Structural insights into the mechanism of c-di-GMP–bound YcgR regulating flagellar motility in Escherichia coli

The motile-sessile transition is critical for bacterial survival and growth. Cyclic-di-GMP (c-di-GMP) plays a central role in controlling this transition and regulating biofilm formation via various effectors. As an effector of c-di-GMP in Escherichia coli and related species, the PilZ domain–containing protein YcgR responds to elevated c-di-GMP concentrations and acts on the flagellar motor to suppress bacterial motility in a brakelike fashion, which promotes bacterial surface attachment. To date, several target proteins within the motor, MotA, FliG, and FliM, along with different regulatory mechanisms have been reported. However, how YcgR acts on these components remains unclear. Here, we report that activated YcgR stably binds to MotA at the MotA-FliG interface and thereby regulates bacterial swimming. Biochemical and structural analyses revealed that c-di-GMP rearranges the PilZ domain configuration, resulting in the formation of a MotA-binding patch consisting of an RXXXR motif and the C-tail helix α3. Moreover, we noted that a conserved region in the YcgR-N domain, which is independent of MotA interaction, is necessary for motility regulation. On the basis of these findings, we infer that the YcgR-N domain is required for activity on other motor proteins. We propose that activated YcgR appends to MotA via its PilZ domain and thereby interrupts the MotA-FliG interaction and simultaneously interacts with other motor proteins via its YcgR-N domain to inhibit flagellar motility. Our findings suggest that the mode of interaction between YcgR and motor proteins may be shared by other PilZ family proteins.

neous flagellar counterclockwise (CCW) rotation forming the bundle, whereas clockwise (CW) rotation induces the cells to tumble upon bundle disassembly (14). This flagellar rotation is driven by a motor, which consists of at least 11 stators across the inner membrane (15), with four copies of MotA and two copies of MotB in each stator (16), along with a rotor at the bottom of the basal body, containing ϳ26 FliG proteins, 34 FliM proteins, and over 100 FliN proteins (17). The electrostatic interactions between the cytoplasmic domain of MotA and the C-terminal domain of FliG convert the chemical energy that is produced by the proton flow across the integral membrane channels of the stator into torque to drive the flagellar rotation (18). The flagellar rotation can be regulated via some regulators interacting with rotor proteins, such as phosphorylated CheY that binds to FliM and FliN to switch CCW rotation to CW in E. coli and Salmonella typhimurium (19,20) and EspE that binds to FliG as a clutch during biofilm formation in Bacillus subtilis (21,22). In contrast, YcgR is proposed to directly interact with different targets, the stator protein MotA or the rotor proteins FliG and FliM, in a brakelike fashion upon c-di-GMP binding by independent studies (7,11,12).
Boehm et al. (7) suggested that YcgR binds to MotA in flagellar basal bodies based on FRET assay in living cells. The FRET signal between fluorescently labeled MotA and YcgR is strong in the presence of high c-di-GMP concentration and weak in the low c-di-GMP concentration, suggesting that the MotA-YcgR interaction correlates with the cellular c-di-GMP concentration (7). They proposed that YcgR interacts with MotA at the MotA-FliG interface, thus interfering the MotA-FliG interaction and slowing down bacterial swimming velocity in a c-di-GMP-dependent manner (7). Paul et al. (11) identified that YcgR could pull down FliG and FliM in crude E. coli extracts. They observed that the mutation of residues in YcgR-N domain (N62W and K81D) weakened the binding ability of YcgR to FliG, not that to FliM, and that the mutation of residues in the PilZ domain (Q223P and I227W) weakened the binding ability to FliM, not FliG, and concluded that YcgR interacts with FliG and FliM via its YcgR-N and PilZ domains, respectively (11). They thus proposed a model that YcgR reduces the efficiency of torque generation and biases the flagella rotation by interacting with both FliG and FliM (11). Meanwhile, they did not rule out the MotA-YcgR interaction and actually proposed a potential MotA-YcgR interaction weaker than the binding of YcgR to FliG and FliM (11). Fang and Gomelsky (12) showed YcgR interaction with FliG based on pulldown and two-hybrid assays. They also observed an interaction of YcgR with FliM in twohybrid assays only when YcgR variant R118D was the partner pair. They proposed that YcgR binds to the central fragment of FliG where FliM binds, thus altering FliG-FliM interaction and biasing the flagella rotation (12). These studies make a great improvement for the knowledge of c-di-GMP regulating flagellar motility. However, these different target proteins do not complete the puzzle of YcgR-mediated motility regulation, and where and how YcgR binds to the motor still needs more investigation to clarify (23,24).
Previously, we undertook the crystallization and preliminary crystallographic analysis of E. coli YcgR in complex with c-di-GMP (25). Here, we find that YcgR could interact with the sta-tor protein MotA in vitro. We determine the crystal structure of YcgR from E. coli in complex with its ligand, c-di-GMP, and provide direct evidence that c-di-GMP activates YcgR to bind to MotA at the MotA-FliG interface via a MotA-interacting patch formed by residue Phe 117 in RXXXR motif and the C-tail helix ␣3 (residues 218 -240) in the PilZ domain. We also demonstrate that some conserved residues in the YcgR-N domain, which is independent of MotA binding, are required for motility regulation. Thus, we discuss how c-di-GMP-activated YcgR regulates the flagellar motility via motor proteins. Further, the interaction mode between YcgR and motor proteins could be shared by other PilZ family proteins acting on their targets.

