Recent Advances in the Development of Polymyxin Antibiotics: 2010–2023

The polymyxins are nonribosomal lipopeptides produced by Paenibacillus polymyxa and are potent antibiotics with activity specifically directed against Gram-negative bacteria. While the clinical use of polymyxins has historically been limited due to their toxicity, their use is on the rise given the lack of alternative treatment options for infections due to multidrug resistant Gram-negative pathogens. The Gram-negative specificity of the polymyxins is due to their ability to target lipid A, the membrane embedded LPS anchor that decorates the cell surface of Gram-negative bacteria. Notably, the mechanisms responsible for polymyxin toxicity, and in particular their nephrotoxicity, are only partially understood with most insights coming from studies carried out in the past decade. In parallel, many synthetic and semisynthetic polymyxin analogues have been developed in recent years in an attempt to mitigate the nephrotoxicity of the natural products. Despite these efforts, to date, no polymyxin analogues have gained clinical approval. This may soon change, however, as at the moment there are three novel polymyxin analogues in clinical trials. In this context, this review provides an update of the most recent insights with regard to the structure–activity relationships and nephrotoxicity of new polymyxin variants reported since 2010. We also discuss advances in the synthetic methods used to generate new polymyxin analogues, both via total synthesis and semisynthesis.


ABSTRACT:
The polymyxins are nonribosomal lipopeptides produced by Paenibacillus polymyxa and are potent antibiotics with activity specifically directed against Gram-negative bacteria.While the clinical use of polymyxins has historically been limited due to their toxicity, their use is on the rise given the lack of alternative treatment options for infections due to multidrug resistant Gram-negative pathogens.The Gram-negative specificity of the polymyxins is due to their ability to target lipid A, the membrane embedded LPS anchor that decorates the cell surface of Gram-negative bacteria.Notably, the mechanisms responsible for polymyxin toxicity, and in particular their nephrotoxicity, are only partially understood with most insights coming from studies carried out in the past decade.In parallel, many synthetic and semisynthetic polymyxin analogues have been developed in recent years in an attempt to mitigate the nephrotoxicity of the natural products.Despite these efforts, to date, no polymyxin analogues have gained clinical approval.This may soon change, however, as at the moment there are three novel polymyxin analogues in clinical trials.In this context, this review provides an update of the most recent insights with regard to the structure−activity relationships and nephrotoxicity of new polymyxin variants reported since 2010.We also discuss advances in the synthetic methods used to generate new polymyxin analogues, both via total synthesis and semisynthesis.

■ INTRODUCTION: POLYMYXINS, CLASSIC BUT CRUCIAL ANTIBIOTICS
Antibiotics are a cornerstone of modern healthcare but their viability is threatened significantly by the emergence of antimicrobial resistance (AMR). 1 This problem is particularly noteworthy in hospital settings where the intense use of antibiotics is increasingly giving rise to nosocomial infections.In 2019, 1.27 million people died globally as a direct consequence of AMR, while an additional 3.68 million deaths were associated with AMR. 2 Among the pathogens contributing to these mortality rates are antibiotic resistant isolates of the Gramnegative species K. pneumoniae, A. baumannii, P. aeruginosa, and E. coli. 2 While integrated health approaches such as monitoring and stewardship can mitigate the infectious potential of such drug-resistant pathogens, it is clear that new antibiotics are required to treat the infections they cause.The development of entirely new antibiotics operating via unexploited mechanisms requires major investment and carries significant risk for drug makers which has led to their reduced enthusiasm for such campaigns in recent decades. 3,4As an alternative, the improvement of clinically used natural product antibiotics has, and continues to be, a proven and fruitful strategy in antibiotic discovery that is likely to continue to be of value in addressing the increasing antibiotics crisis. 5,6e polymyxins are a class of natural product lipopeptide antibiotics that specifically prevent the growth of Gram-negative bacteria.−9 LPS decorates most Gram-negative bacteria on the surface of their outer membrane (OM) and is responsible for many of the systemic toxic and inflammatory effects associated with infections caused by Gram-negative bacteria. 10,11The polymyxins were discovered and brought to the clinic in the so-called golden era of antibiotic discovery (1940−1960). 12,13Many antibiotics discovered in this era, including the polymyxins, subsequently fell out of favor as newer generations of antibiotics with improved efficacy and tolerability profiles were developed.However, with the advent and spread of multidrug resistant (MDR) bacteria, classic antibiotics like the polymyxins face renewed interest.While the toxicities of the polymyxins present a clear drawback to their clinical application, their potent activity against numerous Gram-negative bacteria continues to make them an important last-resort option.Many medicinal chemistry campaigns have been conducted in an attempt to reduce the toxicity of the polymyxins while preserving their potent antibacterial activity.Also of note are recent studies with the octapeptins, a family of lipopeptide antibiotics isolated from B. circulans with structural similarity to polymyxins. 14However, compared to the polymyxins, significantly fewer medicinal chemistry studies have been reported in pursuit of octapeptin analogues with improved properties and as such they are not covered here.In this review we focus on the medicinal chemistry efforts associated with accessing next-generation polymyxins reported since 2010.−17 ■ POLYMYXINS: STRUCTURAL FEATURES, TOXICITY,

