CrtJ Bound to Distant Binding Sites Interacts Cooperatively to Aerobically Repress Photopigment Biosynthesis and Light Harvesting II Gene Expression in Rhodobacter capsulatus *

Expression of light harvesting II genes and of bacteriochlorophyll and carotenoid biosynthesis genes inRhodobacter capsulatus is repressed under aerobic growth conditions by the transcription factor CrtJ. In this study, we demonstrate that the crtA-crtI intergenic region contains divergent promoters that initiate transcription 116 base pairs apart, based on primer extension analyses. DNase I protection assays demonstrate that purified CrtJ binds to one palindrome that overlaps the crtA −10 promoter recognition sequence as well as to a second palindrome that overlaps the −35 crtI promoter recognition sequence. Similar analyses also show that thepuc promoter region contains two distant CrtJ palindromes, with one near the −35 promoter recognition sequence and the other located 240 base pairs upstream. Gel mobility shift and filter retention assays indicate that CrtJ binds in a cooperative manner to these distantly separated palindromes. In vivo expression assays with puc and crtI promoter reporter plasmids further demonstrate that aerobic repression of pucand crtI expression requires both CrtJ palindromes. Thesein vitro and in vivo results indicate that aerobic repression of puc, crtA, and crtIexpression involves cooperative interactions between CrtJ bound to distant palindromes. A DNA looping model is discussed.

Like most species of anoxygenic (non-oxygen evolving) photosynthetic bacteria, Rhodobacter capsulatus represses synthesis of its photosynthetic apparatus under aerobic growth conditions (1). Oxygen repression involves the regulation of photosynthesis gene expression, which is controlled by overlapping regulatory circuits (reviewed in Ref. 2). Anaerobic activation requires a two-component system composed of a sensor histidine kinase (RegB) and a response regulator (RegA) to activate reaction center, light harvesting I, and light harvesting II structural genes that are encoded by the puh, puf, and puc operons, respectively (3)(4)(5). The aerobic repression circuit uses the repressor CrtJ to decrease aerobic expression of bacteriochlorophyll (bch) and carotenoid (crt) biosynthesis genes and of light harvesting II structural genes (puc) (6).
CrtJ-repressed promoters typically exhibit an Escherichia coli-like -70 sequence motif with one or two copies of a con-served CrtJ recognition palindrome TGTN 12 ACA in the Ϫ10 or Ϫ35 region (7)(8)(9). For example, the R. capsulatus bchC promoter has two CrtJ palindromes spaced 8 bp 1 apart, with one spanning the Ϫ10 and the other the Ϫ35 promoter recognition sequences. Binding of CrtJ to these sites is redox-dependent, with in vitro binding much tighter under oxidizing conditions than under reducing conditions (10). As shown in a companion study by Ponnampalam et al. (11), repression of bchC involves cooperative interactions between CrtJ bound to adjacent palindromes. Consequently, mutations in either palindrome disrupt in vivo regulation by CrtJ (12). The spacing between the two bchC palindromes is also critical, as demonstrated by the observation that ϩ6and ϩ11-bp additions between the palindromes abolished CrtJ binding (11). A similar arrangement of CrtJ palindromes spanning the Ϫ35 and the Ϫ10 promoter regions is also found in the Rhodobacter sphaeroides bchC and puc promoters (13,14) and in the Rhodopseudomonas palustris pucB and pucE promoters (15). This suggests that cooperative interactions between CrtJ bound to adjacent palindromes may be a highly conserved mechanism of repression.
