The outer membrane lipoprotein NlpI nucleates hydrolases within peptidoglycan multi-enzyme complexes in Escherichia coli

The peptidoglycan (PG) sacculus provides bacteria with the mechanical strength to maintain cell shape and resist osmotic stress. Enlargement of the mesh-like sacculus requires the combined activity of PG synthases and hydrolases. In Escherichia coli, the activity of the two bifunctional PG synthases is driven by lipoproteins anchored in the outer membrane. However, the regulation of PG hydrolases is less well understood, with only regulators for PG amidases having been described. Here, we identify the lipoprotein NlpI as a general adaptor protein for PG hydrolases. NlpI binds to different classes of hydrolases and can specifically form multimeric complexes with various PG endopeptidases. In addition, NlpI seems to contribute both to PG elongation and cell division biosynthetic complexes based on its localization and genetic interactions. In line with such a role, we reconstitute PG multi-enzyme complexes containing NlpI, the PG synthesis regulator LpoA, its cognate bifunctional synthase, PBP1A, and different endopeptidases. Our results indicate that PG regulators and adaptors are part of PG biosynthetic multi-enzyme complexes, regulating and potentially coordinating the spatiotemporal action of PG synthases and hydrolases. Significance The activity of PG hydrolases may cause lysis of the bacterial cell if left unregulated. Hence, the cell must have ways of regulating and coordinating their activities. Our current understanding of how this occurs is incomplete. In this work, we present the outer membrane (OM) anchored lipoprotein, NlpI, as a scaffold of peptidoglycan hydrolases. We propose that NlpI facilitates the formation of multi-enzyme complexes and that, along with other regulators, it coordinates a safe enlargement and separation of the PG layer in E. coli.

also strong negative interactions with ∆nlpI; Δfitness ratio of -0.30 and -0.21 279 respectively (Fig. 4b). To analyse whether these strong negative genetic interactions 280 were also reflected in the morphology of the cells, all single and double mutants were 281 grown exponentially and imaged by phase contract microscopy. Combining ∆nlpI with 282 ∆mrcB or ∆lpoB led to abnormal cell morphologies, with cells being 30% wider and up 283 to 80% longer (Fig. 4c). This suggests that the NlpI-EPase complexes might be 284 important for facilitating the formation of the PBP1A-mediated PG machinery. This 285 would be consistent with the changes in thermostability of PBP1A and LpoA in ∆nlpI 286 cells (Fig. 1b). Thus, we next tested the in vitro interactions between NlpI and 287 respective EPases with PBP1A and LpoA. We discovered that PBP1A did not directly 288 interact with NlpI but interacts with low nanomolar range affinities with different 289 EPases, including MepS (apparent KD = 91 ± 39 nM), PBP4 (106 ± 44 nM) and PBP7 290 (101 ± 35 nM) ( Fig. 4d and S6a). PBP4 (315 ± 38 nM) and PBP7 (217 ± 93 nM) bound 291 also to LpoA at slightly higher nanomolar ranges ( Fig. 4d and S6b). These interactions 292 between PG synthases and EPases would allow for PG multienzyme complexes to 293 exist as postulated by Hӧltje (Hӧltje, 1998). 294 295 NlpI is part of a PG multi-enzyme complex with PBP4 and PBP1A/LpoA. To further 296 understand the interaction between PG hydrolases and synthases, we characterized 297 in detail the interactions between PBP4 with PBP1A/LpoA and NlpI by MST. We used 298 a fixed concentration MST assay to show that fluorescently labelled PBP1A and LpoA 299 are able to bind a preformed PBP4-NlpI complex ( Fig. 5a and 5b). Whilst the binding 300 of PBP4 and PBP4-NlpI to fl-PBP1A resulted in an increase in FNorm values (which 301 was not the case in the presence of NlpI alone); binding of PBP4 and PBP4-NlpI to 302 LpoA consistently resulted in an enhanced initial fluorescence. This indicated that the 303 ligand was binding in close proximity to the probe and was affecting the local 304 environment of the fluorophore and subsequently its fluorescence yield. Since the 305 change in fluorescence was a property of ligand binding (Fig. S8b), the raw 306 fluorescence data as opposed to the FNorm values were plotted in this instance (Fig.  307 5b). These consistent increases in fluorescence reflect the binding of PBP4 and PBP4-308 NlpI to LpoA and suggest that the presence of NlpI does not prevent the interaction of 309 PBP4 with LpoA (Fig. 5b). To our knowledge, this is the first evidence that PG-310 synthases and PG-hydrolases form multienzyme complexes with regulatory 311 lipoproteins to possibly coordinate PG-synthesis in Gram-negative bacteria. 