A repressor-antirepressor pair links two loci controlling light-induced carotenogenesis in Myxococcus xanthus.

The light-inducible carB operon encodes all but one of the structural genes for carotenogenesis in Myxococcus xanthus. It is transcriptionally controlled by two proteins expressed from two unlinked genetic loci: CarS from the light-inducible carQRS operon, and CarA from the light-independent carA operon. CarA represses transcription from the carB promoter (P(B)) in the dark, and CarS counteracts this on illumination. The CarA sequence revealed a helix-turn-helix DNA-binding motif of the type found in bacterial MerR transcriptional factors, whereas CarS contains no known DNA-binding motif. Here, we examine the molecular interplay between CarA and CarS. We demonstrate the following. (i) Whereas CarS exhibits no DNA binding in vitro, CarA binds specifically to a region encompassing P(B) to form at least two distinct complexes. (ii) A palindrome located between positions -46 and -63 relative to the transcription start point is essential but not sufficient for the formation of the two CarA-DNA complexes observed. (iii) CarS abrogates the specific DNA binding of CarA. CarA is therefore a repressor and CarS an antirepressor. (iv) CarS physically interacts with CarA; thus, the functional interaction between them is mediated by protein-protein interactions.

One of various cellular responses to blue light is the induction of the synthesis of carotenoids. These protect cells against photo-oxidative damage by quenching singlet oxygen and other free radicals produced on illumination (1,2). The Gram-negative bacterium Myxococcus xanthus is a model prokaryotic system for investigating how blue light switches on the network of cellular activities leading to carotenoid synthesis (3). Genetic analyses have revealed a number of regulatory and structural genes involved in this response (Ref. 4; see Fig. 1). One enzyme involved in carotenoid synthesis is encoded by gene crtI, and all the rest by the unlinked carB operon (5,6). Photoinduction of these structural genes is mediated by at least six regulatory genes as follows: the carQ, carR, and carS gene cluster and the unlinked carD, ihfA, and carA genes.
Transcriptional activation of crtI is mediated by the extra-cytoplasmic function-factor CarQ (7)(8)(9) and by CarD, a multifunctional transcriptional factor of considerable resemblance to eukaryotic high mobility group A proteins (10 -12). Light up-regulates crtI expression by triggering the liberation of CarQ from CarR, a membrane-associated protein that sequesters CarQ in the dark (8). The released CarQ is then free to activate transcription from the crtI promoter (P I ). CarQ, in conjunction with CarD and the histone-like protein IhfA, also promotes transcription from its own promoter (P QRS ), leading to increased production in the light of the three proteins encoded in the operon (10,13,14). Photoinduction of the structural genes in the carB operon depends on a different set of regulatory proteins: CarS, encoded by the third gene of the carQRS operon (13), and CarA, produced independently of light from an unlinked operon (Fig. 1). A non-polar deletion within the carA gene leads to light-independent expression of the carB operon, indicating that CarA acts as a negative regulator of the carB promoter (P B ) in the dark. 1 Cells bearing a lack-of-function mutation in carS, on the other hand, do not display light activation of the carB operon; CarS thus functions as a positive regulator of P B in the light (13). However, when carA is mutated, CarS is not required for carB expression (15). These observations taken together have led to the following model for the light-regulated expression of P B . In the dark, CarA would prevent transcription from P B by an as yet unknown mechanism, and this transcriptional blockage would somehow be counteracted by CarS in the light. Derepression of P B is observed when CarS production is increased on illumination or when it is expressed from a constitutive heterologous promoter (13). Hence, the relative levels of CarA and CarS may be important for the latter to exert its antagonistic role. The interplay between CarS and CarA in regulating P B is further manifested by the identification of a gain-of-function mutation in carS (carS1) that leads to constitutive expression of the carB gene cluster (16).
The predicted amino acid sequence for CarS does not reveal any significant sequence homology to other known proteins nor does it suggest the presence of a defined DNA binding domain. By contrast, the amino acid sequence of CarA predicts an N-terminal stretch with high sequence homology to the helixturn-helix DNA-binding motif of the MerR family of gene regulators (17). These transcriptional factors, found in both Gramnegative and Gram-positive bacteria, regulate response to stress such as exposure to toxic compounds or oxygen radicals (18 -21). MerR, the prototypical member of the family, regulates expression of the Tn21 mercury resistance operon merT-P(C)AD that confers resistance to inorganic mercury (Hg(II)). MerR binds to the mer operator to function as a transcriptional repressor in the absence of Hg(II) and as an activator in its presence (reviewed in Ref. 18).
The molecular interplay between CarA and CarS in the regulation of the photoinducible carB promoter is the focus of the present study. We have purified the proteins and examined whether either one or both show specific DNA binding that could underlie their observed functional roles. CarA, but not CarS, was found to bind specifically in the region around P B . Moreover, our experimental data lead to the conclusion that CarS antagonizes CarA by preventing it from binding to its cognate DNA as well as by provoking the dissociation of preformed CarA-DNA complexes. CarA is thus a repressor protein, and CarS functions as its antirepressor partner. By using in vivo and in vitro studies to probe for protein-protein interactions, we demonstrate that the repressor and antirepressor interact physically. The functional relationship between the two regulatory proteins is then most likely bridged by the observed CarA-CarS physical interaction.