In addition to promoters that have two closely spaced palindromes, there are several other CrtJ regulated promoters that appear to contain only one palindrome in the Ϫ10 to Ϫ40 region. For example, the intergenic sequence between the divergently transcribed crtI and crtA genes from R. capsulatus contains two palindromes located 76 bp apart (distance calculated from the axis of dyad symmetry for each palindrome). One palindrome overlaps a potential Ϫ10 region of the crtA promoter, and the other overlaps a potential Ϫ35 region for the crtI promoter (8). Inspection of the puc promoter region also indicates the presence of a single palindrome near the Ϫ35 promoter recognition sequence (16), with a second putative CrtJ palindrome located 240 bp upstream. The presence of two distant CrtJ palindromes in the puc promoter, and in the crtA-crtI intergenic region, is reminiscent of the gal, ara, or lac systems, which require multiple binding sites to obtain full repression of gene expression (17)(18)(19). This raises the possibility that binding of CrtJ to distantly separated palindromes may be a requirement for full repressive activity of CrtJ in this second class of promoters.
In this study, we used DNase I footprint analysis, gel mobility shift, and filter binding assays to demonstrate cooperative binding of CrtJ to the two separated palindromes in the puc and crtA-crtI promoter regions. We also demonstrated that cooperative interactions are required for efficient binding of CrtJ in vitro and for efficient aerobic repression in vivo. The mechanism of cooperativity between distant sites will be discussed.
Cloning and Plasmid Mobilization-For the crtI expression vector, a 2.7-kilobase SmaI-HindIII restriction fragment from pCrtI:Z (6) was cloned into the vector pPHU235 (24) cut by ScaI-HindIII, creating the vector pES2. Amplification of a fragment containing only the crtI palindrome was performed by polymerase chain reaction (PCR) using pES2 as template, and using primers Ose25, 5Ј-CTC GAG TCT GGG TCC CTT GTA AT (which introduced a XhoI site, indicated in boldface letters, in the amplified fragment), and Ose26, 5Ј-AGC AAG CTT GGC TGC AGG TCG. The 1.5-kilobase amplified fragment was purified by using a QIAEX II gel extraction kit (Qiagen), and cloned into the pCR-Script SK(ϩ) plasmid (Stratagene), creating the plasmid pES41. A 1.5-kilobase XhoI-HindIII fragment from pES41 was then subcloned into XhoI-HindIII sites in pPHU235 to construct the plasmid pES42. Plasmids pDN12S and pDN13S, which contain defined puc:lacZ fusions (16), and pPHU235-derived plasmids were mated into recipient strains as described by Nickens and Bauer (16).
␤-Galactosidase Assays-␤-Galactosidase activity was assayed from the cell culture as described by Miller (25) and modified as described by Elsen et al. (26).
Overexpression and Purification of CrtJ-CrtJ fused to an N-terminal tag of six histidine residues was overproduced using BL21(DE3)/ pET28::CrtJ and purified as described by Ponnampalam and Bauer (10). To reduce the amount of inclusion bodies formed, the cells were grown at 34°C with a slow shaking (250 rpm) and induced when cultures reached an A 600 of 0.9, by the addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 1 mM. The purified protein was partitioned into 10-l aliquots and stored at Ϫ80°C. The final protein concentration, which was typically 1.6 mg/ml, was measured with the CrtJ extinction coefficient (⑀180), which was determined using the A205/A280 method described by Scopes (27).
RNA Isolation and Primer Extension-Total RNA was isolated from a 600 ml culture of R. capsulatus SB1003 cells grown photosynthetically in PYS medium using the guanidinium thiocyanate-cesium chloride method of Kocabiyik (28) with modifications described by Kouadio (29).
For primer extension, the primers CrtA-F (5Ј-CGTCGAAACGGAA-CAGGCTGA) and CrtI-R (5Ј-ACAACGGCGCGACCCATACC) were commercially synthesized; these primers are designed to anneal the crtA and crtI transcripts, respectively. 32 P labeling of the primers was performed as described by Jiang and Bauer (30). Primer extension was carried out with 10 g of total RNA per reaction, as previously reported by Kouadio (29). A dideoxynucleotide sequencing ladder was obtained as described by the manufacturer (Sequenase sequencing kit, United States Biochemical) using the same labeled primers and the plasmid pCrtI:Z as a template (6). Prior to loading onto a 6% urea denaturing polyacrylamide gel, the reactions were heated at 90°C for 5 min.
