WO2024123712A1 - Analysis of candidate cytotoxic compositions - Google Patents

Abstract

The presently disclosed subject matter relates to surface plasmon resonance methodologies to record kinetic binding data associated with the interaction of compositions and Gram- negative bacterial cell outer membranes and/or outer membrane vesicles. This approach can be used to study any composition for its ability to bind to or alter the outer membrane barrier of Gram-negative bacteria and the kinetic analysis outlined herein can provide insight into the mechanisms of action and resistance to such compositions.

Description

00B206.1385 ANALYSIS OF CANDIDATE CYTOTOXIC COMPOSITIONS CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Application No.63/385,992, filed December 4, 2022, the disclosure of which is incorporated herein by reference in its entirety. BACKGROUND The outer membrane of Gram-negative bacteria is an essential feature that serves as a permeability barrier to exclude toxigenic molecules, including antibiotics, and to provide physical rigidity to the cell. The principal lipid component of the outer leaflet of the outer membrane is lipopolysaccharide (LPS), a complex glycolipid composed of a conserved lipid A anchor, a solvent-facing core oligosaccharide, and a variable, extracellular O-antigen. These three components combine to produce a dense electronegative hydrogel barrier. The polymyxin class of antibiotics, which include polymyxin B and colistin, are last-resort antibiotics for the treatment of multi-drug resistant Gram-negative bacterial infections and target lipid A. These electropositive amphiphilic antibiotics can establish multiple electrostatic interactions with phosphate groups of lipid A in addition to hydrogen bonds and lipophilic interactions with the hydrophobic anchor and the associated cooperativity leads to enhanced binding avidity. Polymyxin binding to lipid A is thought to lead to aggregation, disruption of outer membrane integrity, and, ultimately, bacterial cell death. Investigations into the binding of polymyxins to bacterial cells have generally relied on whole-cell and plate-based equilibrium approaches. Different methods have produced equilibrium binding constants (KDs) for polymyxin B ranging from 400 nM up to greater than 100 µM, which are, in some cases, multiple orders-of-magnitude higher than the reported cellular potency of these antibiotics. However, these approaches are likely complicated by the rapid kill-kinetics of polymyxins, non-specific binding of polymyxins to plate surfaces, assumptions of simple 1:1 binding with the LPS, an inoculum effect on activity, the inability to reach equilibrium under the assay conditions, and the need for perturbative labels on the polymyxins. The use of purified lipids aimed at mimicking the outer membrane have enabled the use of surface plasmon resonance (SPR), however these have not resulted in a mechanistic model. Given the deficiencies associated with existing methodologies, new strategies for the analysis of candidate cytotoxic compositions, e.g., Active 108465066.1.DOCX 1 00B206.1385 polymyxins, that address the biological complexity of the outer membrane of Gram-negative bacteria would be beneficial in the art. SUMMARY OF THE INVENTION In certain embodiments, the present disclosure is directed to methods of identifying a composition capable of inducing the death of Gram-negative bacterial cells, comprising: contacting the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, with the composition; determining that the composition complexes with a lipopolysaccharide integral with the outer membrane; contacting the complex with a wash; determining that the composition is retained in the outer membrane after contact with the wash; repeating steps a-d; and determining that the composition accumulates in the outer membrane and is retained in the outer membrane, e.g., for at least about 10 minutes; wherein said accumulation of the composition in the outer membrane and its retention in the outer membrane, e.g., for at least 10 minutes, identifies the composition as capable of inducing the death of Gram-negative bacterial cells. In certain embodiments, the outer membrane is an outer membrane of an outer membrane vesicle derived from a Gram-negative bacterial cell. In certain embodiments, the composition accumulates in the outer membrane, e.g., to a mass fraction greater than 10% (weight of accumulated composition / weight of outer membrane). In certain embodiments, the determination that the composition complexes with a lipopolysaccharide integral with the outer membrane is performed by a SPR analysis. In certain embodiments, the determining that the composition is retained in the outer membrane after contact with the wash is performed by SPR analysis. In certain embodiments, the determining that that the composition accumulates in the outer membrane and is retained in the outer membrane, e.g., for at least about 10 minutes, is performed by SPR analysis. In certain embodiments of the methods disclosed herein, the Gram-negative bacterial cell or outer membrane vesicle derived therefrom is immobilized on an SPR sensor substrate. In certain embodiments, the Gram-negative bacterial cell or outer membrane vesicle derived therefrom is immobilized on an SPR sensor substrate via an affinity coupling. In certain embodiments, the affinity coupling comprises amine-coupled polymyxin B immobilized on the SPR sensor substrate. In certain embodiments of the methods disclosed herein, the Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, has lipopolysaccharide with variable length sugars. Active 108465066.1.DOCX 2 00B206.1385 In certain embodiments of the methods disclosed herein, the Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, is polymyxin B resistant. In certain embodiments of the methods disclosed herein, the composition is a candidate antibacterial. In certain embodiments, the candidate antibacterial is a polypeptide or polypeptide analog. In certain embodiments, the candidate antibacterial is a polymyxin. BRIEF DESCRIPTION OF THE DRAWINGS Figures 1A-1E depict SPR-detection of polymyxin B binding to Gram-negative outer membrane vesicles (OMVs). (Figure 1A) Cartoon of SPR chip surface with bound whole cells or OMVs, which present the outer membrane in a near-native state. (Figure 1B) Transmission electron micrograph of wild-type OMVs (wt-OMVs) exposed to polymyxin B (right) or buffer (left). Representative images shown. Scale bar 100 nm. (Figure 1C) Scanning electron micrograph of wt-OMVs affinity captured by polymyxin B pre-coupled to the planar surface. Representative images shown. Scale bar 300 nm. (Figure 1D) SPR binding curves showing a stepwise increase in binding response upon exposure of wt-OMVs to eight serial-doubling concentrations of polymyxin B from 39 nM to 5 µM. (Figure 1E) Same plot shown in (Figure 1C) but illustrating the approximately 1:2 fraction of weakly bound polymyxin B to tightly bound polymyxin B. Figures 2A-2I provide SPR kinetic traces for binding of antibiotics to whole bacterial cells or OMVs. Each serial injection profile was recorded as outlined in Figures 1A-1E with a maximum concentration of 625 nM. Representative SPR sensorgram of polymyxin B binding to (Figure 2A) wt-OMVs or (Figure 2B) resistant-OMVs. Representative SPR sensorgrams of whole cell binding by polymyxin B with (Figure 2C) wild-type E. coli cells or (Figure 2D) polymyxin B-resistant E. coli cells expressing mcr-1. Representative SPR sensorgrams of PMBN binding with (Figure 2E) wt-OMVs and (Figure 2F) resistant-OMVs. Representative SPR sensorgrams of (Figure 2G) polymyxin B and (H) PMBN to wt-OMVs in the presence of 32 mM Mg2+. (Figure 2I) Representative SPR sensorgrams of brevicidine binding to wt-OMVs (black) or resistant-OMVs (blue). All traces are representatives of n=3+ independent SPR runs. Figure 3A-3B illustrates two versions of the three-state model for LPS-targeting antibiotics. The first more approximate version is shown in FIG.3A, Initial State: Affinity network consisting of tightly packed LPS (black circles, L) stabilized by divalent cations (red dots) is shown. State 1: Polymyxin B (PMB, blue) binds to L transiently to form PMB- L. State 2: A fraction of PMB-L slowly form lipophilic interactions with the membrane Active 108465066.1.DOCX 3 00B206.1385 producing a more stable complex, nPMB-L. State 3: LPS phase transition is triggered forming hexagonal clusters when sufficient PMB-L are bound. PMB then dissociates from nPMB-L and liberated PMB phase separates into clusters cPMB that accumulate at the cluster centers, which likely contains phospholipids. Importantly, L becomes free to engage another PMB molecule allowing clusters to continue to accumulate, filling the available compartment volume. These microscopic states are assumed in order to illustrate the three- states that emerge from SPR kinetic analysis, but the application of the kinetic model does not depend on knowing such precise microscopic details. A second version of the three-state model in FIG. 3B, describes the Reaction mechanism for LPS-targeting antibiotics. The annotation of each specie is slightly modified and the enumeration of the rate constants does not match those in FIG. 3A since the process of mass transport of polymyxin B has been added and a transition state has been added to more completely model formation of the polymyxin B clusters. The model describes LPS-catalyzed accumulation of polymyxin B clusters, cPMB through phase separation mediated by transition state intermediates. PMBi is injected polymyxin B; PMB is polymyxin B located at the LPS coated surface; L is LPS; PMBL is a transient polymyxin B-LPS complex; n indicates available membrane insertion sites (open circles); nPMBL is the membrane-inserted polymyxin B-LPS species; tPMB is a transient nPMBL dimer in complex with LL which is a transient LPS dimer; and cPMB is a phase- separated monomer of polymyxin B that exists in clusters containing multiples of this monomer. k_t is a mass transport rate that supplies PMBi to the LPS-coated surface annotated as PMB where the concentration of PBM is less than the concentration of PMBi due to depletion. Depletion results from the surface reactions initiated by binding to L; KD1 is an effective affinity constant for binding of PMB to L. Figures 4A-4C illustrate the three-state model from Fig.3A fitted to experimental SPR binding curves (Fig. 4A- Fig. 4B) and associated simulations (Fig. 4C). Model fit to binding curves for a maximum polymyxin B concentration of 0.625 µM with (Figure 4A) wt-OMVs and (Figure 4B) over resistant-OMVs. A species component analysis of the eight- step binding curves (Fig.4A bottom panel and Fig.4B bottom panel) associated with fitting the three-state model fit of Fig.3A reveals the fraction of each species present over time for polymyxin B. Figure 4A upper panel and polymyxin B nonapeptide (Fig.4B upper panel), Fig.4C Left panel: EC50 curves obtained at steady-state (time = 999 min) and at the onset of dissociation (indicated by arrow). All binding curves were recorded as outlined in Figure 1D and curves with up to 8-concentration steps were time-shifted to ensure that respective Active 108465066.1.DOCX 4 00B206.1385 concentrations were superimposable. Non-saturating behavior of tightly bound clusters, cPMB, was observed at non-physiologically polymyxin B concentrations (>1 µM), likely due to adsorption of polymyxin B aggregates in bulk liquid, and were not fit. Fig.4C right panel: Expected long-term time-dependency of binding. Simulations were performed based on the estimated rate constants in Table 2 assuming k6 = 3.25 x 10-5 M-1s-1 (Table 4). An EC50 model was fit to the steady-state binding response and to the response immediately after dissociation of the rapidly reversible component composed of PMB-L (time = 1001min). Figures 5A-5C depict transmission electron microscopy (TEM) analysis of OMVs. (Figure 5A) Representative TEM images of wt-OMVs exposed to buffer (left), PMBN (middle), or polymyxin (right) for 40 minutes. Scale bars 500 nm. (Figure 5B) Uncropped TEM images from Figure 1B of wt-OMVs exposed to buffer (left) or polymyxin B (right). Scale bars 200 nm (Figure 5C) Additional representative images of wt-OMVs exposed for polymyxin B for 40 minutes. Scale bars 200 nm. White arrows point to sites of micro- vesiculation or tubules. Figures 6A-6G illustrate SPR analysis of mammalian vesicles and bacterial OMVs using lipophilic LP chips and via amine-coupled polymyxin B to the surface of a C1 chips. (Figure 6A) Representative SPR sensorgram of mammalian extracellular vesicles loading onto LP chips and regeneration. (Figure 6B) Representative SPR sensorgram of wt-OMVs loading onto LP chips and regeneration. (Figure 6C) Representative SPR sensorgram of polymyxin (39 nM to 5 µM) over mammalian vesicles on LP chip. “Referenced” indicates subtraction of binding on blank chip surface from binding to EV-loaded surface. (Figure 6D) Representative SPR sensorgram of wt-OMVs loading onto C1 chip via amine-coupled polymyxin B followed by regeneration. (Figure 6E) SPR binding curves showing a stepwise increase in binding response upon exposure of affinity-captured wt-OMVs to eight serial- doubling concentrations of colistin from 39 nM - 5 µM. (Figure 6F) Representative SPR sensorgram of polymyxin B (39 nM to 5 µM) over wt-OMVs on LP chip. (Figure 6G) Representative SPR sensorgram of resistant-OMVs loading onto amine-coupled polymyxin B-C1 chip and regeneration. Figures 7A-7G provide Lipid A R1 MS- extracted ion chromatograms (EIC) of lipid A and modifications thereof as annotated (top). EIC mass ranges, peak intensity, and identification are as follows: (Figure 7A) EIC m/z 884.07 – 884.10 EIC, 3.16E4, unmodified lipid A; (Figure 7B) EIC m/z 898.08 – 898.12, 1.40E5, unmodified lipid A + C2H4; (Figure 7C) EIC m/z 919.60 – 919.64, 2.73E5, unmodified lipid A + C5H10; (Figure 7D) EIC m/z Active 108465066.1.DOCX 5 00B206.1385 945.07 – 949.11, 1.14E5, singly modified lipid A; (Figure 7E) EIC m/z 959.58 – 959.62, 8.40E5, singly modified lipid A + C2H4; (Figure 7F) EIC m/z 1007.07 – 1007.11, 3.65E5, doubly modified lipid A; and (Figure 7G) EIC m/z 1021.09 – 1021.13, 2.08E6, doubly modified lipid A + C2H4. Lipid A R1 MS- mass spectrum exhibiting relative intensities of doubly charged lipid A and modified ions (bottom). Relative ratios of modified versus unmodified lipid A were compared by extracted ion chromatography peak area. Figure 8 provide chemical structures of polymyxin B and a tail-less polymyxin B derivative, polymyxin B nonapeptide (PMBN) that lacks antibacterial activity. Figure 9 provides concentrations of polymyxin B or PMBN bound expressed per 100 RU of wt-OMVs and resistant-OMVs to normalize for loading differences, calculated from double-referenced sensorgrams and assuming 100RU = 1mg/ml. Mean and standard deviation for n=3 replicates are shown for each condition. Figure 10 provides SPR kinetic traces for binding of polymyxin B to wt-OMVs in the presence of the metal chelator EDTA. Each serial injection profile was recorded as outlined in Figure 1 with a maximum concentration of 625 nM. Representative SPR sensorgram of polymyxin B binding to wt-OMVs in the presence of 4 mM of EDTA. Figures 11A-11B depicts measuring apparent kinetic constants for polymyxin B. (Figure 11A) Chaser