The SiaABC threonine phosphorylation pathway controls biofilm formation in response to carbon availability in Pseudomonas aeruginosa

The critical role of bacterial biofilms in chronic human infections calls for novel anti-biofilm strategies targeting the regulation of biofilm development. However, the regulation of biofilm development is very complex and can include multiple, highly interconnected signal transduction/response pathways, which are incompletely understood. We demonstrated previously that in the opportunistic, human pathogen P. aeruginosa, the PP2C-like protein phosphatase SiaA and the di-guanylate cyclase SiaD control the formation of macroscopic cellular aggregates, a type of suspended biofilms, in response to surfactant stress. In this study, we demonstrate that the SiaABC proteins represent a signal response pathway that functions through a partner switch mechanism to control biofilm formation. We also demonstrate that SiaABCD functionality is dependent on carbon substrate availability for a variety of substrates, and that upon carbon starvation, SiaB mutants show impaired dispersal, in particular with the primary fermentation product ethanol. This suggests that carbon availability is at least one of the key environmental cues integrated by the SiaABCD system. Further, our biochemical, physiological and crystallographic data reveals that the phosphatase SiaA and its kinase counterpart SiaB balance the phosphorylation status of their target protein SiaC at threonine 68 (T68). Crystallographic analysis of the SiaA-PP2C domain shows that SiaA is present as a dimer. Dynamic modelling of SiaA with SiaC suggested that SiaA interacts strongly with phosphorylated SiaC and dissociates rapidly upon dephosphorylation of SiaC. Further, we show that the known phosphatase inhibitor fumonisin inhibits SiaA mediated phosphatase activity in vitro. In conclusion, the present work improves our understanding of how P. aeuruginosa integrates specific environmental conditions, such as carbon availability and surfactant stress, to regulate cellular aggregation and biofilm formation. With the biochemical and structural characterization of SiaA, initial data on the catalytic inhibition of SiaA, and the interaction between SiaA and SiaC, our study identifies promising targets for the development of biofilm-interference drugs to combat infections of this aggressive opportunistic pathogen.

(T68). Crystallographic analysis of the SiaA-PP2C domain shows that SiaA is present as a dimer. 48 Dynamic modelling of SiaA with SiaC suggested that SiaA interacts strongly with phosphorylated 49 SiaC and dissociates rapidly upon dephosphorylation of SiaC. Further, we show that the known 50 phosphatase inhibitor fumonisin inhibits SiaA mediated phosphatase activity in vitro. In 51 conclusion, the present work improves our understanding of how P. aeuruginosa integrates 52 specific environmental conditions, such as carbon availability and surfactant stress, to regulate 53 cellular aggregation and biofilm formation. With the biochemical and structural characterization of 54 SiaA, initial data on the catalytic inhibition of SiaA, and the interaction between SiaA and SiaC,55 our study identifies promising targets for the development of biofilm-interference drugs to combat 56 infections of this aggressive opportunistic pathogen. 57

122
SiaABCD is involved in biofilm formation during growth on SDS and glucose 123 Cells of P. aeruginosa preferentially grow as macroscopic suspended biofilms (cellular 124 aggregates in the mm range) in minimal medium with SDS [11]. To investigate whether all genes 125 of the siaABCD operon [70] are involved in biofilm formation under these conditions, we tested 126 individual mutant strains in microtiter plates for macroscopic aggregation and compared it to 127 cultures growing on glucose (Fig 1). 