Dual targeting of the class V lanthipeptide antibiotic cacaoidin

Summary

Antibiotic resistance is reaching alarming levels, demanding for the discovery and development of antibiotics with novel chemistry and mechanisms of action. The recently discovered antibiotic cacaoidin combines the characteristic lanthionine residue of lanthipeptides and the linaridin-specific N-terminal dimethylation in an unprecedented N-dimethyl lanthionine ring, being therefore designated as the first class V lanthipeptide (lanthidin). Further notable features include the high D-amino acid content and a unique disaccharide substitution attached to the tyrosine residue. Cacaoidin shows antimicrobial activity against gram-positive pathogens and was shown to interfere with peptidoglycan biosynthesis. Initial investigations indicated an interaction with the peptidoglycan precursor lipid II_PGN as described for several lanthipeptides. Using a combination of biochemical and molecular interaction studies we provide evidence that cacaoidin is the first natural product demonstrated to exhibit a dual mode of action combining binding to lipid II_PGN and direct inhibition of cell wall transglycosylases.

Graphical abstract

Highlights

• •The lanthidin cacaoidin (CAO) contains a unique disaccharide substitution
• •CAO inhibits peptidoglycan biosynthesis by binding to lipid II
• •CAO binds to transglycosylases and interferes with peptidoglycan polymerization

Microbiology; Bacteriology

Introduction

Ribosomally synthesized and post-translationally modified peptides (RiPPs) represent a family of structurally diverse natural products with a variety of activities including antifungal, antibacterial or antiviral.^1,2,3,4 Lanthipeptides with antibiotic activity, so-called lantibiotics,⁵ are among the best characterized classes of RiPPs. Structurally distinct features of this family include the presence of the thioester amino acids lanthionine (Lan) and/or methyllanthionine (MeLan), as well as dehydroamino acids which result from extensive post-translational modifications.^6,7

The recently discovered cacaoidin (CAO) is the first member of a new family of RiPPs termed lanthidins,⁸ sharing characteristic features of the lanthipeptide and the linaridin family, i.e., a thioether-based lanthionine residue and a linaridin-specific dimethylated N-terminus, respectively. Further analysis of its biosynthetic gene cluster has designated it as the first class V lanthipeptide.^6,7,8,9

CAO is a 23 amino acid peptide produced by Streptomyces cacaoi characterized by the presence of a C-terminal S-[(Z)-2-aminovinyl-(3S)-3-methyl]-d-cysteine (AviMeCys) and a glycosylated tyrosine residue rarely found in natural products, with mannopeptimycins being the only representatives from bacteria reported so far.¹⁰ In contrast to representatives of class I-IV lanthipeptides, CAO contains only a single N-dimethylated Lan residue located at the N-terminus (Figure 1). The latter is a characteristic feature of the linaridins, which in turn lack Lan/MeLan residues. The linaridins represent a small family of linear, dehydrated (arid) peptides presently comprising four members, cypemycin,¹¹ grisemycin,¹² legonaridin,¹³ and salinipeptins.^14,15 After CAO discovery two additional class V lanthipeptide family members pristinin A3¹⁶ and lexapeptide¹⁷ have been identified, which share the N-terminal N,N-dimethylation of CAO. There have been no reports of N-terminal bis-N-methylation for known class I-IV lanthipeptides.

Figure 1.

Structure of cacaoidin (CAO) characterized by the specific N,N-dimethylated lanthionine ring at the N-terminus of the molecule (S∗-Ala; ∗ indicates tentative stereochemistry), the C-terminal S-[(Z)]-2-aminovinyl-(3S)-3-methyl]-d-cysteine (AviMeCys) ring and the glycosylated Tyr-residue. d-aminobutyric acid (d-Abu)

See also Figures S1–S10.

Another rare feature of CAO is the presence of d-amino acids d-alanine and d-aminobutyric acid (Figure 1), only reported for a limited number of RiPPs. Most d-amino acid containing RiPPs are members of the lanthipeptide family, comprising lactocin S,¹⁸ lacticin 3147,¹⁹ carnolysin,²⁰ and bicereucin,²¹ whereas the salinipeptins are the only members of the linaridin subfamily reported to contain d-amino acids.¹⁴ Compared to the lanthipeptide representatives, CAO contains a relatively high number of d-amino acids (a total of seven, compared to one-four in the representatives). In regard to the d-amino acid content of other class V lanthipeptide family members, with one d-amino acid in lexapeptide, and none in pristinin A3^16,17, CAO displays a greater similarity with salinipeptins, that contain nine d-amino acids.¹⁴

Results and discussion

Antimicrobial activity

CAO shows potent activity against methicillin resistant Staphylococcus aureus (MRSA; 0.25 μg mL⁻¹) and good to moderate activity against vancomycin resistant enterococci and Clostridium difficile⁸ but lacks activity against gram-negative organisms (Table S1). A CAO aglycon,²² lacking the sugar moieties connected to Tyr (Figures S1–S10), is characterized by significantly reduced antimicrobial activity (MIC >64 μg/mL; Table S1), indicating that the sugar moieties are involved in the mechanism of action. The antimicrobial activity of lantibiotics (e.g., globular mersacidin (MRS)) is often conferred by interaction with the ultimate peptidoglycan (PGN) precursor lipid II (LII_PGN, undecaprenyl-pyrophosphate-MurNAc-pentapeptide-GlcNAc), thereby blocking cell wall biosynthesis. Flexible lantibiotics, such as nisin, combine LII_PGN-binding and pore formation.^23,24 LII_PGN represents the essential PGN building block that is readily accessible on the outside of gram-positive bacteria, where the lipid intermediate is incorporated into the growing PGN network by glycosyl transferases (GTs) under the release of undecaprenyl-pyrophosphate and subsequent crosslinking by the activity of transpeptidases (TPs).²⁵ Cell wall biosynthesis was identified as the target pathway for CAO, as indicated by pathway specific bioreporters, and the induction of a liaI-lux reporter fusion pointed to interference with the LII_PGN biosynthesis cycle.⁸

Targeting of cell wall building blocks

In accordance, CAO and the LII_PGN-binding lantibiotic MRS added at 1xMIC both induced the expression of liaI-lux as indicated by an increase in the luminescence signal (Figure 2). Pre-incubation with purified LII_PGN in a 2-fold molar excess resulted in the antagonization of liaI induction indicating a direct interaction with the PGN precursor.⁸ LiaI-lux induction was also observed with the CAO aglycon and significant antagonization by LII_PGN was detected (Figure S11A), reflecting the capacity of the CAO aglycon to bind LII_PGN.

