Structural Basis for DNA Recognition by the Two-Component Response Regulator RcsB

ABSTRACT RcsB is a highly conserved transcription regulator of the Rcs phosphorelay system, a complex two-component signal transduction system (N. Majdalani and S. Gottesman, Annu Rev Microbiol 59:379–405, 2005; A. J. Wolfe, Curr Opin Microbiol 13:204–209, 2010, https://doi.org/10.1016/j.mib.2010.01.002; D. J. Clarke, Future Microbiol 5:1173–1184, 2010, https://doi.org/10.2217/fmb.10.83). RcsB plays an important role in virulence and pathogenicity in human hosts by regulating biofilm formation. RcsB can regulate transcription alone or together with its auxiliary transcription regulators by forming heterodimers. This complexity allows RcsB to regulate transcription of more than 600 bacterial genes in response to different stresses (D. Wang et al., Mol Plant Microbe Interact 25:6–17, 2012, https://doi.org/10.1094/MPMI-08-11-0207). Despite increasing knowledge of RcsB importance, molecular mechanisms that drive the ability of RcsB to control transcription of a large number of genes remain unclear. Here, we present crystal structures of unphosphorylated RcsB in complex with the consensus DNA-binding sequence of 22-mer (DNA22) and 18-mer (DNA18) of the flhDC operon from Escherichia coli determined at 3.15- and 3.37-Å resolution, respectively. The results of our structural analysis combined with the results of in vitro binding assays provide valuable insights to the protein regulatory mechanism, demonstrate how RcsB recognizes target DNA sequences, and reveal a unique oligomeric state that allows RcsB to form homo- and heterodimers. This information will help us understand the complex mechanisms of transcriptional regulation by RcsB in bacteria.

T wo-component signal transduction (TCST) systems play important roles in regulating many bacterial processes, primarily in response to altered cellular environments (1). The Rcs phosphorelay system is one of the most studied bacterial TCST systems; it is also one of the most complex (see Fig. S1A in the supplemental material). At its core are three multidomain proteins, two hybrid histidine kinases (RcsC and RcsD) and a response regulator (RcsB). Phosphotransfer within the Rcs phosphorelay system is induced mainly by damage to the cell envelope (2). Phosphorylation of RcsB is thought to induce a functionally active form that facilitates its ability to regulate gene expression (3). The complexity of this phosphorelay is extended by RcsB's ability to associate with other transcription regulators and form protein complexes, in some of which RcsB is phosphorylated and in some it is not (4)(5)(6)(7)(8)(9). For example, phosphorylated RcsB can partner with an unstable RcsA transcription regulator to bind to a DNA sequence (termed the RcsAB box) located in the vicinity of promoters for several genes, including the cps operon, which encodes components involved in capsular polysaccharide biosynthesis (4). A similar sequence is located near the promoter for the flhDC operon, which encodes the master regulator of flagellar biogenesis (5,10). In its phosphorylation-independent state, RcsB can interact with various transcription regulators, including GadE, BglJ, RmpA, PhoP, MatA, and RflM (6)(7)(8)(9). This complexity allows RcsB to regulate diverse bacterial characteristics and cellular processes, many of which affect pathogenesis in humans (3,11,12).
The lack of a crystal structure has been a major obstacle in understanding (i) the molecular mechanisms by which RcsB activates transcription from a wide range of promoters, (ii) the formation of phosphorylation-dependent and phosphorylationindependent protein complexes, and (iii) the mechanisms of RcsB dimerization and DNA binding. RcsB belongs to the FixJ/NarL family of response regulators with two domains connected by a long linker (Fig. S1B). The C-terminal DNA-binding domain (helix-turn-helix [HTH]) of RcsB has a four-helix bundle fold with a classical helix-turnhelix DNA recognition motif (13,14). The N-terminal REC (receiver) domain of RcsB is composed of a five-stranded central ␤-sheet surrounded by five ␣-helices and contains the conserved aspartyl residue (D56) that serves as the phosphoryl acceptor (15). In vivo analysis shows that D56 phosphorylation is essential for activation of transcription from those promoters that require the RcsB homodimer or the RcsB-RcsA heterodimer (10,16). Previously, the structures of full-length RcsB or its complex with DNA site had not been obtained. Instead, structures of the isolated RcsB domains had been determined (14,15).