312 313 Discussion 314 E. coli contains a repertoire of more than 20 periplasmic hydrolases providing 315 specificity to almost every bond present in PG (Chodisetti & Reddy, 2019;Singh et al., 316 2012;van Heijenoort, 2011;Vollmer et al., 2008b;Yunck et al., 2016). However, with 317 the exception of amidases, it is unclear how these hydrolases are regulated to prevent 318 autolysis (Uehara et al., 2009). This study identifies NlpI as a novel scaffolding protein 319 of EPases that might co-ordinate hydrolases within PG synthesis machineries. NlpI 320 seems to also bind to several other hydrolytic enzymes, including some members of 321 the amidase and LTase families. The details of these interactions will be investigated 322 in future work. 323

Deletion of nlpI impacts envelope biogenesis beyond the proteolytic regulation 324 of MepS levels. NlpI interacts with MepS and targets it for digestion via the protease 325
Prc (Singh et al., 2015). ΔnlpI increases the abundance of MepS (Fig. 1a) along with 326 conferring an OM hypervesiculation phenotype (Schwechheimer et al., 2015;Singh et 327 al., 2015). The hypervesiculation is suppressed in ΔnlpIΔmepS mutants, indicating that it is caused by the increased MepS levels (Schwechheimer et al., 2015). Similarly,329 we show that unregulated MepS is also part of the fitness defect of ΔnlpI, as the double 330 ΔnlpIΔmepS mutant has most of its fitness defect restored. Yet, there are several 331 pieces of evidence that NlpI has additional functions in addition to the proteolytic 332 degradation of MepS. 333 First, the cellular morphology of the ΔnlpIΔmepS mutant is not restored. The 334 ΔnlpI and the ΔmepS mutant are thinner and thicker, respectively, compared to 335 wildtype cells; which is in line with MepS playing an important role for cell elongation 336 (Singh et al., 2012) and ΔnlpI cells having this activity uncontrolled. However, the 337 morphology of the double mutant is exacerbated compared to an ΔmepS mutant, with 338 the cells becoming not only thicker but also more elongated (Fig. 3c). This strongly 339 implies that NlpI has additional functions to the proteolytic degradation of MepS. 340 Second, our biochemical evidence (TPP, affinity chromatography) suggest that NlpI 341 binds to and affects a number of PG related processes, including both PG hydrolysis 342 and synthesis enzymes and their regulators. NlpI seems to bind strongly to amidases 343 and their regulators (AmiA,EnvC;Fig. 1b,d), LTases (MltA,MltC;Fig. 1b,d) and 344 EPases (MepM,MepS,PBP4,PBP7;Fig 2,S2,S3) in the context of PG biosynthetic 345 machineries . This raises the possibility that NlpI scaffolds, or even regulates, 346 several classes of hydrolases beyond its function towards EPases. Moreover, to the 347 best of our knowledge, this is the first evidence that NlpI has additional functions during 348 PG synthesis. 349 350

NlpI interacts with several EPases at physiologically relevant concentrations. 351
Immobilized NlpI retained the EPases PBP4 and PBP7, raising the possibility that NlpI 352 interacts with additional EPases along with MepS (Fig. 1c). Especially since MepS was not part of the proteins being pulled down, and is known to bind to NlpI, we 354 decided to investigate this further. Using MST and pull-down assays, we validated 355 interactions between NlpI and 3 other DD-EPases; MepM, PBP4 and PBP7, all of them 356 with apparent KD or EC50 values in the nanomolar range ( Fig. 2c and S2a). We 357 estimated the concentration of these proteins in the periplasm, assuming cell 358 dimensions of 4.77 × 10 -6 m (length) and 1.084 × 10 -6 m (diameter), with a periplasmic 359 width of 21 × 10 -9 m (Banzhaf et al., 2012;Beveridge, 1995) (Fig. 6a). We conclude 360 that the NlpI-EPases interactions identified in the present work are all, in principle, able 361 to occur in the cell (Fig. 6a). Furthermore, our data showed that NlpI could also affect 362 the activity of some of these EPases; for example, the activity of MepM against intact 363 sacculi, was reduced in the presence of NlpI (Fig. 3a). Further, as NlpI facilitates to 364 proteolytic degradation of MepS (Singh et al., 2015), NlpI could be generally restricting 365 the role of elongation EPases (Singh et al., 2012). 