YcgR stably binds to MotA with a molecular ratio of 1:4 in a c-di-GMP-dependent manner
To clarify the specific target(s) to which activated YcgR binds, we tested the interactions between YcgR and flagellar motor proteins of E. coli using analytical size-exclusion chromatography (SEC). The soluble cytoplasmic part of MotA (MotAc, residues 70 -170) was used in the assay, because MotA is a membrane-spanning protein (26,27). In the presence of ligand-free YcgR, MotAc was eluted at a volume of 13.71 ml, corresponding to an apparent molecular weight of 63.4 kDa and suggesting a MotAc tetramer in solution that is in line with the previously reported tetramers of full-length MotA and its transmembrane domain (16,28) (Fig. 1A). YcgR was eluted at a volume of 15.07 ml with an apparent molecular weight of 33.5 kDa (Fig. 1A). Both of them were eluted with elution volumes and apparent molecular weights similar to those of the individual proteins (Fig. S1). With the c-di-GMP-bound YcgR, MotAc was eluted at a volume of 12.87 ml, corresponding to an apparent molecular weight of 94.1 kDa and suggesting a MotAc tetramer bound to a single ligand-bound YcgR monomer with an apparent molecular weight of 30.9 kDa (Fig. 1B). Subsequent analysis using an SDS-PAGE assay and the relatively high absorbance at 254 nm (A 254 ) confirmed that the eluted MotAc was accompanied by both YcgR and c-di-GMP (Fig. 1B). In contrast, neither FliG nor the complex of FliM and FliN (FliMN) could be eluted together with c-di-GMP-free or -bound YcgR (Fig. S2), suggesting that the recombinant FliG and FliMN could not form a stable complex with either YcgR form in the SEC assay.
The molecular weights of those protein samples have also been assayed using an analytic analytical ultracentrifuge (AUC). The results of AUC assays indicated a molecular weight of 30.7 kDa for c-di-GMP-bound YcgR (Fig. 1C), a monomer in line with the result of the SEC assay (Fig. S1). The result of MotAc showed two peaks, 22.3 and 54.4 kDa (Fig. 1C), implying a dimer-tetramer equilibration different from that observed in the SEC assay. The result of the MotAc-YcgR-c-di-GMP complex showed three peaks, 19.4, 33.1, and 86.7 kDa (Fig. 1C), suggesting the existence of MotAc dimer, YcgR-c-di-GMP, and the complex of MotAc tetramer and YcgR. These results indicated that MotAc could interact with YcgR-c-di-GMP.
To further obtain overall information of the complex of MotAc-YcgR-c-di-GMP, we performed small angle X-ray scat-Molecular mechanism of YcgR acting on MotA tering assays (SAXS). The molecular masses of the proteins were calculated through Guinier analysis based on the scattering intensity extrapolated to zero angle (I(0)). All of them (YcgR, 27.7 kDa; YcgR-c-di-GMP, 27.6 kDa; MotAc, 55.6 kDa; MotAc-YcgR-c-id-GMP, 74.2 kDa) ( Fig. 1D) were similar to those estimated by analytical SEC and AUC. The maximum dimension D max of each sample was obtained using the pair distance distribution function P(r) (Fig. 1D and Fig. S3). The ab initio model of YcgR, YcgR-ci-GMP, MotAc, and the complex of MotAc and YcgR-c-di-GMP were built using DAMMIF (29) (Fig. 1D and Fig. S3). These results further corroborate that YcgR binds to MotAc in a c-di-GMP-dependent manner.

