Transmembrane signaling and cytoplasmic signal conversion by dimeric transmembrane helix 2 and a linker domain of the DcuS sensor kinase

Transmembrane (TM) signaling is a key process of membrane-bound sensor kinases. The C4-dicarboxylate (fumarate) responsive sensor kinase DcuS of Escherichia coli is anchored by TM helices TM1 and TM2 in the membrane. Signal transmission across the membrane relies on the piston-type movement of the periplasmic part of TM2. To define the role of TM2 in TM signaling, we use oxidative Cys cross-linking to demonstrate that TM2 extends over the full distance of the membrane and forms a stable TM homodimer in both the inactive and fumarate-activated state of DcuS. An S186xxxGxxxG194 motif is required for the stability and function of the TM2 homodimer. The TM2 helix further extends on the periplasmic side into the α6-helix of the sensory PASP domain and on the cytoplasmic side into the α1-helix of PASC. PASC has to transmit the signal to the C-terminal kinase domain. A helical linker on the cytoplasmic side connecting TM2 with PASC contains an LxxxLxxxL sequence. The dimeric state of the linker was relieved during fumarate activation of DcuS, indicating structural rearrangements in the linker. Thus, DcuS contains a long α-helical structure reaching from the sensory PASP (α6) domain across the membrane to α1(PASC). Taken together, the results suggest piston-type TM signaling by the TM2 homodimer from PASP across the full TM region, whereas the fumarate-destabilized linker dimer converts the signal on the cytoplasmic side for PASC and kinase regulation.


Edited by Karen Fleming
Transmembrane (TM) signaling is a key process of membrane-bound sensor kinases. The C 4 -dicarboxylate (fumarate) responsive sensor kinase DcuS of Escherichia coli is anchored by TM helices TM1 and TM2 in the membrane. Signal transmission across the membrane relies on the pistontype movement of the periplasmic part of TM2. To define the role of TM2 in TM signaling, we use oxidative Cys crosslinking to demonstrate that TM2 extends over the full distance of the membrane and forms a stable TM homodimer in both the inactive and fumarate-activated state of DcuS. An S 186 xxxGxxxG 194 motif is required for the stability and function of the TM2 homodimer. The TM2 helix further extends on the periplasmic side into the α6-helix of the sensory PASP domain and on the cytoplasmic side into the α1-helix of PAS C . PAS C has to transmit the signal to the C-terminal kinase domain. A helical linker on the cytoplasmic side connecting TM2 with PAS C contains an LxxxLxxxL sequence. The dimeric state of the linker was relieved during fumarate activation of DcuS, indicating structural rearrangements in the linker. Thus, DcuS contains a long α-helical structure reaching from the sensory PAS P (α6) domain across the membrane to α1(PAS C ). Taken together, the results suggest piston-type TM signaling by the TM2 homodimer from PASP across the full TM region, whereas the fumarate-destabilized linker dimer converts the signal on the cytoplasmic side for PAS C and kinase regulation.
The fumarate, or C 4 -dicarboxylate (C4DC), sensor kinase DcuS of the DcuS-DcuR two-component system consists of a PAS P (Per-ARNT-SIM) sensor domain, a TM domain composed of the antiparallel transmembrane helices TM1 and TM2, and a cytoplasmic PAS C and the kinase domain (22,23). The sensory PAS P domain is framed by TM1 and TM2 and adopts periplasmic location. For C4DC responsiveness, DcuS requires C4DC transporter DctA or DcuB under aerobic and anaerobic conditions, respectively (24)(25)(26). DcuS forms a homodimer in bacterial membranes (27), and the two α-helices TM1 and TM2 of the DcuS monomer are supposed to facilitate DcuS homodimerization (18). DcuS TM signaling involves TM2, which is linked on the periplasmic side to the PAS P domain and on the cytoplasmic side to the PAS C domain. Binding of a C4DC ligand to the sensor domain (28)(29)(30) results in compaction of the ligand binding site in PAS P and uplifting of the C-terminal α6-helix of PAS p . The uplift and structurally related arrangements upon citrate binding have been shown directly for homologous CitA in a membrane context (6,21). This uplift pulls the periplasmic part of TM2 of DcuS in the direction of the periplasm in a piston-like movement (18). TM signaling by sensor kinases is, however, complex and can change in different segments of the TM region or combine different modes of conformational changes (31). In contrast, TM1 of DcuS exhibits no piston movement and revealed no contribution to TM signaling (18).
