SPR kinetics of antibiotics binding gram-negative membranes

Designing better antibiotics starts at the membrane

Biacore™ SPR kinetics of antibiotics binding gram-negative bacterial membranes

Abstract

Polymyxins, including polymyxin B and colistin, are antibiotics of last resort to treat infection caused by multi-drug-resistant gram-negative bacteria. These natural product antibiotics bind to the lipid A component of lipopolysaccharide in the outer membrane of gram-negative bacteria (1,2). Quantitative binding measurements of polymyxins with lipid A have revealed fundamental insights into the mechanism of action for these antibiotics (3). Here, we describe the novel surface plasmon resonance assay format to study the complex environment of the outer membrane (Fig 1).

1. Introduction

The outer membrane of gram-negative bacteria

The outer membrane of gram-negative bacteria is a unique and essential structure that provides physical rigidity to the cell and serves as a permeability barrier that excludes toxigenic molecules, including many antibiotics (4-7). These critical properties are imparted by the complex glycopeptide lipopolysaccharide (LPS), which makes up the outer leaflet of the asymmetrical outer membrane bilayer. Purification of LPS has enabled studies of this critical outer membrane component in isolation. However, removing LPS from its native outer membrane environment eliminates potentially important interactions between adjacent LPS molecules and with integral outer membrane proteins (OMPs) that make up a large proportion of the outer membrane biomass. Studying interactions with LPS using intact bacterial cells and outer membrane vesicles (OMVs) can capture the complex biology of the outer membrane and provide insight into the mechanisms of action and resistance to antibiotics targeting this structure (Fig 2A)(3).

Polymyxin antibiotics

The conserved lipid A core of LPS is targeted by polymyxins (i.e., polymyxin B and colistin [polymyxin E]) which are natural products that are the antibiotics of last resort for infection caused by multidrug resistant gram-negative bacteria. Polymyxins are acylated, cyclic peptides capable of interacting with both the hydrophobic and polar regions of lipid A (Fig 2B)(1,2). Despite their clinical importance and recent insight into their activities, the precise molecular mechanisms of polymyxin activity remain to be described (3, 8-13).

Polymyxin B has been studied extensively and modifications to both polymyxin B and its lipid A target have provided insight into how these antibiotics function. Polymyxin B nonapeptide (PMBN), which lacks the acyl tail found on polymyxin B, can still bind lipid A and permeabilize the bacterial outer membrane but does not kill bacteria (14-16). These observations suggest that there are multiple cellular consequences to polymyxin binding lipid A.

Several well-characterized covalent modifications to lipid A lead to polymyxin resistance. For example, addition of a 4-amino-4-deoxy-L-arabinose (L-ara4N) or phosphoethanolamine (pEtN) to either of two lipid A phosphate groups allow bacteria to survive in the presence of polymyxins (17,18). However, the outer membranes of bacteria with modified lipid A are still dramatically perturbed by polymyxins (16). This indicates that lipid A modifications do not completely negate polymyxin binding and resistance is more complex than simply blocking this interaction.

Measuring binding of polymyxins to lipid A

Efforts to quantify polymyxin binding to lipid A, the first step in the mechanism of action for these critical antibiotics, have leveraged multiple approaches and yielded vastly differing interpretations. Reported equilibrium binding constants have ranged from 400 nM to greater than 100 µM, which far exceeds the reported minimal inhibitory concentration against bacterial cells (19-23). Many of these approaches required adding perturbative labels to the polymyxin, assumed a simple binding mechanism, and ignored the potential non-specific interactions of polymyxins with various plate surfaces. Surface plasmon resonance (SPR) avoids the need for labels, can measure association and dissociation values in addition to affinity, and measures values over a wide range of concentrations and kinetic values (3,24). Recent applications of SPR have led to a new model for the binding mechanism for polymyxins (3).

Fig 1. Surface plasmon resonance (SPR) workflow to measure polymyxin binding to the gram-negative bacterial outer membrane.