In vitro binding analysis of RcsB-DNA interactions. To understand RcsB function and particularly its mechanisms for DNA recognition, we first determined requirements for RcsB-DNA complex formation in vitro and then determined its cocrystal structures. We performed gel mobility assays (electrophoretic mobility shift assay [EMSA]), which confirmed that RcsB specifically binds the RcsAB box from the flhDC promoter, in the presence or absence of phosphodonors, either carbamoyl phosphate (CPh) or phosphoramidate (PA) (Fig. 1A). RcsB bound DNA in the absence of a phosphoryl donor, but the complex appeared to have reduced stability. The presence of phosphoryl donors appeared to increase binding affinity, as determined by titration EMSA with different RcsB concentrations (Fig. S2). To test whether RcsB forms a dimer in solution or upon binding to DNA and to determine whether phosphorylation affects dimer formation, we performed size exclusion chromatography with multiangle light scattering (SEC-MALS) analysis. In the absence of DNA, RcsB existed predominantly as a monomer in solution even if the phosphoryl donor CPh was present (Fig. 1B). Thus, in the absence of DNA, either CPh does not phosphorylate the RcsB monomer, or phosphorylation does not favor RcsB dimerization. In the presence of DNA, however, RcsB bound DNA and formed a dimer with or without CPh (Fig. 1C). SEC-MALS analysis also confirmed that the unphosphorylated RcsB-DNA complex is less stable and prone to dissociation (Fig. 1C). Based on the increased apparent molecular mass (MM) of the complex, our data suggest that the presence of DNA enhances RcsB phosphorylation and favors dimerization. Surface plasmon resonance (SPR) analysis provided evidence that phosphorylation increases binding affinity of RcsB for its DNA site ( Fig. 1D and E). The RcsB in the presence of a phosphodonor binds more tightly (equilibrium dissociation constant [K D ] of~0.16 M) than in the absence of CPh (K D of~5.34 M).
Structure of the RcsB-DNA complex. To understand the structural basis for DNA recognition by RcsB, we crystallized unphosphorylated RcsB in the presence of DNA that contains the RcsB binding site from the flhDC promoter region of Escherichia coli (Fig. 1F). This binding site can bind both the RcsB homodimer and the RcsA-RcsB heterodimer (5,17,18). We determined two crystal structures of full-length RcsB bound to 22-mer (DNA22) or 18-mer (DNA18) of the flhDC operon from E. coli ( Fig. 2A and Fig. S3). For details of crystallization, data collection, and structure determination, see Text S1 and Table S1 in the supplemental material. The asymmetric unit of the RcsB crystal in complex with DNA22 contained one RcsB dimer bound to one double helix DNA ( Fig. 2A), whereas the RcsB crystal in complex with DNA18 contained an RcsB tetramer (two RcsB dimers) bound to two parallel double helical DNAs (Fig. S3). In both crystal forms, the RcsB dimers displayed similar structures, and the DNA had a classical B form with an average base pair rise and twist of 3.2 Å and 34°, respectively; the DNA was slightly bent along the surfaces of the RcsB HTH domains. The structural similarity between RcsB dimers bound to DNA observed in two different crystal forms suggests that this dimer is likely biologically relevant.
