The PilB-PilZ-FimX regulatory complex of the Type IV pilus from Xanthomonas citri

Type IV pili (T4P) are thin and flexible filaments found on the surface of a wide range of Gram-negative bacteria that undergo cycles of extension and retraction and participate in a variety of important functions related to lifestyle, defense and pathogenesis. During pilus extensions, the PilB ATPase energizes the polymerization of pilin monomers from the inner membrane. In Xanthomonas citri, two cytosolic proteins, PilZ and the c-di-GMP receptor FimX, are involved in the regulation of T4P biogenesis through interactions with PilB. In vivo fluorescence microscopy studies show that PilB, PilZ and FimX all colocalize to the leading poles of X. citri cells during twitching motility and that this colocalization is dependent on the presence of all three proteins. We demonstrate that full-length PilB, PilZ and FimX can interact to form a stable complex as can PilB N-terminal, PilZ and FimX C-terminal fragments. We present the crystal structures of two binary complexes: i) that of the PilB N-terminal domain, encompassing sub-domains ND0 and ND1, bound to PilZ and ii) PilZ bound to the FimX EAL domain within a larger fragment containing both GGDEF and EAL domains. Evaluation of PilZ interactions with PilB and the FimX EAL domain in these and previously published structures, in conjunction with mutagenesis studies and functional assays, allow us to propose an internally consistent model for the PilB-PilZ-FimX complex and its interactions with the PilM-PilN complex in the context of the inner membrane platform of the X. citri Type IV pilus.


Abstract
Type IV pili (T4P) are thin and flexible filaments found on the surface of a wide range of Gram-negative bacteria that undergo cycles of extension and retraction and participate in a variety of important functions related to lifestyle, defense and pathogenesis. During pilus extensions, the PilB ATPase energizes the polymerization of pilin monomers from the inner membrane. In Xanthomonas citri, two cytosolic proteins, PilZ and the c-di-GMP receptor FimX, are involved in the regulation of T4P biogenesis through interactions with PilB. In vivo fluorescence microscopy studies show that PilB, PilZ and FimX all colocalize to the leading poles of X. citri cells during twitching motility and that this colocalization is dependent on the presence of all three proteins. We demonstrate that full-length PilB, PilZ and FimX can interact to form a stable complex as can PilB N-terminal, PilZ and FimX C-terminal fragments. We present the crystal structures of two binary complexes: i) that of the PilB N-terminal domain, encompassing sub-domains ND0 and ND1, bound to PilZ and ii) PilZ bound to the FimX EAL domain within a larger fragment containing both GGDEF and EAL domains. Evaluation of PilZ interactions with PilB and the FimX EAL domain in these and previously published structures, in conjunction with mutagenesis studies and functional assays, allow us to propose an internally consistent model for the PilB-PilZ-FimX complex and its interactions with the PilM-PilN complex in the context of the inner membrane platform of the X. citri Type IV pilus.

Author summary
Bacteria have to adapt their lifestyles to changing environments, at times deciding to hunker down and establish compact multicellular colonies with intricate 3-dimensional structures, called biofilms, that provide protection against hostile conditions; while at other times deciding to go out on their own to explore new habitats. Both of these lifestyles rely on Type IV pili, long extendable and retractable surface filaments that allow the bacteria

Introduction
Prokaryotes have evolved sophisticated surface nanomachines that allow them to colonize a large variety of niches [1]. One such structure is the type IV pilus (T4P), a flexible filament, 4 to 7 nm in diameter and often several micrometers in length, that can extend, attach to surfaces and retract. T4P are involved in a broad range of functions including twitching motility, adhesion, cell orientation, biofilm formation, pathogenicity, natural transformations and bacteriophage infection [2][3][4][5][6][7]. The protein machinery required for T4P biogenesis and function is highly conserved among phylogenetically distant bacterial species [8,9] and is related to the ubiquitous type II secretion systems (T2SS) as well as archaeal flagella [10]. T4P filaments are produced by the polymerization of pilin subunits in a process that depends on PilB, a hexameric ATPase associated with the bacterial inner membrane, while pilus retraction is powered by another hexameric ATPase, PilT [11][12][13]. We have a very rudimentary understanding about these polymerization/depolymerization processes except that, in addition to the above mentioned ATPases, they also require an outer membrane channel formed by the PilQ secretin and an inner membrane platform formed by integral proteins PilC, PilN, PilO, PilP and the cytoplasmic protein PilM [9,[14][15][16][17]. In contrast to these highly conserved structural components, each of the principle model organisms for which T4P have been extensively studied, such as Pseudomonas aeruginosa, Neisseiria spp., Synechocystis spp., Vibrio cholerae, Myxococcus xanthus and Xanthomonas spp., present unique aspects which point to the evolution of a variety of different molecular mechanisms by which the T4P polymerization and retraction can be controlled [8,[18][19][20]. For example, the function of the PilB homolog MshE from V. cholerae depends on the binding of Bis-(3 0 -5 0 )-cyclic diguanylate (cdi-GMP) to its N-terminal MshEN-N domain [21], while Xanthomonas and Pseudomonas PilB proteins lack the conserved amino acid residue motifs required for c-di-GMP binding.
In Pseudomonas aeruginosa and in several phytopathogenic bacteria of the genus Xanthomonas, two regulators of T4P function are PilZ [22,23] and FimX [4,[24][25][26][27]. Pseudomonas and Xanthomonas PilZ are small proteins that do not bind c-di-GMP in spite of belonging to the PilZ superfamily of proteins, many known to be c-di-GMP receptors [23,27,28]. FimX, on the other hand, is a large protein with four domains: REC, PAS, GGDEF and EAL, three of which can be considered degenerate in relation to their canonical functions. The FimX REC domain lacks a site for phosphorylation by a cognate histidine kinase, the GGDEF domain does not possess diguanylate cyclase activity and the EAL domain does not possess phosphodiesterase activity, although it does retain the ability to bind c-di-GMP [24,25,29]. In X. citri, PilZ has been shown to interact with both FimX and PilB and knockouts of these proteins abolish T4P-dependent functions in X. citri [4,23,24] and P. aeruginosa [22,25,26]. PilZ-FimX interactions have been observed for the proteins from X. campestris pv. campestris [30] and X. oryzae pv. oryzae [31] and PilB-PilZ interactions have been observed in the closely related Xanthomonadaceae species Lysobacter enzymogenes [32]. Interestingly, in the more distantly related P. aeruginosa, experiments failed to detect interactions between PilZ and FimX [33]; instead direct interactions between FimX and PilB have been observed [34].
To better understand the three-way interactions between PilB, PilZ and FimX, we determined the crystal structure of the complex between PilZ and a PilB fragment (PilB  ) corresponding to the N-terminal domain of PilB from the phytopathogen Xanthomonas citri that causes canker disease in citrus plants.  is made up of two sub-domains, ND0 and ND1, and is structurally similar to the N-terminal domains of V. cholera MshE (MshEN, residues 1-145) [21] and X. campestris XpsE (XpsE Nt , residues 1-149) [35]. NMR experiments show that PilB  can undergo a large structural rearrangement upon interaction with PilZ. We demonstrate that full-length PilB, PilZ and FimX can interact to form a stable complex, as can PilB  , PilZ and C-terminal FimX fragments FimX EAL or FimX GGDEF-EAL . We also crystallized the complex formed by PilZ Δ107-117 and FimX GGDEF-EAL . Evaluation of PilZ interactions with PilB  and the FimX EAL domain via functional assays of specific mutants allow us to propose a consistent model for PilB-PilZ-FimX ternary complex and its interactions with other T4P components in the inner membrane. This is the first atomic resolution model of a PilB protein in complex with a specific protein regulator and so provides us with unique structural insights into the means by which the activity of a T4P polymerization ATPase can be controlled.

