Introduction

The type IX secretion system (T9SS) is a nanomachine responsible for the delivery of effectors in the medium or at the cell surface, exclusively present in the Bacteroidetes phylum1. The T9SS is comprised of an outer membrane associated ring made of the PorK and PorN proteins, associated to 18 PorL–PorM inner membrane motors that extend through the periplasm to form a birdcage-like structure2,3,4. The PorL–PorM motor is proposed to provide the energy necessary for effector secretion1,5,6.The PorKN ring is also connected to 8 Sov/SprA translocons that are involved in the transport of effectors through the OM4,7,8. The T9SS secretes a broad variety of effectors including virulence factors, adhesins involved in gliding motility, S-layer proteins, and carbohydrate hydrolyzing enzymes9. Studies on the T9SS have been mainly performed in Porphyromonas gingivalis and Flavobacterium johnsoniae. P. gingivalis, an anaerobic, non-motile Gram-negative bacterium, is the major human oral pathogen associated to periodontal diseases10,11. Chronic P. gingivalis infection is also linked to rheumatoid arthritis, heart disease, diabetes, Alzheimer and other systemic diseases12,13,14,15,16,17. The main virulence factors involved in P. gingivalis pathogenicity are secreted through the T9SS. Among them, gingipains are cysteine proteases that are covalently linked to the cell surface or delivered to the extracellular milieu, and that are responsible for tissue colonization and destruction, as well as for host defense perturbation18. F. johnsoniae is an aerobic, non-pathogen bacterium living in soil and fresh water that can move along surfaces at speeds of up to 5 μm s−1 in a process known as gliding motility19. In F. johnsoniae, the T9SS is mainly responsible for the secretion of the SprB and RemA adhesins required for gliding motility20. In addition to its role in effector secretion, the T9SS GldL–GldM inner membrane complex functions as a rotary motor powering F. johnsoniae adhesin motion and hence gliding motility5,6,21,22.

While structural and functional data on the T9SS from both bacterial models accumulate, only information on P. gingivalis T9SS regulation are available23,24,25,26. A comparative genomic study first showed that the porX and porY genes co-occur with genes encoding the type IX secretion apparatus23. It was further shown that PorY and PorX act as cognate sensor histidine kinase and response regulator, respectively, of a two-component system (TCS)23,24. TCS is one of the most common signal transduction mechanisms in bacteria to sense and respond to environmental cues27,28,29,30. TCS is usually constituted of a histidine kinase that autophosphorylates upon a specific stimulus, and consequently transfers the phosphoryl group to the receiver domain (RD) of the response regulator. Once activated, the response regulator elicits the cellular response through its effector domain. This effector domain usually functions as a DNA-binding transcription factor, but can also display RNA-binding, protein-binding or even enzymatic activities31.

In contrast to the majority of TCSs, in which the components are encoded within the same operon, the porX and porY genes distribute at separate loci within the P. gingivalis chromosome. Nevertheless, disruption of the PorXY TCS results in the dysfunction of the T9SS, which manifests as the impaired processing of gingipains, as well as in the down-regulation of essential T9SS component genes such as porT, sov, porKLMN, and porP23. It was shown that PorX and PorY interact with each other24,25, and that PorX interacts with the cytoplasmic domain of the T9SS rotary core component PorL24. PorX–PorL interaction requires a hydrophobic patch at the PorL C terminus24, a situation that is reminiscent of the association of CheY response regulator to the flagellar C-ring FliN protein32. It has been therefore proposed that PorX translates the regulatory signal into a mechanical output24. However, the mode of action of PorX is controversial. Studies showed that PorX does not directly bind to the promoter regions of the T9SS genes24,25 but instead interacts with SigP, a putative extracytoplasmic function (ECF) sigma factor that itself binds to the promoter regions of the porT, porV and porP genes25. In contrast, a recent study detected a direct interaction of PorX on two DNA sequences in the porT gene26. Sequence analysis of PorX showed that it is constituted of an N-terminal RD of the CheY family and a C-terminal alkaline phosphatase-like domain of the PglZ family separated by a linker region24. The C-terminal domain is not involved in T9SS gene regulation, and the linker region was proposed to interact with the PorX DNA target26. Very recently, a study provided further insights onto the structure–function relationship of PorX. It was notably shown that the PorX PglZ domain possesses a phosphodiesterase activity, and that dimerization of PorX, promoted by the presence of zinc, is required for its activity33. Substrates for the PglZ domain were determined, and an interdependence between the RD and PglZ domains through the dimerization surface was established. Furthermore, the crystal structures of the dimeric, activated PorX, as well as its complex with pGpG, were solved33.