Structure of the RcsB homodimer bound to DNA. The RcsB dimer observed in two different crystal forms is asymmetric ( Fig. 2A and Fig. S3). Subunit A (or C) was bound to the first DNA half-site, whereas subunit B (or D) was bound to the second half-site. The subunits in the dimer were not related by a two-fold rotation axis of symmetry. Instead, an approximate two-fold axis was present between each of the RcsB domains of a dimer. However, these two axes met each other at an angle of about 120°. In the RcsB dimer, the domains superimpose with an average root mean square deviation (RMSD) value of 0.9 Å (for 129 C ␣ atoms) and 0.6 Å (for 57 C ␣ atoms) for the REC and HTH domains, respectively. The main differences were found in the position of the partially disordered linker region between helices ␣5 of the REC domain and ␣7 of the HTH domain. The structure of the RcsB asymmetric dimer is unique because it does not correspond to any known structure of full-length TCST response regulators despite the fact that dimer asymmetry of some TCST response regulators in their DNA-bound complexes has been reported previously. Examples include structures of the quorum- Structural Basis for DNA Recognition by RcsB ® sensing transcriptional factor TraR from the LuxR family in the presence of its ligand, the pheromone N-3-oxooctanoyl-L-homoserine lactone (OOHL) (19), and of the response regulators PhoP and KdpE from the OmpR/PhoP family determined in their unphosphorylated state (20,21). In each case, the asymmetry defines the regulator's distinct dimeric structure, its interactions with DNA, and thus, the response regulator's individual functions. The ability of two-component response regulators, including RcsB, PhoP, and KdpE, to bind to DNA in their unphosphorylated state might not be a coincidence and could be important for these proteins' regulatory activities.
RcsB-DNA contacts. In the RcsB-DNA complex, the helices that form the HTH DNA recognition motif of the DNA-binding domain (␣8 and ␣9) bound to the floor of the DNA major groove perpendicular to the DNA helical axis ( Fig. 2A and Fig. S3A). The RcsB-DNA interactions were largely identical at all sites in the structure of RcsB in complex with either DNA22 or DNA18 (Fig. 2B). In these structures, two amino acid residues of RcsB made direct sequence-specific contacts with only three bases of the DNA. K180 and S184, located on helix ␣9, made hydrogen bonds (H-bonds) with G -8 , A Ϫ9 and A Ϫ6 in one-half-site (TTAGGAA) or with G Ϫ30 , A Ϫ31 , and A Ϫ28 in its complement (TCTAAGA) on the other half-site of the DNA (Fig. 2C). Most H-bonds occurred  between phosphates of the DNA backbone and side chains of several residues of the RcsB HTH domain (S152, K154, T169, R177, T181, S183, Q185, K186, and K187). Some main chain atoms from V168, T169, and S178 contacted phosphates of the DNA backbone. The side chain amino group of K192 in the RcsB-DNA22 complex contributes to electrostatic interactions, but it is too far away for a direct H-bond (Fig. 2C). The resolution of the crystal structures did not allow detection of any water-mediated interactions. Biochemical and physiological experiments have shown that K180 is a critical residue for RcsB-dependent regulation of flhDC (22). N-lysine acetylation, which can negate the positive charges of K154 and K180, is reported to reduce RcsB's ability to bind the flhDC DNA site (22,23). Our structures support the importance of both and suggest that the small number of DNA sequence-specific contacts with DNA bases likely permits RcsB to function on multiple promoters.
