Mutational analysis of dimeric linkers in peri- and cytoplasmic domains of histidine kinase DctB reveals their functional roles in signal transduction

Membrane-associated histidine kinases (HKs) in two-component systems respond to environmental stimuli by autophosphorylation and phospho-transfer. HK typically contains a periplasmic sensor domain that regulates the cytoplasmic kinase domain through a conserved cytoplasmic linker. How signal is transduced from the ligand-binding site across the membrane barrier remains unclear. Here, we analyse two linker regions of a typical HK, DctB. One region connects the first transmembrane helix with the periplasmic Per-ARNT-Sim domains, while the other one connects the second transmembrane helix with the cytoplasmic kinase domains. We identify a leucine residue in the first linker region to be essential for the signal transduction and for maintaining the delicate balance of the dimeric interface, which is key to its activities. We also show that the other linker, belonging to the S-helix coiled-coil family, plays essential roles in signal transduction inside the cell. Furthermore, by combining mutations with opposing activities in the two regions, we show that these two signalling transduction elements are integrated to produce a combined effect on the final activity of DctB.

ATP to phosphorylate a conserved histidine residue on the DHp domain [8]. Subsequently, this phosphoryl group is transferred to the N-terminal asparagine residue of the RR, normally resulting in the activation of the RR [4,9,10]. Many RRs are transcription factors, regulating the transcription of relative genes in response to environmental signals [7].
In HK, signal perception normally occurs on its periplasmic sensor domain [11]. The Per-ARNT-Sim (PAS) domain is identified most frequently as the signal perception domain in HKs. The SMART database contains over 44 000 PAS domains in more than 29 000 proteins (December 2013), spanning all kingdoms of life. Most known PAS domains contain a PAS core and flanking regions [12]. The PAS core has a wellconserved tertiary structure of three to four a-helices flanking a central five-stranded anti-parallel b-sheet (figure 1b), which is essential for the ligand binding [12]. The flanking region usually adopts a-helical conformations [12][13][14][15] and serves as a linker between the PAS core and other parts within the larger protein [12]. It was also proposed that the signals originated upon ligand binding to the PAS core would be propagated through the flanking region of the PAS domain to other parts of the protein [12,16]. However, it is unclear through what mechanism the flanking region could transduce the signal.
Immediately following the transmembrane helix is a cytoplasmic linker region that is proposed to play key roles in transducing signals to activate the kinase domain. About three-quarters of all TCS HKs contain helical linkers in the cytoplasmic portion [17]. A conserved HAMP linker has been identified in some TCS HKs [18]. Structural and mutational analysis suggests an essential role of the HAMP linker in signal transduction from the environmental stimuli to the cytoplasmic kinase domains [11,[19][20][21][22][23][24]. Recently, a novel linker family, termed S-helix (signalling helix), has been identified [25]. S-helix has been found in thousands of signalling proteins from bacteria to humans, separating diverse sensor domains from the enzymatic domains, such as in HKs, cNMP cyclases, PP2C phosphatases, AAAþ ATPases and diguanylate cyclases [25]. S-helix linkers are predicted to form dimeric coiled coils [25]. In yeast osmosensor Sln1p and bacterial nitrate sensor NarX proteins, the S-helix linkers were shown to play essential and conserved roles in the signal transduction from the sensor domains to the enzymatic domains [11,26,27].
Coiled coils are common structural features that usually consist of two or more a-helices forming helical bundles. The primary structure of a coiled coil is represented by heptad repeats denoted (abcdefg) n , where n is the number of repeats. Positions 'a' and 'd' are generally hydrophobic residues and they form the 'knobs and holes' at the buried core, responsible for the stability of the coiled-coil structure. Positions 'e' and 'g' are usually hydrophilic residues located on the outer rim of the coiled coil and their interactions further stabilize the coiled-coil structure [28].
