The type IV pilus assembly ATPase PilB functions as a signaling protein to regulate exopolysaccharide production in Myxococcus xanthus

Myxococcus xanthus possesses a form of surface motility powered by the retraction of the type IV pilus (T4P). Additionally, exopolysaccharide (EPS), the major constituent of bacterial biofilms, is required for this T4P-mediated motility in M. xanthus as the putative trigger of T4P retraction. The results here demonstrate that the T4P assembly ATPase PilB functions as an intermediary in the EPS regulatory pathway composed of the T4P upstream of the Dif signaling proteins in M. xanthus. A suppressor screen isolated a pilB mutation that restored EPS production to a T4P− mutant. An additional PilB mutant variant, which is deficient in ATP hydrolysis and T4P assembly, supports EPS production without the T4P, indicating PilB can regulate EPS production independently of its function in T4P assembly. Further analysis confirms that PilB functions downstream of the T4P filament but upstream of the Dif proteins. In vitro studies suggest that the nucleotide-free form of PilB assumes the active signaling conformation in EPS regulation. Since M. xanthus PilB possesses conserved motifs with high affinity for c-di-GMP binding, the findings here suggest that c-di-GMP can regulate both motility and biofilm formation through a single effector in this surface-motile bacterium.

The level of EPS in M. xanthus is regulated by a signal transduction pathway consisting of the Dif chemosensory proteins as well as the T4P machinery (T4PM) 24,[29][30][31] . DifA, DifC and DifE resemble the methyl-accepting chemoreceptor proteins (MCPs), the scaffold CheW, and the histidine kinase CheA, respectively 27 . These three proteins form a transmembrane (TM) signaling complex that positively regulates EPS production through the kinase DifE 32,33 . DifD, a CheY-like substrate of DifE phosphorylation, functions as a phosphate sink to negatively regulate EPS production 29,32 . DifG, a homologue of the CheC phosphatase, is a negative regulator that can dephosphorylate DifD-phosphate 29,32 . The occurrence of T4P correlates closely with that of EPS 31 . T4P − mutants are EPSwhereas the hyperpiliated pilT mutant is EPS + . Since mutations in dif are epistatic to those in T4P or pil genes, T4P has been proposed to function as a sensory apparatus that perceives and transmits signals to the Dif proteins downstream. The communication between T4P and Dif is mediated in part by the negative regulator StkA, a DnaK-like protein 34,35 that acts downstream of T4P but upstream of Dif 36 . Many questions remain concerning the mechanism of EPS regulation by this pathway although the transcription of eps genes does not appear to be the target of this regulation 37 .
We demonstrate here that the M. xanthus T4P assembly ATPase PilB 21 functions in a regulatory capacity in signaling EPS production independently of T4P assembly. Genetic studies uncovered that mutations in pilB can suppress the EPS defects resulting from the deletion of the pilin gene pilA. A mutation known to eliminate the ATPase activity of PilB and its ability to support T4P assembly 21 was found to strongly suppress the EPS defect of the pilA deletion strain. This observation indicates that the role of PilB in EPS regulation can be independent of its role as the T4P assembly ATPase. Analysis in vitro suggests that it is the nucleotide-free or the apo form of PilB that actively signals EPS production. Our results support the conclusion that PilB functions in a signaling capacity with dual roles in the regulation of motility and biofilm formation in M. xanthus.

Results
Isolation of pil mutations suppressing the EPS − phenotype of a pilA deletion. To understand how the T4P filament functions to regulate EPS in M. xanthus, the pil genes encoding the T4P structural proteins were targeted in a genetic screen for suppressors of a pilA deletion (ΔpilA) 17,38 . For this suppressor screen, 11 genes in three clusters at the M. xanthus pil locus (Fig. S1) were mutagenized. These are the pilB, pilT and pilC (pilBTC) genes in one cluster, pilG, pilH and pilI (pilGHI) as well as pilM, pilN, pilO, pilP and pilQ (pilMNOPQ) in two other clusters. We first deleted pilBTC, pilGHI and pilMNOPQ as individual clusters in the WT and the ΔpilA backgrounds. These mutants were confirmed to be defective in both S motility and EPS production (Fig. S2) 31 . Next, three complementation plasmids containing pilBTC, pilGHI and pilMNOPQ were constructed. These plasmids, which can integrate into the M. xanthus chromosome at a phage attachment site 39 , were transformed into and demonstrated to complement its corresponding deletions in S motility and EPS production in the WT background (Fig. S2). Transforming these plasmids into their respective deletion mutants in the ΔpilA background resulted in strains without S motility and EPS (Fig. S2) as expected for a ΔpilA strain 31 .
