Transpeptidase PBP2 governs initial localization and activity of major cell-wall synthesis machinery in Escherichia coli

Bacterial shape is physically determined by the peptidoglycan cell wall. The cell-wall-synthesis machinery responsible for rod shape in Escherichia coli is the processive ‘Rod complex’. Previously, cytoplasmic MreB filaments were thought to govern formation and localization of Rod complexes based on local cell-envelope curvature. However, using single-particle tracking of the transpeptidase PBP2, we found strong evidence that PBP2 initiates new Rod complexes by binding to a substrate different from MreB or any known Rod-complex component. This substrate is likely the cell wall. Consistently, we found only weak correlations between MreB and envelope curvature in the cylindrical part of cells. Residual correlations do not require any curvature-based Rod-complex initiation but can be attributed to persistent rotational motion. Therefore, local cell-wall architecture likely provides the cue for PBP2 binding and subsequent Rod-complex initiation. We also found that PBP2 has a limiting role for Rod-complex activity, thus supporting its central role.


14
The peptidoglycan (PG) cell wall is the major load-bearing structure of the bacterial cell 15 envelope and physically responsible for cell shape (Vollmer et al., 2008). Rod-like cell shape MreB. We thus wondered, whether the cell wall itself could provide a local cue for the 64 initiation of Rod complexes, independently of cell-envelope curvature. Such a local cue 65 would have to be sensed by a protein with a periplasmic domain that can possibly bind the 66 cell wall. 67 An obvious candidate is the transpeptidase PBP2. For its cross-linking activity PBP2 must 68 bring together donor peptides on nascent glycan strands and acceptor peptides in the cell 69 wall. Binding of PBP2 to the existing cell wall could therefore provide an alternative 70 mechanism of Rod-complex initiation. In support of this hypothesis, a PBP2(L61R) mutant 71 shows increased cell-wall synthetic activity and affects the distribution of MreB-actin filament 72 length (Rohs et al., 2018 Our major observation let us conclude that PBP2 determines the initial localization of newly 85 forming Rod complexes, possibly by binding to the cell wall directly. In support of this 86 conclusion, we found that MreB filaments are likely not recruited to regions of particular cell-87 envelope curvature in filamentous and normal cells, contrary to (Ursell et al., 2014). 88 Specifically, we found only weak MreB-curvature correlations once cell poles were excluded 89 from the analysis. Residual correlations were attributed to weak spontaneous cell bending, 90 suggesting that they are caused by persistent rotational motion (Wong et al., 2017). Finally, 91 we also found that fast diffusing molecules cannot contribute to processive Rod-complex 92 activity due to limitations of diffusion, contrary to (Lee et al., 2014). 93

PBP2 enzymes can be quantitatively separated into diffusive and bound fractions 95
To study the role of PBP2 for the formation of Rod complexes we characterized its different 96 states of motion, which are potentially representative of different states of substrate binding 97 and activity. We imaged a functional, N-terminal protein fusion of the photo-activatable 98 fluorescent protein PAmCherry to PBP2 (Lee et al., 2014). The fusion is expressed from the 99 native mrdA locus at a level similar to the wild-type protein according to quantitative mass 100 spectrometry ( Fig. 1-SI Fig. 1D)  We obtained single-molecule tracks by single-particle tracking PhotoActivatable Localization 105 Microscopy (sptPALM) (Manley et al., 2008) in total internal reflection fluorescence (TIRF) 106 mode, which restricts the observation to the bottom part of the cell. We first imaged PBP2 107 molecules at high frequency (intervals of 60 ms). We found both spatially extended 108 trajectories, corresponding to fast diffusing molecules, and trajectories that appeared as 109 localized, corresponding to immobile or slowly moving molecules (Fig. 1A, Fig. 1-Movie 1). 