c-di-GMP binding is essential for YcgR interacting with MotA and regulating motility
In view of the low sequence identities between YcgR from E. coli and other PilZ domain-containing proteins whose structures have been determined (i.e. 18.22% for PP4397 from Pseudomonas putida (30), 16.40% for MrkH from Klebsiella pneumoniae (31,32), and 15.42% for MotI from Bacillus subtilis (33)), we determined the crystal structure of YcgR-c-di-GMP complex from E. coli at a resolution of 2.30 Å (data collection and refinement statistics given in Table S1) to investigate the mechanism of c-di-GMP regulating the binding ability of YcgR to the motor. The structure indicated that YcgR folds into two individual domains, an N-terminal YcgR-N domain and a C-terminal PilZ domain ( Fig. 2A and Fig. S4), similar to that observed in the structures of the homologs PP4397 (30), MrkH (31,32), and MotI (33). The YcgR-N domain (residues 1-110) possesses a ␤-barrel fold, formed by single five-stranded and four-stranded anti-parallel ␤-sheets and clamped by two helices, ␣1 and ␣2, across the top and bottom ( Fig. 2A). The PilZ domain (residues 111-244) shares a similar ␤-barrel fold, with only a helix ␣3 (C-tail helix) situated over the top ( Fig. 2A). This domain contains two conserved c-di-GMP binding motifs, 114 RXXXR 118 and 145 (D/N)XSXXG 150 , comparable with other PilZ domains (10). The two ␤-barrels are bridged by a loop comprising residues 111-119, which contains the conserved motif RXXXR. Two mutually intercalated c-di-GMP molecules bind to this loop as well as the (D/N)XSXXG motif and fit into the groove formed by the two domains ( Fig. 2A).
The structure showed that residues Arg 113 , Arg 114 , Arg 118 , Asp 145 , Ser 147 , Arg 208 , Ser 190 , and Ser 210 are involved in a hydrogen bond network to bind c-di-GMP (Fig. 2B). The roles of them were assessed using mutagenesis experiments. The c-di-GMP affinity of YcgR was determined using an isothermal titration calorimetry (ITC) assay, with a disassociation constant K d of 0.141 M and a stoichiometric ratio of 2 ( Fig. 2C and Fig.  S5A). Mutants R114A, R118A, and D145A retained very limited c-di-GMP affinity ( Fig. 2C and Fig. S5A), did not inhibit bacterial motility (Fig. 2D), and did not bind to MotAc in the analyt- ical SEC assay (Fig. 2E). Notably, a previous study reported that c-di-GMP levels caused by the yhjH deletion were not enough for the binding of BcsA, which binds c-di-GMP with a K d of 8.2 M, to trigger cellulose production (3). We therefore suggested that the c-di-GMP concentration in strain ⌬yhjH could not activate those YcgR mutants, because they maintained similar c-di-GMP affinities to BcsA (R114A with K d of 21.1 M, R118A with K d of 14.33 M, and D145A with K d of 3.98 M; Fig. S5A). Mutants R113A and R208A maintained low c-di-GMP affinity ( Fig. 2C and Fig. S5A) and approximately half of the motility inhibition ability compared with the WT (Fig. 2D), although they were observed to bind to MotAc in the SEC assay with elution plots similar to that of the WT (Fig. 2E). These results suggest that the loss or decrease of c-di-GMP affinity for YcgR leads to the loss or decrease of its ability to interact with MotA and inhibit the motility. Mutant S147A retained a somewhat high c-di-GMP affinity ( Fig. 2C and Fig. S5A) and still bound to MotAc (Fig. 2E), yet only exhibited a fraction of the motility inhibition ability (Fig. 2D). It is possible that the mutation decreased the stability of YcgR and then impaired the motility. Meanwhile, considering that only a single c-di-GMP binds to S147A ( Fig. 2C and Fig. S5A) and that the mutation confers a SEC profile different from that of WT YcgR, we believe that the c-di-GMP dimer binding to YcgR is necessary for maintaining a proper MotAc-binding region and then for motility regulation. Therefore, we conclude that the c-di-GMP binding is essential for YcgR interaction with MotA and regulation of the motility and that the conserved residues Arg 113 , Arg 114 , Arg 118 , Asp 145 , and Ser 147 preferentially function in c-di-GMP recognition rather than MotA interaction.

Molecular mechanism of YcgR acting on MotA An individual PilZ domain is sufficient to interact with MotA upon c-di-GMP binding, whereas both YcgR-N and PilZ domains are requisite for motility regulation
To unravel the roles of the YcgR-N and PilZ domains in the YcgR-MotA interaction pattern, we obtained both the crystal structure of individual YcgR-N and that of PilZ in complex with c-di-GMP (data collection and refinement statistics given in Table S1). These structures are almost identical to those of the two domains in the YcgR-c-di-GMP structure (Fig. 3A), suggesting that the YcgR-N domain would not undertake a significant conformational change upon the binding of c-di-GMP. The roles of the individual YcgR-N and PilZ domains were assessed using the c-di-GMP-binding, MotAc-interacting, and bacterial motility assays. The YcgR-N domain did not bind c-di-GMP in vitro and did not form a stable complex with MotAc in the SEC assay ( Fig. S5B and Fig. 3C). Conversely, the PilZ domain bound c-di-GMP, as reported previously (8), with somewhat higher affinity (K d of 0.099 M) than YcgR (Fig. S5B). Notably, it could bind to MotAc upon c-di-GMP binding (Fig. 3C). Nevertheless, neither domain could restore the motility inhibition ability of YcgR (Fig. 3B). These results clearly indicate that MotA binding is not sufficient for the motility inhibition and that both YcgR-N and PilZ domains are required for motility regulation.

c-di-GMP binding induces the RXXXR motif and the C-tail helix to form a MotA-interacting patch
To identify which residues directly participate in the interaction between YcgR and MotA, we performed multiple-sequence alignment and found two conserved regions in the PilZ domain including residues Gln 112 and Phe 117 in the RXXXR motif and residues Glu 221 , Arg 222 , Gln 225 , Arg 226 , Ile 228 , Phe 229 , Glu 232 , and Glu 234 in the C-tail helix (Fig. 4A). To reveal the local conformational change of these residues induced by c-di-GMP, we attempted to solve the ligand-free structure; however, this was not successful. Thus, we studied the global conformational change using SAXS assay (Fig. 4B) and built a ligand-free rigid-body model (Fig. S4B). Superposition of the ligand-bound and -free models shows a large-scale domain movement (Fig.  4C), which has been also observed in YcgR homologs (30,34). A close inspection indicated that c-di-GMP anchors the side chains of Arg 113 , Arg 114 , Arg 115 , and Arg 118 (Fig. 4C, pink box), exposes the side chain of Phe 117 , and triggers a rearrangement of the RXXXR motif and the C-tail helix ␣3 to form a potential binding region for partners (Fig. 4C, blue box). These residues were then examined using mutagenesis experiments to assay their roles in c-di-GMP affinity, MotA-interacting ability, and motility regulation.
Mutant F117A presented somewhat high affinity to c-di-GMP (Fig. S5C) but lost the ability to interact with MotA and

