mgpS, a complex regulatory locus involved in the transcriptional control of the puc and puf operons in Rhodobacter sphaeroides 2.4.1

A new method has been developed in order to select mutants showing decreased puc operon transcription in Rhodobacter sphaeroides 2.4.1. A transcriptional fusion of a promoterless fragment derived from the sacB gene, encoding the levansucrase from Bacillus subtilis, to the upstream regulatory region of the puc operon has been constructed. With appropriate levels of exogenous sucrose, survivors of a sucrose killing challenge have been isolated. Subsequent analysis revealed the presence of both cis- and trans-acting "down" mutations in relation to puc operon expression. One of the trans-acting regulatory mutations was chosen for further study. The original mutation showed less than 2% of the level of puc operon transcription compared with the wild type under aerobic conditions and an 86% reduction under dark dimethyl sulfoxide conditions. This mutation can be complemented by a 3.9-kb BamHI DNA fragment derived from a cosmid contained within a genomic cosmid bank. DNA sequence analysis of this fragment revealed the presence of a 2.8-kb open reading frame, designated mgpS, which would encode a 930-amino-acid protein. The N-terminal portion of the putative protein product presents homologies to proteins of the RNA helicase family. Disruption of the chromosomal mgpS resulted in decreased transcription of both puc and puf, while the presence of mgpS in multicopy in the wild type, 2.4.1., increased puc expression by a factor of 2 under aerobic conditions. Structural analysis of the mgpS locus revealed that expression of mgpS was likely to be complex. A smaller protein containing the 472 C-terminal amino acids of MgpS is able to act by itself as an activator of puc transcription and is expressed independently of the large open reading frame in which it is contained.

Rhodobacter sphaeroides is a purple, nonsulfur bacterium able to grow under photoheterotrophic or chemoheterotrophic conditions. The pigment-protein complexes which carry out light harvesting and primary photochemistry are the antenna complexes, designated B875 and B800-850, and the reaction center, respectively (6,7,31). They are localized to the intracytoplasmic membrane. The intracytoplasmic membrane, formed by invagination of the cytoplasmic membrane, is induced when the partial pressure of oxygen is lowered below 2.5% relative to that of air (5,20), and its abundance is inversely related to the incident light intensity (24). The puc operon of R. sphaeroides comprises the pucBA structural genes encoding the B800-850 ␤ and ␣ polypeptides as well as an additional gene(s) downstream of pucA which encodes a function essential for posttranscriptional expression of the B800-850 complex (24,27). Two of the three polypeptides of the reaction center (L and M) as well as the ␣ and ␤ polypeptides of the light-harvesting complex B875 are encoded by the puf operon (22,42).
It has been shown that transcription of puc is highly regulated by light as well as oxygen (23). A two-component regulatory system, PrrA/PrrB, sensitive to oxygen partial pressure, has been described. PrrB is the histidine kinase responsive to anaerobiosis. PrrA, the response regulator, positively regulates the transcription of the puc and puf operons (11,12). The descriptions of the PrrA/PrrB system as well as those of other trans-acting effectors of photosynthesis gene expression have been facilitated by the isolation of mutants which display increased expression of photosynthesis gene expression, e.g., App and PpsR (16,17).
Here, we describe an alternative method to the isolation of mutants (those displaying decreased transcription of the puc operon). The promoterless sacB gene of Bacillus subtilis has been transcriptionally fused to the upstream regulatory region of the puc operon. sacB encodes a levansucrase, which when expressed in the presence of sucrose at the appropriate concentration, leads to cell death. SacB hydrolyzes sucrose and synthesizes branched polymers of fructosyl residues which have been shown to be lethal for a variety of gram-negative bacteria (15). This construction has been moved into R. sphaeroides 2.4.1, and when grown on plates containing carefully titrated levels of sucrose, only those cells with reduced synthesis of the sacB product or a defective product can survive. Using this selection, we have isolated and subsequently characterized several mutants with decreased puc operon transcription. By complementation analyses, we have localized and described a new activator of puc and puf operon expression. We have designated this new locus, mgpS, for modulator of genes for photosynthesis.
(A preliminary report of these results was presented at the 95th General Meeting of the American Society for Microbiology, Washington, D.C., 21 to 25 May 1995.)

MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions. The bacterial strains and plasmids used in this study are listed in Table 1. The R. sphaeroides wild type, 2.4.1, and its derivatives were grown at 30ЊC in Sistrom minimal medium supplemented with succinate as the carbon source (8). Escherichia coli strains were grown at 37ЊC in Luria-Bertani medium (28). When appropriate, tetracycline, spectinomycin, streptomycin, and kanamycin were added at 1, 50, 50, and 25 g/ml, respectively, for R. sphaeroides and 20, 50, 50, and 25 g/ml for E. coli.
DNA manipulation and sequence analyses. Plasmid DNA was isolated according to standard protocols or manufacturers' instructions, as were restriction endonuclease and other enzymatic treatments of DNA. Deletion clones for sequencing were prepared with the exonuclease III-mung bean nuclease deletion kit from Stratagene (La Jolla, Calif.). A series of deletions in both orientations differing by approximately 300 nucleotides were made from pUI3104 according to the instructions of the manufacturer. These clones were used to determine the DNA sequence of the insert DNA of pUI3104 with the pBST3 primer (5Ј-CCCTCACTAAAGGGAACAAAAGCT-3Ј) and the EXT7-1 primer (5Ј-AAT ACGACTCACTATAGGGC-3Ј). Part of the sequence was confirmed with the primers 5Ј-GCCCTGAAGGAAGCCGATG-3Ј (Ϫ534), 5Ј-CACCTCGAAC GAATGGCTCT-3Ј (888), 5Ј-ACCGTCACTTCACCCTTGTC-3Ј (1409), and 5Ј-CGAAGGATTTGGGTTGGTTG-3Ј (2713). The numbers in parentheses represent the nucleotide coordinates to which the primers hybridize (see discussion pertaining to Fig. 3  This study a The overhangs were made blunt with the Klenow fragment of DNA polymerase I before cloning. b The overhangs were made blunt with T4 DNA polymerase before cloning. c (Ϫ), insert DNA fragment having an orientation in the opposite direction of the tet and lac promoters. d (ϩ), insert DNA fragment having an orientation in the same direction as the tet and lac promoters.
enzymes, and an ⍀ cartridge encoding resistance to spectinomycin and streptomycin (Sm r /Sp r ) was cloned in their place. PvuII fragments were excised from these constructs and cloned into pSUP202Km, which had been previously treated with ScaI. The resulting plasmids, unable to replicate into R. sphaeroides, were moved from E. coli to R. sphaeroides by standard procedures (9). Presumptive even-numbered crossovers were selected from the exconjugants on Sm/Sp-containing plates, and these were tested for kanamycin resistance. It was assumed that Sm r Sp r Km r cells would have undergone a single crossover, while Sm r Sp r Km s cells would have undergone a double crossover. For several candidates chosen at random, the double crossover event was confirmed by Southern hybridization analysis.
Pulsed-field agarose gel electrophoresis. Genomic DNA was prepared and analysed by pulsed-field agarose gel electrophoresis according to previously described protocols (38) with a transverse alternating field electrophoresis gel apparatus (Beckman Instruments, Inc., Fullerton, Calif.).
Southern DNA analysis. DNA was transferred by capillary action to nylon membranes (Micron Separations, Inc., Westboro, Mass.), and the nucleic acids were bound by UV cross-linking with a UVC1000 (Hoefer Scientific Instruments, San Francisco, Calif.). Hybridizations were performed at 68ЊC, and the final wash was at 68ЊC in 0.1ϫ SSC (1ϫ SSC is 0.15 M NaCl plus 0.015 M sodium citrate)-0.1% sodium dodecyl sulfate (SDS). Labeling of DNA probes with biotin and detection of hybridizing sequences by chemiluminescence were performed with the NEBlot Phototope kit (New England Biolabs, Inc., Beverly, Mass.).
Preparation of cell extracts and assay of ␤-galactosidase. R. sphaeroides cultures used for the measurement of ␤-galactosidase activities were grown aerobically by sparging with gas mixtures or anaerobically in the presence of 70 mM dimethyl sulfoxide (DMSO) as described previously (9). Cells were harvested at an optical density at 660 nm of approximately 0.2. Cell breakage with a French press, preparation of crude extracts, and ␤-galactosidase assays (at 30ЊC for 5 min) with O-nitrophenyl-␤-galactoside hydrolysis were performed as described previously (40). All experiments were done at least three times. Protein determinations were done with the bicinchoninic acid protein assay (Pierce, Rockford, Ill.).
