Allosteric regulation of a prokaryotic small Ras-like GTPase contributes to cell polarity oscillations in bacterial motility

Mutual gliding motility A (MglA), a small Ras-like GTPase; Mutual gliding motility B (MglB), its GTPase activating protein (GAP); and Required for Motility Response Regulator (RomR), a protein that contains a response regulator receiver domain, are major components of a GTPase-dependent biochemical oscillator that drives cell polarity reversals in the bacterium Myxococcus xanthus. We report the crystal structure of a complex of M. xanthus MglA and MglB, which reveals that the C-terminal helix (Ct-helix) from one protomer of the dimeric MglB binds to a pocket distal to the active site of MglA. MglB increases the GTPase activity of MglA by reorientation of key catalytic residues of MglA (a GAP function) combined with allosteric regulation of nucleotide exchange by the Ct-helix (a guanine nucleotide exchange factor [GEF] function). The dual GAP-GEF activities of MglB accelerate the rate of GTP hydrolysis over multiple enzymatic cycles. Consistent with its GAP and GEF activities, MglB interacts with MglA bound to either GTP or GDP. The regulation is essential for cell polarity, because deletion of the Ct-helix causes bipolar localization of MglA, MglB, and RomR, thereby causing reversal defects in M. xanthus. A bioinformatics analysis reveals the presence of Ct-helix in homologues of MglB in other bacterial phyla, suggestive of the prevalence of the allosteric mechanism among other prokaryotic small Ras-like GTPases.


Introduction
Small Ras-like GTPases are ubiquitous in eukaryotes and perform varied functions, including cell signaling and motility. Recently, such GTPases have been identified in prokaryotes too [1,2]. The soil bacterium Myxococcus xanthus exhibits 2 types of motility-adventurous gliding motility and social motility, which are essential for its normal life cycle [3]. The localization of motility complexes, and thereby the direction of movement, is determined by an oscillatory system, which includes Mutual gliding motility A (MglA), a small Ras-like GTPase, and Mutual gliding motility B (MglB), its GTPase activating protein (GAP) as major components [1,4,5]. Mutual gliding motility (mgl) was identified as a locus essential for both social and adventurous motilities in M. xanthus, comprising the 2 genes mglA and mglB [1,3]. Studies on MglA and MglB showed that these proteins exhibit a mutually exclusive localization pattern [4,6,7]. MglA localizes to the leading pole presumably in its GTP-bound state [4], whereas MglB localizes at the lagging pole (Fig 1). These studies [1,[3][4][5][6][7] established the role of the prokaryotic small Ras-like GTPase in cell polarity determination and motility, which are functions analogous to its eukaryotic counterparts [8,9]. A remarkable feature of Myxococcus motility is the frequent reversals in its cell polarity in response to environmental cues [10]. M. xanthus changes its direction of movement by swapping the leading and lagging poles. The reversal frequency is regulated by the chemosensorylike frizzy (Frz) pathway, which relays environmental signals to the downstream MglA-MglB polarity control system [11][12][13] (Fig 1). High-resolution time-lapse experiments revealed that MglA and MglB relocalize sequentially during a reversal [4,7,11]. Importantly, the Frz system does not establish polarity, but its activity is required for inversion of the polarity axis. Another protein involved in the polarity control module is RomR, a protein with a response regulator receiver-like domain [14,15]. It was recently shown that RomR and its interacting partner RomX are essential for recruiting MglA-GTP to the poles [16]. Frz-signaling and RomR stimulates the pole-to-pole exchange of MglA and MglB [11,[13][14][15]. All together, these proteins form a so-called gated relaxation oscillator, which functions to drive the polarity reversals in response to environmental signals [11].
Regulated reversals are centrally controlled by the MglA GTPase cycle. Though it is established that MglB functions as a GAP for MglA [4,7], proteins that function as guanine nucleotide exchange factors (GEFs) [17] for MglA were not identified until very recently [16]. The complex of RomR and RomX has been recently demonstrated to function as a GEF for MglA [16]. In vivo, the RomRX complex recruits MglA to the leading pole at the time of reversal. Between reversals, RomRX dissociates slowly and relocalizes to the lagging cell pole where RomR interacts with MglB [14,15]. Thus, both GAP and GEF activities accumulate at the lagging cell pole, preparing the next reversal event (Fig 1). However, at this stage, the cell will only reverse if signals from Frz are sufficiently strong. When this occurs, 2 response regulators FrzX P (phosphorylated FrzX) and FrzZ P (phosphorylated FrzZ) bind to opposite poles, which trigger the detachment of MglA from the leading cell pole and its relocalization to the lagging cell pole. MglB is subsequently detached and relocated to the opposite pole, completing the switch.
Many questions remain unaddressed concerning the exact sequence of events that take place at the poles. Specifically, the presence of antagonizing GAP and GEF activities at both cell poles requires that this balance be regulated. The accumulation of RomRX at the lagging pole is not sufficient to drag MglA to the lagging cell pole, most likely because the MglB GAP activity predominates. Accumulation of FrzX P at the lagging pole could switch this balance in favor of the GEF and provoke the switch. In this study, we show that MglB is in fact a bifunctional enzyme and that MglB regulation could be key for the control of this enzymatic balance at the pole.
Previously, biochemical and structural studies with homologous proteins from Thermus thermophilus demonstrated that MglB is the GAP of MglA [4,5]. MglA and MglB from T. thermophilus (TtMglA and TtMglB) share 62% and 29% sequence identities, respectively, with the M. xanthus counterparts. Unlike the majority of small Ras-like GTPases, both the active site residues of the GTPase were found within TtMglA itself (Arg53 A and Gln82 A ; residue names with superscript "A" denote residues of MglA) instead of being contributed by the GAP [5]. A dimer of TtMglB interacts with a monomer of TtMglA. The structural studies revealed that β 2 strand of TtMglA undergoes a screw-type movement during the interaction between TtMglA and TtMglB in the presence of GTP analogues [5]. The β-screw movement positions the active site residues favorably for GTP hydrolysis. In contrast to M. xanthus, T. thermophilus is not known to switch polarity and appears to lack most of the other regulatory components (e.g., Frz proteins). Hence, it is imperative to study the Myxococcus proteins for a complete molecular characterization of the oscillatory system.
In this study, we report the biochemical and structural characterization of M. xanthus MglA and MglB (MxMglA and MxMglB). Our investigations led to the discovery of an allosteric pocket on the small Ras-like GTPase fold. Allosteric binding of the C-terminal helix (Cthelix) of MxMglB on MxMglA facilitates nucleotide exchange. Thus, MxMglB plays a dual role in increasing the GTPase activity of MxMglA by orienting the catalytic residues (role of a GAP) and by facilitating exchange of GDP with GTP (role of a GEF). Furthermore, our study reveals that the Ct-helix plays an important role in regulating the mutually exclusive localization of MglA, MglB, and RomR. Accordingly, deletion of the helix in M. xanthus trapped MglA, MglB, and RomR at both poles of the cell and drastically affected motility. Importantly, the study reveals the integration of GAP and GEF activities into a single interacting partner of a prokaryotic small Ras-like GTPase. Our study highlights the mechanistic features of a GEF in a prokaryotic small Ras-like GTPase family.