128 As shown previously [70], growth on SDS led to the formation of macroscopic aggregates in the 129 parental strain but not in the ΔsiaA and ΔsiaD mutants. Likewise, no macroscopic aggregates 130 were observed in cultures of ΔsiaC, whereas ΔsiaB showed increased aggregation. The 131 aggregative behaviour of cells was inversely correlated with the optical density (OD600) of the 132 surrounding medium with higher OD600 values for strains ΔsiaA, ΔsiaC and ΔsiaD (darker colour 133 in normalized pictures) compared the parental strain or the ΔsiaB mutant. In contrast, no 134 macroscopic aggregates were found during growth with glucose for all strains. However, the 135 OD600 of the culture medium of glucose-grown cultures followed a similar pattern as those grown 136 on SDS. Growth on glucose has previously been found to occur mainly as attached or suspended 137 biofilms (aggregates of up to 400 nm in size), which quickly disperse upon carbon starvation in a 138 c-di-GMP-dependent manner [63]. To study, whether differences in the ability to form suspended 139 biofilms could explain the observed OD600 pattern of the sia-mutants, we analysed glucose-140 growing cultures with scanning electron microscopy (SEM) and laser diffraction analysis (LDA). 141 142 SiaABCD regulates the ratio of freely suspended and aggregated cells 143 SEM imaging (Fig 2, 10,000; left panels) showed the presence of three-dimensional structures 144 composed of densely packed microbial cells that projected above the substratum. The 145 multicellular aggregates appeared to be less structured in the ΔsiaA, ΔsiaC and ΔsiaD mutants 146 compared to the parental strain. In contrast, the ΔsiaB mutant formed much larger and more 147 densely packed aggregates than the parental strain. Similar differences in aggregate sizes were 148 also noted at lower magnification (1,300), with the additional observation that aggregates were 149 generally less abundant for the ΔsiaA, ΔsiaC and ΔsiaD mutants. 150 Light diffraction analysis (LDA) can be used to determine the size distribution of PAO1 aggregates 151 in liquid cultures 64 . LDA analysis in the current study confirmed these results for cultures of the 152 wildtype strain, which were dominated by particles > 10 µm (≥ 95% ± 2% of the total bio-volume 153 of all particles). Particles < 10 µm, which include the single cells, represented only a minor fraction 154 (≤ 5% ± 2%). In line with SEM results, the distribution of particles in cultures of the ΔsiaA, ΔsiaC 155 and ΔsiaD mutants were strongly shifted towards smaller sizes. A substantial increase in the total 156 bio-volume of all particles < 10 m for ΔsiaA (≥ 6.6 fold), ΔsiaC (≥ 11.8 fold) and ΔsiaD (≥ 8 fold) 157 was consistently observed. In contrast, for cultures of ΔsiaB the total bio-volume of all particles > 158 10 µm (≥ 99% ± 1%) was considerably higher than for the parental strain. Consequently, almost 159 no particles < 10 µm were detected. 160 161 SiaABCD regulates biofilm formation as a response to carbon availability 162 Biofilm formation is a dynamic process that varies according to the time point used for 163 quantification. The temporal nature of biofilm formation and dispersal can be monitored with a 164 microtiter plate-based biofilm assay [63]. When the impact of the SiaABCD proteins on the biofilm 165 life cycle during growth on glucose was studied using this assay, we found the following temporal 166 pattern of biofilm formation for the wildtype strain: i) a steady increase in biofilm formation while 167 glucose was present in excess; ii) a decrease in biofilm formation upon exhaustion of glucose; 168 and iii) an increase in optical density in the culture supernatant that lagged behind biofilm 169 formation, but continued to increase even after glucose was completely consumed (Fig 3). The 170 latter observation is consistent with the dispersal of cells from the attached and/or suspended 171 biofilm into the liquid phase in response to carbon starvation [63]. The ΔsiaA, ΔsiaC and ΔsiaD 172 mutants exhibited a similar growth pattern on glucose as described for the parental strain (Fig 3  173 SiaB are both enzymatically active on their respective substrates, catalysing a cycle of SiaC 225 phosphorylation and de-phosphorylation in dependency of ATP availability. 226 To determine the exact site of phosphorylation, we performed kinase (SiaB + SiaC) and 227 phosphatase (SiaA-PP2C + SiaC P ) assays and analysed the SiaC/SiaC P protein by PMS after 228 gel purification (Fig 4E and Table S1). For SiaB with SiaC in the presence of ATP, we observed 229 a 1000-fold increase of the phosphorylated T68-containing peptide (peptide [LLYLNTSSIK]) 230 compared conditions in which ATP was omitted from the assay (1.5 x 10 9 vs. 1.3 x 10 6 peak area). 