Chemotaxis and cyclic‐di‐GMP signalling control surface attachment of Escherichia coli

Attachment to surfaces is an important early step during bacterial infection and during formation of submerged biofilms. Although flagella‐mediated motility is known to be important for attachment of Escherichia coli and other bacteria, implications of motility regulation by cellular signalling remain to be understood. Here, we show that motility largely promotes attachment of E. coli, including that mediated by type 1 fimbriae, by allowing cells to reach, get hydrodynamically trapped at and explore the surface. Inactivation or inhibition of the chemotaxis signalling pathway improves attachment by suppressing cell reorientations and thereby increasing surface residence times. The attachment is further enhanced by deletion of genes encoding the cyclic diguanosine monophosphate (c‐di‐GMP)‐dependent flagellar brake YcgR or the diguanylate cyclase DgcE. Such increased attachment in absence of c‐di‐GMP signalling is in contrast to its commonly accepted function as a positive regulator of the sessile state. It is apparently due to the increased swimming speed of E. coli in absence of YcgR‐mediated motor control, which strengthens adhesion mediated by the type 1 fimbriae. Thus, both signalling networks that regulate motility of E. coli also control its engagement with both biotic and abiotic surfaces, which has likely implications for infection and biofilm formation.

binding is mediated by FimH at the fimbrial tip that interacts with mannosides in a force-dependent catch bond fashion, which allows bacteria to remain attached even when exposed to high sheer flow experienced by E. coli during urinary tract infection (Aprikian et al., 2011;Nilsson, Thomas, Sokurenko, & Vogel, 2006). Type 1 fimbriae also bind non-specifically to a variety of abiotic surfaces (Beloin et al., 2008;Chao & Zhang, 2011;Pratt & Kolter, 1998) being important at the early stages of biofilm formation (Monteiro et al., 2012;Wang et al., 2018). Fimbriae-mediated attachment further contributes to the catheter-associated infections (Reisner et al., 2014).
Besides possessing adhesive structures, planktonic bacteria need to establish contact with the surface in order to attach.
Motility in E. coli and other bacteria is controlled by the chemotaxis pathway (Sourjik & Wingreen, 2012;Wadhams & Armitage, 2004) as well as by the biofilm-associated second messenger c-di-GMP (Guttenplan & Kearns, 2013;Hengge, 2009;Jenal & Malone, 2006;Romling, 2013). At the core of the chemotaxis pathway is a histidine kinase CheA that is associated with ligand-specific receptors and becomes inhibited upon binding of attractant ligands.
CheA phosphorylates the response regulator CheY, which -in its phosphorylated state -binds to the motor and induces a switch in the direction of flagellar rotation from the default counterclockwise (CCW) to clockwise (CW), resulting in a rapid reorientation of the cell swimming direction (tumble). CheY is subsequently dephosphorylated by the phosphatase CheZ. Although chemotaxis per se is clearly not essential for E. coli biofilm formation (Pratt & Kolter, 1998, 1999, mutations in the chemotaxis system affect attachment and biofilm formation in E. coli and several other species (Armitano, Mejean, & Jourlin-Castelli, 2013; Barken et al., 2008;Merritt, Danhorn, & Fuqua, 2007;Niba et al., 2007;Schmidt et al., 2011).
Binding of YcgR to flagellar motor in the presence of c-di-GMP affects motility, either by decreasing cell swimming speed (Boehm et al., 2010) or by suppressing CW rotation of flagellar motor (Fang & Gomelsky, 2010;Paul et al., 2010). High levels of c-di-GMP also stimulate synthesis of biofilm matrix factors, such as curli in E.
The goal of our study was to analyse the importance of motility and its regulation by the chemotaxis pathway and c-di-GMP during attachment of E. coli to both abiotic and biomimetic mannosylated surfaces. We demonstrate that fimbriae-mediated surface attachment of E. coli is enhanced by the chemoattractant-induced smooth swimming and by higher swimming velocity in absence of c-di-GMP signalling. These findings suggest that c-di-GMP can act as a negative rather than positive regulator of E. coli surface attachment, and that low c-di-GMP levels and low activity of the chemotaxis pathway promote surface attachment of E. coli in a concerted way.