Protein electrophoresis. E. coli cells grown overnight in Luria-Bertani medium were diluted fourfold with Luria-Bertani medium supplemented with 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) and grown for 4 h. The cells were harvested, washed with Tris-50 mM HCl (pH 7.0), and then resuspended in the same buffer. After a 2-min sonication (Sonifier cell disruptor), the cells were centrifuged for 2 min at top speed in a microcentrifuge, and the supernatant was then centrifuged for 1 h at 200,000 ϫ g. The supernatant (soluble fraction) was removed, and the pellet (membrane fraction) was resuspended in the same buffer. The proteins were separated on an 8% polyacrylamide gel as previously described (35). Accession number. The nucleotide sequence of the mgpS region has been assigned EMBL accession number Z50182.

Isolation of cis-and trans-acting mutations involved in puc operon expression.
To provide positive selection for the isolation of down regulatory mutations involved in the control of puc operon transcription, we transcriptionally fused a promoterless fragment of the sacB gene to the puc upstream regulatory DNA sequence. We isolated a 1,572-bp AccI-EcoRV fragment containing the sacB gene from plasmid pRL271 (4) and filled the unpaired bases with the Klenow fragment of E. coli DNA polymerase I. This fragment was then inserted into the SmaI site of pUC18. The fragment was reisolated with the XbaI and SstI restriction enzymes. The plasmid pPDK1 (26), a pRK415 derivative containing the 699-bp PstI-DraII fragment of the puc upstream regulatory region cloned between the transcription translation stop cartridge ⍀ Sm r /Sp r and the aph gene was treated with XbaI and SstI to remove the aph-containing DNA fragment. The XbaI-SstI fragment containing the promoterless sacB gene was then cloned into the subsequently linearized plasmid. Figure 1 shows a map of the resulting plasmid, pPS400.
When this plasmid was moved into R. sphaeroides, sacB was expressed under control of the puc promoter and levansucrase was synthesized. In the presence of sucrose, this enzyme forms polymers of fructosyl residues lethal for the bacteria. We determined that the minimal lethal concentration of sucrose under aerobic growth conditions on the surface of a petri dish was 0.05%. Under these conditions, only mutants with a reduced level of the sacB product or a defective sacB product can survive. Mutations can occur either within the puc upstream regulatory region or within the sacB gene itself; these are ''cis-mutations.'' Mutations can also occur within the genomic DNA of the host, and these are ''trans-mutations.'' To localize each mutational type, plasmid DNA was isolated from sucroseresistant colonies, mobilized back into the R. sphaeroides wild type, 2.4.1, from E. coli without a sucrose challenge, and the sensitivity of the exconjugants to sucrose was tested under the original conditions of isolation. When the mutation appeared not to be on the plasmid (exconjugants were sensitive to sucrose), we took the original mutant strain, cured it of its resident plasmid, reintroduced pPS400, and tested it for sucrose sensitivity. The insensitivity of the exconjugants to sucrose confirmed that the mutation was localized on the chromosomal DNA of the original isolate. Using 0.05% sucrose, we obtained both cis-and trans-mutations. At higher concentrations of sucrose (0.1 to 5%), the only surviving cells were those presenting deletions into the puc-sacB region of the plasmid. We measured ␤-galactosidase activity of the first eight trans-mutations isolated, with a puc::lacZ transcriptional fusion (pCF200 Km) placed in trans in each of these mutants as well as in the wild type, and grown under aerobic conditions. All of the isolates showed decreased puc expression ranging from 2 to 14% of the value obtained for the wild type (data not shown). We chose one of these mutants, designated M7, for further studies. This mutant presents a pale pink coloration but is still able to grow slowly under photoheterotrophic conditions. coli into M7. The initial phenotype used to score for complementation was the change from a decreased pigmentation to the pigmentation of the wild type. Several cosmids were able to restore the wild-type coloration. Among these, some are known to contain regulatory genes capable of affecting photosynthesis gene expression as characterized by puc operon activation, such as PrrA or App (11,16). However, one cosmid, pUI8256 had not been previously observed in any of our studies. pUI8256 was able to partially restore to the mutant both coloration and ␤-galactosidase activity with pCF200Km in trans. Cosmid pUI8256 was subcloned into pRK415, and all subclones were mobilized into M7 by conjugation. One of the plasmids (pUI3102) containing a 3.9-kb BamHI fragment was able to partially restore wild-type coloration to mutant M7. This fragment was cloned in pBluescript II SK (pUI3104), and deletion clones were obtained in both orientations. These deletion clones were used to sequence the entire 3.9-kb BamHI fragment. The GCG software package was used in conjunction with a R. sphaeroides codon usage table (13) to identify two open reading frames (ORFs) with sizes of 594 and 2,793 bp, respectively, apparently transcribed in the same orientation and separated by 86 bp. pUI3102 was treated with PstI and religated, removing the terminal 878 bp of the 2,793-bp ORF. The resulting plasmid (pUI3103), which contains a 3.1-kb BamHI-PstI fragment from cosmid pUI8256 including the whole 594-bp ORF but only part of the 2,793-bp ORF (Fig. 2), was unable to restore coloration to mutant M7. We therefore concluded that the 594-bp ORF itself was not responsible for restoration of the phenotype to mutant M7.