PilZ interacts with the N-terminal domain of PilB
We have previously shown that PilZ interacts with the hexameric ATPase PilB, required for Type IV pilus biogenesis [23]. Fig 1A shows that the 578 residue PilB protein from X. citri can be divided into 3 domains: an N-terminal domain made up of sub-domains ND0 and ND1, a central ND2 domain and a C-terminal ATPase domain [36]. In order to determine which PilB domains are involved in the interaction with PilZ, we produced three preys consisting of fragments spanning sub domains ND0-ND1 (residues 1-190; PilB  ), domain ND2 (residues 159-306, PilB 159-306 ) and domains ND2-ATPase (residues 159-578, PilB 159-578 ) for use in yeast two-hybrid interaction assays with a PilZ bait. Fig 1B and 1C shows that an interaction was only detected using the PilB  prey. This result allows us to conclude that the PilB-PilZ interaction is mediated via the former´s N-terminal region.
Based on these results, we expressed and purified PilB 1-190 and a smaller PilB N-terminal fragment corresponding to residues 12-163 (PilB  ) that lacks regions predicted to be unstructured at the N-and C-terminal ends of PilB  . Fig 1D shows that PilB 1-190 -PilZ and PilB 12-163 -PilZ complexes can be co-purified after co-expression in E. coli cells (in these experiments the PilB fragments were expressed with a cleavable N-terminal 6xHis tag). 15 N-labelled PilB 12-163 has an 1 H-15 N HSQC NMR spectrum characteristic of a partially disordered polypeptide. However, in the presence of unlabelled PilZ, the 1 H chemical shift dispersion increases which is indicative of a folded protein (Fig 1E). S1 Fig shows the results of typical SEC-MALS (size-exclusion chromatography coupled with multi-angle light scattering) experiments that show that PilB  and PilB  are predominantly monomeric on their own and each can form 1:1 binary complexes with PilZ.
In order to compare the affinities of PilZ for PilB and FimX, we expressed PilZ in a tryptophan auxotrophic strain of E. coli grown in minimal media containing 5-hydroxytryptophan (5OHW). The recombinant PilZ protein, PilZ W69_5OHW , now contains this modified amino acid at position 69, normally occupied by tryptophan. The titration of PilZ W69_5OHW by PilB  was accompanied by a 2.5-fold increase in 5OHW fluorescence and the data indicate that the dissociation constant is in the nanomolar or subnanomolar range (Fig 1F). The calculated K D of 4.1 ± 0.9 nM should be taken as an upper limit since the PilZ protein concentration used in the experiment (1 μM) is between two and three orders of magnitude greater than the calculated dissociation constant. PilZ W69_5OHW was also titrated by a fragment corresponding to the C-terminal EAL domain of FimX (FimX EAL ) [24], in which case a more moderate 25% increase in fluorescence was observed and the dissociation constant was calculated to be 0.2 μM ± 0.1 μM (Fig 1F). These results indicate that the PilZ interaction with PilB 1-190 is at least one or two orders of magnitude stronger than its interaction with FimX EAL . We note that these experiments were performed in the absence of added c-di-GMP and that the W69_5OHW substitution may be affecting PilZ´s interactions with PilB and FimX to different degrees (see below).

The structure and interface of the PilB 12-163 -PilZ complex
Native and selenomethionine PilB 12-163 -PilZ complexes were crystallized as described in Materials and Methods, in space group P2 1 . Two native crystal datasets were obtained, one with the wild type sequences and another with a spontaneous P70S mutation in the PilB 12-163 subunit. Data collection statistics for all three crystals are presented in S1 Table. Initial phases were estimated by single wavelength anomalous dispersion. The asymmetric unit contains two copies of the PilB 12-163 -PilZ complex (chains AC and chains BD) (S2A Fig) and the overall structure of the heterodimeric complex is shown in Fig 1G and 1H. The PilB 12-163 -PilZ interface buries 1026 Å 2 and 1018 Å 2 of surface area for the AC and BD PilB 12-163 -PilZ dimers, respectively, as calculated by PISA [37]. A list of residues found at the interface is presented in S2 Table. Structural alignment of the two PilB 12-163 fragments (chains A and B) and the two PilZ molecules (chains C and D) show that they are very similar with a root mean square deviation (RMSD) of 0.6 Å and 0.4 Å for C α atoms, respectively. No electron density was observed for the last 6 residues of PilB 12-163 and the first 8 residues of PilZ.
PilB  is made up of two sub-domains, a four helix ND0 (α1-α4) and an α/β ND1 (a β1-β3 antiparallel β-sheet surrounded by three helices (α5-α7) and a string of three consecutive short 3 10 helices (η1-η3)) (Figs 1G and S2B). The ND0 and ND1 sub-domains are topologically similar to the MshEN_N and MshEN_C domains, respectively, of the V. cholerae T4P ATPase MshE. It is noteworthy that X. citri PilB sub-domain ND0 lacks the specific amino acid sequence motifs required for c-di-GMP binding found in the MshEN_N domain.
Bioinformatics analysis indicates highly conserved PilB residues at the PilB-PilZ interface. S3 Table lists 51 bacterial species from 51 different genera whose genomes, all annotated in the KEGG database [38], code for all three PilB, PilZ and FimX homologs as well as a homolog of the conserved T4P component PilM. All except one of these genera are from the class Gammaproteobacteria but are distributed among different orders and families. The coincidence of genes coding for close PilB, PilZ and FimX homologs in different genomes could reflect a shared and conserved mechanism of T4P regulation in these species. We note that all of the PilB sequences in this list present both ND0 and ND1 sub-domains and none of the PilB sequences carry residues implicated in the binding of c-di-GMP by MshE and its closest homologs. Fig 2A presents the conservation profile of the N-terminal domains of these PilB proteins using the numbered positions of X. citri PilB as a reference. This conservation profile is mapped onto the PilB 12-163 structure in Fig 2B-2D. Fig 2B presents

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex PilB sequence conservation and interactions. A) Sequence logo generated from an alignment of 50 non-redundant PilB N-terminal sequences that contain ND0 and ND1 sub-domains. The X. citri PilB N-terminal (residues 1-190) sequence is shown below the alignment (highlighted in green). The secondary structure elements observed in the crystal structure of X. citri PilB (residues 12-163) are indicated above the alignment. B-D) Surface residues of PilB 12-163 colored according to degree of conservation. Panel B shows the PilB 12-163 surface oriented towards PilZ in the PilB 12-163 -PilZ complex. � Indicates residues involved in direct contacts with PilZ. Panels C and D show PilB 12-163 after 120 o and 240 o rotations with respect to A. The surface in C presents residues proposed to be involved in the interaction with PilM. E) Possible mode of interaction between PilB , PilZ and PilM. Shown is the superposition of the X. citri PilB 12-163 -PilZ complex with homologous EpsE ND1 -EpsL complex (PDB ID: 2BH1) using the ND1 sub-domains as reference (see S12 and S13 Figs). PilZ and the EpsL cytoplasmic domain (PilM homolog) are shown in ribbon representation colored in yellow and blue, respectively. F) Proposed model for the PilB-PilZ-FimX ternary complex based on the superposition of the PilB 12-163 -PilZ complex and the PilZ-FimX EAL complex (interface 2). PilZ is the reference in the

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex also Fig 1G, 1H and S2 Table). Fig 2C and 2D presents the exposed PilB 12-163 surfaces after rotations of 120 o and 240 o about the vertical axis, respectively. Interestingly, in addition to the conservation of residues interacting with PilZ, the PilB 12-163 surface viewed by the 120 o rotation is also very well-conserved while the surface viewed by the 240 o rotation is relatively poorly conserved. Below, we will show that this second conserved PilB surface is most likely involved in interactions with PilM (Fig 2E).
The superposition of PilZ of the PilB 12-163 -PilZ complex with the previously described crystallographic structures of X. citri PilZ [23] and the PilZ-FimX EAL complex [24] shows that they are very similar, with a RMSD of 0.7 Å and 0.5 Å for C α atoms, respectively (S2D Fig). The most significant difference between these three PilZ structures is that the electron density for the last 11 PilZ residues (residues 107-117) is well-defined in the PilB 12-163 -PilZ complex (Figs 1I and S2E and S2F), while no density is observed for these residues for PilZ alone [23] and in the FimX EAL -PilZ complex [24] (S2D Fig). These C-terminal residues (called motif V, or MV) are highly conserved in X. citri PilZ orthologs, including PilZ PA2960 from P. aeruginosa, and are required for the interaction with PilB but not with FimX EAL [23]. Specifically, in the PilB 12-163 -PilZ complex, the carboxylate group of the C-terminal Met117 residue of PilZ makes a salt bridge with highly conserved Arg103 of PilB while the Met117 side chain fits into a hydrophobic pocket made up of PilB residues Leu72, Phe77, Leu100, Phe101, Phe108 and Ile134 and conserved PilZ residues Phe28 and Trp69 (Fig 1I and 1J and S2 Table). In addition to motif V, many other PilZ residues conserved in the PA2960/XAC1133 orthologous group [23] also participate in interactions with PilB (S2 Table).