| Smooth swimming promotes surface attachment of E. coli
To investigate possible functions of motility and its regulation in surface attachment of E. coli W3110 (Hayashi et al., 2006), we first quantified biomass of cells attached on microtiter plates after 24 hr incubation using crystal violet (CV) staining ( Figure 1a). In line with previous studies that studied effects of flagella and motility on biofilm formation (Genevaux et al., 1996;Niba et al., 2007;Pratt & Kolter, 1999), ∆fliC strain that lacks flagella and ∆motA strain that has paralysed flagella showed strongly reduced surface colonisation under these conditions. Biomass of surface-attached cells was also reduced for ∆cheZ strain that has an increased level of tumbling, but not for the smooth-swimming ∆cheY and ∆cheA strains. Defects in motility apparently impair surface colonisation by reducing initial attachment, as time-resolved microscopy showed a steady increase in the number of attached wild-type cells during the first 12 hr of incubation but nearly no surface attachment for ∆fliC, ∆motA or ∆cheZ cells (Figure 1b,c). Confirming that under our conditions motility rather than flagella themselves are required for attachment, similar numbers of flagella-less and wild-type cells attached when bacteria were artificially forced to the surface using mild centrifugation ( Figure S1a,b in the Supporting information).
Interestingly, in contrast to both our ( Figure 1a) and previous (Pratt & Kolter, 1998) results based on CV staining of cell biomass at the late stages of attachment and biofilm formation, we observed a much better attachment of ∆cheA and ∆cheY strains at the early time points of incubation (Figure 1b,c, and Figure S2a,b in the Supporting F I G U R E 1 Motility and chemotaxis regulate E. coli surface attachment. (a) Relative surface colonisation by motility and chemotaxis mutants. E. coli cultures were incubated on polystyrene Corning Costar tissue culture TC-treated plates for 24 in M9 medium at 30°C. Colonisation was quantified using staining with crystal violet (CV) and normalised to the CV value of the wild type (wt). Shown are mean and standard error of three to eight replicates. (b, c) Time course of the relative surface attachment of indicated motility and chemotaxis mutants. Wild-type cells labelled with cyan fluorescent protein (CFP) or with mCherry were mixed 1:1 with wild-type or mutant cells labelled with yellow or green fluorescent protein (YFP or GFP) and incubated for indicated time in M9 medium at 30°C on polystyrene BD Falcon TC-treated plates. (b) Exemplary images of mixed cultures of wild-type cells labelled in magenta and wild type or mutants in green. Scale bar: 25 µm. (c) Relative numbers of attached cells at different time points. The number of attached cells in each image was normalised to the number of wild-type cells, and the values were normalised again to the wild type/wild type ratio in the same experimental series, to compensate for any growth effects of fluorescent protein expression. Inset shows absolute numbers of attached wild-type and ∆fliC cells. Shown are mean and standard error of three replicates. (d, e) Swimming behaviour at the surface of a glass slide. Cells were grown in planktonic cultures to post-exponential phase and their swimming was analysed in motility buffer. (d) Exemplary images of tracks of wild-type and ∆cheY cells. (e) Quantification of trajectory durations and lengths. Shown are mean and standard error of three replicates. (f) Relative surface attachment in the presence of chemoattractants. Attachment of wild-type and ∆cheY cells in motility buffer with and without 10 mM of methyl aspartate and serine to ibidi uncoated imaging plates was quantified after 20 min incubation (see Experimental procedures), and normalised to the number of wild-type cells that were attached without chemoattactants. Shown are mean and standard error of six replicates. Statistical analyses were performed here and throughout using a two-sample t-test with unequal sample size and unequal variance, with p < .05 (*), p < .005 (**), p < .0005 (***), NS, not significant These results indicate that cell attachment may be directly related to the ability of bacteria to make smooth runs, which is strongly reduced in ∆cheZ but is increased in ∆cheA and ∆cheY strains. Confirming that the enhanced attachment of ∆cheY is motility dependent, no increase in attachment was observed upon deletion of ∆cheY in ∆fliC background ( Figure 1c). Moreover, adhesiveness of ∆cheY cells seemed to be even lower than that of the wild type, judging by their poorer attachment in the centrifugation assay ( Figure S2c in the Supporting information). Smooth swimming may lead to trapping of cells in the proximity of the surface (Berke et al., 2008;Drescher et al., 2011;Frymier et al., 1995;Vigeant et al., 2002), and indeed ∆cheY cells spend more time swimming at the surface than wild-type cells ( Figure 1e) and their trajectories are longer ( Figure 1d,e). Both measures confirm entrapment of ∆cheY cells at the surface, which means that the smooth swimming mutants sample a larger surface area than wild-type cells and therefore have an increased probability of attachment.