Gel Mobility Shift Assays-Three probes were used for studying the effect of cooperative binding of CrtJ to the pucB or crtA/crtI promoter regions. Probes were prepared by PCR amplification using oligonucleotide primers that were 5Ј-end-labeled with 32 P as described above. The amplified DNA segments were then purified by electrophoresis in a nondenaturating 5% polyacrylamide gel and recovered by electroelution.
Amplification of the upstream (Ϫ279 to Ϫ296 bp) puc palindrome FIG. 1. Primer extension analysis of transcripts initiating from the crtA-crtI intergenic region. A and B show primer extension products initiated from the crtI and crtA promoters, respectively. The A, C, G, and T lanes represent dideoxynucleotide sequencing ladders generated with the same primers that were used for primer extension. Nucleotides corresponding to the primer extension bands are indicated on the printed sequence by arrows on the right. Boxed bases are the potential 70 Ϫ10 and Ϫ35 promoter consensus sequences, with the two pairs of half-arrows representing CrtJ-recognition palindromes.
Gel mobility assays were performed by first preparing 5 l of 4ϫ binding buffer (composed of 40 mM Tris-HCl (pH 8.0), 200 mM potassium acetate, 4 mM DTT, 5 l of dH 2 O, and 6 l of different dilutions of purified CrtJ) to which 4 l of the DNA substrate were added (composed of 4 fmol of 32 P-end labeled DNA probe and heparin as a nonspecific competitor at a 500-fold weight excess relative to the probe). Samples were incubated for 30 min at room temperature, loaded on a native 4% Tris-glycine-EDTA-buffered polyacrylamide gel, and electrophoresed at room temperature. The polyacrylamide gel was then dried and autoradiographed overnight at Ϫ80°C with an intensifying screen.
DNase I Footprint Analysis-PCR-amplified DNA segments of the puc promoter and of the crtA-crtI intergenic region that contained both palindromes, as described for the gel mobility shift assays, were used for DNase I footprint analysis. For selective labeling of DNA strands, one of the primers in the PCR was 5Ј 32 P-end-labeled prior to amplification. A 10-l binding reaction mixture was first prepared containing 1 l of DNA (22 fmol), 7 l of H 2 O, and 2 l of 5ϫ footprint binding buffer composed of 125 mM Hepes (pH 8.0), 250 mM potassium acetate, 25 mM magnesium acetate, 10 mM calcium chloride, 5 mM DTT, and 125 g/ml bovine serum albumin. The reaction mixture was then added to a 10-l solution of 1ϫ footprint binding buffer containing various amounts of CrtJ. Digestion with DNase I and subsequent termination of the assay were then carried out as described previously (31). A modified Maxam and Gilbert G ϩ A chemical sequencing reaction was used for determining the location of DNase I protection (32).
Filter Binding Assays-Nitrocellulose filter binding assays were used to determine the fraction of purified CrtJ that was active in DNA binding (10, 33) as well as to determine the effective concentration for 50% response (EC 50 ). For both assays, the binding reaction conditions were performed with the same volume and composition as that used for the gel mobility shift assays, with the exception of the absence of competitor and the addition of 0.1% Nonidet P-40 and 0.5 mg/ml bovine serum albumin in the CrtJ dilution buffer. The binding assays involved a 30-min incubation of the 20-l reaction mixtures at room temperature followed by filtering the mixture through 0.45-m pore size nitrocellulose filters (BA85, Schleicher & Schuell) that had been presoaked in 1ϫ binding buffer for at least 1 h at room temperature. Filters were then washed once with 1 ml of 1ϫ binding buffer, air-dried in scintillation vials, and counted in a Beckman LS 6500 scintillation counter in presence of 1 ml of Bio-Safe II biodegradable counting mixture (Research Products International Corp.).