The 2,793-bp ORF could encode a putative protein with a size of 930 amino acids. Both the DNA sequence and deduced amino acid sequences are shown in Fig. 3. We designated this locus mgpS, for modulator of genes of photosynthesis. When the DNA sequence was examined for the presence of a Shine-Dalgarno sequence, no obvious match was found. A portion of the sequence downstream of the BamHI site was obtained ( Fig. 3) with the cosmid pUI8256 as a template and the oligonucleotide 5Ј-CAACCAACCCAAATCCTTCG-3Ј (2694). Two GC-rich stems at the end of mgpS followed by a series of four T's ( Fig. 3) could act as a rho-independent terminator.
Alignment of mgpS. The DNA and deduced protein sequences of mgpS were compared against sequences in the GenBank database. For the first half of the deduced protein sequence, homologies were found to proteins from the RNA helicase family, i.e., proteins involved in folding, splicing, and translation of RNA. The best matches (Fig. 4) were to the product of the yeast nuclear gene suv3, a putative ATP-dependent RNA helicase which affects mitochondrial posttranscriptional processes (37), and the SKI protein of Saccharomyces cerevisiae, which blocks the expression of viral mRNA (41). Similarities were found in a number of regions of conserved amino acid motifs, like GRAGR, a putative RNA binding domain (37) (Fig. 4). The C-proximal portion of the putative protein presents a region rich in alanine and proline residues in a repeated motif similar to those found in the H8 antigen of Neisseria gonorrhoeae (43) and the dihydrolipoamide acetyl- Disruption of mgpS with an ⍀ Sm r Sp r cartridge: effects on puc and puf transcription and spectral complex levels. An ⍀ Sm r Sp r cartridge (33) was cloned into either the EspI or the PstI site in the 3.9-kb BamHI fragment of pUI302 (Fig. 2). The BamHI fragment was then cloned into pSUP202Km, which is incapable of replicating in R. sphaeroides. Ultimately these plasmids were moved into R. sphaeroides by conjugation, and even-numbered crossover events were selected among the exconjugants as described in Materials and Methods. The structure of each recombinant clone was confirmed by Southern analysis. Two of these were chosen for further study and designated ⍀E for the insertion of the ⍀ cartridge in the EspI site and ⍀P for the insertion in the PstI site. The ⍀P strain showed a decreased coloration, while the ⍀E strain presented a coloration only slightly lighter than that of the wild type. Both of the mutants were able to grow under phototrophic conditions (3 or 10 W/m 2 ) with a rate identical to that of the wild type.
Spectral analyses of membrane extracts of these mutants as well as of those of the original M7 and 2.4.1 were performed, and the results are presented in Fig. 5. The original mutant, M7, contained less than 2% of the B800-850 complex and less than 25% of B875 complex compared with the wild type, while the ⍀P strain revealed ϳ33% of the B800-850 complex and ϳ50% of the B875 wild-type levels in cells grown under dark DMSO conditions (Fig. 5A). Under phototrophic conditions (Fig. 5B), mutant M7 contained levels of B875 identical to those of the wild type, but the quantity of the B800-850 complex was very low. Under the same conditions, ⍀P revealed ϳ25% of the B800-850 levels compared with 2.4.1 but only slightly lower levels of B875. Surprisingly, the ⍀E strain showed no significant difference from the wild type, 2.4.1, when grown under both photosynthetic and dark DMSO conditions.