Since stimulation with chemoattractants transiently inhibits
CheA activity and therefore reduces cell tumbling, we expected it to enhance attachment similar to the effects of cheA and cheY knockouts. Indeed, we observed that stimulation with 10 mM of α-metyl-DL-aspartate and L-serine, strong chemoattractants sensed by two major E. coli chemoreceptors, elevated attachment of the wildtype cells to the levels similar to, and even slightly higher than that of ∆cheY strain (Figure 1f). The origins of such better attachment of the attractant-stimulated wild type compared to ∆cheY mutant need further investigation. The generally reduced attachment of stimulated cells, apparent for ∆cheY but also presumably true for the wild-type cells, is likely to be non-specifically caused by high millimolar levels of amino acids.

| Low levels of c-di-GMP enhance surface attachment
We next observed that colonisation of microtiter plate surface by E. coli is also affected by c-di-GMP signalling. Deletion of the gene encoding DgcE (YegE), the dominant c-di-GMP cyclase under our growth conditions (Sarenko et al., 2017) and the major cyclase controlling motility via c-di-GMP in E. coli (Pesavento et al., 2008) Although ∆dgcE and ∆pdeH strains showed similar surface attachment after 1 hr of incubation in the microscopy assay, which might be due to inherently low c-di-GMP levels at the early phase of culture growth, the effects of the deletions become visible be- To rule out that the effect of ycgR deletion on attachment is due to altered adhesiveness, we again performed the centrifugation assay. Indeed, no differences in adhesion were observed between wild-type and ∆ycgR cells ( Figure S5 in the Supporting information).
Relative attachment of ∆dcgE cells was even mildly reduced, whereas that of ∆pdeH calls was increased ( Figure S5 in the Supporting information), which could be explained by involvement of c-di-GMP in the production of extracellular matrix (Hengge, 2009;Romling et al., 2013;Serra & Hengge, 2019). Since these effects are opposite to the differences in attachment observed for the swimming cells, they do not appear to play major role in attachment under our conditions.
Finally, we observed that when compared to the wild type, not only the number of attached ∆ycgR cells but also the strength of attachment apparently increased ( Figure S6 in the Supporting information).

| C-di-GMP controls swimming speed via YcgR
To get insights into the mechanism of YcgR-mediated reduction of surface attachment by c-di-GMP, we directly compared surface attachment and motility of wild-type and ∆ycgR cells. Even when pregrown in a shaking culture to OD 600 of 0.5 and subsequently imaged in the motility buffer (i.e., in absence of further growth), ∆ycgR cells showed pronouncedly enhanced surface attachment compared to the wild type ( Figure 3a). Thus, the enhanced attachment of ∆ycgR cells does not require growth or differentiation on the surface.
Among the analysed motility parameters, differences in swimming speed between mutants were consistent with the previous report