231 Incubation of SiaA-PP2C with SiaC P resulted in > 10-fold decrease in the phosphorylated peptide 232 (LLYLNTSSIK) compared to the control incubation without SiaA-PP2C added. Hence, the PMS 233 data obtained from these enzyme assays provide direct evidence that T68 of SiaC is a target for 234 phosphatase and kinase activity of SiaA  in these solvent-exposed loops regions. The SiaA-PP2C monomer adopts the canonical α-β-β-α 264 fold first described for the human PPM1A [76] and subsequently found in several other 265 phosphatases [77,78] ( Fig 5A). The core structure of the SiaA-PP2C monomer is a β-sandwich 266 composed of two antiparallel β-sheets each comprising five β-strands. The active site is at the 267 apex of the β-sheet structure (Fig 5A,B and D). This core structure is flanked on the N-terminal 268 side by helices α1-α4 and on the C-terminal end by helices α5-α7, forming a four-layered αββα 269 structure. The two monomers interact extensively with each other through residues projecting 270 from helices α1, α2 and α3, forming a compact homodimer that buries a total solvent accessible 271 area of 1392 Å 2 (Fig 5B). Consistently, in size-exclusion chromatography, instead of monomer 272 (~33 kDa), SiaA-PP2C eluted as a 66 kDa dimer, suggesting SiaA exists mainly as dimer in 273 solution ( Fig 5C). The homodimer presents two concave surfaces that are likely to accommodate 274 the incoming substrates ( Fig 5D). 275 An automated search of 3D structures similar to SiaA-PP2C, using the Dali server 276 (ekhidna2.biocenter.helsinki.fi/dali), returned PP2C-type phosphatases as the top matches, the 277 best of which was Rv1364C (PDB code: 3KE6, chain A) from Mycobacterium tuberculosis with a 278 Z score of 23.6 and a RMSD of 2.2 Å for 210 superimposed α-carbon atoms. Despite a very low 279 overall amino-acid sequence identity (17%), these structures share a conserved fold and several 280 strictly conserved active site residues with SiaA-PP2C: D457, D474, G477, D600, G601 and 281 D653 (Fig 5E and Fig S2). it is possible that Mn 2+ and Mg 2+ both function as co-factor of SiaA-PP2C. Metal-coordinating 289 residues are all conserved except C475 of SiaA-PP2C, which interacts with metal ion M2 through 290 its carbonyl oxygen (Fig 5E and F). 291 Compared to Rv1364C, which only has two bound metal ions, a third Mg 2+ ion (M3) was found in 292 SiaA-PP2C coordinated by D600 and with an incomplete octahedral coordination shell. D604 and 293 D534 assist in stabilization of M3 through formation of hydrogen bond with M3-coordinating water 294 molecules. Given the structural similarity with PP2C [76], an SN2 mechanism is the most plausible 295 whereby the water molecule bridging M1 and M2 (Fig 5E) performs a nucleophilic attack of the 296 phosphorus atom from the phosphate bound to the threonine target residue. However, M3 could 297 also play a direct role in the catalytic activity, as was recently proposed for the human PPM1A 298 protein, a negative regulator of cellular stress response pathways [79,80]. Further study is needed 299 to investigate whether a third metal ion is present in all PP2C. We note that near the entrance of 300 the active site, a hydrophobic pocket lined with residues I554, V564, L603, F619 and A620 301 constitutes a potential site for the design of inhibitors (Fig 5D and G). Another distinct feature of 302 SiaA-PP2C is a flexible loop located between β7 and β8, whereas a Flap subdomain, which has 303 been reported to aid in substrate specificity, is present at the equivalent position in human PPM1A 304 and many other PP2C-type phosphatases [80] (Fig 5H). Therefore, this shorter yet flexible loop 305 might contribute to the substrate specificity of SiaA-PP2C. 306 307 Crystal structure of SiaC reveals similarity with anti-sigma factor antagonists 308 To gain insights about SiaC, its crystal structure was determined in this study. Native crystals of 309 SiaC were grown in C6 condition of the JCSG+ kit from Molecular Dimensions (0.1 M 310 Phosphate/citrate, pH 4.2, 40% (v/v) PEG 300), with space group I222 and diffracted to a 311 maximum resolution of 1.7 Å. The structure (PDB ID: 6K4F) was determined by SAD using SeMet 312 derivative crystals (Table S2). 