DctB is a typical HK sensor, responsible for detecting the major energy and carbon sources C 4 -dicarboxylic acids (DCA) for nitrogen fixation in symbiotic rhizobium. In the presence of DCA, DctB phosphorylates its RR DctD, which in turn activates the transcription of the DCA-transporter-encoding gene dctA [29][30][31]. In particular, the apo form of DctB is at an inactive state ('OFF'), whereas the ligand-binding induces an activated state ('ON') [7]. DctB detects stimuli with its periplasmic sensor domain (residues 55-308, DctBp) and transmits the 27 1b). Structural analysis reveals that succinate binding to the mdPAS domain induces conformational changes mainly at the dimeric level. In this case, each monomer moves as a rigid body, resulting in about a 20 Å distance increase between the two N-termini of DctBp as well as a 208 rotation along the axis perpendicular to the membrane plane (figure 1b) [7,13]. A salt bridge network among residues D89-K110-D89 0 -K110 0 (' indicates residues from the other monomer) is formed in the 'ON' conformation. Perturbing this network through mutations (such as K110D mutation) resulted in constitutively inactive DctB [7,13].
In this study, we investigated the key determinants for the transmembrane signal transduction in DctB. In the periplasmic side, we identified a single leucine residue (L71) on the dimeric interface of DctBp to be essential for maintaining both the 'OFF' state and the signal transduction. Mutagenesis analysis indicates that the integrity of this hydrophobic residue is key to the proper signal transduction. In addition, we show that a delicate balance of interactions at the dimeric interface and the ability to form and dissociate the dimer are essential for response to signal. In the cytoplasmic side, we identified a typical S-helix linker of about 40 amino acids in DctB [25]. Mutating a number of conserved residues in the S-helix linker resulted in hyperactive or inactive phenotypes of DctB, suggesting its essential roles in signal transduction. Combining mutations with opposing activities in the above two regions, we show that these two signalling transduction elements have a combined effect on the final activity of DctB.

A delicate balance of the dimer interface is essential in controlling signal transduction
The signal perceived by DctBp was supposed to be transduced via the dimeric interface of DctBp [7,13]. The dimeric interface is composed of the membrane distal part and membrane proximal part [7,13]. Upon ligand binding, the membrane proximal dimeric interface dissociates while the membrane distal dimeric interface is still maintained (figure 1b). The dissociation of the membrane proximal dimeric interface is essential to induce changes that ultimately lead to the activation of the catalytic domain. In order to understand the molecular basis for the changes at the dimeric interface, we analysed key residues at the interface. Previous structural data show that in the 'OFF' state, the membrane proximal dimeric interface is a coiled coil formed by two a-helices (residues 67-76 in a1 and a1 0 , figure 2a,b). The N-terminal part of the two a1-helices curve away from each other (residues 55 -66), while the C-terminal part of the a1-helix (residues 77 -86) is involved in interactions with other parts of DctBp ( figure 2a,b). The middle section, containing residues 67-76, could therefore play crucial roles in the change from the 'OFF' to the 'ON' state. Specifically, three residues, L67, L71 and A74 appear to form the 'knobs' and 'holes' of the dimeric coiled coil (figure 2b), maintaining the 'OFF' state. SOCKET [32] analysis of DctBp apo structure (PDB: 3E4Q) did identify residues L71 and A74 (and only these two) participating in the classical 'knobs into holes' interactions (figures 2b and 3b). To investigate their roles further, we mutated these two leucine residues to alanine separately and analysed the activity of the mutant proteins in terms of their ability to activate transcription in the absence and presence of an inducing agent, succinate. The activities are measured using a reporter system on a plasmid with a DctB-regulated dctA promoter [7]. DctB-L67A acts similarly to wild-type proteins, while DctB-L71A is constitutively active and no longer responds to inducing signals (figure 2c), suggesting the importance of L71 in maintaining the 'OFF' state conformation.
To further investigate the roles of L71, we carried out additional mutagenesis analysis. Our results show that all small side chain variants, such as DctB-L71A, DctB-L71G and DctB-L71S, are constitutively active in vivo (figure 3a), and could not be further induced, even when the concentration of succinate was as high as 40 mM. On the other hand, two conservative hydrophobic changes, DctB-L71I and DctB-L71V, are inducible (figure 3a). These results suggest that the hydrophobic property of L71 is essential to maintain the 'OFF' conformation under non-inducing conditions. Interestingly, the activities of DctB-L71I and DctB-L71V are higher than that of the wild-type. However, the ratio between the activities of the induced and non-induced DctB (we define this as inducibility in table 1) is relatively similar, suggesting that the system is robust in its response to signals as long as residue 71 is hydrophobic (L, I or V).