Crystal structure of MxMglA is in the GDP-bound conformation
The crystal structure of MxMglA was determined at a high resolution of 1.3 Å (Protein Data Bank [PDB] ID: 5YMX; S1 Table). Fig 2A contains a description of the small Ras-like GTPase fold of MglA with the secondary structure elements labeled. Structure of MxMglA is similar to TtMglA with 2 additional β-strands (β 0 and β 2 � ), compared with the 5 α-helices and 6 βstrands present in the small Ras-like GTPase fold [5]. All the residues of the small Ras-like GTPase fold as well as the C-terminal hexahistidine tag were modeled in the MxMglA structure. Though GDP was not added during protein purification or crystallization, the crystal structure of MxMglA contained GDP in the nucleotide-binding pocket (Fig 2A, S1A Fig). High-performance liquid chromatography (HPLC) analysis of the protein sample reconfirmed that the bound ligand is GDP (S1A Fig, S1B Fig).
At the C terminus, density was absent for residues 131 to146. However, residues 147 to 157 of the MglB 1 protomer could be modeled into the electron density (S1D Fig, S1E Fig), which adopted a helical conformation. Henceforth, this will be referred to as Ct-helix (C-terminal helix of MxMglB). In the MglB 2 protomer, electron density was absent for residues beyond 130. The Ct-helix of MglB 1 bound to a pocket in MxMglA distal from the nucleotide-binding site (Fig 2B and Fig 2D). The binding pocket in MxMglA was formed by the residues of the α 5 helix, β 2 strand, and the loop connecting β 2 and β 3 strands (β 2 -β 3 loop) of MxMglA (interswitch region of small Ras-like GTPases [19]; Fig 2D). The corresponding β 2 to β 3 loop was disordered in the TtMglAB structure [5] (S1C Fig).
The conformational features described below are consistent with the previously determined structures of the corresponding complex from TtMglAB [5] (S1C Fig). Some of the prominent features are the optimal orientation of the active site residues (Fig 2E, Fig 2F) and the β-screw movement (Fig 2G, Fig 2H), which were also observed in the TtMglAB structures [5]. The conformational changes between MxMglA-GDP and the MxMglA-GTPγS bound to MxMglB resulted in movements of switch 1 and switch 2 such that Arg53 A and Gln82 A of MxMglA are now oriented towards the catalytic water ( Fig 2E, Fig 2F). Another interesting observation is the orientation of the residues Thr54 A and Asp58 A of MxMglA, which coordinate the Mg 2+ ion bound to GTPγS (Fig 2F). The β-screw movement exposed hydrophobic residues of MxMglA toward the MxMglB interface and facilitated MxMglAB interaction ( Fig 2C, Fig 2H).

Ct-helix allosterically affected the GTPase activity
To elucidate the functional role of the MglB 1 Ct-helix, we compared features of GTPase stimulation and MxMglA interaction with full-length MxMglB (MglB) and a truncated version of MxMglB lacking the Ct-helix (MglB ΔCt ; S1F Fig). The GTPase activity was monitored by estimation of released GDP using a coupled enzyme assay [20] or phosphate release using malachite green-based assay [21]. The reactions were carried out with excess GTP (substrate) and hence measured multiple turnovers of the GTPase. Both approaches for monitoring enzyme activity showed similar rates of hydrolysis. The GTPase activity of MxMglA on its own was negligibly low, whereas the activity increased in the presence of MxMglB ( Fig 3A, Fig 3B, Fig  3C, S2A Fig). This confirmed the role of MglB as a GAP, consistent with published results [4,5,7]. Control experiments were performed with only MxMglB (no MxMglA) and excess of MxMglB (S2A Fig). Similar GTPase activities were observed for MxMglA in the presence of MxMglB with or without a C-terminal hexahistidine tag (Fig 3C), and hence all further experiments were performed using the construct with C-terminal hexahistidine tag.
We found that the GTP hydrolysis rate was approximately 2-fold lower in the presence of MxMglB ΔCt compared with MxMglB ( Fig 3A, Fig 3B, Fig 3C). The GTPase activity remained low despite using 5 times excess of MxMglB ΔCt (Fig 3A, Fig 3B), suggesting that the effect was not due to a reduced affinity of MxMglB ΔCt to MxMglA. We compared the GTPase activities with that of the active site mutant MxMglA Q82L (referred to as MxMglA Q ) in the presence of MxMglB (S2B Fig, S2C Fig; Fig 3B). The GTPase activity of MxMglAB ΔCt was, however, higher than that of MxMglA Q B.
Next, we proceeded to confirm that the interactions observed in the crystal structure between Ct-helix and the binding pocket of MxMglA are indeed relevant for the observed increase in GTPase activity. Analysis of the crystal structure of MxMglAB-GTPγS complex showed that Asp150 B , Asp151 B , and Asp152 B on the MxMglB Ct-helix formed water-mediated as well as direct interactions with MxMglA ( Fig 2D). Lys181 A and Lys185 A of the α 5 helix of MxMglA interacted with Phe157 B and Asp154 B of the MxMglB Ct-helix, respectively ( Fig 2D). Leu64 A and Ile67 A of MxMglA were on the β 2 to β 3 loop (interswitch region) and were within MxMglB-stimulated GTPase activity of the MxMglA mutants was less than the wild type ( Fig 3D). The GTPase activity remained lower despite 5-fold increase in the concentration of MxMglB ( Fig 3D). The above data demonstrated that mutation of residues in the helix-binding pocket affected rate of nucleotide hydrolysis, even though they were located away from the GTPase active site. This implied that MxMglB Ct-helix played an allosteric role in regulating the GTPase.  Table). The binding affinities of MxMglB to MxMglA K and MxMglA L in presence of m-GNP were also not affected drastically despite the mutations (S2F Fig, S2B Table). Hence, we concluded that the lower GTPase activities observed for these mutant constructs were not due to decreased affinity between the GTPase and MxMglB constructs.
Thus, the Ct-helix appears to be essential for stabilizing the interaction between MxMglB and GDP-bound MxMglA. In the previous study with TtMglB, the binding between TtMglB and TtMglA was observed only in the presence of GNP (or GTP analogues), a characteristic feature of a GAP. Our studies imply that nucleotide hydrolysis by MxMglA is not sufficient for disrupting the MxMglAB complex in M. xanthus. Also, interaction with the GDP-bound state of a GTPase is a characteristic feature of a GEF [17]. Because we observed that MxMglA and MxMglB interact in the presence of GDP, we proceeded to check whether MxMglB could also perform a GEF function.