EXPERIMENTAL PROCEDURES
Bacterial and Yeast Strains, and Growth Conditions-M. xanthus strain DK1050 (22) was the wild-type strain used in this study. Strain MR844 is a derivative of DK1050 bearing a non-polar deletion within carA. 1 The rich casitone-tris (CTT) 2 medium was used for growth of M. xanthus cells (23). Escherichia coli strain DH5␣ was used for plasmid constructions, and strain BL21-(DE3) containing plasmid pLysS was used for protein overexpression. The recipient yeast strain for all yeast two-hybrid experiments was Saccharomyces cerevisiae EGY48 (24). Yeast growth conditions and media were as described elsewhere (25).
DNA Manipulations-Standard protocols were followed for DNA manipulation (26). Each PCR-derived clone was sequenced to verify the absence of any PCR-generated mutations. Detailed information on specific plasmid constructions is given below.
Construction of pMAR191 and pMAR192-A DNA fragment encompassing the carB promoter/operator region was PCR-amplified using as template DNA pMAR140 (which includes 1058 bp upstream of the translation initiation site of the first gene in the carB operon), and as primers the oligonucleotides proB1 (5Ј-CCTGCGATCCACGCCTTCAT-GAGG-3Ј) and proB2 (5Ј-CTTTCCTCCGAAGAACCCGTTCCTTT-GTTTCC-3Ј). PCR was performed using Pfu DNA polymerase to yield a 130-bp blunt-ended DNA fragment spanning positions Ϫ102 to ϩ28 relative to the transcription start point. The amplified product was ligated to EcoRI adapters, and the 5Ј ends were phosphorylated with T4 polynucleotide kinase after removing unbound adapters. The phosphorylated DNA fragments were then cloned into EcoRI-digested pMAR240, which carries a 1.5-kb DNA fragment of M. xanthus DNA, sufficiently long for plasmid integration by homologous recombination, and with no promoter activity (9). Restriction analysis was used to identify a plasmid with the 130-bp fragment inserted in the right (pMAR191) or wrong (pMAR192) orientation to produce a transcriptional fusion to the promoter-less lacZ gene lying downstream of the EcoRI site in pMAR240. pMAR191 and pMAR192 were introduced into M. xanthus by electroporation, and integration of the plasmid was selected for on CTT plates containing 40 g/ml kanamycin. Expression of the reporter lacZ gene under the control of the 130-bp fragment was qualitatively monitored on CTT plates containing 40 g/ml 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (X-gal) and assessed quantitatively by measurements of ␤-galactosidase activity as described previously (27).
Construction of Overexpressing Plasmids, Protein Overexpression, and Purification-The vector pET15b was used in constructs for overexpressing His 6 -tagged CarA, CarS, and CarS1 (28). DNA fragments encoding these proteins were obtained by PCR using M. xanthus genomic DNA as template and cloned into the NdeI-BamHI sites of the vector.
To overexpress proteins, cells cultured in 50 ml of LB/ampicillin at 37°C to an A 600 of 0.6 -1.0 were harvested by centrifugation, resuspended in 50 ml of fresh LB/ampicillin, and inoculated into 1 liter of the same medium. After growth to A 600 of 0.6 -0.8, protein expression was induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 or 4 h (all at 37°C). Expression and solubility of each protein were checked by SDS-PAGE of whole cell extracts or of the supernatant and pellet obtained by sonication and centrifugation of cells from 1-ml cell cultures. His 6 -CarS and His 6 -CarS1 were expressed as soluble proteins. His 6 -CarA was partly soluble at 50 -200 mM NaCl but insoluble at higher salt concentrations.
Cells from 1-liter of induced culture were pelleted and suspended in 80 ml of buffer A (50 mM Tris, 5% glycerol, 5 mM ␤-mercaptoethanol, pH 7.5) containing 1 M NaCl (unless otherwise stated) and 1 mM phenylmethylsulfonyl fluoride and benzamidine. Resuspended cells were lysed by sonication under ice-cold conditions and centrifuged (12,000 ϫ g, Beckman JA-20 rotor, 30 min, 4°C) to separate cell debris and the soluble fraction. Soluble His 6 -tagged protein was purified off TALON metal affinity resin following the accompanying native purification protocol at neutral pH, with imidazole elution and subsequent elimination by dialysis (CLONTECH, Palo Alto, CA). Native His 6 -CarS and His 6 -CarS1 purified in this fashion yielded Ն20 mg/liter cell culture. The His tag was effectively removed by thrombin cleavage (1:1000 molar ratio of thrombin:protein in 150 mM NaCl, 50 mM Tris, pH 7.5, 2 mM ␤-mercaptoethanol, 5 mM CaCl 2 , incubated overnight at 20°C) followed by dialysis. Some native CarA (ϳ50 g of protein/g cell pellet) could be similarly purified by resuspending the cell pellet in buffer A containing 50 mM NaCl and performing cell lysis and purification off TALON in this same buffer.