Our data so far suggest that the hydrophobicity and size of the L71 side chain are important in maintaining the coiledcoil structure in the 'OFF' conformation. As A74 also contributes to the dimer interface (figures 2b and 3b), we questioned whether the size and hydrophobicity of residue 74 are also important. We mutated A74 to a leucine residue, and the results show that the activity of DctB-A74L remains low irrespective of the addition of the inducer (figure 3c), in contrast to L71A which stays constitutively active. These data suggest that a delicate balance of the dimer interface is required to maintain the 'OFF' state and to switch to the 'ON' state. When the interactions at the dimer interface are strengthened, as in A74L, the activity stays low irrespective of the inducing agent, implying that the protein is unable to switch to the 'ON' state. On the other hand, when the dimer interface is weakened, as in L71A, L71G and L71S, the protein switches to the activated state without inducing agent. Indeed, the combined double mutant DctB-L71A/A74L, which presumably results in a dimer interface that has properties in-between the single mutations, has a final activity that is in-between those of individual mutants (figure 3c).

The dissociation of the dimeric coiled coil a1-a1 0 is essential in signal transduction
The above data show that a delicate balance at the dimer interface is essential for signal transduction in DctB. We reasoned that this is due to the required dissociation of the a1-a1 0 coiled coil upon succinate binding in order to induce further downstream changes. To confirm this, we rsob.royalsocietypublishing.org Open Biol. 4: 140023 introduced a disulfide bridge by mutating L71 to a cysteine residue in order to covalently link the two a1-a1 0 -helices, artificially forbidding the dissociation of the a1-a1 0 coiled-coil dimer. We expressed DctB without the kinase domain (residues 1-395, termed as DctB 395 ) to confirm that this mutation indeed induces covalently linked dimers. SDS-PAGE analysis shows that approximately 70% of His 6 -DctB 395 -L71C forms covalently linked dimers (figure 4a). Similar results have also been observed in the full-length DctB (electronic supplementary material, figure S1). We then introduced the same mutation in our in vivo assays and probed the activities of full-length DctB-L71C. The results show that DctB-L71C has similar activities to that of wild-type in the absence of succinate, confirming that DctB-L71C is in the 'OFF' state. However, the inducibility is significantly reduced (figure 4b and table 1), confirming that DctB-L71C is defective in activation. Addition of the reducing agent dithiothreitol (DTT) recovered the activity of DctB-L71C to about three times of that of wild-type under the same condition (figure 4b). As a result, the inducibility of DctB-L71C is recovered to slightly higher than that of wild-type in the same reducing condition (table 1). These results further support the idea that a dissociable dimer interface is essential for signal transduction and induction.

Differential roles of L71 and the salt bridge network in signal transduction of DctB
Previous structural and biochemical analysis [7,13] revealed that a salt bridge network consisting of D89-K110-D89 0 -K100 0 is essential for maintaining the 'ON' conformation (figure 5a). To investigate the interplay between the L71 -L71 0 and the salt bridge network in signal transduction, we constructed a double mutant that perturbs both 'OFF' (L71A) and 'ON' states (K110D) (figure 5b -d). Compared with single mutations, the double mutant has an activity somewhere in-between (figure 5e), in agreement with the idea that the 'ON' and 'OFF' conformations are stabilized by the salt bridge network and the hydrophobic interactions between the leucine residues, respectively. In the absence of one of the two stabilizing interactions, the conformation stabilized by the other one dominates (figure 5b,c). In the absence of both stabilizing interactions (as in the double mutants), it is possible that the protein would adopt a conformation between the 'OFF' and 'ON' conformations or would oscillate between them (figure 5d), with an average activity in-between the two (figure 5e). The inducibility is significantly reduced, suggesting that the system has lost its ability to respond to signal.