Until now, there is no information on T9SS regulation by a TCS in F. johnsoniae. However, PorX homologs are present in all species encoding T9SS components34, arguing that PorX is indispensable for the regulation of T9SS activity. In this study we produced and purified the F. johnsoniae PorX homolog (Fjoh_2906, called hereafter PorXFj) and solved its structure by X-ray crystallography at 2.0 Å resolution. Comparison with the P. gingivalis PorX (PorXPg) structure suggests that both proteins share similar function. Indeed, we confirm that PorXFj has phosphodiesterase activity and further show that PorXFj interacts with the cytoplasmic domain of the PorL homologue GldL, and that this interaction is mediated by the phosphorylated CheY-like domain of PorXFj.

Results

Overall structure of PorXFj

In order to solve the structure of PorXFj, the native protein as well as its Selenomethionine-labeled (SeMet) derivative were produced and purified by immobilized ion metal affinity and size-exclusion chromatographies. Diffracting crystals of both proteins grew in the P212121 space group; the SeMet PorXFj structure was solved using a SAD data set and then used as a model to solve a higher resolution structure of the native PorXFj by molecular replacement.

Similarly to the recently released P. gingivalis PorXPg structure33, PorXFj displays three distinct domains: a N-terminal CheY-like RD (residues 2–121), followed by a three-helix bundle domain (HBD, residues 122–205), and a C-terminal PglZ domain (residues 206–517) (Fig. 1A). Two PorXFj molecules are present in the asymmetric unit. While the three domains of each molecule are perfectly superimposable independently (with an rmsd of 0.24 Å, 0.54 Å and 0.22 Å for the N-terminal domain, HBD and C-terminal domain, respectively), superposition of the two whole molecules yields an rmsd of 1.47 Å. Indeed, when the two molecules are superimposed through their N- or C-terminal domains, a slight shift of the HBD's first helix is observed, which carries over to the rest of the molecule and could reflect some flexibility on either side of the kink present in this helix (Fig. S1).

Figure 1
figure 1

Overall structure of PorXFj. (A) Structure of the PorXFj monomer ; the CheY-like N-terminal, the HBD and the PglZ C-terminal domains are shown in cyan, white and purple, respectively; the N- and C-termini are labelled. (B) Comparison of the PorXFj (left) and PorXPg (right) dimer structures; For each structure, molecules A and B are shown in light and dark grey, respectively.

Full size image

The two PorXFj molecules form an intertwined dimer identical to the PorXPg dimer33, with a buried surface area of 2985 Å2 (representing 12.5% of the total surface area) and a binding energy of − 42.7 kcal mol−1, according to QtPISA35 (Fig. 1B). The dimer is stabilized by a network of electrostatic interactions (21 hydrogen bonds and 6 salt bridges), as well as hydrophobic patches (involving 21 hydrophobic residues from each monomer), principally between the N- and C-terminal domain of each monomer (Fig. S2; Table S1).

Similarly to PorXPg, PorXFj is monomeric in solution and dimerization is induced by phosphorylation with the low-molecular weight phospho-donor acetyl phosphate (AcP) in presence of Mg2+, but not by AcP nor Mg2+ alone, as shown by SEC-MALS results (Fig. 2). No crystal of monomeric PorXPg could be obtained, and its dimeric structure was solved after incubation with BeF3, a compound used to mimic phosphorylation33. As no BeF3 was added to PorXFj, we hypothetize that formation of the dimer was promoted by the crystallization process. Noteworthy, 41 residues have their side chains involved in the stabilization of the PorXFj dimer through electrostatic interactions and hydrophobic contacts, compared to the 29 residues in the PorXPg dimer (Fig. S3). This difference suggests that the PorXFj dimer is more stable and could be more prone to assemble during crystallization.

