Characterization of a CO-responsive transcriptional activator from Rhodospirillum rubrum.

In Rhodospirillum rubrum, CO induces the expression of at least two transcripts that encode an enzyme system for CO oxidation. This regulon is positively regulated by CooA, which is a member of the cAMP receptor protein family of transcriptional regulators. The transcriptional start site of one of the transcripts (cooFSCTJ) has been identified by primer extension. The ability of CooA to bind to this promoter in vitro was characterized with DNase I footprinting experiments using extracts of a CooA-overproducing strain. CooA- and CO-dependent protection was observed for a region with 2-fold symmetry (5'-TGTCA-N6-CGACA) that is highly similar to the consensus core motifs recognized by cAMP receptor protein/FNR family. In vivo analysis in a heterologous background indicates that CooA is sufficient for CO-dependent expression, implicating it as the likely CO sensor.

Exposure of the purple nonsulfur bacterium Rhodospirillum rubrum to CO stimulates the expression of the coo regulon, which consists of at least two transcriptional units. Among the products of this regulon are a carbon monoxide dehydrogenase (CooS), an Fe-S protein (CooF), and a hydrogenase (CooH), where the two former proteins have been purified and characterized (1)(2)(3)(4)(5)(6). This CO-oxidizing system functions under anaerobic conditions to oxidize CO to CO 2 , allowing growth on CO as sole energy source (7). The cooFSCTJ region has been cloned, sequenced, and mutationally characterized, verifying the requirement for the encoded products for oxidation of CO (7,8). 1 The mutational analysis has also indicated that cooFSCTJ is organized in a single transcriptional unit. 1 cooH lies at the 3Ј terminus of the other known CO-regulated transcript, but this transcript has not yet been fully sequenced at the 5Ј end. cooH is located 5Ј of cooF and is separated from it by 450 nucleotides of noncoding DNA (Fig. 1B).
Our previous mutational studies revealed that CooA, 2 which is apparently encoded on it own transcript on the 3Ј side of cooFSCTJ, is essential for the expression of the coo regulon of R. rubrum in response to CO (9). The sequence of CooA predicts that it is a member of the CRP/FNR family of transcriptional regulators (9), with a putative DNA-binding domain that is highly similar to that found in CRP and FNR. Modeling the sequence of CooA on the known CRP crystal structure (10) predicts the presence of four Cys and one His residues adjacent to the region known to bind cAMP in CRP (9). These residues suggest the possibility that CooA contains a metal center at this position, which might be expected if CooA binds CO. A particularly interesting question in this area is how the binding of a molecule as small as CO might induce a similar conformation change in CooA as that caused by cAMP binding in CRP.
To test the model of CooA as a CO-binding transcriptional activator, we have sought evidence for CO-and CooA-dependent DNA binding. The results described herein support the above model, and the assay of DNA-binding activity of CooA will aid in the purification of CooA for more direct analysis.

EXPERIMENTAL PROCEDURES
Growth of Bacterial Strains-R. rubrum strains were grown photoheterotrophically in SMN (supplemented malate-ammonium) medium supplemented with 10 M NiCl 2 in stoppered serum vials with an argon head space (8).
For RNA isolation, cultures to be CO-induced were grown photoheterotrophically to an optical density of 1 at 680 nm, whereupon CO was added to a final concentration of 30%; uninduced culture received no additions. The cultures were agitated under illumination (8) for 6 h.
The CooA-overexpressing strain (UR459) was grown under nif-derepression conditions in malate-glutamate medium (11) to an OD 680 of 2.0, and the expression of nifH promoter was monitored by nitrogenase activity (12).
RNA Extraction-Total cellular RNA was isolated by repeated phenol extraction as described previously (13) with the following modifications. After addition of 1.5 ml of lysis solution to 6 ml of culture, the mixture was boiled for 90 s and extracted with phenol 3 times. Purified RNA was dissolved in 100 l of 10 mM Tris-Cl (pH 8.0).