110 We confirmed the presence of two distinct fractions of diffusing and localized molecules 111 based on the distribution of single-track effective diffusion constants (Fig. 1B). More 112 specifically, the experimental distribution was fit to the prediction from a two-state diffusion 113 model, which contains as a special case a diffusing and an immobile population ( To test whether all or part of the bound molecules were moving persistently we imaged 128 PBP2 molecules at low frequency, taking images with an exposure time of 1 s and intervals 129 of 3.6 s. The long exposure time effectively smears out the fluorescence of fast diffusing 130 molecules, allowing us to detect the positions of individual bound molecules. Using this 131 protocol, we found molecules that moved persistently, were immobile, or showed transitions 132 between these two states ( Fig. 1C-E, Fig. 1-Movie 2, and Fig. 1-SI Fig. 4). Upon overexpression of PAmCherry-PBP2 as above, we found that the persistent fraction 151 remained nearly constant (Fig. 1H). This finding suggests that PBP2 limits the number of 152 active Rod complexes. This viewpoint is consistent with the recent report that a hyperactive 153 PBP2 point mutant (L61R) increased the overall amount of active Rod complexes (Rohs et 154 al., 2018). We will come back to this mutant below. 155

msfGFP-PBP2 fusion confirms findings and demonstrates increased PBP2 binding 156 upon PBP2 depletion 157
We confirmed our findings using a strain that carries a functional msfGFP-PBP2 fusion 158 under IPTG-inducible control as the sole copy of PBP2 (Cho et al., 2016). We chose the 159 induction level (25 uM IPTG) based on measurements of cell diameter ( Fig. 1-SI Fig. 7A). 6 three-fold higher than native PBP2 in the wild type ( Fig. 1-SI Fig. 7B), while near-wild-type 162 levels observed at lower induction (5 uM IPTG) led to loss of rod shape at long times 163 ( Fig. 1-SI Fig. 7A). This discrepancy might be a consequence of reduced enzymatic 164 efficiency of the msfGFP fusion or simply due to the more noisy expression from the 165 inducible promoter. 166 Tracking the msfGFP-PBP2 at high frequency after initial pre-bleaching, we found a similar 167 fraction of bound PBP2 molecules (23.2 ± 9.2 %) ( Fig. 1-SI Fig. 7C, Fig. 1-Movie 3). 168 Similarly to PAmCherry-PBP2, the bound fraction was reduced only slightly upon two-fold 169 increase of protein levels (18 ± 6.1 %), supporting our previous conclusion that the target of 170 PBP2 binding is non-saturated. 171 Among the bound molecules we found a fraction of 80 ± 10 % persistently moving molecules 172 ( Fig. 1-SI Fig. 8A, Fig. 1-Movie 4), which is significantly higher than in the case of the 173 PAmCherry fusion. However, similarly to the PAmCherry fusion, the persistent fraction 174 remained nearly constant over the 6-fold range of expression levels, only showing a mild 175 drop at the highest induction level of 100 uM IPTG ( Fig. 1-SI Fig. 8A). These results 176 therefore confirm that the binding target of PBP2 is non-limiting and that PBP2 actively limits 177 Rod-complex activity. 178 Interestingly, the bound fraction increased almost two-fold after long times of low expression 179 levels (5 uM) ( Fig. 1-SI Fig. 9B), already before rod shape had been lost ( Fig. 1-SI Fig. 9A). 180 This observation suggests a potential feedback mechanism between cell-wall architecture 181 due to reduced Rod-complex activity and PBP2 binding. We will come back to this point 182 further down. 183

A hyperactive PBP2 mutant (L61R) shows increased binding at low expression 184
Initially intended as a further test of our method, we also tracked an msfGFP fusion to the 185 PBP2 point mutant (L61R) recently characterized by (Rohs et al., 2018). Based on slow-186 frequency imaging, they reported that the number of bound msfGFP-PBP2(L61R) was about 187 two-fold increased as compared to msfGFP-PBP2 (Rohs et al., 2018). We confirmed this 188 finding quantitatively ( Fig. 1-SI Fig. 7C). However, we also found that the bound fraction 189 was high only at low protein expression (5 uM IPTG), while it was equal to the bound fraction 190 of the wild-type protein at higher expression (25 uM IPTG). Furthermore, the persistent 191 fraction was reduced in comparison to the wild-type protein ( Fig. 1-SI Fig. 8A). Therefore, 192 the mutant shows higher activity at low expression as previously reported but reduced 193 activity at higher expression. 