Domains formed within the N-terminal region of the quorum-sensing activator TraR are required for transcriptional activation and direct interaction with RpoA from agrobacterium.

TraR, a quorum-sensing activator, induces transcription from its binding site, the tra-box, located upstream of Ti plasmid target promoters. TraR activated expression of a lacZ reporter in Escherichia coli only when RpoAAt from Agrobacterium tumefaciens was co-expressed. As assessed by gel retardation assays RpoAAt, but not RpoAEc, formed a ternary complex with TraR and a tra-box probe in vitro. TraR formed similar ternary complexes with alphaCTDAt but not with NTDAt, the C- and N-terminal segments of RpoAAt. As measured by surface plasmon resonance refractometry, TraR interacted directly with RpoAAt with an affinity about five times greater than that observed for its interaction with RpoAEc. The activator interacted with alphaCTDAt with kinetics and affinities similar to those of the full-sized -subunit. Positive control (PC) mutations at Asp-10 and Gly-123 of TraR did not affect DNA binding but greatly decreased the TraR-RpoAAt interaction. These two residues combine to form two patches on the activator, one of which may be involved in interaction with RpoA. When co-expressed, mutants of TraR with substitutions at Asp-10 complementing mutants with substitutions at Gly-123 for gene activation in an allele-specific manner. Co-expression studies with TraR and its PC mutants, and also with complementary PC alleles of TraR, coupled with three-dimensional structure are consistent with a hypothesis that both Asp-10/Gly-123 patches are required for activator function.

TraR activates expression of the three operons of the tra regulon on the Ti plasmids of Agrobacterium tumefaciens (1)(2)(3) in response to the population density of the donors (3,4). The activator, a member of the LuxR family of quorum-sensing regulators, binds an 18-bp inverted repeat, the tra-box, located upstream of its target promoters (3,5). pTiC58 contains three such boxes, one located between two divergently oriented operons, traAFB and traCDG (6), one in the promoter region of the trb operon (2) and one located upstream of the repABC operon (7,8). TraR activates expression of each of these four gene sets, thereby controlling conjugation and plasmid copy number (4,8,9).
Most members of the LuxR family are activators and require as ligands acyl-homoserine lactone (acyl-HSL) 1 signals (reviewed in Ref. 10). The acyl-HSLs are produced by the bacteria themselves, and the cells gauge their population size by sensing the accumulation of these signals in the environment in which they are growing (11). Ligand binding is required for TraR, LuxR, and LasR to form homodimers (12)(13)(14), a process required for binding DNA (5,15,16). However, other members of the family, including CarR and EsaR, multimerize and bind DNA even in the absence of their acyl-HSL signals (17,18).
Although much is known about the regulatory circuits controlled by quorum sensing, less is known about the mechanisms by which these activators initiate transcription. Purified homodimers of TraR and LuxR bind target promoters containing appropriate binding sites and, with RNA polymerase (RNAP), activate transcription in vitro (16,19). According to the crystal structure of TraR, each protomer of the dimeric activator is composed of two domains joined by a 12-residue flexible linker (20,21). The N-terminal 162 residue segment contains the signal binding site and the primary dimerization domain, whereas the C-terminal 60-residue segment contains secondary dimerization regions and a helix-turn-helix domain, one helix of which makes contacts with specific nucleotides in a tra-box half-site (14,20,21).
Genetic, physiological, and biochemical data suggest that LuxR is an ambidextrous activator and, as such, interacts with the ␣-subunit (22)(23)(24) and 70 (25) of Escherichia coli RNAP holoenzyme. Given its dimeric structure and the location of the tra-box recognition site within target promoters, TraR also likely makes contacts with one or more components of RNAP. Consistent with this hypothesis, alterations at two residues of TraR, Asp-10 and Gly-123, yield a strong positive control (PC) phenotype; both mutant proteins retain tra-box binding activity as measured by a genetic assay but fail to activate transcription (5). Asp-10 and Gly-123 both are located N-terminally to the linker region of the protein. Moreover, because of the asymmetric alignment and the folding patterns of the N-terminal regions of the two protomers, Asp-10 of one subunit combines with Gly-123 of the other to form two composite surface patches (Fig. 1). One such patch is close to the bound DNA, whereas the other is located at the opposite extremity of the dimer. Either or both of these patches could make contact with one or more components of RNAP.
