Activity and Predicted Nephrotoxicity of Synthetic Antibiotics Based on Polymyxin B.

The polymyxin lipodecapeptides colistin and polymyxin B have become last resort therapies for infections caused by highly drug-resistant Gram-negative bacteria. Unfortunately, their utility is compromised by significant nephrotoxicity and polymyxin-resistant bacterial strains. We have conducted a systematic activity-toxicity investigation by varying eight of the nine polymyxin amino acid free side chains, preparing over 30 analogues using a novel solid-phase synthetic route. Compounds were tested against a panel of Gram-negative bacteria and counter-screened for in vitro cell toxicity. Promising compounds underwent additional testing against primary kidney cells isolated from human kidneys to better predict their nephrotoxic potential. Many of the new compounds possessed equal or better antimicrobial potency compared to polymyxin B, and some were less toxic than polymyxin B and colistin against mammalian HepG2 cells and human primary kidney cells. These initial structure-activity and structure-toxicity studies set the stage for further improvements to the polymyxin class of antibiotics.

Introduction

Bacterial sepsis kills millions of people every year. In the United States, patients hospitalized for septicemia or sepsis are more than eight times as likely to die during their hospitalization compared to those hospitalized for other diagnoses,[1] with over 200,000 deaths in 2008.[2] Over one-third of patients with sepsis treated in an intensive care unit die in the hospital.[3] Sepsis is the most expensive condition treated in U.S. hospitals, costing more than $20 billion in 2011.[4] Sepsis is most commonly caused by Gram-positive (G+ve) Staphylococcus aureus and Streptococcus pyogenes and Gram-negative (G-ve) Klebsiella spp., Escherichia coli, and Pseudomonas aeruginosa.[5] Unfortunately, there has been an ominous rise in highly drug-resistant G-ve bacteria that are able to overcome almost every known antibiotic, threatening a return to a “preantibiotic” era if new therapies are not discovered. The current clinical pipeline is mainly populated by G+ve drug candidates.[6] One possible approach to antibiotic development is to revisit antibiotics discovered in the “golden age” of antibiotic discovery (the 1950s–1970s) to see if they can be optimized or improved.

In 1947 two American groups independently discovered antibiotic substances produced by Bacillus polymyxa.[7−9] The same year a group in England described a new antibiotic they named aerosporin isolated from a strain at first thought to be B. aerosporus, but later shown to be identical to the strain B. polymyxa.[10] Many examples of polymyxin derivatives have since been isolated from nature, all of them with various degrees of activity against bacteria.[11−15] A review by Stansly published in 1949 on early human clinical use of four distinct polymyxin antibiotic complexes A, B, C, and D concluded that there is “little doubt that polymyxins have exhibited effective therapeutic activity in certain infections of man produced by Gram-negative bacteria. It is apparent, however, that their application in this field is likely to be limited because of the occurrence of untoward reactions elicited by material used in the clinical trials.”[16] Nonetheless, two of the polymyxins, polymyxin B 1 and polymyxin E (colistin) 2, entered clinical use and in recent years have become last-resort antibiotics for multidrug resistant G-ve infections, fueling a surge of interest in their use.[17−22] The main limitation of the polymyxins is nephrotoxicity, with adverse renal toxicity observed in over 50% of cases in some studies.[23,24] This has led to a systematic effort to improve dosing regimens.[25,26]

Polymyxins 1 and 2 are both positively charged macrocyclic peptide antibiotics composed of a heptapeptide ring and three amino acid exocyclic tail capped by a mixture of two fatty acids (see Figure ), differing only in a d-Phe/d-Leu alteration at position-6. They are highly basic, containing six 2,4-diaminobutyric acid (Dab) residues, including one residue whose side chain is used for lactam formation with the C-terminal Thr residue. While polymyxin B is administered as the parent compound sulfate salt, colistin is almost always used as a polymethanesulfonylated prodrug, colistin methanesulfonate sodium, which spontaneously hydrolyses to release polymyxin E 2. Given that the polymyxins are natural products produced by fermentation, their composition does vary considerably between manufacturers.[27]

Figure 1

Structures of polymyxin B 1 and polymyxin E (colistin) 2.

