Transcription Activation by CooA, the CO-sensing Factor fromRhodospirillum rubrum

CooA, a member of the cAMP receptor protein (CRP) family, is a CO-sensing transcription activator fromRhodospirillum rubrum that binds specific DNA sequences in response to CO. The location of the CooA-binding sites relative to the start sites of transcription suggested that the CooA-dependent promoters are analogous to class II CRP-dependent promoters. In this study, we developed anin vivo CooA reporter system in Escherichia coli and an in vitro transcription assay using RNA polymerases (RNAP) from E. coli and from Rhodobacter sphaeroides to study the transcription properties of CooA and the protein-protein interaction between CooA and RNAP. The ability of CooA to activate CO-dependent transcription in vivoin heterologous backgrounds suggested that CooA is sufficient to direct RNAP to initiate transcription and that no other factors are required. This hypothesis was confirmed in vitro with purified CooA and purified RNAP. Use of a mutant form of E. coli RNAP with α subunits lacking their C-terminal domain (α-CTD) dramatically decreased CooA-dependent transcription of the CooA-regulated R. rubrum promoter P cooF in vitro, which indicates that α-CTD plays an important role in this activation. DNase I footprinting analysis showed that CooA facilitates binding of wild-type RNAP, but not α-CTD-truncated RNAP, to P cooF . This facilitated binding provides evidence for a direct contact between CooA and α-CTD of RNAP during activation of transcription. Mapping the CooA-contact site in α-CTD suggests that CooA is similar but not identical to CRP in terms of its contact sites to the α-CTD at class II promoters.

change that allows a dimer of CRP to bind to a specific 22-base pair sequence at target promoters (consensus sequence is 5Ј-AAATGTGATCTAGATCACATTT-3Ј, in which the most important bases for CRP recognition are in bold) and to activate transcription at those promoters (18). CRP-dependent promoters can be grouped into two classes based on the position of the CRP-binding site relative to the start of transcription as well as on the mechanism for transcription activation (19). At class I promoters, the DNA-binding site for CRP is upstream of that for RNAP and is centered at position Ϫ61.5, Ϫ71.5, Ϫ82.5, or Ϫ92.5. At class II CRP-dependent promoters, to which the CooA-dependent promoters are analogous, the binding site for CRP is centered at Ϫ41.5, overlapping the Ϫ35 region, and the ␣-CTD binds to DNA upstream of the CRP dimer.
Direct interaction between CRP and RNAP plays a pivotal role in transcription activation at both promoter classes (20,21). In particular, transcription activation at class II promoters requires two distinct contacts between CRP and the ␣ subunit and a third contact between CRP and 70 . One interaction is between activating region 1 (AR1) of the upstream subunit of the CRP dimer and the ␣-CTD. This interaction increases initial binding of RNAP to the promoter (22). Recently, residues 285-288 and 317 of ␣-CTD have been shown to comprise the surface that interacts with AR1 of CRP at class II promoters (23). The second contact, between activating region 2 (AR2) of the downstream subunit of CRP and the N-terminal domain of the ␣ subunit, facilitates isomerization of the closed complex to the open complex (24). The residues in AR1 and AR2 of CRP are not conserved in CooA, which suggests that there might be certain differences in the interactions of CooA and CRP with ␣. The third activator region (AR3) in CRP, formed by residues 52-58, interacts with 70 of RNAP (20). Because the AR3 region in CRP is highly similar to an analogous region in CooA, this region in CooA might serve an AR3-like function.
The two CooA-regulated R. rubrum promoters, P cooF and P cooM , contain 2-fold symmetric DNA sequences that serve as CooA-binding sites and are similar to the CRP consensus sequence. This is consistent with the similarity between CooA and CRP in their helix-turn-helix motifs (2,3). The CooAbinding sites lie at the Ϫ43.5 and Ϫ38.5 positions relative to the transcription start sites in P cooF and P cooM , respectively, overlapping with the Ϫ35 region (2,3). This overlap suggests that both CooA-regulated promoters are analogous to class II CRP-dependent promoters. We chose P cooF for this study because it is the stronger promoter based on the amount of primer extension product and level of coo-encoded proteins synthesized in vivo (1,2).