analysis of the stable, long-lived interaction between polymyxin B and wt-OMVs. Representative sensorgram of wt-OMV capture, polymyxin B saturation, incubation, and re-saturation with polymyxin B to visualize decreased occupancy. (Figure 11B) Apparent KD of the reversible, saw-tooth binding interaction of polymyxin B by saturating the stable binding component. A representative sensorgram showing saturation of wt-OMVs with polymyxin B followed by serial injection of increasing concentrations of polymyxin B Figure 12 depicts measuring apparent kinetic constants for Brevicidine and Ogiopeptin. Figure 13A-13H depicts binding of polymyxin B and PMBN to E. coli cells, OMVs, and LPS and fitting to approximate kinetic models. (13A and 13B) Affinity analysis of PMBN binding to (13A) E. coli cells and (13B) OMVs using a 1:1 kinetic boundary model fit (Eqn (S3)) to estimate affinity (13C). PMBN binding to LPS (black curves) and replicated (red curves) after saturation of the surface with 1 mM polymyxin B, as shown in (13D). (13D) Saturation of the LPS surface with polymyxin B repeated at two different LPS densities and fit to a simple 1:1 kinetic binding model (Eqn (S1)). (13E and 13F) Binding of polymyxin B to E. coli cells (13E), and OMVs (13F) fit to a 1:1 two-compartment binding Active 108465066.1.DOCX 6 00B206.1385 model (Eqn (S2)). (13G) Binding of polymyxin B to LPS and fit to a two-state kinetic binding model (Eqn (S6)). (13H) Polymyxin B dissociation curves for cells and OMVs pre- saturated with polymyxin B. Polymyxin B occupancy was obtained by chaser SPR analysis, where repeated PMBN injections report changes in polymyxin B occupancy allowing dissociation to be estimated without interference from baseline drift. Figure 14A-14D depicts the three-state model of Fig.3B fitted to experimental SPR binding curves. Titration of polymyxin B to a maximum concentration of 625 nM with wt- OMVs (14A) and resistant-OMVs (14B). The fitted SPR curves are shown in the upper panels with SPR data (black) and model fit (red) together with decomposition of one of these binding curves into component species (lower panels) with PMBL (pink), nPMBL (turquoise), tPMB (dark purple), cPMB (light purple), composite (black). The fitted model is near superimposable upon the experimental SPR binding curves (14A and 14B, upper panels). (14C and 14D) 2D fitspace analysis associated with each fitted data set are shown (right panels). Binding constants were constrained to global values per curve set and the resulting parameter values, standard error associated with the fit, confidence intervals, and ^^2 values are summarized in Table 2. DETAILED DESCRIPTION The presently disclosed subject matter relates to SPR methodologies to record kinetic binding data associated with the interaction of compositions and Gram-negative bacterial cell outer membranes and/or OMVs. This approach can be used to study any composition for its ability to bind to or alter the outer membrane barrier of Gram-negative bacteria and the kinetic analysis outlined herein can provide insight into the mechanisms of action and resistance to such compositions. For purposes of clarity of disclosure and not by way of limitation, the detailed description is divided into the following subsections: 1. Definitions 2. Outer Cell Membranes & Outer Membrane Vesicles 3. SPR to Monitor the Binding of Compositions to Bacterial Cells and/or OMVs 4. Examples 1. Definitions Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. In case of conflict, the present document, including definitions, will control. Preferred methods and Active 108465066.1.DOCX 7 00B206.1385 materials are described below, although methods and materials similar or equivalent to those described herein can be used in practice or testing of the presently disclosed subject matter. All publications, patent applications, patents and other references mentioned herein are incorporated by reference in their entirety. The materials, methods, and examples disclosed herein are illustrative only and not intended to be limiting. The terms “comprise(s),” “include(s),” “having,” “has,” “can,” “contain(s),” and variants thereof, as used herein, are intended to be open-ended transitional phrases, terms, or words that do not preclude the possibility of additional acts or structures. The singular forms “a,” “an” and “the” include plural references unless the context clearly dictates otherwise. The present disclosure also contemplates other embodiments “comprising,” “consisting of”, and “consisting essentially of,” the embodiments or elements presented herein, whether explicitly set forth or not. For the recitation of numeric ranges herein, each intervening number there between with the same degree of precision is explicitly contemplated. For example, for the range of 6-9, the numbers 7 and 8 are contemplated in addition to 6 and 9, and for the range 6.0-7.0, the number 6.0, 6.1, 6.2, 6.3, 6.4, 6.5, 6.6, 6.7, 6.8, 6.9, and 7.0 are explicitly contemplated. As used herein, the term “about” or “approximately” means within an acceptable error range for the particular value as determined by one of ordinary skill in the art, which will depend in part on how the value is measured or determined, i.e., the limitations of the measurement system. For example, “about” can mean within 3 or more than 3 standard deviations, per the practice in the art. Alternatively, “about” can mean a range of up to 20%, preferably up to 10%, more preferably up to 5%, and more preferably still up to 1% of a given value. Alternatively, particularly with respect to biological systems or processes, the term can mean within an order of magnitude, preferably within 5-fold, and more preferably within 2-fold, of a value. 2. Outer Cell Membranes & Outer Membrane Vesicles In certain embodiments, the presently disclosed subject matter relates to SPR methodologies to record kinetic binding data associated with the interaction of compositions and Gram-negative bacterial cell outer membranes. While the methods of the present disclosure are applicable to the analysis of Gram-negative bacteria outer membranes generally, the methods find particular use in connection with the outer membranes of pathogenic Gram-negative bacteria. For example, but not by way of limitation, pathogenic Gram-negative bacteria that find use in connection with the methods of the present disclosure include, but are not limited to: Escherichia coli, Klebsiella pneumoniae, Active 108465066.1.DOCX 8 00B206.1385 Pseudomonas aeruginosa, Enterobacter cloacae, Enterobacter aerogenes, Burkholderia cepecia, Proteus mirabilis, Salmonella, Vibrio cholerae, and Acinetobacter baumannii. In certain embodiments, the Gram-negative bacteria employed in the context of the methods described herein are wild type (wt) Gram-negative bacteria. For example, but not by way of limitation, such wt Gram-negative bacteria will exhibit one or more wt phenotype. Exemplary wt phenotypes include, but are not limited to, wt antibiotic sensitivities. For example, but not by way of limitation, the wt Gram-negative bacteria can exhibit polymyxin sensitivity. In certain embodiments, the wt-Gram-negative bacteria is sensitive to one or more polymyxin, e.g., polymyxin B. In certain embodiments, the Gram-negative bacteria employed in the context of the methods described herein are Gram-negative bacteria that exhibit one or more mutant, i.e., non-wt, phenotype. For example, but not by way of limitation, such mutation may impact a biosynthetic enzyme, e.g., ADP-l-glycero-d-manno-heptose-6-epimerase (waaD). In certain embodiments, such Gram-negative bacteria will exhibit resistance to one or more antibiotic. For example, but not by way of limitation, such resistant Gram-negative bacteria can exhibit resistance to a polymyxin. In certain embodiments, the resistant-Gram-negative bacteria is resistant to polymyxin B. In certain embodiments, the presently disclosed subject matter relates to SPR methodologies to record kinetic binding data associated with the interaction of compositions and outer membrane vesicles (OMVs). In certain embodiments, the OMVs of the instant disclosure are membrane spheres with diameters of about 20 nm to about 250 nm. While the exact composition of OMVs can vary, e.g., based on species and/or culture conditions, they generally capture both the protein and lipid constituents of the outer membrane from which they are derived. In certain embodiments, the OMVs are prepared from Gram-negative bacteria. In certain embodiments, the methods of the present disclosure are applicable to the use of OMVs derived from Gram-negative bacteria generally, although the methods find particular use in connection with OMVs derived from pathogenic Gram-negative bacteria. For example, but not by way of limitation, the OMVs that find use in connection with the methods of the present disclosure include, but are not limited to, those derived from: Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, Enterobacter cloacae, Enterobacter aerogenes, Burkholderia cepecia, Proteus mirabilis, Salmonella, Vibrio cholerae, and Acinetobacter baumannii. Active 108465066.1.DOCX 9 00B206.1385 In certain embodiments, the OMVs employed in the context of the methods described herein are derived from wild type (wt) Gram-negative bacteria. For example, but not by way of limitation, such OMVs will exhibit one or more wt Gram-negative bacteria phenotype. Exemplary wt Gram-negative bacteria phenotypes include, but are not limited to, antibiotic sensitivities. For example, but not by way of limitation, the wt-OMV can exhibit polymyxin sensitivity. In certain embodiments, the wt-Gram-negative bacteria is sensitive to one or more polymyxin, e.g., polymyxin B. In certain embodiments, the OMVs employed in the context of the methods described herein are derived from Gram-negative bacteria that exhibit one or more non-wt phenotype. For example, but not by way of limitation, such OMVs will exhibit resistance to one or more antibiotic. For example, but not by way of limitation, such resistant-OMVs can exhibit resistance to a polymyxin. In certain embodiments, the resistant-OMV is resistant to polymyxin B. 3. SPR to Monitor the Binding of Compositions to Bacterial Cells and/or OMVs In certain embodiments, the present disclosure is directed to methods of identifying a composition capable of binding to a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom. For example, in certain embodiments, the methods described herein comprise contacting the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, with a candidate composition and determining whether the composition complexes with a lipopolysaccharide or other component integral with the outer membrane. In certain embodiments, the determination of whether the composition complexes with a lipopolysaccharide or other component integral with the outer membrane is made using SPR. In certain embodiments, the present disclosure is directed to methods to monitor the binding of a candidate composition to the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, over time. For example, but not by way of limitation, the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, can be contacted with a candidate compound and two or more determinations of whether the composition complexes with a lipopolysaccharide or other component integral with the outer membrane can be made separated in time. In certain embodiments, the determinations of whether the composition complexes with a lipopolysaccharide or other component integral with the outer membrane are made using SPR. Active 108465066.1.DOCX 10 00B206.1385 In certain embodiments, the present disclosure is directed to methods to monitor the binding of a candidate composition to the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, over time and when the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, and the candidate compound are exposed to physical manipulation, e.g., a wash, between the first time point and a subsequent time point. For example, but not by way of limitation, the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, can be contacted with a candidate compound and two or more determinations of whether the composition complexes with a lipopolysaccharide or other component integral with the outer membrane can be made separated in time, where the physical manipulation, e.g., a wash, occurs between the time of the first determination and the time of the second determination. In certain embodiments, the determinations of whether the composition complexes with a lipopolysaccharide or other component integral with the outer membrane are made using SPR. In certain embodiments, the present disclosure is directed to methods to monitor the binding of a candidate composition to the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, over time, where the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, and the candidate compound are exposed to physical manipulation, e.g., a wash, between the first time point and a subsequent time point. In certain embodiments, the methods disclosed herein encompass determining whether the candidate compound and is retained in the outer membrane over time. For example, the methods of the present disclosure encompass determining whether the candidate compound is retained for up to about 1 to about 120 minutes, up to about 1 to about 90 minutes, up to about 1 to about 60 minutes, or up to about 1 to about 30 minutes. In certain embodiments, the methods of the present disclosure encompass determining whether the candidate compound is retained up to about 1, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, about 10, about 11, about 12, about 13, about 14, about 15, about 16, about 17, about 18, about 19, about 20, about 21, about 22, about 23, about 24, about 25, about 26, about 27, about 28, about 29, or about 30, about 31, about 32, about 33, about 34, about 35, about 36, about 37, about 38, about 39, about 40, about 41, about 42, about 43, about 44, about 45, about 46, about 47, about 48, about 49, about 50, about 51, about 52, about 53, about 54, about 55, about 56, about 57, about 58, about 59, or about 60 minutes, about 61, about 62, about 63, about 64, about 65, about 66, about 67, about 68, about 69, about 70, about 71, about 72, about 73, about 74, Active 108465066.1.DOCX 11 00B206.1385 about 75, about 76, about 77, about 78, about 79, about 80, about 81, about 82, about 83, about 84, about 85, about 86, about 87, about 88, about 89, or about 90 minutes, about 91, about 92, about 93, about 94, about 95, about 96, about 97, about 98, about 99, about 100, about 101, about 102, about 103, about 104, about 105, about 106, about 107, about 108, about 109, about 110, about 111, about 112, about 113, about 114, about 115, about 116, about 117, about 118, about 119, or about 120 minutes. In certain embodiments, the methods described herein encompass determining whether the candidate compound is retained for at least about 1 to about 120 minutes, about 1 to about 90 minutes, about 1 to about 60 minutes, or about 1 to about 30 minutes. In certain embodiments, the methods of the present disclosure encompass determining whether the candidate compound is retained at least about 1, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, about 10, about 11, about 12, about 13, about 14, about 15, about 16, about 17, about 