Amphotericin B and Other Polyenes—Discovery, Clinical Use, Mode of Action and Drug Resistance

Although polyenes were the first broad spectrum antifungal drugs on the market, after 70 years they are still the gold standard to treat a variety of fungal infections. Polyenes such as amphotericin B have a controversial image. They are the antifungal drug class with the broadest spectrum, resistance development is still relatively rare and fungicidal properties are extensive. Yet, they come with a significant host toxicity that limits their use. Relatively recently, the mode of action of polyenes has been revised, new mechanisms of drug resistance were discovered and emergent polyene resistant species such as Candida auris entered the picture. This review provides a short description of the history and clinical use of polyenes, and focusses on the ongoing debate concerning their mode of action, the diversity of resistance mechanisms discovered to date and the most recent trends in polyene resistance development.


History of Polyenes as Antifungal Drugs
Polyenes were, after griseofulvin [1], the first fungal-specific antibiotics on the market and ever since, more than 200 polyene antifungals have been discovered [2], of which amphotericin B, nystatin and natamycin ( Figure 1) are most commonly used in antifungal therapy [3]. In 1949, microbiologist Elizabeth L. Hazen and chemist Rachel F. Brown isolated the first antifungal polyene-fungicidin-later called nystatin after New York State, where the meeting of the National Academy of Sciences in which fungicidin was presented, took place that year [4][5][6]. Nystatin was purified as a fermentation product of Streptomyces noursei, cultured from a soil sample of the farm of W. Nourses, after which the antibiotic producing actinomycete species was named [6]. Nystatin was patented by the E.R. Squibb and Sons Institute and became one of the first antimycotic drugs on the market. Early on, it became apparent, however, that nystatin had poor gastrointestinal absorption and could thus only be used to treat topical mycoses [7]. Therefore, the same research group continued their broad screening of soil-cultivated fermentation broths and in January 1953, a fermentation broth from the Streptomyces nodosus culture M4575, cultured from a soil sample of the Orinoco basin in Venezuela, showed remarkable antifungal activity [5,7]. Two active compounds were isolated: amphotericin A and amphotericin B (AmB), named after their amphoteric properties. These polyenes were chemically similar to nystatin, but the ultraviolet absorption spectrum showed additional maxima at longer wavelengths [7]. After successful

Polyene-Sterol Interactions
Overall, polyenes have an unusual mode of action compared to other antifungal drug classes, as they do not target a specific enzyme but rather interact with a vital molecule-ergosterol [1]. The first indications of the mode of action of polyenes were published in 1958, when Gottlieb et al. [30] discovered that the addition of sterols such as cholesterol, lanosterol and ergosterol could inhibit the fungicidal effect of nine polyenes including filipin, AmB and nystatin on three fungal species. They suggested that polyenes could prevent the synthesis of sterols (as the, then not yet discovered, azole, allylamine and phenylmorpholine antifungals do [1]) or competitively replace the sterols as a cofactor of an essential metabolic reaction [30]. Later, however, it became apparent that polyenes can alter the permeability of the membrane by reacting with sterols [10]. How this process of sterol sequestration works, remains subject to debate. The most studied mechanism of action is the pore forming model (see Section 3.2. Pore Forming Models) in which polyenes interact with ergosterol to form ion-leaking pores in the membrane [19]. Nevertheless, it has been shown that pore formation can also be established in the absence of sterols [31]. Early on, Cotero et al. [31] proposed that sterols have an essential role in the structure of the membrane itself during amphotericin activity, but might not be directly involved in the pore formation [31]. Later, other studies supported this idea [32][33][34][35][36][37]. For example, the polyene natamycin was shown to bind ergosterol without altering the cell membrane permeability [37]. Further research pointed out that natamycin inhibits various ergosterol-dependent membrane proteins and so disturbs essential cellular processes such as glucose transport, amino acid transport [38] and vacuolar fusion [36]. Currently, four models of the polyene mode of antifungal action have been proposed: the pore forming model, the surface adsorption model, the sterol sponge model and the oxidative damage model [39] (see Section 3.2 Pore Forming Models, Section 3.3 Surface Adsorption and Sterol Sponge Models and Section 3.4 Other Modes of Action).