Figure 2
figure 2

PorXFj dimerization is induced by phosphorylation and Zn2+. The purified PorXFj alone or incubated with AcP and MgCl2 (AcP/Mg2+), AcP (AcP), MgCl2 (Mg2+) or ZnCl2 (Zn2+) was analyzed by size-exclusion chromatography-multiangle light scattering (SEC-MALS). Note that the chromatograms of PorXFj alone and incubated with MgCl2 overlap and are hence almost indistinguishable.

Full size image

The CheY-like domain

The CheY-like RD adopts the classical (α/β)5 doubly-wound fold consisting of a central 5-stranded β-sheet, surrounded by five α-helices. The conserved active site is an acidic pocket formed by three acidic residues (D11, E12 and D54), T82, and K104. PorXFj was not activated by phosphorylation before crystallization, and indeed no phosphoryl group attached to the conserved D54 residue is present in the electron density map. The putative binding site of the metal supposed to stabilize the phosphoryl group is clearly occupied by an ion that is octahedrally coordinated by the carboxylate oxygens of D11 and D54, the main chain-carbonyl oxygen of N56 and three water molecules (Fig. 3). The nature of the ion cannot be attributed unambiguously from inspection of the electron density. By comparison with the PorXPg structure, and accordingly to the coordination number and the bond lengths between the ion and its ligands, a Mg2+ ion was modelled in the structure. As no magnesium was added during the purification and crystallization steps, we can therefore speculate that the ion present in the structure was acquired intracellularly. Despite the absence of BeF3 that mimics the activating phosphorylation of the RD, the side chains of the two PorXFj highly conserved ‘switch’ residues T82 and Y101 are oriented towards the active site, in a conformation corresponding to the active state of RDs (Fig. 3). This confirms that binding of the Mg2+ ion is likely sufficient to induce the conformational changes associated to RDs activation, while phosphorylation stabilizes this Mg2+-bound conformation, as previously proposed36.

Figure 3
figure 3

The PorXFj CheY-like N-terminal domain. Overlay of the PorXFj (cyan), active CheY (orange; PDB: 1FQW) and inactive CheY (yellow; PDB: 2CHE) structures. For clarity purpose, only the side chains of the CheY ‘switch’ residues (corresponding to PorXFj T82 and Y101) are displayed. The Mg2+ ion and the water molecules are shown as black and red spheres, respectively, and the hydrogen bonds are indicated by dashed black lines.

Full size image

The HBD domain

The PorXFj HBD folds as a three-helix bundle, and presents the same features as in the P. gingivalis PorXPg structure: a 80° bend of the polypeptide chain at its N terminus, and a kink at the end of the first helix. In PorXPg, this domain was proposed to interact with DNA26. However, the first hits returned by a structural similarity search within the Protein Data Bank (PDB) using the DALI server37 correspond to domains involved in protein–protein interaction: the fibronectin-binding protein RevA38, the plakin domain of plectin39, or the BAG domain40 for instance.

The PglZ domain

The PorXFj C-terminal PglZ domain adopts the classical α–β–α fold of the alkaline phosphatase superfamily, which consists of a catalytic domain with a central β-sheet of six β-strands surrounded by six and four α-helices on each side, and a β-rich cap subdomain at the entrance of the active site (Fig. 4A). The active site coordinates two Zn2+ ions: Zn1 is coordinated by the catalytic T271, D237, D414 and H415 residues while Zn2 is coordinated by residues H364, H499, D360 and a water molecule (Fig. 4A). Of note, the distance between the two ions is 5.4 Å, a distance much greater than the distance reported for any other alkaline phosphatase superfamily members41. In the PorXPg dimer, each monomer presents a distinct conformation of residues 359–367: in one monomer this region folds as an α-helix (conformation HH), while in the other it folds as an extended loop (conformation HL)33. As the HL conformation is not compatible with Zn2 binding, the monomer with this conformation is unlikely to be active. In contrast, the two monomers of the PorXFj dimer adopt the HH conformation with two coordinated Zn2+ ions and are therefore likely to be active. Comparison with the structure of an inactive PorXPg mutant (T272A) crystallized in complex with pGpG reveals that residues interacting with the ligand are strictly conserved in PorXFj, except Y332 that is replaced by T331 (Fig. 4B). However, the electron density around Y332 and its contacting pGpG purine group is poorly defined33, suggesting that this interaction is labile and could be not critical for the catalyzed reaction.