Ct-helix of MxMglB contributed toward nucleotide exchange
As mentioned earlier, the purified MxMglA contained GDP bound to the active site pocket (S1 Fig). To observe the loading of m-GDP to MxMglA, fluorescence intensity of m-GDP was monitored upon addition of MxMglA. An increase in fluorescence intensity indicated increased binding of m-GDP. Exchange was observed to occur spontaneously upon addition of MxMglA but was accelerated upon addition of a mix of MxMglA and MxMglB (Fig 4A, phase I). In contrast, the presence of MxMglB ΔCt instead of MxMglB did not affect the exchange kinetics (Fig 4A, phase I). The trend did not change despite addition of MxMglB ΔCt in a 1:10 ratio (S2G Fig). Next, we monitored the exchange kinetics by competing the bound m-GDP with excess of unlabeled GDP. The rate of loss of fluorescence (k off ) was higher for MxMglA+MxMglB mix compared with MxMglA or MxMglA+MxMglB ΔCt mix (Fig 4A, phase II). This suggested that full-length MxMglB assisted in GDP exchange, whereas the C-terminal deletion mutant failed to stimulate GDP exchange.
We proceeded to carry out the same experiment by competing m-GDP with GTP, a relevant reaction for a GEF activity. MxMglB addition affected the rate of displacement of m-GDP in this experiment too ( Fig 4B; phase II). Loading with m-GNP and competition with unlabeled GTP provided further information on the exchange kinetics ( Fig 4C). Lack of a major increase in fluorescence indicated that binding of m-GNP to MxMglA was very unstable ( These experiments suggested that MxMglA favored binding of GDP over GTP. However, in the presence of MxMglB, GTP bound better than GDP, as evidenced by a higher k on and higher k off for m-GDP and a higher k on but a lower k off for m-GNP ( Fig 4C). In contrast, MxMglB ΔCt did not affect the MxMglA-GDP complex (equal k on for m-GDP for MxMglA and MxMglA+MxMglB ΔCt mix; Fig 4A, phase I). However, it stabilized the GTP-bound conformation of MxMglA (increased intensity for m-GNP exchange; Fig 4C, phase I). This effect is consistent with our observations that MxMglB interacts with both MxMglA-GDP and MxMglA-GTP, whereas MxMglB ΔCt interacts only with MxMglA-GTP ( Fig 3E, Fig 3F). Thus, MxMglB functions not only by stabilizing GTP-bound MxMglA but also by destabilizing the bound GDP, whereas MxMglB ΔCt contributes only toward stabilizing GTP-bound MxMglA.
Our assays measuring exchange kinetics with different combinations of labeled and unlabeled nucleotide pairs clearly implied that the Ct-helix played a role in nucleotide exchange. Consequently, we concluded that MxMglB accelerates MxMglA GTPase not only by appropriately orienting the catalytic residues but also by facilitating GDP release through an allosteric effect via the Ct-helix. This effect increased the overall hydrolysis rate in our GTPase assays, which are multiple turnover enzymatic reactions.

mglB ΔCt does not respond to frz signaling similar to ΔmglB
To test whether the MxMglB Ct-helix is important for MxMglB activity in vivo, we expressed MxMglB and the MxMglB ΔCt variant under the control of mglB promoter at the Mx8 phage attachment site in an mglB deletion mutant of M. xanthus (ΔmglB). MxMglB ΔCt was stably expressed, although its expression level appeared slightly decreased compared to MxMglB (it is also possible that this apparent decrease is because the Ct-helix is an important epitope for the ratio (magenta) at 400 seconds (marked by solid grey line labeled "+ P"), followed by competition of m-GNP (phase II) by addition of excess of unlabeled GTP at 1,800 seconds (marked by solid gray line labeled "+ GTP"). ND for the k on and k off values in panel C stands for "not determined". For panels A to C, phase I (between solid gray line labeled "+ P" and the first dotted line) represents kinetics for association (for k on MxMglB antibody; S3A Fig, S3B Fig). MxMglB ΔCt must nevertheless carry biological activity because it induced a drastic motility phenotype and localized at the poles (see below). MglA was expressed to comparable levels in strains complemented by MxMglB or MxMglB ΔCt (S3A Fig, S3B Fig). The level of MglA is slightly reduced in a ΔmglB mutant because of a weak polar effect but this reduction did not create detectable phenotype [7]. The strain mglB + complemented motility on an agar surface, while mglB ΔCt+ led to profoundly defective motility in colony plate assays ( Fig 5A). The motility phenotype was in fact more pronounced than for the ΔmglB mutant, suggesting that expression of MxMglB ΔCt deeply perturbs the function of MglA. However, single cells were motile, and the defect could be linked, at least partially, to aberrant cell reversals ( Fig 5B, S3C Fig, S3D Fig). In ΔmglB mutant cells, the polarity axis is disrupted and therefore the reversal distribution does not change depending on environmental conditions, contrarily to wild-type (WT) cells in which the reversals depend on the Frz-signaling state [7]. This can be shown in cells in which Frz signaling can be induced by addition of isoamyl alcohol (IAA), which increases the reversal frequency of mglB + cells ( Fig 5B) [7,22,23]. As expected, the ΔmglB mutant showed a reversal frequency distribution that did not change upon addition of IAA ( Fig 5B). Remarkably, mglB ΔCt+ cells also did not respond to the addition of IAA (Fig 5B), suggesting that the polarity axis and Frz-dependent regulations are deeply affected in absence of the MxMglB Ct-helix.

MxMglA, MxMglB ΔCt , and RomR localize to both cell poles in mglB ΔCt+
To further explore if MxMglB ΔCt perturbs cell polarity, we analyzed its localization in single cells of M. xanthus by expressing C-terminal neonGreen (nG) fused [24] to MxMglB and MxMglB ΔCt in mglB deletion backgrounds. MglB-nG localized to the lagging cell pole and oscillated from pole-to-pole with kinetics comparable to previously described yellow fluorescent protein (YFP) and mCherry fusions (Fig 5C top, S3C Fig, S4A Fig) [4,5,11]. In contrast, MxMglB ΔCt -nG mostly localized symmetrically at both the poles and failed to oscillate (Fig 5C  bottom, S3C Fig, S4A Fig). Thus, the Ct-helix is required for ensuring a unipolar localization of MxMglB.
We next tested how the bipolar localization of MxMglB ΔCt affected the other polarity proteins, MxMglA and RomR, by expressing MxMglA-YFP or RomR-GFP, respectively, in the mglB ΔCt+ strain. Remarkably, both MxMglA-YFP and RomR-GFP were bipolar and static ( Fig  5D, Fig 5E [top], S4B Fig, S4C Fig), similar to the localization pattern observed in the ΔmglB deletion mutant [12,13]. Thus, in the presence of MxMglB ΔCt , the polarity axis is disrupted, exactly as observed when MxMglB is deleted. Most strikingly, all 3 proteins MxMglA, RomR, and MxMglB ΔCt co-localize at the poles.

