribulose-bisphosphate carboxylase/oxygenase. Although many chloroplast mRNAs are polycistronic, the message encoding the large subunit is monocistronic and has a 55-base 5"untranslated leader. This mRNA does not contain a Shine-Dalgarno sequence and thus provides a model system in which to delineate the translational initiation signals recognized by ribosomes in one class of chloroplast mRNAs. In the accompanying paper (441, the effects of secondary structural 7494 This is an Open Access article under the CC BY license.

~n i t i~t i o n
Signals on ~h l o~o p l a s t as 7495  TTTTATTATT A A A A T E T C ACCTCAAACTG  1: TTTTATTATT AAAATNNNTC ACCTCAAACTG  2 : TAATACGACT C A C T A~T T TTATTTTATT ATTAAAATAT GTCAC  Oligo 3: CGACTCACTA TAGAATATAA TTTTTATTTT Table It were prepared in the Departments of Pathology and of Microbiology and Immunology and at the Lineberger Cancer Center of the University of North Carolina at Chapel Hill.
Plasmid Preparation and Dunscription-The preparation of the plasmid pRbcN has been described previously (13). A derivative of this construct (pRbcN X2) has had two XbaI sites introduced upstream of the start site of the RbcN open reading frame (15). A derivative of pRbcN X2 was prepared in which a large portion of the 5"untranslated leader sequence was present as the inverse complement. Digestion of pRbcN X2 with XbaI released a 48-base pair fragment from the 5'untranslated leader region. The 48-base pair fragment and the remaining large fragment of the plasmid were separated by electrophoresis on a 2% agarose gel. The large fragment was extracted from the gel using the GeneCleanTM method according to the man~acturer's i n s t~c t i o~. The 48-base pair fragment was recovered from the ge1 by electrophoresis onto Whatman DE81-cellulose paper (16) and was ligated with the large fragment, which had been treated with calf intestine alkaline phosphatase (17). The ligation mixture was used to transform E. coli JMlOl (18,19). A derivative (pRbcN X F ) in which the 48-base pair fragment had been inserted in the reverse orientation was identified by DNA sequencing.
In vitro mutagenesis of the AUG start codon of the rbcL gene was carried out using oligonucleotide 1 ( Table I). During synthesis, this oligonucleotide was doped at the positions designated N with 33% of the wild-type nucleotide and 67% of an equimolar mixtiFe of the other 3 nucleotides. Deletions of portions of the 5'-untranslated leader region of the rbcL gene were carried out using oligonucleotides 2-5 (Table It. The insertion of a Shine-Dalgarno sequence (GGGAG) into the 5'-untranslated leader region and a partial deletion derivative of this construct were prepared using o l i~n u c l~t i d e s 6 and 7. Site-directed mutagenesis was carried out according to the method of K d e l et al. (20,21) and Prior to transcription, plasmids were prepared as described (17), further purified by centrifugation in CsCl gradients, and linearized by digestion with HindIII. In vitro transcription was carried out as described previously (13).
Znitiation Complex Formation-Reaction mixtures using a high salt wash of E. coli ribosomes as the source of initiation factors were prepared and analyzed as described previously (131, except that the concentration of NH,C1 was reduced to 40 m. Reaction mixtures (50 pl) using the partially purified initiation factors were prepared as described (13)

5).
To test the ability of an oligonucleotide containing a Shine-Dalgarno sequence to inhibit initiation complex formation with various mRNAs, the indicated amount of the oligodeo~ucleotide dGGGAGW, which is complementary to the sequence at the 3'-end of the chlo~plast 16 S rRNAfrom E. gracilis, was hybridized to 0.06A260 unit o f c~o~p l a s t 30 S subunits by incubation at 4 "C for 16 h in 5 pf of buffer containing 25 spermidine, and 20% glycerol. Following this incubation, the 30 S subunits were tested for the ability to participate in initiation complex formation as described (13) in reaction mixtures containing 5 pmol of the indicated S A .
Measurement of Off Rate of Danslational Initiation Complexes-Reaction mixtures (50 pl) containing 5 pmol of the indicated message, 0.06 A280 unit of chloroplast 30 S subunits, and 6 pmol of [36SlfMet-tRNA, were prepared as described previously (13). After incubation for 10 min at 37 "C to allow initiation complex formation, 100 pmol of nonradioactive fMet-tRN& in 10 m~ potassium succinate, pH 6.0, were added, giving a final volume of 70 pl. Incubation was continued at 37 "C, and at the indicated times, the reaction was terminated by the addition of 1 ml of cold stop buffer (50 nm Tris-HCi, pH 7.8,40 m NH4Cl, and 10 nm MgCl,). The amount of [~S l~e t -t R N A remaining associated with 30 S complexes was determined as described previously (13).