Purification of His 6 -CarA from the insoluble fraction was carried out at room temperature and denaturing solution conditions. Cells pelleted from a 1-liter culture were resuspended in 40 ml of binding buffer (500 mM NaCl, 20 mM Tris, pH 7.9, 5% glycerol, 2 mM ␤-mercaptoethanol) and sonicated. Inclusion bodies isolated by centrifugation at 20,000 ϫ g for 15 min were solubilized in binding buffer containing 8 M urea. Insoluble material was eliminated by centrifugation at 39,000 ϫ g for 20 min. The supernatant yielded ϳ2 mg of CarA per g of cell pellet when purified off TALON metal affinity resin under denaturing conditions. CarA was renaturated following protocols described by Burgess and Knuth (29): buffer A with 50 mM NaCl was added to dilute urea from 8 to 3 M and protein to Յ0.3 mg/ml. Sarkosyl (N-laurylsarcosine) was then added to 0.2% and the sample dialyzed against buffer A containing 50 mM NaCl. After eliminating any precipitate formed during dialysis by centrifugation, renatured protein was used immediately or stored at Ϫ20°C in 50% glycerol. To determine protein concentrations, absorbance at 280 nm and the following extinction coefficients, ⑀ 280 (M Ϫ1 cm Ϫ1 ), wild-type probes were prepared by PCR as follows. Primer pro B1 (see construction of pMAR191 and pMAR192 above) was labeled at its 5Ј end with [␥-32 P]ATP and T4 polynucleotide kinase and then added to a PCR mix containing the second unlabeled amplification primer. The radiolabeled PCR-amplified fragment was purified off a 2% low-melting agarose gel. 5Ј-Radiolabeled mutant probes were generated employing PCR site-directed mutagenesis by overlap extension (31). Binding was performed in 20-l reaction volume containing 100 mM KCl, 15 mM HEPES, 4 mM Tris, pH 7.9, 1 mM dithiothreitol, 10% glycerol, 200 ng/l bovine serum albumin, 1 g of sheared salmon sperm DNA as nonspecific competitor, 1.2 nM end-labeled double-stranded probe (ϳ13000 cpm), and the indicated amounts of proteins. After incubation at 20°C for 30 min, the samples were loaded onto 4% non-denaturing polyacrylamide gels (acrylamide:bisacrylamide 37.5:1) pre-run at 200 V, 10°C for 30 min in 0.5ϫ TBE buffer (45 mM Tris base, 45 mM boric acid, 1 mM EDTA), and electrophoresed for 1-1.5 h at 200 V, 10°C. Gels were vacuum-dried and analyzed by autoradiography. Experimental conditions for DNase I footprinting matched those used for the gel shift assays except that 10 mM MgCl 2 was included in the reaction mix. After the 30-min incubation at 20°C, the mix was treated with DNase I (0.07 units) for 2 min and then quenched with EDTA. DNA was ethanolprecipitated and run in 8 M urea, 8% polyacrylamide gels against G ϩ A and C ϩ T chemical sequencing ladders of the 130-bp fragment (25).
Yeast Two-hybrid Analysis-Yeast two-hybrid analyses were performed using the LexA-based system (24). N-terminal protein fusions to the LexA DNA-binding protein were constructed in plasmid pEG202, whereas those to the B42 transcriptional activation domain were in plasmid pJG4-5. Genes carA, carS, and carS1 were PCR-amplified using genomic DNA as template and cloned into EcoRI-XhoI doubledigested pEG202 or pJG4-5, and the respective constructs were designated pEG-X or pJG-X, X referring to the gene cloned. The recipient yeast strain EGY48 was transformed by electroporation or by the lithium acetate method. Prior to use in the analysis of protein-protein interactions, self-activation and entry into the nucleus of the LexA fusion proteins were tested. pEG202 and pJG4-5-based constructs were introduced in different pairwise combinations into EGY48 cells bearing pSH18-34. Cells containing all three plasmids were streaked on galactose plates supplemented with or without leucine, and interaction was assessed by monitoring expression of the two reporter genes as follows: (i) by analyzing growth on plates lacking leucine; (ii) by the development of blue color when plates were subjected to the X-gal overlay assay (32). Measurements of ␤-galactosidase activity were done as described previously (33).
Pull-down Assays, Size-exclusion Chromatography, and Analytical Ultracentrifugation-In pull-down assays, 50 l of TALON metal affinity resin in a 1.5-ml tube was washed twice with 500 l of binding buffer (50 mM NaCl, 20 mM Tris, pH 7.9, 5% glycerol, and 2 mM ␤-mercaptoethanol) by centrifuging at 700 ϫ g for 3 min and removing the supernatant. His 6 -CarA (50 l of 5 M protein stock) was then bound to the resin in separate tubes for 1 h at 25°C, after which unbound protein was removed by washing three times with the above binding buffer. Thrombin-cleaved CarS was added in excess (50 l of 35 M stock) to the resin-bound His 6 -tagged protein and to the resin alone (as control). After allowing to bind for 2 h (or overnight) at 25°C, unbound protein was again removed by washing three times with 500 l of binding buffer. Then the protein-bound resin was incubated for 30 min at 25°C with 100 l of binding buffer containing 200 mM imidazole. The supernatant recovered from the resin by centrifugation was then analyzed in a 15% SDS-PAGE gel.