Primer Extension of mRNA-Two synthetic oligonucleotide primers (Operon Technologies Inc.) were used: Primer 1 (5Ј-GATCGGGATTG-GCGTAGATG) is complementary to the coding strand in region ϩ27 to ϩ46 (numbering relative to the translational start codon of cooF) and Primer 2 (5Ј-GAATTAACGCCACCCCTGTTCG) is complementary to the coding strand in region Ϫ154 to Ϫ133. Primers were labeled at their 5Ј-ends with [␥-32 P]ATP using T4 polynucleotide kinase (14). The primer extension was performed as described (15) except that 25 g of total RNA was mixed with 0.25 pmol of ␥-32 P-labeled primer. The extended products were analyzed by electrophoresis on a 6% polyacrylamide-urea gel.
Construction of Plasmids for Heterologous Expression-Plasmid pCO6R (9) was processed through a number of steps to delete coo DNA outside the region of interest (data not shown), creating two derivatives. In both, the 5Ј-end of the coo DNA was at a newly created BamHI site at position Ϫ72 relative to the transcriptional start of cooF. The clone carrying cooFSCTJ had a 3Ј-end 113 bp downstream of the translational stop of cooJ, while the clone carrying cooFSCTJA had a 3Ј-end 272 bp downstream of the translational stop of cooA. The gentamycinresistance ⍀ cassette of pGM⍀1 (16) was then inserted at the BamHI site, and the inserts were excised by PvuII, isolated, ligated into BamHI-cut (Klenow polymerase-blunted) pRK404E1 3 to generate plasmids pCO46R (cooFSCTJ) and pCO47R (cooFSCTJA). These plasmids were subsequently transformed into UQ324 (17) and mated to Rhodobacter sphaeroides 2.4.1 (UR363) by the usual method, 4 with selection for resistance to tetracycline (1 g/ml) and tellurite (K 2 TeO 3 , 10 g/ml) (18), generating strains UR453 (pCO46R) and UR454 (pCO47R).
CO Dehydrogenase Assay-CooS activity with the CO-dependent reduction of methyl viologen was measured by the plate overlay assay (9) or by a spectrophotometric method (8). Briefly, R. sphaeroides strains UR363, -453, and -454 were cultivated aerobically in SMN medium containing 2 g/ml tetracycline then transferred to anaerobic SMN medium supplemented with 10 M nickel for overnight photosynthetic growth. Cultures were then diluted to an OD 680 ϭ 1.0 and cultivated in the presence or absence of CO for 4 h prior to lysis and CO dehydrogenase assays. 50% of the UR453 and UR454 clones remained tetracycline-resistant under these assay conditions.
Overexpression of CooA in R. rubrum-To overexpress CooA in R. rubrum, a plasmid fusing P nifH to cooA was constructed and introduced to R. rubrum in the following manner. pDWS119 (9), a pUX19-derived plasmid that carries cooA, was digested with BamHI and XmnI. The resulting 6348-bp fragment, which carries cooA, a portion of the putative cooA promoter, and the vector portion of pDWS119, was ligated to the 412-bp BamHI-HincII fragment from pUX111 (19) that carries the R. rubrum nifH promoter region (20) and the 5Ј-portion of the nifH coding region of R. rubrum, creating pDWS126.
The P cooA and nifH coding material remaining in pDWS126 (77 bp total) was deleted to create a junction between the ribosome binding site of nifH and the initiation codon of cooA in the following way. A primer, 5Ј-CGATGTTGAAACGAGGCGGCATGGAATCAATCCTTTTCTTCGG-TGATCCGGTCTTAAGGCGGG, (double underline indicates a base change (C to T) from the wild-type P nifH region to create a new AflII site (single underline)) was synthesized (Genosys Biotechnologies) and used for site-directed mutagenesis by a modification of the unique site elimination procedure (21), utilizing a single primer incorporating the desired deletion as well as a selectable restriction site (Bsu36I) loss. The desired plasmid (pDWS131) was identified by the presence of an new site in the plasmid (AflII) derived from the primer. The P nifH region and cooA on pDWS131 were verified by sequencing in one direction. pDWS131, which is a mobilizeable plasmid that does not replicate in R. rubrum, was used to transform Escherichia coli strain S17-1 (17), and the resulting strain was mated with R. rubrum strain UR2 (coo ϩ ). 4 Strains with pDWS131 integrated into the UR2 chromosome by homologous recombination were selected for kanamycin (15 g/ml). A single transconjugant, R. rubrum strain UR459, was used for further study.