Mutations at the PilB-PilZ interface destabilize the complex
We produced mutants in specific residues at the PilB-PilZ interface and tested the stability of these complexes using size-exclusion chromatography (SEC). S3A-S3F Fig shows the results using PilZ mutants in complex with PilB  (this larger PilB fragment was used due to its greater solubility in the absence of PilZ). Mutations in conserved PilZ residues Y22 and F28 did not affect binding to PilB   (S3A and S3B Fig). On the other hand, mutating PilZ residue W69 to alanine severely reduced the stability of the complex (S3C Fig). The PilZ Δ107-117 mutant also failed to make a complex with PilB   (S3E Fig), consistent with previous observations [23]. Since the C-terminal residue M117 makes a number of contacts in the interface (Fig 1J), we produced two more PilZ mutants, one in which M117 was deleted (eliminating both carboxy-terminal and side chain interactions) and another in which it was substituted for a glycine residue (eliminating side-chain interactions). In both cases, interaction with PilB  was abolished (S3D and S3F Fig). We also mutated PilB residues found at the PilB-PilZ interface: F77A, F101A, F108A, R103A and E132A and the double mutant F101A/F108A. S3G-S3K Fig shows that all of the single mutants in PilB  retain the ability to interact with PilZ. However, the F101A/F108A double mutant no longer interacts (S3L Fig). Together these superposition. G) Above: SEC analysis of the mixture of full-length PilB-PilZ complex and full-length FimX shows that these proteins form a ternary complex (black continuous line). Addition of ATPγS and c-di-GMP results in a shift in the elution profile of the ternary complex (red continuous line). Also shown is the elution profile of the PilB K343A/E407A -PilZ and FimX mixture in absence (dashed black line) and presence (dashed red line) of ATPγS and cdi-GMP. In all cases, a 1:1 molar ratio of PilB-PilZ complex to FimX was used. The elution profiles of fulllength FimX (blue continuous line) and the PilB-PilZ complex (green continuous line) are also shown. To facilitate comparison, peak heights were normalized. Below: SDS-PAGE of the fractions of the major peak observed for the PilB-PilZ-FimX ternary complex in presence of ATPγS and cdi-GMP. H) ATPase activity of the PilB-PilZ complex in the absence and presence of full-length FimX. ATPase activity was determined by measuring the production of inorganic phosphate (Pi) at 30˚C using 1 μM of PilB-PilZ complex, 1 mM de ATP in the absence and presence of 1 μM FimX. Where indicated, c-di-GMP was added to a final concentration of 20 μM. When wild-type PilB was substituted with the PilB K343A/E407A mutant, no significant ATPase activity was observed. Each experiment was performed at least three times. Error bars: standard deviation. https://doi.org/10.1371/journal.ppat.1009808.g002

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex mutation studies point to the importance of hydrophobic interactions in the stabilization of the PilB-PilZ interface.

Determination of the relevant PilZ-FimX interface in the PilB-PilZ-FimX complex: PilZ interacts with FimX through a second interface distinct from that between PilZ and PilB
We also crystallized the complex between PilZ Δ107-117 and a larger FimX fragment encompassing its GGDEF and EAL domains (FimX GGDEF-EAL , residues 255-689) in the presence of c-di-GMP (S4 Fig). The PilZ Δ107-117 -FimX GGDEF-EAL complex crystallized in space group P4 2 2 1 2 with one copy of each protein in the asymmetric unit (see S1 Table for data collection statistics). Very clear electron density for PilZ Δ107-117 and the FimX EAL domain (beginning at residue 436) with bound c-di-GMP in the syn/anti conformation were observed (S4C and S4D  Fig). However, no density for the FimX GGDEF domain could be detected (S4D and S4E  Fig), a phenomenon similar to that observed for the crystal structure of the FimX GGDEF-EAL fragment from P. aeruginosa [29] that is only 30% identical to X. citri FimX GGDEF-EAL .
The PilZ Δ107-117 -FimX GGDEF-EAL crystal presents the same two principal modes of crystal lattice contacts between PilZ and the FimX EAL domain observed in the previously published PilZ-FimX EAL crystal structures from X. citri and X. campestris (Fig 3A-3C), in spite of the different fragments used and different space groups [24,30]. We name these two modes of contact interfaces 1 and 2. The most significant difference among the three structures has to do with the orientation of the first helix in FimX EAL domains: in the X. citri PilZ-FimX EAL and PilZ Δ107-117 -FimX GGDEF-EAL complexes (residues 429-454 and 436-454, respectively), the helix is oriented so as to contribute significantly to the interaction with PilZ in interface 1 (362 or 308 Å 2 and 234 Å 2 of buried surface, respectively) while this helix is rotated approximately 140 o , pointing away from PilZ in the X. campestris PilZ-FimX EAL structure [30] (Fig 3D and  3E). If the contribution of this helix is ignored, then the two interfaces bury similar amounts of surface area (Fig 3E). There are significant differences in the nature of the interfaces, however. While interface 1 involves a β-sheet extension and tenuous contacts between PilZ and the c-di-GMP ligand bound to FimX [24,30], interface 2 is dominated by hydrophobic interactions.
S5 Fig shows that the PilZ residues involved in contacts with FimX EAL via interface 1 overlap significantly with PilZ residues involved in its interactions with PilB. Since we have previously reported that a ternary PilB-PilZ-FimX GGDEF-EAL complex can be detected in farwestern overlay assays, and that mutating conserved PilZ residue Trp69 to alanine (located at interface 1) reduces interactions with PilB but does not abolish PilZ-FimX EAL interactions [23,24] we made a set of PilZ mutants to test the relevance of interface 2 for the stability of the PilZ-FimX interaction. Figs 3F and S6 show that mutations in PilZ residues Ile10, Phe49 and Leu51 (mutants I10E and F49E/L51E) significantly reduce the stability of the PilZ-FimX GGDE-F-EAL binary complex while the PilZ interaction with PilB 1-190 is maintained. These mutants have thermal stabilities (66 o C and 61 o C, respectively) very similar to that of wild-type PilZ (61 o C) (S7 Fig). On the other hand, simultaneously mutating non-interfacing PilZ residues D46 and E47 to alanine does not interfere with its interactions with FimX GGDEF-EAL or with PilB   (Figs 3F and S6), even though it presents a lower Tm (48.8 o C) than wild-type PilZ (S7 Fig). We note that the binary complexes employing wild-type and D46A/E47A PilZ seem to be stabilized by the addition of c-di-GMP (Figs 3F and S5). Together, these observations raise the possibility that interface 2 between PilZ and FimX EAL domains may be physiologically relevant. Importantly, this second mode of contact would allow the simultaneous interaction of PilZ with both FimX and PilB as shown in Fig 2F. This hypothesis was tested in the experiments described below.

PilZ bridges PilB and FimX to form a ternary complex
When employing PilB 12-163 or PilB  , ternary complexes can be observed in SEC experiments using PilZ and FimX EAL or FimX GGDEF-EAL fragments, but not when using full-length FimX or FimX PAS-GGDEF-EAL (a FimX fragment (residues 153-689) lacking the N-terminal REC domain) (S8 Fig). We therefore determined if mutations in the PilZ-FimX and PilB-PilZ interfaces affect the stability of the PilB 1-190 -PilZ-FimX GGDEF-EAL complex. Indeed, mutations The principal crystal contacts observed between PilZ and FimX subunits in the A) PilZ-FimX EAL -c-di-GMP complex from X. citri (PDB: 4FOU; FimX EAL colored in orange and PilZ colored in blue), B) PilZ-FimX EAL -c-di-GMP complex from X. campestris pv. campestris (PDB: 4F48; FimX EAL colored in green and PilZ colored in yellow) and C) PilZ Δ107-117 -FimX GGDEF-EAL -c-di-GMP complex from X. citri (this study; FimX EAL colored in magenta and PilZ colored in cyan). The dashed black circles and dashed red boxes delimit the two principal crystal contacts (interface 1 and 2, respectively). Right: Detailed view of interface 2 in the three structures. D) Superposition of one FimX EAL domain with the two contacting PilZ subunits via interfaces 1 and 2 observed in the three structures shown in A-C with the same coloring scheme. Note the different orientations of the first alpha helix in the X. citri and X. campestris pv. campestris FimX EAL domains. This explains the reduced contact surface at interface 1 in the latter structure (see part E). E) Buried surface areas at the two interfaces in the three structures as calculated by the PISA server [37]. F) Size exclusion chromatography of PilZ-FimX GGDEF-EAL binary complexes containing wild-type PilZ (PilZ Wt ) and its mutants (PilZ F49E/L51E , PilZ I10E and PilZ D46A/E47A ). In each chromatogram, the elution profile of the FimX GGDEF-EAL −PilZ mixture (1:1.5 molar ratio) was performed in the absence (black line) and in the presence (red line) of a 2-fold excess of c-di-GMP to FimX GGDEF-EAL . In these experiments, FimX GGDEF-EAL has an N-terminal 6xHis-tag. Each experiment was performed at least three times and representative results are shown. These chromatograms are reproduced in S5A