AND MECHANISMS OF ACTION
Polymyxins from Natural Sources.−23 Among all known polymyxins, the only clinically used agents are polymyxin B and polymyxin E, with the latter also commonly referred to as colistin. 24More recently, genome mining has gained popularity as a means of prospecting for natural products, which led to the identification of macolacin, a polymyxin produced by Paenibacillus xylanexedens. 25Interestingly, the amino acid composition of macolacin is identical with that of polymyxin F which was first reported in the 1970s but for which a full structure was not determined. 22,26ll polymyxins are composed of a peptide macrocycle consisting of seven residues connected to an exocyclic tripeptide that is N-acetylated with a lipid tail (Figure 1A).In general, amino acid numbering in the polymyxin class follows the formalism indicated in Figure 1A.Throughout the text of this review when reference is made to a specific residue position it will be indicated by "Px", with "x" denoting the indicated position.Common to all polymyxin variants is the presence of the nonproteinogenic residue diaminobutyric acid (Dab) at positions 1, 4, 5, 8, and 9.The side chain of Dab4 is connected with the polymyxin C-terminal carboxyl group of Thr10 to create the heptapeptide macrocycle.The other four Dab residues are protonated at physiological pH giving the polymyxins their polycationic character.With the exception of these five Dab residues and Thr2, the other amino acids in the polymyxin scaffold are found to vary among the naturally occurring family members.The P6 position is typically filled with a nonpolar D-amino acid (generally D-Leu or D-Phe).Interestingly, the configuration at P3 is variable as either L-or Dstereochemistry can be encountered at this position among the naturally occurring polymyxins.−23 Typically, fermentative production yields an ensemble of polymyxin isomers usually owing to the incorporation of these different lipid tails. 27As also illustrated in Figure 1A, diversity in the peptide sequence is observed, either as a result of the presence of a specific Biosynthetic Gene Cluster (BGC) in the producing host, or as a consequence of substrate flexibility of the enzymes involved in Nonribosomal Peptide Synthesis (NRPS). 28t is also well documented that while truncated polymyxin derivatives, most notably polymyxin B nonapeptide (PMBN), are much less active than the parent compound, they maintain the ability to disrupt the Gram-negative OM. 29−33 In this way PMBN and related polymyxin derivatives have been widely explored as synergists capable of enhancing the activity of antibiotics that otherwise cannot traverse the OM.It is also noteworthy that PMBN is significantly less nephrotoxic than the parent compound, providing some insights into the structural features that contribute to the undesired effects associated with the polymyxins. 34,35Given that the focus of this review is on polymyxins with direct antibiotic activity and not on polymyxinbased synergists, we direct the reader to another recently published review that provides an up to date view of OMdisrupting synergists. 32olymyxin Toxicity.−42 From an early stage in the their clinical history, serious side effects were noted upon administering polymyxins. 43Upon systemic administration, polymyxins accumulate primarily in the proximal tubule (PT) cells of the renal cortex, as visualized by exposing kidney tissue from polymyxin treated mice to polymyxin specific IgM antibodies. 44Polymyxins have also been shown to accumulate in rat PT cells (NRK-52E) and human PT cells (HK-2), based on the detection of a dansyl-labeled polymyxin analogue.Using this approach, it was determined that the concentration of polymyxin found inside PT cells was 1000-fold higher than the extracellular concentration. 45echanistically, uptake of polymyxins by the kidney occurs primarily via the apical membrane of proximal tubule epithelial cells.The uptake is saturable, suggesting that a transporter is involved. 46Several studies have shown that megalin (also called LRP2), an endocytic receptor 47 involved in the uptake of polybasic drugs, 48 plays a role in the uptake of polymyxins into renal cells. 46,49,50−53 Recent work by Li and co-workers also suggests that the inwardly rectifying potassium channels Kir4.2 and Kir5.1 may function as transmembrane proteins that mediate the uptake of polymyxins in HK-2 cells. 54Notably, the interaction of these channels with polymyxins is proposed to not only allow for polymyxin uptake, but also for inward potassium flux, changes of membrane potential, and cellular depolarization. 54,55Megalin ligands like endogenous cytochrome c and cilastatin (used clinically to inhibit renal dehydropeptidase I) inhibit polymyxin uptake into renal cells.Colchicine, a microtubule polymerization inhibitor required for physiological megalin action, does the same. 49,56For OCTN2, a protective effect is seen upon supplementation with carnitine, effectively reducing the cytotoxic effects of polymyxins. 51For Kir4.2 and Kir5.1, the universal Kir inhibitor BaCl 2 and the specific Kir4.2inhibitor VUF0134992 57 were found to provide protection against polymyxin induced toxicity on HK-2 cells. 54The relative contribution of each of these uptake receptors for polymyxin toxicity remains to be elucidated.
Polymyxin localization in the nuclear region and throughout the cytoplasm has been shown both in rat (NRK-52E) and human (HK-2) PT derived cell lines. 45Association with the ER, but not with lysosomes, has also been shown by means of a similarly dansyl-labeled polymyxin analogue. 58This indicates a pathway wherein endocytosis of polymyxins is followed by endosomal escape, a process observed for many biological substances. 59Intracellular effects of polymyxin accumulation include DNA instability, as observed by double stranded breaks in HK-2 cells treated with polymyxin B. 60 This finding is in line with earlier observations that polymyxins can bind directly to dsDNA strands. 61Increased levels of reactive oxygen species (ROS) have also been found in cells incubated with polymyxins. 62,63ROS are involved in the induction of apoptosis via several pathways. 64Studies in mice have shown apoptosis in renal cells upon polymyxin treatment. 62,65Freshly isolated mouse mitochondria also show a loss of Ca 2+ and membrane potential upon incubation with colistin. 53As a consequence of the involvement of ROS and oxidative stress, many compounds have been screened for their potential to lower oxidative stress upon polymyxin treatment. 66These include antibiotics (minocycline), 67 immunosuppressants (rapamycin), 68 dietary flavonoids (rutin) 69 and curcumin, 70,71 a compound with alleged antioxidant properties. 72hile it may be possible to attenuate polymyxin toxicity by coadministration of agents capable of reducing these effects, a more direct strategy would be to address the toxicity of polymyxins themselves thus avoiding the need for combination therapy and its associated drawbacks.In this regard, derivatization of colistin to the methanesulfonate (CMS) adduct (Figure 1B) was an approach developed in the 1960s in an attempt to reduce its systemic toxicity. 73−76 Upon metabolic conversion to the parent compound in vivo, antibacterial activity is restored, making the therapeutic benefit largely dependent on PK properties.−80 This review is primarily focused on recent advances in the development of next generation polymyxins that address the toxicity issues associated with their use.Before expanding on this topic, we will also address the mechanistic underpinnings of the specific anti-Gram-negative activity of the polymyxins, including a number of recently reported insights.