Although CooA shares some common features with CRP, such as DNA binding properties and effector-induced activation, it displays striking differences from CRP in the effectorbinding domain and in regions AR1 and AR2. We were interested to know whether CooA was necessary and sufficient for CO-dependent activation of transcription and whether the mechanism of activation by CooA was similar to that of CRP. In this work, we used in vivo CooA reporter systems and in vitro transcription assays to examine the properties of CooA in transcription activation. Because of the particular questions we wished to address, we performed the bulk of the work with RNAP from E. coli, so that any differences between CooA and the CRP detected would reflect properties of CooA. The nature of transcription activation by CooA was investigated through the study of the interaction between CooA and RNAP and, in particular, the interaction between CooA and the ␣-CTD.

EXPERIMENTAL PROCEDURES
Construction of a System for CooA Expression and a Reporter of CooA Activity in E. coli-CooA was overexpressed from a vector, pYHA1, created as follows. A 1.9-kilobase PvuI-BamHI fragment containing P tac -cooA-rrnBT 1 T 2 , was isolated from pKK223-3 (8), digested with PvuI, mung bean nuclease, and BamHI, and cloned into the EcoRV and BamHI sites of pACYC184. A plasmid containing a P cooF -lacZ fusion, pYHF4, was constructed by inserting a polymerase chain reactionamplified EcoRI-HindIII fragment, extending from positions Ϫ250 to ϩ70 of P cooF , into plasmid pMSB1 (25), which created pMSBP cooF . The reporter region was recombined from plasmid pMSBP cooF into phage RS468 (26) in strain DH5␣ containing pYHA1. Lysogens were screened by the blue color of plaques on Luria broth (LB) ϩ X-gal plates incubated anaerobically in the presence of CO. The promoter region of integrated P cooF -lacZ fusion in the chromosome was confirmed by DNA sequencing.
In Vitro Transcription Assays-Polymerase chain reaction-amplified EcoRI-HindIII fragments from positions Ϫ250 to ϩ70, Ϫ90 to ϩ70, and Ϫ60 to ϩ70 of P cooF were cloned into pRLG770, which contains transcription terminator rrnBT 1 T 2 downstream of the multicloning site (27), to yield plasmids pYHF1, pYHF2, and pYHF3, respectively. The supercoiled plasmids used as DNA templates for these assays were purified with the Midi Kit from Qiagen. The RNAPs used in these experiments were E 70 purified from E. coli or a Rhodobacter sphaeroides RNAP preparation enriched for the E 70 homolog (28). Standard multiple-round transcription assays (13) were modified as described below to accommodate the requirement of CooA for an anoxic environment to bind CO (7). The sealed tubes containing 25-l reactions (0.2 nM supercoiled plasmid, 3.5 nM RNAP, 40 nM CooA dimer, and a buffer (30 mM KCl, 40 mM Tris acetate, pH 7.9, 10 mM MgCl 2 , 1 mM dithiothreitol, 100 mg/ml bovine serum albumin, 200 mM ATP, 200 mM CTP, 200 mM GTP)) were degassed and filled with argon in the head space. After the addition of dithionite to 1.7 mM to scavenge any free oxygen, CO was added, and the reactions were incubated at room temperature for 15 min. This step served to activate CooA and allow the activated CooA to interact with the promoter and RNAP to form a predicted 21-nucleotide transcript by incorporating ATP, GTP, and CTP. The reactions were then exposed to air, and 10 mM UTP and 5 Ci of [ 32 P]UTP (DuPont) were added to extend the mRNA at room temperature for 20 min. Reactions were terminated and electrophoresed as described (29). The signal intensities of transcripts were quantified using a PhosphorImager (Molecular Dynamics) and ImageQuant software.
