GGDEF domain as spatial on-switch for a phosphodiesterase by interaction with landmark protein HubP

In bacteria, the monopolar localization of enzymes and protein complexes can result in a bimodal distribution of enzyme activity between the dividing cells and heterogeneity of cellular behaviors. In Shewanella putrefaciens, the multidomain hybrid diguanylate cyclase/phosphodiesterase PdeB, which degrades the secondary messenger c-di-GMP, is located at the flagellated cell pole. Here, we show that direct interaction between the inactive diguanylate cyclase (GGDEF) domain of PdeB and the FimV domain of the polar landmark protein HubP is crucial for full function of PdeB as a phosphodiesterase. Thus, the GGDEF domain serves as a spatially controlled on-switch that effectively restricts PdeBs activity to the flagellated cell pole. PdeB regulates abundance and activity of at least two crucial surface-interaction factors, the BpfA surface-adhesion protein and the MSHA type IV pilus. The heterogeneity in c-di-GMP concentrations, generated by differences in abundance and timing of polar appearance of PdeB, orchestrates the population behavior with respect to cell-surface interaction and environmental spreading.


INTRODUCTION
Cyclic mono-or oligonucleotide second messenger molecules are ubiquitously occurring in cells from all three domains of life and are implicated in a wide array of signaling networks at the single or multiple cell level. The second messenger cyclic bis-(3′-5′)cyclic diguanylic acid (c-di-GMP) has emerged as an important signaling molecule in a multitude of bacterial species [1][2][3] . c-di-GMP is synthesized from two molecules of GTP by specific diguanylate cyclases (DGCs), which are characterized by a signature GGDEF pentapeptide motif within their reactive center. Degradation via hydrolysis of c-di-GMP is mediated by phosphodiesterases (PDEs) that contain either signature EAL or HD-GYP domains and convert c-di-GMP to linear pGpG or to GMP, respectively (reviewed in ref. 4 ). These domains may occur as stand-alone proteins, however, more frequently they are found as parts of larger multidomain proteins together with various sensor input or output domains. Often, both GGDEF and EAL domains jointly occur as part of multidomain (sensor)proteins, which may localize in the cytoplasm or may be membrane-bound. The processes c-di-GMPmediated regulation is involved in are highly diverse. In numerous bacterial species this secondary messenger plays a key role in controlling the shift between sessile and planktonic lifestyle. In addition, c-di-GMP is implicated in regulation of the cell cycle, cellular development, differentiation, and virulence 1,4,5 .
Mechanisms that ensure appropriate and specific phenotypic responses to c-di-GMP include timing of PDE/DGC or effector production, effector affinity to c-di-GMP, or spatial control by sequestration of the regulator/effector components into functional clusters or to specific cellular compartments 3,6,7 . In several bacteria, DGCs and/or PDEs are specifically recruited to the cell poles. In Caulobacter crescentus, which has an asymmetric cell cycle in which a sessile stalked cell produces a motile swarmer cell, c-di-GMP is an important regulatory factor of cell cycle control and pole morphogenesis [8][9][10] . One key protein is the DGC PleD, which is localized to and activated at the pole of the sessile stalked mother cell. In contrast, an active PDE, PdeA, is present at the opposite pole of the newborn swarmer cells where it keeps the cdi-GMP concentrations low 10 , leading to a bimodal distribution between mother and daughter cell upon division 11 . A similar bimodal distribution of c-di-GMP was described in Pseudomonas aeruginosa 12 , where the PDE Pch is recruited to and activated at the flagellated cell pole via phosphorylation by the chemotaxis histidine kinase CheA. Accordingly, the Pch-induced asymmetry results in low c-di-GMP concentrations of the already flagellated cell promoting active swimming 12 or detachment of surfaceassociated cells, leaving the still attached offspring behind 13 . This process, termed "Touch-Seed-and-Go", balances spreading of the population and demonstrates how heterogeneity in c-di-GMP concentrations may benefit proliferation of the species.
In Shewanella oneidensis and S. putrefaciens, the phosphodiesterase PdeB is a main regulator of the transition from planktonic to sessile lifestyle in response to a yet unknown environmental signal 14,15 . PdeB consists of a putative periplasmic signal receptor domain, flanked by two transmembrane regions, and cytoplasmic section with a HAMP and a PAS domain followed by a GGDEF and an EAL domain (Fig. 1a). The EAL domain is crucial for overall function of PdeB, however, full activity in vivo requires every single domain of the protein 15 . SpPdeB is directly recruited to the flagellated cell pole by the polar landmark protein HubP. HubP/ FimV orthologs occur in a number of different bacterial species (e.g., P. aeruginosa, Vibrio and Shewanella). In these species these polar landmarks coordinate a range of different cellular processes involving, but not restricted to, flagellation, swimming motility, and chemotaxis [16][17][18][19][20][21][22][23] , type IV pili assembly and activity [24][25][26][27][28][29] , and recruitment of the origin of replication to the cell pole 15,16 . The polar localization of PdeB solely depends on its GGDEF (DGC) domain, which directly interacts with the C-terminal part of HubP, the so-called FimV domain 15 . Assembly of a novel PdeB cluster at the pole of the non-flagellated daughter cell normally occurs long after cell fission, suggesting that PdeB induces an asymmetry with respect to c-di-GMP concentrations between the daughter cells in dependence of environmental signals.
Here, we characterized the interaction of PdeB with HubP between the two proteins and mapped the interaction surface between the GGDEF domain of PdeB and FimV of HubP. In vivo analyses strongly indicate that the interaction with HubP at the cell pole is crucial for PdeB function, which therefore is only fully active upon recruitment to the cell pole. This and the difference in abundance of PdeB enhances the heterogeneity of the c-di-GMP concentrations within the population, thereby promoting a diversity of cellular c-di-GMP-dependent processes.

Mapping the FimV HubP /GGDEF PdeB interaction site
Our previous studies showed that in S. putrefaciens the GGDEF domain of PdeB (GGDEF PdeB ) directly interacts with the C-terminal FimV domain of the polar landmark HubP (FimV HubP ) to mediate polar localization of PdeB 15 . However, as full-length protein in concert with a yet unknown environmental signal is required for PdeB to be fully active, it could not be determined by in vitro approaches whether the interaction has functional consequences for the activity of PdeB. Therefore, we set out to identify the critical residues mediating the interaction between GGDEF PdeB and HubP FimV to i) determine how a specific GGDEF domain interacts with FimV and ii) to create a mutant in which PdeB can be uncoupled from FimV to determine the effect of PDE function in vivo. For this, we used a two-pronged complementary approach: As we showed that other S. putrefaciens GGDEFdomain proteins do not interact with FimV 15 , we reasoned that oneidensis PdeB-sfGFP was tested by fluorescence microscopy and is shown as micrographs. The scale bar equals 5 μm. c Analysis of PdeB-sfGFP localization based on fluorescence microscopy shown as scatter plots. A normalized distance to center of 1 indicates that the strongest fluorescence signal was found at the cell pole, while a value of 0 means that the strongest signal was found at the cell center plane. The normalized maximum fluorescence was determined by subtracting the strongest fluorescence value with the mean cellular fluorescence. d SEC interaction studies of MBP-GGDEF SoPdeB with FimV SoHubP . Both proteins form a complex in vitro. e iSCAMS analysis to determine the stochiometry of the MBP-GGDEF SoPdeB /FimV HubP complex. While at high protein concentrations a large complex is present (see d), iSCAMS assays indicate a complex of 2 FimV to 2 GGDEF. f Binding of MANT-GTP to the GGDEF domain of SoPdeB by fluorescence anisotropy measurements. g Residue D216 of VdcA is crucial for DGC activity, as shown by soft-agar motility assay. This residue usually forms a salt bridge with a conserved C-terminal lysine and all PdeB homologs encode glutamic acid instead of aspartic acid at this position. The relative spreading was determined by dividing the mean spreading of the analyzed strain with the mean spreading of the empty vector control strain. Significance was tested by using the unpaired t-test (*P < 0.05, **P < 0.005, ***P < 0.0005). OE, overexpression. GGDEF PdeB has distinctive features that are not present in canonical GGDEF domains and may therefore be involved in mediating the interaction with FimV HubP . In parallel, we directly mapped the interaction surface by in vitro protein studies.