Molecular mechanism of YcgR acting on MotA
inhibit motility (Fig. 5, A and B), revealing that Phe 117 is directly involved in the MotA interaction, which is essential for motility regulation. Mutant Q112A maintained the c-di-GMP affinity (Fig. S5C) and interacted with MotAc (Fig. 5B). However, it lost the majority of its motility inhibition ability (Fig. 5A), possibly because Gln 112 is close to the YcgR-MotA interface. The roles of conserved residues Glu 221 , Arg 222 , Gln 225 , Arg 226 , Ile 228 , Phe 229 , Glu 232 , and Glu 234 in the C-tail helix were subsequently checked. The substitution of residues at the immediate vicinity of Phe 117 (Fig. 4C), residues Ile 228 , Phe 229 , or Glu 232 , to alanine abolished the bacterial motility inhibition function of YcgR, whereas the substitution of the other residues did not (Fig. 5A). In addition, mutants I228A, F229A, and E232A also abolished the MotAc-binding ability and maintained the c-di-GMP affinity ( Fig. 5B and Fig. S5C), indicating that these three residues, similar to Phe 117 , are directly involved in the interaction with MotA and subsequently in mobility inhibition. These findings demonstrate that residue Phe 117 in the ligand binding motif RXXXR and residues Ile 228 , Phe 229 , and Glu 232 in the C-tail helix form a MotA interacting patch upon c-di-GMP binding, and this patch is essential for YcgR-MotA interaction and motility regulation.
The YcgR interacting site of MotA was consequently investigated. It has been reported that the YcgR-interacting site is in or close to the MotA-FliG interface (residues Arg 90 -Glu 98 of MotA) and includes residues Gly 93 and Ser 96 (7). However, the mutations of those MotA suppressors (e.g. G93R and S96L) resulted in the aggregation of MotAc. A mutagenesis scanning of the MotA-FliG interface (i.e. residues Arg 90 -Glu 98 ) was performed. The SEC assay indicated that mutants R90A, M92A, . c-di-GMP induces the rearrangement of YcgR domains. A, distribution of conserved residues generated by ConSurf Server (54). The conserved residues are labeled, among which the buried residues are marked in blue. The sequence consensus of RXXXR motif and C-tail helix are shown using "Weblogo" (55) at the bottom. The residue numbers below the "Weblogo" are according to YcgR from E. coli. The acidic, basic, nonpolar, and polar residues are shown in red, blue, black, and yellow, respectively. B, SXAS assay of YcgR and YcgR-c-di-GMP. In the left panel, the experimental spectra of SAXS data are shown in gray. The shape reconstruction of YcgR and YcgR-c-di-GMP by DAMMIF (29) (brown) show good agreement with the theoretical data calculated from the YcgR model constructed by SASREF (purple) and the crystal structure of YcgR-c-di-GMP (green). In the right panel, the most probable envelope (gray) of ligand-free/bound YcgR predicted by DAMMIF was superimposed with the YcgR model and the structure of YcgR-c-di-GMP. C, superposition of the c-di-GMP-bound crystal structure (green) and the ligand-free model built by SASREF (purple). The residues in and around the RXXXR motif are boxed in pink dotted lines, and the residues involved in the MotA-interacting patch are boxed in blue dotted lines. The residues forming the MotA-interacting patch are highlighted in blue.

Molecular mechanism of YcgR acting on MotA
G93A, M94A, F95A, and E98A lost their YcgR-binding abilities (Fig. 6), whereas Q91A and S96A remained (Fig. 6). Taken together, YcgR is activated by c-di-GMP to tightly interact with MotA at MotA-FliG interface via the patch consisting of RXXXR motif and C-tail helix in the PilZ domain.

A conserved region of the YcgR-N domain is necessary for regulating motility
The previous results confirmed that both the YcgR-N and PilZ domains are essential for motility regulation. We investigated the possible role of YcgR-N and key residues involved in motility regulation. Sequence alignment of YcgR similarities revealed residues Gln 38 , Ile 40 , Leu 44 , Asp 54 , and Asn 62 as conserved (Fig. 4A). In the structure, they are located at the same side of the YcgR-N ␤-barrel, distant from the MotA-binding patch (Fig. 7A). The substitution of residue Gln 38 , Ile 40 , Leu 44 , Asp 54 , or Asn 62 to alanine decreased the mobility inhibition ability (Fig. 7B), implying their roles in motility regulation. In comparison, the triple-mutant Q38A/D54A/N62A (QDN/ AAA) displayed lower motility inhibition ability than the single mutants (Fig. 7B), but retained the same c-di-GMP affinity (Fig.  S5D) and MotA interacting ability as the WT (Fig. 7C), revealing that these residues are independent from c-di-GMP and MotA binding. The results suggest that these residues may be involved in interactions with other proteins (e.g. the rotor proteins FliG and FliM described in previous research (11,12)) and then regulation of flagella rotation.