366 With regards to activity of EPases, we note that we were unable to observe the DD-367 EPase activity of MepS, previously reported in (Singh et al., 2012) (Fig. 3a). However, 368 while addition of NlpI had no effect on the activity of PBP4 or PBP7, there was a very 369 slight stimulation of MepS activity against isolated muropeptides in the presence of 370 NlpI, following overnight incubation (Fig. 3a). Overall, these results raise the possibility 371 that NlpI could modulate the activity of specific hydrolases along with its role as a 372 scaffolding protein. Since NlpI seems able to bind several of the hydrolytic enzymes 373 simultaneously (at least different EPases); its role in the regulation of activity may 374 become clearer when probed in the context of these multimeric complexes. 375 376 NlpI scaffolds multiprotein complexes with PG hydrolytic enzymes within the 377 context of PG biosynthesis machineries. NlpI is able to form trimeric complexes 378 with different EPases that lack mutual interactions. Examples of such complexes 379 resolved in the present work are MepS-NlpI-PBP4 and MepS-NlpI-PBP7 ( Fig. 2c and  380 2d). Since NlpI has four helix-turn-helix TPR-like repeats per monomer, it remains to 381 be seen whether the different TPR helixes are specific for different binding partners 382 (Wilson et al., 2005) and/or different type of hydrolytic enzymes. Nevertheless, the 383 ability of NlpI to bind multiple ligands simultaneously would be consistent with the idea 384 that TPR domains facilitate the formation of multi-protein complexes (Blatch & Lassle, 385 1999;Cortajarena & Regan, 2006). In this sense, NlpI is more promiscuous in nature 386 than the previously identified amidase regulators EnvC and NlpD, which appear to 387 have specificity to their cognate amidases (Uehara et al., 2009). Despite binary 388 interactions between various EPases and PBP1A/LpoA being able to occur in the 389 absence of NlpI (Fig. 4d), we hypothesize that NlpI could also sequester additional or 390 specific sets of EPases and other hydrolytic enzymes, determining the specificity of 391 such synthetic machineries. Accordingly, our finding that PBP4 is able to 392 simultaneously bind PBP1A/LpoA and NlpI supports the idea that NlpI could 393 specifically scaffold hydrolases at active PG synthases ( Fig. 5a and 5b). The ability of 394 an OM-anchored NlpI to complex EPases and other hydrolases would not only serve 395 to locally concentrate those enzymes near PG-synthesis complexes, but also to 396 maintain the active hydrolases in the space between the PG layer and OM; facilitating 397 cleavage of the mature PG of the sacculus and keeping them at distance to the newly 398 synthesized PG, which emerges between the CM and PG layer and is not subject to 399 turnover. NlpI molecules are outnumbered by the amount of potential binding partners 400 in the periplasm, so it is unlikely that there is an abundance of free NlpI (Fig. 6a). 401 EPase regulation might occur on the level of binding affinity to NlpI and its TPR-like 402 domains. This would see NlpI resembling a "dock" for EPases (and possibly other hydrolases) to make them available for PG synthesis complexes when needed. Such 404 a system would allow for greater flexibility, as NlpI interacts with many hydrolases. In 405 contrast, the specificity could be encoded on the level of the hydrolases. As 406 demonstrated, EPases interact directly with PG-synthases, but those interactions 407 might be specific to them particular EPases (and no other hydrolases) and/or might be 408 subject to environmental cues or to competition for the same binding site. Therefore 409 NlpI could be a more general adaptor of hydrolases, as suggested by its interactions 410 with amidases and LTases (Fig. 1b, d), bringing a set of hydrolases to biosynthetic 411 complexes. This hypothesis will require more work in the future to ascertain. 