Identification of common features in Shewanella PdeB
Potential orthologs to PdeB are present in numerous different species of Shewanella. To first determine whether polar recruitment is a common feature of Shewanella PdeB, we localized PdeB from S. oneidensis (SoPdeB) 14 , which shares 79% identity to SpPdeB over the full length of the proteins (Supplementary Fig. 1). To this end, we chromosomally fused pdeB to sfgfp, leading to production of stable fluorescently labeled SoPdeB-sfGFP (Supplementary Fig. 2b). As in S. putrefaciens, SoPdeB-sfGFP exclusively localized to the cell pole (Fig. 1b, c). To investigate whether SoPdeB also interacts with FimV HubP , we heterologously produced and purified the GGDEF domain of SoPdeB and found that it forms a complex with both the FimV domains of S. oneidensis (SoFimV) and S. putrefaciens (SpFimV) after size exclusion chromatography (Fig. 1d). According to interferometric scattering mass spectrometry (ISCAMS) analysis the complex had a molecular mass of about 145 kDa (Fig. 1e), which would fit to a 2:2 stoichiometry. Based on this finding we hypothesized that PdeB/HubP interaction via the GGDEF and FimV domains is a common feature in Shewanella, indicating that the residues within the FimV-GGDEF interaction surface are conserved between S. oneidensis and S. putrefaciens and potentially further Shewanella species. Therefore, we identified the most homologous GGDEF domains of 50 putative PdeB orthologs from various species of the genus Shewanella, which we defined as the most similar protein with the same domain organization in each species. The GGDEF domains of these proteins exhibited a high conservation (100-79% identity). The amino-acid sequences were used to create a position-based weight map for GGDEF PdeB , which was then compared to that of bona fide cyclases 30 (Supplementary Fig.  2a). The critical residues required for GTP binding and stabilization of the transition state are present in all PdeB homologs, however, the c-di-GMP-binding inhibitory site (I-site) is degenerated (GxxE in PdeB instead of RxxD in canonical GGDEF domains) and the R" region at the N-terminal region is fully missing in GGDEF PdeB ( Supplementary Fig. 2a). Accordingly, fluorescence polarization assays showed that the purified GGDEF domain of S. putrefaciens was able to bind GTP at a K D value of~16.77 ± 4.33 µM, but not cdi-GMP ( Fig. 1f and Supplementary Fig. 2c). To determine whether GTP binding to the GGDEF domain has a role in PdeB activity, we mutated chromosomal pdeB at the corresponding binding site (PdeB GGAAF ). The c-di-GMP concentrations of the mutant bearing the substitution were significantly higher than those of wild-type cells ( Supplementary Fig. 2d), indicating that GTP binding to the GGDEF domain positively regulates PdeB activity.
As a second unusual feature, GGDEF PdeB possesses a glutamic acid residue at a position (e.g., E467 in S. putrefaciens PdeB; SpPdeB) where canonical GGDEF domains have a conserved aspartic acid. This variation is conserved among all PdeB homologs and may induce structural differences since the conserved aspartic acid usually interacts with a lysine (K578 in SpPdeB) at the far C-terminus of the GGDEF domain. To test if the E467 represses a DGC activity in SpPdeB and other GGDEF domains, we introduced a reversing D to E substitution into the DGC VdcA from Vibrio cholerae (VdcA D216E ). This protein was chosen as a canonical DGC, as it has previously been shown to be active in Shewanella 31 . VdcA or VdcA D216E were ectopically produced in S. putrefaciens, and a decrease in flagella-mediated spreading through soft agar was scored as a measure for DGC activity. While production of wild-type VdcA limits flagellamediated spreading of S. putrefaciens through soft agar as expected, the overproduction of the VdcA D216E variant had a significantly smaller effect (Fig. 1g). We thus concluded that the E467 variation negatively affects the enzymatic activity of the GGDEF domain of VdcA and, thus, may be a factor for the inactivity of GGDEF PdeB as DGC in vivo and in vitro 14,15 .