The antimicrobial activity of the polymyxins is initiated by binding to Lipid A, the membrane-anchoring component of lipopolysaccharide (LPS) found in the outer membrane of G-ve bacteria.[17,28,29] Displacement of membrane-stabilizing divalent cations and insertion of the fatty acyl chain expands and weakens the bacterial outer membrane, causing cell leakage and allowing polymyxin entry.[17,30,31] Once inside the cell, further modes of action have been postulated, including the generation of hydroxyl radical production[32] and inhibition of the respiratory enzyme NADH-quinone oxidoreductase.[33,34]

Unfortunately, resistance to the polymyxins is appearing and has been identified in most major pathogens, including A. baumannii, K. pneumoniae, and P. aeruginosa.[17] The main mechanisms of resistance involve alterations to LPS, in particular via addition of phosphoethanolamine and/or l-4-amino-arabinose to Lipid A, which lowers the binding affinity with the Dab residues of the polymyxins due to reduction of negative charge.[17,35,36] This increase in resistance, coupled with the growing importance of polymyxins in treating highly drug-resistant G-ve infections, has led to a number of efforts in recent years to prepare polymyxin analogues to understand structure–activity relationships, improve potency, reduce nephrotoxicity, and overcome resistance. Despite their length of time in use, there have been relatively few structure–activity relationship (SAR) studies conducted to date compared to other classes of antibiotics, with most publications focusing on the readily modified exocyclic tail (Pfizer,[37−39] Novacta Biosystems Ltd./Cab-Tab Anti-Infectives Ltd.,[40−42] Northern Antibiotics,[43−45] Sakura et al.[46]) or selective/nonselective modifications of the multiple Dab side chains (Northern Antibiotics,[43−45] Pfizer,[37−39,47] and Schering-Plough[48]). Biosource Pharm Inc./Cubist developed a method for fatty acid deacylation and linear tail amino acid hydrolysis, followed by reconstruction of the linear chain.[49−54] From this study, CB182,804[55] entered Phase I clinical trials, but development was discontinued in 2009. In one of the few studies to investigate changes within the polymyxin ring system, Li et al. at Monash University patented a set of polymyxin derivatives that retain activity against polymyxin-resistant strains; this group centered their efforts on the variation of hydrophobicity at positions 6 and 7 (numbering as in Figure ).[56−58] Early efforts at polymyxin SAR were summarized by Storm et al. in 1977,[59] while more recent developments were reviewed by Velkov et al. in 2010[30] and Vaara in 2013.[60]

A set of novel polymyxin derivatives developed by Northern Antibiotics, with the exocyclic Dab-1 residue removed/substituted and the fatty acid tail altered, were found to be less toxic than polymyxin B[61] based on cytotoxicity testing in the porcine renal proximal tubular cell line LLC-PK1. It was proposed that this was due to a reduction in the number of positive charges coupled with alteration of hydrophobicity of the fatty acid. Based on these results and the improved activity of the Monash compounds, we decided to investigate the effects of alterations at other positions within the polymyxins, with a particular focus on modifying the basic Dab residues contained within the ring to alter their basicity. Given that the Dab residues are proposed to be integral to binding to Lipid A and that mutation of Lipid A to introduce phosphoethanolamine is a key polymyxin resistance mechanism, we also wanted to examine whether introducing zwitterionic residues or substituents capable of hydrogen-bonding at these positions might help overcome resistance.

Twenty-six new analogues were prepared and assayed. In addition, six key compounds from the literature were also prepared to provide a systematic SAR comparison assessed under common conditions. Some of the compounds retained better antimicrobial activity against polymyxin-resistant strains compared to polymyxin B and colistin. Most importantly, some of the most promising compounds showed less cytotoxicity than polymyxin B and colistin when tested in mammalian cell lines, and less toxicity than polymyxin B, colistin, and gentamicin in a cytotoxicity assay potentially predictive of nephrotoxicity as measured by LDH and GGT release from human primary kidney cells.