18, about 19, about 20, about 21, about 22, about 23, about 24, about 25, about 26, about 27, about 28, about 29, or about 30, about 31, about 32, about 33, about 34, about 35, about 36, about 37, about 38, about 39, about 40, about 41, about 42, about 43, about 44, about 45, about 46, about 47, about 48, about 49, about 50, about 51, about 52, about 53, about 54, about 55, about 56, about 57, about 58, about 59, or about 60 minutes, about 61, about 62, about 63, about 64, about 65, about 66, about 67, about 68, about 69, about 70, about 71, about 72, about 73, about 74, about 75, about 76, about 77, about 78, about 79, about 80, about 81, about 82, about 83, about 84, about 85, about 86, about 87, about 88, about 89, or about 90 minutes, about 91, about 92, about 93, about 94, about 95, about 96, about 97, about 98, about 99, about 100, about 101, about 102, about 103, about 104, about 105, about 106, about 107, about 108, about 109, about 110, about 111, about 112, about 113, about 114, about 115, about 116, about 117, about 118, about 119, or about 120 minutes. In certain embodiments, the binding of the candidate composition induces the death of the Gram-negative bacterial cell. In certain embodiments, the retention of the candidate composition in the outer membrane identifies the composition as capable of inducing the death of Gram-negative bacterial cells. In certain embodiments, such cytotoxic retention is up to about 1 to about 120 minutes, up to about 1 to about 90 minutes, up to about 1 to about 60 minutes, or about 1 to about 30 minutes, e.g., up to about 1, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, about 10, about 11, about 12, about 13, about 14, about 15, about 16, about 17, about 18, about 19, about 20, about 21, about 22, about 23, about 24, about 25, about 26, about 27, about 28, about 29, about 30, about 31, about 32, about 33, about 34, about 35, about 36, about 37, about 38, about 39, about 40, about Active 108465066.1.DOCX 12 00B206.1385 41, about 42, about 43, about 44, about 45, about 46, about 47, about 48, about 49, about 50, about 51, about 52, about 53, about 54, about 55, about 56, about 57, about 58, about 59, or about 60 minutes, about 61, about 62, about 63, about 64, about 65, about 66, about 67, about 68, about 69, about 70, about 71, about 72, about 73, about 74, about 75, about 76, about 77, about 78, about 79, about 80, about 81, about 82, about 83, about 84, about 85, about 86, about 87, about 88, about 89, or about 90 minutes, about 91, about 92, about 93, about 94, about 95, about 96, about 97, about 98, about 99, about 100, about 101, about 102, about 103, about 104, about 105, about 106, about 107, about 108, about 109, about 110, about 111, about 112, about 113, about 114, about 115, about 116, about 117, about 118, about 119, or about 120 minutes. In certain embodiments, such cytotoxic retention is at least about 1 to about 120 minutes, at least about 1 to about 90 minutes, at least about 1 to about 60 minutes, or at least about 1 to about 30 minutes, e.g., at least about 1, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, about 10, about 11, about 12, about 13, about 14, about 15, about 16, about 17, about 18, about 19, about 20, about 21, about 22, about 23, about 24, about 25, about 26, about 27, about 28, about 29, about 30, about 31, about 32, about 33, about 34, about 35, about 36, about 37, about 38, about 39, about 40, about 41, about 42, about 43, about 44, about 45, about 46, about 47, about 48, about 49, about 50, about 51, about 52, about 53, about 54, about 55, about 56, about 57, about 58, about 59, or about 60 minutes, about 61, about 62, about 63, about 64, about 65, about 66, about 67, about 68, about 69, about 70, about 71, about 72, about 73, about 74, about 75, about 76, about 77, about 78, about 79, about 80, about 81, about 82, about 83, about 84, about 85, about 86, about 87, about 88, about 89, or about 90 minutes, about 91, about 92, about 93, about 94, about 95, about 96, about 97, about 98, about 99, about 100, about 101, about 102, about 103, about 104, about 105, about 106, about 107, about 108, about 109, about 110, about 111, about 112, about 113, about 114, about 115, about 116, about 117, about 118, about 119, or about 120 minutes. In certain embodiments, the methods disclosed herein encompass determining whether the candidate compound and accumulates in the outer membrane over time. For example, the methods of the present disclosure encompass determining whether the candidate compound accumulates in the outer membrane to a mass fraction of between about 1% to about 30% (weight of accumulated composition / weight of outer membrane). In certain embodiments, the methods of the present disclosure encompass determining whether the candidate compound accumulates in the outer membrane to a mass fraction of up to about 1%, about 2%, about 3%, about 4%, about 5%, about 6%, about 7%, about 8%, Active 108465066.1.DOCX 13 00B206.1385 about 9%, about 10%, about 11%, about 12%, about 13%, about 14%, about 15%, about 16%, about 17%, about 18%, about 19%, about 20%, about 21%, about 22%, about 23%, about 24%, about 25%, about 26%, about 27%, about 28%, about 29%, or about 30%. In certain embodiments, the methods described herein encompass determining whether the candidate compound accumulates in the outer membrane to a mass fraction of at least about 1% to about 30%, e.g., at least about 1%, about 2%, about 3%, about 4%, about 5%, about 6%, about 7%, about 8%, about 9%, about 10%, about 11%, about 12%, about 13%, about 14%, about 15%, about 16%, about 17%, about 18%, about 19%, about 20%, about 21%, about 22%, about 23%, about 24%, about 25%, about 26%, about 27%, about 28%, about 29%, or about 30%. In certain embodiments, the accumulation of the candidate composition in the outer membrane identifies the composition as capable of inducing the death of Gram-negative bacterial cells. In certain embodiments, such cytotoxic accumulation is up to about 1% to about 30%, e.g., up to about 1%, about 2%, about 3%, about 4%, about 5%, about 6%, about 7%, about 8%, about 9%, about 10%, about 11%, about 12%, about 13%, about 14%, about 15%, about 16%, about 17%, about 18%, about 19%, about 20%, about 21%, about 22%, about 23%, about 24%, about 25%, about 26%, about 27%, about 28%, about 29%, or about 30%. In certain embodiments, such cytotoxic accumulation is at least about 1% to about 30%, e.g., at least about 1%, about 2%, about 3%, about 4%, about 5%, about 6%, about 7%, about 8%, about 9%, about 10%, about 11%, about 12%, about 13%, about 14%, about 15%, about 16%, about 17%, about 18%, about 19%, about 20%, about 21%, about 22%, about 23%, about 24%, about 25%, about 26%, about 27%, about 28%, about 29%, or about 30%. In certain embodiments, the methods disclosed herein encompass determining whether the candidate compound and accumulates in the outer membrane over time. For example, the methods of the present disclosure encompass determining whether the candidate compound accumulates in the outer membrane to a mass fraction of between about 1% to about 30% (weight of accumulated composition / weight of outer membrane). In certain embodiments, the methods of the present disclosure encompass determining whether the candidate compound accumulates in the outer membrane to a mass fraction of up to about 1%, about 2%, about 3%, about 4%, about 5%, about 6%, about 7%, about 8%, about 9%, about 10%, about 11%, about 12%, about 13%, about 14%, about 15%, about 16%, about 17%, about 18%, about 19%, about 20%, about 21%, about 22%, about 23%, about 24%, about 25%, about 26%, about 27%, about 28%, about 29%, or about 30%. In Active 108465066.1.DOCX 14 00B206.1385 certain embodiments, the methods described herein encompass determining whether the candidate compound is retained for at least about 1 to about 120 minutes, about 1 to about 90 minutes, about 1 to about 60 minutes, or about 1 to about 30 minutes. In certain embodiments, the methods of the present disclosure encompass determining whether the candidate compound is retained at least about 1, about 2, about 3, about 4, about 5, about 6, about 7, about 8, about 9, about 10, about 11, about 12, about 13, about 14, about 15, about 16, about 17, about 18, about 19, about 20, about 21, about 22, about 23, about 24, about 25, about 26, about 27, about 28, about 29, or about 30, about 31, about 32, about 33, about 34, about 35, about 36, about 37, about 38, about 39, about 40, about 41, about 42, about 43, about 44, about 45, about 46, about 47, about 48, about 49, about 50, about 51, about 52, about 53, about 54, about 55, about 56, about 57, about 58, about 59, or about 60 minutes, about 61, about 62, about 63, about 64, about 65, about 66, about 67, about 68, about 69, about 70, about 71, about 72, about 73, about 74, about 75, about 76, about 77, about 78, about 79, about 80, about 81, about 82, about 83, about 84, about 85, about 86, about 87, about 88, about 89, or about 90 minutes, about 91, about 92, about 93, about 94, about 95, about 96, about 97, about 98, about 99, about 100, about 101, about 102, about 103, about 104, about 105, about 106, about 107, about 108, about 109, about 110, about 111, about 112, about 113, about 114, about 115, about 116, about 117, about 118, about 119, or about 120 minutesIn certain embodiments of the methods disclosed herein, the Gram-negative bacterial cell, or OMV derived therefrom, is immobilized on an SPR sensor substrate. In certain embodiments, the Gram-negative bacterial cell, or OMV derived therefrom, is immobilized on an SPR sensor substrate via an affinity coupling. In certain embodiments, the affinity coupling comprises amine-coupled polymyxin B immobilized on the SPR sensor substrate. In certain not limiting embodiments, but polymyxin B can be covalently attached to a sensor substrate, e.g., a C1 chip (Series S Sensor Chip C1, Cytiva), using an amine coupling kit (Cytiva). For example, but not by way of limitation, equal amounts of reagents N-hydroxysuccinimide and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride can be mixed and added to the gold chip surface, e.g., for about 2 min, washed, e.g., with distilled deionized water, and dried. Sufficient polymyxin B (e.g., 1 mM in 1 M HEPES buffer, pH 8) can be added to cover the chip surface, incubated at room temperature, e.g., for about 1 h, washed, and then incubated with 1 M ethanolamine hydrochloride-NaOH, pH 8.5, for about 1 min. The chip surface can then be washed with water, dried, immediately loaded into a SPR device, e.g., a Biacore S200, and primed with running buffer (described below) to equilibrate the system. Chips, e.g., C1 chips, can be Active 108465066.1.DOCX 15 00B206.1385 used for multiple runs and discarded upon removal from the SPR device, e.g., a Biacore S200. In certain embodiments the SPR experiments, e.g., experiments using a Biacore S200 SPR device and C1-chips, can be performed in running buffer (e.g., Dulbecco's phosphate-buffered salt solution 1x without calcium or magnesium (Fisher Scientific), pH 7.4, with 0.0005% tween-80 (Sigma) passed through a 0.2 µm filter). MgCl2 can be added as indicated throughout the figures and examples. In certain embodiments, inclusion of tween-80 can prevent loss of candidate compositions, e.g., polymyxins, to the plastics. In certain embodiments, e.g., experiments with brevicidine as the candidate composition, the candidate composition can be suspended in DMSO, and DMSO, e.g., 0.125% DMSO, can be included in running buffer. In certain embodiments, the analysis and compartment temperature of the SPR device can be set to 25°C or 37°C. OMVs can be diluted from frozen stocks to approximately 20-30 µg/ml protein in OMV buffer. Capture of OMVs can be performed at low flow rate, e.g., 5 µl/ml for 300 sec over the test channel(s), followed by a stabilization period, e.g., about 300 sec stabilization period. All subsequent steps can be performed at a higher flow rate, e.g., at a flow rate of 40 µl/ml. For single cycle kinetics, two-fold dilutions can be injected over the channel(s) loaded with OMVs and a reference channel without OMVs, e.g., for about 30 sec contact and about 480 sec dissociation. To regenerate chips, detergent, e.g., 0.5% SDS (desorb 1, Cytiva), can be injected into all channels, e.g., for 60 sec at 30 µl/ml twice, and can involve an extra buffer wash and four carry-over control steps to prevent residual SDS from disrupting the following cycle. In certain embodiments, SPR on whole bacterial cells can be performed and analyzed as described for OMVs on C1 chip, but using a lipophilic chip (e.g., LP (Xantec), or L1 (Cytiva)), and with an additional carry-over wash prior to the capture. Regeneration of the chip after whole cell binding can also include additional steps (e.g., 40 µl/ml flow rate): 1) PBS supplemented with 32 mM MgCl2 for 120 sec, 2) 2.5 M NaCl for 30 sec, 3) 0.5% SDS (Desorb 1, Cytiva) for 60 s with carry over controls between. In certain embodiments, the apparent KD of the reversible binding event occurring between the candidate composition and the Gram-negative bacteria or OMV derived therefrom, single-cycle kinetics were performed. For example, in certain embodiments relating to analysis of OMVs, after OMVs are loaded, an appropriate amount of candidate composition, e.g., 5 µM polymyxin B, is injected, e.g., for about 360 sec (at about 40 µl/ml flowrate) followed by about 120 sec dissociations to saturate the stable-binding population prior to the sample injections. The base-to-peak value of each trace can then be determined Active 108465066.1.DOCX 16 00B206.1385 from double-referenced traces (reference 1: - OMVs/ + compound channel; reference 2: + OMVs/ - compound) exported from the SPR software, e.g., Biacore S200 evaluation software into PRISM 9 software. The change in RU with each pulse can be plotted over the concentration and the KD determined by fitting a non-linear regression, one-site total function with background parameter set to 0 (GraphPad Prism version 9.3.1 for Mac, GraphPad Software, San Diego, California USA, www.graphpad.com). In certain embodiments, determination of the apparent KD can occur without any pre-saturation step prior to kinetic pulses. In certain embodiments, the determination of the residence time of a candidate compound with OMVs can be performed using the ‘chaser method’ to account for drift that can occur in the system over long incubation times. In such instances, OMVs can be loaded onto the chip as described above then the candidate composition, e.g., 5 µM polymyxin B, can be injected using the ‘low-sample consumption’ setting, e.g., for 300 sec (5 µl/ml flowrate), followed by a 240 sec dissociation time. A second dose of the candidate composition, e.g., 5 µM polymyxin B, can then injected, e.g., for 60 sec (30 µl/ml) to assure saturation followed by a 2 h dissociation time prior to the ‘chaser’, which is a subsequent pulse of the candidate composition, e.g., a 60 sec pulse of 5 µM polymyxin B. The RUs of associated the candidate composition can be determined after the start of the dissociation, e.g., after 120 seconds. The change in RUs from before dosing to after the second dose, and the change in RUs from prior to the ‘chaser’ dose and after the start of its dissociation can be used to calculate the fraction occupancy of the