412 Around 20 years ago, Hӧltje hypothesized that growth of the PG sacculus 413 requires both synthases and hydrolases working in tandem to enable a safe and 414 coordinated enlargement (Hӧltje, 1998). However, it has also been suggested that 415 EPases are not necessarily part of multi-protein complexes; as overproduction of three 416 different EPases confers mecillinam resistance (Lai et al., 2017). In this work, we 417 provide the first evidence of interactions between PBP1A/LpoA with PBP4 and 418 hypothesize that interactions between NlpI and other EPases could facilitate their 419 delivery to PG-synthesis complexes during PG growth. The existence of PG multi-420 protein complexes is not necessarily contrasting the idea that EPases and/or NlpI-421 EPase complexes may in part localize outside of such PG assembly machineries. This 422 work, and the work of others supports the idea that PG multi-protein complexes are 423 highly dynamic and driven by transient protein-protein interactions (Pazos et al., 2017). 424 In addition, the existence of such PG multi-protein complexes is in line with the recent 425 isolation of an 1 MDa cell division complex (Trip & Scheffers, 2015). 426 NlpI functions together with the PBP1A/LpoA PG machinery. We studied the 428 localization of NlpI to infer if NlpI scaffolds complexes exclusively for cell elongation or 429 division. The localization pattern of NlpI is spotty and diffusive with no enrichment at 430 the mid-cell (Fig 4a). NlpI was previously shown to be located in the OM of bacterial 431 cells, containing an N-terminal cysteine residue at position 19 that is likely the target 432 for lipoprotein modification (Ohara et al., 1999;Teng et al., 2010). The subsequent N-433 acyl-S-sn-1,2-diacylglyceryl-cysteine is processed, culminating in the tethering of the 434 mature protein to the inner leaflet of the OM (Noland et al., 2017;Wilson et al., 2005). 435 It is hence also possible that interactions between NlpI and hydrolases concentrate 436 and facilitate cleavage from the outer face of the PG layer. Its disperse localization 437 would enable binding of EPases involved in both division and elongation. NlpI was 438 shown to bind a number of essential divisome proteins at high salt concentrations in 439 our affinity chromatography experiment (Fig 1d), suggesting that it is at least a 440 transient member of the divisome. In addition, the negative genetic interaction of nlpI 441 with mrcA (PBP1A) and mrcB (PBP1B) raises the possibility that NlpI can affect both 442 the elongasome and the divisome. This is in line with both PBP1B/LpoB and 443 PBP1A/LpoA complexes showing changes in thermostability in ΔnlpI cells (Fig. 1b). 444 However, because the genetic interactions of nlpI with mrcB (PBP1B) were stronger 445 and led to a near synthetic lethality, we reasoned that NlpI predominantly worked with 446 the PBP1A/LpoA machinery. Cells lacking PBP1B depend on a functional 447 PBP1A/LpoA complex to achieve growth (Yousif et al., 1985). An alternative scenario 448 is that cells with only PBP1A/LpoA are more sensitive to genetic and chemical 449 perturbations in the cell wall than cells with only PBP1B/LpoB, because latter is more 450 efficient. Although we found no direct interaction between NlpI with PBP1A or LpoA 451 by MST assay (Fig. 4d and S6), a complex of NlpI-PBP4-PBP1A could be formed with PBP4 as the linking protein (Fig. 5). The multitude of interactions between 453 PBP1A/LpoA, different EPases and NlpI ( Fig. 4d and S6) could enable the formation 454 of different active synthase-hydrolase complexes under a range of growth conditions 455 or availability of particular proteins (Fig. S6) (Pazos et al., 2017). 456 In conclusion, this work provides the first evidence for NlpI as a novel adaptor 457 of EPases, and possibly other classes of PG hydrolases. NlpI is likely involved in co-458 ordinating PG-multienzyme complexes by way of nucleating complexes between 459 synthases, EPases and NlpI (Fig. 6b). BW25113 transformants carrying a Red helper plasmid were grown in 5-ml 500 SOB cultures with ampicillin and l-arabinose at 30°C to an OD600 of ≈0.6 and then 501 made electrocompetent by concentrating 100-fold and washing three times with ice-502 cold 10% glycerol. PCR products were gel-purified, digested with DpnI, repurified, and 503 suspended in elution buffer (10 mM Tris, pH 8.0). Electroporation was done by using 504 a Cell-Porator with a voltage booster and 0.15-cm chambers according to the 505 manufacturer's instructions (GIBCO/BRL) by using 25 μl of cells and 10-100 ng of 506 PCR product. Shocked cells were added to 1 ml SOC, incubated 1 h at 37°C, and then 507 one-half of the incubation/cells were spread onto agar to select Km R transformants. 