The three plasmids were mutagenized in an Escherichia coli mutator strain 40 , and pools of mutagenized plasmids were isolated and transformed into their respective deletion strains in a ΔpilA background for the suppressor screen. Approximately 20,000 transformants for each plasmid were screened on plates with Congo red where EPS + colonies appear red but EPS − ones are unstained 29,35 . Transformants of the mutated pilBTC plasmid yielded five red colonies but no transformant of the other two plasmids appeared EPS + . The five putative ΔpilA suppressor strains were confirmed to be EPS + by an EPS binding assay using the fluorescent dye Calcofluor White. Genetic mapping by transformation using genomic DNA 39,41 determined that two out of the five isolates likely had suppressor mutations in the mutagenized pilBTC genes as they are linked to the kanamycin resistant (Kan R ) marker carried by the integrative plasmid. The other three, which must have occurred elsewhere 29,36,39,41,42 , were not pursued further in this study.
A Walker B box mutation in PilB suppresses the EPS − phenotype of ΔpilA. The mutations in the pilBTC gene cluster from the two EPS + strains were identified by cloning and DNA sequencing as described in Methods. The same G to A transition mutation in pilB was found in both suppressor strains. This G to A mutation, referred to as pilB* hereafter, occurred at the third position in codon 388 of pilB, resulting in a methionine (M) to isoleucine (I) substitution (M388I) (Fig. S3). PilB is an ATPase with the conserved Walker A (WA) and Walker B (WB) boxes 21,43 and the M388I substitution resides in its Walker B box (Fig. S3A). Since no other mutation in pilBTC was found in the two suppressor strains, they probably originated from the same mutated plasmids.
To verify that the pilB* mutation was solely responsible for suppression of ΔpilA in the suppressor strains, the pilB M388I mutation was reconstructed by targeted mutagenesis in the plasmid containing the WT pilBTC gene cluster. When this plasmid with the M388I mutation was transformed into the pilABTC quadruple deletion (ΔpilABTC) mutant, the resulting strain had its EPS production restored (Fig. 1A). In addition, the M388I mutation was constructed on a plasmid containing pilB only. When transformed into a ΔpilA ΔpilB double mutant, this plasmid restored EPS production to this double deletion strain (Fig. 1B). These results indicate that the pilB* single mutation alone is sufficient to suppress the EPS defect resulting from ΔpilA.
PilB functions upstream of the Dif pathway in EPS regulation. We examined the relationship between PilB and other known EPS regulators ( Fig. 2A) by genetic epistasis (Fig. 2B). stkA encodes a DnaK homologue that negatively regulates EPS production [34][35][36] . A ΔpilB ΔstkA double mutant was constructed and it was found to be EPS + . Likewise, a ΔdifD ΔdifG double mutation restored EPS production to a ΔpilB mutant as it did to ΔpilA 31 . In contrast, a difE − pilB* double mutant was found to be EPS − similar as the single difE − mutant. Since the suppression of ΔpilA by pilB* showed that PilB functions downstream of the T4P filament in EPS regulation, these observations support a model wherein PilB acts as an EPS regulator downstream of the T4PM but upstream of StkA and the Dif pathway 44 .  PilB* is gain-of-function in EPS regulation albeit WT in S motility. A strain with either the ΔpilB mutation or the WT pilB (pilB WT ) in a ΔpilA background is EPS − (Fig. 1), indicating that neither the pilB WT allele nor the null or loss-of-function (LOF) ΔpilB mutation can support EPS production in a ΔpilA background. The pilB* mutation is therefore likely gain-of-function (GOF) in EPS regulation instead of WT or LOF. If so, it should be dominant over pilB WT with regard to the EPS phenotype. To test this, a pilB WT /pilB* merodiploid was constructed in ΔpilA background by transforming the pilB*-containing plasmid into a ΔpilA mutant. As shown in Fig. 1B, the resulting strain with these two pilB alleles was EPS + , indicating pilB* is dominant over pilB WT and confirming that pilB* is indeed a GOF mutation with regard to EPS regulation.