Discussion
Cellular levels of c-di-GMP are related to controlling the transition of bacteria from motile to sessile. YcgR, as a PilZcontaining protein in E. coli and related species, responds to the elevated c-di-GMP concentration by reducing the flagella motor output, slowing down the swimming velocity in a brakelike fashion and facilitating bacterial surface attachment and subsequent biofilm formation (4). Previously, three independent studies suggested the motor proteins MotA, FliG, and FliM as the YcgR partners (7,11,12). These studies proposed different models where YcgR interacted with different targets (i.e. MotA (7), both FliG and FliM (11), or individual FliG (12)) to regulate motility and suggested that activated YcgR alters the MotA-FliG (7,11) or FliG-FliM (12) interaction to decrease the flagellar rotation speed and induce CCW motor bias. These observations suggest that YcgR possibly interacts with multiple proteins within the motor to control the motor output and bias flagellum rotation. Here, we provide novel evidence that both the YcgR-N and PilZ domains are necessary for c-di-GMP regulating flagellar motility. We identify that YcgR could interact with the stator protein MotA via its PilZ domain and propose that a conserved region formed by Gln 38 , Ile 40 , Leu 44 , Asp 54 , and Asn 62 in the YcgR-N domain contributes to the interaction with some other motor proteins.
MotA is proposed as the YcgR partner in the study of Boehm et al. (7), where c-di-GMP-activated YcgR interacts with MotA and reduces flagellar motor function in a brakelike fashion. Previous studies also reported the interaction of two MotA homologues, MotI from B. subtilis (33) and FlgZ from Pseudomonas aeruginosa (35), with MotA or a MotA homolog MotC. Our study provides direct evidence that YcgR interacts with MotA. We show that YcgR could interact with MotA via its PilZ domain to form a 1:4 complex in the presence of c-di-GMP (Fig.  1B), suggesting that one stator complex, MotA 4 MotB 2 , only binds one YcgR molecule. We determined the crystal structure of YcgR-c-di-GMP complex ( Fig. 2A). Because the ligand-free YcgR was too difficult to crystallize, we proposed a ligand-free

Molecular mechanism of YcgR acting on MotA
YcgR model using a SAXS assay based on the crystal structure of YcgR-N domain and the homology model of ligand-free PilZ domain (Fig. 4B). Our structural and SEC assays indicate that c-di-GMP binding induces the RXXXR motif and C-tail helix to form a MotA-interacting patch (Fig. 4C). The residues Phe 117 , Ile 228 , Phe 229 , and Glu 232 that constitute the patch are conserved across different bacteria ( Fig. 4A and Fig. S4), suggesting that YcgR similarities share a common mode of c-di-GMP-activated interaction with MotA. Meanwhile, we identify that residues Arg 90 , Met 92 , Gly 93 , Met 94 , Phe 95 , and Glu 98 of MotA, which situate at the MotA-FliG interface, are involved in the interaction between MotA and YcgR (Fig. 6). This result is in line with the previous observation that mutation G93R of MotA impaired the YcgR inhibition ability (7) and confirms that YcgR binds to MotA at the MotA-FliG interface.
However, YcgR binding to MotA at the MotA-FliG interface is not sufficient for motility regulation, supported by malfunction of individual PilZ domain (Fig. 3B) as well as YcgR mutants harboring the alanine substitution of conserved residues in YcgR-N domain (Fig. 7B). Some conserved residues in the YcgR-N domain, including Gln 38 , Ile 40 , Leu 44 , Asp 54 , and Asn 62 , are involved in motility regulation. The triple mutation QDN/AAA impaired motility regulation ability (Fig. 7B) yet kept the similar affinity with c-di-GMP and still formed a stable complex with MotAc ( Fig. S5D and Fig. 7C). It has been reported that the rotor proteins FliG and FliM are also the targets of YcgR (11,12); thus, we infer that these residues consist of an interaction region for them. Residue Asn 62 has been proposed as the interacting site for FliG (11). Therefore, we thought this conserved region in the YcgR-N domain might be responsible for FliG binding. The other reported residue involved in FliG interaction, Lys 81 (11), actually interacted with c-di-GMP according to our structure and thus was eliminated from the FliG-interacting region. Previously, the role of YcgR in motility regulation was proposed to be conserved because YcgR similarities could be found across many proteobacteria (10,11). Nevertheless, some YcgR similarities do not harbor these conserved residues in the YcgR-N domain, such as MotI from firmicutes (Fig. S7). MotI is proposed to sequester MotA from FliG via binding to MotA in a clutchlike fashion (33) and share the similar folding with YcgR, with an root mean square deviation of 4.72 Å for overall ␣-carbon positions. These findings suggest that YcgR in different bacteria could interact with various targets besides MotA and recruit different motility regulation mechanisms.
YcgR or its mutant R118D could interact with FliM, which has been proposed by Paul et al. (11) and Fang et al. (12). We have defined the roles of conserved motifs of YcgR (i.e. RXXXR and (D/N)XSXXG motifs for c-di-GMP binding, RXXXR motif and C-tail helix ␣3 for MotA binding, and the conserved region of the YcgR-N domain for FliG interacting). We did not find any structural and bioinformatic clues of residues potentially important for direct FliM binding. Fang and Gomelsky proposed that only the YcgR-R118D mutant interacts with FliM (12). Based on our structure and affinity assay, residue Arg 118 is critical for c-di-GMP binding of YcgR; hence, its mutation presumably impairs the FliM interaction due to the loss of the c-di-GMP affinity. Thus residue Arg 118 or its substitution does not seem to interact with FliM directly. Paul et al. (11) proposed that helix ␣3 is important for FliM binding. However, our findings show that this helix is vital in MotA binding, and the key residues, such as Ile 227 , reported in the previous study were not assayed here because the side chain of Ile 227 is buried at the protein interior and could not function as an exterior interacting site. The mutation of I227W possibly results in the