313 Clear electron density maps allowed complete tracing except for residues 101-103, which are 314 located in the flexible loop between the α3 helix and β5 strands. The SiaC protein comprises six 315 β-strands arranged in a mixed parallel/antiparallel fashion and three α-helices ( Fig 6A). A 316 hydrophobic pocket next to T68 is formed by residues L61 and L63 projecting from β4, W95 from 317 β5, I71 and M74 from α2, L107 and F111 from α3 while L66 belongs to the β4-β5 loop ( Fig 6B  318 and C). Dali server was used to search for protein structures that are most similar to SiaC crystal 319 structure, and an anti-sigma factor antagonist from Mycobacterium paratuberculosis (PDB ID: 320 4QTP; gene ID MAP_0380; Fig 6D) was identified as the closest homologue (Z score of 9.0, an 321 amino-acid sequence identity of 10% and a RSMD of 2.8 Å over the -carbon atoms of 117 322 residues). Despite sharing very low sequence identity, a closer examination of both structures 323 reveals a conserved overall topology with several unique features, especially at β3, β4 strands 324 and α3 helix that are longer in 4QTP compared to SiaC. These variations are probably related to 325 differences in the ability to establish protein-protein interactions. Nonetheless, their target 326 phosphorylation sites, T68 for SiaC and S58 for 4QTP, are both located at the N-terminus of helix 327

α2. 328
It is worth mentioning that our SiaC crystal structure not only agrees with the SiaC structure 329 published by Chen et al. [73], but also provides more details at 1.7Å resolution. A 2SiaB:2SiaC 330 complex was suggested by size-exclusion chromatography when SiaB and SiaC were mixed in 331 equal molar ratio and incubated at room temperature for 1 h (Fig 6E). Instead of two individual 332 peaks (one consists of SiaB and one consists of SiaC), a single peak containing the complex was 333 observed. This 2SiaB:2SiaC complex model is consistent with the recently reported crystal 334 structure of the SiaB/SiaC complex [73]. When the SiaB/SiaC mixture was supplemented with 10 335 mM ATP and 20 mM MgCl2 prior to size-exclusion chromatography, the complex disappeared 336 and SiaB and SiaC p eluted as two individual peaks. The phosphorylation state of SiaC P was 337 subsequently confirmed by LC-MS. As such, our data demonstrate that SiaB binds tightly to SiaC 338 but quickly dissociation from SiaC P after phosphorylation. 339 340 Molecular Docking and MD simulation of SiaC P with the phosphatase domain of SiaA 341 As shown above, the SiaB/SiaC complex quickly dissociates upon phosphorylation. Since SiaA-342 PP2C targets SiaC P , we wanted to study the interaction between SiaA and SiaC during the 343 dephosphorylation reaction. Therefore, we performed a MD simulation on SiaA binding to SiaC 344 and to SiaC P . The SiaA/SiaC P complexes were predicted by molecular docking of SiaA-PP2C 345 dimer to SiaC and to SiaC P using ZDOCK, followed by 60 ns of MD simulation (Fig 7A and B). 346 These simulations suggest that the phosphorylated pT68 is required for productive binding to the 347 catalytic site of SiaA-PP2C, where it remains efficiently bound via the catalytic Mg 2+ ion M3 and 348 the bridging water molecules surrounding it (Fig 7C-E). This result is consistent with the 349 experimental observation that only pT68 in SiaC P constitutes the substrate for SiaA-PP2C activity. 350 In contrast, the non-phosphorylated T68 dissociated from the catalytic site in less than 15 ns of 351 simulation, suggesting that the interaction of SiaC with SiaA is weak upon dephosphorylation. 352 Based on the simulated model, several key interactions that could contribute to substrate 353 specificity of SiaA-PP2C were identified. First, R652 stabilizes the positively charged phosphate 354 group of pT68 through the formation of a salt bridge ( Fig 7F). In addition, the flexible loop between 355 α5 and α6 of SiaA inserts into a shallow pocket located between α2, α3, β4, β5 and β6 of SiaC P 356 and establishes hydrophobic contacts with L61, L66, I71, M74, M75, L78, W95, L107 and F111 357 through F612 (Fig 7G). Notably, the salt bridge identified between R618 and E609 of SiaA is likely 358 to restrain the orientation of the flexible loop between α5 and α6, which could favour interaction 359 with SiaC P . Notably, F612 was recently found to be deleted in the SiaA protein from a non- for the interaction with the hydrophobic pocket of SiaC P (Fig 7H1-I2). As such, a stable SiaC P -366 SiaA-PP2C* complex seems unlikely to form, which would explain the non-aggregative phenotype 367 of the ∆siaA mutant strain due to an impaired enzymatic activity towards SiaC P . 