Since PilB is the T4P assembly motor ATPase, the effect of the pilB* mutation on M. xanthus S motility in an otherwise WT background was examined as well. The pilB* and pilB WT alleles were introduced into a ΔpilB single mutant. The resulting strains were examined for S motility (Fig. 3A) and EPS production (Fig. 3B). As indicated by the clear binding of Calcofluor White in EPS assays, both the pilB WT and pilB* restored EPS production to the pilB mutant as expected. Interestingly, pilB* also complemented ΔpilB in S motility as analyzed on a soft agar plate. These results demonstrate that while pilB* is a GOF mutation in EPS regulation, it is WT with regard to T4P assembly. PilB* therefore likely retains sufficient ATPase activity to support T4P assembly and S motility in vivo despite the significant alteration in its function in EPS regulation. These observations suggest that the ability of PilB to signal EPS production and its role as the T4P assembly ATPase may be genetically separable and functionally distinct.
An ATPase-deficient PilB variant supports EPS production but not S motility. The WA and WB boxes of PilB, while separated by 56 amino acids in the primary sequence (Fig. S3A), are packed against each other in the tertiary structure of the protein 43,45 . In the model of M. xanthus PilB (MxPilB) (Fig. 4), based on the crystal structure of Thermus thermophilus PilB/PilF (TtPilB) 43 , the WA box forms the C-terminal part of an α helix and a loop that connects the helix to a β strand. This strand is part of an antiparallel β sheet that lines one side of this α helix. M388 in the WB box is positioned in a strand in the center of this β sheet. The side chain of this methionine packs tightly against the α helix encompassing WA. Substituting this methionine with a bulkier isoleucine in MxPilB or TtPilB is expected to create clashes that would need to be resolved by altering the relative position between WA and WB. Despite its location in the WB box, the M388I substitution therefore may disturb the structure, orientation or conformation of both WA and WB.
It is known that residues in both WA and WB are crucial for the activity of many ATPases because they interact with the bound ATP molecule 43,[46][47][48] . Although the M388I mutation apparently does not eliminate the ATPase activity of MxPilB, it may still affect its ATP binding and/or ATPase activity. To test whether the ATPase activity of PilB has an effect on EPS regulation, we constructed mutations known to have more drastic effect on the ATPase activity of PilB. The strictly conserved residues K327 in WA and E391 in WB (Fig. 4) are known to be required for the ATPase activity of MxPilB in vitro and its ability to support T4P assembly and S motility in vivo 21 . We constructed K327A (pilB WA ) and E391A (pilB WB ) substitution mutations, respectively, and confirmed that neither mutation supported M. xanthus S motility. As shown in Fig. 5A, pilB WA but not pilB WB restored EPS production to a ΔpilA mutant. pilB WA is in fact a more robust suppressor than pilB* because its presence in the ΔpilAB background results in a higher EPS level when compared to pilB* (Fig. S4). In addition, a pilB WA /pilB WT merodiploid strain in a ΔpilA background is EPS + (Fig. 5B), and the dominance of pilB WA indicates that it is a GOF mutation in EPS regulation as pilB*. Because PilB WA fails to support S motility in vivo and showed no ATPase activity in vitro 21 , the results here ( Fig. 5) demonstrate that the role of PilB in EPS production is distinct from its function as the T4P assembly ATPase. That is, in the absence of pilA, the catalytically inactive PilB WA not only supports EPS production, but it is even more potent in doing so than its enzymatically active counterparts PilB WT and PilB* (Figs 1, 5 and S4). The stronger EPS phenotype of the pilB WA mutant further strengthens the conclusion that the activation of EPS production by PilB does not require or can bypass its function in T4P assembly. These The α helix containing WA is colored magenta and the β sheet containing WB is colored orange. The side chain of M388 is shown in yellow and those from the helix residues in close contact with M388 are shown as aqua surfaces. Also shown are the side chains of K327 (green) in WA and E391 (blue) in WB. Because ATP-γ-S and a magnesium ion were present in the TtPilB template structure, we were able to model the ligand (in sticks) and the catalytic magnesium ion (red sphere) as well. observations support PilB as a signaling or regulatory protein with a more direct role in EPS regulation and that the PilB WA variant may assume a more active signaling conformation without the pilus filament.