Molecular mechanism of YcgR acting on MotA
improper protein fold, which could cause a low FliM (as well as MotA/FliG) affinity and a normal motor speed and bias of the strain harboring the I227W mutant (11). Thus, we could not evaluate the role of YcgR interaction with FliM. Additional experimentation will be required to explain the interaction between YcgR and FliM, along with its role in motility regulation.
Considering that both YcgR-N and PilZ domains are necessary, we propose that activated YcgR simultaneously interacts with the stator and rotor via two domains to regulate the flagellar motility. In detail, c-di-GMP binding rearranges the PilZ domain configuration to assemble a MotA-binding patch, including the RXXXR motif and C-tail helix ␣3. Activated YcgR binds to MotA at the MotA-FliG interface, which should increase the resistance and decrease energy transfer between the stator and rotor, thus slowing down bacterial swimming velocity (Fig. 8). The YcgR-N domain is inferred to interact with other motor proteins, such as FliG. Although some details still need further study to fill in, for example, how YcgR biases the flagellar rotation, our study clarified one of the key questions in YcgR-mediated motility regulation, how YcgR interacts with its target MotA. We suggest that YcgR and its similarities could interact with MotA via activated PilZ domain in a common mode and that, further, they could be involved in motility regulation via diverse mechanisms, like a brake or a clutch, depending on the YcgR-N domain. Our study paves the way for  Fig. 4A. B, bacterial motility assay of strains harboring YcgR mutants (fused to a C-terminal HA tag) involved in the conserved region of the YcgR-N domain. The expressions of the mutants in vivo were detected using anti-HA antibody and shown in Fig. S6D. The histogram depicts the mean and error bars representing S.D. from more than five independent tests. C, the MotAc-interacting ability of c-di-GMP (cdG)-free (left) and -bound (right) QDN/AAA assayed by SEC. The elution fractions were assayed by Coomassie-stained SDS-PAGE, as shown at the bottom of the chromatograms. The lines below the SDS-polyacrylamide gel represent the elution volumes of fractions.

Molecular mechanism of YcgR acting on MotA
further study on YcgR-mediated flagellar motility and provides new clues for investigating other PilZ proteins.

Bacterial strains and plasmids
The bacterial strains and plasmids used in this study are listed in Table S2. The genes of YcgR, FliG, FliM, and FliN were cloned from E. coli strain MG1655. The full-length YcgR, the N-terminal YcgR-N domain (residues 1-110), and the C-terminal PilZ domain (residues 111-244) constructs used for expression and purification were generated by subcloning in pET-22b plasmid with a His 6 tag fused to their C terminus. The sequences encoding the soluble cytoplasmic part of MotA (MotAc, residues 70 -170 of MotA) and the full-length of rotor protein FliG were also cloned into the pET-22b plasmid with a C-terminal His 6 tag. Because the rotor protein FliM was inclined to aggregation in cells (36), the DNAs encoding FliM and FliN were cloned into two ORFs of co-expression plasmid pETDuet-1 with an N-terminal His 6 tag of FliM to generate FliMN. The constructs involved in motility assay were generated by subcloning in pBBR1MCS-3 (37) plasmid fused with a C-terminal HA tag. Site-directed mutants were introduced using the standard QuikChange site-directed mutagenesis kit protocol (38). The correctness of the constructs was confirmed by DNA sequencing.
The SeMet derivative of YcgR was expressed in E. coli B834 (DE3) and grown in LB medium at 37°C to an A 600 of 1.0. The cells were harvested, resuspended in M9 minimal medium, and starved of methionine for 1 h. Then 0.25 M SeMet (Anatrace) and 0.05 mM IPTG was added to the culture, and the cells were incubated overnight for 16 h at 16°C. The SeMet derivative of YcgR was purified using the same buffer as above with additional 5 mM ␤-mercaptoethanol in Ni-NTA affinity buffer and 2 mM DTT in SEC buffer. All proteins were confirmed by SDS-PAGE, concentrated by ultrafiltration, and immediately used for crystallization or frozen and stored at Ϫ80°C.

Analytical ultracentrifugation
Purified YcgR-c-di-GMP, MotAc, and MotAc-YcgR-c-di-GMP were concentrated to ϳ1 mg ml Ϫ1 in buffer containing 20 mM HEPES, pH 7.5, and 150 mM NaCl. An AUC assay was carried out on a Beckman Coulter Optima XL-I analytical ultracentrifuge. The experiment was performed at 20°C at a speed of 60,000 rpm for 7 h. Raw data were processed using Sednterp, and the sedimentation coefficients and apparent molecular masses were calculated (39).