Crude Extract Preparation-Cultures were harvested by centrifugation at 4,800 ϫ g for 15 min, and all handling of extracts utilized anoxic technique due to concern over the possible O 2 -lability of CooA. The cell pellets were resuspended in buffer A (25 mM MOPS (pH 7.4), 1 mM phenylmethanesulfonyl fluoride, 1 mg/ml leupeptin, 1.7 mM sodium dithionite, 1 mM dithiothreitol) and passed through a French press at 16,000 p.s.i. The mixture was centrifuged at 13,200 ϫ g for 1 h to remove the debris. The supernatant, to which glycerol was added to 5% (w/v) for storage, was used for the DNA-binding assay. The total protein concentration was determined by the Bradford dye binding procedure (22) using the protein assay dye reagent concentrate (Bio-Rad).
DNase I Footprinting-A 294-bp EcoRV-EagI fragment (Fig. 1B), which contains the P cooF region, was used for both DNase I footprinting and gel retardation assays. This fragment was isolated from a pB-SKS(Ϫ)-derivative (Stratagene), pCO17, which contains a portion of the coo region from pLJC24 (23). The fragment was uniquely 3Ј-end-labeled in the coding strand by filling the EagI end with [␣-32 P]dGTP and Sequenase, followed by purification via polyacrylamide gel electrophoresis (14) and an Elutip Minicolumn (Schleicher & Schuell). DNase I footprinting analysis of the P cooF region was performed as described previously (24) with the following modifications. The sealed tubes used for the protein-DNA binding reaction were degassed, and the head space was filled with argon. The DNA binding reactions were done under stringently anoxic conditions in the buffer B (20 mM Tris-HCl (pH 7.6), 7 mM MgCl 2 , 50 mM KCl, 7 mM dithiothreitol, 50 g/ml bovine serum albumin, 5% (w/v) glycerol) supplemented with 1.7 mM sodium dithionite in the presence or absence of CO. After 5-10 pmol of DNA was incubated with various amount of extracts in a 20-l volume at room temperature for 20 min, the mixtures were treated with 2 units/ml RQ RNase-free DNase I (Promega) for 30 s. DNase I cleavage products were separated on a 6% (w/v) polyacrylamide-urea gel.
Gel Retardation Assay-Radiolabeled DNA fragments were prepared similar to those in the footprinting experiment. The DNA-binding reactions were performed anoxically in the same buffer and at the same temperature used in the footprinting assays. The samples were applied to an anoxic 5% polyacrylamide gel (37.5:1 acrylamide/bis ratio) in standard 1 ϫ Tris borate/EDTA buffer (14) with 1.7 mM sodium dithionite. After prerunning at 180 V for 1.5 h, the upper running buffer was changed once in order to maintain strictly anoxic conditions. The electrophoresis was performed at 180 V for 3 h in 4°C.

RESULTS
Location of a CO-induced 5Ј-End mRNA of cooF-In order to identify the coo promoter upstream of cooF and to determine the effect of CO on its expression, we performed primer extension analysis on coo mRNA from both CO-induced and uninduced wild-type R. rubrum (strain UR2). Initially, we used a primer (Primer 1) that hybridized near the beginning of cooF coding region. The extension product suggested that the 5Ј-end of the cooF mRNA was approximately 250 nucleotides upstream from the translational start site of cooF. This signal was present only in the CO-induced sample, and no other primer extension products were observed (data not shown).
In order to more precisely identify the 5Ј-end of the cooF mRNA, a second primer (Primer 2) was designed that hybridized about 150 nucleotides upstream of cooF coding region. Results with this primer showed that the major transcript from the cooF promoter initiates with the A nucleotide positioned 257 bp upstream from the start codon of cooF. A minor product starting six nucleotides upstream of that site was also ob-4 D. P. Lies, personal communication.

FIG. 1. Identification of the transcriptional start site for cooF.