508 They were colony-purified once non-selectively at 37°C and then tested for ampicillin 509 sensitivity to test for loss of the helper plasmid. 510 511 Eliminating Antibiotic Resistance Gene for the NlpI-HA. pCP20 is an ampicillin and Cm R plasmid that shows temperature-sensitive replication 513 and thermal induction of FLP synthesis (Cherepanov & Wackernagel, 1995). Km R 514 mutants were transformed with pCP20, and ampicillin-resistant transformants were 515 selected at 30°C, after which a few were colony-purified once non-selectively at 43°C 516 and then tested for loss of all antibiotic resistances. The majority of the mutants lost 517 the FRT-flanked resistance gene and the FLP helper plasmid simultaneously. 518 519 Immunolabeling 520 After reaching steady state, the cells were fixed for 15 min by addition of a mixture of 521 formaldehyde (f.c. 2.8%) and glutaraldehyde (f.c. 0.04%) to the shaking water bath 522 and immunolabeled as described (Buddelmeijer et al., 2013) with rabbit polyclonal 523 antibodies against NlpI or against the HA-tag. As secondary antibody, donkey anti-524 rabbit conjugated to Cy3 or conjugated to Alexa488 (Jackson Immunochemistry, USA) 525 diluted 1:300 in blocking buffer (0.5% (wt/vol) blocking reagents (Boehringer,526 Mannheim, Germany) in PBS) was used, and the samples were incubated for 30 min 527 at 37°C. For immunolocalization, cells were immobilized on 1% agarose in water slabs 528 coated object glasses as described (Koppelman et al.,2004)  Phase contrast and fluorescence images were combined into hyperstacks using 542 ImageJ (http://imagej.nih.gov/ij/) and these were linked to the project file of Coli-543 Inspector running in combination with the plugin ObjectJ 544 (https://sils.fnwi.uva.nl/bcb/objectj/). The images were scaled to 15.28 pixel per μm. 545 The fluorescence background has been subtracted using the modal values from the 546 fluorescence images before analysis. Slight misalignment of fluorescence with respect 547 to the cell contours as found in phase contrast was corrected using Fast-Fourier 548 techniques as described in (Vischer et al., 2015). Data analysis was performed as 549 described in (Vischer et al., 2015). In brief, midcell was defined as the central part of 550 the cell comprising 0.8 µm of the axis. From either cell part, midcell and remaining cell, 551 the volume, the integrated fluorescence, and thus the concentration of fluorophores 552 was calculated. The difference of the two concentrations is multiplied with the volume 553 of midcell. It yields FCPlus (surplus of fluorescence). For age calculation, all cell 554 lengths are sorted in ascending order. Then the equation: 555 age = ln(1 -0.5 * rank / (nCells -1)) / ln(0.5) 556 is used, where rank is a cell's index in the sorted array, nCells is the total amount of 557 cells and age is the cell's age expressed in the range 0 to 1. 558 559 Ni 2+ -NTA pulldown assay 560 His-tagged proteins of interest were incubated with untagged or native ligands, in the 561 presence of Ni 2+ -NTA coupled agarose beads (Qiagen). Beads were pre-equilibrated 562 with dH2O and binding buffer (10 mM HEPES/NaOH, 10 mM MgCl2, 150 mM, NaCl 563 0.05% Triton X-100, pH 7.5) by centrifugation at 4000 × g, 4 min at 4⁰C. Samples were 564 incubated overnight on a spinning plate at 4⁰C before beads were washed 3-6 times 565 with 10 mM HEPES/NaOH, 10 mM MgCl2, 150 mM, NaCl 0.05% Triton X-100, 30 mM 566 Imidazole, pH 7.5. Retained material was eluted from Ni 2+ -NTA beads using proteus 567 spin columns and boiling at 100°C in SDS-buffer (50 mM Tris/HCl pH 6.8, 2% SDS, 568 10% glycerol 0.02% bromophenol blue, 10% β-mercaptoethanol). Elutions were 569 diluted 1:1 with dH2O and proteins were separated by SDS-PAGE for analysis. 570 571

Protein overexpression and purification 572