 

 

Figure 2A

Figure 2B

Fig 2. Polyxymins.and whole cells or OMVs immobilization (A) Lipopolysaccharide (LPS) (bottom right), and the conserved lipid A core of LPS (bottom left). OMVs are budded off from whole cells and capture the biological complexity of the outer membrane (right panel). (B) left, Cells or OMVs can be immobilized to the SPR chip surface via hydrophobic interactions (B) right or via amine-coupled polymyxin B. Bacteria or OMVs are prepared and loaded onto prepared chip surfaces prior to measuring binding of antibiotics.

 

2. Materials and methods

OMV isolation

E. coli OMVs were purified from E. coli BW25113 ΔtolQ, a mutant that over-produces OMVs (3, 25, 26). OMVs amounts were normalized based on total protein quantified by Bradford protein assay and stored at -80°C.

Amine coupling of polymyxin to Sensor Chip C1

Sensor Chip C1 (Cytiva) was prepared using Amine Coupling kit (Cytiva). These steps were carried out on the benchtop as this increased the ease and speed of chip preparation (multiple uniform chips can be made using the same batch of reagents), required smaller volumes of reagents compared to preparation in the machine, eliminated the risk of carry-over contamination of the reagent, and avoided introducing a high concentration of polymyxin B, a cationic lipopeptide, into all components of the flow system. Equal volumes of N-hydroxysuccinimide (NHS) and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide hydrochloride (EDC), prepared according to the manufacturer’s instructions, were mixed and immediately applied to the chip surface for two minutes. The chip was then thoroughly washed with distilled deionized water and dried carefully to avoid touching the gold surface. Following surface activation, a sufficient quantity of polymyxin B (1 mM in 1 M HEPES buffer, pH 8) was added to completely cover the chip surface and incubated for a minimum of 1 h at room temperature. The chip was washed with distilled deionized water and then treated with 1 M ethanolamine for approximately 1 min. The chip was washed and dried as described above. The prepared chip was loaded into a Biacore S200 SPR system and equilibrated in running buffer (Dulbecco's phosphate-buffered salt solution [PBS, 1x, without calcium or magnesium] at pH 7.4 with 0.0005% tween-80).

OMV capture

OMVs were diluted to approximately 20-30 µg/mL protein and capture was performed at 5 µL/min for 300 sec followed by a 300-sec stabilization period. Subsequent assay steps were run at a flow rate of 40 µL/min. The chip could be regenerated and used for multiple runs after 2 injections of 0.5% SDS (60 sec, 40 µL/min) followed by extra wash after injection with buffer and 4 carry-over controls, which prevented residual SDS from affecting subsequent cycles. PBS without divalent cations, pH 7.4, with 0.0005% tween-80 was primarily used for experiments described here, however, other buffers, including those with physiological levels of calcium and magnesium are also appropriate and should be tested for each model.

Whole cell E. coli growth and capture on chip

E. coli was grown in LB media to log phase (OD₆₀₀ ~0.4 to ~0.6) and cells were harvested by centrifugation. The cell pellet was resuspended in PBS to a final OD₆₀₀ of ~5. We note that whole cells were less consistent during capture step than OMVs even when highly concentrated. Regeneration of Sensor Chip C1 after whole cell binding proved more demanding than for OMVs, requiring additional wash steps: PBS with 32 mM MgCl₂ for 120 sec, 2.5 M NaCl for 30 sec, 0.5% SDS (Cytiva Desorb 1) for 60 sec. Buffer washes were performed between each step and carry-over controls to prevent residual detergents or salts from affecting subsequent experimental cycles. We also found chip life was reduced when running whole-cell bacteria compared to OMVs. It is important to note that the bacteria are viable and can potentially grow inside the flow cell chamber so machine clean up and sanitization cycle (0.5% SDS, Glycine, pH 9.5, and sodium hypochlorite, as recommended by Biacore system manual) should be run immediately after experiments.

Data analysis

Quantification of binding

The change in RU from double-referenced traces were first converted to concentration using the assumption that 100 RU = 1 mg/mL. This could then be converted to μM using the molecular weight, which also allowed different sized analytes to be directly compared. To normalize to and account for variations in captured OMV levels, the concentration was expressed as μM/100 RU captured OMV of that trace. This method of quantification allows multiple cycles to be directly compared.