Figure 2.

Antagonization of antibiotic-induced liaI-lux stress response by cell wall precursors

CAO (A)- and mersacidin (MRS; B)-induced liaI-lux expression is antagonized when preincubated with purified PGN precursors LI_PGN or LII_PGN at a two-fold molar excess with respect to CAO and MRS. The central lipid carrier undecaprenyl-phosphate (C₅₅P) and undecaprenyl-pyrophosphate (C₅₅PP) did not antagonize antibiotic activity when added in equivalent amounts. The wall teichoic acid precursor LIII_WTA had no influence on MRS-induced liaI-lux expression but reduced the CAO-induced signal three-fold. Representative graphs of three independent experiments. See also Figure S11.

To identify the minimal binding motif of CAO, different undecaprenyl-containing cell wall precursors were tested for their ability to antagonize the CAO-induced response. As observed for LII_PGN, pre-incubation of antibiotics with purified LI_PGN (undecaprenyl-pyrophosphate-MurNAc-pentapeptide) in a 2-fold molar excess almost completely antagonized liaI induction (Figure 2), whereas equivalent amounts of C₅₅PP or C₅₅P had virtually no effect (Figure 2). Even a 20-fold molar excess of C₅₅PP reduced the luminescence signal only by ∼30% (Figures S11B and S11C), indicative for unspecific hydrophobic interactions. The results demonstrate that in contrast to nisin and flexible lantibiotics, interaction with the pyrophosphate moiety is not sufficient for target interaction and that binding of CAO and MRS involves interactions with the first sugar of lipid intermediates. To evaluate the relevance of the nature of the sugar, purified wall teichoic precursor LIII_WTA (undecaprenyl-pyrophosphate-GlcNAc) was tested. Of interest, CAO-induced lia-lux expression was suppressed 3-fold when preincubated with LIII_WTA (Figure 2A), whereas MRS-induced lia-lux expression remained unaffected (Figure 2B), indicating that, in contrast to CAO, the interaction with the MurNAc-pentapeptide sugar of LII_PGN is crucial for MRS-binding, likely involving interactions with the pentapeptide moiety. Indeed, MRS, but not CAO, was able to inhibit amidation of d-glutamate in position 2 of the LII_PGN-stem peptide in vitro (Figure S12).

Isothermal titration calorimetry (ITC) was conducted to determine the strength of the interaction of CAO with LII_PGN. As the target is embedded in the cytoplasmic membrane and contacts with constituting negatively charged phospholipids might impact interaction, CAO binding to DOPG/DOPC (25/75 mol %) model membranes was determined (K_D 42 ± 16 μM) to simulate a more natural environment. The presence of LII_PGN (2 mol %) in model membranes resulted in an approximately 200-fold increased affinity (K_D 0.184 ± 0.014 μM), supporting high affinity binding of CAO to LII_PGN (Figure S13).

Together these results support differential binding modes, involving binding to the pyrophosphate-sugar-peptide portion of LII_PGN for MRS, whereas the interaction of CAO primarily involves binding to the phosphate sugar moiety, with the nature of the first sugar being less relevant. Moreover, interactions with the second sugar in LII_PGN are not mandatory for high affinity binding for both antibiotics, as efficient antagonization was observed with LI_PGN lacking the GlcNAc residue (Figure 2).

Treatment with antibiotics that bind to LII_PGN or inhibit membrane-bound steps of PGN biosynthesis can result in the accumulation of UDP-MurNAc-pentapeptide in the cytoplasm of susceptible bacteria. In contrast to other LII_PGN-binding antibiotics, including vancomycin and MRS, treatment with CAO did not result in accumulation of the soluble PGN precursor UDP-MurNAc-pentapeptide (Figures S14A and S14B), suggesting CAO has membrane perturbating activity or binds to additional target molecules. Absence of UDP-MurNAc-pentapeptide accumulation was also observed for the lantibiotic nisin (Figure S14B) or the LII_PGN binding lassopeptide siamycin.²⁶ In contrast to siamycin, nisin exhibits a dual mode of action. It forms a complex with LII_PGN leading to inhibition of late-stage PGN biosynthesis reactions and in addition uses LII_PGN as a docking molecule for the formation of a defined membrane pore,²⁷ ultimately leading to leakage of cytoplasmic content, thus interfering with UDP-MurNAc-pentapeptide accumulation. However, in contrast to nisin, neither pore formation nor rapid membrane disrupting effects were observed for CAO. Although nisin treatment triggers the release of intracellular potassium ions from Staphylococcus simulans cells, CAO did not lead to potassium ion leakage (Figure S15). Furthermore, the cellular localization of GFP-MinD in Bacillus subtilis was almost unaffected by CAO treatment (Figure S16). MinD accumulates at the cell poles in growing cells to facilitate selection of the mid-cell division site, specific FtsZ placement and formation of the division septum.²⁸ MinD localization is sensitive to alterations of the membrane potential that stimulates membrane binding. Treatment with compounds that dissipate the membrane potential, such as the protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP) or the pore forming nisin, results in rapid MinD delocalization.²⁸ Such effects were not observed for CAO (Figure S16) and the MinD localization pattern at concentrations up to 10xMIC remained unchanged. To further test whether CAO impacts the membrane potential or perturbs membrane integrity, treated cells were incubated with DiBAC₄(3). The fluorescent dye is excluded from cells with a normally polarized membrane whereas it is able to enter depolarized cells.²⁹ No significant difference in fluorescence intensity was detected after treatment with CAO or MRS using concentrations up to 5-fold the MIC (Figure S17).