Prior to purification, plasmids of interest were transformed into E. coli strain BL21 573 (λDE3) and grown overnight in LB agar (1.5% w/v) containing appropriate antibiotic, 574 at 37⁰C. Transformants were inoculated into 50 ml of LB with appropriate antibiotic 575 and grown overnight at 37⁰C, shaking. Pre-cultures were diluted 1:40 in 2 L LB and 576 grown to OD578 0.5-0.6, at 37⁰C. Induction conditions are specified for each respective 577 protein below. After overexpression, cells were harvested by centrifugation at 7500 × 578 g, 15 min, 4⁰C. Pellets were re-suspended in buffer I (25 mM Tris/HCl, 300 mM NaCl, 579 pH 7.5) with the addition of a small amount of DNase (Sigma) and 100 μM P.I.C and 580 PMSF. Cells were lysed by sonication (Branson digital) and the lysate was centrifuged 581 at 14000 × g, 1 h, 4⁰C, before the supernatant was applied at 1 ml/min to a 5 ml 582 chromatography column attached to an ÄKTA Prime plus (GE Healthcare). 583 If desired, the removal of his-tags for tagged constructs, following IMAC steps, 584 was achieved by incubating protein samples with 1 unit/ml of restriction grade thrombin 585 (Novagen). This was carried out overnight at 4⁰C in 25 mM Tris/HCl, 200 mM NaCl, 586 pH 8.0 or 25 mM HEPES/NaOH, 300 mM NaCl, 10% glycerol, pH 7.5, depending on the next purification step. Removal of His-tag was verified by western blot with 588 monoclonal α-His -HRP (1:10000) antibody (Sigma NaCl, pH 6.8. Protein was applied at 1 ml/min to a 5 ml ceramic hydroxyapatite column 635 (BioRad Bioscale TM ) in the dialysis buffer. Fractionation of proteins was achieved by 636 using a linear gradient of buffer 2 (500 mM Potassium phosphate, 300 mM NaCl, pH 6.8) over a 50 ml gradient. Fractions of highest purity and yield were dialyzed overnight 638 against 25 mM HEPES/NaOH, 300 mM NaCl, 10% glycerol, pH 7.5 and concentrated 639 to ~ 5ml using Vivaspin concentrator spin columns (Sartorius). Protein sample was 640 applied to a HiLoad 16/600 Superdex 200 column (GE healthcare) at 1 ml/min pre-641 equilibrated with dH2O and buffer I (25 mM HEPES/NaOH, 300 mM NaCl, 10% 642 glycerol, pH 7.5). Samples were analyzed by SDS-PAGE and fractions containing 643 purified protein was pooled and stored at -80⁰C. 644 645 Purification of PBP7 646 PBP7 overproduction was induced with 1 mM IPTG for 3 h at 30⁰C before being 647 harvested by centrifugation as described above and re-suspended in buffer I (25 mM 648 Tris/HCl, 500 mM NaCl, 20 mM Imidazole, pH 7.5). Following sonication and 649 subsequent centrifugation as described above, the lysate was applied to a 5 ml 650 HisTrap HP column (GE healthcare) and washed with 4 column volumes of buffer I; 651 before bound protein was eluted with buffer II (25 mM Tris/HCl, 300 mM NaCl, 400 652 mM Imidazole pH 7.5). Samples were analyzed by SDS PAGE and dialyzed overnight 653 against 25 mM HEPES/NaOH, 300 mM NaCl, 10% glycerol, pH 7.5 before being 654 concentrated to ~ 5ml using Vivaspin concentrator spin columns (Sartorius) at 4500 × 655 ɡ, 4⁰C. Protein samples were applied to a HiLoad 16/600 Superdex 200 column (GE 656 healthcare) at 1 ml/min pre-equilibrated with dH2O and buffer I (25 mM HEPES/NaOH, 657 300 mM NaCl, 10% glycerol, pH 7.5). Samples were analyzed by SDS-PAGE and the 658 purest fractions with highest yield were pooled and stored at -80⁰C. 659 660
were terminated by boiling at 100⁰C for 5 min and the muropeptide substrates obtained 737 by centrifugation at 10 000 × g for 5 min, RT and taking the supernatant. Reactions 738 were then carried out and prepared for analysis as described above. 739 740

Purification of anti-NlpI 741
This protocol was adapted from a previously published method (Banzhaf et al., 2012). 742 Serum against NlpI was obtained from rabbits at Eurogentec (Herstal, Belgium), using 743 purified oligohistidine-tagged NlpI protein for immunization. For affinity purification of 744 the serum, purified His-NlpI (5 mg The NlpI antibodies were eluted with 10 ml of elution buffer I and mixed with 2 ml of 756 elution buffer II (2 M Tris/HCl, pH 8.0) afterwards. The elution was analyzed by SDS-757 PAGE and glycerol was added to a final concentration of 20% and the purified PBP2 758 antibodies were stored at -20°C. Anti-NlpI was tested for specificity by Western Blot 759 ( Figure S5C). 760

Affinity chromatography 780