Identification and prediction of the FimV HubP /GGDEF PdeB interaction site
To determine the protein interaction surface between GGDEF PdeB and FimV HubP , the GGDEF/FimV complex from S. putrefaciens was purified and analyzed by hydrogen-deuterium exchange mass spectroscopy (HDX-MS) (Fig. 2a). The assay revealed that residues 1047-1107 of HubP exhibited reduced deuterium content in the presence of PdeB GGDEF , indicating that this region serves as the docking site for PdeB. However, no HDX result could be obtained for PdeB GGDEF . As a complementary approach, we therefore applied cross link-mass spectroscopy (CL-MS) to analyze the protein complex of FimV HubP /GGDEF PdeB . This assay was conducted using S. oneidensis FimV HubP /GGDEF PdeB , which could be produced and purified at higher yields. We found that a peptide encompassing the residues 1053-1088 of SoHubP was chemically crosslinked to the peptide encompassing the residues 566-577 of SoPdeB ( Fig. 2b and Supplementary Fig. 3a). The peptide identified by CL-MS in FimV HubP in S. oneidensis matched the region localized by HDX in S. putrefaciens FimV HubP -GGDEF PdeB (Fig. 2b).
The data obtained by the HDX-and CL-MS approaches allowed a definition of the interaction surface between both proteins. The structure of Shewanella FimV HubP , predicted based on that of the highly homologous FimVc-domain from Pseudomonas aeruginosa 26 , showed a distinct negative surface charge of the FimV domain at the putative docking site of GGDEF PdeB (Fig. 2c, d). We therefore assumed that the interaction surface of the domain includes critical positively charged residues. Correspondingly, the GGDEF PdeB peptide that was identified by CL-MS to interact with FimV HubP harbors two lysines, K575 and K577 (see Supplementary  Fig. 2a), which are fully conserved in all Shewanella PdeB proteins. K577 PdeB is variable in canonical GGDEF-domains, while K575 PdeB is a conserved lysine that interacts with a conserved aspartic acid at the N-terminus of the GGDEF motif and is therefore functionally relevant 30 . To identify additional residues potentially involved in polar recruitment of PdeB, we used the position-based weight map of PdeB homologs to screen for further surface-exposed, positively charged amino acids, which are not present in other, canonical GGDEF domains. According to these assumptions, the residue pair K524 PdeB and Q525 PdeB turned out as potential candidate, which in canonical GGDEF domains consists of a negatively charged glutamic acid followed by an arginine. While K524 PdeB is fully conserved in all PdeB homologs, the following residue Q525 PdeB is variable but always a large hydrophilic amino acid that is not negatively charged (Fig. S2a). As, according to structural predictions, both residues are near the degenerated inhibitory site (I-site; Fig. 2e) of GGDEF PdeB , we further included residues within this site to map the FimV/GGDEF interaction surface. Thus, the HDX-, crosslinking and in silico approaches identified several amino-acid residues in GGDEF PdeB that are putatively involved in mediating the interaction with HubP FimV .
The far C-terminal region and degenerated I-site of GGDEF PdeB mediate interaction with HubP FimVc As the next step, we wished to determine whether the prediction of the GGDEF PdeB /FimV HubP interaction surface was correct and performed in vitro interaction studies. Owing to the better purification properties of S. oneidensis GGDEF PdeB and the conserved interaction between GGDEF PdeB and FimV HubP from S. putrefaciens and S. oneidensis, the in vivo studies were conducted using S. oneidensis GGDEF PdeB . To this end, mutant variants of SoGGDEF PdeB bearing amino-acid substitutions at the desired . The sequences of SpHubP and SoHubP were aligned using clustal omega and JalView. c The structure of the C-terminal domain of SpHubP was predicted using the swiss-model algorithm and the peptide found by CL-MS is colored in orange. d The electrostatic surface potential was calculated using the APBS electrostatic tool and pymol. Red coloring indicates a negative surface charge. e The structure of the PAS-GGDEF-EAL region of PdeB was predicted using Phyre2. The N-terminal PAS domain is colored in gray, the GGDEF domain in green and the C-terminal EAL domain in light orange. Residues within the GGDEF domain that we identified to be involved in the PdeB-HubP interaction are colored in cyan. f The dissociation constant for the SoPdeB-SoHubP interaction was determined using BLI assays. The specific binding in nanometer was determined by subtracting the unspecific binding of the analyte in absence of the ligand FimV. g-k Residues that are involved in respective interaction were substituted as indicated and the K D value was determined using BLI assays. All mutations decreased the affinity of SoGGDEF to SoFimV.
T. Rick et al. position were purified. To investigate if K575 and K577 in GGDEF PdeB are involved in mediating interaction to FimV HubP , we used variants in which the appropriate lysine residues were substituted with a serine (K575S and K577S). A potential effect of the residues within or close to the degenerated I-site of GGDEF PdeB was determined by introducing a series of single substitutions (G494A; E497S; Q525S; K524S) and a K524E Q525S double substitution into the protein. To verify structural integrity and functionality of the mutated proteins, we showed that all SoPdeB GGDEF variants elute at the same volume after size exclusion chromatography (SEC) and still bind MANT-GTP with roughly the same affinity as wild-type SoPdeB GGDEF (Supplementary Figs. 3b and 4a-i). Interaction of the wild-type and mutant proteins to SoFimV HubP was then determined by bio-layer interferometry (BLI) (Fig. 2f-k and Supplementary Fig. 3c, d). Wild-type GGDEF PdeB bound to FimV HubP with high affinity at a K D value of around 200 nM (Fig. 2f). Compared to the wild-type version, both substitutions in the conserved lysine residues strongly decreased the affinity to FimVc and resulted in a 100-fold (K575S) and a 10fold (K577S) increased K D -value (Fig. 2g, h). In addition, all substitutions targeting the region around the I-site negatively affected the affinity of the protein variant towards FimV to various extents ( Fig. 2i-k and Supplementary Fig. 3c, d). The SoGG-DEF K527E-Q528S double substitution even completely suppressed FimV HubP -GGDEF PdeB interaction in vitro (Fig. 2i). Determination of the K D value for the FimV-GGDEFK524S variant by BLI assays was not possible due to slight aggregation of the latter protein ( Supplementary Fig. 3e). However, pull-down assays suggest that also this variant does not effectively bind FimV ( Supplementary   Fig. 3f). Taken together, HDX, crosslinking and mutational analysis allowed us to predict the interaction surface between GGDEF PdeB and FimV HubP and to create mutants in which the interaction is diminished.

Mutations in the GGDEF-FimV interaction surface alter PdeB localization dynamics
To further determine whether the decreased affinity between several GGDEF PdeB variants and FimV HubP affects polar recruitment of PdeB by HubP, we performed in vivo localization studies by fluorescence microscopy. To this end, the appropriate mutations were introduced into the chromosome of a S. putrefaciens strain producing functional fluorescently labeled PdeB-sfGFP 15 . All substitutions in GGDEF PdeB that negatively affected the interaction with FimV HubP in vitro were stably produced at normal levels ( Supplementary Fig. 5a-c) but exhibited a significant decrease in polar localization in vivo ( Fig. 3a-d and Supplementary Figs. 5d, e and 6a-f). The PdeB variants bearing the G497A and K527E/Q528S substitutions within the degenerated I-site and the K578S substitution almost completely lost polar localization. To analyze the localization dynamics of PdeB in more detail, we monitored the intracellular diffusion of the protein in vivo using singlemolecule tracking (SMT) and determined the average time PdeB spent in static/slow moving (interacting with FimV HubP ) or an unbound, freely diffusive mode. To this end, the C-terminal sfGFP tag to PdeB was replaced by mVenus. We observed that about 84% of the observed molecules were statically confined to the cell pole leaving only about 16% of the molecules that were freely Fig. 3 Localization behavior of SpPdeB-sfGFP variants. a-d Residues that were shown to decrease the affinity of PdeB to HubP were genomically substituted in S. putrefaciens pdeB-sfgfp. The localization was quantified using BacStalk and is shown in d scatterplots where each dot represents an individual cell. A normalized distance to center of 1 indicates that the strongest fluorescence signal was found at the cell pole, while a value of 0 means that the strongest signal was found at the cell center plane. The normalized maximum fluorescence was determined by subtracting the strongest fluorescence value with the mean cellular fluorescence. Mutated proteins showed decreased polar localization compared to the wild-type version. e-h The diffusion rates of mutants with mutations in the active center of the GGDEF and EAL domain, as well as in the interaction surface with HubP (h) were determined using single-molecule microscopy of PdeB-mVenus. Shown are projections of tracks into a standardized cell of 1 by 3 μm size. Blue trajectories represent freely diffusive molecules, red tracks molecules in confined motion (molecules staying within a radius of 120 nm, equal to three times the localization error), and green tracks represent molecules showing mixed behavior, i.e., switching between diffusive and confined motion. i Bubble blot showing size and diffusion coefficients of the two determined populations, the fast-and the slow-diffusing population, for each tested strain as indicated below. For a better comparison between PdeB and the mutant versions, diffusion constants were determined to best fit to all four proteins (0.095 μm² s −1 ± 0.001 for the slow population, 0.22 μm² s −1 ± 0 for the fast population).  Fig. 5f). In a mutant bearing the K527E/Q528S substitution within GGDEF PdeB the ratio shifted towards the freely diffusing subpopulation (61% static, 39% diffusing) (Fig. 3h, i and Supplementary Fig. 5i). To further determine if the PdeB localization dynamic is affected by GTP binding (by the GGDEF domain) or the PDE activity (conferred by the EAL domain), we also determined the protein diffusion of PdeB-mVenus variants in which the GGDEF or the EAL domain were mutated by residue substitutions (GGAAF and AAL, respectively). The resultant PdeB-mVenus mutant versions exhibited only minor differences in diffusion compared to that of nonmutated PdeB-mVenus ( Fig. 3f, g, i and Supplementary Fig. 5g, h). Thus, PdeB is tightly bound to FimV HubP via its GGDEF domain and therefore is almost exclusively present at the cell pole, and this localization pattern is unaffected by the enzymatic activity of PdeB.