Mechanisms of Action.
As is the case for many lipopeptide antibiotics, the mechanism underscoring the antibacterial activity of the polymyxins includes several aspects, many of which have been covered in previous reviews. 15,16,81,82−86 A basic understanding of the polymyxin mechanism of action includes an initial electrostatic interaction between the cationic Dab residues of the polymyxin and the anionic phosphate head groups of lipids in the bacterial OM, typically lipid A. This association results in the displacement of Mg 2+ and Ca 2+ ions that serve to bridge the phosphate headgroup of adjacent lipid A molecules. 16Upon lipid A binding, the polymyxin molecule is then believed to adopt a conformation that separates its polar and nonpolar features.In doing so, the polymyxin's acyl tail and hydrophobic Leu-D-Phe motif fold toward the lipophilic lipid A tails, while the more polar side chains point toward the sugar and phosphate regions of lipid A. 87,88 Given that LPS/lipid A is highly abundant in the OM, much attention has been focused on this initial interaction as the driving force behind the activity of the polymyxins.However, a variety of LPS precursors, including lipid A variants, are also found in the inner membrane (IM). 11−91 Edwards and co-workers recently investigated the impact of colistin on the cytoplasmic membrane of Gram-negative bacteria. 90,92pecifically, by using spheroplasts derived from E. coli cells, they could assess the effects of colistin on the cytoplasmic membrane isolated from the OM.In addition, studies with polymyxin resistant strains wherein resistance is conferred by mcr plasmids encoding phosphoethanolamine transferases that modify the structure of lipid A, have provided new insights into the relative importance of polymyxin interactions with the OM and IM. 90 Notably, among such resistant strains a significantly larger portion of LPS in the IM was found to be modified by ethanolamine compared to LPS in the OM.When treated with colistin, the OM was still readily permeabilized, but not the IM.This implies that mcr positive strains are resistant toward colistin primarily by virtue of ethanolamine modified LPS in their IM, not in their OM.
In keeping with these findings, the Edwards group also showed that treatment with murepavadin, a P. aeruginosa specific LptD inhibitor, 93 inhibited LPS transport to the OM. 90Upon coadministration of murepavadin and colistin on P. aeruginosa, a synergistic effect was observed.The authors ascribed this effect to the capacity of murepavadin to interfere with the transport of LPS to the OM, leading to an accumulation of LPS in the IM, and in doing so enhancing the IM-specific effect of colistin.Notably, Kahne and co-workers had previously reported that novobiocin analogues that stimulate LptB, one of the proteins involved in LPS transport from the IM to the OM, can also induce more efficient killing by polymyxin. 94It has been postulated that increased LPS transport will lead to enhanced LPS synthesis, yielding increased levels both in the IM and OM. 90 This hypothesis may reconcile the observations of both the Kahne and Edwards groups as it seems that interfering with LPS transport (whether via LptD or LptB) can potentiate the bactericidal action of polymyxins.Supramolecular Effects.Recently, Hiller and co-workers published a study describing the interaction of colistin with outer membrane vesicles (OMVs) derived from both polymyxin-susceptible and -resistant strains. 95This revealed a clear correlation between the efficacy of the antibiotic and the resulting macromolecular features observed on the OMV surface by use of atomic force microscopy (AFM).Upon adding colistin to OMVs derived from polymyxin-sensitive strains, a clear hexagonal or honeycomb-like pattern was observed by AFM imaging, suggesting an ordered rearrangement of OMV associated molecules.This effect was found to be independent of OM proteins and highly dependent on the presence of divalent cations.When OMVs containing truncated LPS were used, the same hexagonal pattern was observed, in line with our group's assessment of the activity of polymyxins on strains bearing truncated LPS variants. 96Interestingly, a polymyxin analogue (D-Thr at P10) with reduced activity (MIC: 32 μg/mL) elicited a similar hexagonal pattern, while a much less active analogue (azido-Dab at P9, MIC: 128 μg/mL) did not.The authors further note that induction of the hexagonal crystalline pattern is associated with lateral membrane expansion, explaining previously observed protrusions in bacterial membranes upon polymyxin treatment. 91In addition to the Hiller group, several others have studied polymyxin-target interactions using AFM.Notably, an increase in cell surface roughness was observed, as well as the formation of micro-and nanoclusters. 91,97The higher resolution achieved in the studies performed by Hiller and coworkers likely provides for a more detailed explanation of the phenomena observed in earlier investigations.It is also worth noting that while OMVs provide a very good representation of the OM, they do not allow the study of effects on the IM.In this light, evidence derived from studies using OMVs alone needs to be considered as providing only part of the picture, as is also the case for insights gleaned from studies using only spheroplasts.
Stereochemical Requirements.The first investigations aimed at assessing the stereochemical underpinnings of the polymyxin mechanism of action employed the enantiomeric form of polymyxin B nonapeptide, ent-PMBN, rather than the full-length natural product. 98In these studies, the Fridkin group prepared ent-PMBN by total synthesis and somewhat surprisingly found that its binding to LPS appeared to be similar to PMBN.These investigations relied on a displacement assay, in which dansyl-PMBN, bound to LPS, was shown to be readily displaced by ent-PMBN.In support of this finding, binding studies recently performed in our group using isothermal titration calorimetry (ITC) revealed identical LPS binding patterns for PMBN and ent-PMBN. 96Interestingly, however, was the observation that ent-PMBN does not potentiate the activity of novobiocin against E. coli, while PMBN does so strongly.This finding led Fridkin and co-workers to conclude that the association of PMBN or ent-PMBN with LPS has relatively low stereochemical requirements, whereas the functional association leading to synergy with novobiocin requires the native stereoisomeric species. 98 recent study by the Reymond group also provided insights into the stereochemical requirements for polymyxin activity by means of stereorandomization, wherein racemic amino acids were introduced into the polymyxin sequence. 99Somewhat surprisingly, the mixture that resulted from full stereorandomization of polymyxin B (yielding a theoretical mixture of 1024 possible diastereomers), still showed activity on E. coli (MIC: 2 μg/mL).Partial stereorandomization, either at positions 1, 3, 4, 7, 9, 10 (64 possible diastereomers) or at positions 2, 3, 4, 9, 10 (32 possible diastereomers) also yielded active mixtures, with the latter achieving an MIC of 0.25 μg/mL on E. coli and 2 μg/ mL on A. baumannii.It should be noted that the relative amount of each possible diastereomer within these mixtures was not quantified, raising the possibility that some are overrepresented, influencing the activity of the mixture.Interestingly, the mixtures of diastereomers were found to be less efficient in LPS binding, as judged by the level of TNF-α expression in RAW264.7 macrophages upon incubation with LPS and the compound mixtures.In addition, no membrane disruption was observed in bacterial cells treated with the diastereomeric mixtures, whereas polymyxin B showed significant membrane disruption.
To more specifically study the stereochemical requirements of the polymyxin mechanism, we recently reported the first total synthesis of the unnatural enantiomer of full-length polymyxin B 4 , ent-PmxB 4 , 96 as well as the previously reported ent-PMBN. 98hen tested against a panel of Gram-negative bacteria we observed essentially no activity for ent-PmxB 4 in MIC assays (MICs ≥ 128 μg/mL).This is in notable contrast to the activity reported for the stereorandomized mixtures of polymyxin B diastereomers described by Reymond and co-workers. 99The above-mentioned findings clearly indicate that while some (combinations of) residues are amenable to stereochemical inversion, doing so for all amino acids is not tolerated.Also of note was our finding that despite lacking inherent activity, ent-PmxB 4 does effectively synergize with rifampicin against E. coli ATCC 25922 (FICI: 0.08).When assessing OM permeabilization by means of NPN assays, we also found that PmxB 4 and PMBN, with natural stereochemistry, were more effective than their corresponding enantiomers.Interesting, however, was the finding that for mcr-1 positive E. coli, the NPN uptake is rather similar for PMBN and ent-PmxB 4 , indicating that for polymyxin resistant strains, the addition of a lipophilic tail to the unnatural enantiomer can compensate for the otherwise reduced interactions with LPS.In the final analysis, it is evident from our studies and those of others, that stereochemical considerations play a key role in the mechanisms and activities of the polymyxins.

WITH REDUCED NEPHROTOXICITY
Given the dose-limiting toxicity associated with the clinical use of polymyxins, much effort has been spent in trying to develop analogues with improved therapeutic indices.In recent years a small number of new polymyxins have entered clinical trials while other analogues also offer insights in the nephrotoxicity of this class of compounds.Here, we summarize these developments in three distinct subsets including: (1) new polymyxins currently undergoing clinical development; (2) recent polymyxin analogues reported with accompanying antibacterial activity and nephrotoxicity data but not (yet) in clinical development; and (3) novel analogues for which structures have recently been disclosed but without extensive antibacterial activity or nephrotoxicity data.For reasons of convenience, we refer to the compounds according to their original codes or numbers used in the source literature cited.To provide the reader with a concise overview of the various methods in the synthesis of the different polymyxins here covered, we include summary Table 1 at the end of the review.Also, to allow for convenient comparison, we provide a tabular summary of the antibacterial activities and associated toxicities for all polymyxin analogues here covered (Table 2).