candidate composition and the residence time/half-life as described. In certain embodiments of the methods disclosed herein, the candidate composition is a candidate cytotoxic composition. In certain embodiments of the methods disclosed herein, the candidate cytotoxic composition is a candidate antibacterial. In certain embodiments, the candidate antibacterial is a polypeptide antibacterial or polypeptide analog antibacterial. In certain embodiments, the candidate antibacterial is a polymyxin. In certain embodiments, the candidate composition is natural product antibiotic. In certain embodiments, the natural product antibiotic is brevicidine, ogipeptin, or pedopeptin. Active 108465066.1.DOCX 17 00B206.1385 4. EXAMPLES A. Introduction Densely packed LPS in the extracellular leaflet of the Gram-negative bacterial outer membrane establishes a permeability barrier. Polymyxins, including polymyxin B and colistin, are last-resort antibiotics that directly bind the conserved lipid A anchor of LPS, but the molecular mechanism underpinning their antibacterial activity remains unresolved. Here, SPR is used to kinetically interrogate the interactions between polymyxins and LPS in bacterial cells as well as OMVs and pure LPS films, which are effective surrogates that capture the natural complexity of the outer membrane and afford more experimental robustness. The information-rich SPR data presented herein enables application of a mechanistic model showing that LPS catalyzes super-stoichiometric accumulation of polymyxins via three-states: (1) transient binding to the lipid A anchor of LPS facilitates (2) membrane insertion of polymyxins and promotes (3) phase-separation of long-lived polymyxin B clusters that reach a high density. A proposed microscopic mechanism is consistent with the kinetic measurements as well as previous observations and is generalizable to other lipid A-targeting antibacterial molecules, including brevicidine. Bacterial killing is not observed when the outer membrane fails to accumulate polymyxin clusters as observed for polymyxin B nonapeptide, which lacks an acyl tail, and when lipid A is modified, leading to polymyxin resistance. Transient binding to LPS is sufficient for outer membrane permeabilization, whereas cluster formation upon lipid A binding is critical for cell killing. OMV-based SPR is a useful platform for dissecting interactions with the outer membrane able to yield mechanistic insight that enables efforts to discover LPS- targeting antibiotics among other cytotoxic compositions. B. Results a. Binding of Polymyxin B to the Outer Membrane by a Complex Kinetic Process Polymyxin B targets the conserved lipid A anchor of LPS and exhibits potent and selective Gram-negative antibacterial activity, which was confirmed by determining the minimal inhibitory concentration (MIC), or the concentration of compound required to completely inhibit bacterial growth, for polymyxin B. An MIC of 0.08 nM for polymyxin B was measured against wild-type E. coli, a model Gram-negative species, whereas this antibiotic had no detectable activity at the highest tested concentration against a Gram- Active 108465066.1.DOCX 18 00B206.1385 positive strain, Staphylococcus aureus, which lacks lipid A and an outer membrane (Table 1). Table 1. Minimum Inhibitory Concentrations (MICs) of polymyxin B and rifampicin potentiation against wild-type and polymyxin-resistant strains. MIC (µM)¹ species background Polymyxin B PMBN² Brevicidine E. coli WT 0.00008 >40 0.313 PmrA^G53E 10 >40 0.625 mcr1³ 2.5-5 >40 0.313 WT+rifampicin⁴ <0.00002 0.078 0.078 PmrA^G53E+rifampicin⁴ 0.001 20 0.078 mcr-1+rifampicin⁴ 0.001 2.5 0.078 S. aureus WT >40 >40 >40 1 Minimal Inhibitory Concentrations (MICs): lowest concentration of antibiotic that completely inhibits bacterial growth 2 PMBN, polymyxin B nonapeptide 3 Strain carrying a plasmid encoding the mcr-1 gene 4 Rifampicin present at 1.56 µM, which has no effect on growth of E. coli or polymyxin-resistant E. coli strains tested In order to characterize the kinetics of the interaction between polymyxin B and LPS in the outer membrane, SPR was employed to monitor binding to either whole E. coli cells or OMVs that were linked to a planar surface (Fig.1A). OMVs are useful surrogates for the Gram-negative outer membrane, and when treated with polymyxin B they exhibited vesiculation and tubules (Fig.1B and Figs.5A-5C) similar to those perturbations observed on polymyxin-treated bacterial cells. While the sensitivity of SPR decays exponentially with distance from the planar surface and is reduced to about 30% at 270 nm, it is sufficient to detect binding to both whole E. coli cells (diameter of approximately 0.5 µm to about 1 µm along their shorter axis) and OMVs (diameters of about 20 nm to about 250 nm) range, which fit entirely within the sensitivity range of the sensor (Fig.1A). SPR optically probes the volume within this sensitivity depth to produce an averaged refractive index change, represented as response units (RUs), that is proportional to a change in average concentration. The response is also weakly sensitive to mass redistribution within this volume because detection sensitivity decays exponentially from the surface. Membrane vesicles from Expi293 cells and Gram-negative bacterial OMVs were initially immobilized non-specifically on the surface of a lipophilic chip, however this approach was not compatible with detergents needed to reduce non-specific interactions, exhibited poor Active 108465066.1.DOCX 19 00B206.1385 regeneration, and was complicated by interactions between polymyxin B and the chip surface (Figs.6A-6C). To overcome these limitations, OMVs were immobilized via amine- coupled polymyxin B to the surface of a C1 chip. OMVs were stably bound to this chip surface in the presence of 0.0005% tween-80 and OMVs bound on the surface could be regenerated by standard methods, leaving the amine-coupled polymyxin B intact for subsequent OMV capture (Fig. 6D). Electron microscopy showed discretely bound wild- type OMVs (wt-OMVs) that retained their spherical shape when attached to the planar surface (Fig.1C). To measure the interactions of polymyxin B with immobilized vesicles and cells, single-cycle kinetic (SCK) injection and multicycle injection formats were employed. SCK finds particular use in cases where there is an accumulation of long-lived bound species as it is possible to obtain a full dose-response range in a single binding curve. Adapting the SPR format, contact time, and dosing regimen allowed the complex kinetic processes of polymyxin B binding to the outer membrane to be resolved (Fig.1D and 1E). Binding response curves for serial injection of increasing concentrations of polymyxin B (Fig. 1D and 1E) or colistin (Fig. 6E) over captured wt-OMVs revealed a binding profile dominated by the accumulation of tightly bound polymyxin B and a superimposed saw-tooth profile associated with transient binding. Importantly, an identical binding profile was observed using OMVs non-specifically immobilized on the surface of a lipophilic chip (Fig. 6F), ruling out any potential effects of immobilization via amine- coupled polymyxin B. This saw-tooth profile was less apparent for low-concentration injections likely because any polymyxin B-LPS complexes that formed must rapidly transition to the tightly bound state that accumulated. Therefore, the observed curvature at low concentrations reflects the effective rate of accumulation of the tightly bound complex. At high concentrations, the binding capacity for tightly bound polymyxin B approaches saturation, allowing the weak polymyxin B binding to OMVs to dominate, with its characteristic saw-tooth profile. At saturation, the tightly bound component accounted for a higher proportion of total binding, approximately 2:1 over the reversible component (Fig. 1D and 1E)). b. Modifications to Lipid A and Polymyxin B Alter the OMV Binding Profile Covalent modifications of lipid A phosphate groups, including pEtN and L-ara4N, lead to polymyxin resistance, presumably by disrupting binding. It was confirmed that E. coli with a mutation in pmrA (PmrA^G53E) or carrying a plasmid with an inducible mcr-1 Active 108465066.1.DOCX 20 00B206.1385 gene (pBAD-mcr-1) were resistant to killing by polymyxin B (Table 1). As previously observed, polymyxin B still potentiated the activity of rifampicin, an antibiotic normally excluded by the outer membrane, consistent with outer membrane disruption (Table 1). OMVs isolated from polymyxin-resistant bacteria (resistant-OMVs) were composed of modified lipid A (Table 3 and Fig.7A-7G) but could still be immobilized via amine-coupled polymyxin B to the surface of a C1 chip (Fig.6G). Compared to wt-OMVs (Fig.2A), serial injection of increasing concentrations of polymyxin B over captured resistant-OMVs exhibited only weak, saw-tooth binding (Fig.2B). The same polymyxin B binding profiles were observed with both wild-type and polymyxin-resistant whole bacterial cells indicating OMVs are a faithful model for studying outer membrane interactions (Fig.2C and 2D). All subsequent experiments were performed with OMVs as they eliminated experiment-to- experiment variability, likely caused by stochastic population differences among growing bacterial cells, eliminated compounding biological considerations, and enabled sufficient loading on the SPR chip surface for quantitative assessments. Table 3. Proportion of modified LPS in OMV preparations used to measure interactions with polymyxin B, PMBN, and brevicidine by SPR Lipid A species O^{MVs Unmodified Single-} Double- m_{odified modified} ^{Total Modified} Wild-type 100 0 0 0 pmrA^G53E batch 1 12.3 27.7 60.0 87.7 pmrA^G53E batch 2 8.8 20.1 71.1 91.2 mcr1 31.3 68.7 0 68.7 A polymyxin B variant lacking its terminal amino acid and acyl tail, polymyxin B nonapeptide (PMBN) (Fig.8), does not exhibit the antibacterial activity of polymyxin B but still permeabilizes the outer membrane to rifampicin (Table 1). Serial injection of increasing concentrations of PMBN over wt-OMVs (Fig. 2E) or resistant-OMVs (Fig. 2F) exhibited only weak, saw-tooth binding. To directly compare the saw-tooth binding fractions, RUs were converted to concentrations, thus normalizing for molecular weight, and this revealed that the saw-tooth fraction binding capacity of OMVs is similar across conditions with the exception of the lowest concentrations of polymyxin B when stable binding dominates (Fig. 9). Active 108465066.1.DOCX 21 00B206.1385 c. Divalent Cations Disrupt Polymyxin B Binding to OMVs Divalent metal cations stabilize the LPS network and compete with polymyxins for the negatively charged phosphates of lipid A. To determine if excess divalent cations compete with or limit binding of polymyxin B to OMVs, single-cycle kinetic binding in the presence of excess magnesium ions (Mg2+) was performed. In the presence of 32 mM Mg2+, the interaction of polymyxin B with OMVs was eliminated or reduced (Fig.2G) and the interaction of PMBN with OMVs was eliminated (Fig.2H). Moreover, EDTA, a metal chelator that disrupts outer membrane permeability by removing divalent cations (4), had no effect on polymyxin binding (Fig. 10) indicating that divalents do not interfere with polymyxin B binding under these conditions. These findings are consistent with a role for charge-based interactions in both the association of polymyxin B with the outer membrane. d. Brevicidine Binding to OMVs is Similar to Polymyxin B Brevicidine was recently identified as a cationic, non-ribosomal, natural product peptide with selective activity against Gram-negative bacteria due to LPS targeting. MICs of 0.313 µM against E. coli and >40 µM against S. aureus, confirmed this activity profile (Table 1). Brevicidine exhibited a complex interaction with wt-OMVs similar to that observed for polymyxin B, suggesting a common mechanism (Fig.2I). Unlike polymyxin B, however, brevicidine exhibits antibacterial activity against strains with polymyxin- resistant lipid A modifications (Table 1). Strikingly, and consistent with its antibacterial activity, brevicidine binding to polymyxin-resistant-OMVs was indistinguishable from its binding to wt-OMVs (Fig.2I). e. Polymyxins Bind to OMVs with a Long Half-Life Binding of polymyxin B to OMVs was too complex for kinetic constants to be extracted using standard modeling. Thus, to determine the off-rate of the tightly bound fraction an established chaser analysis was used, where OMVs were pre-saturated with polymyxin B and then re-saturated after an extended time interval (Fig.7A). The amount of accumulation upon re-saturation relative to the initial saturation step revealed the loss in occupancy and allowed calculation of a residence time ( ^ = 1/k) and half-life for the interaction (t_1/2 = ln2/k). A t_1/2 of >6 hours was determined for polymyxin B binding to OMVs at both 25°C and 37°C (Table 4). The t_1/2 of the interaction between the LPS-targeting natural product brevicidine, also measured by chaser analysis, was approximately 1 hour on both wt-OMVs and polymyxin-resistant-OMVs (Table 4). Active 108465066.1.DOCX 22 00B206.1385 Table 4. Half-life (t_1/2) and k_off of polymyxin B and brevicidine binding to immobilized wt-OMVs and resistant OMVs measured by SPR at 25°C and 37°C with standard deviations. Antibacterial OMVs Temp (°C) t1/2 (h)¹ koff (s^-1)² Polymyxin B wt 25 6.86 (1.28) 3.25x10^-5 (0.66x10^-5) wt 37 7.27 (2.15) 2.65x10^-5 (0.75x10^-5) Brevicidine wt 25 1.09 (0.20) 1.77x10^-4 (0.32x10^-5) Resistant³ 25 1.06 (0.56) 1.82x10^-4 (0.98x10^-5) ¹ t1/2 determined by chaser analysis as described in text. ² koff calculated using the measured residence time as described in text. ³ polymyxin B resistant OMVs isolated from PmrA^G53E strain In order to measure an apparent-KD of the reversible saw-tooth binding step more approximate methods were used. To estimate the KD for the transient species, the stable binding component was first pre-saturated and then polymyxin B was titrated over the surface as before (Fig. 7B). The resulting binding curve displayed steady-state regions at the maxima of each saw-tooth profile, enabling a simple affinity isotherm fit. No pre- saturation was necessary when using resistant-OMVs or PMBN as these lack the stable interaction (Fig.2B, 2E, and 2F). The change in RUs from the base to the plateau of each step was used to construct a standard affinity plot that enabled determination of an apparent- KD of 517 nM for the reversible, saw-tooth binding of polymyxin B (Table 5). The measured apparent-KDs were not significantly different among the treatment groups, suggesting that with reversible, saw-tooth binding events for polymyxin B and PMBN with wt-OMVs and resistant-OMVs are equivalent. Table 5. Equilibrium binding constants (Apparent-KDs) of polymyxin B and PMBN binding to immobilized wt-OMVs measured by SPR with standard deviations. Compound OMVs Apparent KD [nM] (SD) Polymyxin B WT (isolated reversible binding) 517 (449) PMBN WT 288 (153) Polymyxin B Resistant 273 (221) PMBN Resistant 371 (243) f. Mechanistic Studies of Polymyxins Binding to Cells, OMVs, and LPS Additional SPR experiments were performed to measure binding of polymyxin B and PMBN to E. coli cells, OMVs, and purified LPS under physiological concentrations of salts and divalent cations and at 37ºC. Without being bound by theory, these experiments Active 108465066.1.DOCX 23 00B206.1385 are analyzed using the interaction models described herein. For example, the interaction models introduced here (Eqn (S1-S3)) are annotated in terms of binding of polymyxins (polymyxin B [PMB]) to LPS but also apply to any other affinity binding pair. Molar concentrations of bound PMB are converted to SPR response by R = G.MW.