Substitutions within the GGDEF PdeB interaction surface to FimV HubP affect PdeB in vivo activity
The above results suggested that the interaction surface between GGDEF FimV and FimV HubP involves residues with a putative mechanistic function in the GGDEF domain, e.g., the salt bridgeforming lysine residue (K578) or the degenerated I-site (K527Q528; G497). We therefore hypothesized that interaction of the GGDEF domain of PdeB with FimV may result in differences in in vitro function of PdeB. To test this, we determined the intracellular c-di-GMP concentrations of S. putrefaciens wild-type cells and cells bearing substitutions in the residues in the GGDEF domain of PdeB that are critical for HubP/FimV interaction. We found that all substitutions resulted in significantly elevated c-di-GMP concentrations like that of a mutant completely lacking pdeB (Fig. 4a).
To verify these findings at the single cell level, we adopted a plasmid-based c-di-GMP-responsive fluorescent reporter system for Shewanella, which relies on three c-di-GMP-dependent riboswitches mediating the production of destabilized turboRFP AAV 32,33 . Additionally, a second fluorophore is constitutively produced that allows normalization of the c-di-GMP-derived signal to the plasmid copy number ( Supplementary Fig. 7a, b). In accordance to our in vitro results, all cells lacking pdeB or producing PdeB mutated in the GGDEF/FimV interaction surface displayed a higher concentration of c-di-GMP than wild-type cells (Fig. 4b). The results demonstrate that the residues within the GGDEF/FimV interaction surface are critical for in vivo function of PdeB and strongly suggest that interaction of PdeB with HubP at the cell pole stimulates PDE activity.
PdeB promotes heterogeneity in c-di-GMP concentrations As shown above, the c-di-GMP-responsive reporter system revealed that the S. putrefaciens wild type exhibits a, on average, lower level of c-di-GMP than mutants lacking or being mutated in PdeB. Of note, we also observed a broader distribution of c-di-GMP levels within the population, including a subpopulation with a very low c-di-GMP content (Fig. 5a). Thus, the presence and activity of PdeB leads to an increased heterogeneity of c-di-GMP levels of the cells within a population. We reasoned that part of the effect likely stems from the bimodal distribution of FimV HubP during cell division 15 . In addition, during single-molecule tracking we observed a very low number of PdeB-mVenus tracks, and many cells completely lacked fluorescent signals. To determine the number of PdeB-mVenus molecules per cell, we quantified the photon count of single bleaching steps (i.e., single-fluorescentprotein bleaching) toward the end of the signal acquisitions. The total fluorescence intensity at the beginning of the acquisition (not yet bleaching) was then divided by that of the single fluorophores, relative to the cell size 6 . By this, we found that about 25% of the cells did not contain any PdeB molecule. The residual cells harbored one to forty PdeB molecules, with a major population possessing one to seven copies ( Fig. 5b and Supplementary Fig. 8). Thus, in addition to the spatial distribution of active PdeB molecules per cell occurring during cell division, further heterogeneity in c-di-GMP levels is mediated by a wide array of PdeB molecule numbers.
Previous studies on P. aeruginosa showed that an asymmetric division of surface-associated cells results in one daughter cells prone to stay attached due to high concentrations of c-di-GMP and one flagellated daughter cell with low cdi-GMP levels, which is more prone to detach and leave. This behavior was termed "Touch-Seed-and-Go" 13 . The strict polar localization of PdeB and its regulation of surface-adhesion factors in S. putrefaciens strongly suggested that PdeB promotes a similar behavior. We therefore monitored and quantified the detachment behavior of dividing S. putrefaciens cells adhered to a surface. To distinguish between the daughter cell with or without PdeB, we used a strain in which PdeB was C-terminally labeled by sfGFP. We observed in about 80% of the division events that one cell detaches after cell fission. When a PdeB-sfGFP cluster was visible in one cell (in about 50% of the events), it was always this cell detaching Fig. 4 HubP affects the activity of PdeB in vivo. a The intracellular c-di-GMP content of strains with reduced PdeB localization was extracted and quantified by mass spectrometry. Strains with reduced FimV/GGDEF interaction and polar localization of PdeB showed significantly increased c-di-GMP concentration. Significance was tested by using the unpaired t-test (*P < 0.05, **P < 0.005). b The c-di-GMP level of strains with reduced PdeB localization was determined using a fluorescence microscopy-based reporter assay. The tested mutants showed increased cellular c-di-GMP levels, indicating that polar localization of PdeB is required for full PDE activity.
( Fig. 5c, d). In a strain in which the PDE activity of PdeB-sfGFP was disrupted by introduction of a mutation within the EAL active center (PdeB AAL -sfGFP) a much smaller fraction of cells detached after finishing cell division (about 20%). Among these detaching cells, only about a 30% had a visible PdeB AAL -sfGFP cluster. A mutant strain with strongly decreased GGDEF PdeB and FimV HubP interaction (GGDEF PdeB K527E Q528S) exhibited a highly similar phenotype to PdeB AAL -sfGFP with a minority of cells (28%) detaching after division.
Taken together, our data shows that, in S. putrefaciens, abundance and localization of PdeB promote heterogeneity in cellular c-di-GMP concentration within the population including an asymmetry between dividing cells. This facilitates "Touch-Seedand-Go" behavior and thus regulates the spreading and proliferation of the population.
PdeB regulates the abundance and activity of major surfaceadhesion factors in S. putrefaciens Our results show that the activity of PdeB governs S. putrefaciens surface adhesion. However, so far it is unknown which surfaceadhesion factors are directly or indirectly regulated by PdeB. We therefore determined, whether major surface-adhesion factors of S. putrefaciens, the surface-adhesion protein BpfA 34-36 and the mannose-sensitive hemagglutinin (MSHA) type IV pilus 37,38 , are regulated by PdeB.
In the Vibrio cholerae MSHA pilus system, filament extension and retraction can be modulated by direct binding of c-di-GMP to the pilus extension ATPase MshE, which affects near-surface swimming and motile-to-sessile transition [39][40][41] . Amino-acid sequence alignments indicated the presence of the conserved cdi-GMP-binding motif at the N-terminal domain of S. putrefaciens MshE (Supplementary Fig. 9a). To determine potential c-di-GMP binding of that region, full-length MshE as well as its predicted cdi-GMP-binding domain MshE-N (aa 2-145) from S. putrefaciens were purified (Supplementary Fig. 10). Using isothermal titration calorimetry (ITC) and fluorescence polarization (FP) we found that both MshE and MshE-N bind c-di-GMP with high affinity (Fig. 6a,  b). This data indicated that PdeB regulates MSHA pili extension via c-di-GMP concentrations. Therefore, we directly determined MSHA pili formation using a serine to cysteine substitution (S68C) within the major subunit of the pilus filaments MshA. The cysteine residue allowed covalent coupling of maleimide-ligated fluorophors to extended pili 41 and subsequent visualization and quantification of MSHA pilus formation by fluorescence microscopy (Fig. 6c). Compared to wild-type cells, mutants lacking pdeB displayed a significantly higher amount of extended MSHA pili (Fig. 6d). In addition, cells lacking MSHA pili showed increased The number of PdeB-mVenus molecules in S. putrefaciens was measured using single-molecule microscopy and is shown as histogram. Around 25% of the observed cells do not harbor PdeB-mVenus molecules. c, d Polar localization leads to phenotypic heterogeneity and to the formation of a motile mother cell and a sessile daughter cell. Two-hundred sixty-seven cell divisions were observed for each strain and the results are shown in stacked bar plots (c). In the presence of polarly localizing, enzymatically active PdeB, a motile mother cell and a sessile daughter cell is created. An exemplary time lapse is shown in d, the scale bars equal 10 μm. The significance was tested by using the fishers test of independence (*P < 0.05, **P < 0.005, ***P < 0.0005).
To investigate a potential effect of PdeB on the S. putrefaciens Bpf surface-adhesion protein system, we determined the abundance of proteins BpfA and AggC (Sputcn32_3592) in the presence or absence of PdeB. To this end, the corresponding genes on the chromosome were mutated to encode protein versions with a C-terminal FLAG-tag, which allowed the quantification of the protein levels by Western blotting. All tested proteins showed a significantly higher abundance in the absence of PdeB (Fig. 6f).
The results demonstrate that PdeB regulates the presence and activity of major c-di-GMP-dependent surface-adhesion factors, i.e. the MshA pilus and the Bpf surface protein system, and by this governs cell-surface interactions, which in turn affects flagella-mediated swimming and spreading through structured environments.