Although TraR binds to the tra-box in E. coli, it does not activate transcription in the enteric host. VirG, another Ti plasmid-encoded activator, shows a similar defect but will activate a VirG-dependent promoter in E. coli when co-expressed with rpoA from A. tumefaciens (rpoA At ) (26,27). This observation suggests that VirG does not properly interact with RNAP containing the enteric form of RpoA (RpoA Ec ). In this report we show that when co-expressed with RpoA At , TraR will activate expression from a tra-box promoter in E. coli. We also show that TraR interacts with RpoA At in vitro through the C-terminal domain of the ␣-subunit. The Asp-10 or Gly-123 PC mutants bind DNA in vitro but do not interact with the ␣-subunit. In vivo activation was restored by co-expressing certain alleles of TraR containing different substitutions at the Asp-10 and Gly-123 positions in the same cell. However, activation could not be rescued by co-expressing wild-type TraR with any of these mutants or by expressing an allele of TraR containing otherwise complementing substitutions at both locations.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Culture Conditions-A. tumefaciens strain NTL4 (28), a Ti plasmid-cured derivative of the nopaline/agrocinopinetype strain C58, was used in all studies. E. coli strain DH5␣ was used as the host in all cloning experiments, as well as in assays of gene expression, and E. coli strains BL21(DE3)(pLysS) and SG13009 (26) were used for expressing proteins prior to purification. E. coli strains were grown in LB liquid medium, and A. tumefaciens was grown in LB or in MG/L (29) liquid medium. For studies involving defined media, A. tumefaciens was grown in AB minimal medium containing mannitol as sole source of carbon (ABM; Ref. 30) and E. coli was grown in A medium containing glucose as sole carbon source (5). Cultures of A. tumefaciens were grown at 28°C, and E. coli was grown at 28 or at 37°C. All cultures in liquid medium were grown with shaking to ensure adequate aeration. Antibiotics, 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-Gal) and IPTG were included in medium at concentrations described previously (6). N-(3-Oxo-octanoyl)-L-homoserine lactone (3-oxo-C8-HSL; AAI), synthesized as described previously (31), was added to cultures at the concentrations indicated in the text.
Plasmid Isolation and Genetic Methods-Plasmids were isolated by miniprep techniques (6) and introduced into A. tumefaciens and E. coli strains by electroporation (29) or by biparental matings using E. coli strain S17-1 (6).
Plasmid Constructions-Plasmids that express RpoA of Agrobacterium (RpoA At ) were constructed as follows. The rpoA At gene was amplified by PCR from pPS1.3 (26). The amplicons were cloned into pZLQ (5) to produce pZLQRpoA, which expresses wild-type protein, and into pET14b to produce pET14RpoA, which expresses an N-terminal His 6 -tagged version of RpoA At . Plasmid pKKTR2-I41 was constructed by cloning HindIII fragment 4 of pTiC58 containing the TraR-dependent traG::lacZ reporter fusion from pH4I41 (32) into the HindIII site of pKKTR2-I, which contains the traR gene under control of the trc promoter and lacI q (33). This plasmid and pH4I41 were used to monitor activation by TraR and its mutants in E. coli and A. tumefaciens hosts. The repression reporter pPBL1 was used to genetically monitor the DNA binding properties of TraR and its mutants as described previously (5). Plasmid pQKK was constructed by replacing the HindIII-EcoRI region of the polylinker of pKK38 (34) with the HindIII-EcoRI polylinker fragment from pUC18. This substitution removed the trc promoter from pKK38 and yielded a vector that lacks a promoter upstream of the polylinker. Wild-type traR or its single amino acid substitution mutants were amplified by PCR from pZLQ derivatives containing the appropriate genes. The amplicons, which included the upstream trc promoter of pZLQ, were cloned as SacI fragments into the pQKK polylinker. The traG::lacZ reporter was transferred from pH4I41 as a HindIII fragment into the pQKK-traR derivatives yielding plasmids that express traR or its mutants and also a TraR-dependent reporter fusion.