Intersubunit domain and dimer interfaces. In the asymmetric dimer formed upon binding to DNA ( Fig. 2A; Fig. S4A and Fig. S4B), both RcsB domains contributed to the dimer interface, and the subunits in the dimer were stabilized primarily through hydrophobic contacts. In contrast, the linker region between the REC and HTH domains ran along the side of each subunit in both structures and was not involved in dimerization. Very few H-bonds and no salt bridges or water molecules were found in the dimerization interface. Interestingly, within the RcsB tetramer of the RcsB-DNA18 complex, dimers formed asymmetric interfaces between their REC domains ( Fig. S3B and Fig. S4B). We cannot exclude the existence of a higher-MM RcsB-DNA complex, as indicated by the slowest-migrating band (Fig. 1A). This putative complex could correspond to an RcsB tetramer bound to two DNAs as seen in our RcsB-DNA18 cocrystal structure. Further analysis is needed to confirm this supposition. In all crystallized dimers, the HTH domains interacted through helix ␣10 and the loop between helices ␣7 and ␣8 ( Fig. 2A and Fig. S4A). The orientation and binding mode of RcsB HTH domains resemble those seen in the structure of the isolated DNA-binding domain of the NarL regulator in complex with DNA (24). The REC domains in the RcsB dimer interacted through helix ␣1 and the loops between ␤1 and ␣1 and between ␤5 and ␣5 ( Fig. 2A and Fig. S4B). Previously, we reported that isolated RcsB REC domains could form a similar dimer in the crystal and characterized the dimerization interface (15).
The subunits of the RcsB dimer had different conformations and interdomain interfaces that involve a distinct set of contacts ( Fig. 2A and Fig. S4C). The area of surface contacts between the HTH and REC domain were two times larger in subunit B (D) than in subunit A (C). In subunit A (C), helices ␣4 and ␣6 of the REC domain interacted with helices ␣9 and ␣10 of the HTH domain. In contrast, in subunit B (D), helix ␣4 and the ␤4-␣3 loop of the REC domain interacted with helices ␣7 and ␣8 of the HTH domain. The unusual asymmetric oligomeric state(s) of RcsB could present different surfaces for interactions with RNA polymerase (RNAP) and thus extend RcsB's capacity to regulate genes with a variety of promoter architectures. At the same time, this unique asymmetry and the presence of hydrophobic interactions between RcsB subunits may allow unphosphorylated RcsB to form heterodimers with auxiliary transcription regulators (4-9). For example, the transcription regulator RflM in Salmonella enterica serovar Typhimurium increases the binding affinity of unphosphorylated RcsB for the homologous RcsB DNA-binding site located on the flhDC operon (9) and therefore can form an analogous asymmetric dimer. Undoubtedly, our structures will provide important details for investigating interactions between RcsB and auxiliary transcription factors.
Only a small number of full-length FixJ/NarL response regulators, including NarL, VraR, and ChrA, have been structurally characterized (24)(25)(26)(27); however, the structure of a full-length response regulator from the FixJ/NarL family in its DNA-bound state has not been determined previously. Structural comparisons of full-length FixJ/NarL regulators using the Dali server (28) show that the RcsB asymmetric dimer and/or RcsB subunits do not correspond to any existing structure despite the fact that the structural fold and dimerization interfaces of the isolated RcsB HTH and REC domains are consistent with structures of homologous transcription factors (15,(24)(25)(26)(27). The relative orientation of isolated domains, conformation of the interdomain linker, and oligomerization of subunits in the determined RcsB structures are different compared to structures of homologous proteins. These differences likely contribute to RcsB's function.