POLYMYXIN ANALOGUES IN CLINICAL DEVELOPMENT
1.1.SPR206.Spero Therapeutics is currently pursuing clinical development of its lead polymyxin candidate SPR206 which was selected from a series of novel polymyxin analogues bearing structure variations in the exocyclic moiety, specifically modifications of the P1 and P3 residues and N-terminal acyl group (Scheme 1). 100 These analogues were prepared via an established semisynthetic route wherein all five amine groups of polymyxin B were first Boc protected followed by enzymatic degradation using Savinase to yield the corresponding tri-Boc protected polymyxin B heptapeptide PMBH(Boc) 3 . 101Subsequent coupling with the dipeptide building block Cbz-Thr(tBu)-Dap(Boc), followed by selective removal of the Cbz group, provided an intermediate that was then further diversified by addition of different acyl groups followed by Boc group removal.Notable among these polymyxin analogues is the omission of the P1 residue.Rather, the Spero team found that the amine normally present in the side chain of the Dab residue at P1 could instead be incorporated into the acyl group appended to the N-terminus leading to the identification of compound 22 as a particularly potent analogue that also exhibited reduced in vitro kidney cell toxicity against HK-2 cells.However, the reduced cell toxicity of 22 did not translate to a reduced nephrotoxicity when assessed in vivo in mice.By analyzing the accumulation of the compound in the kidney, it was found that the kidney exposure of compound 22 in fact far exceeded that of polymyxin B.
In an attempt to address this liability, the Spero team next explored the effect of shifting the position of the free-amine present in the N-terminal acyl group.This led to analogues 40 and 44, which displayed an improved balance of potency, toxicity and kidney exposure.In particular, compound 44 (now known as SPR206) was found to possess antibacterial activity and kidney exposure on par with that of polymyxin B, while exhibiting a 12-fold lower cytotoxicity toward HK-2 cells.
When tested against a panel of bacteria (200 strains), SPR206 was found to be broadly active, consistently displaying MIC 50 values below 1 μg/mL against all Gram-negative members of the ESKAPE family.In general, SPR206 outperformed polymyxin B and colistin, with MIC values 2-to 4-fold lower.Against a subset of A. baumannii strains (33 strains, including drug resistant isolates), an impressive MIC 90 of 0.125 μg/mL was reported. 102n a neutropenic murine thigh infection model, SPR206 exhibited efficacy on par with that of polymyxin B, resulting in a 7-log 10 (CFU/mL) reduction in viable bacteria when dosed at 0.86 mg/kg.Furthermore, in a neutropenic murine pneumonia model, SPR206 again outperformed polymyxin B, showing a 2log 10 (CFU/mL) reduction of viable A. baumannii NCTC 13301 at a concentration where polymyxin B showed no effect (17.2 mg/kg).Spero recently published the results of a Phase I clinical trial with SPR206. 103When tested in a population of healthy volunteers, no evidence of nephrotoxicity was observed upon dosing at 100 mg q8h for 14 days.These encouraging results lay the foundation for a Phase II trial, currently in the planning stages, wherein the capacity for SPR206 to treat multidrug resistant (MDR) Gram-negative bacterial infections, specifically hospital-acquired and ventilator-associated bacterial pneumonia (HABP/VABP), will be evaluated. 104.2.QPX9003.The groups of Li and Velkov recently reported the development of a novel synthetic polymyxin analogue with promising safety and efficacy against lung infections caused by Gram-negative pathogens. 55Originally assigned the compound code F365, this polymyxin analogue was later renamed as QPX9003 by QPEX Biopharma (recently acquired by Shionogi) who licensed it for clinical development.QPX9003 resulted from a focused SAR study which investigated the effect of modifications of the N-terminal hydrophobic tail, the P3 residue, and the hydrophobic patch associated with the P6/P7 residues.Substitution of residues present in the polymyxin macrocycle preclude semisynthetic approaches and as such QPX9003 and the other analogues prepared were fully assembled by solid phase peptide synthesis (SPPS), starting with Fmoc-Thr(tBu)-OH loaded onto CTC resin (Scheme 2). 105he side chain of Dab at P4 was protected with an orthogonal ivDde protecting group, which is conveniently removed on resin by treatment with hydrazine solution after assembly of the complete linear polymyxin precursor.Subsequent cleavage from the resin using mild acidic condition to maintain all other side chain protection was followed by solution phase cyclization with DPPA/DIPEA.The final step involved global deprotection and HPLC purification to yield the desired polymyxin species.
Antibacterial assays with the polymyxin analogues generated using this approach revealed that amino acids at P6 and P7 could be exchanged for less hydrophobic residues (e.g., by changing Phe or Leu for Val or Abu), with analogues containing the D- Leu-Abu motif at P6/P7 maintaining the activity of polymyxin B. Based on previous reports describing the benefit of modification at P3, 34,100 the Dab residue at this position was replaced by Dap, to give analogues with increased potency and decreased toxicity.Finally, the N-terminal acyl moiety was optimized starting with saturated linear lipids as in analogue F287 bearing an octanoic acid tail found typically in polymyxins.F287 was found to exhibit potent activity, with lower MICs than polymyxin and it was proposed to be less nephrotoxic as a consequence of reduced kidney cell membrane penetration, as predicted by MD approaches. 106However, the maximum tolerated dose (MTD) for F287 was found to be rather low and similar to polymyxin.Analogue F342, with a shorter C6 lipid, was shown to be better tolerated (by ca.2.5-fold), but exhibited reduced activity against A. baumannii strains.This led to evaluation of a new series of analogues including F360, F365, and F477, all bearing alternative aromatic acyl groups, given that similar N-terminal moieties had previously shown benefit in other polymyxin optimization programs. 34,100,107Among the analogues thus prepared, the 2,4-dichloro benzoyl substituted variant (F365/QPX9003) was found to possess the best balance of activity with low acute toxicity.Interestingly, QPX9003 exhibited consistently lower MIC values compared to the highly similar 2,6-dichloro benzoyl analogue F360.Overall, the ratio between MIC and MTD proved to be optimal for QPX9003 and in a blood infection model in neutropenic mice the compound was found to outperform colistin in reducing bacterial burden 4 h post treatment. 55o estimate the efficacy of QPX9003 in a lung infection model, MICs were determined in the presence of Survanta, a natural bovine lung surfactant.Sputum derived biomolecules, including surfactants, have previously been shown to negatively impact the activity of lipopeptide antibiotics. 108Notably, while the MIC values measured for polymyxin B were found to be elevated by as much as 10-fold in the presence of 10% Survanta, no such increase was observed for QPX9003.This effect can be attributed to the reduced hydrophobicity of the Abu residue present at P7 in QPX9003 as opposed to the Leu residue found at the same position in polymyxin B. 55 In vivo studies showed QPX9003 to have lower plasma protein binding and increased urinary recovery relative to polymyxin B. These findings suggest that the compound is taken up by proximal tubule epithelial cell (PTECs) to lesser extent, which might also contribute to its reduced toxicity.Furthermore, MS imaging revealed a relatively limited kidney distribution for QPX9003, primarily restricted to the renal cortex.In addition, genetic analysis on HK-2 cells showed that QPX9003 only impacted the expression patterns for 70 genes, compared to a much larger number of 1282 genes upon treatment with colistin.As shown by Li's earlier work, 54 HK-2 cells with a KNCJ16 (inwardly rectifying potassium channel Kir5.1) knockout are generally protected from the toxic effects of lipopeptides.For QPX9003, no differences are seen for the parent cell line and the knockout, suggesting that the compound does not affect the inwardly rectifying potassium channel Kir5.1. 55he development of QPX9003 represents a promising new polymyxin design strategy to access analogues with enhanced properties.Notably, reduction in overall charge does not appear to be an obligate requirement for achieving reduced nephrotoxicity.A Phase I trial run in 2022 with QPX9003 substantiated the safety of the compound, paving the way for its continued clinical development. 109.3.MRX-8.Recently, Shanghai-based MicuRx Pharmaceuticals disclosed a series of novel polymyxin analogues containing metabolically labile ester moieties in their N-terminal tail moiety (Scheme 3). 110The lead compound being taken forward for clinical development is designated as MRX-8 and while its specific structure has not been confirmed by MicuRx, based on the breadth of patent data specifically provided for analogue 12 (see Scheme 3) it seems likely that this is the structure of the compound.As noted, MRX-8 and related analogues were designed to be metabolically labile.Specifically, by incorporation of an ester linkage in the lipid tail, these polymyxins are designed to be degraded in vivo by blood plasma esterases, resulting in the formation of metabolites with reduced nephrotoxicity.The synthesis of these analogues is achieved by means of an established semisynthetic route wherein polymyxin B is first selectively degraded using the proteolytic enzyme ficin to remove the N-terminal Dab residue and lipid moiety, followed by full Boc protection to yield the PMBN(Boc) 4 building block (Scheme 3). 111Next, the Dab residue is reinstalled at P1 by coupling of the orthogonally protected Cbz-Dab(Boc)-OH, followed by hydrogenation to yield the corresponding intermediate with a free N-terminus.This protected decapeptide in turn allows for convenient diversification through coupling of various acyl groups featuring an internal ester motif. 110he analogues depicted in Scheme 3 were found to exhibit potent antibacterial activity on par with that of polymyxin B. It was subsequently shown that incubation of the compounds with plasma resulted in significant degradation of analogues 7 and 18 with only 15% and 18% remaining respectively after 4 h incubation while no degradation of polymyxin B was detected.By comparison to 7 and 18, analogue 12/MRX-8 showed greater plasma stability with 31% of the compound intact after 4 h incubation.These findings indicate that both the regiochemistry of the ester linkage as well as the flanking substituents play a key role in the stability of these compounds.The (enzymatically) hydrolyzed product (Scheme 3, denoted as compound 20 in the reporting patent) bears a hydroxyethyl motif and retains moderate activity against P. aeruginosa, E. coli, and K. pneumoniae (2−4-fold increase compared to the parent compound MRX-8, Table 2). 110hen tested on a large set (n = 1314) of clinical isolates, MRX-8 was found to have consistently high potency with MIC 90 values of 0.25, 0.25, 1.0, and 1.0 μg/mL against E. coli, K. pneumoniae, A. baumannii, and P. aeruginosa, respectively (versus 0.5, 0.5, 0.5, 1 μg/mL for polymyxin B against the same strain collection). 112Similar efficacy was also observed in a separate study recently reported. 113Based on these promising in vitro data, MRX-8 was taken forward for preclinical development with similarly encouraging results. 114At present, MicuRx Pharmaceuticals is conducting Phase I clinical trials to assess the safety, tolerability, and pharmacokinetic profile of MRX-8 when administered intravenously in healthy subjects. 115hile more than a decade has passed since the last human trials were run with a novel polymyxin, at present three new candidates are under clinical investigation.Prior to these, the last polymyxin to be evaluated in Phase I trials was Cubist Pharmaceutical's polymyxin B analogue CB-182 804, whose development was stopped in 2010 due to toxicity concerns. 116In this light, the clinical candidates currently under investigation have all been carefully scrutinized, particularly for nephrotoxicity.In the case of SPR206, the limited renal uptake is likely to be a major advantage, while QPX9003 appears to have a relatively high tolerated dose with a favorable kidney safety profile.For MRX-8, the short decomposition half-life, and the presumed lower toxicity of the resulting metabolites, provides the foundation for its enhanced therapeutic profile.The progress of these three clinical candidates will be closely watched in the coming years in the hope that they will offer physicians and their patients with safer and effective options for polymyxin therapy.