An AKTA high performance liquid chromatography unit and a Superdex-200 (Amersham Biosciences) column equilibrated with 200 mM NaCl in buffer A containing 5% or 25% glycerol, and with or without 1 mM CHAPS, were used in size-exclusion experiments. 100 l of 5-50 M protein samples were injected, and the elution was tracked by absorbances at 280, 235, and 220 nm at flow rates of 0.2-0.4 ml/min. The column calibration using as standards (all from Sigma) cytochrome c (12.4 kDa), carbonic anhydrase (29 kDa), ovalbumin (43 kDa), bovine serum albumin (66 kDa), yeast alcohol dehydrogenase (150 kDa), and ␤-amylase (200 kDa), blue dextran (2 MDa; to determine void volume, V 0 ) and vitamin B 12 (to estimate total bed volume, V t ) yielded: log M r ϭ 7.91-0.23 V e (correlation coefficient Ն0.99) in buffer A, 200 mM NaCl. V e , the elution volume, was assigned for each peak after verifying its identity by SDS-PAGE and used to determine the molecular weight, M r .
A Beckman Optima XL-A analytical ultracentrifuge, and a Ti60 rotor with 6-sector Epon charcoal centerpieces of 12-mm optical path length, was used for sedimentation equilibrium measurements. 70-l samples (5-50 M protein) in 50 mM NaCl, buffer A were centrifuged at 13,000, 18,000, or 25,000 rpm and 20°C to equilibrium (verified when consec-utive radial scans acquired in 2-h intervals, and monitored at 230, 236, or 275 nm, were superimposable). Apparent weight average molecular weight (M w ) was determined by fitting data (program EQASSOC-Beckman) to the equation for an ideal solution containing a single species (34,35). Partial specific volumes, v , (in ml/g) calculated from the amino acid compositions (36), were 0.7314 for His 6 -CarS, 0.7368 for thrombincleaved CarS, 0.7345 for His 6 -CarS1, and 0.7326 for His 6 -CarA.

RESULTS
A 130-bp DNA Segment Encompassing the carB Promoter Includes All of the Cis-acting Elements Essential for Its Correctly Regulated Expression-A 1058-bp DNA segment upstream of orf1 has been shown to encompass all the cis-acting elements required for the correct in vivo expression and regulation of P B (17). In this study, we narrowed down this DNA segment to a shorter yet functionally competent length of 130 bp, spanning positions Ϫ102 to ϩ28 relative to the transcription start point. Features identified in this fragment are shown in Fig. 2A. Direct in vivo evidence that this 130-bp DNA fragment contains all the cis-acting elements required for the correct expression of P B was obtained as follows. Plasmid pMAR191, which contains the 130-bp fragment-lacZ transcriptional fusion (see "Experimental Procedures"), was introduced into M. xanthus wild-type strain DK1050 by electroporation. Chromosomal integration of the plasmid via homologous recombination was selected for by growth in the presence of kanamycin. On plates containing X-gal, DK1050-derived electroporants showed the light-inducible phenotype expected for P B . Quantitative analysis of ␤-galactosidase activity for darkand light-grown cultures of several of these electroporants provided results directly comparable with those reported previously with the longer 1058-bp stretch of DNA upstream of orf1 ( Fig. 2B; see Ref. 17). On the other hand, pMAR191 introduced into strain MR844, where part of the carA gene is deleted, yielded electroporants that showed the constitutive, light-independent expression of the lacZ reporter gene expected for carA lack-of-function mutants (Fig. 2C). As shown in Fig. 2, B and C, control electroporation experiments with pMAR192 (where the 130-bp fragment is fused to the lacZ gene in the opposite orientation relative to pMAR191) gave rise to electroporants expressing low levels of basal ␤-galactosidase activity, which remained the same irrespective of the genetic background (wild type or MR844) or growth conditions (dark or light). Thus, we conclude that the 130-bp DNA fragment includes all of the cis-acting elements essential for the correct regulation and expression of the carB promoter.
CarA Exhibits Specific DNA Binding at P B -The paringdown experiment discussed above provides a DNA fragment that is sufficiently short for use in DNA binding assays, yet fully functional in vivo. To verify binding of CarA to the carB promoter/operator region, purified His 6 -CarA was used in gel mobility shift assays using as probe the 130-bp DNA fragment (CCR, carB control region). His 6 -CarA was purified under native conditions or under denaturing conditions followed by renaturation (see "Experimental Procedures"). These and other proteins used in this work were Ն95% pure, and their mobilities in SDS-PAGE were those expected based on the molecular weights (Fig. 3A).