To test for piston-type movement or indications for additional modes of TM signaling upon fumarate activation of DcuS, we studied the α-helical arrangement and dimerization of TM2 for the full-length and for the adjacent cytoplasmic region by Cys scanning and oxidative cross-linking (CL). This should define the extension of the α-helix from the periplasmic side across the membrane and its connection to the cytoplasmic PAS C domain, with an emphasis on structural properties that could be relevant for the signal transfer across the membrane. Cysteine scanning and oxidative CL were applied as major methods to study TM2 interactions, as well as structural alterations due to receptor activation by fumarate. All studies were performed in bacteria containing DctA, which is a functional determinant of DcuS function (25).

TM2 of DcuS homodimerizes at specific contact sites regardless of activity state
The role of TM2 and the adjacent cytoplasmic region in DcuS homodimerization was analyzed in aerobically grown Escherichia coli cells, i.e., in the native context with DctA, by Cys CL using membrane-permeant copper(II)-(1,10phenanthroline) 3 , or "Cu 2+ phenanthroline," as an oxidant (18,32,33). In our previous study, we used CL to study interactions between α6 of PAS P and the initial part of TM2 up to residue Ser186 (18). Here, we studied the involvement of every single residue of the complete TM2 helix in DcuS dimerization up to residue Glu213 ( Fig. 1), which is located on the cytoplasmic side in the N terminus of the α1-helix in PAS C (9,34). Each amino acid was individually replaced genetically by a Cys residue in the plasmid-encoded Cys-less variant DcuS Cys0 . In DcuS Cys0 , the natural Cys199 and Cys471 are replaced by Ser, which only moderately decreases fumarate stimulation of DcuS (18,27). The Cys variants retained all 87%-118% of DcuS Cys0 activity in dcuB-lacZ expression and induced dcuB-lacZ in a fumarate-dependent manner as the wild-type (Fig. 2). However, there was a striking drop in dcuB-lacZ expression in a coherent stretch of seven amino acids (F189-L195).
Cys-mediated CL of the DcuS variants was visualized by nonreducing SDS-PAGE and Western blotting with DcuS specific antisera (18), which allowed us to separate DcuS monomers (calculated mass 62 kDa; apparent mass 55 kDa for His 6 -DcuS) from cross-linked DcuS homodimers (calculated mass 124 kDa; apparent mass 170 kDa; Fig. 1A). Reaction of the antiserum with DcuS and the cross-link product was verified by comparing the reaction with purified DcuS (Fig. S1) (35).
The CL efficiency was calculated from the levels of crosslinked versus free DcuS (Fig. 1A) and ranged from 0 to 74%. We ensured that the band intensity was in the linear range of detection to allow a comparison of band intensities (18). The CL efficiencies were high at several positions within TM2 (approx. Ser182 to Leu201), the periplasmic α6 of PAS P , and the cytoplasmic α1-PAS C helices (see Fig. 1B). Lower CL efficiency was observed at the membrane/water interfaces on the periplasmic (residues Trp181-Trp185) and cytoplasmic (Cys199-Lys207) sites. For most regions, a +3/+4 periodicity of high CL efficiency was observed (Fig. 1B), which is characteristic of the interaction between two α-helices. Notably, the CL efficiency and the +3/+4 periodicity were retained in the region from F189 to L195, demonstrating that the region has a supposed α-helicity, which is retained in the Cys replacement mutations.
Remarkably, the CL pattern and yield were essentially identical for fumarate activated versus nonactivated DcuS, with a maximal 12% difference in the CL efficiency. The clearest differences were observed in the Cys199-Lys207 region, corresponding to the TM2-PAS C interdomain or linker region. In this region, a generally low CL efficiency was observed with partial nonhelical spacing of the CL maxima, and relatively large fumarate-induced differences, especially for residues Leu201, Val202, Val204, Leu205, and Lys206.
The CL results suggest that the region from α6 in PAS P to α1 in PAS C forms a continuous α-helix, including the complete TM2 region.