Figure 4
figure 4

The PorXFj PglZ C-terminal domain. (A) The catalytic and cap subdomains are shown in magenta and pink, respectively. The two Zn2+ ions (Zn1 and Zn2) are indicated as black spheres. Enclosed is a close-up view of the Zn2+ ions coordination, with the water molecule and the hydrogen bonds displayed as a red sphere and dashed black lines, respectively. (B) Overlay of the ligand-binding region of PorXFj (magenta) with PorXPg (grey) in complex with pGpG (yellow). The Zn2+ ions present in PorXFj and PorXPg are shown as black and grey spheres, respectively. The side chains of PorXPg residues in contact with pGpG and of the corresponding PorXFj residues are indicated. For clarity purpose, only PorXFj residues are labelled, and the molecules are slightly rotated compared to panel (A).

Full size image

Purified PorXFJ has phosphodiesterase activity

PorXPg was previously shown to have phosphodiesterase activity in vitro33. Enzymatic assays using the purified PorXFj protein showed that it also hydrolyzes bis-p-nitrophenyl-phosphate (Fig. 5). Similarly to PorXPg, this activity requires the presence of a divalent cation, but unlike PorXPg, PorXFj is active not only in presence of Zn2+, but also in presence of Cu2+ and Mn2+ (Fig. 5). The presence of the divalent cation is necessary but also sufficient for PorXFj activity as a comparable activity was measured when the protein was phosphorylated or not prior to the reaction (Fig. 5). Such a broad metal specifity was also observed for the Sinorhizobium meliloti PhnA alkaline phosphatase41. However, comparison of the PorXPg and PorXFj structures highlights no significant structural difference in the ions pocket that could explain the different specificity of the two proteins.

Figure 5
figure 5

PorXFj has metal-dependent phosphodiesterase activity. The phosphodiesterase activity of PorXFj or phosphorylated PorXFj (pre-incubated with acetylphosphate and MgCl2, P-PorXFj) was measured using bis-p-nitrophenyl phosphate (bis-pNPP) as substrate. The average activity (nitrophenol released (measured at A405) per minute, from three independent measurements) is represented as a bar, with the three raw values (closed circles) and standard deviation (red vertical line). Reaction buffer was supplemented with MgCl2 (Mg2+), CaCl2 (Ca2+), CuCl2 (Cu2+), ZnCl2 (Zn2+), MnCl2 (Mn2+) or no metal.

Full size image

SEC-MALS analysis showed that PorXFj incubation with Zn2+ induced dimerization, similarly to PorXPg33 (Fig. 2). Interestingly, dimers induced by phosphorylation (AcP + Mg2+) and Zn2+ eluted at different volumes, suggesting that they adopt different conformations, which confirms SAXS analysis of PorXPg33. Moreover, a dimer/monomer equilibrium was observed when PorXFj was incubated with Zn2+, which could be explained by a dilution effect as the SEC-MALS analysis was carried out with the column equilibrated in PBS only. Therefore, SEC analysis of PorXFj dimerization induced by the different ions necessary for activity (Zn2+, Cu2+, and Mn2+) was carried out with the column equilibrated with PBS supplemented with the corresponding metal solutions. As expected, full dimerization was observed in the presence of Zn2+ and Cu2+ (Fig. 6). A dimer/monomer equilibrium was still observed in the presence of Mn2+ that could reflect a lower affinity of PorXFj for this ion.

Figure 6
figure 6

PorXFj dimerization is induced by different metals. The purified PorXFj alone or incubated with ZnCl2 (Zn2+), CuCl2 (Cu2+) or MnCl2 (Mn2+) was analyzed by size-exclusion chromatography (SEC). Note that the column was equilibrated in PBS supplemented with the corresponding metal solution.