This protocol was adapted from a previously published method (Vollmer et al., 1999). 781 Sepharose beads were activated following the instructions of the manufacturer (GE). 782 Coupling of 2 mg of protein to 0.13 g of activated sepharose beads was carried out 783 overnight at 6°C with gentle agitation in protein buffer. After washing the beads with 784 protein buffer, the remaining coupling sites were blocked by incubation in AC blocking 785 buffer (200 mM Tris/HCl, 10 mM MgCl2 500 mM NaCl, 10% glycerol and 0.25% Triton X-100, pH 7.4) with gentle agitation over-night at 6°C. The beads were washed 787 alternatingly with AC blocking buffer and AC acetate buffer (100 mM sodium acetate, 788 10 mM MgCl2, 500 mM NaCl, 10% glycerol and 0.25% Triton X-100, pH 4.8), and 789 finally re-suspended in AC buffer I (10 mM Tris/maleate, 10 mM MgCl2, 50 mM NaCl, 790 1% Triton X-100, pH 6.8). As control (Tris-Sepharose) one batch of activated 791 Sepharose beads was treated identically with the exception that no protein was added. 792 Affinity chromatography was performed at 6°C. E. coli membrane fraction extracted 793 out of 2 L per sample (see above) containing 50 mM NaCl (or 400 mM NaCl for high 794 salt chromatography) was incubated with gentle agitation over-night. The column was 795 washed with 10 ml of AC wash buffer (10 mM Tris/maleate, 10 mM MgCl2, 50 mM 796 NaCl and 0.05% Triton X-100, pH 6.8). Retained proteins were eluted with 20 ml of 797 AC elution buffer I (10 mM Tris/maleate, 10 mM MgCl2, 150 mM NaCl, 0.05% Triton 798 X-100, pH 6.8) followed by a second elution step with 1 ml of AC elution buffer II 799 (10 mM Tris/maleate, 10 mM MgCl2, 1 M NaCl, 0.05% Triton X-100, pH 6.8). Both 800 elution fractions were stored at -20°C. For the high salt affinity chromatography the 801 AC high salt wash buffer (10 mM Tris/ maleate, 10 mM MgCl2, 400 mM NaCl and 802 0.05% Triton X-100, pH 6.8) and the AC high salt elution buffer (10 mM Tris/maleate, 803 10 mM MgCl2, 2 M NaCl, 0.05% Triton X-100, pH 6.8) were used. 0.1% formic acid. In total 1 μg of peptide was separated with a nanoACQUITY UPLC 810 system (Waters) fitted with a trapping column (nanoAcquity Symmetry C18; 5 μm and were analyzed by electrospray ionization-tandem mass spectrometry on an 815 Orbitrap Velos Pro (Thermo Fisher Scientific). Full-scan spectra from a mass/charge 816 ratio of 300 to one of 1,700 at a resolution of 30,000 full width at half maximum were 817 acquired in the Orbitrap mass spectrometer. From each full-scan spectrum, the 15 818 ions with the highest intensity were selected for fragmentation in the ion trap. A lock-819 mass correction with a background ion (mass/charge ratio, 445.12003) was applied. 820 The raw mass spectrometry data was processed with MaxQuant (v1.5.2.8) (Cox & 821 Mann, 2008)  batch-corrected using the limma package (Ritchie et al., 2015) and then normalized with the vsn package (Huber et al., 2002). Individual normalization coefficients were 837 estimated for each biological condition separately. Limma was used again to test the 838 normalized data for differential expression. Proteins were classified as a 'hit' with a 839 log2 fold change higher than 4 and a 'candidate' with a log2 fold change higher than 840

841
Data availability: The mass spectrometry proteomics data will be deposited to 842 (observed) across replicates. The probability that the two means (expected and 860 observed) are equal across replicates is obtained by a Student's two-sample t-test. 861

Thermal proteome profiling and sample preparation 863
Thermal proteome profiling was performed as previously described in (Mateus et al., 864 2018). Briefly, bacterial cells were grown overnight at 37ºC in lysogeny broth, and 865 diluted 100-fold into 20 ml of fresh medium. Cultures were grown aerobically at 37ºC 866 with shaking until optical density at 578 nm (OD578) ~0.5. Cells were then pelleted at 867 4000 × g for 5 min, washed with 10 ml PBS, re-suspended in the same buffer to an 868 OD578 of 10, and aliquoted to a PCR plate. The plate was subjected to a temperature 869 gradient for 3 minutes in a PCR machine (Agilent SureCycler 8800), followed by 3 870 minutes at room temperature. Cells were lysed with lysis buffer (final concentration: 871 50 µg/ml lysozyme, 0.8% NP40, 1X protease inhibitor (Roche), 250 U/ml benzonase 872 and 1 mM MgCl2 in PBS) for 20 min, shaking at room temperature, followed by three 873 freeze-thaw cycles. Protein aggregates were then removed and the soluble fraction 874 was digested according to a modified SP3 protocol (Mateus et al., 2018). Peptides 875 were labelled with TMT6plex (ThermoFisher Scientific), desalted with solid-phase 876 extraction on a Waters OASIS HLB µElution Plate (30 µm), and fractionated onto six 877 fractions on a reversed phase C18 system running under high pH conditions. 878 879

2D-TPP mass spectrometry-based proteomics 880