Nitrocellulose-filter-retention efficiencies, defined as the fraction of the total protein-DNA complexes retained on the filter, were calculated with 0.2-1 fmol of 32 P-labeled DNA probe and varying amounts of CrtJ (from 0.02 to 90 pmol), and they ranged from 78 to 100%. To determine the percent of active protein, DNA excess assays contained 0.6 or 1 pmol of purified CrtJ incubated with varying amounts (from 0 to 1 pmol) of 32 P-labeled DNA probes of the puc promoter, containing either one or the two palindromes. The percentage of active protein was then calculated from the retention efficiencies of the filter and number of CrtJ binding sites as described in Ponnampalam and Bauer (10). CrtJ purified using the modified expression system described above was typically 8% active. For EC 50 calculation, the CrtJ DNA binding curves were drawn using curve fit values based on the ligand binding equation Cap * L n (K d ϩ L n ), with the GraFit program, version 3.03 (Erithacus Software Ltd.).

Determination of Transcription Initiation
Sites in the crtA-crtI Intergenic Region-We undertook an analysis of transcription start sites in the crtA-crtI intergenic region as a prelude to analysis of CrtJ binding to this region. Primer extension analysis using a crtI specific primer yielded a major product that corresponds to a G residue located 18 bp upstream of the start codon of crtI and two minor products located 9 and 10 nucleotides upstream (Fig. 1A). Inspection of the sequence immediately upstream of the major primer extension product revealed a putative -70 type promoter on the basis of sequences located 10 (AATGCA) and 35 (TTGACG) bp upstream ( Fig. 2A). As indicated in Fig. 2, this region contains a single CrtJ binding sequence (TGTN 12 ACG) that overlaps the Ϫ35 promoter motif. Putative Ϫ10 (TATCAT) and Ϫ35 (TTGTAA) sequences can also be identified upstream from the weak start sites, with the CrtJ binding site overlapping the Ϫ35 region. Primer extension analysis of crtA gave rise to a major product with a 5Ј-end that corresponds to a transcription start site located 23 bp upstream of the crtA translation start codon, as well as minor products located 75-77 bp upstream (Fig. 1B). The region upstream from the major product contains Ϫ10 (AATATC) and Ϫ35 (ATTACA) promoter recognition sequences and a potential CrtJ binding palindrome (TGTN 12 ACA) that overlaps the Ϫ10 recognition sequence (Fig. 2A). Inspection of the sequence upstream from the minor products did not reveal obvious Ϫ10 or Ϫ35 promoter recognition sequences ( Fig. 2A). If the upstream primer extension products do represent a transcription start site, this second crtA promoter is likely to be regulated by CrtJ because the CrtJ palindrome that spans the crtI promoter also overlaps this promoter ( Fig. 2A).

CrtJ Binds to Distant Palindromes in the crtA-crtI Intergenic
Region and in the puc Promoter Region-DNase I protection assays were performed on both strands of the crtA-crtI promoter region to determine whether CrtJ binds to the palindromes that were identified by sequence analysis. The DNase I digestion patterns in Fig. 3 show that CrtJ protects the palindrome that spans the Ϫ10 region of the crtA promoter (best resolved with the top strand in Fig. 3A), as well as the palindrome that spans the Ϫ35 region of the crtI promoter (best resolved in Fig. 3B). Several DNase I hypersensitive sites are present in each of the protected regions.
We also undertook an analysis of CrtJ binding to the puc promoter using similar techniques. Inspection of the sequence of this promoter region (Fig. 2B) revealed the presence of two putative CrtJ palindromes, with one adjacent to the puc Ϫ35 motif (downstream palindrome) and the other one located 240 bp upstream (upstream palindrome). As showed by the DNase I digestion pattern generated with the top strand (Fig. 3C), CrtJ protects a region extending from Ϫ274 to Ϫ308 bp that corresponds to the upstream CrtJ palindrome. DNase I protection patterns generated with the bottom strand (Fig. 3D) exhibit CrtJ protection extending from Ϫ37 to Ϫ62 bp that corresponds to the downstream CrtJ palindrome. Several hypersensitive sites to DNase I digestion are also observed within these protected regions (Fig. 3, C and D).