Results and Discussion

Chemistry

A total of 32 compounds was synthesized (six of these have been reported previously[46−48,53,54,57,58]) with the structures presented in Tables and S1 (Supporting Information). All compounds possessed >95% purity, as determined by LCMS analysis using both ELSD and UV (210 nm) detection. The synthetic routes toward the final compounds in this library are summarized in Schemes , 2, and S2 (Supporting Information). The first solid-phase total synthesis of polymyxin B1 was developed in 1999 by Sharma[62] using Fmoc-Thr(tBu)-SASRIN resin, with the linear peptide cleaved from the resin and cyclized in solution. In the current study we present a new synthetic route to the polymyxins employing an on-resin cyclization that allows for alteration of any substituent position other than Thr-10 (see Scheme for a representative synthesis of 10). The first step toward this library of compounds was the O-allyl ester protection of the C-terminal amino acid Fmoc-l-Thr-OH 3 (Thr10 in the final product). The side-chain alcohol of the protected amino acid 4 was then loaded onto a dihydropyran DHP HM resin (3,4-dihydro-2H-pyran-2-yl-methoxymethyl polystyrene). The peptide sequence was constructed by standard solid-phase peptide synthesis (SPPS) using sequential Fmoc deprotection followed by amino acid attachment using a systematic double coupling protocol, including a capping step with acetic anhydride and pyridine (50:50 v/v) to avoid the formation of deletion products due to unreacted free amino groups. On-resin cyclization was conducted following incorporation of the Fmoc-Dab(Alloc) residue by in situ deprotection of both Thr10 allyl ester and the Dab-4 side chain amine Nγ-Alloc protecting groups (Scheme , step vii), followed by overnight cyclization using DPPA and DIPEA in DMF at room temperature. Fmoc deprotection of the cyclized product then allowed completion of the exocyclic linear tail by further amino acid couplings. Following each step, an analytical sample of resin was cleaved with TFA/Et3SiH/H2O (95:1:4) followed by LCMS analysis to confirm the coupling with the corresponding amino acid. The final coupling capped the peptides with the appropriate fatty acid. Cleavage of the cyclic decapeptides from the resin with concomitant side chain deprotection was achieved with the addition of a solution of TFA/Et3SiH/H2O (95:1:4) for 30 min at room temperature, providing crude compounds that were then HPLC-purified to afford the final polymyxin analogue TFA salts as white powders; total yields were generally between 5 and 30% based on the initial resin loading.

Table 1

Compound Structures

All amino acids are l-isomers unless otherwise indicated. Pmx-B: polymyxin B.

11 = polymyxin B3.

12 = compound 5 in O’Dowd et al.[47]

13 = compound 5 in Leese et al.[54]

15 = compound 18 in Leese et al.[53]

18 is similar to compound 24 (retains PmxB mixture of fatty acids) in Weinstein et al.[48]

28 = dl-OctGly version of l-Ade compound 2 in Li et al.[57]

30 = dl-OctGly version of d-Ade compound 1 in Li et al.[58]

Scheme 1

General On-Resin Cyclization Synthetic Route Exemplified by Synthesis of 10

Reagent and conditions: (i) Cs2CO3, MeOH, H2O, pH 7-8; (ii) allyl bromide, dry DMF; (iii) DHP polystyrene resin, PPTS, dry DCE; (iv) 30% piperidine, DMF; (v) solid-phase peptide synthesis (SPPS) with corresponding amino acid (Table S3), HCTU, DIPEA; (vi) acetic anhydride, pyridine; (vii) Pd(PPh3), PhSiH3; (viii) DPPA, DIPEA, DMF; (ix) 4-Ph-PhCO2H, HCTU, DIPEA; (x) TFA/Et3SiH/H2O (95:1:4).