In every proposed model, the binding of the polyene with ergosterol is key to its antifungal effect [39]. Ergosterol plays an essential role in many cellular processes of fungi, including regulation of membrane proteins, endocytosis, cell division, membrane fluidity and cell signaling [40][41][42]. The specificity of therapeutic polyenes to ergosterol comes from the fact that ergosterol has a significantly different three-dimensional structure compared to mammalian cholesterol, enabling better binding into the hydrophobic "pocket" of polyenes such as AmB, as depicted in Figure 2 [1]. Three interactive forces play a role in the binding of AmB and ergosterol: Van der Waals powers which are highest when both molecules are orientated co-planar and parallel, a hydrogen bond network between the 3β-OH group of the sterol and the polar mycosamine group of AmB and π-π electronic interactions between the ergosterol side chain and the polyene "tail" of AmB [43]. The latter essential "attach point" does not occur when AmB binds to cholesterol [39,43]. Moreover, Van Der Waals interactions are weaker between cholesterol and AmB due to the sigmoidal conformation of the sterol side-chain [39]. The specific binding, along with the higher ergosterol:phospholipid ratio in fungal cell membranes, compared to the cholesterol:phospholipid ratio in mammalian cells, explains the selectivity of most polyenes to fungal cells [1].

Pore Forming Models
The most studied model for polyene action is the pore formation model, in which polyenes and ergosterol interact to form an ion channel-like complex that leaks ions and small organic molecules from the cell, eventually leading to cell death [19]. Based on the amphipathic properties of polyenes, they would orientate in the plasma membrane with their hydrophobic polyene "tail" interacting with ergosterol, directed to the inner lipid environment of the membrane, while the hydrophilic polyol portion would form an aqueous channel, as illustrated in Figure 3A,B. Intermolecular hydrogen bonds between amino and carboxyl groups of the hydrophilic "heads" of neighboring polyene molecules further stabilize this channel [19]. Neutron diffraction studies have confirmed that such an architecture can exist when AmB interacts with ergosterol [44]. Typically, 4 to 12 polyene monomers would form a pore [39]. As the length of AmB is almost equal to the length of a membrane phospholipid on average, two types of channels can be made: a full pore consisting of two ring complexes of polyenes (see Figure 3A) and a "half-pore", containing only one polyene ring (see Figure 3B). Both types essentially have the same structure but the latter would induce a conformational thinning of the lipid bilayer [19]. Which pore is formed would primarily depend on the polyene, and the composition and thus thickness of the membrane [19]. For example, AmB would primarily form half-pores in a membrane mainly composed of dimyristoylphosphatidylcholine (DMPC) [45]. Pores are only formed after a certain threshold of polyene molecules in the membrane is reached. Below this threshold, aggregate complexes termed "non-aqueous pores" or "cation-selective pores" can increase the membrane permeability to monovalent cations, while true pores and half-pores can also transport larger nonelectrolyte molecules [46]. This threshold is significantly lower (by factor 5-10) in ergosterol containing membranes, compared to cholesterol containing membranes [46]. Moreover, patch clamp experiments with artificial AmB channels have shown that the ion transport occurs faster (a shorter channel "dwell time") in ergosterol containing membranes compared to cholesterol containing membranes [47], showcasing their antifungal specificity. The diameter of the pore determines the selectivity of transport or "leakage" out of the cell and depends on the type and concentration of polyene, while sterol type and sterol concentration have minor influences [19,39,48]. This is shown in the study by Yang et al., in which the channel diameter increases 100-fold when the concentration of AmB multiplies with factor 40 in an ergosterol-rich membrane [48]. In general, AmB forms relatively wide pores (approximately 0.46 nm) and can transport molecules as big as sucrose, while nystatin forms smaller pores (approximately 0.36 nm) [19].