Full size image

The CheY domain of PorXFj mediates interaction with GldL

The PorXPg protein was shown to interact with PorL24, a component of the PorLM rotor5,6,21,42, a situation resembling the interaction of CheY with the FliN subunit of the flagellar C-ring. Bacterial two-hybrid (BACTH) assays showed that this is also the case of PorXFj, which interacts with the cytoplasmic domain of the PorL homologue, GldL (GldLC, Fig. 7A). Interestingly, reminiscent of the CheY/FliN interaction, our analyses defined that the PorXFj/GldLC interaction is mediated by the PorXFj CheY-like RD (Fig. 7A). We further tested whether activation of the CheY-like domain of PorXFj is required for GldL interaction. Figure 7B shows that while a phosphomimetic substitution of the phosphorylated D54 residue (D54E) maintains the interaction with GldL, a phosphoablative substitution (D54A) prevents PorXFj–GldL interaction.

Figure 7
figure 7

The activated form of the PorXFj CheY-like domain interacts with GldL. Bacterial two-hybrid assay. BTH101 reporter cells producing the indicated domains/proteins (A, same color code as in Fig. 1: CheY domain, cyan; HBD, grey; PglZ domain, magenta) and CheY D54 phosphomimetic (D54E) and phosphoablative (D54A) variants (B) fused to the T18 and T25 domain of the Bordetella adenylate cyclase were spotted on X-Gal-IPTG reporter LB agar plates. The blue color of the colony reports interaction between the two partners. Controls include T18 and T25 fusions to TssF and TssG, two T6SS proteins from enteroaggregative E. coli that interact but unrelated to the T9SS43.

Full size image

Discussion

In P. gingivalis, the T9SS activity was shown to be regulated in part by the TCS composed of the histidine sensor PorY and the response regulator PorXPg. The mode of action of PorXPg remains elusive, notably on how its phosphodiesterase activity and its phosphorylation status influence T9SS activity. In addition, its potential interaction with DNA is controversial. It was recently shown that PorXPg activation is induced by phosphorylation of its RD, resulting in dimerization of the protein, and the structure of active, dimeric PorXPg was solved33. In this study, we solved the crystal structure of PorXFj, the F. johnsoniae PorX homolog. Comparison with the PorXPg structure strongly suggests that the two proteins share similar functions. Indeed, PorXFj possesses phosphodiesterase activity, it crystallized as a dimer identical to the active PorXPg dimer, the CheY-like RD adopts an active-like conformation, and the active site of the PglZ phosphatase-like domain is highly conserved with PorXPg. Interestingly, crystallization is sufficient to promote dimerization of PorXFj without prior activation by phosphorylation, as previously observed44. We propose that PorXFj dimerization relies on the binding of the Mg2+ ion to the PorXFj RD, inducing its active-like conformation, in conjunction with high concentration conditions occurring during the crystallization process that promote molecular contacts. PorXPg dimerization was not observed in the absence of activation, probably due to a less stable dimer than PorXFj. Unfortunately, there is no monomeric PorX structure that could provide clues to the dimerization process, particularly if conformational changes are involved.