DNase I Footprinting Assays-A DNA fragment containing the P cooF sequence from position Ϫ90 to ϩ70 was polymerase chain reactionamplified with an unlabeled bottom strand primer and a top strand primer labeled with [␥-32 P]ATP and polynucleotide kinase. The amplified fragment was purified by polyacrylamide gel electrophoresis, followed by an Elutip Minicolumn (Schleicher & Schuell). The labeled fragment was incubated with CooA, or RNAP, or both in the presence of CO and under the stringent anoxic conditions described previously (2), except that 40 nM pure CooA and 3.5 nM RNAP were used in a 20-l binding reaction. The reactions were treated with 2 units/ml RQ RNasefree DNase I (Promega) for 30 s. DNase I cleavage products were separated on a 6% (w/v) polyacrylamide-urea gel. Neither heparin nor any nucleotides were added into the reactions.
CO Induction of P cooF Expression in E. coli and Measurement of ␤-Galactosidase Activity in Vivo-Strains with a P cooF -lacZ reporter were grown aerobically in LB medium containing 100 mg/ml ampicillin and 35 mg/ml chloramphenicol for 12 h to reach stationary phase. In stoppered test tubes, 20-l cultures were diluted into 2 ml of LB medium supplemented with 20 mM glucose and the same antibiotics. The air in the head space was replaced by argon and 2% CO, and the cultures were grown anaerobically to an A 600 of approximately 0.45. ␤-Galactosidase activity was determined according to Miller (30).

CooA Is Necessary for CO-dependent Expression in Different
Organisms-CooA has been shown to be necessary for the CO-dependent expression of the two coo operons of R. rubrum (1,2), and purified CooA binds DNA in a CO-dependent manner (7). These data are consistent with the hypothesis that CooA is both necessary and sufficient for sensing and activating transcription in response to CO, but they are not conclusive. We therefore examined the requirement for CooA in two heterologous systems. We chose R. sphaeroides because it is related to R. rubrum, yet does not appear to have the coo system, as judged by the absence of CO dehydrogenase activity and a failure to hybridize to probes of the coo genes (2). In this organism, a pRK404-based plasmid carrying cooFSCTJ with its normal promoter failed to produce detectable CO dehydrogenase activity (the product of cooS) in response to CO. However, when cooA was added to the plasmid in its normal position at the 3Ј end of the cooFSCTJ operon, exposure of the R. sphaeroides strain carrying this plasmid to CO produced easily detectable CO dehydrogenase activity (2).
We then examined CO-and CooA-dependent transcription in E. coli, an organism that is less related to R. rubrum and also lacks any evidence of a coo system. For this test, a reporter system was constructed that contained a P cooF -lacZ fusion in the chromosome and a plasmid overexpressing CooA. In this system, we detected a substantial increase of ␤-galactosidase activity upon CO induction (200 Miller units in the presence of CO and 1.6 units in its absence), suggesting that CooA is sufficient for activating the transcription of P cooF in E. coli. Similar results with CooA reporters in E. coli have recently been reported by others (9). These results establish that CooA is necessary for the transcriptional response to CO and that it is able to associate productively in vivo with RNAPs from both E. coli and R. sphaeroides.
CooA Is Sufficient to Activate the Transcription of P cooF in Vitro-The above results indicate that CooA is necessary, but only in vitro analysis can establish whether it is sufficient for CO-dependent transcriptional activation or whether additional factors are required for this activation. The ability of CooA to activate P cooF was studied by monitoring RNA synthesis in a purified system containing only DNA, RNAP, CooA, nucleotide triphosphates, and the proper buffer. To investigate the nature of CooA-mediated activation at the P cooF promoter, we modified the standard in vitro transcription assay, because CooA is only able to bind CO when reduced. The modified assay, detailed under "Experimental Procedures," was kept anoxic in the presence of CO to support activation by CooA until the formation of a 21-nucleotide transcript from P cooF . At this point, the reaction was exposed to air and extended aerobically for technical convenience. When the reactions were maintained strictly anoxically throughout the entire experiment, a quantitatively similar result was obtained, indicating that the in vitro transcription assay conditions we employed were sufficient for maximal CooA activity (data not shown).