[PMB], where MW is the molecular weight of PMB, and G is a mass to response conversion factor. A full quantitative interpretation of SPR sensitivity has been reported¹. In a general case, a protein concentration of 0.01 (g/L) produces a response of 1 response unit (RU) yielding the proportionality constant G=100 (L/g). Here, we assume protein is distributed uniformly within a 100 nM thick hydrogel, dissolved in buffer at near physiological salt concentrations in the absence of high concentrations of high refractive index substances, such as glycerol and DMSO. This calibration may be adjusted for variability in the refractive index increment when working with buffers or other classes of molecules such as nucleic acids, polysaccharides, or drug-like molecules. In addition, binding interactions confined to planar sensing surfaces (i.e., no hydrogel matrix) results in a 1.3-fold increased response due to minimal decay in SPR sensitivity when close to the sensing surface (e.g., < 10 nm). Change in protein accumulation over time produces responses that are indicated here by a time subscript (e.g., R(t)). A simple 1:1 model for binding of PM to LPS to form an affinity complex PML is given by the simple 1:1 pseudo-first-order model defined by Eqn (S1). DI/dt = (kon.[PMB]i.(Rmax-R(t)) -koff.R(t)) Eqn (S1) Where R is the SPR response for accumulation of affinity complex PMBL. [PMB]i is the injected concentration of PM and Rmax is the saturation response assuming full target occupancy. This model assumes that mass transport of PMB within the flow cell, which governs the rate of exchange of PMB between the bulk liquid and the sensing surface, is non-limiting. However, this assumption does not hold in many cases requiring addition of a mass transport term resulting in the 1:1 two-compartment model as shown in Eqn (S2) d(R)/dt = (kon.[PMB].(Rmax-R(t)) -koff.R(t)) + kt.([PMB]i(t) – [PMB]) Eqn (S2) Where [PMB]i_(t) is the injected concentration profile of PMB with respect to time and the response normalized mass transport constant is kt (units, RU/Ms) = kt’.G.MWPMB.Rmax where kt’ (units, m/s) is the advective mass transport rate constant. This model allows estimation of kinetic parameters to be estimated despite the influence of mass transport limitation. However, when mass transport is entirely dominant then any observable curvature will be due to mass transport limitation and sensitivity to kinetic binding constants Active 108465066.1.DOCX 24 00B206.1385 is lost entirely. In this extreme case Eqn (S2) simplifies to Eqn (S3), which we refer to as a diffusion boundary model². dR/dt = kt.[PMB]i(t) - kt.KD/(Rmax/Rt -1) Eqn (S3) Where the dissociation affinity constant K_D = k_off/k_on Importantly, the kinetics of binding does not appear in Eqn (S3) because the observable kinetic curvature is associated with formation and decay of the mass transport boundary layer and not binding kinetics. In general, Eqn (2) can be used even when mass transport limitation is negligible as in this case kt will tend towards infinity and, therefore, will not influence estimation of the kinetic binding constants. However, fitting of Eqn (2) in cases where binding is completely dominated by mass transport limitation is problematic because the goodness-of-fit, kinetic rate constants, and associated standard error can sometimes appear reasonable but in reality, the returned rate constants may remain entirely driven by boundary layer kinetics and not related to binding rate constants. Conditions requiring application of Eqn (3) can be defined using the Damköhler number (Da) as full transport limitation can be assumed when the binding flux Lr = kon*(Rmax-R(t)) >> kt, which is equivalent to Da = kt /Lr >> 1. Therefore, a high kon combined with a high surface binding capacity (Rmax-R(t)) will likely result in Da >>1. Increasing k_t would overcome this limitation and restore sensitivity to kinetic binding but this is usually not possible in practice because kt cannot be increased by orders of magnitude since it is a function of flow cell geometry (which is fixed), analyte diffusion (which is fixed for a given experimental condition) and is weakly dependent on flow rate³. However, overcoming complete mass transport limitation is possible by lowering Lr by limiting the concentration of target at the sensing surface. Polymyxin B (PMB) binding to whole E. coli cells, OMVs, and LPS forms transient complexes with available LPS in a fully transport limited regime and a fraction of these transient complexes (R1) transform into long-lived complexes (R2). Here, we neglect Eqn (S3) and crudely approximated the observed binding as an isomerization process, using a simple two-state model to provide a qualitative analysis. dR1/dt = kt.[PMBi](t) - kt.KD1/(Rmax1/R1t -1) - (kon2.R1(t) - koff2.R2(t)) Eqn (S4) d(R2)/dt = (k_on2.R1(t) -k_off2.R2(t)) Eqn (S5) Total Response R(t) = R1(t) + R2(t) Eqn (S6) Active 108465066.1.DOCX 25 00B206.1385 This two-state approximation also does not correctly account for the finite capacity of each target to retain PMB for extended periods. Changes in PMB occupancy can be obtained indirectly from SPR binding using chaser SPR analysis (see Section 1D - Multipoint Chaser Binding). Briefly, repeated PMBN injections over a sensing surface that has been pre-saturated with PMB are exploited to report loss in PMB occupancy over an extended time period. Progressive dissociation of PMB causes a corresponding increase in PMBN binding since PMBN binding is proportional to free LPS and therefore varies inversely with PMB dissociation. These time course measurements allow a PMB dissociation curve to be plotted and fit without interference from baseline drift. A simple 1:1 binding interaction produces a single exponential decay curve while biphasic decay curves are observed when there are two concurrent dissociation processes occurring. Here, both mass transport and the association process are neglected and the biphasic dissociation curve are crudely modeled as two-independent sites, where occupancy at each site follows independent 1:1 binding and was fit to the analytic two-site model given in Eqn (S7). Total Response R(t) = Rmax1.Exp(-kd1*t) + Rmax2.Exp(-kd2*t) Eqn (S7) Where Rmax1 and Rmax2 are the saturation responses for each binding site and kd1 and kd2 are the apparent dissociation rate constants. In the experiments described herein, SCK injections have been employed as this produces information-rich titration curves over a range of compound concentrations without requiring surface regeneration and is often preferable for analysis of tightly bound compounds. However, multicycle kinetics (i.e., one concentration per sensorgram) might also be adopted for any compound and is preferable for reversible binding compounds as a full dissociation profile is obtained at each concentration tested, which benefits mechanistic modeling. The multicycle SPR curves of PMBN binding to whole cells and OMVs show kinetic curvature that resembles binding kinetics but is instead caused by the development, and decay, of a mass transport-limited boundary, reporting the binding reaction at quasi- steady-state (Fig. 13A and 13B). The high-quality fit to the boundary layer model (Eqn (S3)) confirms that under these conditions, PMBN (and polymyxin B as described below) binds at an extremely high kinetic rate despite the presence of divalent cations. A finite element based-numerical model shows that such mass transport dominance also applies to single cells, which is relevant to in vivo milieus. Active 108465066.1.DOCX 26 00B206.1385 The binding mechanism that produces both reversible states and more stable polymyxin B bound states by comparing differences in affinity, binding capacity, and stoichiometry after full saturation with polymyxin B was next investigated. The affinity constant and binding capacity for PMBN binding were estimated from the corresponding fitting equation and indicated a two-fold drop in both affinity and binding capacity when LPS was pre-saturated with polymyxin B. This implies that only a fraction of the polymyxin B binding events can transition to the more stable bound state(s) (t1/2 of >6 hours). Polymyxin B binding to LPS exhibited exponential binding kinetics towards a defined saturation limit (Fig.13D) that is consistent with high occupancy of available LPS. Direct binding of polymyxin B to E. coli cells, OMVs, and LPS (Fig. 13E-13G) showed tight polymyxin B binding for all three targets. Indeed, the apparent association rate constant, which is driven by transport kinetics and binding affinity, when averaged for cells, OMVs, and LPS was in good agreement (ka 2.1 (± 0.78) x 105 M-1s-1). At moderate dissociation times (<1400 s), the dissociation is dominated by a moderate rate that was relatively consistent (<3-fold variation) when averaged over all three targets. Tightly bound polymyxin B was observed over LPS (Fig. 13G, high response curve) and again the saturation capacity for the tightly bound component was well below full LPS saturation allowing tightly bound polymyxin B and transiently bound polymyxin B to co-exist, which is observable as the additional reversible saw-tooth-shaped binding profile. Interestingly, repeated serial injections of polymyxin B after pre-saturating the LPS surface with polymyxin B isolated the reversible, transient binding component alone (Fig. 13G, low response curve), implying that saturation of the tightly bound polymyxin B component is limited by another process. Indeed, a PMBN-like binding profile was observed which also resembled resistant-OMVs and resistant E. coli cells exposed to polymyxin B (in the absence of pre-saturation), implying that affinity of polymyxin B for LPS can be estimated by fitting the boundary layer model. To approximate the overall off-rate, dissociation of polymyxin B from cells and OMVs was measured using a multipoint chaser method as described above52, where repeated PMBN injections report changes in polymyxin B occupancy allowing dissociation to be estimated without interference from baseline drift (Fig.13H). This analysis revealed a biphasic dissociation where a practically irreversible component was observed relative to the doubling time of bacterial growth. Apparent binding kinetics to cells and OMVs were nearly identical and binding/unbinding to purified LPS deviated by just 2-fold. Overall, the agreement between all three LPS containing surfaces suggests that OMPs and other Active 108465066.1.DOCX 27 00B206.1385 components of the cell membrane are not required for prolonged retention of polymyxin B. Dissociation of polymyxin B species is heterogeneous due to the presence of nucleates and clusters, which is apparent over short time-courses. However, clusters predominant in the outer membrane over long time-courses. To isolate the apparent retention time, it is important to measure the dissociation over a prolonged period of several hours (Fig.13H). Importantly, the apparent association and dissociation rate constants reported in Fig. 13E and 13F lump all bound states into a single state while those in Fig.13G and 13H assume two states. These models are mechanistically over simplistic and ignore the dominance of mass transport, yet they do allow qualitative comparison of observed kinetic curvature that imply a high probability of shared mechanism. g. Three-State Coupled Phase Separation Model While approximate methods were useful, a mathematical model was pursued to fully quantify the complex interactions between polymyxin B and OMVs. The kinetic differences apparent in polymyxin B-OMV SPR binding curves allowed for a set of assumptions that enabled formulations of a three-state kinetic model (Fig.3). This model, described below, accurately determines additional kinetic constants that could not be defined using approximation methods. This three-state mechanistic model follows mass conservation allowing the fraction of polymyxin B contained in each state to be estimated. It defines the kinetic evolution of the system without requiring a structural understanding of each microscopic state. Without being bound by theory, a microscopic interpretation of the model is presented in Fig. 3A and a more granular version in Fig. 3B. Both models describe lipid A-catalyzed accumulation of polymyxin B clusters, cPMB, in the outer membrane, through phase separation mediated by transition state intermediates. Although the reaction pathways in Fig. 3A and Fig. 3B are mechanistically meaningful, the illustrations depict phase separation events implied by the mechanism but not confirmed structurally. While SPR has been used extensively for evaluating affinity binding, it can also be applied to phase transitions because a change in phase state will be accompanied by a change in stability relative to the previous state which is observable in the kinetic curvature measured by SPR. The true mechanism is therefore already encoded in the curves and the objective of the iterative modeling process is to discover this mechanism. Focusing on the more refined three-state model presented in Fig.3B, an LPS affinity network consisting of an LPS (L) monolayer stabilized by divalent cations is exposed to polymyxin B (PMB). Advection and diffusion transport the injected polymyxin B (PMBi) to the SPR sensing Active 108465066.1.DOCX 28 00B206.1385 surface at mass transport rate k_t, where it will attain concentration [PMB] and react with L to produce a transient complex, PMBL, which is the first bound state. Formation of PMBL competitively displaces divalent metal ions, which lowers the stability of the LPS-metal ion affinity network. The affinity constant (K_D1) replaces the associated transient kinetic rate constants because binding is fully mass transport-limited, which impacts kinetic rate constants but not affinity constants. The weakened LPS-affinity network increases the availability of membrane insertion sites, n, associated with each LPS molecule. PMBL interacts lipophilically with n to produce a membrane-inserted species nPMBL which is the second bound state. The model assumes that all polymyxin B-bound states are described on a monomer basis other than the transition state intermediates that require a dimeric state to trigger phase separation to the third state, cPMB. The transition state begins with self- association of nPMBL complexes through interactions between each respective polymyxin B contained in the dimeric nPMBL. These interactions displace pre-existing interactions between each PMB and its paired L thereby forming transient membrane-inserted polymyxin B dimers (tPMB) and LPS dimers (LL). The transition state intermediate tPMB is fundamentally a form of nucleate and, therefore, it may be expected to share the same dissociation constant (k₄). This was the case when fitting the model as dissociation of the LL intermediate state gated release of cPMB and matched the dissociation of nPMB from the acyl LPS matrix to form PMBL. The rate constant (k8) for dissociation of cPMB from tPMB had no effect on the data and was therefore non-limiting and held constant at an arbitrary high non-limiting value (>1). tPMB does not accumulate significantly because it dissociates irreversibly into phase-separated cPMB. When fitting to polymyxin B binding data for wt-OMVs (Fig.14A), KD1 and k8 were held as constants while fitting rate constants k_t, k₃, k₄ and k₅ (red font) and found that the value of k₅ approached zero, effectively eliminating k5 and k8. The three-state model in Fig.3B was formulated as a set of coupled ordinary differential equations (ODEs), with eight binding rate constants (k_1-8) governing the rates of interchange of each species. Forward rate constants k7 and k9 were eliminated from the equation set as formation of transient species tPMB and LL are irreversible. Analysis constants Time at Injection 1 start = 0 s Sample contact time = 30 s Active 108465066.1.DOCX 29 00B206.1385 Dissociation time interval = 35 s Mass transport rate = kt = 46.51/s Coupled ODEs of Three-State Model d[L]/dt = -(k₁.