Panel A shows the result of primer extension of the region upstream of cooF. A 22-mer oligonucleotide (Primer 2), which is complementary to the coding strand in region Ϫ154 to Ϫ133 relative to the translational start site of cooF was 5Ј-end-labeled and used to prime the reverse transcriptase reaction. The sequencing ladder (G, A, T, C) used the same primer but was not end-labeled. ϩCO refers to RNA extracted from R. rubrum cells induced with CO, while the ϪCO lane refers to RNA from uninduced cells. Asterisks indicate the 5Ј-end of the major and minor RNA species detected. The fact that the primer for the reverse transcription was end-labeled and the one for sequencing was not caused a one-base shift in reading the transcriptional start site. Panel B provides a schematic of the region upstream of cooF. The space between the cooH and cooF genes is 450 bp. The transcriptional start site predicted by the major RNA species from panel A is indicated, as is the region protected in the footprinting experiments shown in Fig. 2A. Boldface letters in the CooA target region represent a 2-fold symmetric sequence that is highly similar to the consensus sequence motif recognized by CRP/FNR (Fig. 2B). The underlined CG and GG residues at Ϫ13, Ϫ14, Ϫ25, and Ϫ26 are characteristic of 54 -dependent promoters.
served. These primer extension products were only detectable in the CO-induced culture (Fig. 1A), indicating that the effect of CO is on the accumulation of coo mRNA. Fig. 1B shows a schematic of the transcription start site relative to other features in the region, including the putative CooA-binding site (see below).
CooA Is Sufficient for CO-dependent Expression in a Heterologous System-As noted above, mutant analysis has shown that CooA is necessary for CO-dependent expression (9). To test the possibility that CooA was sufficient for the response, consistent with the hypothesis that it was actually the CO sensor, two derivatives of plasmid pRK were created. The first carried cooFSCTJ (pCO46R), and the second contained cooFSCTJA (pCO47R). These two constructs were introduced into E. coli and R. sphaeroides and were tested for their ability to produce CO dehydrogenase in response to CO. The E. coli strains showed no detectable CO dehydrogenase activity by the plate overlay assay (data not shown). Likewise, the R. sphaeroides strain (UR453) with pCO46R and the parent strain (UR363) accumulated no CO dehydrogenase activity detectable by either the plate overlay assay or the in vitro spectrophotometric assay. In contrast, the R. sphaeroides strain with pCO47R, upon CO induction, produced detectable CO dehydrogenase (0.23 mol of CO oxidized/min/OD 680 ) in the spectrophotometric assay. Because R. sphaeroides lacks an endogenous coo system, it is likely that CooA is the CO sensor in this heterologous system.
Overexpression of CooA in R. rubrum-We expected that CooA, as a regulatory protein, would be nonabundant and thus anticipated that its activity would be difficult to detect in crude extracts. Consequently, we overexpressed CooA in R. rubrum from the R. rubrum nifH promoter and ribosome binding site. This promoter, which is active when R. rubrum is grown under nitrogen-fixing conditions, has been successfully used in our lab to overproduce nonabundant regulatory proteins (19) and has the potential to produce 1% of total cell protein. Construction of the P nifH ::cooA fusion and its integration into R. rubrum, creating strain UR459, are described under "Experimental Procedures." Extracts of UR459 (P nifH ::cooA) were examined by SDS-polyacrylamide gel for the presence of a protein band corresponding to CooA. The extracts of UR2 and UR407 (cooA::aacC1) (9), grown under the same conditions as UR459, were used as controls, as CooA was deficient in UR407 and is not expected to be detectable in UR2. A band migrating at about 25 kDa, the predicted molecular mass of CooA, was significantly more intense in extracts of UR459 than in those of UR2 or UR407 (data not shown).
CooA-and CO-dependent DNA Binding-To test the hypothesis that CooA is a CO-sensing transcriptional activator, in vitro interactions between CooA and the promoter region of cooF were investigated with or without CO. A 294-bp EcoRV-EagI fragment, which contains the cooF promoter region from position Ϫ248 to ϩ46 (Fig. 1B), was used as a probe in both DNase I footprinting and gel retardation assays. Crude extracts of strains that overproduce CooA (UR459), lack CooA (UR407), and have the normal low level (UR2) were prepared anoxically, due to concern that CooA might be O 2 -labile. All protein-DNA binding reactions were also kept under stringently anoxic and reduced conditions.