The final cellular level of antenna complexes contained within the intracytoplasmic membrane is the result of a series of events including transcription, translation, insertion, and ultimately assembly of the proteins and the pigments in the membrane. To determine the effect of disruption of mgpS at the transcriptional level, we measured ␤-galactosidase activity with two plasmids, pCF200Km and pUI1830, bearing puc::lacZ and puf::lacZ transcriptional fusions, respectively, and one plasmid, pUI1851, bearing a puf::lacZ translational fusion. Under both aerobic and dark DMSO growth conditions, insertion of an ⍀ cartridge into the mgpS locus resulted in a 75% de- crease in puf operon transcription when the insertion was in the PstI site (⍀P) and a 56 to 68% decrease when insertion was in the EspI site (⍀E) ( Table 2). These effects were even more severe than those in the original mutant isolate, M7. These results suggest that mgpS may act in a positive manner in its effect on the puf operon expression under aerobic and anaerobic conditions. When ␤-galactosidase activity was measured with a translational puf::lacZ fusion (Table 2), under aerobic conditions, the results were similar to those obtained with the transcriptional fusion, showing ␤-galactosidase activities approximately 80% lower for ⍀P and 55% lower for ⍀E than those for the wild type. Under anaerobic conditions, ⍀P showed only a 52% decrease and ⍀E was virtually identical to the wild type. Importantly, under aerobic conditions, the original mutant, M7, showed only a 50% decrease in puf expression with the transcriptional fusion but a Ͼ85% decrease with the translational fusion. Thus, the upstream mutation showed no effect on puf operon expression as measured from a puf::lacZ translational fusion under anaerobic conditions, whereas with a transcriptional fusion, the effect was considerable. The nature of and reason for this difference are unknown at the present time. ␤-Galactosidase activity was also measured with a puc::lacZ transcriptional fusion (Table 3). Under anaerobic conditions, the level of puc transcription in the two disrupted strains was approximately 65% lower than that in 2.4.1, showing that under those conditions, mgpS is also a positive effector of puc transcription. Under aerobic conditions, the disruption of the gene at the PstI site resulted in a 90% decrease in puc transcription, while the disruption at the EspI site had no effect. In M7, puc expression is severely depressed in the presence of oxygen. However, the fold induction in both M7 and ⍀P when puc expression is measured anaerobically versus aerobically remains substantial in both mutant strains. On the basis of other studies, given the levels of expression aerobically, we would have predicted an approximately 10-fold induction.
Does mgpS encode one or two proteins of different sizes? The data so far present something of a paradox. On the one hand, analysis of the spectral complex data suggests that the mgpS locus is in reality two separate loci as defined by the ⍀E and ⍀P strains. Similarly, measurements of ␤-galactosidase activity with pCF200Km in each of these strains, under aerobic conditions, reinforces this contention. However, the ␤-galactosidase results determined under anaerobic conditions, as well as the DNA sequence, suggest that mgpS is a single locus. To extend this analysis in more detail, we elected to study the role of the C-terminal portion of the mgpS locus alone on photosynthesis gene expression. From the clones obtained after exonuclease III deletion of pUI3104, we obtained two DNA fragments encompassing the terminal 1,493 bp and the terminal 1,828 bp of mgpS and cloned each in both orientations into pRK415 (Fig. 2). These plasmids were moved into ⍀P, and we examined the restoration of pigmentation in the exconjugants. When plasmid pUI3109 containing the terminal 1,493-bp fragment of mgpS was cloned in the same orientation as the vector tet and lac promoters, the presence of the plasmid was able to restore wild-type pigmentation to ⍀P, while in the opposite orientation (pUI3108), there was no effect on ⍀P coloration. However, the 1,828-bp fragment was able to restore wild-type coloration in either orientation relative to the tet and lac promoters (plasmids pUI3110 and pUI3111). Spectral analysis of membranes of ⍀P containing pUI3110 (fragment in opposite orientation relative to tet and lac promoters) in trans showed that it was able to restore the levels of antenna complexes in the membranes to values similar to those of the wild type, 2.4.1 (Fig. 6). This was not the case with the original mutant, M7. Spectra obtained with M7 bearing plasmid pUI3110 or pUI3102 were identical, showing only partial restoration of the B800-850 complex.