[PMB].[L] - k₂.[PMBL])+2*k₁₀.[LL] (S8) d[PMB]/dt = -(k1.[PMB].[L] - k2.[PMBL]) + kt’.([PMBi](t) –[PMB]) (S9) d[PMBL]/dt = (k1.[PMB]*[L] - k2.[PMBL])-(k3.[PMBL].[n] - k4.[nPMBL] ) (S10) d[n]/dt = -(k₃.[PMBL].[n] - k₄.[nPMBL] ) (S11) d[nPMBL]/dt = (k3.[PMBL].[n] - k4.[nPMBL])-2.(k5.[nPMBL].[nPMBL] - k6.[tPMB].[LL]) (S12) d[tPMB]/dt = (k₅.[nPMBL].[nPMBL] - k₆.[tPMB].[LL] ) – k₈.[tPMB] (S13) d[cPMB]/dt = 2*k₈.[tPMB] (S14) d[LL]/dt = (k5.[nPMBL].[nPMBL] – k6.[tPMB].[LL]) – k10.[LL] (S15) The initial concentrations of each species were as follows [PMB]_t=0 = [PMBL]_t=0 = [nPMBL]_t=0 = [tPMB]_t=0 = [LL]_t=0 = [cPMB]_t=0 = 0 [L]t=0 = [n]t=0 = 1.56 x 10^-4 (units, M) kt’ (units, m/s) was already defined herein. [PMBi](t) is the injected concentration profile and [PMB]_t is the concentration profile at the sensing surface. [PMBi]_(t) follows a serial-doubling concentration of injected analyte, with a concentration profile defined by a serial step function, where each injection step represents a discrete concentration followed by a dissociation step without injected analyte and this is repeated for each concentration in the SCK dosing series. Kintek Explorer V9.5 was used to build and fit the three-state model. This program employs its own numerical integrator, which reports the change in concentration of each species over time. The SPR responses for concentrations of each accumulating PMB-species estimated from Eqn (8 - 15) are summed over time in Eqn (1), repeated here. Active 108465066.1.DOCX 30 00B206.1385 Response_(t) = ([PMBL]_t+[nPMBL]_t+2*[tPMB]_t+m.[cPMB]_t).MW.G. Eqn (1) The concentration of [tPMB]t is multiplied by two to account for its additional mass, being composed of two PMB molecules. Insertion of PMB into the acyl-LPS matrix to form cPMB is associated with changes in the phase state of LPS¹⁰ and is also likely associated with redistribution of mass towards the sensing surface. These changes will manifest as an increase in sensitivity of the SPR binding response when clusters are formed and are included in Eqn (1) using a sensitivity coefficient (m). MW is the molecular weight of the injected analyte, in this case PMB (1203.48 Da). The constant, G, is a unit conversion factor that converts protein concentration (g/L) to response (RU) and is typically 100 (units, RU.L/g). Biacore SPR systems are calibrated such that 1 RU is equivalent to a change of 1x10^-6 refractive index units (RIU), which is equivalent to a 2D concentration of 1 pg/mm² protein when the mass is distributed uniformly within a 100 nM hydrogel. The average height for OMVs bound to the surface is assumed to follow solution phase size measurements of approximately 100 nm and therefore G is assumed to apply as a reasonable approximation. In practice, the value was optimized within ± 6% when model fitting to allow for variations in OMV capture yield and spatial distribution effects. Assuming 1:1 binding stoichiometry, an OMV where LPS is fully saturated with PMB will contain approximately 0.77 ng/mm² PMB, which is equivalent to ~770 RU. Therefore, a PMB binding capacity of 30 RU on an OMV-coated surface, shown in Fig.4, represents just 4% of an equivalent full OMV monolayer, or approximately 156 μM LPS when expressed as a 3D concentration. Extreme depletion of PMB occurs when the surface reaction flux coefficient k1.[L] is high relative to kt’ preventing estimation of the transient rate constants (k1 and k2) associated with PMBL formation. Therefore, the rate equation for the diffusion boundary model (Eqn (S3)) was substituted into Eqn. (S10), effectively

with KD as shown in Eqn (S16) below. d(PMBL)/dt = (k_t’.[PMBi]_(t) - k_t’.K_D.([PMB]_t=0/([PMB]_t=0-[PMB])-1)^-1-(k₃.[PMBL].[n]- k4.[nPMBL]) Eqn (S16) When performing SPR curve fitting, it is common to express the mass transport coefficient in terms of SPR response, where k_t (units, RU/Ms) = k_t’.100.MW_PMB.R_max. The rate constant k₃ for PMBL insertion into an LPS-associated membrane insertion site (n) Active 108465066.1.DOCX 31 00B206.1385 equals the rate constant k₆ governing recovery of nPMBL from the transition state (tPMB). Both processes require membrane insertion and k3 is the rate at which membrane insertion sites become available. Similarly, the rate constant k4 governs dissociation of nPMBL from membrane insertion sites (n) and equals the rate constant k₁₀ (Eqn (S8)) for liberation of LPS during decay of the transition complex (shown as k4 (blue) in Fig.3B). This rate constant defines the stability of membrane insertion for both nPMBL and tPMB. The stability of the model fit was improved by setting k₆ and k₁₀

k₄, respectively. This reduced the number of binding rate constants to be fit to just four. The numbering of rate constants remains as in Eqn (S8 - S15) except for the repeating rate constants mentioned above, which retain the numbering of the first instance. h. Kinetic Characterization of Polymyxin B Binding to OMVs The three-state coupled phase-separation model was fit to binding curve titrations (Fig. 14A and 14D) in order to determine the interaction parameters (Table 2). The goodness-of-fit for each titration was shown by the low ^² (root-mean-square average of the residual difference between the fitted curve and the experimental curve) values and the low standard deviation (SE) a narrow profile intervals indicated that reliable values were returned for each parameter fit. Table 2. Binding parameters for polymyxin B returned from fitting the three-state model (Fig.5A and 5B) with global constraint of all interaction constants. v_alue ¹ _{±SE of fit} ^{98% CI (profile} l_ikelihood)2 wt-OMV³ KD = k₂/k₁ (nM) 126* k3 (M^-1 s^-1) 2233 98 1950-2450 k₄ (s^-1) 0.00488 0.00007 0.00444-0.00524 k₅ (M^-1 s^-1) 29.76 0.02 20-22 kt (ms^-1) 45.3 0.1 44.3-47.1 ^² _(RU ² ₎ ^0.42000 resistant-OMV³ KD = k₂/k₁ (nM) 268 3 243-297 k3 (M^-1 s^-1) 335 1 290-404 k₄ (s^-1) 0.0085 0.0002 0.0078-0.0099 k_t (ms^-1) 44.6 0.4 40.4-48.8 ^² (RU²) 0.47 Active 108465066.1.DOCX 32 00B206.1385 ¹ Values generated calculated from the three-state model with K_D in wt-OMVs (*) fixed to experimentally determined value for wt-OMV. For binding curves with modified LPS (resistant-OMV) the model defaulted to the first two-states. ² Confidence limits were calculated using the brute force method where confidence contours were generated by fitting all parameters to the actual data set while holding a given parameter constant and repeating with values to either side of the optimal fitted value of that parameter. ³ OMVs isolated from wild-type or pmrA^G53E polymyxin-resistant E. coli strains. C. Discussion a. Three-State Model for Polymyxin Binding to the Outer Membrane The importance of LPS binding for the antibacterial activity of polymyxins is well- established, but there is no clear consensus on how binding translates to the permeabilization and bactericidal activities. SPR analysis performed using surface-bound OMVs, a near- native outer membrane mimetic, resolved polymyxin B binding kinetics from seconds to hours and enabled derivation of a three-state mechanistic model. Overall, the interaction of polymyxins with the outer membrane entails coupling of weak, transient LPS binding with phase separation and the formation of tightly bound polymyxin aggregates (Figure 3). In the first step, polymyxin B binds to LPS, approximated as a simple 1:1 complex PMB-L, with a relatively weak apparent KD of approximately 1.3 µM. Formation of PMB-L leads to outer membrane barrier disruption through displacement of divalent metal ions. In the next step, polymyxin liberated through dissociation of PMB-L either exits from the outer membrane or undergoes lipid A-mediated membrane insertion to form nucleates, nPMB-L. The fitted model shows that nucleates remain stoichiometrically associated with LPS such that LPS can co-exist in three populations of various fractions: free L, PMB-L, and nPMB- L. In the third step, nPMB-L coalesce into long-lived, lipid A-free polymyxin B clusters, cPMB. The lifetime of these cPMB clusters, t1/2 >6 hours, exceeds the doubling time of bacteria making them essentially irreversible on the timescale of growth. The binding constants for these steps were well-resolved because of the wide variation in residence time between the three-bound states, which represent transiently, moderately, and highly stable species. The model is consistent with “self-promoted uptake” wherein polymyxin B induces the lipid A layer to act as a catalyst that promotes accumulation of super-stoichiometry concentrations of polymyxin B into clusters. Importantly, weak polymyxin B binding to Active 108465066.1.DOCX 33 00B206.1385 lipid A in the first step is necessary to allow subsequent phase separation as tight binding in this step would trap polymyxin B into a 1:1 stoichiometric complex that would be rendered incapable of forming clusters. This three-state model quantitatively defines the changes in mass of each species in real-time and reveals the total mass of polymyxin bound. Importantly, weak polymyxin B binding to LPS in the first step is necessary to allow the subsequent phase separation as tight binding would trap polymyxin B into a 1:1 stoichiometric complex that would be rendered incapable of forming cPMB clusters. When fully saturated, polymyxin B outnumbered LPS by an approximately 2:1 molar ratio (Fig. 1E, where the reversible fraction represents the total LPS), suggesting that outer membrane stretching might be expected. The excess of cPMB clusters likely cause the vesiculation and tubules observed on cells and OMVs upon polymyxin exposure (Fig. 2B). Thus, the three-state model suggests that polymyxin B clustering is catalyzed from transient interactions between polymyxin B and LPS and formation of clusters correlates with bactericidal activity was observed. The instant model also provides insight into the differences in binding that occur when polymyxin B or lipid A are modified. PMBN lacks a hydrophobic tail and does not kill bacterial cells. This polymyxin B variant exclusively produces rapidly reversible binding. Strikingly, PMBN binding to OMVs is identical to polymyxin B binding when polymyxin B clusters are pre-saturated and this recapitulated with pure LPS, indicating that while PMBN can form PMBL, the lack of an acyl tail prevents further transitions. The absence of a lipophilic anchor reduces the amphiphilic properties of polymyxin B, thereby limiting lipophilic membrane interactions and promoting greater solubility. However, because PMBN can still permeabilize the outer membrane (Table 1), formation of PMBL must disrupt the LPS affinity network, likely through competition for metal ion binding sites. Modifications to the phosphate groups of lipid A impart polymyxin B resistance, and these modifications were observed to eliminate cluster formation. The absence of cPMB clusters observed with resistant-OMVs may be related to an electrostatically enhanced stability of the affinity network that can more effectively resist the outer membrane stretching that is likely needed for accumulation of clusters. Presumably, formation of small nucleates, nPMBL, which were observed for resistant-OMVs, does not require a high degree of energetically costly membrane stretching. Without being bound by theory, cluster formation can promote cell killing by enabling a transmembrane flux of polymyxin B at the stretched phase boundary of cluster sites. Active 108465066.1.DOCX 34 00B206.1385 The generalizability of the instant model was explored by monitoring binding of brevicidine, a distinct lipid A-binding antibacterial natural product. Brevicidine exhibits a binding pattern similar to that observed for polymyxin B with SPR traces showing accumulation of a tightly bound species and a superimposed saw-tooth profile associated with transient binding, suggesting polymyxin B and brevicidine might utilize a common approach for interacting with lipid A in the outer membrane. The complex formed by brevicidine (t_1/2 >1 hour) was less stable than those formed by polymyxin B (t_1/2 >6 hours) and this could account for the less potent antibacterial activity of brevicidine (Table 1). Also distinct from polymyxin B, brevicidine displayed an identical binding pattern with wt- OMVs and resistant-OMVs, consistent with the antibacterial activity of brevicidine against polymyxin-resistant mutants (Table 1). Thus, though the molecular interactions with lipid A likely differ, both the polymyxins and brevicidine form stable, long-lived interaction with the outer membrane, suggesting this could be a necessary event in the mechanism of lipid A-binding antibiotics. However, while the three-state model suggests polymyxin B clustering is a necessary step in cell killing, these findings do not exclude other possibilities for lipid A-targeting bactericidal antibiotics. The three-state model can aid in the identification and design of novel, selective lipid A-targeting antibacterial molecules able to overcome polymyxin resistance and of potentiators able to induce outer membrane permeability. Importantly, how polymyxin B interacts with lipid A in the context of the phospholipid inner membrane, a step proposed to be necessary for cell killing, and with mammalian kidney cells to understand nephrotoxicity associated with polymyxin clinical use, remain to be determined. Overall, the three-state model provides a quantitative understanding of kinetic processes driven by the multifaceted physicochemical properties of polymyxin B that essentially transform the LPS barrier, which has evolved to preserve cell integrity, into a catalyst that promotes accumulation of cytotoxic clusters. b. Microscopic Mechanism of Polymyxin B Binding to the Outer Membrane Although not required to interpret a three-state kinetic model, it is possible to suggest a plausible structural mechanism that might provide further insight into polymyxin activity. In addition to the instant kinetic data, a number of key observations are important for a structural model. First, soluble LPS is highly polymorphic and the oligosaccharide chain length has a significant impact on the packing density of LPS. Second, the area per LPS molecule can range from 8-16 nm depending on applied pressure. Third, LPS has been Active 108465066.1.DOCX 35 00B206.1385 observed in nanocrystalline formations by EM. In the outer membrane, tubules and vesiculations have been observed in the absence of Mg2+ while