Similarly to PorXPg, PorXFj lacks any canonical DNA-binding motif, which further questions about the ability of PorX, or PorXFj, to directly interact with DNA. Rather, PorXPg interacts with the extracytoplasmic function (ECF) sigma factor SigP, that itself interacts with DNA25. Only one homolog of SigP is present in F. johnsoniae (GenBank: PZQ89414.1). Further investigation would be necessary to analyze the putative interaction of PorXFj with the SigP homolog, and more generally to assess its involvement in T9SS gene regulation. In addition to its interaction with SigP, PorXPg was shown to interact with the cytoplasmic domain of PorL, a component of the PorLM motor that uses the proton-motive force (PMF) to energize effector secretion through the T9SS5,6,24,45. Our results demonstrated that this interaction is conserved in F. johnsioniae, suggesting that PorXFj may regulate T9SS activity. In addition, we showed here that the PorXFj CheY-like domain is sufficient to mediate interaction with the PorL homolog, GldL. The interaction of a CheY-like protein with a multiprotein complex is well documented in the case of the regulation of the flagellum rotation, in which the chemotaxis CheY protein interacts with FliM and FliN, two components of the C-ring32,46. Binding of phosphorylated CheY to FliM/N induces a tilting movement of the C-ring that repercutates onto FliG to reverse the direction of motor rotation32,47,48,49,50,51. Interestingly, mutagenesis of the conserved phosphorylable aspartate residue D54 further showed that a substitution preventing D54 phosphorylation prevents PorXFj binding to GldL. Taken together, we propose that the PorXFj phosphorylation status controls association to and dissociation from the GldLM complex. We further speculate that this interaction regulates the activity of the motor as a response of an environmental cues or of PglZ domain activity. In the case of the flagellum, CheY association to the C-ring controls the reversible switch between clockwise and counterclockwise rotation, hence enabling cells to swim towards favorable chemical habitats50,51. In F. johnsioniae, the PMF-dependent activity of the T9SS GldLM rotor powers effector transport through the outer membrane but also the movement of the SprB adhesin at the cell surface, hence supporting gliding5,6,45,52. One may hypothesize that PorXFj may have a function comparable to CheY by switching the T9SS motor to different conformations allowing to control effector secretion or adhesin displacement. PorXFj may thus control the speed of gliding or the gliding direction. Further studies should be performed to better understand the contribution of PorXFj, and of its CheY, HBD and PglZ domain in gliding motility, T9SS gene regulation and SigP interaction.

Experimental procedures

Bacterial strains, media, chemicals and growth conditions

Escherichia coli K-12 strain DH5a was used for all cloning procedures, T7 strain for protein production, and BTH101 for bacterial two-hybrid assays, E. coli cells were grown in Lysogeny Broth (LB) or Turbo broth supplemented with antibiotics when necessary (kanamycin 50 µg mL−1, ampicillin 100 µg mL−1). Expression from pLIC03 and BACTH vectors was induced with 1 mM and 0.5 mM of isopropyl β-d-1-thiogalactopyranoside (IPTG), respectively.

Plasmid construction

The sequence encoding full length PorXFj (Fjoh_2906) was amplified from F. johnsoniae genomic DNA (ATCC17061/DSMZ2064) using the following primers: 5ʹ-CCTGTACTTCCAATCAATGGATAAGATAAGAATACTTTGGGTCG and 5ʹ-CCGTATCCACCTTTACTTTATTATTTAGGGTTAAATACCAAAAACGG. The sequence was cloned into pLIC03 (kindly provided by the BioXtal company, Marseille) using the In-Fusion technology (Takara), following the manufacturer protocol. The pLIC03 expression vector is a pET-28a + derivative (Novagen) carrying a cassette coding for a His6 tag and a Tobacco Etch Virus (TEV) protease-cleavage.

BACTH vectors producing TssF, TssG and GldLC fused to the Bordetella adenylate cyclase T18 or T25 domains have been previously published6,43. BACTH plasmid producing PorXFj and PorXFj domains fused to the T18 and T25 domains were constructed by restriction free (RF) cloning53 using oligonucleotides CGCCACTGCAGGGATTATAAAGATGACGATGACAAGGATAAGATAAGAATACTTTGGGTCGATGATGAG and CGAGGTCGACGGTATCGATAAGCTTGATATCGAATTCTAGTTATTTAGGGTTAAATACCAAAAACGGAATAATCATTTC (T18-PorXFj), CGCCACTGCAGGGATTATAAAGATGACGATGACAAGGATAAGATAAGAATACTTTGGGTCGATGATGAG and CGAGGTCGACGGTATCGATAAGCTTGATATCGAATTCTAGTTATTTTTGGTAATCTAATGTTGTTTTTTCTGTAATCAGTC (T18-CheY), CGCCACTGCAGGGATTATAAAGATGACGATGACAAGGAATTCCGCAAAATCTCGATGGAATTAGC and CGAGGTCGACGGTATCGATAAGCTTGATATCGAATTCTAGTTATTTTGGAGCAAACCAGTCTTCGTAATTTC (T18-HBD), and CGCCACTGCAGGGATTATAAAGATGACGATGACAAGGCAGATAAACCAATTCAATCTCATAATTTATTTAAAGAATTAGTTG and CGAGGTCGACGGTATCGATAAGCTTGATATCGAATTCTAGTTATTTAGGGTTAAATACCAAAAACGGAATAATCATTTC (T18-PglZ) (sequences annealing on the target plasmids in italics). Briefly, the DNA fragment was amplified using primers that introduced extensions annealing to the target vector. The double-stranded product of the first PCR has then been used as primer for a second PCR using the target vector as template. PCR products were then treated with DpnI to eliminate template plasmids and transformed into DH5a-competent cells. Substitutions were introduced by site-directed mutagenesis using complementary oligonucleotides bearing the desired mutation (CTTTGACATTGTTTTTCTTGCCGAAAATATGCCGGGAATG and CATTCCCGGCATATTTTCGGCAAGAAAAACAATGTCAAAG for D54A, CTTTGACATTGTTTTTCTTGAGGAAAATATGCCGGGAATG and CATTCCCGGCATATTTTCCTCAAGAAAAACAATGTCAAAG for D54E; mutagenized codon in italics, mutagenized bases underlined). All plasmids have been verified by DNA sequencing (Eurofins).