A GxxxG motif in TM2: its role for DcuS activity and DcuS homodimerization The sensitivity of DcuS activity to Cys replacement mutation in the F189-L195 region coincides with the presence of a GxxxG motif in this region. GxxxG motifs frequently mediate the interactions of TM α-helices via "ridge into groove" tight packing of two interacting TM helices (36)(37)(38). The Gly residues can be replaced by other amino acid residues with small side chains, such as Ala, Ser, or Thr, resulting in a (small) xxx(small) (39) or glycine-zipper-like motif, (small)xxx(small) xxx(small) (40). The F189-L195 region in TM2 is part of such a glycine-zipper-type motif S 186 xxxGxxxG 194 (Fig. S2) (41,42).
A role of the motif for DcuS function and activity was tested by substituting the motif residues and adjacent residues with Ala or Cys (for Ser), which avoids gross structural rearrangements. Replacement of Gly with Ala in GxxxG motifs disrupts helix-helix interactions, as the Gly residues are important for TM helix dimerization (38,43,44). The impact of the mutations on DcuS function was assayed using the variants in a DcuS-less background by monitoring DcuS-dependent dcuB-lacZ reporter gene expression (Fig. 3). Individual replacements in the signature residues Ser186 and Gly190 of the motif did not significantly affect dcuB-lacZ expression, whereas the activity decreased to 44% of wild-type after replacement of Gly194 by Ala. Mutation of other residues in that region had no effect as well. However, mutation of the core pair of the motif (Gly190 residue together with Gly194) dramatically reduced the activity to 16% of wild-type. The effect is specific since mutation of neighboring residues (W185A together with F189A) had no effect. Remarkably, mutation of Ser186 together with Gly190 showed the same reduction in activity as for the Gly190 plus Gly194 pair. Taken together, the data suggest that the S 186 xxxGxxxG 194 motif in TM2 (particularly Gly190 or the Gly190 + G194 pair) is crucial for DcuS function. Notably, the level of DcuS protein in membranes was similar for wild-type and all variants after expression (Fig. S3A), confirming that the observed differences were caused by the mutations rather than by altered DcuS levels.
As Thr residues can also be a part of (small)xxx(small) motifs (38), we analyzed the involvement of Thr198. However, when both the Thr198 and the Gly194 residues were replaced by Ala residues (Fig. 3), dcuB-lacZ expression was not decreased compared with DcuS(G194A) alone, indicating that Thr198 is not important for DcuS activity.

Role of the S 186 xxxGxxxG 194 motif in homodimerization of TM2
The role of the S 186 xxxGxxxG 194 motif in DcuS homodimerization was tested using bacterial two-hybrid (BACTH) systems. For the full-length DcuS protein, we applied the BACTH system that relies on the restoration of adenylate cyclase activity and cAMP production by fusing the individual domains (T18 or T25) of Bordetella pertussis adenylate cyclase to interacting proteins. Interaction restores the adenylate cyclase activity, which can be monitored as β-galactosidase expression (49,50). The T18 and T25 domains were genetically fused to the DcuS N terminus. When T18-DcuS was coexpressed with T25-DcuS, high β-galactosidase activity was observed for wild-type DcuS (Fig. 4A). When the signature residues Gly190 and Gly194 of the S 186 xxxGxxxG 194 motif were individually mutated, activity (DcuS homodimerizationdependent lacZ expression) decreased to 70% and 55%, of wild-type for the G190A and G194A variants, respectively, but did not significantly decrease for the S186C variant. Combining the Gly190 with the Gly194 mutation to the G190A + G194A double mutant decreased the activity to near background levels. Combination of S186C with the G190A mutation to the S186C + G190A double mutant added no further effect to the single G190A mutant. The levels of T18and T25-fusion proteins and their membrane integration were confirmed to rule out that changes in activity were caused by low expression or lack of membrane integration (Fig. S3). Taken together, these data suggest that the central G 190 xxxG 194 pair within the S 186 xxxGxxxG 194 motif is important for homodimerization of DcuS.