Chaser method to determine residence time

The chaser method was used to determine the residence time (t_1/2) of polymyxin B (27). This technique effectively accounts for drift in the machine that can occur over the extended incubation periods necessary to measure off rates of tightly bound species by using a re-saturation method to measure the free occupancy after a known period of time.

OMVs loaded as previously described were first saturated with polymyxin B (5 µM, 300 sec association time, flow rate 5 µL/min with a 240 sec dissociation time). A second injection (60 sec, 30 µL/min) assured saturation, and bound RU was measured 120 sec after the end of the injection. The total change in RU is the “R_max”. After a 2 h dissociation was the ‘chaser’ pulse (60 sec, 30  of 5 µM polymyxin) and the change in RU measured again 120 sec after the end of the injection, this is the “R_free”. The RUs of associated polymyxin B were determined 120 sec after the start of the dissociation. The fraction occupancy was determined along with residence time and half-life as shown (Fig 5A).

 

R (fraction occupancy) = (R_max-R_free>)/R_max

 

Half-life (t_1/2) = -(ln(2)Δt)/ln(R), t = time (between measured R_max and R_free)

K_off (=ln2/t_1/2)

 

Apparent K_D

OMVs are a relatively complex system compared to most ligand-analyte binding interactions interrogated with SPR. As such the measured K_D likely represents multiple binding interactions and is thus referred to as the “apparent K_D”. For polymyxin the OMVs were first saturated with polymyxin B prior to the SCK dose series to in as much as possible isolate the saw-tooth profile binding species. The apparent-K_D of the saw-tooth binding species was calculated by plotting the RU change from each base to plateau (of double-referenced traces) against the polymyxin concentration in a standard affinity plot or using Biacore S200 evaluation software. Data from a minimum of three runs were averaged, with the value for each run obtained from the average of 2-3 individual cycles.

3. Results

Biacore SPR assay development considerations

The implementation of SPR for studying polymyxin binding to the outer membrane required the evaluation of different sensor surfaces. Lipophilic chips, such as Sensor Chip L1, have a hydrophobic surface designed to bind liposomes and vesicles (Cytiva) (Fig 2B). E. coli OMVs can be easily immobilized on the surface of Sensor Chip L1, however, the lipophilic chip surface presents challenges for studying the interactions of polymyxins with lipid A. The chip surface is incompatible with detergents that are necessary to reduce non-specific binding of polymyxin B to plastic components of the Biacore system (28). Additionally, polymyxin B itself interacts with the hydrophobic chip surface. Thus, with Sensor Chip L1, there was a high background making it difficult to distinguish and resolve the specific interaction of polymyxin B with lipid A in OMVs, particularly at lower concentrations.

To overcome these limitations, we developed an immobilization strategy using Sensor Chip C1, which possesses a planar surface lacking dextran ³. To capture OMVs onto Sensor Chip C1, we amine-coupled polymyxin B directly to the chip surface (Fig 2C). This approach allowed for the stable capture of OMVs in the presence of 0.0005% tween-80 and eliminated the high levels of interaction with the chip surface. The Sensor Chip C1-polymyxin B chip surface could also be regenerated using standard methods (see Materials and methods).

Polymyxin B binding to OMVs

We leveraged E. coli OMVs captured on Sensor Chip C1 to dissect the kinetics of binding of polymyxin B to lipid A in the outer membrane. To obtain single-cycle kinetics, a two-fold dose series of polymyxin B was serially injected at increasing concentrations. The resulting SPR sensorgrams revealed a complex interaction characterized by accumulation of a tightly bound polymyxin B species superimposed with a transient binding event (Fig 3A). At lower concentrations of polymyxin B, the tightly bound species dominated the SPR sensorgram. Our interpretation was that the initial binding between polymyxin B and lipid A led to complexes that rapidly transitioned to a stable, tightly bound state. At higher polymyxin B concentrations, as the binding capacity for the tightly bound polymyxin B approaches saturation, reversible binding, indicated by a saw-tooth profile, becomes more dominant. When the tightly bound state plateaued, the concentration dependent, reversible binding pattern remained (Fig 3A). These sensorgrams enabled generation of a model explaining the mechanism of polymyxin B binding (3).