Scheme 2

General On-Resin Cyclization Synthetic Route with Selective Dab Modification Used for Synthesis of Compounds 18, 19, 22, 23, 31, and 32

Reagent and conditions: (iv) 30% piperidine, DMF; (v) solid-phase peptide synthesis (SPPS) with corresponding amino acid (Table S3), HCTU, DIPEA; (vi) acetic anhydride, pyridine; (vii) Pd(PPh3), PhSiH3; (viii) DPPA, DIPEA, DMF; (ix) n-C7CO, HCTU, DIPEA; (x) 2% H2NNH2 in DMF; (xi) peptide coupling with corresponding amino acid: Boc-l-Arg(Pbf)-OH or Boc-l-Glu(OtBu)-OH, HCTU, DIPEA; (xii) TFA/Et3SiH/H2O (95:1:4).

For some syntheses, an orthogonal protecting group was employed for one of the multiple Dab residues, allowing for selective modification of individual Dab side chains. Thus, side-chain ivDde-protected Dab was used at positions P5, P8, or P9 (see Scheme for generic route; Scheme S1 for explicit synthesis of 18 and 19). The P4 Alloc-protected Dab was deprotected and used for cyclization as before, with the fully protected resin-bound peptide treated with 2% hydrazine to selectively unmask the desired Dab residue. This was then derivatized by coupling with Boc-Arg(Pbf)-OH or Boc-Glu(OtBu)-OH, with global deprotection and purification as previously described.

Biological Activity

The 32 compounds, along with vancomycin, gentamicin, colistin, and polymyxin B, were assessed against a panel of predominantly G-ve bacteria, with S. aureus used as a representative G+ve strain. Five antibiotic-sensitive and resistant ATCC reference strains covering the G-ve ESKAPE pathogens (E. coli, K. pneumoniae, A. baumannii, and P. aeruginosa) were used, along with polymyxin-resistant clinical isolates of K. pneumoniae, A. baumannii, and P. aeruginosa (Table S2). Notably, compounds that were resyntheses of those previously reported demonstrated similar activity to the earlier literature values (footnotes in Tables and 3). Counter-screening for general cytotoxicity was conducted using human liver hepatocellular carcinoma cells (HepG2). Our previous studies have demonstrated that cytotoxicity measurements against these immortalized cells were not predictive of nephrotoxicity,[63,64] but they do provide an indication of potential general cell toxicity.

Table 2

Summary of MIC against a Panel of Sensitive and Resistant Bacteria and Cytotoxicity (CC50) against HepG2 Cells for a Set of Compounds with Diversity in the Exocyclic Chain and Fatty Acid Tail

All amino acids are l-isomers unless otherwise indicated. ND: not determined. PmxR: polymyxin resistant. CI: clinical isolate. MSSA: methicillin sensitive S. aureus. MICs were determined visually after 18 h of incubation at 37 °C, with the MIC defined as the lowest compound concentration at which no bacterial growth was visible. MIC in μg/mL. 2-Chlorophenyl carbamoyl (2-ClPhNHCO), 4-biphenyl-benzoyl (Ph-4-PhCO), 4-phenoxybenzoyl (PhO4-PhCO), 4-biphenyl carbamoyl (Ph-4-PhNHCO), octanoyl (nC7CO).

11 is polymyxin B3. Literature MIC E. coli = 0.5 and P. aeruginosa = 1.0 nmol/mL. E. coli IFO12734, and P. aeruginosa IFO3080 from the Institute for Fermentation, Osaka, Japan.[46]

12 is compound 5 in O’Dowd et al.; MIC E. coli = 0.03, P. aeruginosa = 4, S. aureus = >16 μg/mL.[47]

13 is compound 5 in Leese et al.; MIC90 P. aeruginosa (100 strains) = 2, A. baumannii (81 strains) = 4, E. coli (80 strains) = 2, K. pneumoniae (81 strains) = 4 μg/mL.[54]

15 is compound 18 in Leese et al.; MIC E. coli = 2.5, P. aeruginosa ATCC 27853 = 2.5 μg/mL; MIC values were determined by serial 2-fold broth dilution method.[53]

Table 3

Summary of MIC against a Panel of Sensitive and Resistant Bacteria and Cytotoxicity (CC50) against HepG2 Cells for a Set of Compounds with Diversity in the Heptapeptide Ring

All amino acids are l-isomers unless otherwise indicated. ND: not determined. PmxR: polymyxin resistant. CI: clinical isolate. MSSA: methicillin resistant S. aureus. MICs were determined visually after 18 h of incubation at 37 °C, with the MIC defined as the lowest compound concentration at which no bacterial growth was visible. MIC in μg/mL. 2-Chlorophenyl carbamoyl (2-ClPhNHCO), 4-biphenyl-benzoyl (Ph-4-PhCO), 4-phenoxybenzoyl (PhO4-PhCO), 4-biphenyl carbamoyl (Ph-4-PhNHCO), octanoyl (nC7CO).