Surface Adsorption and Sterol Sponge Models
The second and third models for polyene mode of action both hypothesize that, by adsorption or extraction of ergosterol from the membrane, the phospholipid membrane is destabilized, and essential cellular processes such as endocytosis and regulation of membrane protein function are disturbed [19]. Polyenes could adsorb ergosterol molecules to the "surface" of the phospholipid bilayer as illustrated in Figure 3C, termed the "surface adsorption model" [33,49]. In light of this theory, Anderson and colleagues [32] conducted a series of NMR studies to determine the localization and structure of AmB interacting with ergosterol, and they observed that these complexes are not (always) inserted in the membrane and can form extra-membranous aggregates. They suggested that in such a mechanism, large aggregates of parallelly positioned AmB molecules can form at the membrane, functioning as a "sterol sponge" [32] as illustrated in Figure 3D. Extracting ergosterol from the membrane in such a "sponge" would perturbate a vast array of ergosterol-dependent cellular processes, many of which governed by membrane proteins that directly bind to ergosterol. This might also explain why resistance to polyenes is rarely observed, as in resistant cells an alternative membrane sterol such as lanosterol (a precursor of ergosterol) will probably malfunction in these processes and reduce the fitness and pathogenicity of the cell [32]. Anderson et al. [32] also suggest that the extraction of cholesterol by large extra membranous aggregates of AmB is the primary cause of toxicity of this drug towards mammalian cells and thus, optimizing the binding affinity of AmB derivatives to ergosterol could significantly improve its therapeutic efficacy. Some remarks have to be made regarding the sterol sponge model proposed by Anderson et al. [32]. First, the ergosterol-to-lipid ratio used in their experimental set-up is different from the one observed in natural systems [39]. As the ergosterol-to-lipid ratio is essential regarding the polyene susceptibility of fungi, this might play a vital role. Secondly, it has been proven that, in a cholesterol-saturated environment such as the mammalian cell membrane, the thermodynamical balance between ergosterol-AmB vs. cholesterol-AmB would shift towards the latter, meaning that the "sterol sponge" would be saturated with cholesterol rather than ergosterol. A third argument against this model is that fungi have a rigid hydrophilic cell wall composed of polymers of chitin that might prevent the passage of hydrophobic ergosterol and so the formation of super-aggregates or "sponges" outside the cell wall [39].
Ion channel formation, small membrane-spanning aggregates and large extra-membranous aggregates may exist at the same time [32], although the chemical structure of the polyene probably influences the primary mode of action. Several studies have shown that the elimination of the C35 hydroxyl group of AmB does not alter the sterol-adsorption capacity and cytotoxic effect of the molecule but eliminates its pore-forming capacity [33,50]. It was suggested that the ability to form pores depends on the dimensions of the polyene macrolactone ring and, therefore, polyenes can be divided into non-pore forming polyenes (e.g., natamycin) and the pore-forming polyenes (e.g., AmB and nystatin) [19].

Other Proposed Modes of Action
Several observations point towards oxidative damage as an additional mode of action of AmB [51,52]. One example is the rescue effect of hypoxia, exogenous catalase and super oxide dismutase (SOD) during AmB treatment of C. albicans without hindering AmB induced K + -leakage [52]. Another example is the enhanced resistance to oxidative damage by H 2 O 2 of AmB resistant C. albicans strains [51]. Currently, several studies have provided evidence that polyenes can induce oxidative stress and cause DNA damage, protein carbonylation and lipid peroxidation, eventually leading to or contributing to cell death in fungi [53]. Moreover, metabolomic analysis of C. albicans exposed to AmB pointed out that AmB induced cell death was attenuated through increased production of polyamines such as putrescine, spermidine and spermine which have a role in scavenging reactive oxygen species (ROS) [54]. This is supported by gene expression analysis of C. albicans exposed to AmB, showing an increased expression of stress-related genes besides genes involved in membrane sterol homeostasis [55]. In Cryptococcus neoformans, it was shown that, after addition of AmB, cells become metabolically inactive and encounter a strong oxidative burst suggested to contribute to AmB induced cell death apart from membrane interactions and pore formation [35]. How this oxidative stress is exactly caused is still not clear, although it was suggested that polyene binding to the membrane triggers this response that leads to an apoptotic such as phenotype that includes ROS production or that, since AmB auto-oxidizes and forms free radicals [56], the antifungal itself causes oxidative stress [35]. In the latter model, the oxidative stress effect of polyenes would be distinct from its membrane permeabilization properties, although the free radicals produced would affect the membrane itself through lipid peroxidation [35].