Protein production, purification and analysis

PorXFj was produced in E. coli T7 cells cultured in Turbo Broth medium at 37 °C. At OD600nm of 0.6–0.8, porXFj expression was induced by adding 1 mM IPTG and the bacterial growth was pursued for 18 h at 17 °C. Cells were harvested by centrifugation at 4000×g for 10 min, resuspended in lysis buffer (50 mM Tris–HCl pH 8.0, 300 mM NaCl, 10 mM imidazole, 250 µg mL−1 lysozyme, 1 mM PMSF) and frozen overnight at − 20 °C. After thawing, 20 µg mL−1 of DNase and 1 mM of MgSO4 were added, and cells were lysed by sonication. The pellet and soluble fractions were separated by centrifugation at 16,000×g for 30 min, and the His6-tagged protein was purified from the soluble fraction by immobilized metal ion affinity chromatography using a 5 mL HisTrap crude (GE Healthcare) Ni2+-chelating column equilibrated in buffer A (50 mM Tris–HCl p H8.0, 300 mM NaCl, 10 mM imidazole). The protein was eluted with buffer A supplemented with 250 mM imidazole and further purified by size exclusion chromatography (HiLoad 16/60 Superdex 200 prep grade, GE Healthcare) equilibrated in 10 mM HEPES pH 7.5, 500 mM NaCl. Selenomethionine-labeled (SeMet) PorXFj was produced and purified with the same protocol as the native PorXFj, except that the cells were grown in SeMet minimal medium54.

Dimerization analysis

For the Size-Exclusion Chromatography Multi-Angle Light Scattering (SEC-MALS) analysis, the purified PorXFj (1.6 mg mL−1) was incubated for 1 h at RT with 20 mM acetyl phosphate (AcP) + 10 mM MgCl2, or 20 mM AcP, or 10 mM MgCl2, or 100 µM ZnCl2. The samples were loaded on a Superdex 200 Increase 10/300 GL column (GE Healthcare) equilibrated in PBS at a flow rate of 0.6 mL min−1, using an Ultimate 3000 HPLC system (Fischer Scientific). Detection was performed using an eight-angle light-scattering detector (DAWN8, Wyatt Technology) and a differential refractometer (Optilab, Wyatt Technology).

For the Size-Exclusion Chromatography (SEC) analysis, the purified PorXFj (1.6 mg mL−1) was incubated for 1 h at RT with 100 µM ZnCl2, MnCl2, or CuCl2. The samples were loaded on a Superdex 200 Increase 10/300 GL column (GE Healthcare) equilibrated in PBS supplemented with100 µM of the corresponding metal solution.