Selection of Correct Ranslational Start Codon in RbcN
Construct-In our previous work (131, a clone was prepared carrying the 5'-untranslated leader and exon 1 of the E. gracilis rbcL gene fused to a portion of the neomycin resistance gene under the control of the T7 transcriptional promoter. The mRNA transcribed from this vector (mRbcN) contains the entire 55-base untranslated leader of the rbcL gene ( Fig. 1) and is identical to the in vivo RNA except for the presence of a 5'terminal G residue that facilitates transcription by the T7 polymerase (26). This mRNA does not contain a Shine-Dalgarno sequence upstream of the start codon, but does have a very AICJ-rich fi'-untranslated leader region (Fig. 1).
The RbcN mRNA directs the formation of chloroplast translational initiation complexes (Net-tRNA.mRNA.30 S) in the presence of either E. coli or chloroplast initiation factors (15). Transcripts having deletions of the entire leader through the known AUG start codon are inactive in initiation complex formation, suggesting that the correct start site of this mRNA is being used for initiation complex formation (13). To verify this conclusion, the AUG start codon was mutated using saturation mutagenesis. A number of mutations in the putative AUG start d o n were obtained, and two mutants were seleeted for additional study. The first of these mutants had three changes leading to a UAU codon in place of the original AUG start codon

5"
G[ _" " " " " " (mRbcN UAU; Fig. 1). This message was tested for the ability to participate in initiation complex formation. As indicated in Fig. 2, mRbcN UAU is essentially inactive in this assay. This result strongly confirms the previous conclusion that initiation complex formation with chloroplast 30 S subunits and mRbcN is occurring at the correct translational start site (13). In addition, this result argues that the other 18 AUG codons present in mRbcN are unable to direct initiation complex formation with chloroplast 30 S subunits.
The second mutagenic construct tested also contained a 3-base change in the AUG codon, converting it to a GAC codon (mRbcN GAC; Fig. 1). This mutation created a new AUG codon just 2 nucleotides upstream of the original AUG codon. When mRbcN GAC was tested for the ability to promote initiation complex formation, it was -60% as efficient as the original mRNA (Fig. 2). This result again argues that the AUG codon in the 5"region of mRbcN is the codon selected for initiation complex formation by chloroplast 30 S ribosomal subunits. The somewhat reduced efficiency of mRbcN GAC compared with the wild-type mRNA probably reflects the different context immediately surrounding the start codon ( A A B A versus UAXU). There are a number of cases in which the residues immediately surrounding the AUG codon affect the efficiency of initiation complex formation in bacterial systems (27, 28) and in the eukaryotic cytoplasm (29). Further mutagenesis of the nucleotides adjacent to the start codon will be required to examine the subtle effects of start codon context on the efficiency of initiation complex formation.
Effect of Length of Untranslated Leader on Initiation Complex Formation-In previous work (15), the 55-base 5"untranslated leader of the RbcN construct was modified by mutagenesis to introduce two XbaI sites (mRbcN X 2 ; Fig. 1). The mRNA transcribed from this construct is indistinguishable from the wild-type mRNA in assays of initiation complex formation (15). Deletion of the nucleotides between these XbaI sites in the leader results in the synthesis of a mRNA (mRbcN XA9; Fig. 1) that contains only 10 nucleotides 5' to the normal start codon (9 nucleotides of the natural 5'-untranslated leader and a 5'-terminal G residue). This mRNA has little or no activity in initiation complex formation (Fig. 3) (15). There are two reasonable explanations for this observation. First, the 5'-untranslated leader may have essential sequence information that directs the chloroplast 30 S subunit to the AUG start codon. Second, there may be a minimal length in the 5"region that is essential for ribosomal subunit binding. Since the 30 S subunit would be expected to cover -20 nucleotides upstream of the start codon, deletion of all but 10 nucleotides 5' to the initiation codon could result in a mRNA that cannot interact with the small subunit.