Direct inhibition of cell wall transglycosylases

The distinct cellular effects observed for CAO prompted a more detailed analysis of CAO’s effect on LII_PGN-consuming reactions, catalyzed by S. aureus penicillin binding proteins (PBPs) and monofunctional transglycosylases. To this end the impact of CAO on transglycosylase activity was investigated in vitro (Figures 3 and S18). Bifunctional PBP2 catalyzes the formation of polymeric PGN from monomeric LII_PGN in vitro. To enable individual analysis of the transglycosylase activity a PBP2 variant harboring a loss of function mutation in the transpeptidase domain was used in all experiments (PBP2_mutTP, Figure 3).

Figure 3.

PBP2-mediated transglycosylation of LII_PGN is inhibited by CAO and restored by increasing concentrations of PBP2, as observed for the transglycosylase inhibitor moenomycin (MOE)

Incrementing PBP2 concentrations did not affect the inhibitory effect of LII_PGN-binding MRS. The extent of inhibition relates to enzyme activity in absence of antibiotics at respective enzyme concentrations. Data presented are means from at least two independent experiments as indicated by individual data points and error bars represent the standard deviation (SD).

See also Figures S18 and S20.

Reactions were initiated by the addition of increasing enzyme concentrations. At lower enzyme concentration (0.75 μM) about 50% of the LII_PGN substrate was converted and full inhibition was observed in presence of CAO, MRS, and moenomycin (MOE) (Figures 3 and S18). For MRS and other LII_PGN-binding antibiotics, it is known that binding of LII_PGN leads to steric hindrance, ultimately resulting in interference with PBP2 action.^27,30 In line with the formation of an antibiotic-LII_PGN complex, increasing concentrations of PBP2 were unable to outcompete the inhibitory effect of MRS. In contrast, incrementing enzyme concentrations gradually restored LII_PGN transglycosylation in the presence of the phosphoglycolipid antibiotic MOE, that acts by direct enzyme inhibition.³¹ Of interest, increasing PBP2 concentrations successively restored transglycosylase activity in presence of CAO, pointing to direct inhibition of the PBP2 enzyme as observed with MOE (Figures 3 and S18).

To further characterize transglycosylase inhibition, the impact of CAO on the monofunctional transglycosylase SgtB was analyzed. In absence of antibiotics and at low SgtB concentration (0.2 μM) approximately 50% of monomeric LII_PGN was transglycosylated, whereas full conversion was achieved at 0.8 μM of SgtB (Figure 4). At the lowest enzyme concentrations tested, SgtB was almost completely inhibited by CAO (Figures 4 and S19). With increasing SgtB concentration enzyme activity was gradually restored to the level of the respective control without antibiotic. As observed for the PBP-catalyzed reaction, MOE strongly inhibited SgtB activity at low enzyme concentrations and full recovery of activity was achieved at 1.6 μM (Figures 4 and S19). Once more, the inhibitory effect of MRS remained unaffected with elevating enzyme amounts (Figures 4 and S19), in agreement with the LII_PGN-binding mechanism of this compound.

Figure 4.

Inhibitory effect of CAO and MOE on the activity of the monofunctional glycosyltransferase SgtB is restored by increasing enzyme concentrations, whereas inhibition of MRS remains unaltered

The extent of inhibition relates to enzyme activity in absence of antibiotics at respective enzyme concentrations. Data presented are means from at least two independent experiments as indicated by individual data points. Error bars represent the SD. See also Figure S19.

Binding to S. aureus PBP2

The in vitro data obtained analyzing CAO’s inhibitory effect on GTs, compared to the substrate binding lantibiotic MRS and the enzyme inhibitor MOE, strongly pointed to a direct GT inhibition by CAO. To validate a direct CAO-GT interaction, surface plasmon resonance (SPR) analysis was performed. Because the transmembrane domain of PBPs is critical for binding of MOE,³² experiments were conducted with full-length PBP2_mutTP (Figure S20A). A dose-dependent binding response of MOE to PBP2 was observed (Figure 5A) confirming the interaction. Kinetic analysis was performed by fitting of the curves to a 1:1 interaction model (Figure S21A) resulting in a K_D of 0.728 μM (Table S2), comparable to steady state kinetics previously determined (0.393 μM).³² Furthermore, the K_D was in the range of 10⁻⁷ M, correlating with the reported inhibitory concentration of MOE for the transglycosylation reaction.^32,33 To investigate the interaction of CAO with PBP2, a series of CAO concentrations were analyzed, resulting in a dose-dependent binding response (Figure 5B). Kinetic analysis was fitted to a 1:1 interaction model (Figure S21B), that proved to be the most suitable to describe the observed interaction (Table S2) and resulted in a K_D of 16.1 μM. Compared to MOE, full dissociation of CAO was not achieved after injection of higher concentrations (≥16 μM). Furthermore, the association curve at these higher concentrations differed compared to lower concentrations. This observation may indicate a more complex interaction pronounced at higher concentrations.

Figure 5.

Binding analysis of MOE, CAO and the CAO aglycon to PBP2

Dose-dependent SPR binding response of MOE (A) and CAO (B) to immobilized full-length S. aureus PBP2_mutTP indicates a direct interaction. For kinetic evaluation, the experimental data was fitted to a 1:1 interaction model using the global data analysis option available within BIAevaluation4.1 software. For CAO a K_D of 16.1 μM and for MOE a K_D of 0.728 μM was determined. The absence of a dose-dependent SPR response in case of the CAO aglycon (C) indicates the importance of the CAO disaccharide moiety for the interaction with PBP2. See also Figures S20 and S21, and Table S2.