Given the complexity of OMVs and the interactions with polymyxins, analysis of these traces was more complicated compared to analyzing binding to a purified protein product. Importantly, the biological variability in size, weight, and composition of each OMV prevents normalization of the traces by R_max. We observed that the degree of OMV capture varied widely, and this was affected by, among other variables, the concentration of OMVs used in the capture step and the number of previous cycles performed with the chip. Across a range of OMV capture levels, we observe that the amount of polymyxin B necessary to saturate the tightly bound binding species (before the reversible, saw-tooth binding was observed) increased in proportion to the amount of OMVs captured on the chip (Fig 3A, high concentration 5 μM versus high concentration 0.625 μM dose series). These observations were first made qualitatively. To quantify the amount of tightly bound and reversibly bound polymyxin, the change in RU from double-referenced sensorgrams was first converted to concentration using the conversion factor 100 RU = 1 mg/mL, then converted to μM using the molecular weight. This value was expressed as μM/100 RU captured OMVs to normalize the amount of polymyxin bound to the amount of captured OMVs in that cycle thus allowing for comparisons across runs (Fig 3B).

Fig 3. Polymyxin B interacts with the gram-negative outer membrane via multiple binding modes (A) Sensorgrams of single-cycle kinetics of polymyxin B binding to wild-type E. coli OMVs measured at different levels of OMV capture. Highest dose in 8 step, 2-fold dilution series and RU of OMV captured indicated in graph legend (B) Quantification of stable and reversible binding amount quantified per 100 RU of OMVs.

 

E. coli cells

E. coli are rod-shaped bacteria that measure 0.5-1 μm along their shorter axis (29). Although considerably larger than most ligands bound for SPR, bacterial cells still lie within the detectable range of the SPR sensor (30). To determine if OMVs accurately recapitulate the outer membrane of an intact cell, whole E. coli cells (non-pathogenic K12 strain) were captured onto the amine-coupled polymyxin Sensor Chip C1 surface. We found the cells grown to log phase and washed in PBS were captured more readily on the surface, although we noted the capture levels were significantly variable between cycles. Importantly however, the complex binding profile of polymyxin B observed with OMVs was recapitulated on whole E. coli cells, supporting the use of OMVs as a model for understanding interactions with the OM (Fig 5A).

The observation that whole E. coli cells and OMVs resulted in the same conclusions supports OMVs as a biological relevant model for the interaction of polymyxins with the outer membrane. Experimentally, the use of OMVs in SPR experiments offered several advantages compared to whole cells or even purified LPS. OMVs can be produced in large batches and frozen. They can be isolated from various genetic backgrounds, including the polymyxin-resistant strains, and from potentially any gram-negative species, allowing for the investigation of LPS types that are not commercially available. Moreover, purified LPS fails to capture additional interactions, such as with outer membrane proteins (OMPs), that might be important for binding the outer membrane. From a technical standpoint, OMVs displayed reduced variability over whole live cells, and provided more consistent, and sufficient, capture levels for quantitative analysis (Fig 4A and 4B). Thus, OMVs provide an easily obtained, tractable, reliable, and scalable platform for monitoring outer membrane interactions by SPR.

Fig 4. (A) Representative sensorgram of polymyxin B binding to whole E. coli cells, 260 RU captured. (B) Representative sensorgram of PMBN binding to wild-type OMVs exhibiting reversible, saw-tooth binding. 510 RU captured. (C) Representative sensorgram of polymyxin B binding to OMVs isolated from a strain of E. coli that is resistant to polymyxins due to covalent modification to lipid A. 785 RU captured.

 

PMBN and polymyxin-resistant OMVs

PMBN lacks the final amino acid and acyl tail of polymyxin B resulting in a molecule that no longer kills gram-negative bacteria but is still able to bind to lipid A and permeabilize the outer membrane barrier (14, 16). To compare binding of PMBN with that of polymyxin B, we performed single-cycle kinetics with PMBN and OMVs on Sensor Chip C1 as described above for polymyxin B. PMBN exhibited a binding profile distinct from that of polymyxin B in that it completely lacked the accumulation of a tightly bound population and displayed only reversible, saw-tooth binding (Fig 4). This observation indicates that different parts of the polymyxin B molecule are likely responsible for distinct steps in the interaction with the outer membrane.