| Type 1 fimbriae mediate swimming-speed dependence of surface attachment
Since flagella apparently have only a minor role in attachment under our conditions, we tested the potential involvement of type 1 fimbriae, one of the major and best characterised adhesins of E. coli (Korea et al., 2011). Indeed, ∆fimA strain that lacks a major subunit of fimbria showed clearly reduced attachment (Figure 4a,b). In contrast, deletion of csgA encoding the subunit of curli fibres, major component of E. coli biofilm matrix (Vidal et al., 1998), had no significant effect on attachment under our conditions (Figure 4a,b). Same results were observed when the surface colonisation was assessed for these strains after 24 hr ( Figure S8 in the Supporting information), which is consistent with the previously observed higher importance of type 1 fimbriae for early biofilm formation (Monteiro et al., 2012).
Strikingly, lack of fimbriae also abolished the effect of the YcgR deletion, with attachment of ∆fimA and ∆fimA∆ycgR being identical.
This indicates that fimbriae are responsible for the increased number of attached cells seen for the ΔycgR mutant, which was confirmed by similar attachment of fimbria-less strains with and without ycgR.
In contrast, ∆cheY cells attached better even in absence of fimbriae ( Figure 4b), although this enhancement was weaker than in the wildtype background. This suggests that prolonged swimming enhances attachment independently of the adhesin involved. Comparable results were obtained when the experiment was performed at higher temperature, 30°C ( Figure S9 in the Supporting information).
Specificity of ∆fimA effects on attachment could be confirmed by complementation with FimA expression from a plasmid ( Figure S10 in the Supporting information). Moreover, the activity of type 1 fimbriae fimD promoter was not affected by ycgR deletion ( Figure S11a in the Supporting information), favouring the hypothesis that attachment mediated by type 1 fimbriae is directly promoted by cell swimming speed. Similarly, no differences in fimD promoter activity were observed in ∆dgcE cells ( Figure S11b in the Supporting information). The effect of c-di-GMP was also observed for the specific attachment on mannosylated surface, which is strictly dependent on type 1 fimbriae (Figure 4c,d and Figure S12 in the Supporting information). For a relatively brief (20 min) incubation, the attachment to this surface was strongly promoted by motility (Figure 4c,d), although flagella-less ∆fliC cells attached somewhat better than ∆motA cells, possibly because immobile flagella partially hinder the attachment. Importantly, the attachment of ΔycgR cells under these conditions was almost twice of that observed for the wild type.
In contrast, attachment of ΔycgR∆fliC and ∆fliC cells was indistinguishable, confirming that the effect of YcgR on attachment is motility-dependent. Strongly increased attachment was also observed for the smooth-swimming ∆cheY cells. However, motility became less important for attachment at later time points, with attachment of ∆motA and ΔycgR cells becoming similar to that of the wild type after 1h, and the advantage of ∆cheY cells being reduced ( Figure S12 in the Supporting information). At this later time point, the lack of flagella reduces the number of attached cells, as ∆fliC cells attached less efficiently than ∆motA cells, pointing to a possible role of flagella as (non-specific) secondary adhesins that can stabilise the fimbriae-mediated primary adhesion.

| D ISCUSS I ON
Although multiple studies have demonstrated that flagella and motility play an important role for bacterial attachment to biotic and abiotic surfaces Friedlander et al., 2015;McClaine & Ford, 2002;Pratt & Kolter, 1999;Zhou et al., 2013), the importance of motility control through cell signalling in this transition from planktonic to the sessile lifestyle remains unclear. Here, we could show that both established networks that post-translationally regulate swimming of E. coli, the chemotaxis pathway and the c-di-GMP signalling pathway, also control initial stages of cell adhesion.