Analysis of the CAO and MOE binding curves revealed clear differences in association and dissociation rates, with slower association and dissociation for CAO. Indeed, the half-life of the CAO-PBP2 complex is approximately 10-fold higher (12 min) than for the MOE-PBP2 complex (1.1 min; Table S2). Nevertheless, the determined K_D of CAO at equilibrium is about 22-fold higher. The binding strength of MOE to PBPs from S. aureus, E. faecalis and E. coli, was shown to correlate with antibiotic activity.^32,33 We did not observe a correlation of K_D (16.1 μM) and MIC value (0.106–0.847 μM) for CAO. Whether this difference results from synergistic effects resulting from the dual mechanism of action that combines binding to LII_PGN and direct transglycosylase inhibition (Figure 6) and may further be affected the simultaneous binding to additional undecaprenyl pyrophosphate-containing target molecules from other cell wall biosynthesis pathways remains elusive. It is tempting to speculate that the latter mechanism relies on the unique disaccharide moiety of CAO mimicking and thus competing with the natural transglycosylase substrate, i.e., the β-1,4-linked MurNAc-GlcNAc-chain. Corroborating, the CAO aglycon did not bind to PBP2_mutTP (Figure 5C), indicating that the disaccharide in CAO is crucial for PBP2 binding and further supports specificity of direct enzyme inhibition. MOE represents the only natural product acting specifically as an inhibitor of the GT active site. Crystal structure analysis revealed that the antibiotic binds to the glycosyl donor site of GTs, involving interactions with the disaccharide and the phosphoglycerate portion of the MOE pharmacophore.^31,34,35,36 Semi-synthetic GT inhibitors harboring β-1,2-, β-1,3- or β-1,4-linked disaccharides moieties were shown to facilitate inhibition of PGN synthesis in vitro^37,38 and to exhibit antimicrobial activity.^39,40,41,42 Of interest, neither the positioning of the glycosidic bond nor the identity of the disaccharide hexoses were determinants for the overall activity of the inhibitors, supporting the involvement of the β-1,3-linked 6-deoxygulopyranosyl-α-rhamnopyranoside disaccharide of CAO in GT binding in a similar fashion.

Figure 6.

Proposed model for the dual mode of action of CAO, given the example of S. aureus PBP2 activity

PBP2 is a membrane-anchored bifunctional PBP catalyzing the extracellular steps of bacterial PGN biosynthesis. The glycosyltransferase (GT) domain polymerizes the disaccharide pentapeptide moieties from LII_PGN and the transpeptidase domain subsequently crosslinks the polymer via their peptide moieties. Our proposed mode of action comprises the direct binding of CAO to the GT domain (CAO:PBP2) and binding of the transglycosylase substrate LII_PGN, exerting steric hindrance of enzymes catalyzing LII_PGN-consuming reactions (CAO:LII).

To identify the motifs crucial for binding to GTs and LII_PGN and for rational SAR analysis, determination of the crystal structures of the individual complexes would be of importance. Further investigation of CAO variants with modified sugar moieties could provide valuable information on the role of the disaccharide unit for GT interaction.

Limitations of the study

In this study, we report that CAO exhibits a dual mode of action by binding to the peptidoglycan precursor LII_PGN and by directly inhibiting murein transglycosylases. The fact that substrate and enzyme of the same reaction are inhibited prevented this study to dissect to which extend the two mechanisms contribute to the cellular mode of action of CAO in vivo. Given the much higher affinity of CAO to LII_PGN as compared to PBP2 suggests that this feature is most important for the antibacterial activity of CAO. However, binding to LII_PGN may affect binding to the murein transglycosylase and vice versa. Although this is likely to lead primarily to synergistic effects, it cannot be completely excluded that either mechanism is sufficient for the activity of CAO in vivo.

Significance

CAO represents the first natural product with a dual mode of action comprising binding of the PGN synthesis intermediate LII _PGN and direct inhibition of murein transglycosylases ( Figure 6 ), both representing validated target structures of successful antibiotics. The unique structural features of CAO could guide the synthesis of hybrid antibiotics with resistance breaking properties because of multi-targeting.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Staphylococcus simulans 22, clinical isolate	Bierbaum and Sahl43	–
Staphylococcus aureus	Sass and Bierbaum44	SG511
Staphylococcus aureus COL (MB5393)	Gill et al.45	SACOL
Staphylococcus aureus Mu50	ATCC, Kuroda et al.46	ATCC 700699, SAV
Bacillus subtilis	Kunst et al.47	168
Bacillus subtilis MinD-GFP	Marston et al.48	1981
Bacillus subtilis PliaI-lux	Radeck et al.49	TMB1617
Enterococcus faecalis	Jacob and Hobbs50	JH2-2
Enterococcus faecalis, VanA-resistant (pIP819)	Leclercq et al.51	BM4223
Escherichia coli, antibiotic sensitized	Merck Collection, Kodali et al.52	MB5746
Chemicals, peptides, and recombinant proteins
Undecaprenyl-MPDA monophosphate (NH4+)2	Larodan	62–1055
Undecaprenyl-DPTA Diphosphate (NH4+)3	Larodan	68–1100
Mersacidin	Hoechst GmbH	–
Moenomycin complex	Cayman Chemical	Cay15506
1,2-dioleoyl-sn-glycero-3-phospho-(1horac-glycerol) (sodium salt), chloroform solution	Sigma-Aldrich	840475C
1,2-dioleoyl-sn-glycero-3-phosphocholine, chloroform	Sigma-Aldrich	850375C
Software and algorithms
Fiji	Schindelin et al.53	https://imagej.net/software/fiji/
Plug-In MicrobeJ	Ducret et al.54	https://www.microbej.com/
GraphPad Prism	Dotmatics	https://www.graphpad.com/scientific-software/prism/
Illustrator CS5	Adobe	https://www.adobe.com/products/illustrator.html
BIACORE Control Software 4.1.1	Cytiva	https://www.cytivalifesciences.com/en/us/shop/protein-analysis/spr-label-free-analysis/software?sort=NameAsc&chunk=1
MicroCal PEAQ-ITC Analysis Software Version 1.20	Malvern Panalytical	https://www.malvernpanalytical.com/de/support/product-support/software/microcal-peaq-itc-family-analysis-software-update-v121
Other
Bio-ScaleTM Mini NuviaTM IMAC Ni-Charged Cartridge, 1 mL	Bio-Rad	–
Sensor chip for protein - peptide/small molecule interaction	XanTec	HC1500M
Microplate reader	Tecan	Spark 10 M
Mini-Extruder	Avanti
AxioObserver Z1 equipped with a HXP 120C lamp, an αPlan-APOCHROMAT 100×/1.46 oil objective and an AxioCam MRm camera	Zeiss	https://www.zeiss.com/microscopy/en/products/light-microscopes/widefield-microscopes/axio-observer-for-life-science-research.html
Biacore 3000	Cytiva	https://www.cytivalifesciences.com/en/us/shop/protein-analysis/spr-label-free-analysis/systems?sort=NameAsc&chunk=1
MicroCal PEAQ-ITC Automated microcalorimeter	Malvern Panalytical	https://www.malvernpanalytical.com/en/products/product-range/microcal-range/microcal-itc-range/microcal-peaq-itc-automated