CrtJ Binds Cooperatively to the crtA, crtI, and puc Palindromes-We next addressed whether cooperative interactions may occur among CrtJ bound to the palindrome sites located in the crtA and crtI promoters or to the distantly separated puc binding sites, using gel retardation assays with probes that contained one or both of the relevant palindromes. As shown in Fig. 4A, lanes 1-4, a complete shift of a DNA probe that contained both the crtA and crtI palindromes was observed with as little as 0.5 g of CrtJ. In contrast, a DNA probe containing only the crtA palindrome (Fig. 4A, lanes 5-8) or only the crtI palindrome (lanes 9 -12) required 2 g of CrtJ to bind all the probe. (Note that the crtA and the crtI probes (Fig. 4A,  lanes 5-12) are contaminated with another DNA fragment, amplified during the PCR step and co-purified with the specific probes. The absence of a shift with these contaminating DNA fragments indicates good specificity for CrtJ binding to the target DNA).
Similar gel retardation assays were performed with DNA probes made from the puc promoter region. Using a probe containing both palindromes, we observed a complete shift with the lowest amount (0.35 g) of CrtJ used (Fig. 4B, lanes 1-4). When more CrtJ was added, the shifted fragment exhibited even slower mobility, presumably caused by nonspecific binding of CrtJ or by the formation of highly ordered DNA-protein complexes (Fig. 4B, lanes 3 and 4). When CrtJ was incubated with probes containing only the upstream (Fig. 4B, lanes 5-8) or downstream (Fig. 4B, lanes 9 -12) palindrome, a shift of the entire probe required 1.4 g of purified CrtJ.
To measure CrtJ-DNA binding affinities, we performed nitrocellulose filter binding assays with probes that contained one or both of the palindromic sites in the crtA-crtI and puc promoter regions. For this analysis, the same probes that were used for the gel mobility shift assays were analyzed for filter retention with various amounts of CrtJ. The resulting binding curves ( Fig. 5A and B) indicate that probes that contained single palindromes produced EC 50 values 6 -10-fold higher than for probes that contained two palindromes. When corrected for the percentage of active protein (see under "Materials and Methods"), the calculated EC 50 value for a probe with both the crtA and crtI palindromes was 2.6 ϫ 10 Ϫ9 M, versus 3.2 ϫ 10 Ϫ8 and 2.4 ϫ 10 Ϫ8 M for probes that contained only the crtA or crtI palindromes, respectively. For experiments that measured CrtJ binding to the puc promoter region, the calculated EC 50 value for the probe containing both palindromes was 3.2 ϫ 10 Ϫ9 M, compared with 1.6 ϫ 10 Ϫ8 and 2 ϫ 10 Ϫ8 M for probes that contained only the upstream (Ϫ279 to Ϫ296) or downstream (Ϫ39 to Ϫ56) palindromes, respectively. Fig. 6 summarizes the observed EC 50 values for the various individual binding sites that were analyzed in this study, as well as the observed EC 50 value for the bchC upstream palindrome that is reported in the companion study by Ponnampalam et al. (11).
The in Vivo Requirement for Two Distant Palindromes for Activity of the Repressor CrtJ-The in vivo requirement for two distant palindromes for CrtJ-mediated repression of gene expression was studied using lacZ reporter fusions in R. capsulatus. For analysis of puc expression, we used plasmid pDN13S, which has an extended 1051-bp puc promoter segment contain-ing both palindromes, and plasmid pDN12S, which has a truncated 284-bp puc promoter segment containing only the downstream (Ϫ39 to Ϫ56) palindrome (Fig. 7A) (16). When assayed for ␤-galactosidase activity in aerobically grown wild-type R. capsulatus cells (SB1003), we observed that cells harboring the construct with only one palindrome (pDN12S) have a significantly higher level of activity (144%) over cells that contain the construct that has two palindromes (pDN13S) (Fig. 7B). This elevated activity observed with pDN12S is clearly a result of the inability of CrtJ to effectively repress puc expression, as the level of activity is similar to that observed with the crtJ-disrupted strain DB469 carrying either pDN12S or pDN13S (Fig. 7B).