EMSA analysis of CarA binding to probe CCR was characterized by two retarded bands that appeared as a function of CarA concentration (Fig. 3B). Only the higher mobility species (lower band) was apparent at the lowest concentrations of CarA used in these assays (4 -8 nM; lanes 2 and 3, Fig. 3B). The lower mobility complex (upper band) appeared with increasing CarA concentration and became the predominant species at the highest concentrations of CarA used (Fig. 3B, lanes 6 and 7). This concentration-dependent appearance of the lower mobility band may reflect different modes of CarA binding to CCR, as a consequence of distinct binding sites on the DNA and cooperativity between these binding modes. It could also reflect specific DNA binding by higher oligomeric form(s) of CarA that could be increasingly populated as the protein concentration is raised. That the binding of CarA is probe-specific was demonstrated by the fact that addition of excess cold CCR probe effectively competed in EMSA. It may be noted that these results obtained with His 6 -CarA purified under native conditions were reproducible with purified, renatured His 6 -CarA. Hence, CarA manifests the DNA binding ability predicted from its sequence analysis, and moreover, it binds specifically within the region of the P B promoter shown to be essential in vivo.
Dissection of the CarA DNA-binding Site-The 130-bp CCR probe used above contains a palindromic DNA sequence upstream of the Ϫ35 region which, by analogy with MerR proteins, could be a potential binding site for CarA. It also includes two direct repeats (one overlapping with the palindrome and the other lying between the Ϫ35 and Ϫ10 regions, Fig. 2A), an arrangement reminiscent of the operator for the Bacillus subtilis DeoR repressor of the dra-nupC-pdp operon (37,38). Therefore, we analyzed in further detail CarA binding around P B .
We first examined DNA binding by EMSA analysis with a 64-bp probe corresponding to the segment Ϫ39 to Ϫ102. This probe contains the palindromic sequence and the direct repeat that overlaps with the 3Ј-half of the palindrome but lacks the other direct repeat located between the Ϫ10 and Ϫ35 regions. In striking contrast to the behavior observed for CarA binding to the longer 130-bp probe (Fig. 3B), only a single retarded band was observed with this probe for an equivalent range of protein concentrations (Fig. 4B, lanes 2-7, probe a). The observed differences in the gel-shift mobility pattern suggest that elements downstream of the 64-bp segment are necessary for the formation of both CarA-DNA complexes detected with the longer probe. The presence of two retarded bands may therefore be the consequence of increasing occupation of two possible binding sites on the DNA around P B . The single retarded band detected with the 64-bp probe could then be the result of CarA binding to the intact palindrome still present in this probe. To verify this possibility, we mutated either one or both of the palindrome half-sites. The intensity of the single retarded band  was considerably lowered when either one of the inverted repeats in the 64-bp probe was mutated (Fig. 4B, lanes 8 and 9,  probe b; lanes 10 and 11, probe c). No retarded band could be detected when both the inverted repeats were mutated (Fig.  4B, lanes 12 and 13, probe d). Thus, mutating either or both of the inverted repeats leads to a drastic reduction in the DNA binding affinity of CarA. The palindromic sequence is therefore a specific CarA DNA-binding site but, on its own, is not capable of promoting the formation of the two CarA-DNA complexes that could be observed with the 130-bp DNA probe. In other words, the participation of additional downstream elements may also be important. This inference is supported by the following observation: mutations in both halves of the palindrome in the 64-bp probe d that led to undetectable DNA binding (Fig. 4B, lanes 12 and 13) resulted in two retarded bands in the 130-bp probe e (Fig. 4C, lanes 5 and 6), but of considerably lower intensity relative to wt, the 130-bp CCR probe (Fig. 4C, lanes 2 and 3).
CarA-DNA binding was further analyzed by DNase I footprinting of probes wt and a labeled at the upper strand. Binding of CarA to the 64-bp probe a protected positions Ϫ70 to Ϫ41, spanning the inverted repeats, the DNA between them, and 6 bases flanking each side of the palindrome (Fig. 4D, lanes 8 and  9). Two hypersensitive sites were also observed, one lying at the 5Ј-end of the left inverted repeat (position Ϫ63) and the other between the two half-sites (position Ϫ55). These results additionally support the inference that the palindrome constitutes a CarA-binding site and define the footprint features that characterize its occupation by CarA. To determine whether CarA is capable of occupying additional sites downstream of the palindrome, as suggested by EMSA, we performed DNase I footprinting with the 130-bp probe wt (Fig. 4D, lanes 1-7). With increasing concentrations of CarA, a footprint that extended beyond that observed with the 64-bp probe became apparent.
At the highest CarA concentration used (where the low mobility species predominates in EMSA), at least an additional 22 bp (positions Ϫ42 to Ϫ19) were protected. Included in the expanded footprint are the Ϫ35 promoter element and part of the 3Ј-direct repeat. In sum, the DNase I footprinting results reinforce our conclusion that the low mobility species in EMSA corresponds to CarA bound to the palindrome and to additional downstream elements.