Drug Resistance to Polyenes
Although polyenes have been used for many decades [5], polyene resistance is still rare compared to resistance to other antifungal drugs [20]. One explanation for this is the fact that it is often associated with severe fitness trade-offs [20]. Another hypothesis is that, compared to most other antifungals, polyenes target a major cell membrane component instead of an essential enzyme. Nevertheless, several mechanisms of resistance have been proposed ( Figure 4) as described in Section 4.1. Molecular Mechanisms of Resistance, and several pathogenic fungi species are known to possess intrinsic resistance or are able to acquire resistance to polyenes in the clinic, as described in Section 4.2. Epidemiology of Polyene Resistance.

Molecular Mechanisms of Resistance
The most common mechanism of acquired AmB resistance is attributed to alterations in the sterol composition of the fungal cell membrane [57,58]. Several mutations in genes of the ergosterol biosynthesis pathway (ERG genes) have been associated with this mechanism. In C. albicans, the loss of function of ERG11 and ERG3 genes (lanosterol 14α-demethylase and C-5 sterol desaturase, respectively) leads to the exchange of ergosterol for alternate sterols such as lanosterol, eburicol and 4,14-dimethyl-zymosterol in the membrane [20,58]. In other reports, AmB resistance in C. albicans is associated with a substitution in ERG11 and loss of function of ERG5 (C-22 sterol desaturase), again associated with an alternate membrane sterol composition [20,59]. In other Candida spp., inactivation of ERG6 (C-24 sterol methyl-transferase) [60,61] and ERG2 (C-8 sterol isomerase) [61] were found to have a similar effect. A mutation in ERG2 resulting in its inactivation, is one of the only described mechanisms of AmB resistance in C. neoformans [62]. Nevertheless, high AmB resistance (MIC 20 µg/mL) without significant alterations in ergosterol biosynthesis have been reported in experimentally evolved C. neoformans, suggesting that sterol-independent mechanisms of AmB resistance also exist in Cryptococcus spp. [63]. Isolates of the Candida haemulonii species complex (C. haemulonii, C. haemulonii var. vulnera and C. duobushemulonii), known for their increased AmB resistance, have been shown to possess a cell membrane with an altered sterol profile similar to AmB resistant species with ERG11, ERG3 ERG2 and ERG6 mutations [64]. Given the severe fitness deficits of mutations in the ergosterol biosynthesis pathway [20], AmB resistance is rarely found in combination with resistance to other antifungal drugs, although certain polyene resistance inducing ERG mutations can result in cross resistance to azoles [58, 65,66]. A well-known example of this occurs upon inactivation of the target of azoles ERG11 and ERG3, resulting in the abrogation of the biosynthesis of ergosterol and inhibition of the synthesis of the toxic sterol 14-methyl-3,6-diol, respectively [58].
The growing interest in the role of oxidative stress in the fungicidal effects of AmB has led to an increase in research concerning stress related mechanisms of resistance. It is hypothesized that a reduction in polyene induced oxidative stress might allow the cell to better tolerate AmB exposure [53,67]. In intrinsically AmB resistant fungal pathogens such as A. terreus, this mechanism of resistance appears to be more important than an altered membrane sterol composition [67]. Compared to AmB susceptible Aspergillus fumigatus, intrinsically resistant A. terreus shows only a slight increase in membrane ergosterol levels, but a significant increase in catalase levels [68]. As research shows, a catalase-based mechanism of resistance counteracting AmB-induced oxidative stress that would otherwise result in cell death, is likely [35,[51][52][53]. Catalase activity might break down the harmful ROS produced during AmB treatment and so protect the cell [69]. Further evidence supporting this hypothesis was given by Blatzer et al. [70], who found a severe induction of mitochondrial ROS generation upon AmB exposure in rare susceptible isolates of A. terreus, compared to resistant isolates [67]. Another mechanism of stress tolerance and/or resistance upon AmB exposure is likely governed by the molecular chaperone heat shock protein 90 (Hsp90), a highly connected, central factor in cellular stress responses that possibly potentiates the acquisition of resistance by regulating a myriad of client proteins [71,72]. Cowen et al. [72] hypothesize the involvement of Hsp90 in the emergence and/or maintenance of resistance via different possible mechanisms: (i) the disruption of Hsp90 might allow the persistence of new genetic variations, (ii) active Hsp90 might stabilize mutated cell regulators that have the ability to induce resistance but are prone to misfolding or (iii) Hsp90 might chaperone unmutated regulators of cell signaling (e.g., calcineurin), thereby allowing the development of adaptive phenotypes [72]. The connection between Hsp90 and antifungal resistance can be investigated through the use of Hsp90 inhibitors such as geldanamycin and radicicol [73,74]. In fluconazole resistant strains, modest inhibition of Hsp90 leads to abrogation of resistance but has no effect on cell growth [72]. Interestingly, inhibition of Hsp90 in AmB resistant cells does lead to inhibition of growth, even in the absence of AmB [20]. Blatzer et al. [75] found a correlation between AmB resistance in A. terreus and another family of chaperones: the Hsp70 family. They identified higher basal levels of expression of several Hsp70 genes in resistant strains and observed an instant activation of the majority of Hsp70 genes in the resistant A. terreus isolates upon exposure to AmB. Inhibition of Hsp70 only had limited effects on AmB susceptible A. terreus isolates, while it led to a considerable increase in susceptibility for resistant isolates [75]. Overall, these observations indicate a possibly essential role for the molecular chaperones of the Hsp90 and Hsp70 family in the acquisition and/or maintenance of AmB resistance [20,75].