Analogues with Reduced Charge.
Based on the hypothesis that the polycationic nature of the polymyxins contributes significantly to their toxicity, researchers at the Finnish company Northern Antibiotics explored a series of polymyxin analogues containing a reduced number of cationic residues. 117These analogues were prepared by solid-phase synthesis.Common to some of the more potent analogues  identified by the Northern Antibiotics team was the presence of D-Ser rather than Dab at P3, a substitution also found in a subset of the natural polymyxins including polymyxin D, 118 S, 119 and macolacin 25 (see Figure 1A above).Further reduction of the overall cationic charge by removal of the Dab residue at P1, in combination with the D-Ser for Dab substitution at P3, gave polymyxin B analogue NAB739 (Figure 2) that was found to retain significant antibacterial activity.Against Enterobacteriaceae, NAB739 exhibits activity similar to polymyxin B, while against strains of Acinetobacter and Pseudomonas polymyxin B was found to be superior. 117Also of note was the finding that the hydroxy group of D-Ser at P3 is essential for potency, with a D-Ser to D-Abu substitution as in NAB7061 (Figure 2) resulting in a large drop in activity. 120The impact of the reduced net positive charge of these polymyxins on their nephrotoxicity was also assessed.These studies showed that they exhibit reduced binding to the brush border membrane of the rat renal cortex and also display significantly reduced toxicity against HK-2 cells (Table 2). 117,121hereas NAB739 bears all 3 cationic charges in its macrocycle and none in the exocyclic moiety, novel analogue NAB815 (prepared by total synthesis), containing D-Thr at P3, was designed to contain one positive charge in the exocyclic region and only two in the macrocycle (Figure 2). 121,122AB815 was found to have the same antibacterial activity as NAB739 (MIC of 2 μg/mL against E. coli ATCC 25922, Table 2) with both being slightly less active than polymyxin B (0.5 μg/ mL on E. coli ATCC 25922).Surprisingly, the CC 50 value determined for NAB815 with HK-2 cells (334 μg/mL) was found to be about 20-fold higher than that of polymyxin B (18 μg/mL) and 1.4 times higher compared to NAB739 (237 μg/ mL).
Also of note is the finding that the NAB analogues are effectively secreted via the urine, a property not observed for most natural polymyxins. 122This feature was hypothesized to be a consequence of their lower uptake by PTECs resulting in a reduced accumulation in the kidneys.The accompanying increase in renal secretion might also qualify these compounds for treatment of urinary tract infections.Despite their higher in vitro MICs relative to polymyxin B, both NAB739 and NAB815 demonstrated superior performance in a pyelonephritis infection model with E. coli, requiring only about 10% of the amount of polymyxin B to achieve the same effect. 123.2.Analogues Containing P3 Substitutions and N-Terminal Biaryl Moieties.Researchers at Pfizer employed a total synthesis approach to generate a series of novel polymyxin B analogues with various amino acids at P3 along with the introduction of different biaryl moieties at the N-terminus conjugated via amide, urea, or sulfonamide linkages.34 These studies led to the important finding that replacement of the P3 Dab residue normally found in polymyxin B with the shorter Dap resulted in both increased antimicrobial activity as well as a >2-fold reduced toxicity toward PTECs.Subsequent screening of various biaryl moieties at the N-terminus resulted in a panel of novel analogues with potency equivalent to polymyxin B. While many of these analogues also exhibited CC 50 values similar to polymyxin B against PTECs, a subset was found to display >2fold reduced toxicity, including compounds 5r, 5t, 5u, and 5x (Figure 3, Table 2).Among these, compound 5x was identified as the clear favorite, displaying a CC 50 value >5-fold higher than that of polymyxin B. Compound 5x was also found to be equipotent to polymyxin B, with IC 90 values of 2 μg/mL against E. coli, P. aeruginosa, and A. baumannii, while against K. pneumoniae its IC 90 of 1 μg/mL was lower than that of polymyxin B (2 μg/mL).Also noteworthy is the activity of 5x against drug-resistant A. baumannii isolates where its IC 90 value was found to be 16 μg/mL, compared to >64 μg/mL for polymyxin B. Disappointingly, while 5x showed reduced toxicity toward PTECs and in exploratory rat studies, results from dog studies did not demonstrate superiority to polymyxin, leading to the termination of its development.

Analogues Assessing Substitution of All Positions and N-Terminal
Acyl Group Variation.The group of Blaskovich and Cooper reported a comprehensive polymyxin structure−activity study assessing the contribution of each variable amino acid position (excluding Dab at P4 and Thr at P10 as they are involved in ring closure) and variations of the Nterminal acyl moiety. 124Using a total synthesis approach, a large number of new polymyxin analogues were prepared.While many of the amino acid substitutions sampled resulted in reduced potency, some modifications to the lipophilic tail, such as the inclusion of aromatic biphenyl moieties, led to analogues that retained potent activity.In line with the findings of the Pfizer team (see previous section) the introduction of D-Ser and D-Dab at P3 were also found to be well tolerated.In addition, a novel analogue bearing Gly at P3 was found to exhibit the same activity as polymyxin B against Enterobacteriaceae strains.By comparison, P5 and P9 were found to be least amendable to modification, resulting in significant loss of activity upon the introduction of relatively minor structural changes.Notable was the effect of increased hydrophobicity of the side chains at P6 and P7 which resulted in a slight enhancement of activity against polymyxin resistant strains.These analogues, however, also exhibited decreased CC 50 values on HepG2 cells, suggesting nonbacteriospecific cytotoxicity.The residue at P8 was also found to be somewhat tolerant of modification with an Arg for Dab substitution at this position resulting in an analogue with activity comparable to polymyxin B.
Based on the results of antibacterial assays, analogues 10, 14, 15, and 38 (Figure 4, Table 2) were selected for further cytotoxicity testing on HEK293 cells which indicated that they are not more toxic than polymyxin B. In addition, Lactate Dehydrogenase (LDH) and γ-Glutaryl Transferase (GGT) release assays on primary cells revealed significantly increased CC 50 values for compounds 10, 14, 15 and 38 compared to polymyxin B. Specifically, the compounds were all found to The same group also evaluated a series of lipononapeptide polymyxin B analogues lacking the Dab at P1, prepared via a total synthesis strategy similar to that used for their previous fulllength analogues. 124,125In doing so a number of lipidated PMBN variants were generated by incorporating various Nterminal acyl moieties.Omitting the Dab at P1 along with Nterminal acylation with octanoic acid led to a 2−4-fold increase in MIC values against a panel of Gram-negative pathogens tested.The incorporation of structurally diverse acyl groups gave rise to novel analogues such as compound 12 (Figure 4) that against some strains showed antibacterial activity approaching that of the parent polymyxin.However, the most active lipidated PMBN analogues thus identified were also found to exhibit significantly higher cytotoxicity toward HK-2 cells compared to polymyxin B.