Crystallization, data collection, and structure determination
Prior to crystallization trials, purified YcgR and PilZ from E. coli were mixed with c-di-GMP (BioLog) at a 3:1 ligand/protein molar ratio. Crystallization was performed using the vapor diffusion hanging-drop method at 20°C. The protein concentrations of YcgR-c-di-GMP, YcgR-N, and PilZ-c-di-GMP used in crystallization were 17, 8, and 8 mg ml Ϫ1 , respectively. The best crystals of YcgR-c-di-GMP were obtained after 3 days in 0. Prior to flash freezing in liquid nitrogen, the crystals were soaked for 30 -60 s in the mother liquor solution containing 10 -25% (v/v) glycerol. Diffraction data of YcgR-c-di-GMP were collected to a resolution of 2.3 Å using an ADSC Q315r CCD detector at beamline BL-17A, KEK (Photon Factory, Tsukuba, Japan) with a wavelength of 0.9791 Å. The data were integrated and scaled by MOSFLM (40) and SCALA from the CCP4 software suite (41). The space group was determined to be R3:H with one molecule per asymmetric unit. The data of SeMet-derivative crystals were collected for experimental phasing at beamline BL17U1 (42), Shanghai Synchrotron Radiation Facility (SSRF), using an ADSC Q315r CCD detector with a wavelength of 0.9793 Å. The data were processed by MOS-FLM and SCALA. The phasing and the initial model building were performed by Phenix Autosol (43) using the single-wavelength anomalous diffraction method. The model was then used in Phaser from the CCP4 software suite for molecular replacement against the native data. The model was improved by iterative cycles of refinement and manual building using CNS (44), Phenix.refine (45), and Coot (46). The final model contained 226 amino acids, except the residues 123-124, 156 -160, and 195-203 between two ␤-strands and 244 at the C terminus because of poor electron density, and two c-di-GMPs intercalated to each other.
Diffraction data of PilZ-c-di-GMP were collected at beamline BL17U1 (42) at SSRF with a wavelength of 0.9793 Å and also processed with MOSFLM (40) and SCALA (41). The crystals belonged to space group P3 1 21 and contained one copy in the asymmetric unit. The structure was solved at 2.3 Å by molecular replacement with Phaser_MR in the Phenix suite using the coordinates of C-terminal domain structure of YcgRc-di-GMP as the search model. The final structure was obtained through iterative cycles of manual building and refinement by Phenix.autobuild (43), Phenix.refine (45), and Coot (46). The model consisted of 129 amino acids, including the missing part in the YcgR-c-di-GMP model and two intercalated c-di-GMPs.
Diffraction data of YcgR-N were collected with a wavelength of 1.5418 Å using a Rigaku MM007 CCD944ϩHG system at the National Laboratory of Biomacromolecules, Institute of Biophysics, and Chinese Academy of Sciences. The data were processed by MOSFLM (40) and SCALA (41). The crystals diffracted to 1.77 Å resolution and also belonged to the trigonal space group R3:H. The N-terminal domain structure of YcgR-c-di-GMP was used for molecular replacement with Phaser_MR. The refinement and manual model building were taken by Phenix.refine (45) and Coot (46). The model contained 110 amino acids, including residues 4 -111, leucine, and glutamic before the His tag.
The statistics of data collection and refinement are summarized in Table S1. The structural figures were prepared with PyMOL.

Binding studies by size-exclusion chromatography
Analytical SEC experiments were performed at 16°C on a Superdex 200 10/300 GL column (GE Healthcare) to determine the interactions between YcgR and flagellar motor proteins. Purified YcgR was mixed with MotAc, FliG, or FliMN at a 1:1 molar ratio, incubated for 30 min on ice, and then loaded on the column in a final volume of 500 l. For the YcgR-c-di-GMP complex, 3-fold excess of c-di-GMP were first added to YcgR solution on the ice for 30 min, and then the complex was incubated with flagellar motor proteins for another 30 min. The samples were eluted in SEC buffer at a flow rate of 0.5 ml min Ϫ1 with monitoring of the absorbance at 280 and 254 nm. The eluted peaks were analyzed on SDS-polyacrylamide gels and stained with Coomassie Blue. The apparent masses were determined using Gel Filtration Calibration Kits (GE Healthcare) with the standard procedure on the Superdex 200 10/300 GL column. PilZ, YcgR-N, and the mutant derivatives of YcgR were assayed in the same way.

Isothermal titration calorimetry assays
ITC experiments were performed using a MicroCal iTC-200 calorimeter (Malvern) at 25°C. c-di-GMP was dissolved and diluted in the SEC buffer. Each titration consisted of 21 or 31 injections, the first one 0.5 l (not used in data fitting) and all subsequent injections of a 2-or 1.2-l volume. 100 -200 M c-di-GMP solution was titrated into 5-10 M YcgR (WT), PilZ, Q112A, QND/AAA, F117A, I228A, F229A, or E232A solutions because of the relatively high c-di-GMP affinity of those proteins. For those proteins with relatively low c-di-GMP affinity, such as R113A, R114A, R118A, D145A, S147A, R208A, or YcgR-N, 1-2 mM c-di-GMP solution was injected into the sample cell containing 30 -70 M protein solution. The titration data were integrated and fitted to a one-site model using the Origin program provided by the manufacturer. The binding constant (K a , K d ϭ 1/K a ) and the stoichiometry (n) were extracted directly from the fit.