Alternative mechanism of transcriptional regulation by RcsB. One of the features of our crystal structures is that RcsB forms an asymmetric dimer that is identical in two different crystal forms. In one structure, unphosphorylated RcsB exists as a dimer bound to one double-stranded DNA; in another, two RcsB dimers form a tetramer bound to two parallel double-stranded DNAs. These structures combined with in vitro binding assays reveal that unphosphorylated RcsB can form a dimer upon binding to DNA and emphasize that the presence of a phosphoryl donor in the presence of DNA favors dimerization and enhances RcsB binding affinity. It is widely accepted that phosphorylation facilitates formation of the RcsB dimer, which can then bind its DNA site and thus regulate transcription. Under environmental conditions that favor phosphorylation of RcsB either by its cognates RcsC/RcsD or by the small molecule phosphoryl donor acetylphosphate, formation of the phosphorylated dimer could be favored prior to binding its DNA site. However, because RcsB in its unphosphorylated state can bind to DNA and can activate expression of some genes (29), we hypothesize that the RcsB unphosphorylated state is important for function and that an alternate mechanism may also exist. In this alternative, unphosphorylated RcsB binds DNA as observed in the structures, albeit as an unstable RcsB-DNA complex that can be stabilized by phosphorylation of RcsB or through interactions with its cognates RcsC and/or RcsD, as observed for the OmpR regulator (30). We further hypothesize that phosphorylation would induce conformation changes in the receiver domain, propagating structural rearrangements within intersubunit interfaces of the RcsB homodimer that subsequently lead to stabilization of RcsB-DNA interactions. Since the unphosphorylated complex is less stable, reversible phosphorylation may provide an extra opportunity to regulate recruitment and stabilization of RNA polymerase and thus transcription. In this case, phosphorylation of RcsB could regulate the level of gene expression. In its lower-affinity unphosphorylated state, RcsB could also easily dissociate from DNA or form heterodimers with other transcription regulators. This mechanism combined with asymmetry of the RcsB dimer and low specificity to DNA-binding sites may explain RcsB's ability to regulate a diverse set of genes. At the same time, we anticipate that phosphorylation may enhance formation of a different RcsB dimer or another RcsB oligomer prior to or upon binding to DNA. Therefore, future experiments are needed to explore these possibilities.
Conclusions. Here, we described the crystal structures of unphosphorylated fulllength RcsB from E. coli in complex with one of its DNA sites. Our structural data support the hypothesis that an RcsB dimer forms upon binding to DNA. The data emphasize that phosphorylation enhances RcsB binding affinity and presumably facilitates interaction with RNAP and thus activates transcription. The structure yields the first view of the RcsB homodimer in its unphosphorylated state bound to the DNA and provides an explanation for how RcsB could regulate large numbers of genes. The RcsB-DNA complex structures provide insights into the structural basis of the mechanism(s) of transcriptional regulation by RcsB and offer a platform for design of novel antimicrobial compounds against Gram-negative pathogens.
Methods. (i) Gel mobility shift assays. DNA strands of 22 or 18 bp (as defined in Text S1 in the supplemental material) 5= end labeled with IRDye 700 or not labeled were annealed according to the general methodology for infrared electrophoretic mobility shift assay (LI-COR Biosciences). DNA (4.15 nM) was then incubated in 1ϫ binding buffer (10 mM Tris, 40 mM KCl, 1 mM dithiothreitol [DTT] [pH 7.5], 10 mM MgCl 2 , 5% glycerol, and 0.1 g/l of bovine serum albumin) in the presence or absence of 169 nM purified RcsB in a total volume of 20 l for 20 min at room temperature. The samples (10 l per well) were loaded onto a 7% native acrylamide/bisacrylamide (29:1) 1ϫ Tris-glycine-EDTA (TGE) gel and ran in 1ϫ TGE buffer for 40 min in a cold room. After electrophoresis, the gel was scanned on an Odyssey imager (LI-COR Biosciences Inc.) For competition reactions, 100-fold excess of unlabeled DNA22 or DNA18 or mutated DNA22 or DNA18 (as defined in Text S1) in the binding buffer (described above in a total volume of 20 l) was added along with labeled DNA (415 nM). For phosphorylation, phosphoramidate (PA) or carbamoyl phosphate (CPh) was added to a final concentration of 20 mM to the binding reaction mixture and incubated for 30 min at 37°C in a heating block prior to the addition of the DNA. For titration experiments, increasing concentrations of RcsB in the binding reaction mixtures were used (as given in the legend to Fig. S2), and 10 l of each reaction mixture was loaded onto a 8% native gel. The conditions of the binding reaction and electrophoresis were kept the same. All assays were repeated four times to confirm reproducibility.