Analogues Based on
Previously Uncharacterized Natural Polymyxins.The group of You and Li undertook an extensive comparative analysis of the activity and toxicity of a panel of natural polymyxin variants including a number of previously unexplored derivatives. 126,127Of the 29 natural polymyxins studied, polymyxin T variants, containing Leu at P10, showed the weakest antibacterial activity, together with high toxicity toward Vero cells.All other natural polymyxins were found to exhibit potent activity (MICs on E. coli: ≤ 2 μg/ mL; K. pneumoniae: ≤ 2 μg/mL; A. baumannii: ≤ 2 μg/mL and P. aeruginosa: ≤ 4 μg/mL).Interestingly, the natural polymyxin B variants included in the study showed the highest toxicity toward Vero cells while the previously unreported polymyxin A 2 , D 2 , and S 2 showed the lowest toxicity.The latter three analogues contain the same branched 6-methyl heptanoyl tail, indicating that the slightly shorter lipid offers a benefit in reducing toxicity compared to polymyxin variants that bear the 6-methyl octanoyl moiety.These studies highlight the value of exploring scaffolds other than polymyxin B, as has traditionally been the case in medicinal chemistry approaches aimed at finding improved analogues.This assertion is supported by the discovery that polymyxin S 2 , containing D-Ser at P3 and Thr at P7 (Figure 5, Table 2), was found to exhibit potent antibacterial activity comparable to that of polymyxin B but with a 7-fold lower cytotoxicity.Unnatural analogues of Polymyxin S 2 were subsequently prepared to explore the impact of changes to the P6 position as for compound 14, where D-norleucine was incorporated at P6, as well as the effect of linearization of the Nterminal lipid as in analogue 16 (Figure 5).While 14 and 16 both exhibit potent antibacterial activity, they were not found to be less toxic than the parent polymyxin S 2 .Based on these findings, polymyxin S 2 was further evaluated in a murine model of systemic infection with NDM-1 positive K. pneumoniae.This revealed an effective dose (ED 50 ) of 0.9 mg/kg for polymyxin S 2 compared with 2.2 mg/kg for polymyxin B, in agreement with their respective in vitro MIC values. 127Currently under development as ASK0912, polymyxin S 2 was also recently evaluated in a murine model of A. baumannii septicemia, where it again showed an overall better therapeutic potential (lower required dose, better bacterial clearance) compared to polymyxin B. 128 2.5.Analogues Designed to Explore the Roles of Conserved Thr at P2 and P10.Huang and co-workers reported a study focused specifically on characterizing the contributions of the Thr residues found at P2 in all known natural polymyxins and at P10 in the vast majority of polymyxins. 129In doing so they relied on both semisynthesis and total synthesis approaches.These investigations revealed that the Thr at P2 is only modestly tolerant to substitution, with most P2 substituted analogues showing increased MIC values against most of the bacterial strains tested.The one exception was in the case of P. aeruginosa, where a select few P2 substitutions were found to yield polymyxin analogues that showed the same activity as polymyxin B. This effect is rather amino acid specific with analogue 32, bearing Leu at P2 (Figure 6, Table 2), maintaining potent anti-P.aeruginosa activity along  with slightly elevated MICs against the other species tested, while analogue 28 with Tyr at P2 exhibited a significantly reduced activity against all strains.In contrast, the Thr residue at P10 was found to be amenable to substitution by a broader range of amino acids (Cys, Val, Leu, Phe), while keeping good activity against both P. aeruginosa and A. baumannii (MICs < 2 μg/mL).These analogues were, however, found to be somewhat less active toward Enterobacteriaceae (MICs 2−16 μg/mL).For example, the introduction of Cys at P10 (analogue 43, Figure 6) resulted in MICs of 0.06−0.5 μg/mL against the strains of P. aeruginosa tested, while against E. coli MICs of 2−4 μg/mL were measured.Interestingly, P10 substitution with Dab (analogue 42, Figure 6) resulted in reduced antibacterial activity across the board with MICs of 8−32 μg/mL against P. aeruginosa and 8−16 μg/mL against E. coli.These findings are generally in line with those of the Li group who also recently examined the contribution of the P10 residue to antibacterial activity and toxicity. 88A systematic screening of a series of residues at P10 revealed a strong preference for Thr/Ser or hydrophobic residues at this position.While most natural polymyxins contain Thr at P10, there are some variants that instead contain Leu at this position (see Figure 1A above).Notably, the Leu at P10 can be exchanged for a number of other hydrophobic residues, e.g., Ala, Abu, Val, or Nva, without significantly impacting antibacterial activity. 88Also of note is the finding that polymyxin analogues bearing P10 substitution do no exhibit appreciably reduced toxicity, either in cell based assays (HK-2 cells) 129 or in mice (based on observed MTD). 88sing a semisynthetic approach, the Huang group also explored the impact of esterifying the hydroxyl groups of Thr at P2 and/or at P10 in polymyxin B with a variety of amino acids.Rather surprisingly, the activities of the esterified analogues thus prepared were reported to be on par with those of the parent polymyxin B species, even when derivatized with bulky amino acids such as Tyr at P2 or acidic residues like Asp at P2 and P10.Notably, analogue 76 (Figure 6, Table 2), featuring a Pro ester at P2 was found to exhibit exceptionally low MIC values, at least one serial dilution better than the parent polymyxin B 2 while also displaying reduced toxicity toward HK-2 cells.Hydrolysis of these esterified modified polymyxins was also found to occur rapidly upon of incubation at 37 °C in phosphate buffer at pH 7.0, leading the authors to suggest that analogues like 76 may in fact be operating as polymyxin B prodrugs. 129.6.Analogues via Novel Semisynthetic Approaches to Access Variation at P6.A new semisynthetic strategy to further diversity, the polymyxin B scaffold, was described by Brown and colleagues. 130This approach provides convenient access to side chain variants extending upon the P6 D-Phe residue found in the natural product.Specifically, by treating polymyxin B with N-bromosuccinimide and BF 3 , it was found that the aromatic ring of the D-Phe at P6 can be brominated at the para-position.Subsequent Boc protection of the five Dab residues followed by treatment with Savinase (as in previously reported semisynthetic strategies 96 ) provides access to the corresponding PMBH[P6-p-Br-D-Phe](Boc) 3 building block 4, (Scheme 4). 101,130Further elaboration was then achieved by introducing different amino acids at P2 and P3, installed as the appropriately protected dipeptides, followed by Suzuki crosscoupling for diversification at P6. Selective deprotection of the N-terminus and subsequent acylation with a variety of moieties, followed by global deprotection and purification, provided access to a range of novel semisynthetic polymyxin analogues (Scheme 4).Via this approach analogues such as 9a were prepared which in some cases were found to exhibit potent antibacterial activity along with slight reductions in cytotoxicity relative to polymyxin B. Also reported by the same authors is the finding that the P6 D-Phe benzene core in polymyxin B can be fully reduced using PtO 2 under H 2 atmosphere to give the corresponding D-cyclohexyl alanine analogue. 101,130This in turn provided access to analogues such as 13b, which was also found to exhibit potent antibacterial activity and reduced cytotoxicity.