Nucleotides encoding the N-terminal (residues 1-166) and C-terminal (residues 167-336) segments of RpoA At were produced as fragments generated by PCR using Pfu polymerase and pZLQRpoA as template. The N-terminal end of each fragment was constructed to contain an NdeI site. The coding regions were cloned as NdeI-EcoRI fragments into pZLQ and pET14b to generate pZLQRpoAN (␣NTD), pZLQRpoAC (␣CTD), pET14RpoAC (His 6 -␣CTD), and pET14RpoAN (His 6 -␣NTD). Clonings at the pZLQ NdeI site generated an ATG initiation codon that is properly spaced with respect to the vector transcriptional and translational signals to ensure high levels of expression of the protein.
Mutagenesis-Site-specific mutations were introduced into the traR gene using the QuikChange kit (Stratagene), and the resulting DNA samples were transformed into E. coli. Plasmid DNA samples from randomly chosen isolates were subjected to DNA sequence analysis, a representative containing the desired substitution mutation was retained, and the fragment bearing the mutant allele was recloned into the appropriate expression vector. The complete double-stranded sequence was determined for all mutant clones using automated methods by the Keck Center for Biotechnology at the University of Illinois at Urbana-Champaign.
Western Analyses-The stability and expression levels of mutant TraR proteins in vivo were determined by Western analysis using anti-TraR antiserum as described previously (12).
Overexpression and Purification of Proteins-His-tagged RpoA At, as well as His-tagged forms of ␣CTD At and ␣NTD At were over-expressed in E. coli strain BL21(DE3)(pLysS) as described previously (12). N-terminal His 6 -RpoA Ec from E. coli was over-expressed from E. coli strain SG13009(pREP4) as described by Lohrke et al. (26). The His-tagged forms of RpoA At and RpoA Ec remained soluble in cell extracts and were purified under nondenaturing conditions. The two subfragments of RpoA At were insoluble and were solubilized by treatment with 6 M urea. In both cases the proteins were purified by Nickel affinity chromatog- raphy using a 6-mm (id) ϫ 16 cm column as described by the manufacturer (Novagen). Following purification, His 6 -␣CTD At and His 6 -␣NTD At were renatured by gradual removal of the urea using the protocol described by Novagen.
As assessed by SDS-PAGE all proteins were more than 95% pure. All proteins were stored at Ϫ20°C in TEDNG50 buffer (20 mM Tris-HCl, pH 7.9, 1 mM EDTA, 1 mM dithiothreitol, 0.15 M NaCl, 50% glycerol). Protein concentrations were determined using the Coomassie Plus protein assay kit (Pierce).
Gel Retardation Assays-Gel retardation assays were performed as described by Luo et al. (33). A 251-bp probe containing the entire TraR-dependent divergent tra promoter system from pTiC58 (6) was prepared by PCR amplification using pH4I41 as the template. The probe was labeled at its 3Ј-ends with digoxigenin-11-dideoxyUTP using terminal transferase according to protocols provided by the supplier (Roche Applied Science). The concentrations of TraR as well as the two His-tagged RpoA proteins or the His-tagged N-and C-terminal fragments of RpoA At used in the reactions are described in the text or in the appropriate figure legends.
Surface Plasmon Resonance Refractometry-Surface plasmon resonance sensograms were determined using a Biacore 3000 optical biosensor. Reactions were conducted at 25°C at a flow rate of 20 l/min. Interaction experiments were initiated by injecting a preparation of purified native TraR or one of its substitution mutants in EB at concentrations as noted under "Results" into the buffer stream of both flow cells 1 and 2. Buffer flow was continued, and sensograms were recorded for a period up to 15 min, allowing analysis of association and disassociation reactions. The reference sensor signal from flow cell 1 was used to correct by subtraction the experimental binding data observed in the reaction occurring in flow cell 2. Following completion, chips were regenerated by injecting the following solutions in succession into the buffer stream: two 10-l volumes of 0.3 M imidazole, two 10-l volumes of RB, and a single 10-l volume of 0.3% SDS. Both flow cells of the chip were then saturated with NB in preparation for the next run.