Values in parentheses are reported MIC for compound 2 = l-Ade version of 28 in Li et al.[57]

Values in parentheses are reported MIC for compound 1 = d-Ade version of 30 in Li et al.[58]

18 similar to compound 24 (retains polymyxin B fatty acid) in Weinstein et al.; MIC E. coli = 0.16, P. aeruginosa = 0.03, S. aureus = 2 μg/mL.[48]

Alterations to the fatty acid tail (Table ) had minimal effect on activity against the reference strains, but biphenyl or biphenyl ether substituents (10, 14, and 15) did promote improved activity against polymyxin-resistant strains, albeit accompanied by greater cytotoxicity. In the linear exocyclic peptide tail (Table ), replacement of Dab-3 with d-Dab (36, 16) or Gly (38) had minimal effect. Analogue 38 is particularly interesting as it retained MIC activity among the reference strains (0.5–2 μg/mL for most strains) with low cytotoxicity seen in HepG2 cells (CC50 > 300 μM). This is the only polymyxin B example where Dab-3 has been replaced by a nonbasic amino acid, where the amino acid replacement also potentially impacts on the final structural geometry. The effects of substitution of Dab-3 with d-Ser depended on the fatty acid; with the heptanoic acid 37 activity was slightly decreased (vs 11), but with o-chlorophenylurea 17 most activity was lost. The inversion of l-Dab-1 to d-Dab-1 in 40 abrogated most activity, a much larger effect than seen with the inversion of l-Dab-3 to d-Dab-3 in 36/16. Finally, replacing the linear chain Dab-Thr-Dab sequence with Gly-Dab-Thr in 41 was also detrimental to antibacterial activity.

For alterations within the heptapeptide ring system (Table ), replacement of d-Phe/d-Leu-6 or Leu-7 with the more lipophilic residues dl-Ade or Bip/d-Bip (27–30) replicated the results of Velkov et al.,[56] where these substitutions restored activity against polymyxin-resistant strains. However, these changes also caused increased cytotoxicity against HepG2 cell lines. The effects of altering the ring Dab residues varied significantly depending on which Dab was altered. Replacement of Dab-5 with homoserine (Hse), Arg, or 4′-amino-Phe (33–35) or acylation with Arg or Glu (31, 32), almost completely abolished activity (16 to >32 μg/mL against all tested strains). The same modifications at Dab-9 (18–21) gave more variable results, retaining moderate levels of activity (4–16 μg/mL for most strains). The Dab-8 position was most tolerant of alterations, with even neutral Hse 24 and zwitterionic Dab(Glu) 23 retaining good activity (1–8 μg/mL for most strains). Unfortunately, the hypothesis that a zwitterionic substituent might overcome Lipid A resistance modifications was not correct, with all of the Dab(Glu) substitutions inactive against the polymyxin-resistant strains. The Dab-8 replacement with Arg-8 26 (Table ) had the best activity of all the analogues made and showed some activity against the polymyxin-resistant strains. While not as potent as the Arg-8 26 analogue, the 4′-amino-Phe-8 analogue 25 also showed good activity; this is an interesting result due to the aromatic nature of the replacement for the otherwise aliphatic amino acid Dab. Further investigation to elaborate on the SAR trends observed with these type of compounds is in progress.