The stability of the PilB* and PilB WA is diminished in vivo. We explored the possibility that the EPS + phenotype of pilB* and pilB WA mutations may have been due to altered the protein level of PilB by performing immunoblotting using anti-PilB antibodies 21 . The levels of PilB in the WT and the ΔpilA backgrounds were virtually indistinguishable (Fig. 6A), suggesting that the ΔpilA mutation itself does not alter PilB expression or stability. However, the levels of both PilB* and PilB WA were lower than PilB WT in an isogenic ΔpilA background (Fig. 6B). Since the experiments were conducted with strains where all pilB variants were expressed from the same promoter 49 , it is unlikely that transcription or translation are affected by these mutations. Instead, it is more likely that the stability of the protein in vivo was changed by these mutations. Regardless, since a pilB null mutation does not suppress the EPS − phenotype of ΔpilA, only an increase in PilB protein level could explain the suppression of ΔpilA by pilB* and pilB WA , but not a decrease. We propose that PilB* and PilB WA mutations changed the structure of the protein to favor its active signaling conformation and that this conformational change may have reduced the stability of PilB in vivo coincidentally.
PilB WA is altered in its structure and response to ATP. To gain insights into the mechanisms of PilB in EPS regulation, we attempted to purify and examine the properties of PilB WT , PilB WA and PilB WB proteins in vitro. These three PilB variants were chosen because they reflect the three distinct phenotypes with regard to T4P assembly and EPS production in vivo. PilB* was not included in the in vitro studies because it resulted in an intermediate EPS phenotype. Purified MxPilB exhibits measurable but very low ATPase activity 21 and it tends to form protein aggregates (unpublished). We henceforth expressed and purified the T. thermophilus PilB WT , PilB WA and PilB WB equivalents (Fig. S3A) 50 for in vitro studies.  Differential scanning fluorimetry (DSF), which measures the thermal denauration of a protein over a temperature gradient 51,52 , was used to first examine the structural differences among PilB WT , PilB WA and PilB WB . As shown in Fig. 7A, PilB WT and PilB WB behaved relatively similar in this assay: both show biphasic unfolding profiles with two melting temperatures or transition midpoints (T m ) around 64-65 °C and 79-81 °C, respectively. In contrast, PilB WA unfolded over a broader temperature range with an apparent T m around 76 °C. The obvious differences in thermal stability indicate that the structure of PilB WA , but not that of PilB WB , is significantly altered in comparison with PilB WT . We also examined the effect of AMP-PNP, a non-hydrolyzable ATP analogue 53, 54 , on PilB thermal stability (Fig. 7A). This nucleotide stabilized both PilB WT and PilB WB , resulting in the disappearance of the first unfolding transition. In contrast, the stability of PilB WA remained virtually unchanged by the addition of AMP-PNP. These results demonstrate that PilB WA no longer responds to its ligands and that its structure is distinct from those of PilB WT and PilB WB .
The circular dichroism (CD) spectra of these protein variants were collect to further explore their structural differences. In the far ultraviolet (UV) range (200-240 nm), all three proteins showed similar spectra with or without ATP or ADP (Fig. S7). As the signals in this range are sensitive to changes in protein secondary structure 55 , the results indicate that the secondary structure content is unaffected by the mutations in PilB WA and PilB WB or by ligand binding. In the near UV range (260-320 nm), changes in the CD spectra are related to protein conformation 55 . The signals in the 260-285 nm range are contributed primarily by phenylalanine and tyrosine residues 55 , which the TtPilB contains 14 each (Fig. S3A). There is little signal beyond 285 nm because the protein lacks tryptophan (Figs 7B and S3A). All three protein variants, PilB WT , PilB WA and PilB WB , showed similar CD spectrum without ligand. While the addition of ATP shifted the spectra of all three proteins downward in the 260-280 nm range (Fig. 7B, left panel), the shift for PilB WA is less pronounced compared to those seen for the other two. The addition of ADP also shifted the spectra of PilB WT and PilB WB , but not that of PilB WA (Fig. 7B, right panel). These observations further underscored that PilB WA is diminished in its response to nucleotides when compared to PilB WT and PilB WB . In particular, PilB WA more closely resembles the ligand-free rather than the ligand-bound form of PilB in the presence of a nucleotide.