Values for kinetic parameters including the association rate constant, k a , and the disassociation rate constant, k d , were determined using the BIAevaluation 3.1 program (Biacore). The data exhibited best fit using the simple 1:1 Langmuir binding model available in the program. Deviations between experimental and fitted data normally were less than Ϯ 2 resonance units.

RESULTS
TraR Requires Agrobacterium RpoA to Initiate Transcription in E. coli-In E. coli TraR binds to the tra-box sequence but does not initiate transcription from a native TraR-dependent promoter (5). However, TraR strongly activates transcription of the same promoter in A. tumefaciens. Moreover, purified TraR, when combined with RNAP from A. tumefaciens, activates transcription from tra-box promoters in vitro (19). These observations suggest that TraR cannot form an activation complex with enteric RNAP. Given that some activators interact with the ␣-subunit of RNAP (reviewed in Ref. 35), we tested whether expression of rpoA At would restore the activator function to TraR in an E. coli host. As expected, TraR failed to activate the traG::lacZ reporter in E. coli DH5␣ (Table I). However, when rpoA At was co-expressed with traR, the reporter was induced 6-fold in the enteric host (Table I), and activation was depend- a traR was expressed from pKKTR2-I41, which also codes for the traG::lacZ reporter. b rpoA from A. tumefaciens and from E. coli, as well as the NTD and CTD fragments of RpoA At , were expressed from derivatives of pZLQ constructed as described under "Experimental Procedures." c Expressed as units of ␤-galactosidase activity/10 9 cfu as described under "Experimental Procedures." The strains were assayed three times, and results from a representative experiment are shown.  (Table I).
TraR and RpoA Form a Complex With the tra-Box Promoter Region-We determined whether RpoA At and RpoA Ec , its ortholog from E. coli, could interact with the TraR-promoter DNA complex using gel retardation assays. Purified TraR formed detectable complexes when mixed with a 251-bp fragment containing the tra-box and all of the cis-acting promoter elements located in the intergenic region between the traAFB and traCDG operons from pTiC58 (Fig. 2, A and B). On its own, purified N-terminal His 6 -RpoA from either bacterium did not detectably bind to the DNA probe (Fig. 2, B and C). However, although the addition of His 6 -RpoA Ec to the mixture of TraR and probe DNA had no effect on the mobility of the TraR-DNA complex (Fig. 2C), the addition of His 6 -RpoA At yielded a new, slower migrating complex, with the intensity of the new band increasing with increasing amounts of the ␣-subunit (Fig. 2B). We take this to mean that RpoA At , but not the ␣-subunit from E. coli, interacts with the TraR-promoter complex to form a ternary structure.
TraR Interacts Directly with RpoA from Agrobacterium-We used surface plasmon resonance refractometry (SPRR) to assess whether TraR can bind RpoA in the absence of DNA. The sensograms clearly indicated that TraR in the mobile phase interacted with His 6 -RpoA subunits from both A. tumefaciens and E. coli (Fig. 3, A and B). However, the association rate constant, k a , for the interaction between TraR and His 6 -RpoA At was almost 4-fold greater than that calculated for the interaction between TraR and His 6 -RpoA Ec (Table II). Following completion of the binding reaction, the sensograms indicated that both TraR-RpoA complexes dissociated at a slow but measur-able rate (Fig. 3, A and B). Moreover, the dissociation rate constants, k d , for the two TraR-RpoA complexes were not significantly different (Table II).
TraR Interacts with the C-Terminal but Not the N-Terminal Domain of RpoA At -Several well studied activators interact with the C-terminal half of the ␣-subunit (␣CTD) (35). Using PCR, we constructed His-tagged expression clones of the ␣CTD and ␣NTD domains of RpoA At . Strains of E. coli co-expressing either of these two constructs along with traR failed to activate the reporter (Table I). When tested by gel retardation and SPRR, the His 6 -␣NTD fragment did not interact with TraR in either assay (Fig. 4A and data not shown). However, the His 6 -␣CTD fragment altered the mobility of the TraR-DNA complex in the gel retardation assay (Fig. 4A) and, as judged by SPRR, interacted directly with TraR (Fig. 4B). The rate constants of the binding reactions with His 6 -␣CTD approximated those calculated for the interaction between TraR and full-sized RpoA At (Table II).  Positive Control Mutants of TraR Bind DNA but Are Defective in Their Interactions with RpoA At -We previously described three PC mutants of TraR, each of which represses gene expression but no longer activates transcription in vivo (5). Each mutant contains a single amino acid substitution, Asp-10 3 Asn (D10N), Gly-123 3 Arg (G123R), or Gly-123 3 Glu (G123E), at a position located N-terminal to the linker region of the protein.