DISCUSSION
Previous studies have provided evidence that the polar activity of enzymes involved in c-di-GMP synthesis and turnover affects important cellular processes, such as cell cycle regulation in C. crescentus 9 or mediating the dynamics of a surface-associated or planktonic lifestyle in P. aeruginosa [11][12][13] . In these systems, bimodal distribution of c-di-GMP during cell division is usually controlled by abundance (e.g., through spatiotemporally regulated degradation) or specific activation of the corresponding proteins (e.g., through phosphorylation) 1,12 .
Here, we find that the S. putrefaciens GGDEF PdeB domain serves as a spatial on-switch regulator for PdeB upon direct interaction with the FimV domain of the polar marker protein HubP, thereby restricting full PdeB activity to the flagellated cell pole. c, d The number of extended MSHA pili in presence and absence of pdeB were quantified using fluorescence microscopy. The scale bar represents 5 μm. Significantly more extended MSHA pili were observed in the absence of pdeB, as shown in the histogram (d). Significance was tested using the chi-squared test (P < 0.005). e The influence of MshE on the spreading phenotype of ΔpdeB was tested using soft-agar motility assays. Since we previously showed that PdeB regulates the lateral flagellar system (15), these experiments were done in strains lacking the genes for the lateral flagellins (ΔL). The values were normalized to the ΔL strain. Significance was tested by using the unpaired t-test (*P < 0.05, ****P < 0.00005). f Absence of pdeB also leads to increased amounts of Bpf proteins, as shown by immunoblot assays. The amount of two different proteins, BpfA and BpfD, were determined in the presence and absence of pdeB. Note that two unrelated lanes were removed, the original Western blot is added as Supplementary Fig. 12.
PdeB function depends on interaction with FimV As in PdeB, in about two thirds of the proteins containing an EAL domain, this domain occurs as a GGDEF-EAL pair at the C-terminal end of the corresponding protein 42 . In contrast to some of these hybrid proteins [43][44][45] , PdeB does not function as a DGC in vivo or in vitro under the conditions tested. This is at least partly due to a glutamate residue conserved in PdeB orthologs (E467 in SpPdeB) replacing a non-variable functionally crucial aspartate in canonical GGDEF domains 30 . We therefore conclude that the GGDEF domain of PdeB solely acts as a localizing and regulating domain.
Despite its inactivity as a DGC, GGDEF PdeB is still able to bind GTP to its active center, which is crucial for PDE function. Previous work on different DGC-PDE hybrids (e.g., Caulobacter crescentus CC3396, P. aeruginosa RbdA and RmcA) demonstrated that GTP binding to an otherwise inactive GGDEF domain can regulate the downstream EAL domain's PDE activity [46][47][48][49] . Notably, the domain organization and structure prediction of PdeB closely resembles that of RbdA (Fig. 2e). Recent studies on RbdA indicate that GTP binding results in structural rearrangements of an RbdA dimer. These rearrangements release autoinhibitory interactions between the GGDEF and the EAL domains, which are then capable of binding and hydrolyzing c-di-GMP 48 . Similarly, GTP binding to the GGDEF domains of an RmcA dimer is proposed to unlock and activate the EAL domains 49 . The unlocking mechanism is based on a rearrangement of an alpha helix (hinge helix) that connects the DGC and PDE domains, which is also predicted to be present in PdeB. Based on the structural similarities, we hypothesize that PdeB phosphodiesterase activity is allosterically regulated by GTP in a similar fashion.
In addition to GTP binding, PdeB function in vivo requires the interaction with the C-terminal FimV domain of the polar landmark protein HubP. We found that tight interaction of the two proteins occurs between a negatively charged surface area of FimV and a positively charged surface area of the GGDEF PdeB domain, which also includes the degenerated inhibitory (I-)site of the latter. Substitutions within these residues drastically lower the affinity between FimV and GGDEF PdeB and shut down in vivo PDE function without affecting GTP binding. In functional dimers of canonical DGCs, c-di-GMP can bind to the I-site of one GGDEF domain and a second inhibitory site, the N-terminal R" site, of the second GGDEF domain. This locks the dimer in an inactive conformation in an end-product inhibition 30 . However, as the R" site is absent in all PdeB orthologs, we hypothesize that FimV binding to GGDEF PdeB does not immobilize the domains in the dimer but rather induces conformational changes that promote the formation of a complex competent for activation of the C-terminal EAL domains. Further structural studies are required to elucidate the exact underlying mechanism. Notably, the DGC DgcP Recently was shown to be directly recruited to the cell pole in P. aeruginosa by FimV, the orthologous polar landmark protein to HubP in Pseudomonas 50 . DgcP affects surface-adhesion/biofilm formation and twitching motility in P. aeruginosa and the DGC activity was decreased in cells lacking FimV. Thus, as PdeB, DgcP may also be activated through direct interaction with a polar marker protein, but so far direct evidence for this lacking. The GGDEF domains of DgcP and PdeB are not conserved in the critical residues required for PdeB recruitment and activation, which should be expected as both proteins possess opposite enzymatic activities.
In cells of S. putrefaciens and S. oneidensis PdeB mediates a considerably increased heterogeneity with respect to c-di-GMP concentrations. This heterogeneity can be explained by at least two closely connected factors, (i) the strictly monopolar PDE activity in concert with a non-uniform appearance of PdeB at the new cell pole and (ii) the wide range of a comparably small number of PdeB copies per cell. Thus, by regulating flagella-mediated motility and at least two Shewanella c-di-GMP-responsive surface-adhesion factors PdeB promotes a wide range of cell behaviors with respect to spreading and attachment in each population. It is yet unknown how production of Shewanella PdeB is regulated to result in such a pronounced heterogeneity in abundance and timing of appearance after cell division.