Characterization of the S-helix linker of DctB
The signal perceived by DctBp would be transduced to the kinase module via a cytoplasmic linker. A PSI-BLAST search and sequence alignments indicate that the linker region of DctB is a typical S-helix linker, with members including Sln1P from Saccharomyces cerevisiae and NarX from Escherichia coli (figure 6a). Highly conserved residues include hydrophobic residues L358, V362, L369, L376, hydrophilic residues E359 and R365 and a polar residue T366 (figure 6a). Sequence analysis of the S-helix linker using coiled-coil prediction programs results in two different coiled-coil packing models, one continuous coiled coil and one with a discontinuity at residue V362 (figure 6a and the electronic supplementary material, figure S2) [28,33,34]. In either model, all hydrophobic residues mentioned above are predicted to be in an 'a' or 'd' position in the coiled coil (figure 6a and the electronic supplementary material, figure S2), serving as core residues to stabilize the dimeric coiled coil. To probe the importance of these conserved residues, we mutated L369 and L376 to A, R365 to E and T366 to either L (hydrophobic), R ( positively charged) or E (negatively charged). Interestingly, L369A has little effect on its activity while L376A results in mainly inactive proteins (electronic supplementary material, figure S3), suggesting that these two leucine residues may play different roles. R365E mutation resulted in the loss of function (electronic supplementary material, figure S3), highlighting the potential importance of this positively charged residue. Importantly, T366, a conserved polar residue that is predicted to occupy an 'e' or 'a' position in the two different coiled-coil models, showed drastically different phenotypes depending on the type of mutations. When T366 is mutated to a hydrophobic L, I or V, the activity of DctB is significantly reduced; when it is mutated to a charged residue (either R, K or E, D), it becomes constitutively active (figure 6b). The mutagenesis results are consistent with the discontinuous coiled-coil model which places T366 in the position 'a' (figure 6a and the electronic supplementary material, figure S2). The polar residue T366 in the coiled coil therefore serves as a key modulator for signal transduction. Similarly to the periplasmic coiled-coil linker, a delicate balance of the cytoplasmic S-helix linker is thus also required for DctB's activity. It is also notable that although T366R and T366K have similar levels of induced activities as wild-type DctB, their uninduced activities have elevated significantly, which result in reduced inducibilities (table 1, about four times less than wild-type).

Periplasmic DctBp and cytoplasmic S-helix linker have combined effects on transmembrane signal transduction of DctB
Our data here show that both the flanking helix in periplasmic DctB and the S-helix linker in the cytoplasmic side can modulate signal transduction and both transducers are delicately balanced to allow a regulated response in signal transduction.
To investigate the influence of these two elements in the signal transduction, we took advantage of the mutants available on both the periplasmic and the cytoplasmic sides. rsob.royalsocietypublishing.org Open Biol. 4: 140023 We constructed double mutants that are dominantly negative in the periplasmic domain and constitutively active in the cytoplasmic domain or vice versa. For example, the mutation K110D from DctBp, which resulted in a significant loss of activity, was combined with the mutation T366E from the S-helix linker, which results in hyperactive protein ( figure 7). The hyperactive mutation from the DctBp (L71A) was combined with the defective mutant from the S-helix (T366L) (figure 7). The activities of these double mutations are in-between the activities of single mutants (figure 7b,c), suggesting that defects caused by one domain can be rescued by the other to some extent. These results show that both periplasmic and cytosolic signal transduction elements are important and the final activity is a result of a combined signal transduction process (figure 7a). Indeed, the activities suggest that both signal transduction elements act as amplifiers arranged sequentially and the final activity is thus an integrated result of both amplifiers.

Discussion
The cell membrane is considered as the biggest barrier for signals to be transmitted into a cell. Transmembrane HKs and their RRs play crucial roles in signal transduction in bacteria and eukaryotes [5]. In this report, through mutagenesis aided by previous structural work, we have identified a leucine residue, L71, that is essential for maintaining the apo 'OFF' state of DctBp. Interestingly, adding one extra hydrophobic residue by mutating A74 to leucine results in the constitutively 'OFF' state, inhibiting the signalling process. These data suggest that a delicate balance of the interactions at the dimer interface is required to maintain DctBp in the compact dimer conformation in the absence of ligand, but to enable the dimer to dissociate upon ligand binding. This notion is further supported by the in vivo analysis of DctB-L71C. Results show that by covalently linking the two monomers, the signalling process is inhibited. This process can be restored when the disulfide bridge is disrupted under reducing conditions ( figure 4). Interestingly, DTT has an additive effect with succinate, even with wild-type protein (figure 4b). No clear binding between DTT and DctBp has been observed (electronic supplementary material, figure S4). The exact reasons for this enhanced activity are unclear, but it is unlikely to be due to unspecific binding of DTT to the protein as DTT has no effects in the absence of succinate (table 1).