To examine the importance of the length and sequence of the 5"untranslated leader of mRbcN, several additional deletion mutants were prepared. The first deletion mutation (mRbcN A20; Fig. 1) has the sequence from positions -55 to -21 deleted, leaving 20 nucleotides of the 5'-untranslated leader from the wild-type sequence and the additional G residue at the 5'-end. The mRbcN A20 mRNA has significantly reduced activity in initiation complex formation (Fig. 3). It promotes Met-tRNA binding to 30 S subunits only about one-third as well as the normal mRNA. The second mutation (mRbcN A28; Fig. 1) has residues -55 to -29 deleted. This mRNA contains a 5"untranslated leader with 28 residues from the wild-type sequence and the additional 5'-terminal G residue. This mRNA also shows a reduction in the ability to promote initiation complex formation (Fig. 3). Indeed, it is no more efficient than the mRNA with the 5"untranslated leader that has 20 bases present. Another deletion mutation lacks residues -55 to -41 (mRbcN A40; Fig. l). This mRNA has a 5'-leader containing 40 of the 55 residues of the wild-type sequence. However, it is only 30% as efficient as the full-length mRNA in promoting Met-tRNA binding to 30 S subunits (Fig. 3). These observations suggest that the AUG start codon can be selected in the mRNA with only 20 residues of the 5'4eader present. However, maximal activity in initiation complex formation requires the full 5'-untranslated leader. There are two possible interpretations for the observations described above. First, it is possible that the inherent length of the 5'-untranslated leader in the mRNA (55 base pairs) is an important determinant for providing maximal efficiency of initiation. A second interpretation is that the sequences between positions -40 and -55 are important as a translational enhancer. To distinguish between these two possibilities, two internal deletion constructs were prepared. In the first, nucleotides between positions -1 and -25 were deleted, providing a 31-nucleotide B'-untranslated leader carrying the residues originally between positions -40 and -55. The mRNA prepared from this construct (mRbcN A1,25; Fig. 1) has only -30% of the ability of the wild-type mRNA to participate in initiation complex formation (Fig. 4). The second mutant carries an internal deletion from positions -21 to -40 in the 5"untranslated leader (mRbcN A21,40; Fig. 1). The mRNA prepared from this deletion mutant contains a 35-residue 5'-untranslated leader from the rbcN message. This mRNAalso shows a significant reduction in its ability to participate in initiation complex formation (mR-bcN A21,40; Fig. 4). Thus, internal deletions in the 5'-untranslated leader cause a reduction in activity similar to that caused by the 5'-end deletions (Fig. 3). These results argue that the length of the 5"untranslated leader is important for maximal activity of mRbcN in initiation.
Role of Primary Sequence of B'-Untranslated Leader in Initiation-The results summarized above indicate that the full length of the 55-base leader of mRbcN is essential for maximal activity in initiation complex formation. However, these results do not clearly indicate whether there is essential sequence information in the 5'-untranslated leader that strongly influences the initiation process. Rather, the effects of the deletions summarized above appear to arise from changes in the length of the mRNA itself. To address the question of whether there are elements of essential sequence information in the leader, cassette mutagenesis was carried out to prepare a new mutant using the mRbcN X2 construct (Fig. 1) as the starting plasmid. The presence of the XbuI sites (positions -57 to -52 and -9 to -4) allowed the excision of a 48-base pair segment from the plasmid and its insertion in the opposite orientation (mRbcN XF; Fig. 1). Transcription of this plasmid results in the synthesis of a mRNA with a 5"leader region that is the inverse complement of the mRbcN X2 sequence between positions -9 and -56. The 5'-untranslated leader from this construct has base pair changes at 24 out of 58 positions scattered throughout the leader region (underlined in Fig. 1) and has about the same percent AIU found in mRbcN X2. Interestingly, mRbcN XF is -60% as active as mRbcN when tested for the ability to participate in initiation complex formation (Fig.  5). These results suggest that the primary sequence of the 5'-untranslated leader region is not the major determinant in specifying a particular AUG codon as a translational start site.