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to, and will be fulfilled by the lead contact, Tanja Schneider (tschneider@uni-bonn.de).

Materials availability

Request concerning the provision of cacaoidin should be addressed to Olga Genilloud (olga.genilloud@medinaandalucia.es), since production and purification were conducted in this work group. This study did not generate new unique reagents.

Experimental model and subject details

Bacterial strains used in this study:

 Staphylococcus simulans 22⁴³

 Staphylococcus aureus SG511,⁴⁴ S. aureus COL (MB5393, MRSA),⁴⁵ S. aureus Mu50 (VISA)⁴⁶

 Bacillus subtilis 168,⁴⁷ Bacillus subtilis 1981 MinD-GFP,⁴⁸ Bacillus subtilis P_liaI-lux⁴⁹

 Enterococcus faecalis JH2-2 (WT),⁵⁰ E. faecalis BM4223 pIP819 (VRE)⁵¹

 Escherichia coli MB5746⁵²

Method details

Antibacterial assay

MIC values were determined by broth microdilution with Mueller-Hinton broth using bovine serum albumin (BSA)-coated polypropylene plates as previously described .⁵⁵ The use of polypropylene plates instead of polystyrene and BSA-coating should prevent the interaction of compounds with the plate surface. Wells were coated with BSA by incubation of each well with 200 μL of 1% BSA in phosphate-buffered saline (PBS) buffer (w/v) for 30minat 37°C and subsequently washed with PBS buffer.

Induction and antagonization of liaI-lux cell wall stress response

B. subtilis PliaI-lux⁴⁹ was grown in Mueller−Hinton broth supplemented with chloramphenicol (5 μg mL⁻¹) to OD₆₀₀ = 0.5. The antibiotics CAO and MRS (Hoechst GmbH, Frankfurt/Main, Germany), as well as the CAO aglycon were added in a BSA-coated Greiner LUMITRAC^TM 96-well-microtiter plate at a final concentration of 8 μg mL⁻¹,4 μg mL⁻¹ and 100 μg/mL respectively prior to addition of the reporter strain. Induction of cell wall stress indicated by an increase in luminescence was measured using the Tecan Spark 10 M microplate reader for 400minat 30°C.

Antagonization of CAO, MRS and CAO aglycon induced lial-lux luminescence response was used as a method to investigate the direct interaction of the antibiotics with putative antagonists as previously reported.^8,26 To this end, purified PGN (LI_PGN, LII_PGN) and WTA precursors (LIII_WTA) were first preincubated with CAO and MRS at a two-fold molar excess with respect to the antibiotics. The central lipid carrier undecaprenyl-phosphate (C₅₅P) and undecaprenyl-pyrophosphate (C₅₅PP) were tested in equivalent amounts, but additionally in a 20-fold molar excess with respect to the antibiotics. After preincubation, the experiment was performed, and luminescence was measured as described above. At least three independent biological replicate experiments were conducted.

Overexpression and affinity purification of recombinant proteins

Purification of the GatD-His₆/MurT complex was performed as previously described and specified below.⁵⁶E. coli strain BL21 (DE3) was used as a host for the recombinant expression of S. aureus proteins GatD-His₆/MurT (pET21b-murT-gatD), PBP2_mutTP-His₆ (pET21b-pbp2_mutTP), SgtB-His₆ (pET28a-sgtB), and PBP4-His₆ (pET21b-pbp4). Cells were grown at 37°C in lysogeny broth (LB; Oxoid) supplemented with the appropriate antibiotic (100 μg mL⁻¹ ampicillin or 50 μg mL⁻¹ kanamycin). Proteins were expressed in mid-log phase cultures (OD₆₀₀ 0.6) after induction with 1 mM IPTG for 4hat 37°C.

Cells expressing GatD-His₆/MurT were harvested and resuspended in buffer (50 mM Tris-HCl, 300 mM NaCl, 10 mM imidazole, pH 7.5) supplemented with lysozyme (250 μg mL⁻¹), DNase (50 μg mL⁻¹), and RNase (10 μg mL⁻¹) and incubated for 30 min on ice. After sonification the cell debris was spun down and the supernatant was applied to Ni-NTA-agarose slurry. This mixture was gently stirred at 4°C for 2 h and then loaded onto a column support. After washing with lysis buffer, to remove weakly bound material (50 mM Tris-HCl, 300 mM NaCl and 20 mM imidazole, pH 7.5), His-tagged recombinant proteins were eluted with buffer containing 50 mM Tris-HCl, 300 mM NaCl and 100–200 mM imidazole, pH 7.5.