Resistance to polymyxins is typically gained through covalent modifications to one or both of the lipid A phosphates by pEtN or L-ara4N (17, 18). OMVs were isolated from an E. coli polymyxin-resistant mutant (E. coli pmrA^G53E) to determine the impact of lipid A modification on polymyxin B binding (3). Polymyxin B and PMBN (not shown) exhibited only the reversible, saw-toothed binding pattern to the resistant OMVs (OMV-R), with little to no accumulation of the tightly bound species (Fig 4). It is worth noting that while bacteria with modified lipid A are not killed by polymyxin B, this antibiotic is still able to disrupt the outer membrane permeability barrier (16). These results suggest that the reversible, saw-tooth binding of polymyxins (including PMBN), describes the interaction, which is responsible for disruption of the outer membrane, while the acyl-tail portion is necessary for the tightly bound species. Thus, SPR enabled dissection of the interactions of an important antibiotic with the gram-negative outer membrane and revealed previously undescribed interactions.

Determination of apparent-K_D and residence time

To determine the kinetics of the two primary observed binding species, we first isolated the reversible, saw-tooth’, binding observed with polymyxin B interaction by saturating wt-OMVs with polymyxin B prior to running SCK of 8 two-fold increasing doses (Fig 5A). Plotting the change in RU over concentration gives a standard affinity plot from which K_D can be determined, but as OMVs (and the interaction) are likely more complicated than the usual pure ligand-analyte interactions studied with SPR, this affinity constant is referred to as the apparent K_D. We determined an average apparent K_D of 517 nM for the reversible binding interaction of polymyxin B on wt-OMVs. Notably this was not significantly different than that observed for PMBN or interactions with OMVs derived from polymyxin-resistance bacteria (3).

The observed stable binding of polymyxin B to wild-type cells and OMVs is novel and suggests a tight, long-lived interaction. This interaction was quantified by a chaser approach that is able to measure interactions with slow dissociation rates (Fig 5B)(27). Polymyxin B binding to OMVs exhibited a half-life of over 6 hours at both 25°C and 37°C. In log phase growth, E. coli can double every 20-30 minutes and thus, the interaction of polymyxin B with the outer membrane is essentially irreversible on the timescale of bacterial cell division.

Fig 5. Illustration of methods to determine binding kinetics of polymyxin B (A) The reversible, saw-tooth binding of polymyxins to OMVs is present even when the stable, long-lived species is saturated. An effective K_D modeled from this isolated, reversible binding (red square) is 517 nM (SD=499, n=4). (B) Chaser method experiment to determine the half-life (t1/2) of the interaction between polymyxin B and lipid A in OMVs. Sensorgram example highlighting the total initial polymyxin B loading of OMVs (R_max) and the remaining free space after 2 h of dissociation measured after a pulse chase with polymyxin (R_free).

 

4. Conclusions

The rapid spread of antibiotic resistance is a global threat to human health. A better understanding of the mechanisms of action for antibiotics could help improve our ability to address resistance and facilitate discovery approaches for novel therapeutics. Despite over half a century of clinical use, many gaps remain in our understanding of polymyxins. Modeling based on recent SPR studies examining the interactions between polymyxin B and the outer membrane provide a new perspective on the mechanism of killing by polymyxins and how lipid A modifications lead to resistance. These insights could help in the design of approaches to identify molecules able to overcome the limitations of polymyxins, the antibiotics of last resort to treat infections by gram-negative bacteria.

More broadly, these studies highlight the value to a deeper kinetic understanding of the interactions required for bacterial cell killing. For gram-negative bacteria, all antibiotics must interact with, or overcome, the outer membrane barrier in order to access targets inside of the cell. Additionally, while our work was focused on E. coli, this method should be amenable to OMVs or cells from any gram-negative bacterial species of interest. Thus, leveraging our OMV-SPR approach provides a method to study this first step in the action of any antibiotic with this important group of bacteria.