We applied the GALLEX two-hybrid system to test TM2 homodimerization in the absence of other, potentially interfering, domains of DcuS (51,52). In the GALLEX system, the C termini of isolated TM domains are genetically fused to the maltose-binding protein (MBP) that ensures proper membrane integration of the TM helix. At the TM2 N terminus, the DNA-binding domain of the LexA protein was fused (Fig. 4B). Dimerization of DcuS TM2 would result in dimerization of the DNA-binding domain, which then binds to the sulA promoter, resulting in repression of the downstream lacZ reporter gene (53) and decreased β-galactose activity (Fig. 4B). The strongly interacting TM region of human glycophorin A (GpA) serves as a positive and the weakly interacting variant GpA(G83I) as a negative control (51). First, a 16-amino-acid helix (Ile183-Thr198) was established as the optimal length of DcuS TM2 for GALLEX analysis, and the expression, membrane insertion, and proper membrane topology of the expressed fusion protein were confirmed (Figs. S4 and S5). Expression of the wild-type TM2-LexA strongly inhibited lacZ expression, meaning that the isolated TM2 of DcuS strongly homodimerized with about 70% of GpA wild-type (Fig. 4B). However, the observed interaction (measured as inhibition of the β-galactosidase activity) was lost in the case of DcuS TM2 double-mutant G190A + G194A but also S186C + G190A (Fig. 4B). The activities of the double mutants even fell below the negative control of the GpA (G83I) variant. These results confirm that TM2 of DcuS homodimerizes, which is in agreement with the BACTH and CL results, and

Transmembrane signaling by DcuS sensor kinase
show that the S 186 xxxGxxxG 194 .motif plays an important role in the homodimerization. Though the S186C + G190A mutation did not dramatically affect dimerization of fulllength DcuS in the BACTH assay (Fig. 4A), homodimerization of isolated TM2 was affected nearly as dramatically as in the case of the G190A + G194A mutant (Fig. 4B). The retained interaction in the S186C + G190A mutation in the BACTH assay indicates that other DcuS regions further stabilize the TM2-mediated interaction in the full-length protein. Overall, the BACTH and GALLEX data demonstrate an important role of the S 186 xxxGxxxG 194 motif, and particularly G 190 xxxG 194 , in homodimerization of isolated TM2 and full-length DcuS.
The CL data and the interaction studies indicate a TM2/ TM2ʹ homodimer that is stabilized by the S 186 xxxGxxxG 194 motif, which forms the TM2/TM2ʹ interaction surface. Calculation of the TM2/TM2ʹ homodimer by the web-tool PREDDIMER (54) provided as one of the top-scoring structures a homodimer with an interaction surface that involves exactly the residues S 186 xxxGxxxG 194 motif (Fig. 5). Thus, based on the experimental data and predicted structure, the TM2 helix forms a right-handed dimer with a crossing angle of -55 . Residue Thr198 that is located on the subsequent helical turn of the dimer also exhibits a high degree of Cys-CL (Fig. 1B), and Thr residues can be part of (small)xxx(small) motifs (38). Mutation studies did not prove direct involvement of Thr198 in TM2 homodimerization and function, and Thr198 was not prominently involved in formation of the dimer interface in the structural model (Fig. 5). Therefore, Thr198 appears to be in spatial proximity, as shown by Cys-CL, but not directly involved in helix dimerization, which is in line with the crossing angle of 55 (Fig. 5).

The TM2-PAS C linker is involved in regulation of DcuS activity
The linker connecting the C terminus of TM2 (18) with α1 of PAS C (34) ranges from Cys199 to Gly211 (Fig. 1). This linker (C 199 ILVKVLKKILFG 211 ) includes the LxxxLxxxL sequence. Sequences with Leu spacings of this type are found in interactions between proteins regulating diverse cellular functions (55). The linker is located at the membranecytoplasm interfacial region of TM2. Leu205 and Leu209 of the linker attracted attention when the piston mode of TM signaling by TM2 was identified (18). Upon the activation of DcuS by fumarate, Leu205 becomes inaccessible and Leu209 accessible to labeling by hydrophilic reagents, which was interpreted to reflect structural reorganization such as a move into or out of the hydrophobic membrane core region (18). Remarkably, three Lys residues are present in the linker, which is in agreement with the cytoplasmic location and positive inside rule (18,56). As seen in Figure 1, the residues of the linker could be cross-linked when individual residues were replaced by Cys. When the linker residues were mutated individually to Ala, a major loss in DcuS function (in the presence of fumarate) was observed for Leu209, and to a lower extent also for the adjacent Phe210, in the dcuB-lacZ reporter gene assay (Fig. S6). In contrast, substitution of Lys206 by Ala resulted in a clear ON phenotypes in the dcuB-lacZ reporter gene assay in the absence of fumarate, which was not observed when mutating the remaining residues, apart from a weak ON phenotype of the V202A substitution. Remarkably, the Cys substitutions in the linker used for the CL assay were not conspicuous in the DcuS activity assay (Fig. 2), suggesting that Cys substitutions are more neutral than Ala substitutions in this region. Cys replacements in the region connecting the TM region of the chemotaxis protein Tsr with the signal  V172C  T173C  Q174C  Q175C  I176C  N177C  D178C  S179C  R180C  W181C  S182C  I183C  I184C  W185C  S186C  V187C  L188C  F189C  G190C  M191C  L192C  V193C  G194C  L195C  I196C  G197C  T198C  C199  I200C  L201C  V202C  K203C  V204C  L205C  K206C  K207C  I208C  L209C  F210C  G211C  L212C  Transmembrane signaling by DcuS sensor kinase transducing HAMP domain, or "control cables," also showed a remarkable tolerance to Cys substitutions (57). The strong effects observed upon mutation of Lys206 and Leu209 suggest that the linker is important for controlling DcuS activity. This assumption is further supported by the observation that the CL efficiency in the linker (Cys199-Lys207) varied to some extent in the presence and absence of fumarate (Fig. 1B).