F I G U R E 3
Faster swimming of ∆ycgR cells promotes attachment to surfaces. (a) Relative surface attachment of wild-type and ∆ycgR cells in motility buffer. Planktonic cultures were grown to the indicated OD 600 . Wild-type cells labelled with mCherry were mixed 1:1 with wildtype or ∆ycgR cells labelled with YFP and incubated in ibidi uncoated imaging plates at room temperature for 1 hr. The number of attached cells in each image was normalised to the number of wild-type cells, and the values were normalised again to the wild type/wild-type ratio in the same experimental series. Shown are mean and standard error of five to six replicates. (b-d) Swimming behaviour at the surface of a glass slide. Planktonic cultures were grown to the indicated OD 600 , and swimming speed (b), tumbling rate (c), and trajectory lengths (d) of wild type, ∆ycgR, ΔdgcE and ΔpdeH cells swimming at the surface were quantified. Shown are mean and standard error of three replicates. Statistical analyses were performed here and throughout using a two-sample t-test with unequal sample size and unequal variance, with p < .05 (*), p < .0005 (***), NS, not significant First, we confirmed previous observations (Pratt & Kolter, 1998, 1999) that flagella-driven motility strongly promotes initial colonisation of the abiotic as well as of the mannosylated surface by E.
coli. For both types of studied surfaces, the cell adhesion is primarily fimbriae. This YcgR-dependent regulation of attachment apparently correlates with perturbations of global cellular c-di-GMP levels, primarily controlled by the major diguanylate cyclase/ phosphodiesterase pair DgcE/ PdeH. Although this contrasts with a previous study that instead reported an increased attachment of E. coli at high levels of c-di-GMP (Fang & Gomelsky, 2010), this discrepancy could be explained by the observed YcgR-independent effects of c-di-GMP on cell surface adhesiveness, which might have had stronger effect on attachment in the study of Fang and Gomelsky. Notably, it has been previously reported in Salmonella that the effect of dgcE knockout could be different from deletions of other cyclases (Ahmad et al., 2011).
Our results suggest that the most likely cause of better attachment of ΔycgR cells is their faster swimming. In agreement with a previous report (Boehm et al., 2010), we indeed observed that YcgR negatively regulates swimming speed in presence of high levels of c-di-GMP. In contrast, we observed no evidence for the YcgR-mediated control of cell tumbling rate, as could be expected given previously reported influence of c-di-GMP on the CW rotation bias of flagellar motor (Girgis, Liu, Ryu, & Tavazoie, 2007;Paul et al., 2010). One possible explanation of this discrepancy is that the effect of YcgR on flagellar motor rotation is load-dependent, so that the CW bias is only sensitive to YcgR binding at high load such as present in tethered cells, but not in cells swimming in a liquid with low viscosity. We further observed that the effects of chemotaxis activity and c-di-GMP levels on attachment were multiplicative. The negative regulation of surface attachment by c-di-GMP suggests that this second messenger might have dual function during biofilm formation dependent on its stage (Romling, 2012), with low levels of c-di-GMP being important for efficient early attachment, while high levels being required for matrix production in mature biofilms (Hengge, 2009;Romling et al., 2013).
At least in the case of mannosylated surface, enhanced attachment of faster bacteria could be explained by the known forcedependence of mannose binding by FimH (Aprikian et al., 2011;Sauer et al., 2016). This catch-bond mechanism is normally assumed to promote attachment under high shear force, for example, in the urinary tract, but the same effect could also strengthen attachment of faster cells. Indeed, the effects of high fluid flow and fast swimming on cell interaction with the surface may be similar, increasing tensile force on adhesins interacting with the surface. And while the mechanism of type 1 fimbriae interaction with abiotic surfaces has not being established, it is at least possible that the same large force-induced conformational change in FimH is also responsible for better attachment of faster cells to these surfaces. In contrast, the residual attachment to abiotic surface that was observed in ΔfimA cells and is apparently mediated by one or several of other E. coli adhesins that can attach to plastic (Korea et al., 2011), was not regulated by YcgR. Thus, faster swimming bacteria with low c-di-GMP levels could have an advantage when attaching to biological surfaces, such as bladder epithelial cells, as well as to abiotic surfaces such as catheters via type 1 fimbriae.

| Bacterial strains and plasmids
All strains and plasmids used in this study are listed in Table S1.