The binding affinity of PilB for ATP was examined using the ATP analogue MANT-ATP 56 . The fluorescence of this nucleotide is enhanced by a more hydrophobic environment frequently associated with its binding to a protein. Figure 8 shows the difference in fluorescence (ΔF) in the presence and absence of PilB variants over varying concentrations of MANT-ATP. A binding isotherm was fit to the data to estimate the binding affinity of the three PilB protein variants to this ATP analogue. The dissociation constants (K d ) for PilB WT and PilB WB were similar with values of 0.17 μM and 0.19 μM, respectively. The K d for PilB WA is 1.23 μM and this increase in K d represents a significant reduction in its nucleotide binding affinity. These results suggest that the diminished response of PilB WA to nucleotides is attributed to its reduced affinity for its ligand. The in vitro studies (Figs 7 and 8) suggest that the nucleotide-free conformation of PilB may actively signal EPS production in M. xanthus.

Discussion
This study identified the T4P assembly ATPase PilB as a regulator of EPS downstream of the T4P filament and upstream of the Dif signaling proteins in M. xanthus. A genetic screening was devised to first isolate pil mutations that could restore EPS production to a ΔpilA strain, leading to the discovery of pilB* (Fig. 1). Further genetic analysis indicated that PilB functions downstream of T4PM but upstream of the StkA and Dif proteins in EPS regulation (Fig. 2) 44 . Targeted mutagenesis of conserved residues revealed that the involvement of PilB in EPS regulation can be separated from and is independent of its function as the T4P assembly ATPase (Fig. 5). That is, the PilB WA mutant variant, which does not hydrolyze ATP or support S motility 21 , may be locked in a conformation that actively signals EPS production in a ΔpilA or a WT background (Fig. 5). Both pilB* and pilB WA are dominant over pilB WT (Figs 1 and 5), indicating that they are GOF mutations. In vitro studies using TtPilB indicated that the PilB WA variant showed diminished conformational response to the addition of nucleotides because of reduced binding affinity (Figs 7 and 8). These results lead to a model wherein the M. xanthus PilB ATPase functions as a signaling protein in EPS regulation and its nucleotide-free state may correlate with its actively signaling conformation.
As far as we are aware, XB24 57 and the Na/K ATPase (NKA) 58,59 are the two other cases where ATPases have been proposed as signaling proteins. XB24 is a small cytoplasmic ATPase in rice with a role in immunity and defense against bacterial pathogens. It physically associates with and modulates the activity of the kinase XA21, which is a receptor for recognition of pathogen-associated molecular patterns (PAMPs). The ATPase activity of XB24 is required for its function in vivo and it is proposed that the signaling activity of XA21 is regulated by XB24. NKA is ubiquitous in the plasma membrane of all animal cells. It establishes and maintains the ion homeostasis essential for cell viability. All NKAs contain a binding site specific for steroid inhibitors such as ouabain and digoxin. There is evidence that NKAs can associate with and affect the activity of signaling proteins such as the Src kinase 60,61 . It has been proposed that NKAs are receptors and signal transducers to regulate downstream targets in a signal transduction pathway.