Full-length PilB-PilZ-FimX complexes
Recombinant full-length X. citri PilB containing an N-terminal polyhistidine tag (theoretical MW of 64.5 kDa) is insoluble when expressed on its own in E. coli (Fig 5A) but is soluble when co-expressed with PilZ (theoretical MW of 12.4 kDa) (Fig 5B). This full-length PilB-PilZ complex is stable and can be purified by affinity and size-exclusion chromatography (Figs 2G and 5C). The complex elutes as a broad peak with molecular weight, estimated by SEC-MALS analysis, that varies between 100 kDa and 140 kDa (Fig 5D). However, in the presence of ATPγS, the elution volume of this peak shifts slightly with an estimated molecular weight of 120-160 kDa (Fig 5D). These molecular weights are suggestive of a dynamic equilibrium between PilB-PilZ heterodimers (theoretical MW of 77.4 kDa) and (PilB-PilZ) 2 heterotetramers (theoretical MW of 154.8 kDa). Furthermore, the addition of ATPγS results in the appearance of a minor higher molecular weight peak with a large molecular weight distribution (between 350 and 600 kDa), indicative of larger aggregates (Fig 5D). These observations are consistent with previous reports on isolated PilB homologs from other systems that were observed as monomers, dimers and hexamers [12,[39][40][41][42][43][44][45].
Full-length X. citri FimX on its own has a molecular weight estimated by SEC-MALS of 145 and 147 kDa, in the absence and presence of its c-di-GMP ligand, respectively (Fig 5E). These molecular weights are consistent with dimer formation (theoretical MW of 152 kDa) as previously reported for the homologous protein in P. aeruginosa [29]. In contrast to results using PilB N-terminal fragments, when using full-length PilB, we are able to clearly observe ternary PilB-PilZ-FimX complexes containing full-length FimX (Fig 2G). These results indicate that the complete PilB protein is required for the stable incorporation of full-length FimX into the complex. SEC-MALS analysis of the PilB-PilZ-FimX complex again points to a heterogeneous mixture with molecular weight varying between 500 and 650 kDa (S9A Fig). The addition of ATPγS and c-di-GMP seems to stabilize this higher molecular weight complex (Figs 2G and S9A). We note that a PilB-PilZ-FimX stoichiometry of 6:6:2 corresponds to a theoretical molecular weight of approximately 620 kDa (see Discussion below).

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex were used to test their effect on ternary complex stability. In all the cases, a 1:1:1 PilB 1-190 :PilZ:FimX GGDEF-EAL molar ratio was used and the elution profile of the mixture is shown as a black line. All the mutants, except PilZ F28A and PilZ D46A/E47A , affect the stability of the FimX GGDEF-EAL -PilZ-PilB 1-190 ternary complex. In all of the panels, the elution volume for the ternary complex is delimited by the vertical red broken lines. In the first panel, the elution profiles for wild type complex (black line) is shown together with the profiles for the three individual components FimX GGDEF-EAL (red line), PilB  (blue line) and PilZ (green line). Each experiment was performed at least three times and representative results are shown. B) Top row: Fluorescence microscopy images of the edges of the twitching zones at the interstitial surface between the agar medium and the glass base of the microscopy chamber. For visualization, X. citri strains were transformed with the pBBR2-GFP plasmid. Wild-type cells, that are able to twitch, can separate from the main body of the colony and migrate on their own or in small groups, producing a rough, less organized boundary between the dense interior of the colony and the surrounding medium. Cells with mutations that compromise T4P function are not able to separate from the main body of the colony and so the colony border is much smoother and well-defined. Bottom row: Phage FXacm4-11 infection assays. Dark plaques are indicative of phage-induced bacterial lysis in a confluent culture background. For both twitching motility and bacteriophage infection assays: X. citri wild type (Xac_Wt), Xac_ΔpilZ XAC1133 (ΔpilZ) and Xac_ΔpilZ XAC1133 complemented with a plasmid (pURF047) directing the expression of the wild-type PilZ protein (ΔpilZc) or one of the following PilZ mutants: PilZ I10E

FimX enhances PilB ATPase activity
PilB proteins from other bacterial species have been shown to exhibit ATPase activity in vitro [12,40,46]. Fig 2H shows that the X. citri PilB-PilZ complex can hydrolyze ATP in vitro with an intrinsic activity of approximately 8.5 nmol ATP hydrolyzed/min per mg PilB-PilZ complex, under the conditions tested. Upon addition of full-length FimX, an approximately 3-fold increase in ATPase activity was observed (22.5 nmol/min per mg PilB-PilZ complex; Fig 2H). As expected, no significant activity was observed for the PilB-PilZ and PilB-PilZ-FimX complexes containing PilB K343A/E407A in which two active site residues, located within the ATPase domain, were mutated to alanine (Fig 2H). Fig 5F shows that this PilB mutant retains the ability to form a binary complex with PilZ but no mobility shift or evidence of higher order oligomers is observed upon addition of ATPγS. Finally, Figs 2G and S9B show that ternary complexes containing the PilB K343A/E407A mutant, PilZ and FimX do form; but the higherorder oligomers seem to be less stable. These results show that i) FimX stimulates PilB ATPase activity, ii) the incorporation of FimX into the complex is not absolutely dependent on PilB ATPase activity and iii) PilB ATPase activity stabilizes higher order oligomeric forms of the PilB-PilZ binary and PilB-PilZ-FimX ternary complexes.

Unipolar colocalization of PilB, PilZ, FimX and PilQ at the leading edge of twitching X. citri cells
In order to study the localization of the X. citri Type IV pilus components, we produced strains in which the FimX, PilZ, PilB and PilQ genes were replaced with msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP fusions, respectively. Figs 6 and S10A show that FimX, PilZ and PilB all are localized predominantly at a single cell pole. The fraction of cells exhibiting polar foci was 54% for FimX, 32% for PilZ but only 6.5% for PilB under the conditions tested (Fig 6D). The outer membrane secretin subunit PilQ is also found mostly at one pole (79%) with a small fraction of cells showing foci at two poles simultaneously (Fig 6C and 6D). In the cases where bipolar PilQ localization was observed, one of the foci was almost always much more intense than the other (see S10A In order to determine if protein localization is influenced by whether the cells are grown in liquid (no twitching) or semisolid media (twitching) we grew mCherry-FimX and PilQ-msfGFP cells in liquid medium and observed them immediately after transfer to KB-agarose slabs (before they have a chance to begin twitching, time = 0 hours) and 6 hours after transfer during which time they are exhibiting twitching behaviour. S10B Fig shows that no mCherry-FimX foci are observed in cells immediately after transfer from liquid culture and that the foci appear only after growth for some time on semisolid media. In contrast, it was very common to observe PilQ-msfGFP foci at both cell poles (bipolar localization) immediately after growth in liquid medium; but after several hours growth on semi-solid media, most cells exhibited unipolar PilQ-msfGFP foci (S10B Fig). S11 Fig and S1-S4 Movies show that these X. citri strains all exhibit normal twitching motility and can all be infected with bacteriophage FXacm4-11, X. citri phenotypes that are dependent on a functional T4P [4].
Time-lapse fluorescence microscopy images of actively twitching X. citri msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP strains reveal that the most intense foci are found at the leading poles of the cells at the leading edge of the group of migrating cells (Fig  6B and S5-S8 Movies). These strains were then used to introduce deletions in specific Type IV pilus components. Fig 6A and 6D shows that deletion of pilZ or pilB genes results in a drastic reduction in the number of cells with fluorescent msfGFP-FimX foci. Likewise, deletion in the pilZ or pilB genes result in drastic reductions in the unipolar localization of msfGFP-PilB
To confirm that FimX, PilZ, PilB and PilQ are in fact all localized at the same pole, mCherry-FimX fusions were used to substitute the FimX gene in the msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP strains to produce cells in which two T4P components could be simultaneously localized by fluorescence microscopy. These three strains are all able to be infected by FXacm4-11, indicative of normal T4P function (S11B Fig). Fig 7 shows fluorescence microscopy images of groups of X. citri cells in which most of the cells displaying mCherry-FimX foci also display msfGFP-PilB, msfGFP-PilZ or PilQ-msfGFP foci at the same pole. These results are consistent with studies on the co-localization of FimX, PilB and PilQ homologs in P. aeruginosa [25,26,34,47].