Allosteric regulation in MglA homologues and other Ras-like GTPases
The regulation of MxMglA GTPase activity and MxMglAB interaction by MxMglB Ct-helix prompted us to explore whether analogous mechanisms exist among other prokaryotic and eukaryotic small Ras-like GTPases. estimation), phase II (between solid line labeled "+ GDP/GTP" and final dotted line) represents kinetics for dissociation (for k off estimation). A schematic representation of the observations and probable explanations at each phase is shown on the right side, following the same color scheme for the font as in the plots. The events directly observable through fluorescence intensity change of the mant-labeled component in the experiment are highlighted in bold. The panels include representative plots of multiple repeats, and the k on and k off values (in units of s -1 ) represent the average values of the multiple repeats. The numerical data for all the figure panels have been provided in the respective sheets in S1 Data. Ct-helix, C-terminal helix; m-GDP, Allosteric regulation of a prokaryotic small Ras-like GTPase Earlier, it has been shown that MglA and MglB, which exist in almost all phyla of eubacteria, can be categorized into 5 phylogenetically related groups [25]. Genes encoding MglBlike proteins are either associated with genes encoding MglA-like protein in the same operon, called coupled MglBs, or without an associated MglA gene (orphan MglBs) [25] (S3 Table). We analyzed these sequences and found that 66 coupled and 25 orphan MglBs possessed a Cterminal extension (more than 15 amino acids beyond the Rbl/LC7 fold; S3 Table). A total of 56 of the 66 coupled MglB sequences showed a predicted helical region containing 2 to 4 aspartates or glutamates (D/E rich), whereas the C-terminal extensions of the remaining MglB sequences contained proline-rich stretches with no predicted secondary structure (Fig 6A, S3  Table).
Out of these 56 coupled sequences, the α 5 helix (which interacts with the Ct-helix) of 48 MglA sequences possessed positively charged residues (either lysines or arginines; Fig 6B). These positively charged residues were not conserved in other MglA sequences ( Fig 6B). This suggested that the Ct-helix of MglB and its binding pocket on MglA have coevolved. Interestingly, we found MglB having the D/E rich Ct-helix to be present in all the 5 phylogenetic groups of MglA-like proteins, including a variety of bacteria and also in the archaea Methanobacterium paludis (S5 Fig). This indicated that the mechanism of allosteric regulation of the MglA GTPase by the MglB Ct-helix could be widespread in prokaryotic Ras-like GTPases and not limited to Myxococcus. Because the allosteric regulation by full-length MglB has not been characterized for any other prokaryotic MglA-like and MglB-like proteins, the mechanism proposed here would be relevant toward understanding their function.

MxMglB Ct-helix facilitates nucleotide exchange allosterically
Our results revealed that the Ct-helix is essential for MxMglB to interact with MxMglA in the GDP-bound state. Deletion of the helix resulted in the formation of MxMglAB complex in the presence of only GTP but not GDP. This prompted us to identify the structural features that enable MxMglB to bind MxMglA-GDP. The β-screw movement (flipping of MxMglA β 2 strand) appears to be essential for the interaction of MxMglB with MxMglA. This conformational change exposes residues that form a major part of MxMglAB interface.
An inspection of the MxMglAB-GTPγS structure revealed that the switch 1 loop and the binding pocket of the Ct-helix are at the N-and C-terminal ends of the β 2 strand of MxMglA, respectively ( Fig 2H). Hence, we propose that the flipped state of the β 2 strand that favors MxMglB binding can be achieved in 2 ways-by interaction of either the γ phosphate of GTP with the switch 1 loop or the Ct-helix with the β 2 to β 3 loop (Fig 2H). The flipped state achieved by the Ct-helix interaction and the consequent stabilization by the Rbl/LC7 fold of MxMglB results in the formation of a complex with MxMglA-GDP. As a consequence of the flipped state of the β 2 strand, MxMglA in the MxMglAB complex will potentially possess a nucleotidebinding pocket that favors binding of GTP over GDP, consistent with our observations from the nucleotide exchange assays (Fig 4). Thus, the proposed model provides an explanation for field views, in which "n" represents the total number of cells analyzed. The fluorescence intensity profiles for each cell are compiled and shown in S4B Fig MxMglAB-GDP complex formation and a mechanistic insight into the allosteric action of Cthelix for facilitating nucleotide exchange (Fig 7A).
Interestingly, like MglB, many eukaryotic GEFs such as DENN-1B domain, TRAPP-I, and Mon1-Ccz1, have the Rbl/LC7 fold [26]. Furthermore, we found that the intersubunit orientation in MglAB is conserved in the GTPase-GEF complexes of Rab35 with DENN-1B domain [27], Rab Ypt1p with TRAPP-I [28], and Ypt7 with Mon1-Ccz1 [29] (S6 Fig). However, unlike in MglAB complex, the mechanism of GEF action in these eukaryotic systems is based on a glutamate or aspartate residue interacting directly with switch 1 or switch 2 (S6A Fig, S6B Fig). These residues are present in extensions from the Rbl/LC7 fold or are contributed by other interacting partners that form the GEF complex [26].

Mechanism of dual GAP and GEF action by MxMglB
We propose that MxMglB increases the rate of GTP hydrolysis by MxMglA through 2 complementary ways. Firstly, Rbl/LC7 domain of MxMglB stabilizes the flipped β 2 strand interface, which orients the MxMglA catalytic residues favorably. Secondly, its Ct-helix, a flexible extension from the main Rbl/LC7 fold, retains the flipped state of the β 2 strand post GTP hydrolysis and, consequently, stabilizes the MxMglAB-GDP complex. This interaction results in a nucleotide-binding pocket that accelerates release of GDP from the complex and favors GTPbinding. Thus, by facilitating GDP-to-GTP exchange MxMglB enhances the rate of GTP hydrolysis by MxMglA, increasing its potency as a GAP. Indeed, our in vitro results show that MxMglB ΔCt exhibits reduced GAP activity, which we attribute to a defect in nucleotide exchange.

