Real time monitoring of peptidoglycan synthesis by membrane-reconstituted penicillin binding proteins

Peptidoglycan is an essential component of the bacterial cell envelope that surrounds the cytoplasmic membrane to protect the cell from osmotic lysis. Important antibiotics such as β-lactams and glycopeptides target peptidoglycan biosynthesis. Class A penicillin binding proteins are bifunctional membrane-bound peptidoglycan synthases that polymerize glycan chains and connect adjacent stem peptides by transpeptidation. How these enzymes work in their physiological membrane environment is poorly understood. Here we developed a novel FRET-based assay to follow in real time both reactions of class A PBPs reconstituted in liposomes or supported lipid bilayers and we demonstrate this assay with PBP1B homologues from Escherichia coli, Pseudomonas aeruginosa and Acinetobacter baumannii in the presence or absence of their cognate lipoprotein activator. Our assay allows unravelling the mechanisms of peptidoglycan synthesis in a lipid-bilayer environment and can be further developed to be used for high throughput screening for new antimicrobials.

FRET-based assay to follow in real time both reactions of class A PBPs reconstituted in 23 liposomes or supported lipid bilayers and we demonstrate this assay with PBP1B homologues 24 from Escherichia coli, Pseudomonas aeruginosa and Acinetobacter baumannii in the 25 presence or absence of their cognate lipoprotein activator. Our assay allows unravelling the 26 mechanisms of peptidoglycan synthesis in a lipid-bilayer environment and can be further 27 developed to be used for high throughput screening for new antimicrobials.

INTRODUCTION
FRET increases fast and reaches a high level. 137 To confirm that the formation of peptide crosslinks is required to produce substantial 138 FRET in the absence of LpoB, we analysed the PG synthesised by PBP1B Ec from 139 radioactively labelled lipid II and the two fluorescent lipid II analogues ( Figure 1E-G). We cross-linked muropeptide dimers, but not the rate of lipid II consumption (peak 2) ( Figure   146 1G). Overall, we conclude that the FRET assay is capable of reporting GTase activity alone, 147 but the overall FRET signal is dominated by the TPase activity. supplement 2C-D) revealed that SLB-reconstituted PBP1B Ec produced crosslinked PG while, 269 importantly, the wash fraction contained no PG products, confirming that the PG synthesis 270 occurred on the SLBs and this PG remained attached to the bilayer. The fraction of 271 membrane-attached radioactivity was almost the same (33%) when PBP1B Ec was not present 272 in the bilayer, indicating that PBP1B Ec did not affect lipid II-binding to the bilayer. GTase-defective PBP1B Ec version (E233Q) was used ( Figure 4C). In addition, the FRET 291 signal was abolished when the muramidase cellosyl was added after the PG synthesis reaction 292 ( Figure 4C). These results imply that the FRET signal detected by microscopy is primarily 293 due to the transpeptidase activity of PBP1B Ec , in agreement with the results obtained on 294 liposomes ( Figure 2C). As our experiments confirmed that the PG synthesized by PBP1B Ec on SLBs remained 298 attached to the bilayer, we next analysed the lateral diffusion of lipid II-Atto647n and its products during PG synthesis reactions. We first analysed the recovery of fluorescence 300 intensity after photobleaching to monitor the diffusion of lipid II-Atto647n during PG 301 synthesis ( Figure 4D). Only when crosslinking was permitted (absence of ampicillin), the 302 diffusion coefficient of lipid II-Atto647n decreased 2 to 3-fold in a time-dependent manner.

303
The time needed to reach the minimum diffusivity value (~10 min) was similar to the lag 304 detected in the increase of FRET efficiency ( Figure 4B). The fraction of immobile lipid II-

313
This shows that the membrane fluidity was not altered by the PG synthesis reaction and 314 therefore was not the cause of the change in lipid II diffusivity upon transpeptidation. As the 315 immobile fraction of labelled lipid II did not increase after PG synthesis and the diffusion 316 was reduced only 2 to 3-fold, we concluded that lipid II-Atto647n was incorporated into 317 small groups of crosslinked glycan chains which can still diffuse on the bilayer.