PdeB regulates the transition between sessile and planktonic lifestyle
As for many bacterial species, c-di-GMP-mediated signaling is crucial for regulating the switch from sessile to planktonic lifestyle in Shewanella sp., and PdeB has an important role in balancing planktonic and surface-associated states of the cells. In S. putrefaciens, PdeB stimulates the formation and activity of the lateral flagella system, which positively affects swimming and spreading of the cells 15,51,52 . The effect on the lateral flagella is likely mediated by the c-di-GMP-responsive transcriptional flagellar regulator FlrA 2 and the c-di-GMP-binding flagellar motor regulator MotL 15,53,54 . Here, we demonstrated that PdeB also negatively regulates abundance and activity of two major factors that directly mediate cell-surface interactions, the surfaceadhesion protein BpfA and the mannose-sensitive hemagglutinin (MSHA) type IV pilus. The MSHA pilus system has previously been demonstrated to be necessary to initiate cell-surface interactions of S. oneidensis 37,38 and to affect the ability of this species to colonize the Zebrafish gut 55 . Here, we found that the MSHA pili negatively affect the spreading of S. putrefaciens through soft agar, which accounts for a large part of the PdeB-related spreading phenotype of the polar flagella of S. putrefaciens and likely also in S. oneidensis. In Vibrio cholerae, the pilus extension ATPase, MshE, is stimulated by binding c-di-GMP 40,56 , and active VcMSHA pili also limit cell spreading through polysaccharides and assist in initiating surface adhesion by actively swimming cells 39,41 . Our results strongly suggest that the S. putrefaciens MSHA pilus system functions in a similar fashion. The BpfA surface protein and the corresponding transport and control system of Shewanella are largely homologous to the well-characterized Lap system of Pseudomonas fluorescens 57 and have an important role in mediating cell-surface interactions 34,36 . In S. putrefaciens, BpfA production and cell-surface abundance are regulated by c-di-GMP concentrations, e.g., via the flagellar master regulator FlrA 1 and the oxygen-responsive DGC DosD 35,53 ; our results demonstrate additional regulation of BpfA abundance by PdeB. Notably, both LapA and VcMSHA pili are thought to be involved in mediating cdi-GMP-dependent attachment and detachment of cells to a substratum 41,57 , which we now propose to similarly occur for the corresponding Shewanella systems.
The signal PdeB is responding to is yet unknown. In P. aeruginosa the PDE Pch is recruited by the main chemotaxis histidine kinase CheA, which in addition governs Pch activity by phosphorylation 12 . By this, P. aeruginosa elegantly links the perception of a wide range of external signals via the chemotaxis sensor array to cellular behavior in planktonic as well as in surfaceassociated cells. The asymmetry of c-di-GMP levels in dividing cells mediated by Pch also affects the surface attachment/detachment response of P. aeruginosa, which depends on polar recruitment of the c-di-GMP-responsive pilus regulator FimW upon surface association 13 . In S. putrefaciens and S. oneidensis, PdeB appears to be constantly but heterologously produced irrespective of the media conditions 15 , and polar recruitment of PdeB occurs independently of surface attachment. However, PdeB appears to be solely active in complex media 14,15 . Our data strongly suggests that PdeB activation requires GTP binding and may therefore be dependent on the cellular GTP concentrations. As previously discussed, the cellular GTP level is correlated with that of the alarmone (p)ppGpp and thus also reflects the cellular nutrient levels, so that PdeB activity is shut down under conditions of low nutrients 46,48 . In addition, PdeB possesses two putative sensor domains, the N-terminal periplasmic domain and the cytoplasmic PAS domain. It has been suggested that sensor PDE proteins (e.g., RmcA and RbdA) may function as rheostats that integrate several signals to tune the proteins activity 48,49 similar to the sum of different signals transmitted by the chemotaxis array to Pch of P. aeruginosa via CheA. The periplasmic domain of PdeB has no striking homology to other sensing domains but is required for full PDE function 15 . The domain's localization hints at an external signal, such as a small molecule or peptide. Notably, compared to the remaining protein, this hypothetical sensing domain is less well conserved among Shewanella sp. As members of this genus occur in wide range of different environments 58 , it is conceivable that Shewanella PdeB orthologs have evolved to respond to different environmental signals. The second potential sensing domain of PdeB, the cytoplasmic PAS domain, is also required for full PdeB function 15 , and PAS domains could be involved in the perception of a wide range of potential signals 59 . As PdeB orchestrates population heterogeneity and behavior, ongoing current studies are directed at identifying the yet elusive signal(s).

Plasmid and strain constructions
Plasmids are listed in Supplementary Table 2 and were generated essentially as previously described 15 . The used oligonucleotides are listed in Supplementary Table 3. Plasmids were constructed by PCR amplification of the DNA fragments of interest and subsequent Gibson assembly with the appropriate vector backbone 60 . All plasmids were verified by sequencing. To modify the genome of Shewanella species flanking regions of 500 to 750 nucleotides upstream and downstream of the integration site were amplified by PCR and inserted into the suicide vector pNTPS-138-R6K 61 . The resulting plasmids were transferred into Shewanella from E. coli WM3064 by conjugation. Correctly mutated strains after sequential homologous recombination were identified by plating the cells on either kanamycin-or sucrose-containing LB medium plates. Successful integration or deletion was verified using colony PCR.

Soft-agar motility assay
Overnight LB cultures of the appropriate strains were diluted to an OD 600 of 0.02 and grown to mid exponential growth phase (OD 600 of 0.5). Two microliters of respective culture were dropped on LB soft-agar plates (0.25% (w/v) agar). The plates were incubated at room temperature overnight. The plates were scanned and the swim radius was quantified. Cultures of strains to be directly compared were always placed on the same plate.

Fluorescence microscopy
The strains to be imaged were cultivated overnight in appropriate medium and then subcultivated until reaching exponential growth phase (OD 600 of 0.2) or another desired OD 600 . Three microliters of culture were spotted on an agarose pad (LM medium solidified with 1% (w/v) agarose). Fluorescence images were recorded using a Leica DMI 6000 B inverse microscope (Leica, Wetzlar, Germany) equipped with an sCMOS camera and an HCX PL APO 100×/1.4-numerical-aperture objective using VisiView software (Visitron Systems, Puchheim, Germany). The obtained images were further processed using ImageJ 62 (https://imagej.nih.gov) to adjust contrast, to change grayscale to color, to define representative areas for display, and to add scale bars. Adobe Illustrator CS6 was used to add an outline of cells where required and to create the final assembly of panels. Cells from at least three independent cultures were imaged.

Fluorescent labeling of MSHA-pili
To fluorescently label MSHA type IV pili, we adopted pili labeling protocols for Vibrio cholerae 41,63 and flagella labeling in S. putrefaciens 64 . Overnight cultures of the desired strains in LB were diluted to an OD 600 of 0.02 and re-cultivated to the desired OD 600 of 0.2. Cells were harvested by centrifuging 100 µl of each culture at 1,500 × g for 5 min at room temperature. The pellets were washed with 50 µl phosphate-buffered saline (PBS; NaCl, 137 mM; KCl 2.7 mM; Na 2 HPO 4 , 10 mM; KH 2 PO 4 , 1.8 mM; pH 7.4). In all, 1.5 µl CF®488A Maleimide was added to reach final concentration of 25 µg ml −1 . The cells were incubated for 15 min in darkness to prevent bleaching. Afterwards the cells were again sedimented by centrifugation at 1500 × g for 5 min at room temperature and washed with 100 µl PBS. The washing step was repeated and the cells were resuspended in 100 µl LM medium. Three microliters of the stained culture was dropped on a 1% (w/v) LM-agar pad for microscopy.

Quantification of polar PdeB-sfGFP variants by fluorescence microscopy
To quantify the signal intensity of polarly localized PdeB-sfGFP variants, overnight cultures were subcultivated as described above and grown at 30°C until OD 600 of 0.5 was reached. Aliquots were then used for microscopy as described. The obtained images were analyzed using the BacStalk software 65 to obtain data for the cell length, mean fluorescence, maximum fluorescence and distance of the maximum fluorescence to the cell center. Following, at least 300 cells were selected at random and a new parameter was calculated called "normalized distance to center" by using following formula: Norm:distance to center ¼ distance of maximum fluorescence to cell center cell length=2 ð Þ Norm: max fluorescence ¼ maximum fluorescence À mean fluorescence The cluster intensity was normalized by subtracting the mean fluorescence from the maximum fluorescence. Both values were then plotted against each other where each point represents an individual cell. The color of each point represents the cell length of the respective cell.

c-di-GMP in vivo fluorescence reporter assay
The c-di-GMP reporter plasmid pMMB-Gm-Bc3-5 AAV was kindly provided by the group of Fitnat Yildiz 33 . The plasmid was introduced in Shewanella via conjugation. Overnight cultures supplemented with 5 µg ml −1 gentamycin were diluted to an OD 600 of 0.02. Cultures were grown to approximately OD 600 0.6, diluted and 3 µl were dropped on LM-agar pads. Pictures with exposure time of 100 ms were taken using the microscopic set-up described above. The relative fluorescence intensity (RFI) was then determined by obtaining the ratio of turboRFP/AmCyan as quantification of the amount of intracellular c-di-GMP. The fluorescence images were analyzed using BacStalk 65 and parameters such as cell length, maximum fluorescence intensity, mean fluorescence intensity, and distance of maximum fluorescence to cell center were obtained. For downstream analysis we selected at least 500 cells by random with a cell length between 2 and 6 µm. The obtained data was then plotted using the programming environment R and the software Rstudio (Rstudio Inc.). To verify that the reporter system is suitable for the used Shewanella species, we plotted the red fluorescence intensity against the blue fluorescence intensity and determined the expected linear correlation ( Supplementary  Fig. 6). The quotient of red fluorescence divided by blue fluorescence was used as parameter for the intracellular c-di-GMP level and was plotted against the cell length as additional control (Supplementary Fig. 6).