Role of Ct-helix and exchange activity of MxMglB in the polarity oscillation cycle
In vivo, we propose that the MxMglB Ct-helix is critical for the balance between GAP and GEF activities at the lagging pole. In absence of the Ct-helix, all 3 proteins-MglA, MxMglB ΔCt , and RomR-localize to both cell poles. Thus, the MxMglB Ct-helix is absolutely required to enable the cascade of interactions that establish polarity and permit its switch. Co-localization of both MxMglA and MxMglB ΔCt suggest that MxMglB ΔCt does not support its GAP function in vivo, despite exhibiting only a reduction in GAP activity in vitro. In between reversals in wild-type cells, MxMglB and RomRX stably co-localize at the lagging pole [11,14,15]. MxMglA is efficiently excluded from the lagging pole, suggesting that the GAP activity of MxMglB predominates over the GEF activity of RomRX. Thus the MglB Ct-helix possibly overcomes the action of RomRX.
How could this function at the molecular level? Analysis of RomR revealed a stretch of negatively charged residues (Glu-rich) akin to that of MxMglB Ct-helix (S7A Fig). Thus, similar to MxMglB, the RomRX complex could facilitate GDP exchange from MxMglA by acting through the allosteric pocket on MxMglA. In this process, the C-terminal Glu-rich region of implicated in interaction with the allosteric pocket on MglA are highlighted by " � " in the sequence conservation logo. The fractional occupancy of amino acids in the sequence alignment is shown for each amino acid position. The numbers within brackets denote the number of sequences in the alignment. The residues are colored in 20 different shades. (B) Comparison of residue conservation in the α 5 helix of MglA sequences coupled with MglB sequences possessing a negatively charged C-terminal extension versus that of all coupled MglA sequences. The lysine residues that potentially interact with the Ct-helix of MglB are highlighted by " � " in the sequence conservation logo. The fractional occupancy of amino acids in the sequence alignment is shown for each amino acid position. The numbers within brackets denote the number of sequences in the alignment. Ct-helix, C-terminal helix; MglA, Mutual Gliding motility A; MglB, Mutual Gliding motility B. https://doi.org/10.1371/journal.pbio.3000459.g006 Allosteric regulation of a prokaryotic small Ras-like GTPase RomR might thus assist nucleotide exchange, whereas RomX could assist in stabilizing the flipped β 2 strand of MxMglA. We tested this hypothesis by replacing the Ct-helix of MxMglB with the Ct-helix from RomR (S7A Fig, S7B Fig) and carrying out GTPase activity measurements in the presence of the chimeric MglB (MglB Rhelix ). Indeed, we observed that the chimeric construct accelerated the GTPase activity of MglA similar to that of wild-type MxMglB (S7C Fig), supporting the hypothesis that the Ct-helix of RomR is capable of binding at the allosteric pocket of MxMglA. At the lagging pole, the MxMglB Ct-helix and the RomR Glurich helix might compete for binding at the MxMglA helix-binding pocket, which would displace RomRX from MxMglA. When the MglB Ct-helix is deleted, RomRX could therefore interact with MxMglA thus blocking MxMglB ΔCt from exerting its GAP activity. This would allow RomRX to exert its GEF activity at the lagging pole too and stabilize the anchoring of MglA to both the poles.
Additional regulation of the Ct-helix must take place to determine the nucleotide specificity of MxMglB binding to MxMglA. Additional regulators in vivo will potentially determine the Allosteric regulation of a prokaryotic small Ras-like GTPase accessibility of the Ct-helix and thus the localization of MxMglA and MxMglB. At the pole, the exact molecular organization of the signaling complexes is not known. Other proteins that participate in the polarity switch and that are present at the lagging pole, for example, the MglBlike protein MglC, can potentially regulate the Ct-helix availability [30]. During the switch, action of FrzX P could, for example, block the inhibitory effect of MglB on RomRX and thus initiate the switch. This possibility could be explored in the future. Further characterization of other polarity determining components and their interactions with MxMglAB are therefore essential to discriminate the GAP and GEF-active phases of MxMglB during the polarity reversal cycle.
Insights gained from the prokaryotic small Ras-like GTPase MxMglA and its interacting partner MxMglB highlight the dual role of MxMglB both as a GAP and a GEF. The novel mechanism of GEF action is based on an allosteric interaction with the small Ras-like GTPase. Based on sequence analysis, it appears to be a conserved mechanism among prokaryotic small Ras-like GTPases. The study also opens up a new avenue for design of compounds or factors that can modulate the enzyme activity by targeting the newly discovered allosteric binding pocket of the universal Ras-like fold.

Protein expression and purification
Protein constructs, cloning, and overexpression. Genes corresponding to mglA and mglB were amplified from M. xanthus genomic DNA (obtained from DSMZ, Germany, catalogue number 16526) using suitable primers (S4 Table) and cloned into pHis17 vector (obtained from Löwe lab, MRC LMB, Cambridge; refer Addgene plasmid #78201 for vector backbone) between the restriction enzyme sites NdeI (New England Biolabs Inc) and BamHI (New England Biolabs Inc). Single-point mutants of MxMglA and deletion constructs of MxMglB were generated using a site-directed mutagenesis strategy utilizing PCR-based methods followed by DpnI (New England Biolabs Inc) digestion, transformation, and screening of positive clones. All clones were confirmed by sequencing. The list of clones and the primers used to generate them are summarized in S5 and S4 Tables. All chemicals were procured from Sigma Aldrich, unless otherwise mentioned.
Protein expression. Plasmids containing the gene of interest were transformed into suitable Escherichia coli strains, and culture was grown at 37˚C and at 30˚C post induction. E. coli strain BL21-DE3 was used for expressing MxMglA wild type and its mutants. The cultures were grown in LB media containing 50 μg/ml of ampicillin and induced with 0.5 mM IPTG at OD 600 value of 0.8, whereas MxMglB and its mutants were expressed in the strain BL21-AI and induced with 0.02% L-arabinose (SRL Chemicals, India) at OD 600 value of 0.6. Selenomethionine-labeled proteins for MxMglB L156M and MxMglB I148M have been expressed by the feedback inhibition method [31].
Purification of MxMglA(His 6 ). For purification, harvested cells were resuspended in the lysis buffer L (50 mM Tris, 200 mM NaCl [pH 8.0], and 10% glycerol) and then spun at 39,000g for 45 minutes at 4˚C. Supernatant was loaded on to a 5-ml HisTrap (GE Lifesciences) column because the presence of hexahistidine tag at the C-terminus facilitated the binding of overexpressed protein of interest to the column. The column was equilibrated with binding buffer (Buffer A: 50 mM Tris [pH 8.0], 200 mM NaCl) prior to loading the supernatant and the bound protein was washed and eluted with a step gradient of 2%, 5%, 10%, 20%, 50%, and 100% of Buffer B (Buffer A containing 500 mM imidazole). The fractions containing the protein were pooled, concentrated, and loaded onto Superdex75, 10/300 (GE Lifesciences). The protein was eluted into 50 mM Tris (pH 8.0), 50 mM NaCl. Fractions containing the protein were pooled, concentrated, flash frozen, and stored at −80˚C. MxMglA mutants, i.e., MglA K , MglA L , and MglA Q , were also purified using a similar protocol.