The in vitro transcription assays were performed with a supercoiled DNA template (pYHF1) containing the P cooF and extending Ϫ250 base pairs upstream and ϩ70 base pairs downstream of transcription start site. The reactions were carried out in the presence or absence of CO. As shown in lanes 2 and 4 of Fig. 1, CO-dependent transcripts were detected using RNAP from both R. sphaeroides and E. coli. The observed size of the transcript from P cooF correlated well with the predicted size of 240 nucleotides. In the absence of CO, no transcripts from P cooF were seen, whereas transcription of the control (RNA-1) was not affected by CO (Fig. 1, lanes 1 and 3).
These results demonstrate that CooA is sufficient for CO-dependent transcriptional activation and that no other factors are required. The results with the E 70 from E. coli indicate that P cooF can be recognized by 70 when activated by CooA.
CooA-activated Transcription Requires the ␣-CTD of RNAP-Because the ␣-CTD makes specific DNA contacts at some promoters and specific protein contacts with a number of transcription activators including the CooA homolog CRP (13,20), we wished to test whether there were similar contacts between ␣-CTD and either CooA or P cooF . We first examined whether the C-terminal domain of the RNAP ␣ subunit is required for transcription activation of P cooF by CooA. To address this question, we assayed the ability of a reconstituted E 70 containing a truncation of the C-terminal domain of ␣ to direct transcription from CooA-dependent promoter P cooF in vitro. Use of E. coli RNAP with the ␣-CTD truncation resulted in a dramatic reduction of transcription from P cooF (Fig. 1, lanes  5 and 6). The enzyme activities of wild-type and mutant RNAPs were similar as demonstrated by transcription from the RNA-1 promoter. With a longer exposure of the x-ray film, we were able to detect a low level of a CO-dependent transcript from P cooF using RNAP with ␣-235 (data not shown). Because the ␣-CTD is known to make contacts with activators and UP elements, the ineffectiveness of the ␣-CTD-truncated RNAP for P cooF transcription suggests that the ␣ truncation disrupts the binding of ␣ to CooA, to a UP element, or to both. CooA Utilizes Protein-Protein Contacts with ␣-CTD to Facilitate RNAP Binding to P cooF -To analyze potential proteinprotein interaction between CooA and RNAP and to specifically address interactions between ␣-CTD and CooA, we performed DNase I footprinting experiments using wild-type RNAP or the ␣-CTD-truncated RNAP in the presence or absence of CooA. As shown in Fig. 2, lanes 2-4, CooA alone protected P cooF on the top strand from positions Ϫ55 to Ϫ31 relative to the start site of transcription, in agreement with our previous observations (2,7). In contrast, neither the wild-type nor the ␣-CTD-truncated RNAPs alone protected P cooF from DNase I (lanes 5 and  7). This result indicates that RNAP does not form a stable complex at P cooF in the absence of CooA. When both CooA and wild-type RNAP were incubated together with the DNA fragment, the protected region extended from the CooA-binding site in both directions, upstream to Ϫ68 and downstream to at least Ϫ1 (lanes 8 and 9). A similar result was obtained when R. sphaeroides RNAP was used, indicating that the RNAPs from these two different organisms function similarly in interacting with CooA at P cooF (lanes 10 -12). In contrast to the results obtained with the wild-type RNAP, the ␣-CTD-truncated E. coli RNAP showed no evidence of forming a stable complex even in the presence of CooA (lane 6). This suggests that CooA activates transcription by enhancing the initial and stable binding of RNAP to P cooF through direct protein-protein contact with the ␣-CTD.