Drawing analogy between G-proteins 62, 63 and signaling ATPases 57, 58 , we hypothesized initially that it was the PilB in the ATP bound form that actively signaled EPS production. This was tested by mutating E391 in WB, a key catalytic residue for ATP hydrolysis 47,48 (Fig. 4). Since PilB WB has the same binding affinity for ATP as the PilB WT (Fig. 8) but is inactive as an ATPase 21 , it likely exists in its ATP-bound form in the cell. The pilB WB mutation, however, failed to suppress ΔpilA (Fig. 5), arguing against the ATP-bound PilB as the actively signaling conformation. On the other hand, the highly conserved K327 in WA is known to be critical for ATP binding 47,48 (Fig. 4). When mutated to an alanine, the resulting pilB WA turned out to be a robust suppressor of ΔpilA in EPS regulation (Figs 5 and S4). Binding assays indicated that PilB WA binds to ATP with much reduced affinity as expected (Fig. 8).
Biophysical studies also suggested that the structure of PilB WA resembles the apo-form of PilB WT with or without its nucleotide ligands (Fig. 7B). We therefore propose that it is the apo form of PilB that actively signals EPS production in M. xanthus. It remains to be seen how prevalent signaling ATPases are in different biological systems and whether they share a common signaling conformation with M. xanthus PilB.
The regulation of EPS is the key aspect of bacterial biofilm formation [64][65][66] . The paradigm that emerged from the studies of biofilm in flagellated motile bacteria is the mutual exclusivity or inverse regulation of the motile state vs the biofilm state [64][65][66] . That is, the regulation is such that the motility of cells in a biofilm is inhibited whereas motile cells either exit from or exist mostly outside of biofilms in bacteria with flagellated motility. The small signaling molecule c-di-GMP is well known as the "master" regulator of the transition between the motile and the biofilm states. The overarching conclusion from studies of this signaling molecule is that it simultaneously enhances biofilm formation and inhibits flagellated motility. The known effectors or targets of c-di-GMP in biofilm regulation are diverse, ranging from biosynthetic enzymes and riboswitches to transcriptional activators and posttranslational regulators 65 . Nevertheless, EPS, the major constituent of bacterial biofilms, is the ultimate target of this regulation in most cases. For the regulation of flagellated motility, c-di-GMP generally targets the expression or activity of flagellar proteins. In V. cholerae, for example, such regulations are achieved by the c-di-GMP receptors and transcriptional regulators VpsR and FlrA/FliQ among others 67 . VpsR, which directly activates the expression of EPS or VPS (Vibrio polysaccharides) genes, binds c-di-GMP with high affinity. FlrA/FliQ is the master regulator of flagellar genes and the binding of c-di-GMP impairs its ability to activate the transcription of the flagellar operon.
More recently, a new c-di-GMP binding motif was discovered by the studies of V. cholerae MshE, the PilB equivalent in the Msh pilus system [68][69][70] . Sequence alignment with MshE shows convincingly that PilB from M. xanthus and many other T4P systems contain this c-di-GMP binding motif at their N-termini (Fig. S3B) 68 . This motif by itself and a few proteins with it have been verified or demonstrated to bind c-di-GMP with high affinity 68 . These include a PilB from Clostridium perfringens 71 . It is therefore reasonable to assume that M. xanthus PilB is a functional c-di-GMP effector. With this in mind, the finding of M. xanthus PilB as an EPS regulator here suggests that c-di-GMP manages biofilm formation and T4P-mediated S motility through PilB as a direct target in M. xanthus. Drawing analogies with flagellated bacteria, we envision a similar working model for the regulation of EPS production and S motility by c-di-GMP in M. xanthus. We propose that c-di-GMP promotes EPS production and biofilm formation at high concentrations whereas the T4P-dependent S motility is favored at low or basal levels of c-di-GMP. In this model, when c-di-GMP is present at low or basal levels, PilB is active as the T4P assembly ATPase while EPS is produced only at a basal level to allow M. xanthus S motility to function 26,72,73 . When the cellular concentration of c-di-GMP is high, it binds to PilB to signal EPS production and to inhibit its activity as the T4P assembly motor simultaneously. We suggest that the binding c-di-GMP results in conformational changes in PilB mimicked or represented by the PilB WA mutant protein, and such conformational changes lead to the inhibition of ATPase activity and stimulation of EPS signaling.