TraRD10N and TraRG123R purified from cells grown with the acyl-HSL signal bound the 251-bp tra-box probe in a manner indistinguishable from that of wild-type TraR (Fig. 5). Interaction with the promoter probe validates the repressionbased genetic screen that we developed to assess the DNAbinding properties of TraR and its PC mutants (5). Although a mixture of wild-type TraR and His 6 -RpoA At formed a ternary complex with probe DNA, no such complex was detected in reactions containing probe DNA, either of the two mutant TraR proteins, and His 6 -RpoA At at any concentration tested (Fig. 5). When analyzed by SPRR, the two PC mutants were not bound significantly by the immobilized His 6 -RpoA At (data not shown).
Alanine Substitutions At Asp-10 and Gly-123 of TraR Result In a Positive Control Phenotype-We constructed alanine substitution mutations at Asp-10 and Gly-123 to determine whether the PC phenotype of the original mutants is specific to the substituted residues. As assessed by Western analysis, each mutant protein is as stable in vivo as wild-type TraR (data not shown). Each mutant was tested in vivo for DNA binding using the repressor assay in E. coli and for transcriptional activation of the TraR-dependent reporter fusion in A. tumefaciens, all in cells grown with or without 3-oxo-C8-HSL. Wildtype TraR activated the traG::lacZ reporter and repressed the chimeric tra-box::lacZ reporter, whereas the original D10N and G123R PC mutants (5) failed to activate transcription but retained strong repressor activity (Table III), all in a quormonedependent manner (data not shown). The D10A and G123A mutants also exhibited PC properties; both strongly repressed the tra-box::lacZ reporter but did not significantly activate the traG::lacZ reporter (Table III). When co-expressed with TraR, like the original PC mutants, TraRD10A and TraRG123A exhibited dominant negativity; both strongly inhibited activation of the reporter by the wild-type activator (Table IV).
When Co-expressed the Asp-10 and Gly-123 Positive Control Mutants Complement Each Other-Because Asp-10 of one protomer and Gly-123 of the other combine with each other to form two surface patches on the TraR dimer ( Fig. 1; Refs. 20 and 21), we tested whether combinations of the two PC mutants could complement one another when co-expressed in the same cell. Strains expressing either allele of traR alone did not significantly activate the reporter fusion (Table V). However, when co-expressed TraRD10A and TraRG123A cooperated to strongly activate transcription of the reporter (Table V). Similarly, a combination of TraRD10A and TraRG123R strongly activated transcription, whereas co-expressing TraRD10N with TraRG123A yielded only weak activity. When co-expressed, TraRD10N and TraRG123R failed to activate the reporter. In all cases activation was dependent on growth of the cultures with the acyl-HSL signal.
Because certain alleles of the two PC mutants complement one another, we constructed and tested a derivative of TraR with complementing substitution mutations at both positions. The double mutant TraRD10A/G123R failed to activate the traG::lacZ reporter (Table III). However, the double mutant retained repressor activity, and repression required the acyl-HSL (Table III). When co-expressed with wild-type TraR, the double mutant exerted strong dominant negativity (Table IV). , or 560 nM (lane 13) were subjected to electrophoresis; protein-DNA complexes were detected as described under "Experimental Procedures." I is free probe, II is the TraR-probe complex, and III is the TraR-probe-RpoA complex.

DISCUSSION
Based on the SPRR studies we conclude that TraR, like CAP, FNR, and Mor (36), associates directly with RpoA by binding to the C-terminal end of the ␣-subunit. These constitute the first studies using purified proteins to show that a quorum-sensing transcription factor directly contacts a component of RNAP.