194 The classification into different motion states at the sub-trajectory level allowed us to extract 196 average transition rates between immobile and persistently moving states, kip and kpi, 197 respectively ( Fig. 2A-B). Depending on the fluorescent-protein fusion, we found values of kip 198 between 0.015-0.06 s -1 , and of kpi between 0.006-0.02 s -1 . The msfGFP-PBP2 fusion shows 199 less frequent arrests and faster transitions from immobile to persistently moving states, in 200 agreement with its higher fraction of moving molecules ( Fig. 1-SI Fig. 8A). 201 A persistent run of PBP2 terminates either due to an arrest of PBP2 (persistent-to-immobile 202 transition) or due to an unbinding event (persistent-to-diffusive transition). As an upper 203 bound of the unbinding rate, we measured the transition rate from the aggregate bound state 204 (persistent and immobile states) to the diffusive state, kbd (Fig. 1A). Specifically, we acquired 205 distributions of track lengths for two different imaging intervals of 1 s and 12 s ( To test whether MreB filaments were required for PBP2 binding we first imaged the spatial 224 distribution of an mCherry-PBP2 fusion on the cell contour in untreated or A22-treated 225 conditions, in the same manner as we imaged MreB-msfGFP (Fig. 3B, Fig. 3-SI Fig. 1B). 226 We found that mCherry-PBP2 showed a spotty pattern. Since these images were taken with treatment using single-molecule tracking in TIRF. The bound fraction remained close to the 232 value of untreated cells (Fig. 3E). Yet, the apparent fraction of persistently moving molecules 233 among all bound PBP2 enzymes nearly vanished (Fig. 3F). This is consistent with the arrest 234 of MreB rotation ( Fig. 3-Movie 2) and with previous bulk measurements of  activity (Uehara and Park, 2008). To follow the bound fraction during two mass-doubling 236 times (Fig. 3E), we used an intermediate concentration of A22 (20 uM), which did not affect 237 growth (Fig. 3-SI Fig. 2A). Together, our findings suggest that MreB polymers are neither 238 the substrate of PBP2 binding nor do they affect the rate of PBP2 binding and unbinding. 239 MreB depolymerization did also not elicit a rapid change of the diffusion constant of the 240 freely diffusing molecules (Fig. 3E), suggesting that MreB does not significantly constrain the 241 movement of diffusive PBP2 molecules to an area close to the MreB cytoskeleton (Strahl et 242 al., 2014). Therefore, if PBP2 binds its substrate from the diffusive state, the locations of 243 PBP2 binding are also independent of MreB. 244 A22 treatment already demonstrates that PBP2 binding is independent of Rod-complex 245 activity. We confirmed this finding using the PBP2-targeting beta-lactam Mecillinam, which 246 binds covalently to the active site (Spratt, 1975). Mecillinam did not cause a reduction of the 247 bound fraction (Fig. 3E), demonstrating that PBP2 binds its binding target with a moiety 248 different from its active site. 249 At long times of treatment with Mecillinam or A22 (120 min) the bound fraction increased 250 and the diffusion constant decreased (Fig. 3E), which coincides and is potentially caused by 251 the loss of normal cell-wall architecture during loss of rod-like cell shape (Fig. 3-SI Fig. 2C), 252 similar to the increase of the bound fraction at sustained low induction levels of msfGFP-253 PBP2 ( Fig. 1-SI Fig. 9B). 254

PBP2 binds to its substrate at locations that are independent of MreB localization 255
To demonstrate that PBP2 binding was indeed independent of MreB filaments or Rod-256 complex activity as suggested by Fig. 3 we still needed to show that PBP2 molecules 257 interchange between diffusive and bound states during A22 treatment. We already found a 258 low upper bound for the transition rate from bound to diffusive states in non-treated cells (kbd 259 < 0.03 s -1 ) (Fig. 2C), and we expect inverse transitions to occur even more rarely. We suggesting that fast diffusing molecules re-entered the observation window within less than 265 half a minute but did not quickly bind their substrate. Within about 2-4 min the bound fraction 266 recovered, yielding a transition rate from diffusive to bound states of kdb = (4.5 ± 2)*10 -3 s -1 . 