Analysis of CooA-regulated Promoter P cooF -As noted above, the presence of CooA and wild-type RNAP, but not the ␣-CTDtruncated variants, caused DNase I protection to extend upstream and downstream of that region protected by CooA alone. The region upstream of the CooA-binding site is A ϩ T-rich (Fig. 3) relative to most of the R. rubrum genome, reminiscent of the A ϩ T-richness of UP elements in E. coli promoters that contact the ␣-CTD to increase transcription (13,38). To test whether ␣ interacts with this region in a sequencespecific manner, we created P cooF constructs differing only in the extent of that region. P cooF sequences from positions Ϫ90 to ϩ70 and Ϫ60 to ϩ70 were cloned into the transcription assay vector pRLG770, resulting in the constructs pYHF2 and pYHF3, respectively. These plasmids were tested for CooA-dependent in vitro transcription activity with wild-type RNAP. The level of transcripts from those constructs was quantitatively compared with that of the construct pYHF1 containing the P cooF sequences from Ϫ250 to ϩ70. As shown in Fig. 4, these three templates yielded similar amounts of CooA-dependent transcripts, indicating that the specific sequences upstream of position Ϫ60 in P cooF do not contribute significantly to promoter activity. This also suggests that the upstream A ϩ T-rich sequence of P cooF does not act as a UP element. Although we cannot exclude the possibility that some portion of the protection upstream of the CooA-binding site is due to a conformational change in CooA induced by the presence of RNAP, the interaction of ␣-CTD with DNA upstream of CooA is consistent with the sequence-nonspecific interactions between DNA and ␣-CTD observed at class II CRP-dependent promoters (23).
Identification of Residues in ␣-CTD That are Critical for CooA-dependent Activation at P cooF -We wished to determine whether CooA made similar contacts with the ␣-CTD as seen with its homolog, CRP, and also to test whether the extended DNase I protection upstream of CooA reflected the direct contact of ␣-CTD with DNA. To address these questions, we measured CooA-activated transcription using a CooA reporter system containing a P cooF -lacZ fusion integrated into the E. coli chromosome, a plasmid expressing CooA, and a library of plasmids encoding single alanine substitutions throughout the entire ␣-CTD (23). The effects of the ␣-CTD mutants in this screen were small and somewhat variable (data not shown), but they did identify a few potential mutants that were then assayed for their effects on CooA-dependent transcription in vitro.
We tested reconstituted RNAPs in vitro containing single alanine substitutions for Thr 285 , Val 287 , Glu 288 , and Arg 317 (the patch on ␣ that interacts with CRP at class II CRP-dependent promoters (23)), Arg 265 (the residue most important for interaction of ␣ with DNA (14)), and a few additional mutants suggested by the in vivo screen, including Val 306 , Leu 307 , and Ser 313 .
The results of the in vitro transcription analysis with a subset of the variant RNAPs are shown in Fig. 5. The activity of each RNAP preparation was normalized to the transcription from the RNA-1 promoter. The R265A, L307A, and V287A RNAPs were defective in CooA-dependent transcription of P cooF , providing 13, 34, and 37% of wild-type RNAP activity, respectively (Fig. 5, lanes 2, 3, and 7). Because R265A (14) and L307A 2 affected UP element-dependent transcription and RNAP extended the DNase I protection upstream of the CooAbinding site, this strongly suggested that ␣-CTD makes contacts with DNA upstream of CooA and that these contacts are important for CooA-meditated transcription activation. V287A was also defective in CooA-dependent transcription, although the other ␣-CTD variants important for class II CRP-dependent transcription (T285A, E288A, and R317A) had little or no effect. These results suggest that the CooA contact site in the ␣-CTD of RNAP shares some determinants of, but is not identical to, the site for CRP contact at class II-type promoters. RNAPs containing ␣ mutants V306A and S313A, which were also suggested by the in vivo screen, were not defective in CooA-dependent transcription in vitro (data not shown).

DISCUSSION
Many transcription activators directly contact the ␣ subunit of RNAP (16,17). In this study, we determined that CooA is sufficient for directing RNAP to initiate transcription of P cooF and that CooA-activated transcription of P cooF requires the C-terminal domain of the RNAP ␣ subunit. Consistent with these observations, CooA facilitates the binding of wild-type RNAP to P cooF but not that of ␣-CTD-truncated RNAP. These observations suggest that direct protein-protein contact between CooA and the ␣-CTD of RNAP plays an essential role in transcription activation of P cooF .