Discussion
The crystal structures of the N-terminal domains of four relatively distant PilB homologs are available in the Protein Data Bank (PDB): V. cholerae MshE Nt (PDB: 5HTL), the X. campestris T2SS protein XpsE Nt (PDB: 2D28), V. vulnificus T2SS GspE Nt (PDB: 4PHT) and V. cholerae T2SS EpsE Nt (PDB: 2BH1), the latter two in complex with the cytoplasmic domains of the PilM paralogs GspL and EpsL, respectively. These structures show a very similar topology in spite of rather poor sequence alignment with X. citri PilB 12-163 (S12 and S13 Figs). MshE Nt and XpsE Nt contain both ND0 and ND1 sub-domains and align with X. citri PilB 12-163 with r. m.s.d. of 3.3 Å and 3.1 Å, respectively. Interestingly, V. cholerae MshE Nt binds the cyclic dinucleotide c-di-GMP [21,48]. However, residues involved in nucleotide binding [21] are not conserved in X. citri PilB and its homologs found in the genera listed in S3 Table. The N-terminal domains of GspE from V. vulnificus and EpsE from V. cholerae both lack the ND0 sub-domain and the structural alignment of their ND1 sub-domains with PilB 12-163 have r.m.s.d.s of 1.9 Å and 2.5 Å, respectively (S12C Fig).
The NMR spectrum of X. citri PilB  indicates that it is at least partially disordered on its own and becomes fully folded upon interaction with PilZ. This conformational heterogeneity is consistent with the following observations: i) The XpsE N-terminal domain has been crystallized in two conformations in which the ND0 sub-domain is found in an open, less-compact configuration or in a closed configuration [35] that more closely resembles that observed in X. citri PilB 12-163 in complex with PilZ. ii) The crystal structures of full-length PilB proteins from Geobacter metallireducens and G. sulphurreducens lacked observable electron density for their ND0/ND1 domains [35,36,44]. Both proteins are predicted to bind c-di-GMP, though the cyclic nucleotide was not included during crystallization. iii) The cryoEM structures of PilF from Thermus thermophilus only produced maps of very low resolution for the second and third ND0/ND1 domains while the first was not visible at all [49]. Again, two of the PilF ND0/ in wild type and mutants strains with deletion in indicated T4P regulatory components. The X. citri strains expressing the fluorescent fusion proteins were imaged by phase contrast and epifluorescence microscopy and analyzed at least five times independently with similar results. In these experiments, cells were grown on KB-agarose pads (1.5% w, using 0.2% casamino acids as nitrogen source) supplemented with 2 mM CaCl 2 . Scale bar, 3 μm. B) Representative time lapse images showing that msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP localize to the leading pole in wild type X. citri cells undergoing twitching motility. The time lapse interval (h) is indicated for each frame. Images were taken by using an epi-fluorescence light microscope. Scale bar, 5 μm. See also S5-S8 Movies for time lapse images at shorter intervals. C) Graphical representation of the fluorescence intensity profile over the length of X. citri cells expressing msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP. The fluorescence intensity profile was obtained from 100 individual X. citri cells for each case (cell length was normalized). The profiles of all individual cells are shown in S10A Fig. Insets show the foci localizations. A and B show that msfGFP-FimX, msfGFP-PilZ and msfGFP-PilB foci exhibited unipolar localization. PilQ-msfGFP exhibits a mixture of both unipolar and bipolar localization; in the latter case one of the foci was almost always much more intense than the other (see S10A Fig). D) Comparison of the frequency of polar localization of msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP foci in wild-type, ΔfimX, ΔpilZ and ΔpilB backgrounds. https://doi.org/10.1371/journal.ppat.1009808.g006

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex ND1 domains have been shown to bind c-di-GMP [21,50,51] but the cyclic nucleotide was not included during cryo grid preparation. The physiological relevance of PilZ-induced folding of the N-terminal domain is not clear at the moment since we do not know whether the PilB-PilZ interaction is static or if PilZ dissociates and reassociates at some stage in the ATPase/polymerization cycle or during a signaling pathway that leads to assembly or disassembly of the PilB hexamer. Our estimation of the dissociation constant of the PilB 1-190 -PilZ W69_5OHW complex in the nanomolar range makes PilZ dissociation from PilB unlikely.

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex In addition to its interactions with PilB, X. citri PilZ also interacts with the C-terminal EAL domain of FimX. Interface 2 observed in the different crystal forms of the X. citri PilZ-Fim-X EAL and PilZ Δ107-117 -FimX GGDEF-EAL complexes and the X. campestris PilZ-FimX EAL complex (Fig 3) allowed us to propose a model for the PilB-PilZ-FimX ternary complex (Fig 2F) that is consistent with a large amount of structural, biochemical, cellular and genetic data described above. We note that while homologous PilZ-FimX interactions have been observed for the highly similar proteins in X. campestris pv. campestris [30] and X. oryzae pv. oryzae [31], in vitro assays were not able to detect interactions between FimX and PilZ from P. aeruginosa [33]; instead, direct interactions between FimX and PilB were reported, though the precise domains of the two proteins involved in this interaction were not identified [34]. We therefore speculate that there are probably strong parallels in the Pseudomonas and Xanthomonas systems but that binary interactions between purified components from the different species may have different in vitro affinities that may make them difficult to detect out of the context of the native T4Ps.
At this point it is worth considering the orientation of the PilB-PilZ-FimX complex with respect to other T4P components embedded in or associated with the inner membrane (the so-called inner membrane platform). The orientation of the PilB hexamer and its T2SS homologs has been proposed based on crystal structures of the G. metallireducens PilB ND2-ATPase fragment hexamer in the presence of ADP and the non-hydrolysable ATP analogue AMP-PNP [34,36], the GspE-GspL complex from V. vulnificus [52] and cryo-tomography maps of the M. xanthus T4P [9] and the Legionella pneumophila T2SS [53]. In all of these models, the plane of the PilB hexameric ring is parallel with that of the membrane, with the C-terminal ATPase domain oriented towards the cytosol and the ND2 domain pointing towards the membrane. Since the ND0/ND1 and ND2 domains are connected by what is expected to be a negatively charged flexible linker (see below), the ND0/ND1 domain is expected to be free to make contacts with interaction partners in the inner membrane platform. One strong candidate to mediate such interactions is PilM since interactions between PilB and PilM homologs have been observed for T4P of M. xanthus [54], T. thermophilus [41], P. aeruginosa [55], the bundle-forming pilus machinery of enteropathogenic E. coli [56] and the T2SS of X. campestris [35]. Furthermore, complexes between PilB and PilM homologs in T2SS have been crystallized: full-length ATPase GspE with the cytosolic domain of GspL from V. vulnificus [52] and EpsE ND1 domain with the cytosolic domain of EpsL from V. cholerae [57]. S13 Fig shows the superposition of the X. citri PilB 12-163 -PilZ complex with that of the GspE-GspL and EpsE ND1 -EpsL complexes using the ND1 sub-domains as reference. In both GspE-GspL and EpsE ND1 -EpsL structures, the PilM homolog (GspL or EpsL) binds to a surface of the PilB homolog (GspE or EpsE) which we name the PilM interface of PilB. Importantly, the PilM interface does not overlap with the PilB surface to which X. citri PilZ binds (called the PilZ interface in Figs 2E and S13). The residues on this ND1 surface are particularly well conserved in PilB proteins from organisms that also code for PilZ and FimX homologs (Fig 2C).
The above considerations allow us to envision how the PilB-PilZ-FimX complex can interact with the inner membrane platform via PilM which is in turn bound to the cytosolic N-terminal portion of the integral membrane protein PilN [55,58] as depicted in Fig 8. Studies in M. xanthus suggest that PilM and PilN form a dodecameric ring around a concentric PilB hexamer [9]. We used this architecture to incorporate our model of the PilB ND0/ND1 -PilZ-Fim-X EAL complex. In this model, the central hexameric core made up of PilB ND2 and ATPase domains is surrounded by six copies of the PilM-PilN-PilB ND0/ND1 -PilZ-FimX EAL complex.
Here, we point out that the PilB ND1 sub-domain is separated from the ND2 domain by a flexible linker of approximately 30 amino acids that is rich in acidic residues (Fig 8B). For example, the sequence from positions 161 to 193 in X. citri PilB