267 The same experiment in non-treated cells did not reveal recovery of the bound fraction 268 ( Fig. 4-SI Fig. 1), presumably because bound molecules left the field of view through 269 persistent motion. 270 Since the bound and diffusive fractions did not change by more than 10% after A22 271 treatment, transition rates from the diffusive into the bound states must be matched by 272 reverse transitions from the bound to the diffusive state with a rate kbd = kdb(1-b)/b = 273 0.02 ± 0.01 s -1 , where b is the bound fraction. We confirmed this expectation through 274 independent lifetime measurements of A22-treated cells, similar to those in Fig. 2 (Fig. 4C) which makes this scenario highly unlikely. Therefore, we suggest that PBP2 binds to a site in 297 the cell envelope and then recruits MreB filaments (Fig. 4D). Thus, PBP2 determines the 298 locations of newly forming Rod complexes. 299

PBP2-binding substrate is none of the known Rod-complex components but likely the 300 cell wall 301
Proteins different from MreB could still provide the binding substrate or be required for 302 binding -specifically the putative Rod-complex components MreC, MreD, RodA, RodZ, and 303 PBP1a. We therefore constructed depletion strains for RodA, RodZ, or MreCD in a 304 background strain expressing either native levels of PAmCherry-PBP2 (for RodZ, MreCD) or 305 overexpressing PAmCherry-PBP2 (for RodA). Without repression, all of the strains showed 306 normal growth rate (Fig. 5-SI Fig. 1), cell shape (Fig. 5A, Fig. 5-SI Fig. 2), bound fractions 307 (Fig. 5C), and persistent fractions (Fig. 5D). Only the RodZ-depletion strain showed slightly 308 higher bound and persistent fractions. 309 Within 3-5 h after depletion, cells started to lose their rod-like shape (Fig. 5-SI Fig. 2). After 310 growing cells for about two additional doubling times (according to OD), we measured bound 311 and persistent fractions (Fig. 5C-D). At this point, repressed protein levels were reduced well 312 below wildtype levels according to mass spectrometry (Fig. 5-SI Fig. 3A) and cell shape was 313 severely perturbed (Fig. 5B, Fig. 5-SI Fig. 2) while growth rate remained unperturbed 314 ( Fig. 5-SI Fig.1). PBP2 levels remained close to native levels during all experiments except 315 for RodA depletion, where PBP2 remained overexpressed (Fig. 5-SI Fig. 3B). upon RodZ depletion over time, we found that the drop already occurred within about two 326 hours ( Fig. 5-SI Fig. 4), suggesting that RodZ might modulate the rates of PBP2 binding or 327 unbinding. However, after 6 h of depletion, the bound fraction started to increase steadily, 328 demonstrating that RodZ is not strictly required for PBP2 binding. 329 To test whether RodZ might influence the location of PBP2 binding we measured the degree 330 of co-localization of mCherry-PBP2 and RodZ-GFP along the contour of cells expressing 331 both fusions as sole copies of the respective proteins ( Fig. 6A-C, Fig. 6-SI Fig. 1). PBP2 332 and RodZ did only show weak or no visible co-localization, which was also reflected by low also studied the density of RodZ-GFP peaks upon treatment with A22 ( Fig. 6E-F). We found 337 that the pattern of RodZ changed in a similar manner as the pattern of MreB-msfGFP 338 (Fig. 3), while mCherry-PBP2 did not (Fig. 3) (Fig. 5-SI Fig. 5), similarly to long-term treatment with A22 or Mecillinam (Fig. 3E). 353 In all these experiments Rod-complex activity is inhibited or reduced, which changes cell- a Gauss filter of σ = 80 nm) was nearly straight (Fig. 7A). These regions still showed 397 variations of cell-envelope curvature due to bulges or indentations (Fig. 7F). However, we 398 did not find any significant correlations between MreB and contour curvature (Fig. 7D). 399 Therefore, all residual MreB-curvature correlations after removal of poles and septa can be 400 attributed to weak cell bending, while bending-independent bulges and indentations do not 401 affect MreB localization. 402 We confirmed our findings with two alternative approaches: First, we subtracted from the 403 local contour curvature the curvature contribution due to cell bending (Fig. 7-SI Fig. 2A-B).  (Fig. 1F). 