Crystallization and structure determination
MxMglA crystallization. About 500 conditions of commercially available screens (Molecular dimensions, Hampton Research) were screened using Mosquito crystallization robotic system, using drop sizes consisting of 100 nl of protein at a concentration of 10 mg/ml and 100 nl of crystallization cocktail, in 96-well sitting drop plates (MRC plate, SWISS-SCI). Initial hits were obtained in many of the conditions, and, further, it was reproduced and optimized to get well-diffracting crystals. For crystallization of MxMglA, protein was diluted to a final concentration of 10 mg/ml in A50 buffer (50 mM Tris [pH 8.0], 50 mM NaCl). Diffraction quality crystals were obtained using the following conditions in a drop ratio of 1:1 volumes for protein and crystallization condition: (i) 0.1 M sodium cacodylate (pH 6.5), 40% v/v 2,4-methyl pentane diol, and 5% PEG 8000; (ii) 0.1 M imidazole (pH 8.0), 30% w/v 2,4-methyl pentane diol, and 10% w/v PEG 4000; (iii) 0.1 M sodium citrate (pH 5.6) and 35% w/v tertiary-Butanol; 20% ethylene glycol was included as cryoprotectant in the parent condition during crystal freezing.
MxMglAB crystallization. MxMglA and MxMglB were mixed in a ratio of 1:1 (considering monomeric molecular weight of MglB) in A50 buffer containing 5 mM MgCl 2 and 2 mM GTPγS (Sigma). The concentration of the proteins used to crystalize was 4 mg/ml of each. Screening for crystallization hits yielded a few conditions that were further optimized to get the desirable crystals. The crystals were obtained in 2 different conditions, namely, (i) PEG 4000 (8%) and ammonium sulphate (200 mM) and (ii) PEG 3350 (12%) and ammonium sulphate (200 mM). Diffraction quality crystals were obtained at a drop ratio of 1:1 volumes of protein to crystallization buffer. Crystallization conditions containing 20% PEG 400 were used as cryoprotectant during freezing of crystals. Crystals of MxMglA and selenomethioninelabeled MxMglB complex were obtained in the same crystallization conditions. Structure solution and refinement. Diffraction data from the crystals were collected at the home source using Rigaku Micromax 007 X-ray generator, and higher resolution data and anomalous data were collected at the synchrotron sources at Diamond Light Source, Harwell, UK, and ESRF, Grenoble. Data reduction was performed using IMOSFLM [32] or XDS [33], and scaling using AIMLESS [34] in CCP4 package [35]. MxMglA and MxMglB structures were solved by molecular replacement using the TtMglAB structure (PDB code: 3T12). Molecular replacement was performed using PHASER [36] available in CCP4 package. Refinement was carried out using PHENIX package [37] and model building using Coot [38]. The refined structures have been deposited in the PDB with accession numbers 5YMX (MxMglA-GDP) and 6IZW (MxMglAB-GTPγS).
In order to confirm the registry of the amino acids belonging to the Ct-helix of MxMglB, selenomethionine-labeled protein for mutant constructs of MxMglB where Ile148 B and Leu156 B , respectively, were mutated to methionines, were purified, and selenomethioninelabeled MxMglB was used for obtain MxMglAB crystals in the same crystallization condition. The anomalous data from the crystals were collected, and it was confirmed that the correct amino acids were modeled into the electron density. The anomalous signal from the methionines in these 2 mutants, respectively, confirmed the registry of the amino acids of the MxMglB Ct-helix (S1D and E Fig).

Fluorescence anisotropy experiments
Binding studies using fluorescent anisotropy measurements. Binding of fluorescentlylabeled nucleotides, mant-GDP (m-GDP) or mant-GppNHp (m-GNP) (Jena Bioscience) with the proteins were monitored by measuring the change in anisotropy [39]. The excitation and emission wavelengths used for monitoring the mant-labeled nucleotide fluorescence were 360 nm and 440 nm, respectively. The experiments were performed on Fluoromax-4 (Horiba), with a sample volume of 200 μl in a cuvette, and excitation and emission slit widths of 5 nm.
Nucleotide binding of MxMglA and its mutants were estimated by measuring the change in fluorescence anisotropy of the mant-nucleotide upon titration with increasing concentrations of the protein. Protein samples (MxMglA or its mutants) were titrated against a fixed concentration of mant-labeled nucleotide (100 nM). Each reading corresponds to 10 averaged single point anisotropy values. The initial value of anisotropy of the mant-labeled nucleotide was subtracted from all the values. GraphPad Prism was used to plot the values against the concentration of protein (concentration of MxMglA) and to fit the data to the binding equation for estimation of K D . Because it is known that MxMglA has a single binding site for nucleotides, equation for one site-specific binding, as given below, was used to fit the data points. The equation is y = B max × x � (K D + x), where y represents the bound fraction (fluorescence anisotropy based readout), B max is the maximum value obtained, x is the concentration of MxMglA, and K D is the binding constant.
The values were normalized by dividing each of them by the maximum value of anisotropy, B max . The normalized anisotropy values were plotted to obtain a binding curve. Because MxMglA was purified in the GDP-bound form, the measurements from this experiment are not absolute affinities of MxMglA (or its mutants) to the nucleotides but a reliable measure of exchange of the prebound GDP with m-GDP or m-GNP.
Similarly, anisotropy measurements for increasing amounts of MxMglB/MxMglB ΔCt with nucleotide-bound MxMglA provided information about the binding affinity of MxMglB to nucleotide-bound MxMglA. A mix of 400 nM of mant-labeled nucleotide with 2 μM of MxMglA was titrated with increasing concentrations of MxMglB/MxMglB ΔCt . The same calculations as described above were performed to obtain the binding affinities (K D ). In this case, the x-axis represents the concentration of MxMglB dimer. Because MxMglA binds to one dimer of MxMglB, equation for one site-specific binding was used to fit the data points.
Each data point is an average of at least 3 independent measurements, and the standard error is shown in which each data represent 3 to 5 repeats. Error bars represent standard error.
Nucleotide exchange assays. Intensity of fluorescence emission by mant-labeled nucleotide (GDP or GMPPNP [GNP]; Jena Bioscience) at 440 nm was monitored after the excitation at 360 nm. All the kinetic measurements for MxMglA, MxMglA with MxMglB, and their mutants were performed on Fluoromax-4 (Horiba), with a sample volume of 200 μl in a cuvette and excitation and emission slit widths of 2 nm. m-GDP/GNP (800 nM) was added with Buffer A50 (50 mM Tris, 50 mM NaCl, 5mM MgCl 2 [pH 8.0]) in a quartz cuvette (10 × 2 mm path length), and the fluorescence intensity was monitored initially for 400 seconds. The protein, i.e., MxMglA or the mix of MxMglAB/MxMglAB ΔCt (3 μM of MxMglA and 6 μM MxMglB/MxMglB ΔCt monomer), was added in the cuvette at 400 seconds after stabilization of the signal from only mant-nucleotide. Consequently, the fluorescence was recorded for 1,400 seconds. The binding was estimated in terms of increase in fluorescence intensity. At 1,800 seconds, the mant-labeled nucleotide was competed out with excess of unlabeled nucleotide (GDP or GTP; 500 μM), resulting in a decrease of fluorescence intensity because of release of mant-labeled nucleotide from the protein. For plotting the relative intensities from the measurements, each value was divided by the average of first 200 readings (400 seconds). These accumulation and decay reactions were fitted to exponential binding equations as given below to estimate the k on and k off values. P + N 1 ! PN 1 + N 2 ! PN 2 + N 1 , where P represents protein, N 1 is the labeled nucleotide, N 2 is the unlabeled nucleotide, and PN denotes the protein-nucleotide complex.
For estimation of k on , PN t = PN max (1 − e −kt ). For estimation of k off , PN t = PN max − PN min (e −kt ) + PN min Here, PN t represents the amount of the complex at time t, PN max is the maximum amount of the complex upon association, and PN min is the minimum amount of the complex after dissociation.
GTP hydrolysis assay NADH coupled assay. GTPase activity measurements were made using a coupled enzyme based assay in which the conversion of NADH to NAD + is coupled to the utilization of GDP produced [20]. The GTPase or ATPase hydrolyses the nucleotide while pyruvate kinase (PK) uses GDP (or ADP) and phosphoenol pyruvate (PEP) to produce GTP (or ATP) and pyruvate. Lactate dehydrogenase (LDH) utilizes NADH and pyruvate to produce NAD + and lactate. Because the utilization of NADH is equivalent to the GDP produced by the GTPase reaction, the actual read out is decrease in absorbance of NADH, which in turn estimates the GDP produced by the GTPase. The amount of NADH utilized in the reaction was measured by monitoring the NADH absorbance at 340 nm using a multiplate reader, Varioskan Flash (Thermo Scientific). A master mix was prepared in Buffer A50 containing PEP (1mM; Sigma), GTP (1mM; Jena Bioscience), NADH (600 μm; Sigma), and PK/LDH mix (approximately 25 U/ml; Sigma). All the components of NADH reaction were mixed in a total reaction volume of 200 μl and added in a 96-well plate. The reaction was initiated by addition of MxMglA or MxMglAB mix and their relevant mutant constructs. MxMglA and its mutants were used at 10 μM and MxMglB and its mutants at concentrations 20 μM and 200 μM (in a molar ratio of 1:2 and 1:10 considering monomeric molecular weight of MxMglB).
Readings were recorded at the interval of 20 seconds for 2 hours. The absorbance value of buffer containing GTP was subtracted from all the values. Also, the absorbance at the 0 time point for each reaction was subtracted from all readings of the reaction. The absorbance readings were represented as concentration of GDP released, by using a conversion factor based on the standard graph obtained by plotting the absorbance values for known concentrations of NADH. k cat values were calculated and plotted using GraphPad Prism. Readings from 0 to 7,200 seconds were fitted to a line using linear regression, and the slope of the fitted line was calculated to obtain the amount of GDP released (in μM) per unit time. k cat was calculated as the amount of GDP released per unit time per unit enzyme concentration. This value was used further to compare the enzyme activities among all the mutants. The significance was calculated using GraphPad Prism, using one-way ANOVA with Tukey test of 95% confidence interval, and significance was marked on the scatter plot.
Phosphate release measurements using malachite green assay. This assay was performed using MxMglA at a concentration of 10 μM, and MxMglB in molar ratio 1:2 (20 μM, considering monomeric molecular weight of MxMglB) with excess of GTP, i.e., 1 mM GTP in buffer A50 (50 mM Tris, 50 mM NaCl [pH 8.0]) containing 5 mM MgCl 2 . All the enzyme activity assays were performed at 30˚C. Readings were taken at different time points, i.e., 0, 15,30,45,60,75,90,105, and 120 minutes. To stop the reaction, the sample was heated at 65˚C and then malachite green reagent (prepared as described in [21]) was added. The absorbance was recorded at 630 nm after 20 minutes of incubation, using a plate reader (Varioskan, Thermo Scientific). Amount of Pi (inorganic phosphate) released in μM was calculated using a standard curve, obtained by mixing a set of known concentrations of NaH 2 PO 4 with malachite green solution. The experiments were repeated multiple times, and the slope was calculated from every repeat. Readings from 0 to 120 minutes were fitted to a line using linear regression, and the slope of the fitted line was calculated to obtain the amount of Pi released (in μM) per unit time. k cat was calculated as Pi released per μM enzyme concentration for each reaction. k cat values were plotted, and significance was calculated on GraphPad Prism, using one-way ANOVA with Tukey test of 95% confidence interval, and significance was marked on the scatter plot.