Fig 8. The PilB-PilZ-FimX complex in the context of the inner membrane platform of the Type IV pilus. A)
Model for interactions between the PilB ND0/ND1 -PilZ-FimX EAL complex and the PilM-PilN N-terminal complex. PilB ND0/ND1 (green) and PilZ (yellow) are shown as determined in the PilB 12-163 -PilZ structure (present study). FimX EAL (blue) is placed as observed in interface 2 of the X. citri PilZ Δ107-117 -FimX GGDEF-EAL (this study), X. citri PilZ-FimX EAL [24] and the X. campestris pv. campestris PilZ-FimX EAL [30] complexes (see Fig 3). The homology model for X. citri PilM (see Materials and Methods) was oriented with respect to PilB ND0/ND1 based on the V. vulnificus GspE-GspL [52] and V. cholerae EpsE-EpsL [57] crystal structures (see Figs 2E and S13). B) The PilB ND0/ND1 domain is connected to the hexameric PilB core (made up of ND2 and ATPase domains) via a highly acidic and glycine-rich linker (approximately 30 residues; see Fig 2A). The homology model of the X. citri PilB hexameric core was built based on the G. metallireducens PilB core structure as described in Materials and Methods. C) Depiction of possible interactions of the PilB-PilZ-FimX EAL complex with the proposed PilM-PilN dodecamer and PilC dimer based on the cryo-electron tomography model of the M. xanthus T4P [9]. The homology model of the PilC dimer (depicted in orange) was built as described in Materials and Methods. Coloring scheme for the other subunits is the same as in part A. Top and side views are shown. Note that the REC, PAS and GGDEF domains of FimX are not shown and that X. citri (this study) and P. aeruginosa FimX is a homodimer, probably due to interactions between N-terminal REC domains [29]. Here the model assumes a PilB-PilZ-FimX stoichiometry of 6:6:6. However, since the PilB hexamer may exhibit C2 symmetry, stoichiometries of 6:6:4 or 6:6:2 are also possible (see main text). https://doi.org/10.1371/journal.ppat.1009808.g008

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex (DDEEGMGDLDVSAGDEDMGAGGDSGVDAKGDDT) contains 8 glycines, 10 aspartates, 3 glutamates and only one positively charged lysine (Fig 2A). This suggests that the orientation of the ND0/ND1 domain of PilB with respect to the central hexameric core made up of ND2 and ATPase domains could be highly variable or at least susceptible to modulation by other regulatory components. Furthermore, each PilN-PilM-PilB ND0/ND1 -PilZ-FimX EAL complex is shown to alternate with another PilM-PilN complex (Fig 8C). Finally, the central lumen of the hexameric core of PilB is thought to accommodate the cytosolic domains of the dimeric integral membrane protein PilC (Fig 8C) to which conformational changes induced by ATP hydrolysis are coupled so as to promote the incorporation of pilin subunits into the base of the growing pilus [36,59,60].
The interactions depicted in the model shown in Fig 8 immediately suggest possible schemes by which FimX and PilZ may regulate T4P function. Such mechanisms would eventually take into account the binding of ATP by PilM, PilM binding to PilN and PilB (as well as PilT and PilC [55]), the binding of c-di-GMP to FimX, the possibility of alternative PilZ-FimX binding modes (via interfaces 1 and 2), the conformational flexibility of the PilB N-terminal domain and the conformational flexibility afforded by the acidic and glycine-rich linker between ND1 and ND2, the structural transitions induced during ATP binding and hydrolysis by PilB [36] and PilB interactions with PilC. In this model we assume that one PilZ binds to each PilB ND0/ND1 domain, but the stoichiometry of FimX binding to the PilB-PilZ complex is not at the moment clear. One possibility is that one FimX binds to each PilZ-PilB resulting in a PilB-PilZ-FimX stoichiometry of 6:6:6 as depicted in Fig 8C. However, the crystal structures of most PilB homologs are hexameric rings with C2 symmetry, in which three different subunit conformations are adopted, one by each pair of opposing subunits. This opens up the possibility that FimX has different affinities for each of the three conformational states of the PilB-PilZ units in the hexamer and, assuming that FimX dimers are maintained, alternative stoichiometries such as 6:6:4 and 6:6:2 are possible. Structural transitions induced by c-di-GMP binding have been observed in FimX from P. aeruginosa [61] as well as in the present study for X. citri (Fig 5E). Thus, c-di-GMP-induced transitions in FimX could in principle be transferred, through PilZ and the PilB ND0/ND1 domain, to the ND2-ATPase domains in the internal PilB hexameric ring and in this way modulate ATPase activity, its interactions with PilC and the rate of pilus subunit incorporation at the base of the growing pilus. Our observations that FimX promotes the polar localization and ATPase activity of the PilB-PilZ complex is consistent with this general idea. Since depolymerization of the pilus probably requires the partial dissociation of the PilB-PilZ-FimX complex from the T4P platform to give way to the homo-hexameric complex of PilT (and/or PilU) [55,62], one could envision a scheme by which structural transitions in the PilB-PilZ-FimX interface could be transmitted to the PilB-PilM interface so as to promote or impede PilB docking into the platform.
Most bacterial species for which T4P function has been studied in detail, such as M. xanthus, Neisseiria spp., Synechocystis sp., V. cholerae, T. thermophilus, Clostridium perfringens and P. aeruginosa seem to be regulated by different mechanisms [18,19,40,[63][64][65]. We have characterized important interactions between the X. citri T4P ATPase PilB and its regulators PilZ and FimX. At least 50 different bacterial genera from the Gammaproteobacteria class have species whose genomes code for homologs of X. citri PilZ, FimX and PilB (only one species from each genus is included in S3 Table in order to avoid redundancy). We note that i) all of the PilB homologs in the list have both ND0 and ND1 sub-domains even though many from T4P or T2SS in other species lack an ND0 domain, ii) none of the ND0 sub-domains have the conserved motifs that mediate binding to c-di-GMP in other more distant PilB homologs, iii) all of the PilZ homologs belong to the PA2960/XAC1133 orthologous group [23] and iv) all FimX homologs have the same domain architecture as that of X. citri and P. aeruginosa

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex FimX proteins. These observations suggest that in many of these genera, we can expect that PilZ and FimX orthologs interact with PilB and participate in T4P regulation in a manner similar to that described here for X. citri.

Materials and methods
Bacterial strains, bacterial growth and cloning S4 Table shows all strains used in this study. Primer sequences and plasmids are listed in S5 Table. Escherichia coli DH5α was used for DNA cloning. The coding sequence for the different constructions of FimX XAC2398 , PilZ XAC1133 and PilB XAC3398 were amplified by PCR from genomic DNA from X. citri pv. citri strain 306 genomic DNA and cloned into the expression vector pET28a or pET3a [66] in the NdeI and BamHI sites. The full-length PilB and PilZ were cloned into the co-expression vector pETDuet. The coding sequencing for PilB fragment (PilB  , PilB 156-307 and PilB 156-578 ) was amplified by PCR from X. citri genomic DNA (primers listed in S3 Table) and cloned into the two-hybrid bait vector pOBD [67] in the NcoI and XhoI sites. A series of substitutions of amino-acid residues in the PilZ and PilB were carried out by via a single-step PCR protocol using the QuickChange Site-Directed Mutagenesis kit (STRATAGENE) (primers listed in S5 Table). The mutants were used for expression in E. coli or for complementation of X. citri strains. The PilZ wild type and F28A, W69A and Δ107-117 mutants were previously described [23,24]. All clones were confirmed by DNA sequencing. Antibiotics were used at the following final concentrations: kanamycin 50 μg/mL, ampicillin 100 μg/mL. In order to complement the ΔpilZXAC1133 knockouts, fragments coding for wild type PilZ (XAC1133) and its mutants M117G, ΔM117, I10E, F49E, F49E/L51E, F49A/L51A were amplified and cloned into pUFR047, as described previously [23].