Single-molecule tracking
Cultures were grown to an OD 600 of 0.5-0.6 in LB medium. Five microliters of the culture were placed on a cover slide and cells were immobilized with an agar pad (1% (w/v) agarose in LB). The microscope used for the single-molecule tracking was a Nikon Eclipse Ti microscope (100x oilimmersion objective, NA = 1.49). Cells were imaged with the central part of a 514 nm-laser (TOPTICA Beam Smart, maximum power = 100 mW), using about 160 to 200 W cm −2 ; single-molecule level was reached rapidly after 100 frames. Movies with 2000 frames were taken with an EM-CCD camera (ImageEM X2 EM-CCD, Hamamatsu). Movies were cropped so that the first 100 frames were removed, performed in Fiji software 66 . Cell outlines were determined by the software Oufti 67 , and afterwards trajectories were determined with the software UTrack 2.2.1 68 . SMTracker 1.5 6 was used for all further the analyses of protein dynamics. Diffusion constants were determined using squared displacement (SQD)-analyses. Trajectories were categorized into confined motion, for which a radius of 120 nm was set, or as freely diffusive, or having mixed behavior.

Fluorescence-based quantification of molecule number
For molecule quantification, movies of PdeB-mVenus-producing cells, or of cells lacking any mVenus fusion (for determination of auto-fluorescence), were acquired in the same medium, on the same day. Eleven complete movies (including the first 100 frames) of each strain were taken and for each strain; one video of the empty slide was captured for background correction. The analysis was carried out as described in SMTracker 2.0 (see https://sourceforge.net/projects/singlemoleculetracker/) was used for the determination. Initial fluorescence in cells was divided by determined single bleaching steps of PdeB-Venus, to calculate numbers of fluorophores per cell (for which the length is known from Oufti software, see above) and thereby average molecule number.

Protein expression and purification
The plasmid pET24c was used for heterologous protein expression in E. coli BL21(DE3) (New England Biolabs, Frankfurt, Germany). Cells were grown in LB medium supplemented with 50 µg ml −1 kanamacin to an OD 600 of 0.75 at 37°C and vigorous shaking. Prior to induction with D-(+)-lactosemonohydrate (12.5 g l −1 ) cells were chilled for 10 min in an ice bath and expression was carried out overnight at 16°C under vigorous shaking. The cells were harvested the next day by centrifugation (4400 × g, 10 min, 4°C) and the resulting cell pellet was stored at -20°C until further use. Frozen cell pellets were resuspended in the appropriate buffer required for downstream applications supplemented with 0.2 % Tween20 and 40 mM imidazole and lysed by sonification (Bandelin Sonoplus). The resulting lysate was centrifuged for 30 min at 47,000 × g to remove cell debris and the supernatant was filtered with a 0.45 µm filter before loading onto an equilibrated 5 ml His Trap HP column (GE healthcare) connected to an ÄKTA pure 25 system. Unspecifically bound proteins were removed by washing with 10 column volumes (CV) lysis buffer and proteins were eluted with a linear gradient of buffer supplemented with 600 mM imidazole. Eluted proteins were further purified and analyzed by size exclusion chromatography (SEC). The HiLoad Superdex 200 pg column (GE Healthcare, Chicago) was used for preparative runs while the Supedex200 increase column (GE Healthcare, Chicago) was used for analytical and semipreparative runs. Respective SEC columns were equilibrated with the appropriate assay buffer required for downstream applications without other additives. Proteins were isocratically eluted and the concentration was determined using a spectrophotometer (NanoDrop Lite, Thermo Scientific) before snap freezing in liquid nitrogen and storage at -80°C. Protein size was analyzed using a calibration curve of proteins included in the high molecular weight calibration kit (GE Healthcare, Chicago). To test frozen samples for aggregation, SEC was used as described above. To purify protein complexes, both interaction partners were purified using affinity chromatography. Fractions containing the two proteins of interest were then mixed and incubated for at least 1 h at 4°C. The sample was then further purified using SEC as described above to remove unbound proteins.

Protein pull-down assays
The interaction of GGDEF PdeB variants with FimV HubP were tested using amylose resin-based pull-down assays. The purified MBP-tagged GGDEF domains (10 µM) were mixed equimolarly with purified FimV and then incubated overnight with equilibrated amylose resins (New England biolabs, UK). The beads were then washed five times with high-salt Trisbuffer (50 mM Tris-HCl, 500 mM NaCl, 100 mM KCl, 5 mM MgCl 2 , pH 8). Then, proteins were then eluted using SDS-polyacrylamide gelelectrophoresis (SDS-PAGE) sample buffer and the eluate was analyzed using SDS-PAGE).

Western blotting
Protein fusions were tested for stability and expression using Western blot analysis. Cells were grown in liquid culture until exponential phase to generate lysates. The concentration was adjusted to OD 600 of 10, and 10 µl were loaded onto an SDS-gel and immunoblot detection was performed as described previously 15 . Polyclonal antibodies against GFP were used to detect proteins fused to sfGFP or mVenus. Respective antibodies harbored an HRP fusion, and SuperSignal West Pico chemiluminescent substrate (Thermo Scientific, Schwerte, Germany) was used to generate a luminescence signal. The signal was then detected using a Fusion-SL chemiluminescence imager (Peqlab, Erlangen, Germany). All blots and gels shown derive from the same experiments and were processed in parallel. The original data file of the cropped and reassembled Western blot is shown in Fig. 6f is provided as Source Data file.

Biolayer interference assays
To analyze the effect of in frame amino-acid substitutions on the affinity of SoFimV to SoGGDEF, the dissociation constant K D was determined by biolayer interference assays. SoFimV was used as ligand and purified in HEPES buffer (20 mM HEPES, 250 mM NaCl, 50 mM KCl, 5 mM MgCl 2 ) before biotinylation using EZ-Link NHS-PEG4-biotin (Thermo Fisher Scientific, Waltham, Massachusetts) in equimolar concentration. Free biotin was removed using zebra spin desalting spin columns (Thermo Fisher Scientific, Waltham, Massachusetts).
The assay was carried out using the BLITZ system (ForteBio, UK) with streptavidin biosensors. For experiments that determined only the association medium salt HEPES buffer (50 mM HEPES, 250 mM NaCl, 50 mM KCl, 5 mM MgCl 2 , 0.02% (v/v) Tween20, pH 7.5) was used. However, the ionic strength of medium salt buffer was not high enough to induce sufficient dissociation, and, therefore, high-salt Tris-buffer (50 mM Tris-HCl, 500 mM NaCl, 50 mM KCl, 5 mM MgCl 2 , 0.02% (v/v) Tween20, pH 8) was used for association-dissociation experiments. The sensors were hydrated for at least 10 min in assay buffer and runs started with 30 s equilibration until 4 µl of the ligand protein was bound to the sensor for 2 min followed by 30 s washing. Then the association of the analyte protein was measured in real time for 300 s. For association-dissociation kinetic experiments this step was followed by 300 s washing with high-salt Tris buffer (50 mM Tris-HCl, 500 mM NaCl, 100 mM KCl, 5 mM MgCl 2 , 0.2% (v/v) Tween20, pH 8) and the BLITZ PRO software was used to determine the kinetic constants. For each interaction pair this was repeated at least eight times in increasing concentrations until the K D value was stable. Data was further analyzed using the GraphPad PRISM software.