318
In summary, we report the incorporation of active PBP1B Ec into supported lipid 319 bilayers, where we could track a decrease in the diffusion of the protein and its substrate 320 during PG synthesis reactions. Using this system we detected an increase in FRET upon 321 initiation of PG synthesis, only occurring when transpeptidation was not inhibited.  For all PBPs and conditions tested, FRET increased when only the GTase domain was active 331 (i.e. when FRET occurred between probes incorporated along the same strand), but the FRET 332 signal was always higher when transpeptidase was active (Figures 1, 2 and Figure 2 -figure supplements 6 and 7). For detergent-solubilised PBP1B Ec , the FRET curve closely followed 334 the rate of the production of cross-linked PG as determined by HPLC analysis of the products 335 ( Figure 1E-G). These results suggest that inter-chain FRET (arising from both fluorophores 336 present on different, adjacent glycan chains) was a main component of the total FRET signal.

337
Why is this the case? FRET depends on the distance and orientation of the two probes. It 338 might be sterically unfavourable that two large Atto550 and Atto647n containing lipid II 339 molecules simultaneously occupy the donor and acceptor sites in the GTase domain (van't 340 Veer, 2016), preventing the incorporation of probes (and high FRET) at successive subunits 341 on a single glycan chain. Indeed, for all PBPs tested either in detergents or liposomes, the 342 incorporation of labelled lipid II into glycan chains was more efficient when unlabelled lipid 343 II was present and for most enzymes an activator was required to polymerize glycan chains 344 using labelled lipid II in the absence of unlabelled lipid II. We thus hypothesize that the 345 TPase activity brings glycan chains to close proximity, reducing the distance between probes 346 sufficiently to produce high levels of FRET ( Figure 5).  et al., 2005). Interestingly, LpoB-activated PBP1B produces a hyper-359 crosslinked PG (Typas et al., 2010, Egan et al., 2018, suggesting that LpoB stimulates the 360 TPase more than the GTase. In the cell, a protein associated with the Tol system, CpoB, showed that both, LpoP Ab and LpoP Pa significantly activated their cognate PBP1B.

373
Interestingly, LpoP Ab stimulated the GTase and not TPase of PBP1B Ab while LpoP Pa 374 stimulated both activities in PBP1B Pa which may illustrate how different species have tailored 375 their activators to their specific needs. Importantly, PBP1B Ab TPase activity was higher in 376 liposomes than in detergents, which serves as a reminder that detergents are not always 377 neutral solubilising agents and they can affect the activity of membrane proteins.

379
Towards single-molecule PG synthesis 380 We also adapted the FRET assay to supported lipid bilayers and super resolution microscopy 381 to study how PBP1B Ec polymerizes PG on SLBs ( Figure 4). As with the liposome assays, we 382 detected an increase in FRET signal upon triggering PG synthesis that correlated with 383 transpeptidation. Importantly we could follow the diffusion of the reaction products, which 384 indicates that PBP1B Ec does not completely cover the surfaces with a layer of PG but instead 385 produced smaller patches of cross-linked glycan chains. We attribute this to the fact that 386 PBP1B Ec was reconstituted at a very low density in order to ensure the homogeneity and   (Kumar et al., 2014). Moreover, it is uses radioactivity detection and is not amenable to microscopy, in contrast to methods based on fluorescently-labelled substrates. An 402 important advantage of our new assay over other real-time TPase assays is that it uses natural 403 substrates for transpeptidation,, i.e. nascent glycan strands, instead of mimics of the 404 pentapeptide, and its ability to measure the activities in a natural lipid environment.

405
Our new FRET assay can potentially be adopted to assay PG synthases in the 406 presence of interacting proteins, for example monofunctional class B PBPs in the presence of 407 monofunctional GTases (cognate SEDS proteins or Mtg proteins) or interacting class A PBPs 408 (Meeske et al., 2016, Bertsche et al., 2006, Sjodt et al., 2020, Derouaux et al., 2008, Banzhaf 409 et al., 2012, Sjodt et al., 2018. In addition, our assay has the potential to be adopted to high   (Breukink et al., 2003, Bertsche et al., 2005. Lipid II-Atto550 and Lipid II-  Digested protein was applied to a 5 mL HiTrap Q HP column (GE Healthcare) at 0.5 mL/min.     Dy647 was reconstituted in EcPL SLBs at a 1:10 6 (mol:mol) protein to lipid ratio and was 1306 tracked using single-molecule TIRF before or after the addition of 1.5 µM lipid II. Images 1307 were taken with a rate of 62 ms per frame.

1311
Membranes were incubated with 5 µM lipid II in the presence or absence of 1 mM ampicillin.

1312
To detect FRET, the fluorescence of the acceptor Atto647n was bleached within a region. In 1313 the subsequent frame the fluorescence of Atto550 increased indicating the presence of FRET.