Effect of Shine-Dalgarno Sequence in 5'-Untranslated Leader-As noted above, chloroplast mRNAs in E. gracilis fall into two major classes. The first class of mRNAs appears to have a Shine-Dalgarno sequence comparable to that observed in prokaryotic systems. The second class of mRNAs, including mRbcN, does not have a recognizable Shine-Dalgarno sequence. It was of interest to determine the effect of a Shine-Dalgarno sequence on the ability of mRbcN to promote initiation complex formation. To examine this question, oligonucleotide-dir~ted mutagenesis was used to replace a short stretch of nucleotides between positions -7 and -11 with the strong E. gracilis chloroplast Shine-Dalgarno sequence (GGGAG). The transcript designated mRbcN SD (Fig. 1) was then prepared and tested for its ability to direct translation initiation. As indicated in Fig. 6, the presence of the Shine-Dalgarno sequence 5' to the AUG start codon enhances the ability of the mRNA to promote initiation complex formation -1.4-fold. This prokaryote-like initiation signal is thus only modestly helpful to the mRNA in enhancing initiation. This result is in stark contrast to the critical role played by the Shine-Dalgarno sequence in prokaryotes, in which it generally enhances initiation by 1~100-fold.
To determine whether the presence of a Shine-Dalgarno sequence could offset the effect of a reduction in the length of the 5'-untranslated leader, a deletion of this construct was prepared that would direct the synthesis of a mRNA containing 20 nucleotides 5' to the start codon (mRbcN SDA20; Fig. 1). The Shine-Dalgarno sequence is still positioned between positions -7 and -11 in this transcript. As indicated in Fig. 6, the deletion derivative is -3-fold less effective in initiation complex formation compared with the full-length mRNA carrying a Shine-Dalgarno sequence. This reduction in activity is comparable to that observed with the normal RbcN mRNA and its deletion derivatives (Figs. 3 and 6). These results indicate that a long 5'-untranslated leader region is probably important for maximal effkiency in initiation whether or not a chloroplast mRNA contains a Shine-Dalgarno sequence.
Effect of ~~~~e o~y n u c l e o t i d~ Competition on Initiation Complex Formation-In prokaryotes, the Shine-Dalgarno sequence in the mRNA hydrogen-bonds to the polypyrimidine sequence located near the 3'-end of the 16 S rRNA. This interaction places certain constraints on the path of the mRNA along the small ribosomal subunit in the initiation complex. It was of interest to ask whether a mRNA that does not contain a Shine-Dalgarno sequence binds to the 30 S subunit in the same way as a mRNA having a Shine-Dalgarno sequence. In an effort to examine this question, the effect of an oligodeoxynucleotide that can bind to the polypyrimidine sequence at the 3'-end of the small subunit rRNA on the ability of mRbcN to form initiation complexes was tested. This oligodeoxynucleotide contains a W A G sequence that can hydrogen-bond to the 3'-end of the E. gracilis chloroplast 16 S rRNA. Chloroplast 30 S ribosomal subunits were incubated overnight with increasing concentrations of the oligodeoxynucleotide to allow it to bind to the small subunit. The 30 S subunit-oligodeoxynucleotide complexes were then tested for their ability to form initiation complexes with mRbcN. Small subunits incubated under these same conditions overnight in the absence of the oligodeoxynucleotide retained >80% of their activity. As indicated in Fig.  7, a 100-fold excess of the Shine-Dalgarno oligodeoxynucleotide over 30 S subunits has almost no effect on the formation of initiation complexes with mRbcN. Preincubation of 30 S subunits with a 500-fold excess of the oligodeoxynucleotide results in a small reduction (<25%) in the ability of the 30 S subunits to participate in initiation complex formation with mRbcN. This observation suggests that this chloroplast mRNA can bind to the small subunit without any close interaction with the 3'-end of the 16 S rRNA. This conclusion is in agreement with the idea that there is no Watson-Crick hydrogen bonding between the rRNA and the 5"untranslated leader region of chloroplast mRNAs that do not contain a Shine-Dalgarno sequence. The effect of the oligodeoxynucleotide containing the Shine-Dalgarno sequence on the ability of mRbcN SD to form initiation complexes was also tested. This mRNA has a Shine-Dalgarno sequence equivalent to that present in the competing oligodeoxynucleotide. As indicated in Fig. 7, mRbcN SD is somewhat more efficient in initiation complex formation compared with the starting mRbcN message. In the presence of a 100-fold excess of the competing oligodeoxynucleotide, the ability of the Shine-Dalgarno sequence-containing mRNA is reduced essentially to the level observed with mRbcN. At higher concentrations of the competing oligodeoxynucleotide, mRbcN and mRbcN SD behave identically. These results suggest that the small enhancement in the efficiency of initiation complex formation with the Shine-Dalgarno sequence-containing mRNA probably arises from a direct hydrogen bonding between the small subunit rRNA and the Shine-Dalgarno sequence in the mRNA. The oligodeoxynucleotide appears to be able to compete with this rRNA-mRNA interaction, leaving mRbcN SD with an ability to form initiation complexes equivalent to that observed with the wild-type leader. These data again suggest that while mRbcN SD may take advantage of the Shine-Dalgarno sequence to increase the efficiency of initiation slightly, the leader of this mRNA has an intrinsic ability to direct initiation in the absence of any classical rRNA-mRNA hydrogen bonding. It should be noted that Canonaco et al. (30) have observed that an oligodeoxynucleotide carrying a prokaryotic Shine-Dalgarno sequence inhibits the binding of an artificial message carrying a Shine-Dalgarno sequence to E. coli 30 S subunits more strongly than it inhibits the interaction of a corresponding mRNA lacking a Shine-Dalgarno sequence. Thus, the results obtained here are quite compatible with those observed in the bacterial system.