MANT-GTP-binding assays
To verify the functionality of mutated SoGGDEF proteins they were tested for GTP binding using MANT-GTP (Jena Bioscience, Germany). Proteins were diluted to 1 µM and 100 µl of each sample was added in triplicates into a white round bottom 96-well plate before MANT-GTP with a final concentration of 1 µM was added. The plate was then incubated for 30 min at room temperature on a plate shaker before the fluorescence (excitation: 295 nm, emission: 448 nm) was measured in a Tecan Infinite M200 plate reader (Tecan, Switzerland).

Isothermal titration calorimetry assays
Isothermal titration calorimetry (ITC) was performed with MicroCal ITC2000 (Malvern Pananalytical, Kassel, Germany). The proteins were rebuffered five times into a tenfold volume of assay buffer and then concentrated using Vivaspin 6, 10,000 MWCO (Sartorius, Göttingen, Germany) at 4900 x g and 4°C. The rebuffered samples were degassed immediately before ITC with MicroCal ThermoVac (Malvern Pananalytical, Kassel, Germany). The experimental set-up was as follows: 18 injections of 2.0 µl 50 µM c-di-GMP were added to 200 µl 50 µM protein solution with an injection time of 4 and 180 s spacing between each injection at 25°C, and with constant stirring at 750 rpm. The first injection was set to 0.4 µl to remove air and mixed reactants from the tip. For each experiment, we performed a ligand-to-buffer (LtB) titration as well as a buffer-to-protein (BtP) titration with the same experimental set-up to correct for possible heat of dilution introduced by the ligand or protein. The data was analyzed using MicroCal Analysis (OriginLab). The first injection peak was discarded and the isotherms of the LtB and BtP controls were subtracted from the experimental isotherm. We then performed a nonlinear fit with one set of binding sites.

iSCAM assays
Proteins were purified in low-salt HEPES buffer (20 mM HEPES, 50 mM NaCl, 50 mM KCl, 5 mM MgCl 2 , pH 7.5) as described above using Ni-NTA affinity chromatography and SEC. Mass spectra were acquired on a Refeyn OneMP mass photomoeter (Refeyn, UK). The instrument was calibrated with Native Mark standards (ThermoFisher) at room temperature before use. We used high precision glass coverslips, 24 × 50 mm 2 , No. 1.5H, (Marienfeld, Germany) together with self-adhesive CultureWellTM gaskets (Grace Bio-Labs). Before measurements, samples were prepared at 1 µM. Before each measurement, the instrument was focused using 18 µl of room temperature buffer. Two microliters of were then added to the drop and mixed before measurement. We acquired 6000 frames for each measurement. Movies were analyzed using the DiscoverMP software (Refeyn, UK).

Protein crosslinking
The used proteins were purified by Ni-NTA affinity chromatography in HEPES buffer (20 mM HEPES, 250 mM NaCl, 50 mM KCl, 5 mM MgCl 2 ). Fractions containing the proteins of interest were then separated into two batches before SEC where one sample was further purified in HEPES buffer and the other in PBS. The next day, 10 µM proteins were crosslinked by addition of 1 mM DSBU and samples were incubated for 30 min at room temperature before the crosslinking reaction was stopped by adding Tris-buffer to a final concentration of 10 mM. The crosslinking was verified by SDS-PAGE (the nonprocessed original gel shown in Supplementary Fig. 3a is supplied as Supplementary Fig. 11). Crosslinked proteins were digested by the addition of 1 µg trypsin (Promega) and overnight incubation at 30°C. A final of 0.5% sodiumdeoxycholate was spiked to the samples followed by reduction (Tris-2-Carboxyethyl phosphine, 5 mM, 90°C, 10 min) and alkylation (Iodoacetamide, 10 mM, 25°C, 30 min), and an additional digestion step using 1 µg trypsin for 2 h. Finally, peptides were desalted using C18 reversed phase solid phase extraction cartridges (Macherey-Nagel).

Liquid chromatography-mass spectrometry (LC-MS/MS) analyses and identification of crosslinked peptides
LC-MS/MS analysis of digested crosslinked protein complexes was performed on Q-Exactive Plus mass spectrometer connected to an electrospray ionization source (Thermo Fisher Scientific). Peptide separation was carried out using a Ultimate 3000 nanoLC-system (Thermo Fisher Scientific), equipped with a C18 resin column (Magic C18 AQ 2.4 µm, Dr. Maisch) packed in-house. The peptides were first loaded onto a C18 precolumn (preconcentration set-up) and then eluted in backflush mode with a gradient from 98% solvent A (0.15% (v/v) formic acid) and 2% solvent B (99.85% acetonitrile, 0.15% formic acid) to 35% solvent B over 60 min. The flow rate was set to 300 nl min −1 . The data acquisition mode for the initial LFQ study was set to obtain one highresolution MS scan at a resolution of 70,000 (m/z 200, MS1) and 17,500 (m/z 200, MS2) with scanning range from 375 to 1500 m/z followed by MS/MS scans of the 10 most intense ions. To increase the identification efficiency of MS/MS attempts, the charged state screening modus was adjusted to exclude unassigned, singly, and doubly charged ions. The dynamic exclusion duration was set to 30 s. The ion accumulation time was set to 100 ms (both MS and MS/MS). The automatic gain control (AGC) was set to 3 × 10 6 for MS survey scans and 1 × 10 5 for MS/MS scans. Collision was induced by high collision dissociation (HCD) with an NCE 23. MS raw files were converted into mgf files and analyzed using MeroX (v2.0) 74 in default settings.

In silico protein analyses
Protein sequence alignments and position-based weight maps. The NCBI and Pfam database were accessed to obtain the amino-acid sequences of PdeB homologs and other GGDEF domains. Alignments were created using the Clustal omega tool 75 and were further analyzed with JalView 76 . The position-based weight map of PdeB homologs from different Shewanella species was created with the Seq2Logo tool 77 using the P-Weighted Kullback-Leiber clustering algorithm.

Protein structure prediction and visualization
The protein structure of PdeB (PAS-GGDEF-EAL) was predicted using the Phyre2 tool 78 , while the structure of the C-terminal region of HubP was modeled using the Swiss Model algorithm 79 . The obtained data was then analyzed using Pymol (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC) and the included APBS electrostatics tool.

Extraction and analysis of cyclic dinucleotides
Extraction and quantification of nucleotide messengers was conducted as described 80 . Briefly, cells were grown to exponential phase before harvesting by centrifugation. PDEs and DGCs were inactivated by heating for 10 min at 95°C and nucleotides were then extracted using a mixture of acetonitrile, methanol, and water (2:1:1). The protein amount was quantified using BCA assays with the residual pellet and extracted nucleotides were dried using vacuum centrifugation and stored at −20°C before further use. To quantify the extracted nucleotides, mass spectrometry was operated in positive ionization mode. MS was performed with 50 µl each of calibrators, quality controls and biological samples and analyte separation was done accordingly to the HPLC gradient method. The obtained data of the calibrators was used to generate a calibration curve and the MS/MS signals obtained by the biological samples were analyzed by calculating the peak areas and calculating the ratios compared to the calibration signals to obtain precise measurements of the nucleotide concentrations.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

DATA AVAILABILITY
The datasets generated during and/or analyzed during the current study are available from the corresponding authors on reasonable request.