Methods
Growth Conditions. Myxococcus xanthus strains used in this study are listed in Table 1. They were grown and maintained at 32 °C on Casitone-yeast extract (CYE) agar plates or in CYE liquid medium 74 . XL1-Blue (Stratagene) and Rosetta (Novagen), the Escherichia coli strains used for plasmid construction and protein expression, were grown and maintained at 37 °C on Luria-Bertani (LB) agar plates or in LB liquid medium 75 . Unless noted otherwise, plates contained 1.5% agar. Kanamycin and ampicillin at 100 μg/ml and oxytetracycline at 15 μg/ ml were added to media for selection when appropriate.
Plasmid used for strain construction. Three sets of plasmids were generated to construct M. xanthus strains. The first set were constructs to delete single or multiple pil genes; these include pWB525 (ΔpilA), pWB581 (ΔpilB), pWB555 (ΔpilBTC), pWB556 (ΔpilAGHI), pWB605 (ΔpilGHI) and pWB557 (ΔpilMNOPQ). The second were for the expression of pil gene clusters and pilB alleles in M. xanthus; these constructs, which integrate at Scientific REPORTS | 7: 7263 | DOI:10.1038/s41598-017-07594-x Mx8 att site 76 , include pWB559 (pilGHI), pWB565 (pilMNOPQ) and pWB566 (pilBTC) as well as pWB571 (pilB), pWB572 (pilB*), pGD5 (pilBWA) and pGD6 (pilBWB). The third includes the plasmid pWB606, which was used for the replacement of wild-type pilB with pilB*. The first and third sets were derivatives of pBJ113 77 . The second set used pWB425 as the cloning and M. xanthus expression vector 39 . pWB425 contains a BspHI/ApoI fragment from pZero-2 (Invitrogen) with the kanamycin resistance (Kan R ) gene and its promoter. The expression of all pil genes cloned into pWB425 is driven by this promoter because the multicloning site is immediately downstream of the Kan R gene in this plasmid.
Plasmid construction and description. The details for the construction of plasmids used for M. xanthus strain construction are described here. Fragments with in-frame deletion alleles of pilA, pilBTC, pilGHI, pilAGHI and pilMNOPQ were generated by a two-step overlap PCR as described previously 29 . These fragments were cloned into pBJ113 77 to construct pWB525 (ΔpilA), pWB555 (ΔpilBTC), pWB556 (ΔpilAGHI), pWB605 (ΔpilGHI) and pWB557 (ΔpilMNOPQ). In these plasmids, the ΔpilA allele deleted the codons from 7 to 218 of pilA, ΔpilBTC from 8 of pilB to 409 of pilC, ΔpilAGHI from 7 of pilA to 250 of pilI, ΔpilGHI from 5 of pilG to 250 of pilI and ΔpilMNOPQ from 7 of pilM to 896 of pilQ, respectively. The pilB deletion allele from DK10416 18 was PCR amplified and cloned into pBJ113 to produce pWB581. For the allelic exchange of wild-type pilB with pilB*, the fragment from pWB572 (see below) was cloned into pBJ113 to create pWB606.
Plasmids that can integrate into M. xanthus chromosome at Mx8 phage attachment site (att) 78 for ectopic expression were constructed using pWB425 39 as the vector. pWB559, pWB565 and pWB566, which contain the respective pilGHI, pilMNOPQ and pilBTC gene clusters from pDW79 79 or pSWU257 17 in pWB425, were also used for mutagenesis in E. coli. With reference to the coding regions (Fig. S1), pWB559 (pilGHI) contains DNA from 73 bp upstream of pilG to 328 bp downstream of pilI, pWB565 (pilMNOPQ) from 120 bp upstream of pilM to 19 bp downstream of pilQ, and pWB566 (pilBTC) from 22 bp upstream of pilB to 5 bp downstream of pilC. In addition, pWB571 (pilB), which contains from 22 bp upstream to 142 bp downstream of pilB, was derived from  Fig. S3).
Strain Construction. YZ690 through YZ1888 in Table 1 are the M. xanthus strains constructed in this study.