LuxR also most likely contacts RpoA; physiological and genetic studies have shown that this transcription factor requires RpoA that is wild type in its C-terminal domain, to activate transcription (23,24).
Two lines of evidence suggest that TraR and RpoA interact at a tra-box promoter region and that this interaction is required for RNAP to initiate transcription. First, TraR only activates transcription in E. coli when RpoA At is co-expressed (Table I). This observation also suggests that whereas TraR interacts with RpoA At , the activator does not productively interact with RpoA Ec , a conclusion that is consistent with the binding kinetics observed between TraR and the RpoA subunits from A. tumefaciens and E. coli (Table II). Second, in the absence of TraR, the RpoA subunits from A. tumefaciens and E. coli do not detectably bind to the probe DNA (Fig. 2B). However, combining RpoA At with the activator and probe DNA yielded a ternary complex. Significantly, such a complex was not detected when RpoA Ec was added to the reaction (Fig. 2B). These results suggest that TraR assists in the establishment or maintenance of an association between RpoA and the promoter.
That TraR exhibits specificity for the ␣-subunit of A. tumefaciens is intriguing but not unexpected. VirG, the response regulator of the two-component system that controls the Ti plasmid vir regulon, also activates transcription in E. coli only when co-expressed with rpoA At (26,27). These observations suggest that the ␣-subunits of A. tumefaciens, an ␣-proteobacterium, and E. coli, a ␥-proteobacterium, although functionally equivalent, differ in their activator recognition properties. Based on studies with activators in E. coli, three domains of RpoA, the 265, the 261, and the 287 determinants, are important for activation of transcription by CAP (reviewed in Ref. 36). The critical residues in the 265 and 261 determinants, which apparently are involved in DNA binding and UP element recognition and interaction with 70 , respectively, are strongly conserved between the ␣-subunits from the two bacteria. How- a traR and its mutant alleles were expressed from derivatives of pZLQ constructed as described under "Experimental Procedures." b Activation by TraR and its mutant alleles was assessed in A. tumefaciens NTL4 harboring the reporter plasmid pH4I41 and the pZLQ derivative expressing the traR gene grown in AB minimal medium containing mannitol, IPTG, and 3-oxo-C8-HSL at a final concentration of 25 nM.
c Expressed as units of ␤-galactosidase activity/10 9 cfu as described under "Experimental Procedures." Each strain was tested a minimum of three times, and the results shown are from a representative experiment.
d -Fold activation is calculated as ␤-galactosidase activity in the test strain minus ␤-galactosidase activity in strain lacking TraR divided by ␤-galactosidase activity in strain lacking TraR. e Repression by TraR and its mutant alleles was assessed in E. coli DH5␣ harboring the reporter plasmid pPBL1 and the pZLQ derivative expressing the traR gene grown in A medium containing glucose, IPTG, and 3-oxo-C8-HSL at a final concentration of 25 nM.
f -Fold repression is calculated as ␤-galactosidase activity in the strain lacking TraR divided by ␤-galactosidase activity in the tested strain.
g This strain harbors a derivative of pZLQ coding for the double mutant traRD10A-G123R and pH4I41, which supplies the traG::lacZ reporter. a The second copy of traR or one of its mutant alleles as indicated was expressed from derivatives of pZLQ constructed as described under "Experimental Procedures." b Expressed as units of ␤-galactosidase activity/10 9 cfu as described under "Experimental Procedures." Each strain was tested a minimum of two times, and the results shown are from a representative experiment.
c AAI was added to a final concentration of 25 nM at the beginning of growth.
d -Fold inhibition is calculated as the ␤-galactosidase activity in the strain expressing only wild-type traR divided by the ␤-galactosidase activity in the strain expressing both wild-type and mutant traR. e NA, not applicable. f This strain harbors a derivative of pZLQ coding for traRD10A and a derivative of pQKK-I41 coding for traRD10N, which also provides the TraR-dependent traG::lacZ reporter.