Samples were analyzed with liquid chromatography coupled to tandem mass 881 spectrometry, as previously described (Mateus et al., 2018). Briefly, peptides were 882 separated using an UltiMate 3000 RSLC nano LC system (Thermo Fisher Scientific) 883 equipped with a trapping cartridge (Precolumn C18 PepMap 100, 5µm, 300 µm i.d. x 884 5 mm, 100 Å) and an analytical column (Waters nanoEase HSS C18 T3, 75 µm x 25 885 cm, 1.8 µm, 100 Å). Solvent A was 0.1% formic acid in LC-MS grade water and solvent B was 0.1% formic acid in LC-MS grade acetonitrile. After loading the peptides onto 887 the trapping cartridge (30 µL/min of solvent A for 3 min), elution was performed with a 888 constant flow of 0.3 µL/min using a 60 to 120 min analysis time (with a 2-28%B elution, 889 followed by an increase to 40%B, and re-equilibration to initial conditions). The LC 890 system was directly coupled to a Q Exactive Plus mass spectrometer (Thermo Fisher 891 Scientific) using a Nanospray-Flex ion source and a Pico-Tip Emitter 360 µm OD x 20 892 µm ID; 10 µm tip (New Objective). The mass spectrometer was operated in positive 893 ion mode with a spray voltage of 2.3 kV and capillary temperature of 320°C. Full scan 894 MS spectra with a mass range of 375-1200 m/z were acquired in profile mode using 895 a resolution of 70,000 (maximum fill time of 250 ms or a maximum of 3e6 ions 896 (automatic gain control, AGC)). Fragmentation was triggered for the top 10 peaks with 897 charge 2 to 4 on the MS scan (data-dependent acquisition) with a 30 s dynamic 898 exclusion window (normalized collision energy was 32), and MS/MS spectra were 899 acquired in profile mode with a resolution of 35,000 (maximum fill time of 120 ms or 900 an AGC target of 2e5 ions). 901 902

2D-TT data analysis 903
Protein identification and quantification. Mass spectrometry data were processed as 904 previously described (Mateus et al., 2018). Briefly, raw mass spectrometry files were 905 processed with IsobarQuant (Franken et al., 2015) and peptide and protein 906 identification were performed with Mascot 2.4 (Matrix Science) against the E. coli 907 (strain K12) Uniprot FASTA (Proteome ID: UP000000625), modified to include known 908 contaminants and the reversed protein sequences (search parameters: trypsin; 909 missed cleavages 3; peptide tolerance 10ppm; MS/MS tolerance 0.02Da; fixed 910 modifications were carbamidomethyl on cysteines and TMT10plex on lysine; variable 911 modifications included acetylation on protein N-terminus, oxidation of methionine and 912 TMT10plex on peptide N-termini). 913 Thermal proteome profiling analysis. Data analysis was performed in R, as 914 previously described in (Mateus et al., 2018). Briefly, all output data from IsobarQuant 915 was normalized using variance stabilization (vsn) (Huber et al., 2002). Abundance and 916 stability scores were calculated with a bootstrap algorithm (Becher et al., 2018), 917 together with a local FDR that describes the quality and the reproducibility of the score 918 values (by taking into account the variance between replicates). A local FDR<0.05 and 919 a minimum absolute score of 10 were set as thresholds for significance. Abundance 920 and stability scores of knocked out genes were discarded. 921 Data availability: The mass spectrometry proteomics data will be deposited to 922 the ProteomeXchange Consortium via the PRIDE partner repository during the 923 publication process.         the soluble components were labelled, combined and quantified by LC-MS, using the 63 published 2D-TPP protocol (Mateus et al., 2018). Here shown volcano plots of two 64 replicates: protein abundance (a) and thermostability (b  NlpI moderately affects the activity of MepM, but less so the activity of MepS, PBP4 107 and PBP7. Representative HPLC chromatograms of assays containing the 108 respective EPase with or without NlpI. Control samples were incubated with no 109 enzyme. Muropeptides were separated and the PG profiles determined as previously 110 described (Glauner, 1988). Fixed concentration MST assays with LpoA-PBP4-NlpI showed differences in initial 181 fluorescence. These differences were eliminated upon boiling of samples in SDS 182 (which abolishes ligand binding), which suggests that the initial differences were due 183 to ligand binding and not due to inaccurate pipetting, validating the use of the raw 184 fluorescence data to plot binding curves.