We suspect that the ␣-CTD/CooA interaction plays a role similar to that between AR1 of CRP and ␣-CTD at class II CRP-dependent promoters. In class II CRP-dependent promoters, ␣-CTD makes nonspecific contacts with the DNA segment immediately upstream of the CRP site (16,23). In this study, we found that wild-type RNAP, but not RNAP with the ␣-CTD truncation, extends DNase I protection upstream of that protected by CooA alone. Furthermore, we determined that ␣ R265A and L307A, which decrease UP element-dependent transcription, also affect P cooF activity. Because the specific DNA sequence upstream of the CooA-binding site is not critical for P cooF activity, we propose that ␣-CTD makes nonspecific contacts with the DNA immediately upstream of the CooA site and that this protein-DNA interaction is important for CooAactivated transcription of P cooF .
Val 287 is also important for CooA-activated transcription, although the other tested ␣ residues that contact CRP at class II promoters are not critical for CooA-dependent transcription of P cooF . Therefore, a different set of side chains, but probably the same region of ␣-CTD, might be important for activation by CRP and CooA. This hypothesis is consistent with the low similarity between CRP and CooA in activating region 1 (22). In the DNase I footprinting analysis, RNAP does not completely protect the DNA downstream of P cooF ϩ1, even in the presence of CooA and in the absence of heparin (Fig. 2). In addition, the majority of the CooA-RNAP-P cooF complexes detected in gel-shift assays are heparin-sensitive (data not shown). This partial downstream protection and the heparin sensitivity suggest that the detected ternary complex (CooA-RNAP-P cooF ) may be a mixed population of closed and open complexes. Although most CRP-dependent promoters, such as lacP1, melT, galP1, form stable open complexes with RNAP in the presence of CRP (32)(33)(34)(35), other promoters, such as rrnB P1 even in the presence of its activator protein Fis (36), form an unstable open complex with RNAP (37).
Because basal level transcription of P cooF in the absence of active CooA is not detectable, it is formally possible that the deletion of ␣-CTD affects basal level transcription of P cooF and not its activation by CooA. However, the requirement for at FIG. 3. Sequence of P cooF region. The figure shows the upper strand sequence of P cooF . The transcription start is at ϩ1. The underlined sequences represent the putative Ϫ10 region, the 2-fold symmetrical CooA-binding site, and the A ϩ T-rich sequence upstream of CooA-binding site as marked, respectively. The region protected by CooA in DNase I footprinting experiments is indicated by the filled bar. The open bar represents the region protected by wild-type RNAP plus CooA.

FIG. 4.
In vitro transcription analysis of P cooF deletions. The transcription reactions were performed in the same manner as described in Fig. 1. All of the reactions contained wild-type RNAP from E. coli, CooA, and CO. The supercoiled DNA templates used in the assay differ only in the extent of the upstream sequence of P cooF . Lane 1, pYHF3 containing P cooF sequences from Ϫ60 to ϩ70; lane 2, pYHF2 containing P cooF sequences from Ϫ90 to ϩ70; lane 3, pYHF1 containing P cooF sequences from Ϫ250 to ϩ70. The positions of transcripts initiated at the P cooF and the RNA-1 promoter are indicated by arrows. coli RNAPs were reconstituted with either wild-type (WT) ␣ subunit or the ␣ subunits containing alanine substitutions at the positions indicated in the figure. All of the reactions were carried out in the presence of CooA, CO, and the supercoiled DNA template with P cooF sequences from Ϫ250 to ϩ70. The positions of transcripts initiated at P cooF and the RNA-1 promoter are indicated by arrows. Based on the quantitation of transcript accumulations, R265A, L307A, and V287A show 13, 34, and 37% of wild-type RNAP activity for P cooF , respectively. These percentages represent the averages of three independent experiments in which all values were within 18% of the average. least one residue in ␣-CTD (Val 287 ) that has no effect on DNA binding (23) suggests that the ␣-CTD requirement involves, at least in part, a protein-protein interaction between CooA and ␣-CTD.