2.7.Analogues Containing Disulfide Linked Peptide Macrocycles.Rabanal and co-workers reported a unique class of polymyxin analogues, wherein the peptide macrocycle is closed via a disulfide linkage. 131This approach was inspired by the somatostatin analogue lanreotide and other clinically used peptide drugs containing disulfide-linked macrocycles.In the approach developed by the Rabanal group, the L-Thr at P10 and Dab at P4 were replaced by L-Cys or D-Cys, facilitating disulfidemediated ring closure (Figure 7).In addition, the impact of side chain substitutions at P6 and P7 were also explored.Somewhat surprisingly, analogues 38 and 39 were found to have both potent anti-Gram-negative activity, on par with polymyxin B, as well as activity against S. aureus, against which polymyxin B is essentially inactive.Analogue 38, featuring a decanoic acid tail was further evaluated for acute toxicity in CD-1 mice.An LD 50 value of 283 mg/kg was obtained for 38, almost 5-fold higher compared to polymyxin B (59.5 mg/kg). 131Despite this lower acute toxicity, all disulfide analogues bearing a cysteamide motif at P10 showed nephrotoxicity patterns similar to polymyxin B. Interestingly, data disclosed in a subsequent patent filed by the same group showed that the nephrotoxicity of the disulfide analogues could be reduced by replacing the C-terminal amide moiety with a β-hydroxy group to recapitulate the Thr side chain found at P10 in most polymyxins.Representative analogue 6 contains this modification as well as replacement of the acylated P1 moiety found in natural polymyxins with D-2-amino decanoic acid (Figure 7).The results disclosed in the patent describe these analogues as possessing potent in vitro antibacterial activity as well as being less nephrotoxic than polymyxin B in vivo based on the number and severity of lesions observed by histopathological inspection of renal samples. 132Data demonstrating the in vivo efficacy of these analogues in treating bacterial infections has yet to be reported.
2.8.Analogues with Reductively Labile N-Terminal Lipid Tails.With the aim of generating less toxic analogues, our group recently developed a series of novel polymyxin B variants bearing N-terminal lipids containing a reductively labile disulfide bond as well as modifications at P1 and P3 (Figure 8). 133,134ased on previous studies indicating that the lipid tail contributes significantly to the toxic effects associated with the polymyxins, 35 we envisioned intracellular cleavage of the disulfide-linked lipid as a means of addressing this toxicity.Given the highly reducing environment found inside kidney cells, 135 we hypothesized that upon renal uptake, subsequent disulfide reduction could in turn lead to formation of less toxic and/or more rapidly cleared metabolites.In our initial series of analogues, we examined the replacement of L-Dab at P1 with L-Cys or D-Cys, which provided a thiol handle for introducing the disulfide.In doing so, the γ-amino moiety of the Dab side chain normally present at P1 is replaced by an α-amine, bringing it closer to the backbone of the peptide.We therefore also explored the impact of adding a Gly motif to this α-amine to assess the relative importance of the spacing between the primary amine at P1 and the backbone.Using a convenient semisynthetic approach making use of the previously reported PMBN(Boc) 4 , 111 we prepared a range of analogues by coupling different L-Cys-or D-Cys-based building blocks containing disulfide linked lipids of varying length and branching.
In vitro assessment of these disulfide-linked analogues revealed that the structure of the lipid plays an import role in achieving an optimal balance between antibacterial activity and toxicity toward kidney cells (PTECs).To this end, compound   18b and 23a were identified as particularly interesting analogues with antibacterial activities comparable to polymyxin B but with reductions in cytotoxicity of 4.7-and 9.5-fold, respectively (Table 2).Also of note is the importance of the stereochemistry of the P1 Cys residue in these analogues.In the case of 18b where the α-amine moiety is not modified, it is essential to have D-Cys at P1 as the corresponding L-Cys analogue exhibits significantly reduced antibacterial activity.However, when a Gly residue is coupled to the α-amine moiety of the Cys residue at P1, the stereochemical requirement is reversed.In this case, analogue 23a, containing L-Cys at P1, was found to be more potent and less nephrotoxic than the analogue containing D-Cys at the same position.
To establish the impact of the disulfide motif on the reduced toxicity observed for these analogues we synthesized and tested 45, the all-carbon analogue of 18b.While 45 was found to exhibit antibacterial activity on par with 18b, it was also found to be 2.4 times more toxic toward PTECs (CC 50 of 82 μM for 45 vs 192 μM for 18b).This would suggest that reductively labile disulfide linkage in the lipid tail does have a positive effect in reducing toxicity.We also tested the hypothesis that the disulfide linkage incorporated in these polymyxin analogues could be selectively cleaved under different reducing conditions.To this end compounds 18b and 23a were incubated with high (5 mM) and low (5 μM) glutathione concentrations meant to mimic the glutathione concentrations found inside proximal tubular cells and in blood. 135,136While 18b and 23a were both rapidly degraded under conditions mimicking the higher intracellular concentration of glutathione (undetectable after 1 h), in the low glutathione conditions the compounds were found to be 70− 80% intact after 24 h incubation at 37 °C.These findings suggest that the inclusion of disulfide-linked lipids may indeed provide a means to tune the metabolic stability of polymyxins in pursuit of analogues with reduced toxicity.Building from these findings we also recently reported a second generation of disulfide linked analogues based on 18d wherein the P3 residue was varied. 134his led to the identification of analogue 8d, bearing L-Dap at P3, as an improved variant with potent antibacterial activity and even lower cytotoxicity toward PTECs (Table 2).
2.9.Analogues Bearing N-Terminal Lipids Containing Thioether and Ester Linkages.Another class of polymyxin analogues bearing modified N-terminal lipid moieties was recently reported by the group of Harris and Brimble wherein structural variation was introduced by employing thioether linked modifications. 137In these analogues, a number of Salkylated Cys-derived building blocks were coupled to the Nterminus of the polymyxin decapeptide (Figure 9).Among the analogues thus prepared, those containing both the thioether and a hydrolytically labile ester linkage in the N-terminal lipid, similar to the polymyxins developed by MicuRx, were found to be most promising.Specifically, analogues 9g and 13c (Figure 9, Table 2) showed activity comparable to polymyxin B against a panel of Gram-negative bacteria while exhibiting reduced toxicity in an in vitro kidney organoid model.Notably, the use of organoid-based models presents a new approach to assessing polymyxin toxicity and may be of improved predictive value in recapitulating some of the spatial features found in renal tissue.
The same group also explored the effect of introducing additional S-linked lipid moieties at P6 and P7 in combination with those incorporated at the N-terminus.Representative analogue 10a includes a P7 substitution with a sulfur-derivatized Cys, alkylated with the same ester-containing lipid appended to the N-terminus (Figure 9, Table 2).While this compound was found to perform reasonably well against Enterobacteriaceae (MICs 2−4 μg/mL), it showed reduced activity against P. aeruginosa and A. baumannii.These observations, in combination with the finding that 10a is not significantly less toxic than analogue 9g, indicate that S-linked substitutions at P6 and P7 offer less benefit compared to the effects associated with N-terminal modification.Further assessment of 9g revealed that while it was relatively stable in organoid medium (ca.60% remaining after 24 h), in plasma it was rapidly degraded (9% remaining after 4 h with near complete degradation within 24 h).Whether this strategy can lead to optimized polymyxins with a desired activity-toxicity-stability balance in vivo remains an open question and will likely require further fine-tuning of the ester linkage in the N-terminal lipid moiety as for the MicuRx analogues. 110

