Rational approaches towards inorganic and organometallic antibacterials

Abstract

The occurrence of drug-resistant bacteria is drastically rising and new and effective antibiotic classes are urgently needed. However, most of the compounds in development are minor modifications of previously used drugs to which bacteria can easily develop resistance. The investigation of inorganic and organometallic compounds as antibiotics is an alternative approach that holds great promises due to the ability of such molecules to trigger metal-specific mechanisms of action, which results in lethal consequences for pathogens. In this review, a selection of concepts to rationally design inorganic and organometallic antibiotics is discussed, highlighting their advantages by comparing them to classical drug discovery programmes. The review concludes with a short perspective for the future of antibiotic drug development and the role metal-based compounds will play in the field.

Introduction

The emergency of pathogenic bacteria resistant to the current arsenal of antibiotics is one of the greatest threats to human health worldwide. We are at the verge of losing the fight against bacterial infections, risking to reach a situation paralleling the so called “pre-antibiotic era” where minor injuries could end deadly. Up to the early 1900s, infectious diseases were the leading cause of death and accounted for up to 25% of England’s mortality rate (Smith et al. 2012). With the commercialisation of antibiotics, this number dropped to under 1% in England. Unfortunately, increasing mortality rates caused by infectious diseases are not scenarios from the past. In 2015, almost 700,000 cases of antibiotic-resistant infections were reported in Europe causing over 33,000 deaths (Cassini et al. 2019). In the United States, more than 2.8 million people acquire drug-resistant bacterial infections every year causing 35,000 deaths as a result (Centers for Disease Control and Prevention (U.S.) 2019). The Review on Antimicrobial Resistance in the UK suggests that if drug resistance continues to rise at the current rate, infections caused by resistant bacteria would account for up to 10 million lives each year by 2050, overtaking cancer as the leading cause of death (AMR-Review 2016). A particular worry is the continuous appearance of multidrug-resistant (MDR) and pan-drug-resistant (PDR) bacteria in the clinics, also known as “superbugs” causing infections that are almost untreatable with existing antibiotics.

Although there is a pressing need to develop new antibiotics, the clinical development pipeline is dry. In 2019, only 32 antibiotics in clinical development where identified by the WHO that address the priority pathogens and only six of them were considered to be novel (WHO|Antimicrobial resistance 2020). Although 27 antibiotics reached market approval by the FDA between 2000 and 2020, this is a very small number compared to the total FDA approved drugs of 1090 in the same period. As antibiotic drug discovery has a high failure rate, coupled with low profit margins due to the fact that approved drugs are kept as last resort options, several pharmaceutical companies have withdrawn from antibiotic development (Wetzel et al. 2021).

With the dwindling supply of novel antibiotics and the increasing challenge of fighting drug-resistant bacteria, there is the urgent demand to move beyond classical drug discovery programmes. An alternative approach was recently highlighted in a review article by Wetzel et al. (Wetzel et al. 2021). Historically, drug discovery programs have focused on designing small organic molecules to act and inhibit a single molecular target, usually a protein. As complex diseases, such as Parkinson’s disease, and Alzheimer’s disease, can hardly be combated with an one-drug-does-it-all solution, researchers have intensified the development of multi-target drugs that act on multiple cellular or molecular targets of a disease pathway (Bolognesi 2013; Morphy et al. 2004; Ramsay et al. 2018). Since 2000, the FDA has approved fourteen multi-target antibiotics (Wetzel et al. 2021). Metal-based compounds offer exactly this possibility of designing molecules with multiple mechanisms of action that are less accessible with pure organic molecules (Boros et al. 2020).

The research area of biological inorganic chemistry is a steadily growing field that develops medicinally relevant inorganic and/or organometallic molecules and studies their behaviour and activity in the biological context of a disease model (Gasser and Metzler-Nolte 2012; Patra and Gasser 2017). Metal-based compounds have extensively been studied for various diseases but so far most research has focused on anticancer agents (Englinger et al. 2019; Gasser et al. 2011; Hartinger et al. 2013; Jaouen et al. 2015; Johnstone et al. 2016; Patra and Gasser 2017). However, the study of organometallic compounds in other fields such as antiparasitic (d’Orchymont et al. 2018; de Almeida et al. 2013; Glišić and Djuran 2014; Hess et al. 2015, 2016a,b, 2017; Ong et al. 2019; Patra et al. 2013; Simpson et al. 2015) or antibacterial (Albada and Metzler-Nolte 2017; Betts et al. 2020; DuChane et al. 2018; Frei et al. 2020b; Mazzei et al. 2019; Patra et al. 2012; Schatzschneider 2019; Sierra et al. 2019; Stringer et al. 2017; Szczupak et al. 2020; Wenzel et al. 2013) drug discovery is gaining traction. Due to their unique features, metal-based compounds offer an untapped source for future antibacterial drug discovery programmes.