hexagonal LPS clusters were seen in the presence of Mg2+. Each hexagonal cluster in the lattice is composed of an LPS molecule on each side with an LPS-free center. Strikingly, the lattice constant of 14 nm matches that of the hexagonal LPS lattice in the presence of polymyxin recently observed by AFM. Finally, the LPS network becomes destabilized due to polymyxin displacement of divalent cations, while a Mg2+-dependent LPS transition to a crystalline lattice has now been shown to be highly destabilizing in vitro. Taking these into account, it is proposed that at the onset of exposure, polymyxin B will first displace metal ions bound to the core saccharide while leaving lipid A-bound Mg2+. Polymyxin B interactions with the core saccharide should thus lower charge repulsion, increase the excluded volume, and thereby lower solvation, which destabilizes the LPS barrier function to increase permeability. Surface energy minimization triggers a rapid phase change that rearranges LPS from a linearly interlinked network into clusters that pack densely into a hexagonal lattice. Once formed, the lipid A region of the LPS crystal lattice is less permeable to polymyxin B, preventing dissolution of the lattice through competition with the remaining Mg2+. The LPS phase transition causes the outer membrane to expand laterally and thin, facilitating the accumulation of polymyxin B monomers unbinding from LPS. Less dense LPS cluster centers can act as sinks to accumulate high local concentrations of polymyxin B that enable polymyxin B aggregates to form via phase separation. Without being bound by theory, once formed these large polymyxin B clusters are not expected to escape being trapped due to electrostatics, lipophilic interactions and size exclusion imposed by the tightly packed LPS core and oligosaccharides. The finite volume available for cluster accumulation accounts for the outer membrane structural deformities observed on cells and OMVs treated with polymyxin B (Fig.1B and 5A-5C). It is proposed that cluster formation promotes cell killing by enabling a transmembrane flux of polymyxin B aggregates at the stretched LPS layer, allowing access to the periplasm and inner membrane. Loss of the outer membrane barrier alone does not result in cell death, however the catastrophic damage caused by excess polymyxin accumulation in the outer membrane can contribute to the bactericidal activity of these antibiotics. c. Role of hexagonal Lattice Structures Recently, AFM enabled visualization of hexagonal LPS lattice structures in the outer membrane upon polymyxin treatment. These experiments were performed with OMVs Active 108465066.1.DOCX 36 00B206.1385 assembled on mica, which unlike the native state provides solid phase support. Nevertheless, the study showed that removal of all divalent cations, using EDTA, prevented polymyxin- induced lattice formation and excess Mg2+ actually competed with polymyxin in triggering the LPS phase transition. Interestingly, both polymyxin B and PMBN also induced hexagonal lattice formation, at odds with the conclusions that formation of these structures is a pre-requisite to the bactericidal activity of LPS-binding antibacterial molecules as PMBN does not kill cells. The data presented herein, on the other hand, demonstrate that formation of long-lived clusters does correlate with cell killing (Fig.2 and 4, Table 1 and 2), and the instant SPR-based mechanistic model indicates that it is phase separation of polymyxins, cPMB formation, that is a pre-requisite step for cell killing. The instant model assumes that the LPS phase change occurs instantaneously, and it can have different morphologies with respect to the specific antibiotic, or environmental conditions, being tested and therefore will influence the kinetic rate constants obtained for the polymyxin- containing species. Therefore, this coupling is absorbed into the rate constants (k3-k6) avoiding the need to explicitly add an independent LPS conformation change step in the instant model, though adding such a step may be of value in studying this coupling effect more quantitatively. d. Mechanism of Polymyxin Resistance Modifications to the phosphate groups of lipid A impart polymyxin B resistance, and it was observed that these modifications eliminate cPMB formation (Fig.2B, 2D, and 2F). Consistent with this, hexagonal lattices were not observed by AFM with modified, polymyxin B-resistant, lipid A. SPR did reveal the formation of the moderately stable nucleates, nPMB-L, by polymyxin B with resistant-OMVs, indicating that nucleation requires lipophilic interactions between the fatty acid tail of polymyxin B and the outer membrane but this is insufficient for cluster formation. An LPS phase change can enable a sufficient expandable volume for nucleates to aggregate into large clusters that occupy a considerable volume. It is proposed that the absence of cPMB clustering observed with resistant-OMVs is related to an electrostatically enhanced stability of the affinity network that can more effectively resist the outer membrane stretching required for accumulation cPMB (Fig. 2B). Specifically, addition of pEtN or L-ara4N to one phosphate of lipid A introduces an electropositive amine that can electrostatically bridge with a neighboring lipid A via its unmodified phosphate. Structural NMR studies demonstrate that polymyxin B and PMBN bind to the lipid A phosphate groups, yet here it was observed in this work that polymyxin B was able to Active 108465066.1.DOCX 37 00B206.1385 bind resistant-OMVs and form nucleates despite the absence of divalent metal ion stabilized lipid A (Fig 2B). This indicates that polymyxin B is capable of interacting with other available sites, such as phosphorylated sugars in the core oligosaccharide, and it is these interactions that are sufficient to induce outer membrane permeability. This also indicates that divalent cation stabilization is not fully inhibited by polymyxin B. Thus, polymyxin induced outer membrane permeability through core oligosaccharide binding explains why the reversible binding events for polymyxin B and PMBN with wt-OMVs and resistant OMVs are indistinguishable (Fig. 2 and 4), rationalizes the ability of polymyxins to permeabilize the outer membrane of lipid A-modified polymyxin-resistant strains, and explain the enhanced outer membrane permeability of deep rough LPS mutants. e. Displacement of Divalent Cations by Polymyxin B in the Outer Membrane Polymyxin B and Mg2+ are competitive in forming transient interactions with LPS yet concurrent fractional occupancy is assured since these interactions re-equilibrate in real- time due to the short half-life of these complexes where relative occupancy of each will be driven by the affinity constants and the local concentrations within the LPS film. Polymyxin B and PMBN bind to the phosphate groups of the lipid A, however, conditions within a native LPS film favor a heterogeneous ensemble of bound states that changes over time, involving phosphates in the core saccharide. Polymyxin B induced Mg2+ displacement from lipid A can be hindered by increased avidity for Mg2+ ions due to lipid A bridging and diminished access due to tight packing around the lipid A region. f. Generalizability of the Three-Step Model for LPS-Binding Antibiotics Polymyxin B and colistin are both polymyxins, thus it was not unexpected that they would utilize a common mechanism (Fig.2A and 6E) However, brevicidine, a distinct LPS- binding antibacterial natural product, also exhibits a similar binding pattern. Accumulation of a tightly bound species and a superimposed saw-tooth profile associated with transient binding were observed (Fig.2I), suggesting polymyxin B and brevicidine share a common mechanism of action and indicate the generalizability of this three-state model. The apparent long-lived clusters formed by brevicidine (t1/2 >1 hour) were less stable than those formed by polymyxin B (t_1/2 >6 hours) (Table 4) and this may account for its less potent antibacterial activity (Table 1). Also distinct from polymyxin B, brevicidine displayed an identical binding pattern with wt-OMVs and resistant-OMVs (Fig. 2I), consistent with the antibacterial activity of brevicidine against polymyxin-resistant mutants (Table 1). Active 108465066.1.DOCX 38 00B206.1385 PMBN exclusively produces rapidly reversible binding, but, strikingly PMBN binding to OMVs was almost identical to polymyxin B binding when the loading capacity of cPMB was fully pre-saturated (Fig.9). This observation indicates that PMBN binds LPS in a similar manner to polymyxin B, but its acyl tail allows polymyxin B to aggregate, a requirement for cluster formation. However, because PMBN can still permeabilize the outer membrane (Table 1), formation of PMB-L must disrupt the LPS affinity network. Additionally, PMBN-induction of the LPS hexagonal crystal lattice observed by AFM cannot in itself cause cell killing. Thus, the instant model enables mechanistic dissections with novel LPS-binding antibacterials as well as variants of existing molecules. D. Conclusion The proposed three-state model for polymyxin activity provides an important framework for understanding the mechanism of these last-resort antibiotics. The model supports “self-promoted uptake” wherein polymyxin B induces the LPS layer to act as a catalyst to promote accumulation of super-stoichiometric concentrations of polymyxin B. Importantly, transient binding of polymyxin B to LPS is a necessary first step that enables subsequent phase separation as tight binding would saturate available LPS and prevent catalytic accumulation. Importantly, how polymyxin B interacts with LPS in the context of the phospholipid inner membrane, a step proposed to be necessary for cell killing, remains to be determined. Moreover, while cluster-driven antibacterial activity is supported by these findings, the possibility that LPS-targeting antibiotics that do not require clustering can exist or can be developed cannot be excluded. In conclusion, the instant data has shown that biophysical methods and associated kinetic modeling enabled mechanistic characterization of LPS-binding antibiotics and can be applied to the development of LPS-binding inhibitors and outer membrane permeabilizers able to overcome multidrug resistance, e.g., the methods described herein have been applied to the natural antibiotics Brevicidine and Ogiopeptin to identify accumulation as illustrated in Figure 12. E. Methods a. Minimal Inhibitory Concentration and Antibiotic Potentiation Assays Bacterial strains and plasmids are listed in Table 6. LB growth media was prepared according to manufacturer’s instructions and supplemented with 0.2% arabinose and carbenicillin (50 μg/mL) for the strains containing pBAD24-mcr1 plasmid. Cultures were started by inoculating 3 ml LB with 1-2 colonies from fresh overnight plates and grown at 37°C until in log phase. Modified minimal inhibitory concentration (MIC) assays were Active 108465066.1.DOCX 39 00B206.1385 performed in 96-well round bottom polystyrene plates (Corning) with a final volume of 100 µl in LB supplemented with tween-80 at 0.0005%. For outer membrane permeability assays, rifampicin was added at 1.56 µM (1/4x the MIC). Polymyxin B sulfate (TCI Co. Ltd), polymyxin B nonapeptide (PMBN) (Sigma), and brevicidine (Genentech) were diluted directly in the assay plates. Bacteria were diluted in LB with tween-800.0005% and added to a final OD600 of 0.0005. Growth was measured via OD600 on a SpectraMax M5 plate reader after overnight static growth at 37°C with humidity. Table 6. Strains and plasmid used in this study Strain Resistance gene Source GNEID E. coli BW25113 n/a Baba et al.¹ 115 E. coli BW25113 pmrA^G53E n/a In-house 5164 E. coli BW25113ΔtolQ Kan-R Baba et al.¹ 6119 E. coli BW25113ΔtolQ pmrA^G53E Kan-R In-house 6120 S. aureus USA300 n/a ATCC (BAA- 23 1556) Plasmid Description Resistance gene Source pBAD24-mcr1 Constitutive expression of mcr1 Carb-R in house ¹ Baba et al.2006. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol Syst Biol 2:473–11. b. Outer Membrane Vesicle Preparation OMVs were isolated as follows.1 L cultures of bacteria were grown in LB overnight (approximately 16 h) at 37°C with aeration. Cells were pelleted by centrifugation at 15,000 rpm for 30 min at 4°C. Supernatants were filtered through a 0.45 µM PVDF filter (VWR) and concentrated via tangential flow filtration to a volume of approximately 50 mL. OMVs were pelleted by ultracentrifugation at 40,000 rpm for 2 h at 4°C. OMV pellets were washed in 50 ml of OMV buffer (phosphate buffered saline [PBS, Fisher Scientific] plus 200 mM NaCl, 1 mM CaCl2, and 0.5 mM MgCl2) by ultracentrifugation as described above. The washed OMV pellet was resuspended in 1.5 ml of OMV buffer and passed through a 0.45 µM PVDF syringe filter. OMV preparations were quantified using a standard Bradford protein assay. Aliquots were stored at 4°C or at -80°C To compare the composition of the OMVs to outer membranes, E. coli ΔtolQ, E. coli ΔtolQ pmrA^G53E, and E. coli ΔtolQ with pBAD24-mcr1 were grown as described for OMV isolation and cell pellets frozen at -20. The cell pellet was brought up in ice-cold Active 108465066.1.DOCX 40 00B206.1385 25 mM HEPES buffer, pH 7.4, with 1x protease inhibitor cocktail (cOmplete mini, Roche), lysed using a LVI Microfluidizer homogenizer (Microfluidics), and centrifuged for 10 min at 4000 xg in a tabletop centrifuge. Supernatants were centrifuged at 250,000 xg for 1h at 4°C (Beckman TLA 120.2). Pellets were washed in HEPES buffer plus protease inhibitors. To solubilize the inner membrane, pellets were suspended in 25 mM HEPES buffer, pH 7.4, with 2% sodium lauroyl-sarcosinate (Sigma), incubated with rotation at room temperature for 30 min, and centrifuged as before. The outer membrane protein containing pellet was suspended in OMV buffer and quantified using Bradford protein assay. 