␤-Galactosidase activities with pCF200Km, which contains a transcriptional puc::lacZ fusion in trans, were also measured in ⍀P, also containing various subfragments of mgpS in trans, in order to gain insight into the effect of the distal portion of the mgpS locus on photosynthesis gene expression ( Table 4). The presence of pUI3110 increased puc expression in ⍀P by a  The fact that the 1,828-bp fragment when cloned in either orientation in pRK415 is able to complement the insertion mutation in ⍀P, whereas the 1,493-bp fragment can only complement ⍀P when it is under control of the tet and lac promoters, suggests that the biological activity associated with mgpS is contained within the 1,493-bp fragment but can be expressed from promoter sequences present within the 1,828bp fragment. The presence of a putative Shine-Dalgarno sequence just prior to the GTG codon localized 74 bp downstream of the beginning of the 1,493-bp fragment makes it a potential candidate for a start codon (Fig. 3). These findings support the contention that mgpS is actually composed of two loci. However, DNA sequencing of both DNA strands on the exonuclease III deletion clones has been performed several times and has been reconfirmed with internal primers with the intact gene as the template (see Materials and Methods), and no evidence has been obtained for the existence of a stop codon by which the end of a putative upstream ORF could be identified.
To further analyze the structure of mgpS, a series of chromosomal deletions were constructed, as described in Materials and Methods and depicted in Fig. 2. The strains MGP 1, MGP 2, MGP 3, and MGP 4, which have had various portions of the 1,493-bp sequence deleted, revealed a decreased coloration identical to that of the ⍀P strain. Mobilization of PUI3110 into each of these strains was sufficient to restore the normal red pigmentation. Strain MGP 5, which has a deletion in the proximal portion of mgpS but still contains the distal 1,632 bp of the locus, revealed the same phenotype as ⍀E; i.e., a red coloration slightly less intense than that observed for 2.4.1.
Expression of the mgpS product in E. coli. As already described, in order to sequence the 3.9-kb BamHI fragment, we cloned this fragment into pBluescript II SK and performed an exonuclease III deletion of the insert DNA. E. coli DH5␣phe was subsequently transformed with the altered plasmids.
Among the clones obtained, one bears a plasmid (pUI3106) in which the proximal 1,070 nucleotides of the 3.9-kb BamHI fragment have been deleted (Fig. 2). The remaining fragment starts 11 nucleotides upstream of the putative ATG start codon of mgpS. In this plasmid, the translation initiation site of mgpS is in frame with the translation initiation site of the ␣ peptide of ␤-galactosidase, present in pBluescript II SK. Another derivative was used in the construction of pUI3109. Plasmid pUI3107 contains the distal 1,493-bp fragment, and in this case too, the sequence derived from mgpS is in frame with lacZ. Soluble and membrane protein extracts of the strains containing the plasmids pUI3106 and pUI3107, as well as that containing the original pBluescript II SK plasmid, grown in the presence of IPTG, were prepared and separated by SDS-polyacrylamide gel electrophoresis (PAGE) (Fig. 7). The membrane fraction derived from the strain bearing the pUI3107 plasmid (1,493-bp fragment) contained two proteins with apparent sizes of 65 and 60 kDa not present in the strain bearing pBluescript II SK. On the basis of the nucleotide sequence, presuming there is one translation product from each of the ␣ translation initiation sites and the translation initiation site identified for the 1,493-bp fragment, two proteins with sizes of 54 and 51 kDa are expected. The presence of the induced proteins in the membrane fraction rather than in the soluble fraction suggests that the protein encoded by the 1,493-bp fragment is a membrane-bound protein.
From the strain bearing pUI3106, we would expect the synthesis of a 105-kDa protein. Neither a protein of this size nor one of a smaller size was observed in either the membrane fraction or the soluble fraction of this strain. The 45-kDa protein, which seems to be present in a larger amount in the soluble fraction (Fig. 7A, lane 3) appears to be the protein which is found in the membrane fraction in extracts of cells bearing pBluescript II SK or pUI3107 (Fig. 7B, lanes 1 and 2). This has been observed several times in separate experiments. This does not seem to be the result of contamination of the soluble extract by membranes, since other proteins present in large amounts in the membrane fraction are not visible in the soluble fraction. We have no explanation for this phenomenon, although it may represent the ability of the protein to partition itself to different cellular compartments.