Nephrotoxicity

Traditional cytotoxicity assays have been confirmed as improper indicators for nephrotoxicity screening.[63,64] Therefore, in addition to general cytotoxicity screening assessed using HepG2 cells, we selected four compounds with promising antimicrobial potency and reasonable cytotoxicity against the HepG2 cell lines for further assessment for toxicity using primary kidney cells freshly isolated from human kidneys (see Table ). This assay, measuring LDH and GGT release, has been reported to be a better in vitro model for predicting in vivo nephrotoxicity.[63−65] We also assessed cytotoxicity against immortalized HEK293 cells. Colistin, polymyxin B, and gentamicin were employed as control nephrotoxic compounds and clearly showed the greatest toxicity in both primary renal cell assay readouts (Figure A,B), in contrast to their apparent lack of cytotoxicity in the HepG2/HEK293 assays. Compounds 10 and 14, with nonalkyl fatty acids, had shown greater cytotoxicity than polymyxin B and colistin, but induced almost no release of LDH or GGT at 100 μg/mL in the primary kidney cells. Compound 38 was the least cytotoxic (HepG2/HEK293 assays) of the four new analogues tested, but possessed intermediate levels of “nephrotoxicity” against primary kidney cells, as did 15 which had similar cytotoxicity against HepG2 cells as 10 and 14. The results support our previous assessment of in vitro predictors of nephrotoxicity, in that there is little correlation between general cytotoxicity measurements and the human kidney results.[63,64]

Table 4

Summary of MIC, Cytotoxicity CC50 Data (HepG2 and HEK293 Cell Lines) and Nephrotoxicity CC50 Data Measuring LDH (A) and GGT (B) Release from Primary Human Kidney Cells Using Lead Compounds (10, 14, 15, and 38) Compared to Control Compounds (Colistin, Polymyxin B, and Gentamicin)

sample	K. pneumoniae ATCC 13883	K. pneumoniae ATCC 700603, MDR	K. pneumoniae ATTC BAA-2146, NDM-1 pos	HepG2 CC50 (μM)	HEK293 CC50 (μM)	LDH release CC50 (μg/mL)	GGT release CC50 (μg/mL)
gentamicin	0.25	8	>32	>300	>300	33	33
1 polymyxin B	1	0.5	0.25	>300	>300	23	177
2 colistin	0.25	0.25	0.25	>300	>300	25	48
10	0.5	1	0.5	298	297	>128	>128
14	0.5	0.5	0.5	287	296	>128	>128
15	2	2	1	316	>300	>128	>128
38	0.5	1	0.5	>300	>300	122	>128

Figure 2

In vitro nephrotoxicity studies measuring LDH (A) and GGT (B) release from primary human kidney cells using lead compounds (10, 14, 15, and 38) compared to control compounds (colistin, polymyxin B, and gentamicin).

Conclusions

A new solid-phase synthesis of the polymyxin class of antibiotics has been developed based on attachment to the resin through the side chain of Thr-10, allowing for on-resin cyclization, variation at any other position, and selective derivatization of the side chains of individual Dab residues. Thirty-two analogues were prepared and tested against a panel of reference and clinical isolates, including polymyxin-resistant MDR strains. Cytotoxicity screening was performed against the HepG2 cell line. The results confirmed that more lipophilic groups at P6/P7 improved activity against polymyxin-resistant strains but increased cytotoxicity against mammalian cells. Similar effects were observed by replacement of the alkyl fatty acid with biaryl substituents. Further testing of these compounds by assessing cytotoxicity in primary human renal cells as a potential surrogate for nephrotoxicity conversely found reduced toxicity compared to the increased general cytotoxicity seen in an immortalized cell line. Large differences were found when varying the five free side-chain Dab residues: Dab-5 could not be replaced or altered without loss of activity, while alterations at Dab-9 also severely affected potency, though to a slightly less extent. In contrast Dab-8 tolerated a range of basic, neutral, or zwitterionic replacements. Similarly, Dab-3 could be replaced by nonbasic Gly or Ser with minimal effects, while inversion of the configuration of Dab-1 abolished activity. Further studies of this class of compounds will investigate combinations of promising substituent alterations, particularly those that reduce the overall basic character, improve activity against polymyxin-resistant bacteria and demonstrate reduced potential for nephrotoxicity in human kidney cells.