In addition to alterations of the membrane sterol composition and regulation of oxidative stress, some reports show a correlation between AmB resistance and alterations of the fungal cell wall. In 1999, Seo et al. [76] acquired an experimentally evolved Aspergillus flavus isolate that was able to grow in concentrations of up to 100 µg/mL AmB [76]. An alteration of the cell wall was suggested to be an important mechanism contributing to this resistance. The main change in the fungal cell wall composition was reported to be an increase in the 1,3-α-glucan fraction [76]. More recently, an AmB resistant C. tropicalis isolate was also found to have an enlarged cell wall with increased levels of 1,3-β-glucan [77]. Although the increased glucan production could be the result of regulatory mechanisms affected by AmB exposure or resistance [76], another hypothesis is that the glucan molecules can physically inhibit the penetration of AmB through the cell wall [76]. In addition to increased AmB resistance, the cells with an enlarged cell wall also elicited a different immune response compared to susceptible cells. Mesa-Arango et al. [77] suggest that the increased 1,3-β-glucan is proportional to the strength of the immune response and contributes to prolonged survival of resistant clones since an exaggerated proinflammatory response can reduce the efficiency of immune clearance of a systemic infection [77,78].
The active efflux of drugs, a common mechanism of resistance for azole drugs, does not appear to play a role for polyene resistance. Although one report indicates an increased AmB tolerance in C. albicans upon the overexpression of drug efflux pump Cdr1 [79], this was later contradicted by Niimi et al. [80], who did not find a correlation between Cdr1 overexpression and polyene resistance.
Although multiple mechanisms of resistance have been proposed, the molecular mechanisms driving a reduced polyene susceptibility are not always clear. This is showcased by the emergent C. auris, in which an AmB resistance prevalence of up to 30% [81][82][83] has been found among clinical isolates, but still sequencing efforts cannot identify driving mutations. Differential expression of ERG genes might be involved, as Munoz et al. [84] showed an upregulation of ERG13, ERG6, ERG2 and ERG1 upon AmB exposure in AmB resistant C. auris isolates, compared to susceptible strains [84]. Escandon and colleagues [18] sequenced AmB resistant C. auris strains of Colombia and identified a mutation in a gene encoding a homolog of C. albicans FLO8. Flo8 is a transcription factor that has multiple downstream effectors, and it has been proven to positively regulate ERG11 expression in yeast [85]. A recent study by Carolus et al. [86] also identified a mutation in FLO8 in an AmB resistant C. auris strain that was experimentally evolved. Although this mutation did not significantly alter the AmB MIC, it raises questions regarding the link between FLO8 and AmB resistance in C. auris. The same strain from Carolus et al. [86] also acquired nonsense mutations in both ERG11 and ERG3, significantly decreasing the AmB susceptibility with cross resistance towards azoles [86]. Additionally, a mutation was found in a gene predicted to encode the DNA damage checkpoint protein Mec3 [86], part of the Rad17p-Mec3p-Ddc1p sliding clamp [87]. As AmB evokes oxidative damage to DNA that leads to apoptosis [53], DNA damage checkpoint proteins might be involved in AmB resistance, although this link has not been confirmed. Due to these and other loose ends, further research into the mechanisms of polyene resistance, especially in the context of pan resistance as found in C. auris [81,88,89], Fusarium [90][91][92] and Scedosporium spp. [93][94][95][96], is highly needed [97,98].