Mapping  has been shown that this bacterium possesses two chromosomes and five endogenous plasmids (38,39). All of the known photosynthesis structural genes have been mapped to a region of the large chromosome designated the photosynthesis gene cluster. The localization of mgpS was accomplished by restriction analysis of ⍀P with rare cutting enzymes defined for R. sphaeroides DNA (e.g., AseI, DraI, and SnaBI). The ⍀ Sm r Sp r cartridge introduced new AseI and DraI sites into the genome. After digestion with AseI, it was observed that the original 1,105-kb AseI A fragment of the large chromosome (39) disappeared and two new bands with sizes of approximately 500 and 600 kb appeared. When a double digestion was performed with SnaBI and Ase I, the 370-kb SnaBI fragment gave rise to two new fragments with sizes of 60 and 310 kb. This means that mgpS is localized 60 kb from one of the ends of the 370-kb SnaBI fragment. Because the digestion of ⍀P genomic DNA with DraI gave an electrophoresis pattern similar to the one from the wild type (no significant difference in size for the 660-kb fragment), we concluded that mgpS maps 2,110 kb clockwise from the puf operon in chromosome I of R. sphaeroides 2.4.1, far from the photosynthesis gene cluster and other known regulatory elements.

DISCUSSION
The expression of sacB encoding the levansucrase of B. subtilis is lethal for gram-negative bacteria in the presence of sucrose. A plasmid containing a fusion of the promoterless sacB gene to the upstream regulatory region of the puc operon was constructed and transferred into R. sphaeroides 2.4.1. The exconjugants able to grow on plates containing a previously defined concentration of sucrose showed mutations or deletions in the puc-sacB DNA fragment on the plasmid (cis-mutations) or mutations in the genomic DNA of the host (transmutations). One of the trans-mutants, M7, contained very small amounts of both types of antennae but was still able to grow photosynthetically, although slowly. The residual level of puc transcription, measured with a puc::lacZ transcriptional fusion, was less than 2% of that observed for the wild type in cells grown in the presence of oxygen and only 14% of that of the wild type in dark DMSO cultures (Table 3). A 3.9-kb BamHI fragment derived from cosmid pUI8256 (10) was able to restore partial pigmentation and full puc expression to mutant M7. A 2,793-bp ORF designated mgpS was identified within this fragment.
When considered together, the data indicate that the original M7 mutant may contain more than a single mutation. On the one hand, mutations in mgpS and M7 have similar, although not identical, effects on puf and puc transcription, and in both, puc expression is more severely compromised. Likewise, either mgpS or subclones derived therefrom in trans can have a substantial effect on puc operon transcription. However, none of these DNA fragments in trans can fully restore to M7 spectral complex levels identical to those of the wild type, although mutants derived from mutations in mgpS can be fully complemented. M7 is also unusual in that of the numerous photosynthesis mutants we have isolated, this is the first in which we have observed such a dramatic difference between the spectrum obtained under dark DMSO conditions and that at 10 W/m 2 . This may also speak to the presence of more than a single mutation in M7. However it is possible, because of the unusual structure of the mgpS locus, that a complex mutational event entirely within that locus is responsible for the unusual phenotype of M7. mgpS has been interrupted by insertion of an ⍀ cartridge at two different sites, namely EspI and PstI, localized in the N-proximal and C-proximal parts of the predicted protein product, respectively (Fig. 2). Both of the resulting strains, ⍀E and ⍀P, showed decreased puf transcription under aerobic and anaerobic conditions and decreased puc transcription under anaerobic conditions (Tables 2 and 3). These results demonstrate that the mgpS locus is probably an activator of both the puc and puf operons. However, under aerobic conditions, ⍀P showed less than 10% of the wild-type level of puc transcription, while ⍀E showed no difference from the wild type, 2.4.1 (Table 3). Moreover, the amount of antennae present in the membrane of ⍀E was identical to that in the wild type, while ⍀P showed two-thirds less B800-850 complex and ϳ75% of the B875 complex under dark DMSO conditions (Fig. 5A). One salient feature of these spectral data is that the B800-850 complex is more severely restricted than the B875 complex in both the original mutant and ⍀P. At the transcriptional level, ⍀P is more highly compromised than M7 for puf operon expression, while the reverse is true for puc operon expression, although both puf expression and puc expression are substantially reduced in ⍀P.
To further assess the hypothesis that mgpS comprises two loci, we decided to further study the corresponding C-proximal portion of mgpS alone. The results convincingly demonstrated that the C-terminal region of mgpS encodes a protein which has a significant effect upon puc and puf operon expression and that this portion of the locus is driven by its own promoter. The GTG codon localized 74 bp downstream of the beginning of the 1,493-bp fragment is a good candidate for a start codon, since it is preceded by a good Shine-Dalgarno sequence. Even in the wild type, 2.4.1, this region in multicopy gave rise to a 2.6-fold increase in puc transcription. The existence of two loci seems to be confirmed by the fact that strains which had various portions of the 1,493-bp fragment deleted (strains MGP 1, MGP 2, and MGP 3 [ Fig. 2]) always resulted in a decreased coloration identical to that of the ⍀P strain, which was restored to normal in the presence of pUI3110, while strain MGP5, which had a deletion in the N-proximal part but still contained the terminal 1,632 bp of mgpS, had a nearly red coloration only slightly less intense than that observed for 2.4.1.