g This strain harbors a derivative of pZLQ coding for traRG123R and a derivative of pQKK-I41 coding for traG123A, which also provides the TraR-dependent traG::lacZ reporter. ever, the 287 determinants of the two orthologs, which in RpoA Ec mediates interaction with CAP, differ at four of the eight critical residues. Significantly, two of these residues of RpoA Ec , Val-287 and Arg-317 (corresponding to Ala-286 and Glu-316 of RpoA At ), are among the six residues of this region that make direct contacts with CAP (37). We speculate that these divergent residues define specificity differences between the ␣-subunits of A. tumefaciens and E. coli. Consistent with this interpretation, the eight residues of the 287 determinant of RpoA At , including those four that vary in RpoA Ec , are virtually invariant in the ␣-subunits of all of the other ␣-proteobacteria for which DNA sequence information is available.
Clearly Asp-10 and Gly-123 are required for activation by TraR. These two residues are surface-exposed (Fig. 1), and, because of the asymmetric folding pattern and the antiparallel monomer-monomer interactions of the N-terminal domains, the Asp-10 of one protomer is adjacent to the Gly-123 residue of the second in the active, ligand-bound dimer ( Fig. 1 and Refs. 20 and 21). Thus, the two residues combine to form two surface patches, one located close to the DNA-binding domain and the second some distance away from the promoter binding site (Fig.  1). We propose that at least one of these patches interacts with the C-terminal domain of RpoA At . Consistent with this hypothesis, substitution mutations at either residue abolish activation without affecting DNA binding, and neither mutant dimer interacts significantly with RpoA At (Fig. 5 and data not shown).
Neither Asp-10 nor Gly-123 of TraR is absolutely conserved within the LuxR family. However, these two residues are conserved in all alleles of TraR known to be active. This is of little significance among TraR proteins from Ti and opine-catabolic plasmids; the seven known active alleles share almost 90% amino acid identity over their entire lengths. However, when the active forms of TraR from Ri plasmids and from several large plasmids of rhizobial origin are included in the alignment only 46 residues, Asp-10 and Gly-123 among them, are strictly conserved among all of the proteins (data not shown). This conservation at Asp-10 and Gly-123 coupled with the very strong conservation of the 287 determinant of the RpoA subunits from the host bacteria of these traR genes supports an important role for the patches formed by these two residues in transcriptional activation by TraR.
The Asp-10 and Gly-123 mutants by themselves all are inactive and exert strong dominant negativity when co-expressed with wild-type TraR (Tables III and IV). However, co-expressing the D10A mutant with either the G123A or G123R mutant restored activation in vivo (Table V). These two observations together suggest that both of the Asp-10/Gly-123 patches are required for activation by TraR. Such inter-protomer comple-mentation is allele-specific; the D10A-G123A and D10A-G123R pairs yield active TraR, whereas the D10N-G123R pair is inactive (Table V and Fig. 6). However, when two otherwise complementing substitutions are in cis, the double mutant homodimer exhibits a PC phenotype, retaining repressor activity but exhibiting no significant activator function (Table III). We suggest that D10A, but not D10N, trans-suppresses the G123A and G123R mutants, thereby forming active heterodimers. This conclusion is consistent with the observation that the D10A-G123R double mutant fails to activate transcription but exerts strong dominant negativity (Tables III and IV), suggesting that at least one patch must be wild-type whereas the other must retain some function by allelic complementation (Fig. 6).
Although our data clearly show that TraR interacts with RpoA At , the positioning of the tra-box upstream and adjacent to the Ϫ35 promoter element suggests that this transcription factor is an ambidextrous activator such as CAP (36) and interacts with additional components of RNAP. Given that our genetic analysis points to a requirement for both Asp-10/Gly-123 patches for activation, it is conceivable that one interacts with the ␣-subunit whereas the second interacts with another subunit of the polymerase. Consistent with such an interaction, we have recently observed that co-expression of the A. tumefaciens 70 in addition to RpoA At in E. coli allows TraR to activate transcription from a tra-box promoter to levels comparable with that observed in Agrobacterium. 2 Similarly, genetic analyses suggest that LuxR interacts with 70 of RNAP in E. coli (25), an observation consistent with the recent report that purified LuxR stabilizes the interaction of RNAP holoenzyme with the lux promoter in vitro (16).