Crystallization and structural determination of the PilB 12-163 -PilZ complex
The crystallization of unlabeled and selenomethionine-labeled PilB 12-163 -PilZ complexes was carried out in sitting-drop plates by mixing equal volumes (1 μL) of 4 mg/ml protein and well solution (0.1 M Tris pH 8.5; 2.0 M ammonium sulfate) and growing at 18˚C. For the selenomethionine-labeled crystal, X-ray radiation at 0.97889 Å was used (corresponding to the peak of the fluorescence spectrum). The native crystal dataset was collected using 1.4587 Å radiation for the crystal containing the PilB P70S mutation and 1.5418 Å radiation for the crystal containing the wild-type PilB sequence (S1 Table). The data were indexed and integrated using the XDS program [70]. Selenium sites were found using AutoSol [71]. The resulting electron density map was used to construct a preliminary polyalanine model with ARP/wARP [72]. Interpretation of the electron density maps and construction of the missing residues was performed with COOT [73]. Structural refinement of the model was performed using REFMAC [74], Phenix [75] and COOT. Water molecules were added automatically using REFMAC and manually using COOT. Details of the refinement data statistics are shown in S1 Table. Structural alignments and figures were produced using Chimera [76]. The atomic models and experimentally determined structure factors have been deposited in the protein data bank with pdb codes 7LKM, 7LKN and 7LKO.

Crystallization and structural determination of the PilZ Δ101-117 -FimX GGDEF-EAL complex
The crystallization of native PilZ Δ101-117 -FimX GGDEF-EAL complex was performed in sittingdrop plates by mixing equal volumes (1 μL) of 10 mg/ml protein and well solution (0.1 M MES pH 6.5; 1.6 M Mg 2 SO 4 ) and growing at 18˚C. The data were indexed and integrated using the HKL2000 software [77]. Phases were calculated by molecular replacement using the model of the X. citri PilZ-FimX EAL complex (PDB ID 4FOU; [24]). Structural refinement was performed as described above using REFMAC, Phenix and COOT. Details of the refinement data statistics are shown in S1 Table. Structural alignments and figures were produced using Chimera. The atomic model and experimentally determined structure factors have been deposited in the protein data bank with pdb code 7LKQ.

Size-exclusion chromatography (SEC)
A Superdex S200 10/300 (GE Healthcare) column was used to investigate protein-protein interactions between the PilB 12-163 or PilB  , PilZ and the different constructions of FimX. The column was previously equilibrated with 50 mM Tris-HCl pH 8,0, 50 mM NaCl, 5mM MgCl 2 . Protein samples, for binary complex (PilB 1-190 and wild type PilZ or their mutants) and ternary complex (PilB 12-163 -PilZ and PilB 1-190 -PilZ complex with FimX fragments), were mixed at an equimolar ratio (60-100 μM) in a final volume of 100 μL and applied to the SEC column. In SEC experiments studying the binary complex formed between FimX GGDEF-EAL and wild type or mutant PilZ, a 1:1.5 molar ratio was employed (100 μM FimX GGDEF-EAL and 150 μM PilZ). Where indicated, a 2-fold excess of c-di-GMP to FimX was added. In SEC experiments studying complexes containing full-length PilB, PilZ and FimX, a Superose S6 10/300 (GE Healthcare) column was used, equilibrated in the same buffer. In this case, protein samples were mixed in equimolar ratio (30-50 μM) in a final volume of 100 μL and applied to the SEC column. Where indicated, two-fold molar excess of ATPγS and/or c-di-GMP was added to the sample.

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex absence and presence of unlabeled PilZ at molar ratio of 1:1.5, at 298 K on a Bruker Avance III spectrometer operating at 800 MHz ( 1 H frequency) and equipped with a cryogenic TCI probe. NMR spectra were processed with NMRPipe [79] and analyzed using CCPNMR Analysis 2.4.1 [80].

ATPase assay
The ATPase activity of PilB-PilZ complex was analyzed by malachite green assay as previously described [81]. Briefly, 1 μM of PilB-PilZ complex was used for each reaction in the absence and presence of 1 μM FimX. The assay was carried out in reaction buffer containing 25 mM Tris-HCl pH 7,0, 50 mM NaCl, 2 mM MgCl 2 , 2 mM β-mercaptoethanol and 1 mM ATP (and 20 μM c-di-GMP where indicated) and incubated for 12 h at 30˚C. An aliquot (50 μL) of each reaction was transferred to a 96-well plate and the reaction was stopped by the addition of 200 μl of the phosphate assay reagent. Phosphate assay reagent was freshly prepared using 3 volumes of 0.045% malachite green hydrochloride (Sigma-Aldrich), 1 volume of ammonium molybdate (4.2% in 4 N HCl) and 1/100 volume of 1% Triton X-100. After 10 min of incubation at room temperature, the optical density was measured at 650 nm using a 96-well microplate reader (Spectramax-Molecular Devices). Reactions lacking PilB-PilZ complex or ATP were included as negative controls. The phosphate released during the reaction was measured using a standard curve of 1 to 150 μM KH 2 PO 4 .

Fluorescence titration
The formation of the PilB 1-190 -PilZ W69_5OHW and FimX EAL -PilZ W69_5OHW complexes were monitored by changes in the 5-hydroxytryptophan fluorescence emission spectra of PilZ-W69_5OHW upon addition of different amounts PilB  or FimX EAL . The assay was carried out in buffer containing 50 mM Tris-HCl pH 8,0, 50 mM NaCl, 1 mM MgCl 2 , 1 mM β-mercaptoethanol and the initial concentrations of PilZ W69_5OHW was 1 μM. The samples were equilibrated for 2 min before each measurement. Titration experiments were performed using a RF-6000 fluorescence spectrophotometer (SHIMADZU). The excitation wavelength was 310 nm (bandwidth: 5 nm), and the emission spectra were recorded between 325-445 nm (bandwidth: 5 nm). Dissociation constants were calculated assuming a simple 1:1 bonding model, as previously described [82] using the SigmaPlot 11 software (Systat Software Inc.).

Cloning of constructs for genomic deletions and insertions
Primers, plasmid and strains used for cloning and PCR verifications are listed in S4 and S5 Tables. Genomic deletion and insertion in the X. citri (strain 306) genome were introduced by two-step allelic exchange method [83] with small modifications as described [84]. Briefly, for gene deletions, two fragments of approximately 1000 bp corresponding to up-and downstream from the region of interest were amplified by PCR (Phusion polymerase, Thermo Scientific) using primers containing either homology region to the pNPTS138 vector backbone or the up or down-fragment and cloned into the pNPTS138 suicide vector by Gibson assembly (NEB). For the gene insertion of fluorescent reporter, the msfgfp or mcherry genes was amplified from pDHL1029 [85] and pDHL-mCherry [86], respectively, introducing a flexible N-or C-terminal liker (Ser-Gly-Gly-Gly-Gly). Separately,~1,000 bp fragments of the upstream and downstream regions from the star (N-terminal fusion) or stop (C-terminal fusion) codon were amplified by PCR (Phusion polymerase, Thermo Scientific) from X. citri genomic DNA, using primers containing either homology to the pNPTS138 vector backbone or the msfgfp or mcherry gene (S1 Table). The three fragments were cloned into the pNPTS138 vector using Gibson Assembly (NEB). The resulting plasmid was used to transform the appropriate X. citri

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex strain by electroporation (2.0 kV, 200 O, 25 μF, 0.2 cm cuvettes; Bio-Rad). A first recombination event was selected for on LB plates containing 50 μg/ml kanamycin. Transformants were streaked for single colonies on kanamycin plates after which several single colonies of the merodiploids (Kan R , Suc S ) were streaked on NaCl-free LB-agar supplemented with sucrose (10 gl −1 tryptone, 5 gl −1 yeast extract, 60 gl −1 sucrose and 15 gl −1 agar), selecting for a second recombination event creating either a wild-type (reversion) or mutant allele. After confirmation of the loss of the kanamycin resistance cassette together with sacB, a PCR was performed using primers that hybridize outside of the homology regions to identify the genomic deletions or insertion and confirmed by Sanger sequencing of the PCR product.