| Surface mannosylation
In order to create surfaces containing mannose residues, 1% BSA was added to 96-well imaging uncoated plates from ibidi® and incubated for 30 min at room temperature. Surfaces were then washed to remove the unattached BSA and 0.1 mg/ml of 4-Methylumbelliferyl α-D-mannopyranoside (Sigma Aldrich, USA) was added. Plates were incubated for 15 min under UV light using an Ebox VX5 system (Vilber Lourmat, France) in order to activate the fluorophore as previously described (Belisle, Correia, Wiseman, Kennedy, & Costantino, 2008). Wells were then washed and cells were added as explained for attachment of planktonic cells. Colonised surfaces for imaging were grown as described above (growth conditions) and imaged in tethering buffer (10 mM KPO 4 , 0.1 mM EDTA, 1 µM methionine, 10 mM lactic acid, 67 mM NaCl, pH 7) after washing with buffer for 3 times. Cells for fluorescent imaging were labelled with eYFP, eCFP, eGFP or mCherry expressed from plasmids pVS147, pVS130, pVM42 or pOB2. Images from wide-field microscopy were analysed using ImageJ (http:// imagej.nih.gov/ij/).

| Centrifugation-enforced cell attachment
Samples of mixed cultures labelled with different fluorophores were prepared as for surface colonisation experiments on TC-treated BD Falcon™ imaging plates and the plates were centrifuged for 2 min at 650 g. After centrifugation, cells were washed with tethering buffer and imaged in motility buffer.

| Tracking experiments
Cells were grown to OD 600 ~ 0.6 or OD 600 ~ 1.05, as indicated, following the same protocol as for attachment experiments, har-  (Sbalzarini & Koumoutsakos, 2005) running as an ImageJ plugin, and extracted trajectories were analysed using custom-made algorithms running as ImageJ plugins, to evaluate tumbling rate, run speeds and trajectory durations and lengths.

| Tracking analysis
The trajectories were analysed as follows. For trajectories longer than 0.5 s, discrimination between Brownian and stuck particles on the one hand and swimmers on the other hand was performed using a radius of gyration criterion: the measure of diffusivity max t,t � r t − r t � ∕t traj must be larger than 0.2 px 2 ∕fr for the particle to be considered a swimmer, where t traj is the trajectory duration and t and t′ are any couple of times within the trajectory. The tumbles were identified using the following criteria. The veloc-

ity of the cells is computed as
� � Δ⃗ r � � is computed over the same duration, and the ratio r = | | ⃗ v | | ∕d is then evaluated. This quantity is close to 1 during the runs and low (typically less than 0.6-0.7) during the tumbles. A threshold r th is computed which maximises Shannon's entropy, following Huang's fuzzy thresholding method (Huang & Wang, 1995), using all the tracks of swimmers in one movie as the sample. This threshold is used to separate each trajectory into putative run and tumble segments.
To reduce the noise, the variance v run of r during the run segments

| Swimming speed measurements
The same culture growth and cell harvesting protocols were used as for tracking experiments, except that the cells were resuspended to a final OD 600 of 0.5. The sample preparation was the same. It was observed under the same microscope. The motion was then recorded for 100 s at 100 fps. The movie was then analysed using the Differential Dynamic Microscopy algorithm (Wilson et al., 2011), which evaluates the average swimming speed of the population of cells, as well as the standard deviation of swimming speeds within the population and the fraction of swimming cells. The fraction of swimming cells was above 80% in any case, and the standard deviation about 30% of the mean.

| Flow cytometry
Activity of fimD promoter was assessed using a plasmid-based egfp reporter from E. coli promoter library (Zaslaver et al., 2006). Samples were grown and prepared as described for measurements of planktonic attachment. Cells were diluted to an OD 600 of 0.4 in motility buffer and their fluorescence was measured. Alternatively, to investigate promoter activity in swimming and attached cells, 500 µl of this cell suspension was seeded in wells of 8-well ibidi® imaging plates. After 1 hr incubation at room temperature, supernatants were taken and attached cells were scraped in 200 ml of fresh motility buffer. Fluorescence levels of both fractions were measured with BD LSRFortessa SORP cell analyzer (BD Biosciences, Germany).

ACK N OWLED G EM ENTS
The authors thank K. Thormann and O. Schauer for insightful discussions, and Silvia González Sierra for help with flow cytometry. This work was supported by grant DFG SO421/12-1 within the priority program SPP1617 from the Deutsche Forschungsgemeinschaft and by the Max Plank Society.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon request.