Cells expressing PBP2_mutTP-His₆, SgtB-His₆ or PBP4-His₆ were harvested by centrifugation (12 min, 7000 xg, 4°C), washed once with buffer (50 mM Tris-HCl, 500 mM NaCl, pH 7.5), and resuspended in buffer supplemented with 1% (v/v) Triton X-100, lysozyme (250 μg mL⁻¹), DNase (50 μg mL⁻¹), RNase (10 μg mL⁻¹) and 0.5% N-laurylsarcosine. After incubation on ice for 60 min, cells were sonicated and centrifuged (20 min, 15000 x g, 4°C), the cleared lysate was subjected to Ni²⁺-affinity chromatography. In case of the purification of SgtB-His₆, the cleared lysate was incubated for 1 h with NiNTA-agarose (1 mL NiNTA-agarose per 10 mL supernatant) under gentle stirring and transferred onto a column support. After washing of the NiNTA-agarose matrix with buffer W (50 mM Tris-HCl, 500 mM NaCl, 1% (v/v) Triton X-100, pH 7.5), subsequent rinsing with buffer W supplemented with 10 mM and 20 mM imidazole removed weakly bound proteins. SgtB-His₆ was eluted with buffer containing 300 mM imidazole. Elution fractions were collected and stored in 30% (v/v) glycerol at −20°C.

Since PBP2_mutTP and PBP4 were utilized for SPR measurements, these proteins had to be of particularly high purity. The required purity was ensured by two successive purification cycles via a fast protein liquid chromatography (FPLC) system (BioRad) with a 1 mL Bio-Scale^TM Mini Nuvia^TM IMAC Ni-charged cartridge. After the cleared lysate was loaded to the column, the resin was rinsed with buffer (50 mM Tris-HCl, 500 mM NaCl, 0.06% (v/v) Triton X-100, pH 7.5). Subsequently, an imidazole gradient (up to 50–100 mM) was applied in order to remove weakly bound proteins. The protein of interest (PBP2_mutTP or PBP4) was eluted with 500 mM imidazole containing buffer. Purity of elution fractions was assessed by SDS-PAGE (NuPAGE; Invitrogen) analysis and relevant fractions were pooled. Those fractions were dialyzed against buffer without imidazole (50 mM Tris-HCl, 500 mM NaCl, 0.06% (v/v) Triton X-100, pH 7.5) and concentrated using a spin column with molecular weight cut off (MWCO) of at least half the proteins size before the second run. After the second FPLC run (performed in the same way described above), elution fractions were assessed for purity, pooled and concentrated as mentioned above. Finally, proteins used for SPR coupling were dialyzed against coupling buffer (10 mM sodium maleate, pH 5.6). Protein concentrations were measured using Bradford reagent (BioRad).

In vitro reactions of PGN biosynthesis in presence of CAO

In all in vitro assays, the PGN intermediate LII_PGN was used as a substrate and CAO was added in molar ratios with regard to LII_PGN. MRS and MOE were used as control antibiotics. LII_PGN was synthesized using partially purified UDP-MurNAc-pentapeptide and membranes containing the enzymes MraY and MurG and purified by HPLC as described previously .⁵⁷ The polyprenyl containing products and non-processed substrate were extracted from the reaction mixture with an equal volume of n-butanol/pyridine acetate, pH 4.2 (2:1; v/v), and analyzed by thin-layer chromatography (TLC) using chloroform/methanol/water/concentrated ammonium hydroxide (88:48:10:1, v/v/v/v) as solvent⁵⁸ and phosphomolybdic acid staining.⁵⁷ Quantification was carried out using ImageJ 1.52a software.

In vitro amidation was assayed by incubating 2 nmol of LII_PGN in 160 mM Tris-HCl, 40 mM MgCl₂, 50 mM KCl, pH 7.5, 0.26% Triton X-100, 6 mM ATP, and 6.6 mM glutamine in a total volume of 30 μL. The reaction was initiated by the addition of 3.8 μg of the purified GatD-His₆/MurT complex and incubated for 2hat 30°C. In case of reactions containing antibiotics, CAO and MRS were added in a two-fold molar excess with regard to LII_PGN, and preincubated for 20 min prior to the addition of enzyme. The MRS containing reaction further included 1.25 mM CaCl₂. The experiment was performed in at least two biological replicates.

To investigate the impact of CAO on in vitro transglycosylase activity generating polymeric PGN by the consumption of monomeric LII_PGN, the activity of the bifunctional PBP2 and monofunctional SgtB of S. aureus were analyzed. In the case of the bifunctional PBP2, the PBP2_mutTP (S398G) variant^59,60 was used in all experiments to ensure analysis of transglycosylase activity exclusively. Enzymatic activity of PBP2_mutTP-His₆ was determined in 20 mM 2-(N-morpholino)-ethane sulfonic acid (MES) buffer, 2 mM MgCl_2, 2 mM CaCl₂ and 0.06% TritonX-100, pH 5.5. CAO or MRS were added at a two-fold molar excess (120 μM) with respect to the substrate LII_PGN (2 nmol; 60 μM). MOE containing reactions were supplemented with 0.75 μM MOE corresponding to the lowest PBP2_mutTP concentration tested. After pre-incubation of the antibiotics with LII_PGN for 20 min, the transglycosylation reaction was initiated by addition of the enzyme added at incrementing concentrations of 0.75, 1.5, and 3 μM, respectively. For each enzyme concentration tested, a control reaction without antibiotic was prepared. After incubation for 2hat 30°C, the reactions were stopped and analyzed as described above and the amount of remaining LII_PGN was quantified to indirectly determine enzyme activity. Control reactions were supplemented with respective amounts of antibiotics after the reactions were stopped to match the extraction properties. The total signal of 2 nmol unprocessed LII_PGN substrate served as a reference for calculation of enzyme activity, calculated as 100-100/[total LII_PGN signal]∗[LII_PGN signal of respective reaction].

The enzymatic activity of SgtB-His₆ was analyzed in 20 mM MES and 10 mM CaCl₂, pH 5.5. The reactions contained 2 nmol (60 μM) LII_PGN substrate. CAO was added in a three-fold molar excess (180 μM), MRS in a two-fold molar excess (120 μM) towards LII_PGN and MOE was added at a final concentration of 12 μM. The antibiotics were preincubated for 20 min with LII_PGNprior to addition of SgtB. The amount of SgtB was titrated from 0.2 to 1.6 μM, while the concentration of respective antibiotics remained unchanged. All reactions were stopped after 2 h incubation at 30°C, analyzed and quantified as described above.