POLYMYXIN ANALOGUES WITH ONLY ANTIBACTERIAL ACTIVITY REPORTED
3.1.Analogues with Increased Hydrophobicity.Ichikawa and co-workers reported a series of polymyxin derivatives focused on revealing the effect of increasing hydrophobicity at both the N-terminus as well as at P6 and P7. 138Using a total synthesis approach a number of novel analogues were prepared containing branched aliphatic or biaryl moieties at these positions.Antibacterial activity was found to be abolished upon incorporation of long linear (C 18 ) or branched (bis-C 8 ) lipids at the N-terminus.By comparison, some activity was retained upon incorporation of biaryl lipids at the Nterminus (MIC against E. coli of 8 μg/mL), in line with previous reports. 125However, the introduction of biaryl substituents at P6 or P7 in combination with the incorporation of similar motifs at the N-terminus led to a significant loss of activity (compounds 4d and 4e, Figure 10, Table 2).Given that an assessment of the toxicity of these analogues was not reported it is difficult to assess their actual potential.

Analogues with N-Terminal
Bis-Lipidation.Another series of polymyxins with increased N-terminal hydrophobicity were reported by the group of Schweizer wherein both the N-terminal α-amine and the side chain amine of the P1 Dab residue were acylated with fatty acids of varying hydrophobicity. 139When tested against a panel of Gram-negative pathogens, the bis-C 4 lipidated polymyxin B analogue 1 (Figure 11, Table 2) was found to have little-to-no activity (MIC ≥ 64 μg/mL on all strains) indicating that two N-terminal C 4 lipids do not recapitulate the effect of a single C 8 lipid.In keeping with this finding, bis-C 8 lipidated analogue 2 showed improved antibacterial potency, especially against the strains of P. aeruginosa tested (MICs 2−8 μg/mL).This enhanced anti-P.aeruginosa activity was also observed for analogues containing bis-biphenyl or bis-adamantyl motifs at the same positions.In search of an explanation for the observed specificity of these analogues against P. aeruginosa, the authors identified a potential role for reduced bacterial efflux.This was based on the finding that the MIC of 1 was significantly reduced (from 128 μg/mL to 2−4 μg/mL) on a P. aeruginosa strain that lacks the MexAB-oprM efflux pump.In contrast, the inherently more potent analogue 2 did not perform better on the same efflux mutant strain, leading to the conclusion that bis-lipidation with the appropriate length of lipid may serve to circumvent efflux.It is also noteworthy that these bis-lipidated polymyxins were in general found to be slightly more hemolytic than the natural polymyxins. 139In order to assess their potential more fully, further studies in relevant models of kidney injury are needed.

■ CONCLUSIONS AND OUTLOOK
The polymyxins remain a clinically important class of lipopeptide antibiotics given their potent activity against problematic Gram-negative bacteria.This activity is driven by their ability to target lipids that are abundant and exclusive to the Gram-negative OM, most notably the LPS anchor Lipid A. 7−9 To date, a range of polymyxin variants from natural sources have been identified and, using the tools of modern biotechnology and chemistry, are accessible via fermentation or total synthesis.At present, polymyxin B and colistin are the only polymyxins used clinically.−53 Potassium channels have also recently been implicated in the nephrotoxic effects of the polymyxins. 54ltimately, excessive polymyxin accumulation throughout kidney cells results in oxidative stress, in turn leading to apoptosis. 62,65iven the potent antibacterial activity of the polymyxins and their increasing use in treating infections due to multidrug resistant bacteria, a strong case can be made for developing nextgeneration analogues with improved therapeutic indices.In    bearing an N-terminal aromatic acyl group, the same Dap for Dab substitution at P3, and a slightly less hydrophobic side chain at P7. 55 Distinct from these is MicuRx's metabolically labile polymyxin B analogue MRX-8 which aims to employ a different strategy to address polymyxin's toxicity, namely in vivo detoxification by enzymatic hydrolysis. 110enefiting from decades of investigation and effort on the part of many groups, novel polymyxin analogues can today be generated by a range of robust and reliable strategies.Depending on the desired region of structural variation, robust semisynthesis and total synthesis methods offer a wide range of possibilities.Notably, the optimized and highly automated SPPS approaches recently developed provide convenient access to new polymyxins in quantities suitable for explorative studies.Whether such SPPS approaches can be viable for the much larger-scale synthesis of a clinically used next-generation polymyxin remains to be seen.With respect to the individual amino acids found in the natural polymyxins, the Dab residues at P1 or P3 are particularly amenable to modification leading to analogues with enhanced activity and/or reduced toxicity. 34,117y comparison, the Dab moieties at P5 and P9 are much less tolerant of substitution, leaving the P6/P7/P8 region of the polymyxin macrocycle as targets for structural change.Generally, variations to P6/P7/P8 are tolerated so long as the hydrophobic character of this region is maintained. 124hile significant gains have been made in generating SAR insights related to the antibacterial potency of polymyxin analogues, predicting their in vivo toxicity remains a challenging task.Whereas cellular uptake is typically accounted for in cell culture assays, long-term accumulation effects are not observable given the shorter time course of such experiments.Despite this limitation, screens using HK-2 cells or PTECs are still frequently employed in the initial estimation of polymyxin toxicity. 34,100,110,121,133As an alternative, acute toxicity in mice can be used as a proxy for toxicity in humans, recapitulating not only the nephrotoxicity but general toxicity as well. 55,88,140owever, results from animal-based assays do not offer guaranteed transferability to human medicine. 34In this regard, toxicity assessment based on the use of organoids derived from human cells may offer an additional option for assessing polymyxin toxicity, underscoring the value in further establishing such models. 137t is clear that the polymyxins will continue to be of clinical relevance in the coming decades as will interest in the pursuit of new analogues with improved safety profiles.The steadily growing wave of resistance among Gram-negative pathogens, along with the lack of new LPS/lipid A targeting modalities in the pipeline, has led to a renewed appreciation for the polymyxins as critical last resort antibiotics.In this light, the successful development of next-generation polymyxins with potent antibacterial activity and reduced toxicity will critically depend on sustained efforts firmly grounded in medicinal chemistry.

Figure 1 .
Figure 1.(A) General structure of the naturally occurring polymyxins, with residue numbering indicated and variable features highlighted in blue.*For polymyxin C, the stereochemistry of the constituent amino acids has not been reported.(B) Structure of the clinically used colistin methanesulfonate, a prodrug form of polymyxin E developed in an attempt to reduce the toxicity associated with polymyxins.

Scheme 1 .
Scheme 1. Semisynthetic Route Used by Spero Therapeutics to Access Novel Polymyxin B Analogues a

Scheme 2 .
Scheme 2. Synthesis of QPX9003 and Related Analogues a

Scheme 3 .
Scheme 3. Synthesis of N-Terminal Lipid Ester-Linked Polymyxin Analogues Developed by MicuRx Pharmaceuticals a

Figure 2 .
Figure 2. Polymyxin analogues developed by Northern Antibiotics featuring a reduced net positive charge at physiological pH.

Figure 3 .
Figure 3. Polymyxin B analogues developed by Pfizer containing P3 Dap for Dab substitution and various biaryl moieties at the N-terminus.

Figure 4 .
Figure 4. Polymyxin B analogues developed by Blaskovich and Cooper to investigate substitution at P3 and variation of the N-terminal acyl group.

Figure 5 .
Figure 5. Polymyxin S 2 analogues (containing D-Ser at P3 and Thr at P7) developed by the group of You and Li.

Figure 6 .
Figure 6.Polymyxin analogues prepared by Huang and co-workers to probe the roles of Thr residues at P2 and P10.

Scheme 4 .a
Scheme 4. Semisynthesis Route Developed by Brown and Coworkers Providing Access to P6 Modified Polymyxin Analogues a

Figure 7 .
Figure 7. Polymyxin analogues developed by Rabanal and co-workers containing a disulfide-linked macrocycle.*Analogue 6 described in the patent literature.

Figure 8 .
Figure 8. Polymyxin analogues developed by the Martin group containing reductively labile disulfide-linked N-terminal lipids.

Figure 9 .
Figure 9. Polymyxin analogues reported by Harris and Brimble bearing N-terminal lipids containing thioether and ester linkages.

Figure 10 .
Figure 10.Polymyxin analogues prepared by Ichikawa and co-workers to investigate the impact of introducing alternative hydrophobic groups at the N-terminus as well as P6 and P7.

Table 1 .
Summary of Recent (2010−) Strategies in Polymyxin Synthesis, Encompassing Both Semi-and Total Synthesis Approaches

Table 2 .
100parative Overview of Selected Polymyxin Analogues, Including Representative MIC Values and Nephrotoxicity Data recent years three new clinical candidates have been developed each with a different approach for reducing the toxicity of polymyxins.Spero pharmaceuticals' SPR206 (Scheme 1) is a P1 truncated analogue that contains an optimized N-terminal acyl group and a Dap for Dab P3 substitution.100Bycomparison, Shiongi's QPEX9003 is a full length polymyxin E analogue