Cross-linking dynamics of TM2 and the linker
Cys-CL of TM2 and the linker (Fig. 1) was performed under relatively strong oxidizing conditions (16,32) to obtain clear information on α-helicity. The CL indicated a decrease in efficiency in the linker region compared with TM2 and some differences in the sensitivity to fumarate for the linker and the TM2 regions (Fig. 1B). The differences were resolved in the following experiments in more detail (Fig. 6). To this end for each residue, optimal oxidation conditions were adjusted by optimizing the concentration of Cu 2+ phenanthroline and temperature in order to achieve a decrease in oxidative reactivity (see Experimental procedures section). This enabled monitoring CL efficiency over time and comparison for the presence or absence of fumarate. DcuS variants with high CL yield (Fig. 1B) were chosen for analysis to enable quantitative evaluation even under weakened oxidizing conditions. For the TM2 region, amino acids Ser186, Gly190, Gly194, and Thr198 were selected (Fig. 6A). For positions Gly190, Gly194, and Thr198, high CL levels were observed within short time (<1 min) after addition of the oxidant, and the kinetics were not affected by the presence of fumarate. Thus, the interacting Cys residues are similarly positioned in both the inactive and fumarate-activated state. DcuS(S186C) was also rapidly cross-linked with and without fumarate, but the CL efficiency increased slowly in the absence of fumarate.
The CL kinetics were clearly different for residues in the linker region (Fig. 6B). For Cys residues at Leu205, Ile208, and Leu209, the CL efficiency was very low after 0.5-3.0 min in the presence of fumarate, and only slightly increased, if at all, within 20 min. However, in the absence of fumarate, the CL efficiency was also low at the beginning but increased over time. After 20 min, the CL efficiency of the fumarate-deficient state was increased by a factor of 2-7 compared with the fumarate-activated state. The increase was particularly evident for the C-terminal part of the linker (residues 208 and 209). Therefore, the CL conditions allowed a clear differentiation of efficiency for the individual residues in response to fumarate. At position L201, where the linker emanates from TM2 close to the membrane/cytoplasm interface, the CL response was different from those seen for other positions in the linker. CL in the presence of fumarate was low but reached maximal levels very early after 1 min, similar to the neighboring TM2 region (Fig. 6A).
Therefore, the data suggest major differences for homodimerization of TM2 and the linker region, as residues of the linker region (Leu201, Leu205, Ile208, and Leu209), but not the TM helix, exhibited a marked fumarate-induced difference in CL efficiency. These findings suggest fumarate-induced effects on helix dimerization and restructuring of (at least) the extra-membranous part of the linker.
PAS C starts at Leu212 with a long helical region (α1) and nearly parallel arrangement in the PAS C dimer (9,34). Since α1 directly connects the linker to PAS C , the CL efficiency and fumarate response were tested and compared with the response of the linker. When residues of the α1 region  (Table S1) (45)(46)(47)(48). The effect is shown for single (center) and double (right) mutants. Growth was performed under anaerobic conditions in eM9 medium with glycerol plus DMSO with or without 20 mM disodium fumarate. Activities were normalized to the wild-type control of DcuS in the fumarate-activated state. Three biological replicates were tested and the error bars indicate SD.