Intracellular accumulation of UDP-MurNAc-pp

Analysis of the cytoplasmic UDP-MurNAc-pp precursor pool was performed as described before⁶¹ with some modifications. S. simulans 22 cells were grown at 37°C in Mueller-Hinton broth to OD₆₀₀ 0.5 and supplemented with 130 μg mL⁻¹ chloramphenicol. After an incubation time of 15 min, respective antibiotics were added and incubated for another 30 min. The concentrations of the antibiotics were chosen to inhibit growth without significant lysis of the cells. The cells were harvested and subsequently extracted with boiling water for 30 min. The cell extracts were acidified (pH 2) by addition of phosphatic acid (20%; v/v) and sterile filtered prior to high-performance liquid chromatography (HPLC) analysis (100-C18 column).

Membrane disruption-potassium release from whole cells

Potassium release from S. simulans 22 whole cells was performed as previously described⁶¹ in presence and absence of 10 mM glucose. Cells were grown at 37°C in Mueller-Hinton broth and harvested at an OD₆₀₀ 1.0. After washing with cold choline buffer (300 mM choline chloride, 30 mM MES, 20 mM Tris, pH 6.5), cells were resuspended to OD₆₀₀ 30.0, kept on ice, and used within 30 min. Prior to each measurement the cells were diluted to OD₆₀₀ 3.0 with choline buffer (25°C). Compound-induced potassium leakage was plotted relative to the total amount of potassium release after the addition of 1 μM (5xMIC) nisin (NIS, 100%, positive control); non-treated cells were used as the negative control. CAO was added at 0.2 μM and 1 μM (1x and 5xMIC, respectively). Both compounds were added after 20 s and potassium release was monitored.

Measurement of membrane depolarization

S. aureus SG511 was grown in MH Medium supplemented with 1.25 mM CaCl₂ at 37°C and 120 rpm until cultures reached OD₆₀₀ = 0.5. Aliquots were treated with the respective compound at the respective concentration for a total of 30 min. For the last 15 min of incubation, DiBAC₄(3) was added to a final concentration of 100 μM. Cells were washed three times in MH Medium supplemented with 1.25 mM CaCl₂ and placed on a 1% agarose pad mounted on a microscopy slide. Widefield fluorescence microscopy was performed on a Carl Zeiss AxioObserver Z1 equipped with a HXP 120C lamp, an αPlan-APOCHROMAT 100×/1.46 oil objective and an AxioCam MRm camera. Visualization of DiBAC₄(3) was achieved using Carl Zeiss filter set 38 (450–490 nm excitation, 495 nm beam splitter and 500–500 nm emission). Image Analysis was performed using Fiji (ImageJ) Version 2.0.0-rc-69/1.52p; Java 1.8.0_172 [64-bit]⁵³ and the ImageJ Plug-In MicrobeJ Version 5.13L (20) – beta.⁵⁴ Statistical analysis was performed using GraphPad Prism 8.0.2 (GraphPad Software) (263). Individual cell data was extracted from raw images with exclusion of unsharp regions and unspecific artifacts. Mean fluorescence intensity data of individual cells was background corrected and subjected to statistical analysis.

Determination of GFP-MinD localization

Fluorescence microscopy to analyze the cellular localization of the GFP-MinD fusion protein was performed as previously described.²⁸ B. subtilis HS17 cells were treated with increasing concentrations of CAO (1xMIC to 10xMIC, corresponding to 2 μg mL⁻¹ to 20 μg mL⁻¹) or 100 μM CCCP (positive control) before mounting on microscope slides covered with a thin film of 1% agarose. Imaging was carried out within 2 min after addition of CAO or CCCP. Fluorescence microscopy was performed as stated above. Images were processed with Adobe Illustrator CS5.

SPR measurements of PBP2 interaction

All SPR measurements were performed on a Biacore 3000 apparatus (Cytiva) with research grade sensor chips (HC1500M, XanTec) and evaluated using the BIACORE Control Software 3.1.1.

Both, PBP4-His₆ and PBP2_mutTP-His₆, were immobilized on the respective flow cell via amine coupling⁶² according to the manufacturer’s instructions (XanTec): PBP2_mutTP-His₆ was immobilized on the active flow cell, S. aureus PBP4-His₆ on the reference cell, sincein vitro carboxypeptidase activity of PBP4 was unaltered in presence of CAO (data not shown). Briefly, the chip surface was conditioned with elution buffer (1 M sodium chloride, 0.1 M sodium borate pH 9.0; 3 min, 40 μL min⁻¹) and washed with deionized water until a stable baseline was reached. The surface of the reference cell was activated with freshly prepared activation mix (100 mM N-hydroxysuccinimide in 50 mM MES buffer at pH 5.0, supplemented with 26 mM N-ethyl-N’-(dimethylaminopropyl)-carbodiimide hydrochloride; 7.5 min, 40 μL min⁻¹). After a brief wash with deionized water (5 min, 40 μL min⁻¹), the PBP4 protein was injected (22 μg mL⁻¹; 7.5 min, 40 μL min⁻¹). To complete the coupling reaction, the surface was rinsed with deionized water for 40 min. The remaining active NHS esters were quenched with quenching buffer (1 M ethanolamine hydrochloride, pH 8.5; 15 min, 40 μL min⁻¹). The matrix of the analytic flow cell was activated as described above and PBP2_mutTP-His₆ protein solution (10 μg mL⁻¹) was injected (7.5 min, 40 μL min⁻¹). After rinsing the surface with deionized water for 40 min, the residual active ester groups were deactivated as described above.