CarS Shows no DNA Binding in Vitro but Abolishes Specific CarA-DNA Binding at P B -Having established that CarA exhibits specific DNA binding to probe CCR, we next determined whether, under similar experimental conditions, this applied also to CarS. As shown in Fig. 5 (lane 2), we did not detect any specific binding of CarS to probe CCR even at CarS concentrations over 2 orders of magnitude greater than those at which specific binding of CarA to CCR could be observed. These results strongly suggest that CarS cannot exert its antagonistic role by directly competing with CarA for binding to the region surrounding and including P B . CarS, nevertheless, could accomplish its function through the inactivation of CarA so that it no longer binds DNA. Fig. 5 shows an order-of-addition gel shift assay performed with probe CCR at a fixed concentration of CarA (60 nM) and in the absence or presence of increasing concentrations of CarS (0.4 -11 M). It may be noted that the protein concentrations are expressed in terms of the monomer, and the protein stocks are assumed to be fully active. It is evident from Fig. 5 3 with lanes 4 -8 and 9 -13). When CarA and CarS were added to the reaction simultaneously, about a 15-fold excess of CarS relative to CarA was more than sufficient to completely abrogate CarA DNA binding (lanes 4 -8). However, when CarA was first allowed to bind CCR and CarS subsequently added, even a 180-fold excess of CarS could not completely disrupt the  1-7) and 64-bp probe a (lanes 8  and 9). CarA concentrations are indicated. Lanes G ϩ A and C ϩ T are chemical sequencing ladders of the 130-bp fragment. Protection against DNase I is shown by solid lines (left side, probe wt; right side, probe a) and DNase I-hypersensitive sites by arrowheads on the right. On the left, the palindrome is shown by oppositely facing arrows, and the direct repeats by the two arrows pointing in the same direction.
CarA-DNA complexes (lanes 9 -13). Interestingly, before complete neutralization of CarA-DNA binding by CarS was achieved, CarS promoted a shift in the relative distribution of the two CarA-retarded species, from the lower mobility species to the higher mobility one (compare lane 3 with lanes 4 or 9 -13). This effect was particularly obvious in those reactions where CarS was added to pre-formed CarA-CCR complexes, where the lower mobility band was "converted" into the high mobility species before DNA binding was completely eliminated at higher CarS concentrations. Hence, it appears that CarS is more effective against the formation of the lower mobility complex. CarS therefore acts as an antirepressor by preventing free CarA from binding to its cognate site and, less efficiently, by "disrupting" pre-established CarA-DNA complexes. As we have also demonstrated, this antagonistic role of CarS does not involve any CarS-DNA binding.
CarA-CarS Physical Interactions Mediate the Functional Interplay between Them-Possible CarA-CarS interactions were scored in vivo using the yeast LexA-based two-hybrid system (24). In this system, the N terminus of one of the protein pair is fused to the LexA DNA-binding domain (the "bait"), whereas the N terminus of the other protein is fused to the B42 activation domain (the "prey"). Expression of the prey protein is controlled by the GAL1 promoter, which is repressed by glucose and strongly activated by galactose. When both fusion proteins are expressed in yeast strain EGY48 bearing plasmid pSH18-34 (which contains the lacZ gene), physical interaction between the bait and prey results in activation of the reporter genes LEU2 and lacZ. We found that yeast cells producing the LexA-CarA and B42-CarS fusion proteins were able to develop colonies on galactose plates lacking leucine; moreover, the colonies acquired an intense blue color 30 -60 min after the plates were overlaid with X-gal. By contrast, control cells producing only the LexA-CarA fusion protein were unable to grow on plates lacking leucine, and colonies grown on plates containing leucine remained white even 24 h after addition of X-gal. These effects are illustrated by measurements of the level of expression of the lacZ reporter gene as shown in Fig. 6A, which demonstrates galactose-dependent induction of lacZ expression and the absence of any such effect in the control cells. Considering the values of ␤-galactosidase activity attained after just a 2-h induction in galactose, the data indicate that a strong physical interaction exists between CarA and CarS. The reverse experiment, where CarS was fused to the LexA protein and CarA to the activation domain, rendered qualitatively similar results. However, the LexA-CarS construct did not fully satisfy the criteria to pass the self-activation and entry into the nucleus controls. Consequently, we did not proceed with an actual quantitative estimation in this case.
We further probed the interaction between CarA and CarS using the purified proteins in pull-down assays. In these assays, interactions are probed by tethering one of the proteins to a solid matrix, and then checking its ability to pull down a possible interacting partner that is incapable of binding to the matrix. Given that the His 6 tag in purified His 6 -CarS could be completely cleaved off by thrombin (this was less efficient with His 6 -CarA), we examined the ability of TALON-bound His 6 -CarA to pull down CarS lacking its His tag ("CarS") (see "Experimental Procedures"). We observed that TALON-bound His 6 -CarA was capable of pulling down "CarS" in amounts sufficient to be detected in Coomassie-stained SDS-PAGE (Fig.  6B, lane 3) relative to a control of CarS passed through TALON resin alone (Fig. 6B, lane 1). This demonstrates that the two proteins do interact physically and with significant strength, as suggested by the yeast two-hybrid analysis. Moreover, these data also suggest that CarS by itself does not proteolyze CarA, because there was no loss of CarA in the course of the experiment. An ϳ1:1 mix of His 6 -CarA and CarS compares well with the relative intensities of the two proteins in the pull-down assay (compare lanes 3 and 4, Fig. 6B). This suggests, but does not prove on its own, that the interaction may occur with this stoichiometry, because most of the unbound proteins in the pull-down assay are expected to be removed in the repeated washes.