Transmembrane signaling by DcuS sensor kinase immediately following the linker were individually mutated to Cys, a +4 pattern was observed in CL in accordance with the presence of canonical α-helices. The CL pattern was retained in the inactive and fumarate-activated state of DcuS (Fig. 1B). Therefore, the linker located between TM2 (Ser182 to Cys199) and PAS C starting at Leu212 differs from both regions by its CL efficiency and its response to fumarate. The linker has to be considered as a separate region with distinct properties.

Discussion
In the present study, we show that DcuS contains a long continuous α-helix connecting the periplasmic PAS P sensor domain with PAS C on the cytoplasmic side of the membrane. The studies were performed in bacteria to ensure the availability of functional DcuS, which requires the presence of transporter DcuB or DctA (24,25). The α-helix starting at Val172 in PAS P and ranging to Glu213 in PAS C comprises 9-10 helical turns. The major part (Val172-Cys199) remains homodimeric as shown by Cys scanning and CL study for the full length of the α-helix, regardless of the presence or absence of fumarate, indicating a rigid dimer from PAS P across the membrane to the linker. This TM2 helix dimer is stabilized by the S 186 xxxGxxxG 194 motif, which forms the TM2/TM2ʹ interaction surface. The role of the motif was consistently verified by mutational, functional, and interaction studies and structural modeling of the region. The data show also that the core of the motif (G 190 xxxG 194 ) is extended to obtain the tandem S 186 xxxGxxxG 194 motif. The stability of the TM2/ TM2ʹ dimer suggests its parallel movement upon fumarate activation. It is not known whether both PAS P domains of the dimer are fumarate-bound for activation. For full activation, high levels of fumarate (50 mM) were applied, indicating that both monomers are fumarate-bound, given the app. K M values of approx. 2 mM for C4DCs (28).
Piston-type movement toward the periplasmic side upon fumarate activation has been shown for the N-terminal portion of TM2 (18). The stability and persistence of the TM2 helix dimer demonstrated here strongly suggest a piston-type shift of the entire TM2 dimer for the full TM region upon activation. The persistence of the dimer contact sites is in line with a parallel piston movement. It provides strong arguments against rotational movement of TM2 in TM signaling and against major scissor-like movements in TM2. Scissoring requires kinking and reorganization of α-helices, which is not observed here. Furthermore, Pro residues, which support kinking and bending (14,16,58,59), are lacking. Therefore, the data strongly support a piston-type mode of TM signaling by the TM2 homodimer from the periplasmic α6-PAS P to the linker on the cytoplasmic side of the membrane.
The linker (C 199 ILVKVLKKILFG 211 ) connecting TM2 with PAS C comprises three positively charged Lys residues and an LxxxLxxxL sequence. The linker is α-helical similar to TM2, but differs from TM2 in the extent of helix dimerization, the response to fumarate, and helical spacing. In the linker, the CL efficiency responded to the presence of fumarate (Fig. 6B), implying structural changes upon fumarate addition. The impact of fumarate differed between the N-and C-terminal parts of the linker; the response of Ile208 and Leu209 in the C terminus was different from that of residues Leu201 and Leu205 in the N terminus (Fig. 4B). This suggests that  [16] ; sequence in Fig. S4). The bacteria were cultivated aerobically in LB to an OD 578 of 0.5 and β-galactosidase activities determined. The dimerization of TM2 DcuS [16] (G190A + 194A), which exhibited the weakest inhibition of β-galactosidase expression, was set to 0%, and all other values were normalized accordingly. The constructs TM GpA [17] (Wt) and TM GpA [17] (G83I) served as controls for interaction in the GALLEX system (51). Three biological replicates were tested in (A) and (B) and the error bars indicate SD.