Two crtI:lacZ fusions were also constructed to assess the requirement for distantly separated CrtJ-binding sites in the crtA-crtI intergenic region. Plasmid pES2 contains the entire crtA-crtI intergenic region, including both the crtA and the crtI palindromes, whereas plasmid pES42 carries a truncated intergenic region in which the crtA palindrome had been deleted (Fig. 7C). As observed in Fig. 7D, the wild-type cells harboring the construct containing only the crtI palindrome (pES42) had elevated aerobic ␤-galactosidase activity (133%) over cells that contained the plasmid that has both palindromes (pES2). This reflects the impairment of CrtJ to repress crtI expression if only one palindrome is present. Furthermore, when the crtA palindrome was deleted, no regulation of crtI expression by CrtJ was retained, as evidenced by similar ␤-galactosidase activities in SB1003 and the crtJ-disrupted strain DB469 carrying plasmid pES42 (Fig. 7D).
Thus, we conclude that the two CrtJ palindromes separated by 240 bp in the puc promoter region and by 76 bp in the crtA-crtI intergenic region are required for efficient CrtJ-mediated repression of these genes in vivo. DISCUSSION This study demonstrates that CrtJ binds to two palindromes that are located 76 bp apart in the crtA-crtI intergenic region and to two palindromes located 240 bp apart in the puc promoter region. Efficient CrtJ repression of crtA, crtI, and puc expression involves cooperative interactions between CrtJ bound to these palindromes. Cooperative interaction has also been demonstrated for CrtJ repression of the bchC promoter, which has two closely spaced (neighboring) palindromes (10,11). Thus, a common characteristic of DNA binding by CrtJ appears to be a requirement for two binding sites that are either closely spaced, as for the bchC promoter, or distant, as for the puc and the crtA and crtI promoters.
Quantitative filter binding analyses indicate that CrtJ has similar affinities for the individual binding sites that are in the puc, crtA and crtI promoters, as demonstrated by EC 50 values that range from 1.6 ϫ 10 Ϫ8 to 3.2 ϫ 10 Ϫ8 M. Surprisingly, CrtJ has the same affinity for the TGTN 12 ACA palindromes as for  the variant TGTN 12 ACG that is present in the crtI promoter (Fig. 6). These values are very similar to that reported for the upstream bchC palindrome (7.8 ϫ 10 Ϫ8 M) by Ponnampalam et al. (11) in the companion study. In contrast, CrtJ exhibits a significantly reduced affinity (Ͼ1 M) for the downstream bchC palindrome (11) despite the fact that this palindrome still exhibits a very good consensus sequence (Fig. 6). The only significant difference is that this sequence exhibits a substitution of an A for a T two nucleotides upstream from the ACA that is conserved in the other palindromes. This indicates that a T at this position may be critical for effective CrtJ binding.
Compared with individual sites, the affinity of CrtJ for DNA fragments that contain the two puc palindromes, or both the crtA and crtI palindromes, is 6 -11-fold higher (3.2 ϫ 10 Ϫ9 and 2.6 ϫ 10 Ϫ9 M, respectively). This increase of affinity is similar to the observed 26-fold increase in CrtJ binding to two palindromes in the bchC promoter region over a single binding site (11). Even though there must be some fundamental differences in the interactions among CrtJ bound to these two different classes of promoters (i.e. the potential involvement of DNA looping between distant palindromes versus direct interaction between adjacent palindromes), the relative binding affinities and level of repression are conserved.