Mutagenesis and suppressor identification. pWB559, pWB565 and pWB566 were mutagenized by propagation in NR9458, an E. coli mutD5 mutator strain 40 . Cells were initially grown on plates with 1 × Volgel-Bonner salts minimal media containing 0.4% glucose, 50 μg/ml proline and 5 μg/ml thiamine to minimize the mutation rate 40 . Plates containing approximately 100 colonies for each plasmid to be mutagenized were pooled and inoculated into LB broth to increase the mutation rate. Mutagenized pools of plasmid DNA were prepared from overnight cultures and used to transform the appropriate pil deletion strains. Potential EPS producing suppressor mutants were identified as reddish-orange colonies on CYE plates containing Congo red (30 μg/ ml) 35 . Genomic DNA from potential suppressor mutants was electroporated into the original parental strain and selected on plates with kanamycin and Congo red to examine the link of the EPS phenotype with the integrated plasmid 39,41 . Genomic DNA from suppressor mutants was cut with PstI, religation and transformation into E. coli XL1-Blue to recover the plasmid and pil mutations were identified by DNA sequencing.
Phenotypic analysis. Log phase cells grown in CYE were harvested and resuspended in MOPS buffer (10 mM morpholinopropanesulfonic acid [pH 7.6], 2 mM MgSO 4 ) at 5 × 10 9 cells/ml. 5 µl of this cell suspension was spotted onto CYE plates with 0.4% agar and regular CYE plates with Calcofluor white (50 µg/ml) for S-motility and EPS analysis, respectively. Plates were incubated at 32 °C for 5 days before documentation under white light for S motility and 365 nm UV light for EPS production.
Protein purification. Plasmids pWB750, pWB751 and pWB752 (see SI Materials and Methods) were constructed and maintained in the E. coli strain XL1-Blue initially. For protein purification, they were transformed into the E. coli Rosetta strain containing pREP4 (Qiagen). The appropriate expression strains were grown at 35 °C to an OD 600 of 0.5-0.6 in 1 liter of LB plus ampicillin (100 μg/ml) and kanamycin (25 μg/ml). Protein expression was induced by addition of IPTG (Isopropyl β-D-1-thiogalactopyranoside) to a final concentration of 0.1 mM, followed by incubation at 30 °C for 4-5 hrs. Please see SI Materials and Methods for more details for the purification.

Differential Scanning Fluorimetry (DSF).
Proteins were diluted to a final concentration of 8 μM in 1× stock buffer (10 mM HEPES [pH 8], 50 mM KCl, 10 mM MgCl 2 , 0.5 mM EDTA, 1 mM β-ME and 10% glycerol) using a 5× stock buffer. Sypro Orange (Ex. 490 nm, Em. 530 nm) from a 5,000× stock (Invitrogen) was diluted to 50× in deionized water and used at a 5× concentration in DSF assays. For examining the effects of ligand, adenosine 5′-(β,γ-imido)triphosphate (AMP-PNP) (Sigma-Aldrich) was added to a final concentration of 0.4 mM. The experiment was carried out using a Bio-Rad CSX96 Real-Time System (Bio-Rad), starting at 25 °C with temperature increments of 0.5 °C to a final temperature of 99 °C. Samples were held at each temperature for 30 sec prior to measurement of fluorescence. Data were analyzed and Tm's were obtained using Bio-Rad's CFX Manager software. The data points starting at 45 °C were presented in this paper. Circular Dichroism (CD). Protein samples were adjusted to a final concentration of about 6 µM for far UV (200-260 nm) by diluting in storage buffer. For near UV (260-320 nm), proteins were concentrating using Amicon stirred cells to 230 µM and 170 µM for the experiment with ATP and ADP, respectively. CD spectra were generated on a Jasco J-815 Spectropolarimeter equipped with the Jasco PFD-425S temperature-control unit in a 1-mm path-length quartz cell at 25 °C. Each spectrum was from three accumulated scans with a 1 second response time at a scan speed of 100 nm per minute. Scans were performed with a bandwidth of 1 nm in far UV and at 0.5 nm increments in near UV. Data was recorded using Spectra Manager software (Jasco). For examining the effects of ligand, ATP and ADP (Sigma-Aldrich) was added at a final concentration of 0.1 mM. Data was plotted after background subtraction of identical scans of controls without the protein. ATP and ADP were used in this experiment because these T. thermophilus PilB variants has no ATPase activity at 25 °C (data not shown).