HPLC for bound nucleotide estimation
Purified MxMglA was diluted to 2 to 3 mg/ml concentration in buffer A50 and then heated at 65˚C. The heated protein sample was spun at 21,000g for 10 minutes at 4˚C. Then supernatant was filtered with 0.22 μm cellulose acetate filter (Corning), and the sample was loaded on the DNAPac 200 ion-exchange column (Thermo Fisher). Buffer A (2 mM Tris [pH 8.0]) was used as binding buffer. The runs were performed at a flow rate of 0.8 ml/min, with steps of 100% Buffer A for 3 minutes, followed by linear gradients of 0% to 30% Buffer B (2 mM Tris [pH 8.0], 1.25 M NaCl) for 10 minutes, and 30% to 100% B for 12 minutes. In S1B Fig, the peak profile obtained for 18 μM of MglA was compared with that of 50 μM GDP solution, with 35 μl of sample injection volume for both. The absorbance at 260 nm, indicative of the presence of GDP, was plotted against retention time.

Thermal shift assay [40]
In a 25-μl reaction volume containing 2 μM of protein in buffer A50, SYPRO Orange dye (Sigma; available as a 5,000× concentration in DMSO) was added to a final concentration of 5×, in 96-well multiwell PCR plates. Biorad CFX96 Real-Time System was used to monitor the change in fluorescence of SYPRO Orange (Sigma-Aldrich). The mix was heated from 4˚C to 90˚C with an increment of 1.2˚C per minute. Readings were recorded at intervals of 0.4˚C. The melting temperature (T m ) was calculated based on the differential of the fluorescence intensity versus temperature plot (dF/dT). This gave an estimate of the stability of the protein.