Our analysis of transcription start sites in the crtA-crtI intergenic region indicates that one of the CrtJ-binding sites overlaps a -70 type Ϫ35 motif for the crtI promoter. The other binding site overlaps the Ϫ10 motif for the crtA promoter. Thus, cooperative binding of CrtJ to these two palindromes provides a mechanism of coordinately repressing these divergent promoters. What is the significance of co-repressing the crtA and crtI promoters? Previous studies have postulated that crtI is the first of a two-gene operon that also contains crtB (7,34). crtI and crtB are known to code for phytoene dehydrogenase and phytoene synthase, respectively, which catalyze two of the earliest steps in carotenoid biosynthesis (35)(36)(37)(38). It has also been demonstrated that the translational stop codon for crtA overlaps the translational start codon for bchI; thus, the transcripts initiating at the crtA promoter should transcribe the downstream bchI gene (34,39). bchI codes for a subunit of magnesium chelatase, an enzyme involved in inserting Mg 2ϩ into protoporphyrin IX forming the first intermediate in the bacteriochlorophyll branch of the tetrapyrrole biosynthetic pathway (34,40). Thus, coordinate repression of the crtA and crtI promoters could influence the flow of metabolic intermediates into both carotenoid and bacteriochlorophyll biosynthetic pathways.
Disruption of crtJ caused increased aerobic expression of bch and crt genes, but a residual level of anaerobic activation was still observed, suggesting that these genes are subject to aerobic repression by CrtJ and anaerobic activation by unknown factors (inducers) (6,12). Similarly, expression of the puc operon is repressed by CrtJ under aerobic growth conditions and induced under anaerobic growth conditions by the twocomponent regulatory proteins, RegB and RegA (4 -6, 9). Recently, Du et al. (41) isolated a mutant version of RegA, called RegA*, and demonstrated that RegA* binds to the puc promoter, protecting two regions extending from Ϫ52 to Ϫ69 and Ϫ73 to Ϫ80. Thus, RegA binds to a region of the puc promoter that partially overlaps the downstream binding site of CrtJ (from Ϫ37 to Ϫ62). Consequently, the two proteins may compete in vivo for their respective binding sites. Furthermore, deletion and mutational analyses of the bchC promoter region indicate that this promoter is activated under anaerobic conditions by a cis-acting site (AT-rich region) located just upstream of the -70-like sequence of the promoter (12). This region overlaps the bchC upstream palindrome of CrtJ (10). It is therefore possible that, like the puc and bchC promoters, other crt and bch genes are controlled by CrtJ competing with a transcriptional activator for overlapping binding sites.
As indicated above, cooperative interactions between CrtJ bound to widely separated sites on the DNA presumably involves the formation of a DNA loop (Fig. 8). Protein-mediated looped complexes are important regulatory elements in many systems. Examples include the regulation of prokaryotic and eukaryotic gene expression, site-specific recombination, and DNA replication (reviewed in Ref. 42). Different factors are known to facilitate the formation and stability of DNA-protein looped complexes. This includes the presence of intrinsically bent DNA sequences or the binding of proteins that bend DNA. For instance, integration host factor (IHF) (43), HU (44), and cAMP receptor protein (45,46) can stabilize protein-mediated looped complexes.
There is no direct evidence that cooperative binding by CrtJ to distant sites occurs via a DNA loop. However, there is one observation that favors a DNA looping in CrtJ-mediated repression of gene expression. Indeed IHF, which is a sequencespecific DNA-binding protein that induces a severe bend (Ͼ160°) in the DNA helix (47), has recently been shown to bind between the two CrtJ palindromes on the puc promoter (16). Sequence scanning has revealed putative IHF-binding sites in the crtA-crtI intergenic sequence (data not shown). Thus CrtJinduced looped DNA could be stabilized by IHF in vivo. The possible involvement of IHF in the mechanism of CrtJ-mediated repression is currently under investigation.