In DNase I protection experiments ( Fig. 2A), a specific pattern of protection was detected only with extracts from the CooA-overproducing strain in the presence of CO. No protection was observed in the same extract in the absence of CO, nor in the extract of the cooA mutant (UR407) regardless of the presence of CO in the binding reaction. This protection was not detectable with extracts of wild-type in presence of CO (data not shown), presumably due to low levels of CooA.
The site protected by CooA covers 28 bases (position Ϫ27 to Ϫ54) and contains a sequence of 2-fold symmetry (Fig. 1B) that is highly similar to the consensus sequence motif recognized by CRP/FNR in E. coli (Fig. 2B).
Gel retardation analysis with the same DNA fragment and extracts also revealed a DNA-protein complex whose presence requires both CooA and CO (data not shown). This CooA-and CO-dependent complex is very large; it remained in the wells of a 5% polyacrylamide (19:1 acrylamide/bis ratio) gel but entered a 5% polyacrylamide (37.5:1 acrylamide/bis ratio) gel, suggesting the presence of additional proteins in this complex.

DISCUSSION
Our previous mutational studies and sequence analysis of cooA led us to predict that CooA is a CO-sensing transcriptional activator similar to CRP and FNR. The work presented in this paper and elsewhere strongly supports the hypothesis that CooA is a CO-sensing protein responsible for controlled expression of the coo region in a fashion reminiscent of the action of CRP: (i) Northern blot analysis 1 and primer extension experiments demonstrate that CO affects mRNA accumulation; (ii) CooA is sufficient for CO-dependent expression in R. sphaeroides; (iii) DNA binding appears to be CooA-and CO-dependent in vitro; (iv) the detected CooA target site is very similar to the CRP/FNR consensus binding site; and (v) CooA is very similar to CPR and FNR in the helix-turn-helix DNA binding domain (9).
We initially looked for the CooA-binding site by both footprinting and gel retardation assays in the 250-bp EagI-BsmI region immediately upstream of cooF (Fig. 1B), in part because a strain (UR284) with an insertion at the EagI site (Fig. 1B) displayed CooS activity in a CO-dependent manner (8). We now believe that the observed expression in UR284 reflects transcription from P cooF through the Kan r insertion; the Kan r gene is derived from pUC4K and apparently lacks transcriptional terminators. No CooA-or CO-dependent DNA binding could be found between the EagI site and cooF (data not shown), however, and the transcriptional start site identified in this paper is clearly the physiologically significant one in vivo.
While the detected transcription start lacks the Ϫ10 and Ϫ35 sequences expected at a typical E. coli 70 promoter, a typical E. coli 54 recognition sequence is present, with GC and GG at Ϫ13 and Ϫ25, respectively (25) (Fig. 1B). An interaction of CooA with 54 would be interesting as no CRP-or FNR-controlled promoters are known to be recognized by 54 .
The center of the two-fold symmetry of the CooA binding site is at Ϫ43.5 with respect to the transcriptional start point of cooF. This distance is similar to the location of the CRP sites in class II CRP-dependent promoters (e.g. galP 1) and FNR sites in the FNR-dependent promoters (26,27). In CRP, the specific interaction between the side chains of the protein and a given base within the core motif 5Ј-TGTGA-3Ј have been reviewed (28,29): Arg-180 and Glu-181 directly contact the 5Ј-G and the nucleotide of complementary to the 3Ј-G, respectively. It is possible that Arg-177 of CooA, which is in the homologous position of Arg-180 of CRP, contacts the 5Ј-G of the CooA target site 5Ј-TGTCA-3Ј. The absence of a Glu in CooA corresponding to Glu-181 in CRP is consistent with the fact that there is a 3Ј-C instead of a 3Ј-G in the CooA target site (Fig. 2B).
We are currently employing the described gel shift as a functional assay for the purification of CooA. Analysis of the purified protein, together with the eventual determination of the CooA-regulated promoter upstream of cooH, will significantly increase our understanding of this regulatory response.