It is interesting to note that although residue L67 is predicted to be at position 'd' in the dimeric a1-a1 0 coiled coil, DctB-L67A mutation did not result in significant changes compared with the changes of DctB-L71A (figure 2c). Structural analysis shows that at L67, the two a1-helices start to curve away from each other, and L67 and L67 0 are only loosely associated at the DctBp dimeric interface (electronic supplementary material, figure S5 and figure 2b). Consequently, only leucine (L71 -L71 0 ) and alanine (A74 -A74 0 ) pairs act as 'knobs and holes' to stabilize the central region of the dimeric a1-a1 0 coiled coil.
Many HKs contain more than one PAS domain, acting as the sensor domains [12,35]. Previous studies suggested that the interactions between tandem PAS domains could potentially affect the signal amplification/integration [36]. In DctB, the linker between the tandem PAS domains is a5 [7,13]. Structural analysis suggested that a5 possibly interacts with a1 to form a four helical bundle at the dimer interface in the 'OFF' state (electronic supplementary material, figure S5). It is therefore possible that a5 can influence the dimerization of a1-a1 0 and hence affect the ligand-induced activation. Signals detected by the mdPAS domain of DctBp could be possibly integrated/amplified with the membrane proximal PAS (mpPAS) domain through a5 via torque forces [36]. The changes in a5 could subsequently affect the dimerization of a1-a1 0 via interactions between a5 and a1.
The oligomerization and dissociation of coiled-coil structures have been used as a means to regulate biological functions [37][38][39]. Dissociating a typical leucine zipper (with multiple pairs of 'knobs' and 'holes') is an energy-demanding process [39,40]. In the dimeric coiled-coil structure of a1-a1 0 , only a single leucine pair accompanied by an alanine pair act as the 'knobs' and 'holes' and no other potential hydrophilic interactions between residues at 'e' and 'g' positions have been detected in the crystal structures. Clearly, the system shown here is much weaker than the classical coiled coil, enabling it to dissociate readily. The hydrophobic strength of the a1-a1 0 coiled-coil dimer is therefore carefully balanced to regulate the dimeric states of DctBp upon ligand binding.
It is widely accepted that intact HK is a dimer and the major dimerization determinants are the DHp domains,  [35]. The membranes could provide additional restrictions to strengthen the dimer of the sensor domains. However, it is possible that the intrinsic weak dimer of HK domains (such as the sensor domains) has evolved for a specific signal transduction system. From the data here, we observe that the conformational changes in the dimeric level of DctBp are essential for the signal transduction. It is thus possible that the intrinsic weak dimerization of the sensor domains is necessary to maintain an appropriate response to the environment in other HKs as well.
For most membrane-associated HKs in TCS, the signal perceived in the periplasmic domain is transmitted across the membrane to the cytoplasmic kinase module via a conserved linker [5]. The mutagenesis analysis of conserved residues in the S-helix highlights the importance of this region and the potential functional and structural roles played by these conserved residues. Particularly interesting is T366, which displays opposing effects when mutated to hydrophobic or charged residues, pinpointing to a pivotal role in controlling the switch from 'OFF' to 'ON' conformations. Bioinformatics analysis suggests that the S-helix has a large propensity to form a dimeric coiled coil. These conserved residues could play important roles in the dimer interface, with T366 situated at pivotal points for the conformational switches during signal transduction. The mutagenesis results of T366 showed that when T366 is mutated to a hydrophobic residue, the protein becomes mainly inactive, while when T366 is mutated to a charged residue, it becomes hyperactive. These results suggest that the S-helix coiled coil also modulates the signal transduction and the activation of DctB. The predicted 'a' position of T366 implies that a hydrophobic residue would strengthen the coiled coil, while a charged residue would significantly weaken the coiled coil. The results of T366 in figure 6 are therefore in agreement with the idea that the delicate balance of the dimeric coiled coil formed by the S-helix serves as another modulator for the signal transduction. When the coiled coil is too stable as in T366V, T366I and T366L, the protein is in the 'OFF' state irrespective of inducing signals; when the coiled coil is weakened as in T366K, T366R, T366D and T366E, the protein is constitutively in the 'ON' state.