However, DNA sequencing analyses, repeated several times and confirmed with internal primers, did not reveal the existence of a stop codon which could terminate the first ORF. It has been shown that ribosomes can ''slip'' during translation (2) and possibly generate a stop codon out of frame (32) and thereby terminate the translation process. It is also possible that if the N portion of the mRNA encodes an RNA helicase, it can affect translation of the mRNA corresponding to this complex locus.
To gain further insight into the coding capacity of mgpS, we cloned the fragment containing the last 1,493 bp of mgpS and a fragment starting 11 bp upstream of the putative ATG start codon for mgpS into pBluescript II SK and expressed these constructions in E. coli. In both cases, the DNA sequences were in frame with the sequence of the ␣ peptide. With the plasmid containing the 1,493-bp fragment, we obtained the synthesis of two membrane-bound proteins with sizes of 65 and 60 kDa. These sizes are higher than those expected for proteins expressed from the ␣ peptide translation initiation site and the translation initiation site identified for the 1,493-bp fragment (54 and 51 kDa, respectively). However, it has been shown that the presence of the highly rich alanine-proline repeat regions is responsible for anomalous electrophoretic mobility in other proteins like lipoylacetyltransferase or the bovine ␤-crystallin B1 chains (3,18). When the plasmid containing the entire sequence of mgpS was expressed in E. coli, no protein products were observed. The band with a size of approximately 45 kDa, VOL. 178, 1996 mgpS: A COMPLEX REGULATORY LOCUS 43 which seems to be present in a larger amount in the soluble fraction (Fig. 7, lane 3A) appears to be a protein which is present in the membrane fraction in the control and which can partition to two cellular compartments. If mgpS is composed of two separate loci and the distal locus is expressed from its own promoter sequence contained within the 1,828-bp fragment, we would have to conclude that both of these loci encode different proteins, each able to serve as puc and puf effectors, since disruption of the proximal mgpS region at the EspI site also resulted in altered puc and puf transcription (Tables 2 and 3). A more likely hypothesis is that mgpS is a single locus encoding two proteins with sizes of 930 and 472 amino acids derived from two different mRNAs, with the sequence of one (distal) contained entirely within the sequence of the other and with both terminated by the same UGA stop codon. A similar but not identical phenomenon has been observed with transposon Tn5, in which two proteins, one a transposase and the other an inhibitor of transposition, are synthesized from the same ORF (same promoter but two different translation initiation sites). The two proteins share the same amino acid sequence, except that the first one has an additional 40 residues at its N terminus (19,34). If this is the case, both protein products might be expected to activate transcription of the puc and puf operons, since we saw a decrease in transcription with both ⍀P and ⍀E disrupted strains. It is also possible that the RNA helicase activity can affect translation of one or both of the mRNA species, such that the protein products are altered and therefore modulate any effect it might have on transcription.
The homology found with RNA helicase proteins which are involved in modifications of mRNA suggests that MgpS could also be involved in posttranscriptional control of photosynthesis gene expression or, in fact, control of other nonphotosynthesis genes. To further examine this hypothesis, we compared the ␤-galactosidase activities obtained with a puf::lacZ transcriptional fusion and a puf::lacZ translational fusion present in the disrupted strains ⍀P and ⍀E ( Table 2). The results suggest that the disruption of mgpS had a positive effect on posttranscriptional events involved in puf operon expression. MgpS could therefore act as an activator of photosynthesis genes at the transcriptional level but could also, under some conditions, negatively regulate their expression at the posttranscriptional level. Whatever the case, disruption of the RNA helicase domain of the mgpS locus had no discernable effect on cell growth under the conditions employed here. Additional studies may reveal a unique role for this activity. The finding and initial characterization of the mgpS locus raise some interesting possibilities regarding photosynthesis gene expression. Both the behavior of the original mutant and that of the derived mutants reveal the lack of an absolute linkage between transcriptional and posttranscriptional expression of the photosynthetic apparatus. These observations are the subject of ongoing studies.