Stability of Initiation Complexes Formed with Various mRNAs-The data provided above suggest that the 5"untranslated leader of mRbcN contains essential information that directs the 30 S ribosomal subunit to the AUG start codon. The data also argue that a section of the leader >20 nucleotides is important for maximal initiation and that the addition of a Shine-Dalgarno sequence has a small enhancing effect on the ability of this mRNA to participate in initiation. There are several possible stages during initiation complex formation in which variations in the 5'-leader sequence might influence the initiation process. The formation of the initiation complex occurs by a rather complicated sequence of events, many of which may be important for efficient initiation (8). Changes in the 5"untranslated leader might also lead to initiation complexes that are inherently less stable, resulting in a more rapid rate of dissociation. In an effort to determine whether the differences in the amount of initiation complex observed with several of the mRNAs used here resulted from changes in the rate of formation of the complexes, efforts were made to examine the time courses for initiation complex formation with a number of the mRNAs. However, the amount of initiation complex observed in every case had reached a maximal value within 2 min after assembly of the reaction mixtures and remained constant for -20 min. Since the 2-min time point was the earliest one that could be measured accurately, potential changes in the rate of initiation complex formation could not be measured directly with the various mRNAs available. The data obtained indicate that the meas~ements reported here represent equilibrium values. Differences in the amount of initiation complexes observed with the various mRNAs probably are related to the equilibrium constants governing the formation of the various complexes.
Although it was not practical to obtain a direct measure of the forward rate constants for initiation complex formation, it was possible to gain some insight into the stability of the various complexes as measured by the off rate of the bound m e t -tRNA. To observe the off rate, 30 S initiation complexes were assembled with the appropriate mRNA and labeled met-tRNA. The reaction mixtures were then diluted in the presence of unlabeled met-tRNA, and the amount of labeled complex remaining as a function of time was measured. The unlabeled m e t -t~A was added in a 17-fold molar excess to mask the reassociation reaction. The dissociation of the labeled initiation complex formed with mRbcN can clearly be observed following the addition of the unlabeled competitor (Fig. 8). With this mRNA, nearly half of the original complex is dissociated within 5 min, and <lo% remains after 10 min of incubation. The stability of the complex formed with the mRbcN A20 deletion mutant is essentially the same as that formed with the natural mRNA. This observation suggests that it is the formation of the initiation complex rather than its dissociation that controls the amount of complex present at equilibrium. A similar analysis was carried out with mRbcN SD (Fig. 8). Again, the off rate observed is essentially identical to that found with mRbcN, indicating that the somewhat enhanced initiation complex formation observed with this m~A p r o b a b l y arises from changes in the rate of formation of the complex rather than from a slower rate of dissociation of the complex once assembled.
Role of Chloroplast Initiation Factors in Initiation Complex Formation-The experiments described above were performed using chloroplast 30 S ribosomal subunits and initiation factors from E. coli. In general, it is believed that the recognition of the translational start site is carried out by the small subunit of the ribosome rather than by the initiation factors in prokaryotetype systems (7, 8). To make sure that the observations described above were not significantly affected by the use of the bacterial initiation factors, several of the constructs were tested for activity in initiation complex formation using 1@"2,hl and 1F-3chj rather than the crude E. coli initiation factor preparation. As indicated in Table 11, initiation complex formation could be observed using the chloroplast initiation factors and mRbcN. When mRbcN SD was tested in this assay, an -2-fold enhancement in activity was observed compared to that obtained with mRbcN (Table 11). This result is essentially the same as that observed with the purified E. coli factors. Finally, the deletion derivatives mRbcN A20 and mRbcN SDA20 were tested and showed about the same -fold reduction in activity as observed in the previous experiments ( Fig. 6 and Table 11).