Experimental Section

Synthetic Materials and Methods

Experimental procedures are described in the Supporting Information. All chemicals were obtained from commercial suppliers and used without further purification. LC–MS analyses were conducted using Agilent Technologies 1200 Series Instrument with a G1316A variable wavelength detector set at λ = 210 nm, 1200 Series ELSD, 6110 quadrupole ESI–MS, using an Agilent Eclipse XDB-Phenyl column (3 × 100 mm, 3.5 μm particle size, flow rate 1 mL/min, the mobile phases 0.05% formic acid in water and 0.05% formic acid in acetonitrile). Compound purification was done using a Agilent 1260 Infinity Preparative HPLC with a G1365D multiple wavelength detector set at λ = 210 nm and an Agilent Eclipse XDB-Phenyl column 21.2 × 100 mm, 5 μm particle size. Ultraviolet/visible spectra were recorded with a Varian Cary 50 Bio spectrophotometer. Identities of final products were confirmed by MS/MS spectra, obtained using an API QSTAR Pulsar Hybrid LC–MS/MS System, high resolution mass spectrometry (HRMS), performed on a Bruker Micro TOF mass spectrometer using (+)-ESI calibrated to sodium formate, and by 1H (600 MHz) and 2D NMR spectra, obtained using a Bruker Avance-600 spectrometer equipped with a TXI cryoprobe in D2O, referenced externally with NaOAc (δH 1.90 and 8.44; 10 mg in 500 μL D2O) and then internally with the HDO resonance at δ 4.77.[66] Final purity of more than 95% for all compounds was confirmed by LC–MS analysis using both ELSD and UV (210 nm) detection.

Minimum Inhibitory Concentration (MIC) Determination

Bacteria were either obtained from American Type Culture Collection (ATCC; Manassas, VA, USA) or independent academic clinical isolate collections, as listed in Table S2. Bacteria were cultured in nutrient broth (NB; Bacto Laboratories, catalog no. 234000) or Muller Hinton broth (MHB; Bacto Laboratories, catalog no. 211443) at 37 °C overnight with shaking (∼180 rpm). A sample of each culture was then diluted 50-fold in fresh MHB and incubated at 37 °C for 1.5–3 h with shaking (∼180 rpm). Compound stock solutions were prepared as 0.64 or 2.56 mg/mL in water. The compounds, at twice the final desired concentration, were serially diluted 2-fold across the wells of 96-well plates (Polystyrene, Corning, catalogue No. 3370). Mid log phase bacterial cultures (after 1.5–3 h incubation) were diluted to a final concentration of 5 × 105 colony forming units (CFU)/mL, and 50 μL was added to each well giving a final compound concentration range of 32 μg/mL to 0.015 μg/mL (128–0.06 μg/mL for colistin and polymyxin B against the polymyxin resistant clinical isolates). MICs were determined visually after 18 h of incubation at 37 °C, with the MIC defined as the lowest compound concentration at which no bacterial growth was visible.

Cytotoxicity Assay (Resazurin Assay)

Cytotoxicity to HEK293 and HepG2 cells was determined using the resazurin assay.[67,68] In brief, HEK293 and HepG2 cells were seeded at 4000 cells/well in black-wall, clear-bottom 384-well plates (Corning, Australia) and incubated for 24 h at 37 °C, 5% CO2. Compounds were then added into each well. After 24 h incubation, 5 μM resazurin was added per well and incubated at 37 °C for 2 h. Fluorescence intensity was then read using a Polarstar Omega with excitation/emission 560/590. The data were then analyzed using GraphPad Prism 6 software. CC50 values were determined using GraphPad Prism 6 software (Inc. La Jolla, CA).

Nephrotoxicity Assay (Lactate Dehydrogenase (LDH) and Gamma Glutamyl Transferase (GGT) Release in Primary Cells Freshly Isolated from Human Kidneys)

Primary human renal proximal tubule epithelial (hPT) cells were isolated and used for in vitro prediction of drug-induced nephrotoxicity[63,64,69] by measuring the release of metabolism enzymes, LDH and GGT, as previously described.[70]