368 With our study, we demonstrated that the protein phosphatase SiaA and its kinase counterpart 396 SiaB target the protein SiaC at threonine 68 (T68) and provide evidence that the SiaABC proteins 397 make use of a partner switch mechanism to balance cellular aggregation in response to carbon 398 availability. A partner switch system (PSS) typically involves a PP2C phosphatase (SiaA), an anti-399 sigma factor with kinase activity (SiaB) and anti-sigma factor antagonist representing the target 400 for phosphorylation (SiaC). In line with such a mechanism, the phosphorylation status of the 401 proposed anti-sigma factor antagonist SiaC is crucial for the switch to occur. In addition to the 402 sequestration of an alternative sigma factor, partner switch systems have recently been found to 403 regulate important physiological traits also by affecting enzymatic activities involved in cellular 404 signalling cascades through direct protein-protein interactions. 405 In P. aeruginosa, the HptB-HsbR-HsbA system was shown to promote DGC activity of HsbD 406 following its interaction with the phosphorylatable anti-sigma factor antagonist HsbA to regulate SiaA would shift the equilibrium of SiaC/SiaC P towards the unphosphorylated form of SiaC, which 420 subsequently interferes with SiaD activity, leading to increased c-di-GMP production and 421 subsequently biofilm formation. Such a scenario would explain the strong aggregative phenotype 422 of the SiaB mutant observed in this and previous studies [99]. As a negative feedback response, 423 SiaC P has an increased binding affinity towards SiaA, which would lead to a negative impact on 424 SiaD activity through the dephosphorylating of T68 and increased binding to SiaB.

Strains and growth conditions 468
Bacterial strains (Table S3)  Cre recombinase, the pCre1 plasmid was introduced into the P. aeruginosa mutant strain 480 harbouring the ISphoA/hah insertion by conjugation using biparental mating with S17-λpir. An 481 overnight culture of the P. aeruginosa recipient strain was grown in 10 mL LB at 42 o C with shaking 482 at 50 rpm, while the E. coli donor strain was grown in 10 mL LB at 37 o C with shaking at 150 rpm. 483 Optical density measurements were used to provide 2 × 10 9 cells of the recipient strain and 1 × 484 10 9 cells donor strain, where an OD600 = 1 is approximately 1 × 10 9 cells/mL. Cells were washed 485 twice in 1 mL pre-warmed LB with centrifugation for 30 sec at 10,000 × g at room temperature 486 Experiments were performed as at least 6 independent replicates and representative images are 535 shown. 536

Analytics of supernatants 537
For glucose quantification, the liquid cultures were filtered through a 0.22 µM PES filter (PN 4612, 538 Pall) and quantified using the GO assay kit (GAGO20, Sigma Aldrich). Briefly, one volume of 539 sample or standard solution (20 µL) was mixed with two volumes of assay reagent (40 µL) in a 540 96 well plate and incubated statically at 37C for 30 min. Subsequently, the reaction was stopped 541 by the addition of 12 N H2SO4 (40 µL) to the reaction mixture. Glucose concentrations were 542 quantified at 540 nm using a microplate reader (Infinite 200 pro, Tecan). Zero-80 µg/mL glucose 543 standard solutions were used for the calibration curve. 