Small-angle X-ray scattering measurement and modeling
Synchrotron SAXS data of YcgR, YcgR-c-di-GMP, MotAc, and the complex of MotAc and YcgR-c-di-GMP were collected on beamline BL19U2 at the National Center for Protein Science Shanghai (NCPSS) at 4°C. Scattering curves were recorded at a wavelength of 1.03 Å on a Pilatus 1M detector over an angular range of 0.013 Ͻ q Ͻ0.4546 Å Ϫ1 , where q ϭ (4sin)/, and 2 is the scattering angle. The buffer used in SAXS experiment was the SEC buffer. Twenty consecutive frames of 1-s exposure time were recorded and averaged. The raw data were processed using PRIMUS (47) for background subtraction, concentration scaling, and curve merging. The data were checked for radiation damage and concentration-induced aggregation. The curve of 2 mg ml Ϫ1 YcgR and the curve of low-q data from 0.6 mg ml Ϫ1 YcgR-c-di-GMP merged with higher-q data from 4.8 mg ml Ϫ1 YcgR-c-di-GMP, the curve of low-q data from 5 mg ml Ϫ1 MotAc merged with higher-q data from 10 mg ml Ϫ1 MotAc, and the curve of 4 mg ml Ϫ1 MotAc-YcgR-c-di-GMP were used in subsequent analysis. Guinier analysis and Primus distance distribution analysis of Primus qt from the ATSAS program suite (48) were performed to calculate the radius of gyration (R g ), scattering intensity extrapolated to zero angle I(0), maximum dimension D max , and pair distance distribution function P(r). Twenty independent ab initio models were com-Molecular mechanism of YcgR acting on MotA puted with the simulated annealing ab initio bead-modeling programs DAMMIF (29). The 20 models were compared using DAMSEL from the ATSAS program suite, and the most probable one was chosen to be the final SAXS envelope. The simulated scattering curves fitted to the experimental data with a discrepancy factor 2 of 0.841 for YcgR-c-di-GMP, 2 of 1.032 for YcgR, 2 of 0.910 for MotAc, and 2 of 1.057 for MotAc-YcgR-c-di-GMP (Fig. S3). The fit of the theoretical scattering curve between the crystal structure of YcgR-c-di-GMP and the experimental data was obtained using CRYSOL (49) with 2 of 1.231. The SAXS envelope of YcgR-c-di-GMP was superposed to the crystal structure of YcgR-c-di-GMP by SUPCOMB (50), and the superimposed model was provided using PyMOL.
The YcgR model was constructed according to the SAXS experimental data. The molecular model ensemble that best fitted the SAXS data was performed with SASREF (51), which generated molecular models from domains or subunits with known structure by rigid-body movements and rotations. The crystal structure of YcgR-N and the PilZ model without c-di-GMP, which was simulated from the C-terminal domain of the crystal structure of MrkH (Protein Data Bank code 5KED) (31) using the automated protein structure homology-modeling server SWISS-MODEL (52,53), were used in SASREF. Contacts were defined to guarantee the connection between the C terminus of YcgR-N and the N terminus of PilZ. The SASREF modeling was repeated more than 10 times to obtain the best model that fitted the SAXS scattering data with 2 of 1.595. The YcgR SASREF model and the SAXS envelope of YcgR were superimposed by SUPCOMB (50) and visualized using PyMOL.

Motility assays
DNAs encoding YcgR-N, PilZ, YcgR, and its derivatives were cloned into pBBR1MCS-3 and transformed into ⌬yhjHycgR (12). Swimming motility was assayed on soft agar plates containing 0.3% agar, 1% tryptone, and 0.5% NaCl. 1.5 l of cell culture grown overnight at 37°C in LB medium was spotted onto the freshly prepared plates and incubated for 6 h at 37°C. The cultures were normalized to the same cell density of 2.0 at 600 nm. The diameters of the swimming zone around the inoculation sites were measured. At least five independent experiments were performed, and the arithmetic means Ϯ S.D. were plotted by GraphPad Prism 6.
The expression levels of YcgR variants fused with a C-terminal HA tag in vivo were assayed using anti-HA antibody. The bacteria containing the plasmids of YcgR mutants were normalized to the same cell density of 2.0 at 600 nm, harvested using centrifugation, and lysed by BugBuster (Novagen) protein extraction reagent. The supernatants were examined by SDS-PAGE (15% acrylamide), transferred to nitrocellulose membrane, and immunoblotted with an anti-HA antibody (Cell Signaling Technology). The exposure intensity of the bands was calculated using ImageJ.

Accession numbers
Atomic coordinates and structure factors of YcgR-c-di-GMP, PilZ-c-di-GMP, and YcgR-N have been deposited in the Protein Data Bank with accession codes 5Y6F, 5Y6G, and 5Y6H, respectively. The SAXS data and structural models have been deposited at SASBDB with accession numbers SASDEB9 (YcgR) and SASDEC9 (YcgR-c-di-GMP).