References

1. Velkov T, Thompson PE, Nation RL, Li J. Structure−Activity Relationships of Polymyxin Antibiotics. Journal of Medicinal Chemistry. 2010;53(5):1898-1916. doi:10.1021/jm900999h

2. Mares J, Kumaran S, Gobbo M, Zerbe O. Interactions of Lipopolysaccharide and Polymyxin Studied by NMR Spectroscopy. Journal of Biological Chemistry. 2009;284(17):11498-11506. doi:10.1074/jbc.m806587200

3. Buchholz KR, Reichelt M, Johnson MC, et al. Potent activity of polymyxin B is associated with long-lived super-stoichiometric accumulation mediated by weak-affinity binding to lipid A. Nature Communications. 2024;15(1). doi:10.1038/s41467-024-49200-5

4. Guest RL, Rutherford ST, Silhavy TJ. Border Control: Regulating LPS Biogenesis. Trends in Microbiology. 2021;29(4):334-345. doi:10.1016/j.tim.2020.09.008

5. Silhavy TJ, Kahne D, Walker S. The bacterial cell envelope. Cold Spring Harbor Perspectives in Biology. 2010;2(5). doi:10.1101/cshperspect.a000414

6. Nikaido H. Molecular Basis of Bacterial Outer Membrane Permeability Revisited. Microbiology and Molecular Biology Reviews. 2003;67(4):593-656. doi:10.1128/mmbr.67.4.593-656.2003

7.Rojas ER, Billings G, Odermatt PD, et al. The outer membrane is an essential load-bearing element in Gram-negative bacteria. Nature. 2018;559(7715):617-621. doi:10.1038/s41586-018-0344-3

8. Khadka Nawal K, Aryal CM, Pan J. Lipopolysaccharide-Dependent Membrane Permeation and Lipid Clustering Caused by Cyclic Lipopeptide Colistin. ACS Omega. 2018;3(12):17828-17834. doi:10.1021/acsomega.8b02260

9. Hsia CY, Chen L, Singh RR, DeLisa MP, Daniel S. A Molecularly Complete Planar Bacterial Outer Membrane Platform. Scientific Reports. 2016;6:32715. doi:10.1038/srep32715

10. Berglund NA, Piggot TJ, Jefferies D, Sessions RB, Bond PJ, Khalid S. Interaction of the Antimicrobial Peptide Polymyxin B1 with Both Membranes of E. coli: A Molecular Dynamics Study. van der Spoel D, ed. PLOS Computational Biology. 2015;11(4):e1004180. doi:10.1371/journal.pcbi.1004180

11. Nguyen HT, O’Donovan LA, Venter H, et al. Comparison of Two Transmission Electron Microscopy Methods to Visualize Drug-Induced Alterations of Gram-Negative Bacterial Morphology. Antibiotics. 2021;10(3):307-307. doi:10.3390/antibiotics10030307

12. Manioglu S, Modaresi SM, Ritzmann N, et al. Antibiotic polymyxin arranges lipopolysaccharide into crystalline structures to solidify the bacterial membrane. Nature Communications. 2022;13(1):6195. doi:10.1038/s41467-022-33838-0

13. Sabnis A, Hagart KL, Klöckner A, et al. Colistin kills bacteria by targeting lipopolysaccharide in the cytoplasmic membrane. eLife. 2021;10. doi:10.7554/elife.65836

14. Vaara M, Vaara T. Polycations sensitize enteric bacteria to antibiotics. Antimicrobial Agents and Chemotherapy. 1983;24(1):107-113. doi:10.1128/aac.24.1.107

15. Tsubery H, Ofek I, Cohen S, Fridkin M. Structure-function studies of polymyxin B nonapeptide: implications to sensitization of gram-negative bacteria. Journal of Medicinal Chemistry. 2000;43(16):3085-3092. doi:10.1021/jm0000057

16. MacNair CR, Stokes JM, Carfrae LA, et al. Overcoming mcr-1 mediated colistin resistance with colistin in combination with other antibiotics. Nature Communications. 2018;9(1). doi:10.1038/s41467-018-02875-z

17. Aghapour Z, Gholizadeh P, Ganbarov K, et al. Molecular mechanisms related to colistin resistance in Enterobacteriaceae. Infection and drug resistance. 2019;12:965-975. doi:10.2147/IDR.S199844