544 Concentrations of succinate, ethanol and 2,3-butanediol were quantified by a HPLC method using 545 an Aminex HPX-87H 300 x 7.8 mm ion-exchange column (BioRad, Munich, Germany) heated to 546 60°C. The eluent was 5 mM H2SO4, which was delivered to the column by a LC-10ATvp pump 547  The purification process was assessed with SDS-PAGE gel (Fig S2) and the final protein batch 699 was aliquoted into smaller fractions, frozen in liquid nitrogen and stored at -80°C. 700 Phosphatase assays with purified SiaA-PP2C and fumonisin B1 701 The phosphatase activity of SiaA-PP2C was assessed using para-nitrophenyl phosphate (pNPP) 702 or as the target substrate. Phosphatase activity was determined via the production of a 703 colorimetric para-nitrophenolate product that absorbs strongly at 405 nm. Measurements were taken at 620 nm using a microplate reader. Readings were then converted 725 to phosphate measurements using a standard curve. For ADP-Glo™ assays, 20 uL of the ADP-726 Glo™ reagent was added to 20 uL of the sample to stop the reaction and deplete remaining ATP. 727 If enzyme reactions were carried out in the absence of Mg 2+ , MgCl2 was supplemented to the 728 mixture to a final concentration of 20 mM. The mixture was incubated at room temperature for 40 729 min. A second kinase detection reagent was added to the mixture to convert ADP consumed to 730 ATP and for conversion of the ATP signal to luminescence signals and further incubated for 45 731 min. Luminescence was measured using a microplate reader and converted to ATP consumed in 732 kinase reaction using a standard curve generated. At least two independent biological replicates 733 with technical duplicates were carried out for each enzymatic reaction. To improve resolution, the His-tag of SiaA protein was removed by TEV cleavage, prior to the 766 native crystal screen. A native SiaC crystal was obtained in 0.1 M Phosphate/citrate, pH 4.2, 40% 767 v/v PEG 300 at 1.7 Å. SAD phasing and initial model building was processed with AutoSol from 768 the Phenix package, followed by structure refinement with Phenix Refine and PDB deposition 769 Homology modelling, molecular docking and molecular dynamics (MD) simulations 771 The homology model of SiaB was generated with the I-TASSER online Server by specifying the 772 amino acid sequence of SiaB. The homology model of SiaA-PP2C* was generated with SWISS-773 MODEL by providing the amino acid sequence of SiaA-PP2C* and specifying the SiaA-PP2C 774 crystal structure as template. 775 SiaC P model was generated by substituting T68 with pT68 using psfgen in VMD [111]. The model 776 was then subjected to conjugate gradient energy minimization for 10000 steps using NAMD 2.11 777 [112]. SiaC P was docked to the SiaA dimer by ZDOCK 3.0.2 [113] by specifying pT68 as 778 interacting residues. Among the top 10 complexes, complex 2 was chosen as the best model 779 because the pT68 was closest to the bridging water. 780 SiaC P -SiaA and SiaC-SiaA dimers were subjected to all-atom, explicit solvent MD simulation 781 using NAMD 2.11. Each dimer model was simulated in a water box, where the minimal distance 782 between the solute and the box boundary was 15 Å along all three axes. The charges of the 783 solvated system were neutralized with counter-ions, and the ionic strength of the solvent was set 784 to 150 mM NaCl using VMD. The fully-solvated system was subjected to conjugate-gradient 785 minimization for 10,000 steps, subsequently heated to 310 K, and a 10 ns equilibration with 786 protein backbone atoms constrained using a harmonic potential of the form U(x) = k (x-xref) 2 , 787 where k is 1 kcal/mol Å -2 and xref is the initial atom coordinates. Finally, 60 ns production 788 simulations were performed without constraints. All simulations were performed under the NPT 789 ensemble assuming the CHARMM36 force field for the protein and the TIP3P model for water 790 molecules. PDB files of the homology models and docking results used for the study are provided 791 as zipped Supplementary files. 792

Statistical analysis 793
When needed, experiments with n ≥ 3 were analysed using a two-sided students t-tests (α = 0.05) 794 with p-values < 0.05 interpreted as being significantly different. 795

796
The authors further declare that the research was conducted in the absence of any commercial 797 or financial relationships that could be construed as a potential conflict of interest. Janosch 798 Klebensberger would like to thank Prof. Bernhard Hauer for his continuous support. We also 799 acknowledge beam time allocation at the SOLEIL synchrotron. The MD simulations for this article 800 were performed on ASPIRE 1 of the National Supercomputing Centre, Singapore 801