(ii) Size exclusion chromatography with multiangle light scattering. The molecular weights of RcsB (with and without CPh), as well as their respective complexes with DNA22 and DNA18 fragments used in cocrystallization experiments, were determined by conducting size exclusion chromatography with multiangle light scattering (SEC-MALS) experiments using Agilent 1260 high-performance liquid chromatography (HPLC) system (Agilent Technologies) equipped with a Dawn HeleosII 18-angle MALS detector, Optilab T-rEX (refractometer with extended range) refractive index detector, WyattQELS quasielastic (dynamic) light scattering (QELS) detector, and ASTRA software (all four from Wyatt Technology Europe GmbH). All runs with DNA and/or unphosphorylated RcsB were conducted in 10 mM Tris-HCl buffer at pH 8.3 with 150 mM NaCl and 0.5 mM Tris(2-carboxyethyl)phosphine (TCEP) on a Superdex 75 10/300 GL column (GE Healthcare) preequilibrated with the same buffer at a flow rate of 0.5 ml/min at 22°C. Runs with phosphorylated RcsB were conducted in 10 mM Tris-HCl buffer at pH 8.3 with 150 mM NaCl, 10 mM MgCl 2 , and 0.5 mM TCEP. To phosphorylate RcsB, 50 mM CPh and 10 mM MgCl 2 were added to the protein solution prior to experiments. A total of 270 l of protein solution containing RcsB (both in the presence and absence of phosphodonor) and respective RcsB complexes with DNA22 and DNA18 was injected onto the column. The final concentration of RcsB and DNA22 or DNA18 used in experiments was 1 mg/ml. A refractive index increment (dn/dc) value of 0.1600 was used for DNA samples, and a value of 0.178 was used for proteins and their complexes with DNA. Lysozyme from chicken egg white (Sigma-Aldrich Corp.) was used as a control.
(iii) Surface plasmon resonance. Binding of RcsB (with and without CPh) to DNA22 and DNA18 fragments used in cocrystallization trials was measured using a fourchannel Reichert surface plasmon resonance (SPR) instrument (Reichert Technologies). Biotinylated DNA22 or DNA18 (0.5 mg/ml) was captured onto NeutrAvidin planar mSAM surface (catalog no. 13206065; Reichert Technologies, Depew, NY) using highaffinity binding of biotin to avidin (at a flow rate of 33 l min Ϫ1 for 3 min) until a signal increase of approximately 600 micro-refractive index units (RIUs) was achieved. The biotinylated DNA fragments were synthesized by IDT Inc. The protein, DNA, and running buffer solutions used in all experiments contained 10 mM Tris-HCl (pH 8.3) with 150 mM NaCl and 0.5 mM TCEP. Phosphorylation of RcsB was achieved by adding 50 mM CPh and 10 mM MgCl 2 into the protein solution. All SPR experiments were performed at 25°C. RcsB (0.25, 0.5, 1, and 1.5 M) and RcsB phosphorylated with carbamoyl phosphate (0.125, 0.25, 0.375, and 0.5 M) were injected over the chip at a flow rate of 30 l min Ϫ1 for 2.5 min. Dissociation was observed for 5 min, and regeneration of the chip was carried out by a 1.5-min injection of 1 M Tris-HCl (pH 4.0) at 40 l min Ϫ1 . Binding was detected as a change in the refractive index at the surface of the chip, as measured by response units (RIUs). A reference flow channel was used to record the background response, and background was subtracted from each sample injection. Equilibrium dissociation constant (K D ) values were calculated as ratios of association rate (k a )/dissociation rate (k d ) determined from kinetic experiments. Each experiment included duplicates of each solute concentration obtained at two channels.
Data models were fit using TraceDrawer data analysis software available through Reichert Technologies.
Data accessibility. The atomic coordinates and structure factors for RcsB-DNA22 and RcsB-DNA18 have been deposited into the Protein Data Bank (http://www.rcsb.org) with accession numbers or codes 5W43 and 5VXN, respectively. The diffraction images (target identifier [ID] IDP91817) are available at the CSGID website (http://www.csgid .org/csgid/pages/home).