Crystallization, data collection and processing

The purified PorXFj and SeMet PorXFj were concentrated to 10 and 12 mg mL−1, respectively. Crystallization trials were performed using the sitting-drop vapor-diffusion method at 293 K in 96-well Swissci-3 plates, with Stura Footprint (Molecular Dimensions), Wizard I and II (Rigaku), Structure I and II (Molecular Dimensions) and JCSG + (Qiagen) screens, and using Tecan and Mosquito (TTP Labtech) robots to fill in the plates and dispense the drops, respectively. PorXFj crystals appeared in condition No. 20 from Structure I screen (0.2 M calcium acetate, 0.1 M sodium cacodylate pH 6.5, 18% PEG 8000). SeMet PorXFj crystals appeared in several condition, and after optimization55, the final crystallization conditions were 0.2 M ammonium sulfate, 0.05 M sodium acetate, 0.05 M sodium citrate pH 5.0–6.0, 10–30% PEG 2000. Crystals were mounted in cryo-loops (Hampton CrystalCap Magnetic) and were briefly soaked in crystallization solution supplemented with 20% (v/v) polyethylene glycol. The crystals were flash-cooled in a nitrogen-gas stream at 100 K using a home cryocooling device (Oxford Cryosystems).

Native diffraction data of PorXFj and single wavelength anomalous dispersion (SAD) data of SeMet PorXFj were collected to 2.0 Å and 2.3 Å resolution, respectively, on beamline Proxima-1 at the Soleil synchrotron (Paris, France). The data sets were integrated with XDS and scaled with SCALA56 from the CCP4 suite57. Heavy atom substructure determination of SeMet PorXFj, phase calculations and density modification were performed using HySS58, Phaser59 and Parrot60, respectively, as implemented in the Phaser SAD pipeline from the CCP4 suite. A partial SeMet PorXFj model was built automatically in Buccaneer61, completed manually in COOT62, and was subsequently used as model for molecular replacement with MOLREP63 to solve the structure of native PorXFj. Refinement, correction, and validation of the structure were performed with autoBUSTER64, COOT, and Molprobity65, respectively. Data collection and refinement statistics are reported in Table 1.

Table 1 Data collection and refinement statistics of PorXFj.
Full size table

Bacterial two-hybrid assays.

Bacterial two-hybrid assays were conducted as previously described43. The proteins or domains to be tested were fused to the isolated T18 and T25 catalytic domains of the Bordetella adenylate cyclase. After introduction of the two plasmids producing the fusion proteins into the BTH101 reporter strain, plates were incubated at 28 °C for 24 h. Three independent colonies for each transformation were inoculated into 600 μL of LB medium supplemented with ampicillin, kanamycin, and IPTG (0.5 mM). After overnight growth at 28 °C, 15 μL of each culture were spotted onto LB plates supplemented with ampicillin, kanamycin, IPTG, and X-Gal and incubated at 28 °C. Controls include interaction assays with TssF and TssG, two T6SS protein partners from enteroaggregative E. coli43 unrelated to the T9SS. The experiments were done at least in triplicate and a representative result is shown.

Phosphodiesterase activity assay

PorXFj phosphodiesterase activity was measured using bis-p-nitrophenyl phosphate (bis-pNPP, Sigma-Aldrich) as substrate, essentially as previously published33 (Schmitz et al. 2022). Briefly, purified PorXFj was phosphorylated by incubation for 1 h in the presence of 20 mM of acetyl-phosphate and 10 mM of MgCl2, before being desalted on a 7-kDa Zeba™ Spin desalting column (ThermoFischer Scientific). 50 μL of 2 μM purified phosphorylated or non-phosphorylated PorXFj in 50 mM Tris–HCl pH8, 150 mM NaCl were mixed with 50 μL of 50 mM Tris–HCl pH8.5, 150 mM NaCl, 10 mM bis-pNPP supplemented, or not, with 100 μM of MgCl2, 100 μM of CaCl2, 100 μM of CuCl2, 100 μM of MnCl2 or 100 μM ZnCl2 in a 96-well microplate (transparent, flat bottom, Nunc). The release of p-nitrophenol was monitored for 180 min at 37 °C by measuring the absorbance at 405 nm (A405) using a TECAN microplate reader. A control experiment with bis-pNPP only was also performed to measure the spontaneous hydrolysis, and the values of bis-pNPP spontaneous hydrolysis were subtracted to the measures of PorXFj activity. Bis-pNPP activity was calculated from the slope and reported as A405 per minute. The experiments were done in triplicate.