fumarate-induced signaling involves complex structural rearrangements of the linker, including attenuated helix interactions or separation of the helical region in the presence of fumarate. Thus, the linker is predicted to convert the piston movement of TM2 into structural rearrangements and to control PAS C structure and activity. In DcuS, the basis for TM signaling is the movement of a piston (TM2) that crosses the complete membrane as a stable homodimer (Fig. 7). Dimerization of TM2 is central to signaling and maintained in the fumarate-activated and the inactive state. In the closely homologous CitA, movement of the piston is triggered by the binding of citrate to the PAS P domain. Compaction of the binding site then drives an uplift of helix α6-PAS P that directly runs into TM2 (6,21). By analogy to CitA, a similar scenario has been suggested for DcuS (6,61) to trigger the piston movement of TM2. On the cytoplasmic side, the TM2 piston movement is perceived by the linker. In the linker the α-helical structure is modified in stability and +3/+4 spacing of interacting residues in response to fumarate availability. Thus, the mode of signal transduction likely changes in this region from piston mode to reorganization (or destabilization) of the helices in response to fumarate binding (Figs. 7 and S10). The structural changes in the linker are supposed to subsequently reorganize the PAS C dimer, which then controls the kinase domain and activity as suggested previously (9,10). Sensor kinase BvgS from B. pertussis is also composed of a periplasmic signal receiver domain followed by a TM helix and a cytoplasmic 28 aa linker connecting TM2 and a PAS domain. BvgS shows a comparable mechanism in TM signaling by TM2 and signal conversion of the linker by helical coiled-coil disruption (62). Piston-type displacement of BvgS TM2, however, is small, and TM2 is lacking a GxxxG motif.
For other sensor kinases, TM signaling involving a pistontype shift in helices has been shown or suggested, though its importance appears to be less in the entire mechanism. Remarkably, a GxxxG motif stabilizing TM helix dimers is apparently missing in other sensor kinases that also contain two TM helices for TM signaling, such as NarQ, PhoQ, or even the closely related CitA (Fig. S1). Therefore the S 186 xxxGxxxG 194 motif, as important as it is for DcuS, it is apparently not widespread in sensor kinases with piston-mode TM signaling. TM signaling by DcuS resembles, in some aspects, that of the nitrate sensor NarQ. The homodimeric NarQ contains an antiparallel four-stranded coiled-coil structure in the membrane, linking the periplasmic sensor to a cytoplasmic HAMP domain (7). Nitrate binding displaces the sensory domain and shifts the TM1 helices to the periplasmic side relative to TM2. However, a Pro-kink in the C-terminal part of TM2 is crucial for transferring the conformational rearrangement to the HAMP domain in a lever-like mechanism (7). Overall, nitrate binding causes helical rotation and diagonal scissoring of the α-helices connected to the sensor domain, which then leads to a piston-like motion in the TM region (31), demonstrating that TM signaling may apply a combination of conformational changes for signal transduction. The PhoQ-PhoP two-component system of enteric bacteria senses divalent cations and antimicrobial peptides (63,64). Signal transfer by the PhoQ kinase involves scissor-like diagonal displacement of TM helices, or a combination of scissoring and rotational movements, which is suggested by Cysscanning disulfide CL and modeling (3,14,16). In the distantly related sensory rhodopsin from Natronobacterium pharaonis, light excitation induces a rotary motion of the second TM helix that controls finally kinase activity (65). It appears therefore that membrane-bound sensor kinases apply Transmembrane signaling by DcuS sensor kinase various types of conformational changes for transmembrane signaling, and various modes can be combined.

Bacterial growth and molecular genetics
The E. coli, K12 derivatives, plasmids, and primers are listed in Tables S1 and S2. Molecular genetic techniques were performed according to standard procedures (66). The plasmids encoding the GALLEX constructs were generated by inserting TM1 and TM2 sequences obtained by PCR from pMW181 into pBLM100 via the SacI and SpeI sites. For all Cys CL experiments, cells were grown aerobically in 700 μl LB medium (with or without 50 mM sodium fumarate) and expression of dcuS induced after 2 h with 1 mM IPTG (18). For the dcuB-lacZ reporter gene assays, the cells were cultivated anaerobically to an OD 578 of 0.5-0.8 in enriched mineral medium (eM9) containing glycerol (50 mM) and dimethyl sulfoxide (DMSO; 20 mM) (18,30). Sodium fumarate (20 mM) was used as an effector as indicated. For BACTH assays, bacteria were grown in 48-well plates with vigorous shaking to an OD 578 of 0.5-0.8 (18). To measure interactions with the GALLEX system, freshly transformed cells were used to inoculate 5 ml of overnight culture in LB medium. The next day, 5 ml medium was inoculated with 0.1 ml of the overnight cultures and grown aerobically to an OD 578 of 0.6 in the presence of ampicillin and 5 μM IPTG. β-Galactosidase activity was measured as described elsewhere ( [10,67]; supplemental material). For the dcuB-lacZ reporter gene assay, strain IMW260 was transformed with derivatives of pMW181 (DcuS Wt). To test the BACTH (49, 50), E. coli BTH101 was cotransformed with pairs of plasmids encoding DcuS with N-terminal fusions of T18 and T25 fragments (68). In the GALLEX TM-dimerization assay (51), E coli SU101 was transformed with plasmids encoding the GpA and DcuS TM2 fusion proteins.