424 To test whether PBP2 diffusion was fast enough for the biological cross-linking rate, we 425 conducted computational simulations of freely diffusing PBP2 molecules and measured the 426 average encounter rate between any simulated molecule and a given target site 427 representing a Rod complex (Fig. 8A, B). We found that the rate between encounters was at 428 least three times lower than the rate of cross-linking, even if a single enzyme had a high 429 chance of returning to the same site and conduct multiple cross-linking reactions on average 430 (Fig. 8C). Therefore, free diffusion cannot account for physiological cross-linking rates. 431 As an extension of the model we considered the possibility that PBP2 molecules underwent 432 facilitated diffusion by preferentially diffusing along MreB filaments, which could possibly 433 serve to increase encounter rates between PBP2 and cell-wall-insertion sites. One-434 dimensional diffusion along filaments indeed increased the encounter rate (Fig. 8D). 435 However, due to the preferentially circumferential orientation of MreB filaments, facilitated 436 diffusion would also lead to reduced and asymmetric diffusion (Fig. 8E). On the contrary, we 437 did not observe asymmetric diffusion in our experiments (Fig. 8F). Therefore, Rod-complex 438 activity requires the stable association between transglycosylase and transpeptidase for 439 multiple cross-linking events. Our findings suggest, that only the persistently moving fraction 440 of PBP2 molecules substantially contributes to cross-linking. We found that the bound fraction of PBP2 molecules remained nearly constant upon A22 or 463 Mecillinam treatment (Fig. 3). We thus reasoned that persistently moving and immobile 464 molecules are likely bound to the same substrate. In Gram-negative E. coli active Rod 465 complexes are thought to insert nascent glycan strands in between template strands (Höltje, 466 1998), even if deviations from perfect alignment are reported (Turner et al., 2018). During 467 cell-wall insertion, Rod complexes might therefore stay connected to the local cell wall 468 through associations between PBP2 and a template strand, independently of enzymatic 469 activity. In the future, it will be interesting to study potential interactions at the molecular 470 level. These might then also reveal structural features of the cell wall potentially responsible 471 for stable PBP2 association. 472 PBP2 molecules transition only slowly between bound and diffusive states (Fig. 2). Possibly, 473 PBP2 is found in two different molecular states that facilitate stable binding or allow for 474 overexpressed by two-to four-fold ( Fig. 1-SI Fig. 1D). Yet, the number of persistently 498 moving molecules per cell (bound fraction times persistent fraction) dropped by less than 499 two-fold between wildtype and RodA repression. With a wild-type stoichiometry between 500  (B) Transition rates between immobile and persistently moving states for different protein fusions and expression levels. Circles: biological replicates.
(C) Fluorescence-lifetime distributions of msfGFP-PBP2 trajectories with imaging intervals of 1 s (black solid line) and 12 s (red solid line). Dashed lines represent a joint fit of the two curves to a model of photobleaching and bleaching-independent track termination, the latter comprising unbinding and persistently molecules leaving the TIR field of view (bleaching probability per frame pb = 0.39 ± 0.08, apparent track termination rate ka = 0.035 ± 0.007 s -1 ). Based on a model for persistent motion, we obtained an upper limit of the unbinding rate of kbd < 0.03/s ( Fig. 2-SI Fig. 1). Shaded region: Standard deviation between replicates.  (D) Cartoon of suggested Rod-complex initiation through PBP2: PBP2 (blue) binds to a target site in the cell envelope (white circle) independently of MreB filaments or PBP2 activity. PBP2 then recruits an MreB filament through diffusion and capture (green) or through nucleation, and also recruits other rod-complex components (magenta).  Comparison between correlations obtained from full contours (black) and side walls (green) (C) and from side walls (green) and straight cell segments (magenta) (D).
(E-F) Distributions of contour-curvature values corresponding to correlation plots in (C-D). Shaded region: Standard deviation between replicates.