Among pathogenic molds, Aspergillus spp. are well known for their extensive AmB resistance. For A. fumigatus, the most common cause of invasive aspergillosis [143][144][145][146], recent studies have shown the ability to efficiently acquire AmB resistance with rates ranging from 43% [144] to 96.4% [146]. Widespread AmB resistance is also found among other common Aspergillus spp. such as A. flavus (resistance rate ranges from 12.5% [147] up to 87% [144]) and intrinsic AmB resistance is well recognized in A. terreus [67]. Furthermore, intrinsically resistant isolates have been suggested for several other Aspergillus spp. such as A. flavus [145,148] and Aspergillus lentulus [145,149]. It should be noted that some studies [150] report a low incidence of AmB resistance in Aspergillus spp. Baddley et al. [151] report, for example, low AmB MICs for 93.3% of the investigated Aspergillus spp. isolates with increased MICs only in A. terreus [151]. Other filamentous fungi such as Scedosporium apiospermum and Lomentospora prolificans (formerly Scedosporium prolificans) have also been identified as intrinsically AmB resistant and, therefore, often require complex antifungal treatments [93][94][95]152,153]. Additionally, some Fusarium spp. have been reported as intrinsically resistant to AmB [90][91][92]154]. While AmB is usually recommended for treating invasive fusariosis [13], Fusarium solani¸a prevalent Fusarium spp. often shows increased AmB MICs [155,156]. Other members of the Fusarium genus are generally susceptible to AmB. Sporothrix schenkii is usually observed to be susceptible to AmB although some resistant isolates have been found [157]. Furthermore, several reports indicate increased AmB MICs for the hyphal form of this pathogen compared to the yeast form [158][159][160], complicating clinical interpretation [158]. Histoplasma capsulatum exhibits very low MICs for AmB, making it a preferred treatment for this pathogenic mold [161,162]. Although AmB resistance is rare for isolates of Mucorales [163][164][165], some Rhizopus and Cunninghamella strains show decreased AmB susceptibility [103,166], with one report showing decreased AmB susceptibility in 3 out of 19 and 5 out of 8 isolates tested, respectively [167].
A major challenge in investigating the epidemiology of polyene resistance is the lack of clear susceptibility breakpoints for most fungal pathogens. Park et al. [168] showed that the commonly used breakpoint of 2 µg/mL for Candida spp. does not always correlate with the clinical outcome of AmB treatment. Rex et al. [169] showed similar results and even observed an inversely proportional trend, where lower MIC values were associated with increased failure of treatment in the clinic. Nonetheless, some studies report a positive correlation between AmB MIC and treatment success [170,171]. For most fungal species, no clinical breakpoints have been determined, thus forcing researchers to resort to epidemiological cutoff values to distinguish resistant from susceptible isolates. These conflicting observations complicate and question the interpretation of resistance epidemiology and its clinical significance.

Conclusions
Although discovered over 70 years ago, and despite their inherent toxic nature, polyene antifungals are still part of the first-line treatment for various types of invasive fungal infections. This showcases both the predicament of the antifungal drug discovery pipeline and the potency of polyenes as antifungal drugs. Polyenes have been used in the clinic over several decennia but still, debates are ongoing regarding their modes of action, and the mechanisms of resistance are arguably the least understood compared to other commonly used antifungal drugs such as azoles and echinocandins. Emergent polyene resistant pathogens such as C. auris, for which hardly any reports on resistance mechanisms exist despite a high incidence of resistance in the clinic, urge to fill those knowledge gaps. A multitude of recent reports show the interest in further development of AmB formulations, including oral drug development, to facilitate use and increase treatment efficacy. Especially due to their broad spectrum, rare incidence and relatively weak potency of resistance development, polyenes could still be described as a "diamond in the rough" in the current antifungal drug resistance crisis. Nonetheless, for polyenes to be on the forefront of this battle, further research into their mode of action, drug resistance and clinical application is key.