It is interesting to note that the DcuS from E. coli, which also detects DCA signals, uses a cytoplasmic PAS domain instead of S-helix to transduce the signal in the cytoplasmic regions [41]. Monzel et al. [41] suggested that the conformational changes of the cytoplasmic PAS domain of DcuS at the oligomeric level are essential for the kinase activation. It is thus possible that although different structural elements are involved in signal transduction, changes at the tertiary structural level, either through conformational changes and/or oligomeric transitions, are used in signal transductions.
Our data using the combined point mutations of the two signalling elements show that the functional state of DctB is determined by a combined net effect of the peri-and cytoplasmic domains. Signal perceived at the ligand-binding site is first transduced by the periplasmic coiled-coil region consisting of L71 and A74, and further transduced via the transmembrane helix and subsequently the S-helix linker containing T366. Our results show that the activities of the hyperactive 'ON' mutation in the periplasmic linker are dampened down by the inactive mutant in the S-helix. Likewise, the reduced activity of the 'OFF' mutant in the periplasmic linker is to some extent restored by the hyperactive 'ON' mutant in the S-helix ( figure 7). The effects of individual modules/domains therefore do not dominate, but are instead combined to give rise to the final signal (figure 7a). Indeed, our data are in agreement with the two linker regions acting as two transducers/amplifiers, signals perceived by the periplasmic PAS core being thus processed by the these two components, possibly through a dimer to monomer transition and/or tertiary structural changes at the dimer interface, to activate DctB cytoplasmic kinase domains.
Interestingly, although mutations could reduce the induced activities or increase both non-induced and induced activities (table 1), the inducibility (responsiveness) is almost invariantly reduced with the exceptions of L67A, L71I and L71V, suggesting that DctB has a relatively high level of sensitivity to succinate signals. The addition of reducing agent DTT has increased the inducibility of both the wild-type DctB and its L71C mutant, however the inducibility between  rsob.royalsocietypublishing.org Open Biol. 4: 140023 these two is still similar (that of DctB-L71C is about 20% higher than that of DctB wild-type in the same condition, see table 1). Once the delicate balance of the transduction elements has been affected by modifications (such as L71A, T366R, etc.), the responsiveness to succinate is affected, highlighting its importance in the response to signals. Although the two signal transducers discussed here could amplify or dampen signals and the final signal can be modulated by balancing these two components, the responsiveness to signal appears to be sensitive to mutations. It is thus tentative to suggest that the delicate balance of the system has evolved to respond to specific environmental signals. Considering that HKs are conserved in modularized architecture, usually containing a sensory domain and a conserved cytoplasmic linker, it is possible that this highly specific and sensitive amplifier model could apply to other HKs to generate a system that responds to specific environmental stimuli.

Coiled-coil analysis of DctB S-helix linker
The cytoplasmic linker region of DctB (residues 341-410) was used to search against a non-redundant protein database (NCBI) by iterative position-specific BLAST (PSI-BLAST, http://blast.ncbi.nlm.nih.gov/Blast.cgi). Six iterations identified 250 proteins with significant similarities to the DctB linker with an E-value less than 0.001. The conservation pattern of the central region is similar to that of the previously identified S-helix linker [25]. Therefore, the central region of the cytoplasmic linker of DctB (residues 350-395) was further analysed by PSI-BLAST and identified as an S-helix linker. Subsequent sequence alignments were carried out by MAFFT v. 5.0 [43] and JALVIEW v. 2.5 [44]. Secondary structure predictions of DctB S-helix were performed in JPRED [45]. The coiled-coil analysis of DctB S-helix linker was carried out by COILS [34], PAIRCOIL2 [33] and MARCOIL [46].

Genetic manipulations
All constructs for in vivo activity assays are based on sitedirected mutagenesis of pKU4021 (table 2). For constructs with a single mutation, site-directed mutagenesis was based on the wild-type dctABD operon construct generated by PCR that incorporated the appropriate mismatches. For constructs with double mutations, site-directed mutagenesis was based on the mutated dctABD operon by PCR with appropriate mismatches. For in vitro analysis of disulfide bridges, the dctB fragment containing amino acids 1-395 of DctB was cloned and inserted into pBAD-myc/HisA with digesting sites BglII and HindIII. All constructs were confirmed by sequencing (BGI, Beijing, People's Republic of