Following the immobilization, the whole system was rinsed thoroughly with freshly filtered running buffer (50 mM Tris-HCl, 150 mM NaCl, 0.005% Tween20, pH 7.5) at high flow rates. The known and characterized peptidoglycan glycosyltransferase inhibitor MOE³² was used as positive control to analyze the interaction with PBP2. MOE complex, which is a mixture of moenomycins A, A12, C1, C3 and C4 (Cayman chemicals), was freshly dissolved in 100% dimethyl sulfoxide (DMSO) to a concentration of 10 mg mL⁻¹ (6.31 mM). CAO was freshly dissolved in deionized water to a concentration of 10 mg mL⁻¹ (4.24 mM). For the SPR measurements, the compounds were diluted in running buffer at the respective concentrations of 0.25 μM–4 μM for MOE and 0.5 μM–32 μM for COA. The interactions were investigated under constant conditions (40 μL min⁻¹ flow rate). Immobilization and data collection was performed at 25°C. The direct binding assays were performed as a multicycle. Each cycle consisted of 10 min stabilization time, a 2.5 min injection and a dissociation phase of 30 min. Binding response (RU) was recorded continuously and presented as a sensorgram. Running buffer injections were used at the beginning of a run and in between cycles where higher concentrations of compounds were injected (≥4 μM). CAO in a concentration of 8 μM was injected in duplicate to monitor the reproducibility of the binding response. Different kinetic models were applied to the crude data and examined for consistency between crude data and the fit, but kinetic evaluation by fitting the data to a 1:1 Langmuir binding model with drifting baseline created the lowest deviation of data and fit (data not shown).

Isothermal titration calorimetry

The phospholipids 1,2-dioleoyl-sn-glycerol-3-phospho-(1′-rac-glycerol) (sodium salt) (DOPG) and 1,2-dioleoyl-sn-glycerol-3-phosphocholine (DOPC) were purchased from Sigma-Aldrich (Taufkirchen, Germany). Formulation of large unilamellar vesicles (LUVs): Phospholipids were dissolved in chloroform to make 10–30 mM stock solutions, LII_PGN in chloroform/methanol 1:1 for 0.7–1 mM stock solutions and stored in −20°C. Appropriate volumes of the stock solutions were mixed and the organic solvents evaporated under a stream of nitrogen at 35–40°C. The resulting dry lipid films were hydrated with buffer, 50 mM Tris, 100 mM NaCl, pH 7.0 and homogenized by 5 cycles of freezing (−196°C) and thawing (35–40°C) to produce vesicle suspensions with a final concentration of 10 mM total lipid. The suspensions were passed through 2 oppositely directed Whatman polycarbonate membranes with a final pore size of 0.2 μm (Sigma Aldrich, Taufkirchen, Germany) 11 times at room temperature with an Avanti mini extruder (Avanti Polar Lipids Inc., Alabaster, Alabama USA) to yield homogeneous LUV formulations. LUV suspensions of DOPG/DOPC 1:3 with or without 2 mol % LII_PGN (10 mM total lipid) were titrated into a freshly prepared peptide solution in the same buffer. Control titrations included the titration of LUVs into buffer. All binding experiments were performed using a Micro-Cal PEAQ-ITC Automated microcalorimeter (Malvern Panalytical Ltd, Malvern, UK). The samples were equilibrated to 25°C prior to measurement and titrations conducted at 25°C under constant stirring at 1000 rpm. Each experiment consisted of an initial injection of 0.3 μL followed by 25 separate injections of 1.5 μL into the sample cell of 200 μL. The time between each injection 180 s, except noted otherwise and the measurements were performed with the reference power set at 5 μcal s-1 and the feedback mode set at “high”. The calorimetric data obtained were analyzed using MicroCal PEAQ-ITC Analysis Software Version 1.20 (Malvern Panalytical Ltd, Malvern, UK). ITC data fitting was made based on the “One set of sites” model of the software. The best fit is defined by chi-squared minimization. Thermodynamic parameters are reported as the average of two experiments with the standard deviation.

Quantification and statistical analysis

All data are represented as mean ± S.D. or representative images and graphs are shown as indicated in the corresponding figure legends. The number of independent experimental replicates is specified in each figure legend.

Analysis of in vitro PGN biosynthesis reactions was performed in at least two biological independent replicates and quantified using ImageJ 1.52a software.

Induction of PliaI-lux expression in B. subtilis and experiments analyzing antagonization of the cell wall stress response were conducted in at least three independent biological replicates. Data visualization was performed using GraphPad Prism version 6 (GraphPad Software).

For DiBAC₄(3) fluorescence microscopy experiments, statistical analysis was performed using GraphPad Prism version 8 (GraphPad Software). The experiment was repeated three times with biologically independent replicates. Statistical analysis was performed on the population mean values of these replicates. Minimum number of cells used for calculating population mean values as well as details of statistical analysis and data visualization are given in the figure legend.

SPR experiments were analyzed using the BI-Aevaluation software Version 4.1.1 (Cytiva). All binding curves were aligned to the start of injection, and portions of the plot that lay outside of the relevant association and dissociation regions were removed. The baseline was set to zero at a stable region at the injection start. Double referencing was applied by subtracting running buffer injections, and injection spikes were deleted. To align all curves horizontally, a Y-axis transform was performed. For the evaluation of the kinetic constants, curve fitting with simultaneous k_a/k_d was chosen. The data range of association and dissociation was selected, and different kinetic models for ligand-analyte interaction, including 1:1 Langmuir binding, 1:1 binding with drifting baseline, 1:1 binding with mass transfer, bivalent analyte, heterogeneous ligand-parallel reactions, and two-state reaction, were evaluated for the best fit (closest overlap), while considering the residual plots and Chi-squared values that indicate the deviation between the experimental and fitted data. Steady-state affinity analysis was performed by preparing the data as described above and selecting the saturated plateau region of the curves. The average maximal binding responses at saturation were plotted against the tested concentrations, and the K_D was evaluated using the BIAevaluation software. For every compound tested, at least one concentration was injected in duplicate to monitor the reproducibility of the binding response.

The calorimetric data obtained were analyzed using MicroCal PEAQ-ITC Analysis Software Version 1.20 (Malvern Panalytical Ltd, Malvern, UK). ITC data fitting was made based on the “One set of sites” model of the software and the best fit is defined by chi-squared minimization. Thermodynamic parameters are reported as the average of two experiments with the standard deviation.

Published: March 11, 2023

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this work will be shared by the lead contact upon request.