0.5 µg of each sample (prepared with BOLT LDS sample buffer and reducing agent (Invitrogen)) was separated on a 4-12% NuPAGE gel in 1x MOPS buffer (Invitrogen) and stained for 1h with InstantBlue Protein Stain (Novus Biologicals). c. C1 Chip Preparation and OMV Capture Polymyxin B was covalently attached to a C1 chip (Series S Sensor Chip C1, Cytiva) using an amine coupling kit (Cytiva). Briefly, equal amounts of reagents N- hydroxysuccinimide and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride were mixed as per manufacturer’s instructions and immediately added to the gold chip surface for 2 min, washed with distilled deionized water, and dried. Sufficient polymyxin B (1 mM in 1 M HEPES buffer, pH 8) was added to cover the chip surface, incubated at room temperature for 1 h, washed as before, and then incubated with 1 M ethanolamine hydrochloride-NaOH, pH 8.5, for 1 min. The chip surface was washed with water, dried as previously described, immediately loaded into a Biacore S200, and primed with running buffer (described below) to equilibrate the system. C1 chips were used for multiple runs and discarded upon removal from the Biacore S200. All SCK C1-chip SPR experiments were performed in running buffer (Dulbecco's phosphate-buffered salt solution 1x without calcium or magnesium (Fisher Scientific), pH 7.4, with 0.0005% tween-80 (Sigma) passed through a 0.2 µm filter). MgCl2 was added as indicated in the figure legends Inclusion of tween-80 prevented loss of polymyxin to the plastics. For experiments with brevicidine, which was suspended in DMSO, 0.125% DMSO was included in running buffer. Analysis and compartment temperature was set to 25°C or 37°C as indicated in figure legends. OMVs were diluted from frozen stocks to approximately 20-30 µg/ml protein in OMV buffer. Capture of OMVs was performed at low flow rate (5 µl/ml) for 300 sec over the test channel(s) followed by a 300 sec stabilization period. All subsequent steps were at a flow rate of 40 µl/ml. For single cycle Active 108465066.1.DOCX 41 00B206.1385 kinetics, two-fold dilutions were injected over the channel(s) loaded with OMVs and a reference channel without OMVs for 30 sec contact and 480 sec dissociation. To regenerate chips, 0.5% SDS (desorb 1, Cytiva) was injected into all channels for 60 sec at 30 µl/ml twice, with an extra buffer wash and four carry-over control steps to prevent residual SDS from disrupting the following cycle. All single-cycle kinetic SPR traces shown were double-referenced (unless indicated) within the Biacore S200 Evaluation Software 1.0, exported as a .txt file, and imported (decimated) into GraphPad Prism (version 9.3.1 for Mac, GraphPad Software, San Diego, CA). For clarity, in some cases individual outlier data points (at buffer transitions) were removed when this would not change the overall data or conclusions. d. Whole cell SPR Bacterial strains were grown in LB or LB with 0.2% arabinose and carbenicillin (50 μg/mL) to log phase (OD600 approximately 0.4-0.6), centrifuged for 10 min at 3500 rpm, and re-suspended in PBS to a final OD600 of approximately 5. SPR on whole bacterial cells was performed and analyzed as described for OMVs on C1 chip, but with an additional carry-over wash prior to the capture. Regeneration of the chip after whole cell binding also required additional steps (40 µl/ml flow rate): 1) PBS supplemented with 32 mM MgCl2 for 120 sec, 2) 2.5 M NaCl for 30 sec, 3) 0.5% SDS (Desorb 1, Cytiva) for 60 s with carry over controls between. Capturing sufficient RUs of whole bacterial cells required additional injection optimization and exhibited high experiment-to-experiment variability. e. LP Chip LP chips (2D carboxymethyldextran surface, partially alkyl derivatized, Xantex Bioanalytics) were cleaned with two-20 sec pulses of 40 mM CHAPS (3-((3- cholamidopropyl) dimethylammonio)-1-propanesulfonate) with 10 sec dissociation at 30 µl/ml flowrate as recommended by the manufacturer. All LP-chip experiments were performed in 0.2 µm filtered 1x Dulbecco's PBS without CaCl2 or MgCl2 (Fisher Scientific), pH 7.4. Tween-80 was not compatible with this system, increasing the loss of polymyxin B due to non-specific binding. Analysis and compartment temperature were set to 25°C. Mammalian vesicles from Expi293 cells were isolated and diluted to 0.1 mM in PBS. OMVs diluted in OMV buffer (described above) or mammalian vesicles were captured onto the chip for 60 sec (5 µl/ml flow rate) followed by a 300 sec stabilization period. Single cycle kinetics were performed as described above for the C1 chip. To regenerate the chip, 40 mM CHAPS was injected to all channels for 180 sec (40 µl/ml), washed, 50 mM NaOH for 60 sec (40 µl/ml), buffer washed again, and finally four carry-over control steps. Active 108465066.1.DOCX 42 00B206.1385 Regeneration the LP chip was incomplete, leading to poor regeneration and build-up of material which necessitated disposal of chip. Sensorgrams could not be normalized for bound ligand (i.e., the OMVs on the chip surface) using a calculated ‘Rmax’, and, as such, analyzed traces, including those in the figures, were selected to have similar levels of loaded OMVs. All single-cycle kinetic SPR traces shown were double-referenced (unless indicated) within the Biacore S200 Evaluation Software 1.0, exported as a .txt file, and imported (decimated) into GraphPad Prism (version 9.3.1 for Mac, GraphPad Software, San Diego, CA). For clarity, in some cases individual outlier data points (at buffer transitions) were removed when this would not change the overall data or conclusions. f. Binding and Kinetic Analysis of SPR-OMV Assays To determine the apparent KD of the reversible binding event of polymyxin B, single-cycle kinetics were performed as described above. After OMVs were loaded, 5 µM polymyxin B was injected for 360 sec (40 µl/ml flowrate) followed by 120 sec dissociations to saturate the stable-binding population prior to the sample injections. The base-to-peak value of each trace was determined from double-referenced traces (reference 1: - OMVs/ + compound channel; reference 2: + OMVs/ - compound) exported from the Biacore S200 evaluation software into PRISM 9 software. The change in RU with each pulse was plotted over the concentration and the KD determined by fitting a non-linear regression, one-site total function with background parameter set to 0 (GraphPad Prism version 9.3.1 for Mac, GraphPad Software, San Diego, California USA, www.graphpad.com). To determine the apparent KD of nonapeptide and polymyxin B on resistant-OMVs the same approach was taken as described above but without any pre-saturation step prior to kinetic pulses. To determine the residence time of the test compound with OMVs, the ‘chaser method’ was used because it accounts for drift that can occur in the system over long incubation times. OMVs were loaded onto the chip as described above then 5 µM polymyxin B was injected using the ‘low-sample consumption’ setting for 300 sec (5 µl/ml flowrate), followed by a 240 sec dissociation time. A second dose of 5 µM polymyxin B was then injected for 60 sec (30 µl/ml) to assure saturation followed by a 2 h dissociation time prior to the ‘chaser’, a 60 sec pulse of 5 µM polymyxin B. The RUs of associated polymyxin B were determined 120 sec after the start of the dissociation. The change in RUs from before dosing to after the second dose, and the change in RUs from prior to the ‘chaser’ dose and 120 sec after the start of its dissociation were used to calculate the fraction occupancy of the material and the residence time/half-life as described. Active 108465066.1.DOCX 43 00B206.1385 g. Mechanistic Studies of Polymyxins Binding to Cells, OMVs and LPS Bilayers OMV-coated and whole cell-coated C1 chips were prepared as already described with the following changes and LPS-coated surfaces were prepared identically but with rough-LPS extracted from E. coli F583 (Rd mutant, Sigma #L6893). For LPS, cloudy colloid suspension (above its CMC, ~1 mg/ml) contains LPS micelles that were injected and captured. For all formats, sample buffer and running buffer were composed of Dulbecco's phosphate-buffered salt solution containing 0.0005% tween-80 (Sigma), 1 mM calcium and 0.5 mM magnesium (Fisher Scientific), pH 7.4. Analysis and compartment temperatures were both set to 37°C. Freshly prepared OMV-coated or whole cell-coated sensing surfaces were employed for each injection of polymyxin B while LPS-coated sensing surfaces could be fully regenerated by injecting 50 mM CHAPS. Reversibly bound PMBN could be reinjected over all surfaces without requiring regeneration as it fully dissociated within a few minutes. Serial doubling dilutions of PMBN were injected from 2.5 mM to 0.156 mM over whole cells and OMVs at 50 mL/min for 1 min (Fig.13A and 13B). This was repeated over LPS-coated sensing surfaces and in this case, serial doubling dilutions of PMBN were injected from 5 mM to 0.625 mM for 30 seconds at 50 mL/min, immediately prior to injection of 1 mM polymyxin B for 200s at 50 mL/min (Fig. 13D). The PMBN injection series was immediately repeated (Fig. 13C) after a single concentration of polymyxin B (325 nM) was injected, which saturated each sensing surface (Fig.13D). Fresh sensing surfaces coated with whole cells and OMVs were prepared and 325 nM polymyxin B was injected at 50 mL/min for 120s over both. This injection was repeated after a long delay, and without regeneration, in order to allow free sites to become available thereby resulting in two polymyxin B binding curves for each sensing surface (Fig.13E and 13F). A fresh LPS-coated surface was prepared and polymyxin B was injected using SCK injection mode over a serial doubling dilution range from 625 nM to 39 nM. This resulted in surface saturation and the same injection series was immediately repeated resulting in the two curves shown in (Fig. 13G), where the second injection series fails to generate prolonged retention. PMBN injections performed immediately after saturation of a sensing surface with polymyxin B reports loss in polymyxin B occupancy because they are proportional to the PMBN binding response and therefore report the increasing fraction of free LPS sites that become available through polymyxin B dissociation. Here 10 mM PMBN was injected for Active 108465066.1.DOCX 44 00B206.1385 30s at 50 mL/min at regular time intervals of 850s, over sensing surfaces coated with OMVs or whole cells and the steady-state response regions were then plotted as a function of time and fit to a dissociation model. All coated sensing surfaces were paired to uncoated sensing surfaces providing a reference response for data analysis. Evaluation without referencing showed the C1 sensor chip surface had minimal to no non-specific binding to polymyxin B or PMBN. i. Topography SEM of OMVs on SPR Chips OMVs were loaded onto all channels of a C1 in a Biacore S200 as described above. The chip was then removed from the machine and fixed and stored in modified Karnovski's fixative (2.5% paraformaldehyde and 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.2) for at least 24 hours at 4°C. SPR chips were then washed with ultrapure water stained 4% (w/v) uranyl acetate for 15 min at room temperature. The chips were then washed again, dehydrated in a series of ascending ethanol concentrations, and finally incubated twice for 10 min at room temperature with hexamethyldisilazane (HMDS) as the final dehydration steps. The HMDS was then allowed to slowly evaporate. Air dried chips were then coated with a 2 nm thin layer of gold-palladium using a sputter coating device and examined in a Zeiss Gemini 300 Scanning electron microscope. h. Negative Staining and TEM Imaging OMVs were incubated in OMV buffer (as above) plus 0.0005% tween-80 and then polymyxin B, PMBN, or an equal volume of OMV buffer, added for a final ratio of 0.5:1 polymyxin B (or PMBN) to LPS for either 1 min or 40 min. The amount of LPS in the OMV preparations was determined using fluorescently labeled LPS. The suspensions were then adsorbed to the surface of formvar and carbon coated TEM grids (100 Mesh) for 60 sec, quickly rinsed in ultrapure water and stained twice for 60 sec with 2% aqueous uranyl acetate. Excess staining solution was blotted off and grids were air dried. Imaging was with JEOL JEM-1400 transmission electron microscope (TEM) operated at 80kV using magnification form 5000x to 100.000x. i. Lipid A Extraction and Mass Spectrometry Analysis Lipid A was extracted from OMVs by using the following method. To begin, 50 µl (~50-75 µg) of the OMV preparation was suspended in 5 ml PBS and Lipid A extracted. Dried samples were brought up in 1 ml of 0.25% n-Dodecyl-B-D-Maltoside (DDM) detergent in water by heating to 42°C and bath-sonicating. Lipid A analysis was performed on an LC-ESI-MS/MS instrument with a Thermo qExactive orbitrap mass spectrometer with Active 108465066.1.DOCX 45 00B206.1385 an electrospray ionization source in negative mode. The LC was a Thermo Ultimate 3000 and LC eluent was split between the qExactive MS and a Thermo charged aerosol detector (CAD). The LC separation was performed at 40 °C on a Phenomenex Luna 5 μm C8100 Å, 50 x 2 mm column. A 15 minute gradient utilized solvent A: 10 mM Ammonium Acetate in H2O and solvent B: isopropyl alcohol:acetone:ethanol 2:1:1. A linear gradient with the following proportions (v/v) of solvent B was applied: 0 – 5 minutes at 1%, 5 – 15 minutes at 99%, 15 – 20 minutes at 99%, 20 – 20.25 minutes at 1%, and 20.25 – 25 minutes at 1%. Ratios of unmodified, single-modified and double-modified lipid A were evaluated in MS- mode extracted ion chromatograms. The extracted ion chromatogram peak areas were integrated and reported. The intensity of all integrated EIC peaks were within the linear range of the mass spectrometer. Lipid A analogues of methylene extensions were detected of the singly, doubly and unmodified lipid A. Active 108465066.1.DOCX 46

Claims

00B206.1385 What is claimed is: 1. A method of identifying a composition capable of inducing the death of Gram-negative bacterial cells, comprising: a. contacting the outer membrane of a Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, with the composition to form a complex; b. contacting the complex with a wash; c. determining that the composition is retained in the outer membrane after contact with the wash; d. repeating steps a-d; and e. determining that the composition accumulates in the outer membrane and is retained in the outer membrane for at least about 10 minutes; wherein said accumulation of the composition in the outer membrane and its retention in the outer membrane for at least 10 minutes identifies the composition as capable of inducing the death of Gram-negative bacterial cells. 2. The method of claim 1, wherein the outer membrane is an outer membrane of an outer membrane vesicle derived from a Gram-negative bacterial cell. 3. The method of claims 1 or 2, wherein the composition accumulates in the outer membrane to a mass fraction greater than 10% (weight of accumulated composition / weight of outer membrane). 4. The method of any of claims 1-3, wherein the determination that the composition complexes with a lipopolysaccharide integral with the outer membrane is performed by a surface plasmon resonance (SPR) analysis. 5. The method of any of claims 1-3, wherein the determining that the composition is retained in the outer membrane after contact with the wash is performed by SPR analysis. 6. The method of any of claims 1-3, wherein the determining that the composition accumulates in the outer membrane and is retained in the outer membrane for at least about 10 minutes is performed by SPR analysis. Active 108465066.1.DOCX 47 00B206.1385 7. The method of any of claims 4-6, wherein the Gram-negative bacterial cell or outer membrane vesicle derived therefrom is immobilized on an SPR sensor substrate. 8. The method of claim 7, wherein the Gram-negative bacterial cell or outer membrane vesicle derived therefrom is immobilized on an SPR sensor substrate via an affinity coupling. 9. The method of claim 8, wherein the affinity coupling comprises amine-coupled polymyxin B immobilized on the SPR sensor substrate. 10. The method of any of claims 1-9, wherein the Gram-negative bacterial cell, or an outer membrane vesicle derived therefrom, is polymyxin B resistant. 11. The method of any of claims 1-10, wherein the composition is a candidate antibacterial. 12. The method of claim 11, wherein the candidate antibacterial is a polypeptide or polypeptide analog. 13. The method of claim 11, wherein the candidate antibacterial is a polymyxin 14. The method of claim 11 wherein the candidate antibacterial is brevicidine. 15. The method of claim 11 wherein the candidate antibacterial is ogipeptin. 16. The method of claim 11 wherein the candidate antibacterial is pedopeptin. Active 108465066.1.DOCX 48