Twitching motility assay
Twitching motility was assayed by the stab-inoculation method in KB medium in a microscopy chamber covered with glass slides [4]. Briefly, X. citri strains were grown on LB agar (1.5% wt/vol) supplemented with the appropriate antibiotic at 28˚C for 48 hours. Using a sterile toothpick, X. citri cells were collected from an isolated colony and stabbed through KB-agar (1%wt/vol) supplemented with 2 mM CaCl 2 to the plastic surface of microscopy chamber (Nu155411; Lab-Tek, NUNC). Chambers were statically incubated in a humidified chamber at 28˚C for 48 hours. Twitching zone was visualized by an inverted fluorescence microscope (Nikon Eclipse Ti-E) with specific excitation and emission filters for GFP. The images obtained were analyzed with Fiji (ImageJ) software [87]. For time-lapse experiments, overnight cultures of X. citri strains were grown in 2xTY medium at 28˚C with shaking at 200 rpm. After a first overnight growth period, cells were transferred at a 100-fold dilution into 2xTY media for a second overnight growth to synchronize growth. Cells were then diluted 100-fold in fresh media and 1 μL of cell suspension was spotted on a thin slab of KB-agar (1% w/v) supplemented with 2 mM CaCl 2 , after 4-to 6-hour growth, phase contrast images were obtained with a Leica DMi-8 epifluorescent microscope. The KB-agar slabs were constructed as described [88].

Fluorescence microscopy
Overnight cultures of X. citri strains were grown as described above. Cells were then diluted 100-fold in KB fresh media (using 0,2% casamino acids as nitrogen source) and 1 μL of cell suspension was spotted on KB-agarose (1,5% w/v, using 0,2% casamino acids as nitrogen source) supplemented with 2 mM CaCl 2 and incubated at 30˚C for 6 hours before imaging. The KB-agar slabs were constructed as described [88]. Phase contrast and msfGFP and/or mCherry emission images were obtained with a Leica DMi-8 epifluorescent microscope. Fluorescence emission of msfGFP and mCherry were captured using 2000 and 3000 ms exposure times at maximum excitation light intensities. The microscope was equipped with a DFC365 FX camera (Leica), a HC PL APO 100x/1.4 Oil ph3 objective (Leica) and excitation-emission band-pass filter cubes for GFP (Ex.: 470/40, DC: 495, Em.: 525/50; Leica) and mCherry (Ex.: 540/80, DC: 585, Em.: 592-668; Leica) foci. To determine the polar localization of foci and quantify the polar fluorescence intensities, the images were analyzed using the MicrobeJ software package [89].

Bacteriophage plaque assay
Overnight cultures X. citri wild-type and mutant strains were grown in 2×TY medium, collected by centrifugation, and resuspended in fresh medium at an OD 600nm of 0.3. X. citri strains were mixed with warm liquefied KB agar (0.7% wt/vol) supplemented with 2 mM CaCl 2 to form the top agar layer poured into petri dishes carrying previously solidified KB

Western blot assay
Western blot assays were performed using total protein extract of X. citri wild type or ΔpilZ strains carrying the pUFR047 vector directing the expression of wld-type or mutant PilZ proteins (I10E, F49E, F49E/L51E, F49A/L51A and ΔM117). X. citri cells were grown in petri dishes on KB-agar 1% supplemented with 2 mM CaCl 2 and 10 μg mL -1 of gentamicin and incubated at 28˚C for 3 days. Bacteria were collected using a plastic spatula and resuspended in 15 mL of 1xPBS and the OD 600nm for all strains was measured and adjusted to 1.5. Bacterial cells were collected by centrifugation and the bacterial pellet washed three times with 10 mL of 1xPBS. After this, the bacterial pellet was resuspended in 400 μL of 1xPBS and 100 μL of denaturing sample buffer (5x) and incubated for 5 min at 95˚C. Samples (20 μl) were resolved in 10-well Tricine-SDS-PAGE gels, transferred to nitrocellulose membrane (Bio Rad), and blocked for 12 h using 5% skimmed milk in 1×PBS. Primary antibodies produced in rabbit against PilZ (AbPilZ; 1:5000) [23] and VirB8 (AbVirB8; 1:8000) [90] were used. Secondary goat anti-rabbit IgG-IRDye 800CW (Li-Cor 32211; 1:8000) were used for AbPilZ and visualizing using a ChemiDoc MP Imaging System (Bio Rad), and secondary goat anti-rabbit IgG-AP conjugate (Bio Rad 1706518; 1:8,000) was used for and AbVirB8 with BCIP (VWR 0885) and NBT (Sigma-Aldrich N6876) for detection.

Thermal denaturation of PilZ and mutants
The thermal denaturation of wild-type PilZ and its mutants (I10E, D46A/E47A and F49E/ L51E) was accompanied by differential scanning fluorescence (DSF) as previously described [91] using a QuantStudio 3 Real-Time PCR (Thermo Fisher Scientific) instrument. Briefly, 50 μL aliquots of purified protein dissolved in 50 mM Tris-HCl pH 8.0, 50 mM NaCl, 5 mM MgCl 2 , 125x concentrated SYPRO Orange dye (Thermo Fisher Scientific) were distributed into optical 96-well plates (Life Technologies). The final protein concentrations were: PilZ Wt : 240 μM; PilZ F49E/L51E : 240 μM; PilZ I10E : 240 μM; and PilZ D46A/E47A : 60 μM. The plates were sealed with real-time compatible adhesive film (Life Technologies). Fluorescence data were collected using the built-in ROX filter while the temperature was raised from 25˚C to 95˚C at a rate of 3˚C/minute. Data were subsequently processed using the Protein Thermal Shift software (Thermo Fisher Scientific), and the Tm values were determined based on the maxima of the first derivative of the fluorescence versus temperature plot. The reported Tm values represent the mean and standard error of three different experiments.

PLOS PATHOGENS
Type IV pilus PilB-PilZ-FimX regulatory complex showing the common interface residues in both complexes as sticks. Note that in this figure, the interaction interface (interface 1) between FimX EAL and PilZ is as described previously [24]. An alternative mode of interaction (interface 2) is proposed and tested as described in the main text and detailed in  line in A)). Note that the addition of cdi-GMP to the PilZ wt (wild type PilZ)-FimX GGDEF-EAL and PilZ D46A/E47A -FimX GGDEF-EAL mixture results in a shift in its elution profile. The elution profiles for FimX GGDEF-EAL alone is shown in blue in A. The elution profiles for PilB  and PilZ mutants on their own are shown in blue and red respectively in B. In these experiments, FimX GGDEF-EAL and PilB   The fluorescence intensity profiles of 100 individual X. citri cells expressing msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB or PilQ-msfGFP during growth on KB-agarose (1.5% w, using 0.2% casamino acids as nitrogen source) supplemented with 2 mM CaCl 2 . Cell lengths were normalized. B) Left: Fluorescence microscopy images of X. citri cells expressing mCherry-FimX or PilQ-msfGFP when grown in liquid culture (0h) or after 6 h growth (6h) on KB-agarose (1.5% w, using 0.2% casamino acids as nitrogen source) supplemented with 2 mM CaCl 2 . Right: Graphical representation of the fluorescence intensity profile over the length of X. citri cells expressing mCherry-FimX or PilQ-msfGFP. Note that mCherry-FimX foci are observed when grown on agarose (conditions leading to twitching) but not in liquid culture. On the other hand, PilQ-msfGFP foci are observed under both conditions with bipolar localization more common during growth in liquid media. (TIF) S11 Fig. Type IV pilus-dependent phenotypes are maintained in X. citri strains carrying fluorescent chimeras. A) Time-lapse of in X. citri msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB and PilQ-msfGFP strains exhibiting twitching motility. The time lapse interval (h) is indicated for each frame. Images were taken by using a phase-contrast microscope. Scale bar, 20 μm. Also see S1-S4 Movies. B) Phage FXacm4-11 infection assays for X. citri msfGFP-FimX, msfGFP-PilZ, msfGFP-PilB, PilQ-msfGFP, mCherry-FimX/msfGFP-PilZ, mCherry-FimX/ PilQ-msfGFP and mCherry-FimX/msfGFP-PilB strains. Dark plaques are indicative of phageinduced bacterial lysis in a confluent culture background. Western blot assays using polyclonal antibodies (Ab) against PilZ (above) and VirB8 (below). The first lane contains total extract from wild type X. citri strain containing the pUFR047-PilZ WT vector. The following lanes contain total extracts from X. citri ΔpilZ cells carrying the empty pUFR047 vector and the vector directing the expression of PilZ Wt , PilZ I10E , PilZ F49E , PilZ F49E/L51E, PilZ F49A/L51A , PilZ ΔM117 . The same amounts of total protein were loaded on the gel. Detection of the X. citri VirB8 protein (XAC2621) was used as a control. Experiments were repeated three times with similar results. (TIF) S1