CarA Interacts with Itself whereas CarS Is a Monomer-CarA may form oligomers, given the sequence characteristics of its DNA-binding site, and the observation in gel mobility shift assays of two retarded bands whose relative intensities varied with protein concentration. We therefore investigated whether CarA interacts with itself by using the yeast two-hybrid system. The results summarized in Fig. 7A show this to be the case; yeast cells producing the LexA-CarA and B42-CarA fusion proteins were able to develop colonies on galactose (but not glucose) plates lacking leucine, and the colonies acquired an intense blue color when incubated for at least 6 h after the X-gal overlay. This was not the case with control cells producing only the LexA-CarA fusion protein. Significant levels of LacZ accumulation in cells with the LexA-CarA/B42-CarA fusion constructs required overnight induction (Ն12 h), in contrast to cells with LexA-CarA/B42-CarS fusion constructs where much shorter times (2 h) were sufficient. This suggests that CarA-CarA interactions exist but may be weaker than those between CarA and CarS.
A biophysical characterization of the oligomeric state of CarA was attempted using size-exclusion high performance liquid chromatography and analytical ultracentrifugation. This, however, has not been possible thus far because of the loss of material (signal) observed at the micromolar concentrations required and used in these experiments. Size-exclusion experiments in the presence of a mild detergent (1 mM CHAPS) and the use of a lower pH (6.5, about two units less than the theoretical pI) or the presence of a higher glycerol concentration (25%) have proved unsuccessful so far. We have no clear explanation for this other than that CarA may have a relatively low solubility. On the other hand, His 6 -CarS and its thrombincleaved product CarS were characterized by both these biophysical methods. Size-exclusion analysis indicated that both these proteins are predominantly monomeric (Fig. 7B); the apparent molecular masses (in kDa) of His 6 -CarS and CarS of 16.6 and 15.5, respectively, compared well with the corresponding calculated values of 14.4 and 12.5 (and confirmed by mass spectrometry). This was also confirmed in analytical ultracen-trifugation carried out at rotor speeds of 18,000 and 25,000 rpm (shown for His 6 -CarS in Fig. 7C). Here the weight average molecular mass (in kDa) obtained by fitting the equilibrium radial distribution to the equation for a single ideal species was (20 Ϯ 1) for His 6 -CarS and (16 Ϯ 1) for CarS, and the residuals of the fits (small and randomly scattered) were indicative of a single species. Thus, on current evidence, CarA is a dimer or higher order oligomer, whereas CarS is monomeric.
Protein-Protein Interactions Involving Truncated CarA and CarS-In a pilot attempt to localize the regions of CarS and CarA involved in the interactions, we examined truncated forms of each protein. A gain-of-function mutation in carS (carS1) has been identified which provokes light-independent expression from the normally light-inducible P B promoter (16). The carS1 gene product, CarS1, is a truncated form of CarS lacking the last 25 amino acids (13). CarS1 was purified as native His 6 -tagged protein and was found to be monomeric by gel filtration (apparent molecular mass of 15.4 kDa relative to the calculated value of 11.5 kDa, Fig. 7B) as well as by analytical ultracentrifugation (weight average molecular weight of 17 Ϯ 1 kDa). His 6 -CarS1, like CarS, was capable of abolishing the specific DNA binding of CarA (data not shown). We also found that CarS1 matched CarS in physically interacting with CarA when probed by the yeast two-hybrid analysis (see Table  I). These results demonstrate that the CarA-binding domain maps to the first 86 N-terminal residues of CarS.