Oxidative in vivo cysteine cross-linking
To test Cys CL (18), E. coli C43(DE3) was transformed with single-Cys variants encoded by pMW336. Bacteria (OD 578 of 2.5-3.0) were treated for the standard procedure with Cu(II)-1,10-phenanthroline as described (18). The reaction was stopped with 12.5 mM EDTA and 12.5 mM N-ethylmaleimide, and the bacteria evaluated by SDS-PAGE under oxidative conditions (18). For time-resolved CL, the reaction was terminated after 0.5, 3, 10, and 20 min. Oxidant concentrations and temperatures were adjusted in the following way: 0.5 mM CuSO 4 , 1.625 mM 1,10-phenanthroline, and 4 C for variants S186C, G194C, L208C, and L209C; 2 mM CuSO 4 , 7.5 mM 1,10-phenanthroline, and 4 C for G190C; 4 mM CuSO 4 , 13 mM 1,10-phenanthroline, and 4 C for L205C; 4 mM CuSO 4 , 13 mM 1,10-phenanthroline, and 20 C for L201C (all concentrations refer to the reaction mixture). For quantitative analysis of CL efficiency, the X-ray films from the western blots were scanned to obtain a digital image. In the image processing software ImageJ (69), the scanned images were converted to black and white. The individual lanes of each Cys variant per X-ray film were marked by identical squares. The band size and density were plotted as integrals by ImageJ. The relative integral values were measured using the Wand For key residues G190C, G194C, and L209C, the CL kinetics was performed in repeat, and error bars are given. The ratio of CL products to the total amount of DcuS was calculated after scanning the anti-DcuS western blots and evaluating the band intensities in ImageJ software. The western blots are shown in Figure S9.
(tracing) Tool. These values were used for calculating the ratio of the CL product to total DcuS.

DcuS expression and membrane localization
Bacteria (100 or 200 μg cell protein) were sedimented, dissolved in 2× SDS-PAGE sample buffer, and applied to polyacrylamide gels. For oxidative Cys-CL, dithiothreitol was omitted from buffers. For SDS-PAGE (70), 12% gels were used; for Cys CL experiments, gradient gels (5-12% acrylamide) were used. Western blotting was performed (18) with rabbit polyclonal antiserum raised against PAS P of DcuS and peroxidase-coupled goat anti-rabbit-IgG secondary antibody. To test GALLEX constructs, rabbit anti-MBP antiserum (New England Biolabs) was used. Bands were visualized using WesternBright ECL/peroxide and X-ray films (Advansta). The DcuS bands were quantified by measuring the area of the band and the staining intensities in scanned blots. Quantitative evaluation was performed using ImageJ software (69).

Data availability
Data are contained within the article or the Supporting information. Quantitative data from western blotting (ImageJ) are available from the corresponding author on request.   Figure S10. DcuS was modeled as a composition of the structures of PAS P , PAS C , and modeled TM2 and the linker from CL studies. UCSF Chimera was used to fuse the structures obtained by homology modeling (PAS P , PAS C ) and predicted structures (TM2, linker). The structure of the apo PAS P monomer was obtained using the structure of DcuS homolog CitA as a template (6, PDB ID:2V9A). The PAS C dimer structure was modeled using the structure of CitA as a template (PDB ID: 5FQ1). The helix dimer from Val172 to Glu213 was derived by structural analysis of the C-terminal region of PAS P (α6) and the N-terminal region of PAS C (α1) (3,9,34), and the CL data (18, this work) for V172 to E213. The cytoplasmic N-terminal coil, TM1, and the TM2-PAS C linker were predicted as single subdomains by the server I-TASSER (60). The kinase domain is not shown.