These observations indicate that the results obtained with the bacterial factors are a direct reflection of the ability of the chloroplast translational system to use the mRNA derivatives tested here.

DISCUSSION
In E. coli, several features of the mRNA appear to play an important role in translational initiation. The first and most striking feature is the Shine-Dalgarno sequence. Mutations in this region have a drastic effect on the efficiency of initiation (31-33). In addition, statistical analysis indicates that nucleotides -20 to +13 are not random, suggesting that this whole region has a modulating effect on the efficiency of initiation (32,34,35). Finally, it is clear that secondary structural features can play a critical role in initiation. Start sites with little or no secondary structure are generally much more efficient than those in which the Shine-Dalgarno sequence or AUG codon is base-paired (36,37).
In this work, a number of mutants were constructed to delineate the region of the 5'-untranslated leader in the E. grucilis chloroplast rbcL mRNA that is important for specifying the AUG start codon. This mRNA does not have a Shine-Dalgarno sequence and hence cannot use the major bacterial determinant for specifying the start codon. The start site of the rbcL mRNA is probably specified by several features, all of which contribute to the selection process. One component that appears to be important is a 5"untranslated leader of at least 20 nucleotides. This observation is quite reasonable considering data showing that the 30 S ribosomal subunit probably covers -20 residues 5' to the start codon (38). These 20 residues do not have to represent immediate upstream sequences since an internal deletion mutant shows almost the same activity as 5'end deletion mutants. The results presented here indicate that maximal initiation complex formation is dependent on the presence of the full-length 55-base leader sequence. This chloroplast mRNA may thus have two areas through which it contacts the 30 S subunit: one near the AUG codon and the other involving more distal regions of the mRNA. This observation is compatible with those obtained in prokaryotes, where it is believed that regions of the mRNA beyond the ribosome-binding site contribute to the interaction of the mRNA with the 30 S subunit (31, 36,39). For example, Borisova et al. (40) observed that 53 nucleotides upstream of the AUG codon are required for initiation complex formation with the replicase cistron of MS2. It has been suggested (31) that during initiation, the 30 S subunit interacts with a larger region of the mRNA than is protected in the final complex formed. The ribosome-binding site is thus only a part of the true initiation region (41). Details are not yet available on all of the precise contacts between the prokaryotic 30 S subunit and mRNA. However, it

Initiation Signals on
has been suggested that the mRNA interacts with the 30 S subunit via a U-shaped trough or channel (42). It is likely that this channel is formed from parts of both the rRNA and the ribosomal proteins. In E. coli, at least two proteins (SI and S21) along with several stretches of the rRNA have been implicated in facilitating the binding of the mRNA to the 30 S subunit. Recent evidence (42) indicates that the mRNA does not fill the entire channel, but, depending on its specific sequence and structure, makes contacts with groups lining the channel. The analysis of the translational start site of the rbcL mRNA presented here argues that the leader region does not contain a specific sequence that is absolutely required for determining the start site of translation. For the E. gracilis chloroplast 30 S subunit, the major interactions between the mRNA and the subunit may involve the ribose phosphate backbone of the mRNA rather than the bases themselves. Thus, the overall length and structure (or lack of structure) of this region may be the essential features present. An analysis of the effects of structural elements in the mRNA on initiation is presented in the following paper (44).
Finally, it should be noted that the insertion of a strong Shine-Dalgarno sequence 5' to the start codon leads to only a modest improvement in the ability of the RbcN mRNA to participate in initiation complex formation. This result is quite different from E. coli messages, which show a much stronger requirement for the presence of a Shine-Dalgarno sequence in the mRNA (7). It is also compatible with the presence of two classes of mRNAs in E. gracilis chloroplasts, some having Shine-Dalgarno sequences and others lacking this sequence. The observation that the competing Shine-Dalgarno oligodeoxynucleotide overcame the slight advantage in initiation observed with the mRbcN SD argues that when a Shine-Dalgarno sequence is present, it probably results in a slightly altered placement of the mRNA in the RNA-binding trough on the 30 S subunit. This idea is compatible with the proposal that the RNA-binding channel is rather wide. Certain messenger RNAs may sit in this trough in a slightly different way than do other mRNAs (43).