18. Liu YY, Wang Y, Walsh TR, et al. Emergence of plasmid-mediated colistin resistance mechanism MCR-1 in animals and human beings in China: a microbiological and molecular biological study. The Lancet Infectious Diseases. 2016;16(2):161-168. doi:10.1016/S1473-3099(15)00424-7

19. Thomas CJ, A Surolia. Kinetics of the interaction of endotoxin with polymyxin B and its analogs: a surface plasmon resonance analysis. FEBS Letters. 1999;445(2-3):420-424. doi:10.1016/s0014-5793(99)00150-7

20. McInerney MP, Roberts KD, Thompson PE, et al. Quantitation of Polymyxin–Lipopolysaccharide Interactions Using an Image-Based Fluorescent Probe. Journal of Pharmaceutical Sciences. 2016;105(2):1006-1010. doi:10.1016/j.xphs.2015.10.028

21. French S, Farha M, Ellis MJ, et al. Potentiation of Antibiotics against Gram-Negative Bacteria by Polymyxin B Analogue SPR741 from Unique Perturbation of the Outer Membrane. ACS Infectious Diseases. 2019;6(6):1405-1412. doi:10.1021/acsinfecdis.9b00159

22. Clausell A, Rabanal F, Garcia-Subirats M, Alsina MA, Cajal Y. Membrane Association and Contact Formation by a Synthetic Analogue of Polymyxin B and Its Fluorescent Derivatives. The Journal of Physical Chemistry B. 2006;110(9):4465-4471. doi:10.1021/jp0551972

23. Velkov T, Soon RL, Chong PL, et al. Molecular basis for the increased polymyxin susceptibility of Klebsiella pneumoniae strains with under-acylated lipid A. Innate Immunity. 2012;19(3):265-277. doi:10.1177/1753425912459092

24. Nicolò Paracini, Clifton LA, Skoda A, Lakey JH. Liquid crystalline bacterial outer membranes are critical for antibiotic susceptibility. Proceedings of the National Academy of Sciences. 2018;115(32). doi:10.1073/pnas.1803975115 , 201803975–8 (2018)

25. Halima Chutkan, MacDonald I, Manning A, Kuehn MJ. Quantitative and Qualitative Preparations of Bacterial Outer Membrane Vesicles. Methods in molecular biology. Published online November 7, 2012:259-272. doi:10.1007/978-1-62703-245-2_16

26. Bernadac A, Gavioli M, Lazzaroni JC, Raina S, LloubèsR. Escherichia coli tol-pal Mutants Form Outer Membrane Vesicles. Journal of Bacteriology. 1998;180(18):4872-4878. doi:10.1128/jb.180.18.4872-4878.1998

27. Quinn JG, Pitts KE, Steffek M, Mulvihill MM. Determination of Affinity and Residence Time of Potent Drug-Target Complexes by Label-free Biosensing. Journal of Medicinal Chemistry. 2018;61(12):5154-5161. doi:10.1021/acs.jmedchem.7b01829

28. Karvanen M, Malmberg C, Lagerbäck P, Friberg LE, Cars O. Colistin Is Extensively Lost during Standard In Vitro Experimental Conditions. Antimicrobial Agents and Chemotherapy. 2017;61(11):e00857-17. doi:10.1128/AAC.00857-17

29. Size Limits of Very Small Microorganisms. National Academies Press; 1999. doi:10.17226/9638

30. Abbas A, Linman MJ, Cheng Q. Sensitivity comparison of surface plasmon resonance and plasmon-waveguide resonance biosensors. Sensors and Actuators B: Chemical. 2011;156(1):169-175. doi:10.1016/j.snb.2011.04.008

 

* Kerry R. Buchholz, Scientist 4, Department of Infectious Diseases and Host-Microbe Interactions, Genentech, Inc.

Steven T. Rutherford, Director and Senior Principal Scientist, Department of Infectious Diseases and Host-Microbe Interactions, Genentech, Inc.

John G. Quinn, Distinguished Scientist, Department of Biochemical and Cellular Pharmacology, Genentech, Inc.