In vivo experiments
Myxococcus strains, plasmids, growth conditions, and genetic constructs. Primers, plasmids, and strains used for this study are listed in S4 Table, S5 Table and S6 Table, respectively. In general, M. xanthus strains were grown at 32˚C in CYE rich media as previously described by Bustamante et al. [23]. Plasmids were introduced in M. xanthus by electroporation. Complementation, expression of the fusion, and mutant proteins were obtained by ectopic integration of the genes of interest at the Mx8-phage attachment site [7] under the control of their own promoter in appropriate deletion backgrounds. E. coli cells were grown under standard laboratory conditions in Luria-Bertani broth supplemented with antibiotics, if necessary.
Construction of Myxococcus strains. Following an initial step of PCR amplification of the relevant gene fragments for complementation, the fragments were cloned into the pSWU19 vector by the one-step sequence-and ligation-independent cloning (SLIC) method reported previously by Jeong et al. [41]. Specific details of PCR amplification are given below, whereas S4 Table, S5 Table and S6 Table contain the list of primers, plasmids, and strains used.
pSWU19 mglB ΔCt , pSWU19 mglB ΔCt -neon green, and pSWU19 mglB-neon green for complementation of the mglB deletion were constructed by amplifying 200 bp upstream the mglB coding sequence (promoter) with the mglB ΔCt gene (417 bp of mglB gene), mglB ΔCt gene (417 bp) and the linker with neon green, and the mglB gene and the linker with neon green, respectively. We chose to use neonGreen because this fluorescent protein (FP) is exceptionally bright and strictly monomeric, preventing possible artifacts due to the formation of weak FP dimers [24].
Phenotypic assays. Soft agar motility was performed as previously described by Bustamante et al. [23]. In general, cells were grown up to an OD 600 between 0.4 to 0.8 and concentrated at OD 600 = 5. Then they were spotted (10 μl) on 0.5% agar in CYE (soft agar). Colonies were photographed after 48 hours.
Fluorescence imaging and fluorescence intensity measurements. Protocol followed is as reported previously by Guzzo et al. [11]. For phase-contrast and fluorescence microscopy, cells from exponentially growing cultures were concentrated to OD 600 = 2 by centrifugation of 1 ml of culture and resuspended in the corresponding volume of TPM buffer (10 mM Tris-HCl [pH 7.6], 8 mM MgSO 4 , and 1 mM KH 2 PO 4 ). Then a drop of 2 μl was deposited on a coverslip and covered with a 1.5% agar pad with TPM buffer. Microscopic analysis was performed using an automated and inverted epifluorescence microscope TE2000-E-Perfect Focus System (PFS; Nikon, France) with a 100×/1.4 DLL objective and a CoolSNAP HQ2 camera (Photometrics). Images were recorded and processed with Metamorph software (Molecular Devices). All fluorescence images were acquired with a minimal exposure time to minimize bleaching and phototoxicity effects.
Reversal scoring assay. Protocol followed is as reported previously by Guzzo et al. [11]. For this assay, cells were grown at 32˚C in CYE medium. After overnight incubation, 1 ml of cells was centrifuged for 5 minutes at 7,000 rpm, and pellets were resuspended in TPM medium (10 mM Tris [pH 7.6], 8 mM MgSO 4 , 1 mM KH 2 PO 4 ) to OD 600 of 2 units. A total of 2 μl of cells were spotted on coverslip below a TPM 1.5% agar pad with addition of 0.075% IAA in melted agar before pouring the pad, allowing modulation of Frz-signaling intensity with IAA. IAA solutions were made in TPM buffer containing 1mM CaCl 2 . After 10 minutes of incubation at room temperature, time-lapse experiments were performed using an automated and inverted epifluorescence microscope TE2000-E-PFS (Nikon, France), with a 40x/ 0.75 DLL "Plan-Apochromat" objective and an ORCA-Flash4.0 LT PLUS Digital CMOS camera (#C11440-42U30 Hamamatsu Photonics). The microscope is equipped with a PFS that automatically maintains focus. Images were recorded with NIS-Eléments AR 4.60 software (Nikon).
Cell tracking. Image analysis was performed with a newly developed MicrobeJ/FIJI-based tracking procedure. Cells were detected on MicrobeJ by thresholding the phase-contrast images after stabilization. Cells were tracked using MicrobeJ on a minimum of 20 frames by calculating all object distances between 2 consecutive frames and selecting the nearest objects. The computed trajectories were systematically verified manually, and, when errors were encountered, the trajectories were removed. Reversals were scored when cells switched their direction of movement with a distance of reversal at least 0.3 μm long and the angle between the segments was less than 60˚. The reversal frequency was calculated by dividing the number of reversals by the duration of the track to generate a reversal frequency per hour. Trajectories with a "distance mean" less than 0.3 μm were removed from the data set. For each strain, 2 movies at least per biological replicate acquired on 2 different days were analyzed and plotted on box plots with MicrobeJ.
Cluster counting. Image analysis was performed on FIJI with the plugin cell counter; fluorescence images were acquired with a 1-second exposure time for the strains expressing MxMglA-YFP and MxMglB-neonGreen (MglB-nG). A 200 ms exposure time was used for the strain expressing RomR-GFP. The fluorescent intensity profiles of the different strains were generated using MicrobeJ. Briefly, cells were straightened, and the pole of the cells with the strongest fluorescence signal was chosen to orient the cells. Cells with abnormal shapes, unusual fluorescence signals, or dividing were removed from the data set. For each strain, the fluorescence intensity profiles of hundreds of single cells were thus plotted on a same graph.
Western blot. Samples were grown at 32˚C in CYE medium. When optical density at 600 nm (OD 600 ) reached 0.4 to 1, a volume of culture equivalent to 1 ml was centrifuged for 5 minutes at 7,000 rpm. The pellet corresponding to whole cells was resuspended in 1X SDS-PAGE loading buffer containing β-mercaptoethanol for getting 10 OD 600 units and heated for 10 minutes at 99˚C. Proteins samples equivalent to 1 OD 600 units were separated by SDS-PAGE. Electrophoresis was performed at 180 V for 50 minutes at room temperature using 10% SDSpolyacrylamide gel. For Western blotting, proteins were transferred from gels onto nitrocellulose membranes. The membranes were blocked 1 hour at room temperature in Tris-buffered saline (pH 7.6), 5% milk, 0.2% Tween 20 (for MglA) or in Tris-buffered saline (pH 7.6), 2% milk, 0.2% Tween 20 (for MglB) and incubated with primary antibodies directed against MxMglA (dilution at 1:5,000) or MxMglB (dilution at 1:2,500) in blocking buffer overnight at 4˚C. Details of primary antibodies used are available in the work by Zhang et al. [7]. After 2 washings of 5 minutes with Tris-buffered saline (pH 7.6), 0.2% Tween, 20 membranes were incubated with Goat Anti-Rabbit IgG (H + L)-HRP Conjugate (#1706515 BioRad) in the respective blockings buffers. The peroxidase reaction was developed by chemiluminescence (SuperSignal West Pico Chemiluminescent Substrate #34080 Thermo Scientific) scanned and analyzed with ImageQuant LAS 4000 and TL analysis software (GE Healthcare life sciences).

Sequence and structure analyses
The sequences were downloaded using the list given in the work by Wuichet and Søgaard-Andersen [25] from UniProt. The presence of the Ras-like GTPase and Rbl/LC7 domains, respectively, for the MglA and MglB sequences were confirmed using SMART (Simple Modular Architecture Research Tool) [42] analysis. The sequences were analyzed using JalView [43]. The MglB sequences were sorted according to their lengths to identify the sequences with Cterminal extension (longer than 15 amino acids beyond the last amino acid of the Rbl/LC7 fold). The coupled MglA sequences corresponding to MglB sequences with C-terminal extension were identified and analyzed for the presence of positively charged residues in α 5 helix of MglA. Conservation-based logos were obtained from the sequence alignments using Skylign [44]. Phylogentic trees were generated using the default settings in Clustal Omega [45]. The tree data were analyzed using iTol [46]. The structures of small Ras-like GTPases and interactors containing Rbl/LC7 domain fold were identified using DALI server [47] by giving MxMglA and MxMglB structures as input.
(green), mix of MxMglA and MxMglB in 1:2 ratio (dark purple), mix of MxMglA and MxMglB ΔCt in 1:2 (magenta, solid line) and 1:10 (magenta, dotted line) ratios. m-GDP was added into the cuvette, and the fluorescence intensity monitored for 500 seconds, when either MxMglA or MxMglA+MxMglB or MxMglA+MxMglB ΔCt were added. The time point of addition at 500 seconds is marked by a solid gray line labeled with "+ P." The numerical data for all the figure panels have been provided in the respective sheets in S1 Data.