Because we had observed that CarS was less effective against pre-formed complexes of CarA and the CCR probe, we reasoned that the same region(s) of CarA could be involved in the specific binding to DNA as well as to CarS. So we tested whether CarA truncated to its first 78 N-terminal residues, CarA(Nter), was involved in any protein-protein interactions. This fragment of CarA was chosen because the homologous stretch in MerR proteins includes the helix-turn-helix motif and the two "wings" implicated in DNA binding (39). We found that yeast cells producing the LexA-CarA(Nter) and B42-CarS fusion proteins developed colonies on galactose plates lacking leucine that acquired an intense blue color 30 -60 min after the plates were overlaid with X-gal. This was not seen with control cells producing only the LexA-CarA(Nter) fusion protein. These results parallel those described earlier for CarA-CarS interactions. Yeast two-hybrid analysis also indicated that CarA(Nter) does not interact with CarA (Table I). However, LexA fused to residues 80 -288 of CarA, CarA(Cter), interacted with the B42-CarA fusion but not with the B42-CarS fusion. These results suggest that the protein domains involved in physical interactions between CarA and CarS are localized to the first 78 and 86 N-terminal residues of the two proteins, respectively. CarA regions required for interactions with itself are located within the last 209 C-terminal residues (residues 80 -288) and so are distinct from those involved in interactions with CarS. DISCUSSION The Specific DNA Binding of CarA at P B -Our results show that CarA acts by specifically binding to DNA in the region around the promoter, and this is antagonized by CarS through direct physical interaction with CarA. CarA and CarS thus constitute a repressor-antirepressor pair. Given the nature of the binding sites, and that CarA is an oligomer, it is reasonable to infer that the specific CarA-DNA binding must involve at least dimers. The two distinct types of specific CarA-DNA complexes that we observe with increasing protein concentration could then be attributed to a lower and a higher order oligomeric form of CarA. Or it may be that an increasing number of sites on the DNA are being occupied as the protein concentration is raised. Our EMSA and DNase I footprint analyses indicate that a palindrome upstream of the Ϫ35 promoter region is involved in binding to CarA. However, our data suggest that additional elements downstream of the palindrome are also involved. A conspicuous feature of the DNA used in our EMSA analysis is that its sequence also includes two direct repeats, one of which overlaps with the 3Ј-half of the palindrome, and the other is located between the Ϫ10 and Ϫ35 promoter elements. A similar arrangement of a palindrome and two direct repeats that occurs in the promoter region of the dra-nupC-pdp operon in B. subtilis has been shown to constitute the operator for the octameric DeoR repressor (37,38).
We propose that binding to the palindrome may serve as a beacon for a more effective homing-in of CarA to additional site(s). This could account for the two distinct types of specific CarA-DNA complexes observed as the concentration of CarA increases. This proposal does not exclude the possibility that different oligomeric states of CarA may also be involved. If prior binding of CarA to the palindrome then fosters binding to the direct repeats, this would provide a simple and effective mechanism for the repression of transcription. Because one of the direct repeats lies in the spacer between the Ϫ35 and Ϫ10 regions, its complex with CarA could block promoter access to RNA polymerase to repress transcription in the manner of most classical repressors (40). DNase I footprinting does show protection by CarA of the Ϫ35 promoter element that extends to at least the 5Ј end of the downstream direct repeat. Because this occurs at CarA concentrations where the lower mobility complex predominates, it is attractive to speculate that this would be the functionally operative CarA-DNA complex in vivo. Remarkably, the lower mobility complex is also the one that is more easily dismantled by CarS. A detailed analysis of these proposals is currently being pursued.
Possible Mode of CarS-mediated Antirepression of CarA-Since the involvement of two contrasting elements in transcriptional regulation was first suggested by Oppenheim et al. (41), several antirepressor-repressor systems have been reported. Distinct mechanisms for how the antirepressor antagonizes repressor activity include the following: (a) direct protein-protein association without any DNA binding by the antirepressor (42)(43)(44)(45); (b) exclusion of the repressor by DNA binding of the antirepressor (46); and (c) proteolysis of the repressor promoted by the antirepressor (47). Our data have revealed that the monomeric CarS, which does not itself bind DNA, physically interacts with CarA. We also have no evidence for any CarS-  The straight line is the calibration curve using the molecular weight standards, ϩ, indicated under "Experimental Procedures." C, analytical ultracentrifugation of His 6 -CarS. The observed radial distribution (black dots) was fit to the equation for a single ideal species (shown by the line) and yields an apparent molecular mass, the fitted parameter, of 20 Ϯ 1 kDa. The residuals of the fitting are displayed at the top. mediated degradation of CarA. On a per molecule basis, CarA-CarS binding may involve a 1:1 stoichiometry. This, as well as the stoichiometry of CarA-DNA binding, needs to be corroborated by additional experiments currently underway. We observe that CarS relieves DNA binding by CarA, being more effective in abolishing the lower mobility CarA-DNA complex. When simultaneously added with CarA, CarS lowers the effective concentration of CarA available for DNA binding. Preformed CarA-DNA complexes are more refractory to the action of CarS, suggesting that CarS interacts more readily with CarA that is free in solution. Disruption of the pre-bound CarA-DNA complex by CarS would then be dictated by the kinetics of dissociation of the complex and the subsequent trapping of freed CarA by CarS. Thus, the primary mechanism for CarSmediated inactivation of CarA appears to be in binding to and blocking its DNA-binding domain. Consistent with this are a number of other observations. CarS is acidic (theoretical pI ϭ 4.76) and could conceivably be an effective competitor for DNAbinding regions on CarA. Significantly, CarS1 is even more acidic (theoretical pI ϭ 4.09) than CarS; of the 8 Arg in CarS (which also has 2 Lys), 6 are located in the C-terminal stretch of 25 residues that is absent in CarS1. The carS1 phenotype described earlier could then be rationalized in terms of the greater affinity for CarA of the more negatively charged CarS1. Finally, a 78-residue N-terminal segment of CarA containing its putative DNA-binding site (but not the remaining 209residue C-terminal stretch) physically interacts with CarS. A more detailed analysis of the CarA-CarS-interacting regions is beyond the scope of the present study and would, among other things, be aided by a knowledge of the three-dimensional structures of the proteins involved.