The use of compounds containing metals and metalloids to fight bacterial infections is not new. In the early 1910s, Paul Ehrlich’s lab discovered a molecule (Arsphenamine, Salvarsan or Ehrlich 606) featuring the metalloid arsenic, which was later found to be effective in eradicating syphilis, a bacterial infection caused by Treponema pallidum spirochete (Bosch and Rosich 2008). Melarsoprol (Arsobal^®) is another arsenic-based medicine, used to treat African trypanosomiasis (sleeping sickness) and listed as one of the Essential Medicines by the WHO (WHO|Model List of Essential Medicines 2019). Yet, only a few metal-based compounds are currently used in the clinics to treat infections. Pravibismane (MBN-101), a bismuth-thiol compound developed by Microbion, is an antimicrobial compound in clinical development that has completed clinical safety and activity studies for the treatment of infected bone sites and for topical application in patients with moderate to severe diabetic foot infection (DFI) (Microbion Corporation 2020a,b). It is also in development as an inhalable agent for the treatment of cystic fibrosis (CF) related pulmonary infections, and Microbion was recently awarded $1.8 million by the Cystic Fibrosis Foundation to speed up Pravibismane’s development (Microbion Corporation Receives Additional Funding Support from the Cystic Fibrosis Foundation to Speed Advancement of Inhaled Pravibismane 2021). Recently, Blaskovich, Frei and colleagues published their screening results of 906 metal-based compounds, which were supplied by several research groups and were screened by the Community for Open Antimicrobial Drug Discovery (CO-ADD) for their antimicrobial activity. In this high-throughput-screening (HTS)-like approach, the researchers reported the results of 30 molecules that showed activity against both Gram-positive and/or Gram-negative bacteria including activities against a methicillin-resistant Staphylococcus aureus (MRSA) strain, showcasing the potential to find new active candidates among metal-based compounds (Frei et al. 2020b).

The main focus of this mini-review lies on approaches that have been used to rationalise the development of metal-based antibacterial candidates. These include the incorporation of organometallic moieties as structural scaffolds to diversify compound geometries, the utilisation of structural biology data to guide compound design, the usage of metallofragment libraries for a fragment-based approach, light-activated compounds in antimicrobial Photodynamic Therapy (aPDT), or the use of catalytic properties of metals inside pathogens to activate prodrugs.

Plantensimycin – organometallic moieties as structural scaffolds

The incorporation of organometallic moieties as structural inert scaffolds was applied to design metal complexes in various fields of biological inorganic chemistry (Meggers 2007, 2009; Schlotter et al. 2005). In this concept, a metal centre is used as a unique building block almost like “a hypervalent carbon” allowing for the spatial organisation of its substituents in three-dimensional space. With the metal centre acting as a bio-isostere for a challenging organic group, coordination complexes allow to build three-dimensional geometries that are difficult to achieve with organic molecules and simultaneously introduce elements of chirality that can facilitate molecular recognition and interaction (Meggers 2007). While a carbon atom with four different substituents has only two different enantiomers, an octahedral metal complex with six different ligands has 30 different stereoisomers. Meggers and co-workers reported the synthesis of numerous ATP-competitive ruthenium-based kinase inhibitors, where the metal portion substituted a complex organic scaffold of the natural product staurosporine. The facilitated synthesis allowed the generation of a multitude of different metal complexes and led to the discovery of selective protein kinase inhibitors with picomolar affinities (Meggers 2009; Mulcahy and Meggers 2010).

Similarly, Patra, Gasser and Metzler-Nolte used this approach to design various organometallic analogues of the natural antibacterial plantensimycin. The inhibition of FabF, an enzyme which plays a crucial role in the fatty acid biosynthesis in bacteria, is the key target of plantensimycin (Wang et al. 2006). For this work Patra et al. used the co-crystal structures of plantensimycin and its protein target (PDB code 2GFX, Figure 1B) to guide compound design. The polar portion of plantensimycin is engaged in an extensive hydrogen bonding network in the malonyl binding pocket of the FabF enzyme. Therefore, the relatively lipophilic tetracyclic unit of plantensimycin was the ideal point for substitution. Extensive structure-activity relationship (SAR) studies around this scaffold led to four different types of achiral and chiral organometallic derivatives (Patra et al 2009, 2010, 2011). The achiral arene chromium tricarbonyl derivative of plantensimycin was the most promising derivative and docking experiments suggested that the half-sandwich moiety was capable of occupying the same enzyme pocket as the tetracyclic unit (Figure 1A).

Figure 1:

Organometallic derivatisation of plantensimycin and postulated binding with the FabF enzyme.

(A) Chemical structure of plantensimycin and the (η⁶-pentamethylbenzene)Cr(CO)₃ plantensimycin derivative where the portion of organometallic derivatisation is highlighted in dark cyan. (B) Active site interactions of plantensimycin with Thr307, Ala309, Thr270 and Ser271 of FabF enzyme (PDB 2GFX), with the lipophilic tetracyclic unit of plantensimycin coloured in dark cyan.

Yet, the whole bacterial cell assay displayed only a modest inhibitory activity of this compound against the Gram-positive pathogen B. subtilis with an MIC of 80 μg mL⁻¹ (0.15 mM) (plantensimycin MIC = 0.2 μg mL⁻¹ (0.5 μM)) (Patra et al. 2009). In order to pinpoint potential target proteins and investigate the mechanism of action, a proteomics response study was performed. While proteome analysis after plantesimycin treatment showed fatty acid biosynthesis inhibition, the (η⁶-pentamethylbenzene)Cr(CO)₃ derivative did not inhibit proteins relevant for fatty acid biosynthesis. Rather, the metal-based derivative induced a similar proteome profile to membrane-active antibiotics, suggesting a different mechanism of action (Wenzel et al. 2011). In addition, while plantensimycin is known to be only bacteriostatic, the (η⁶-pentamethylbenzene)Cr(CO)₃ analogue was found to be bacteriolytic. Thus, introduction of the (η⁶-pentamethylbenzene)Cr(CO)₃ core drastically changed the mechanism of action, leading to unprecedented lethality.

Structure-informed design of organometallic β-lactam antibiotics

Since the discovery of penicillin by Alexander Fleming in 1928, a variety of synthetic and semi-synthetic β-lactam antibiotics have since reached the market (Fleming 1929, 1980). This class of antibiotics relies on the inhibition of penicillin binding proteins (PBPs), which are key enzymes in the last steps of the bacterial cell wall biosynthesis (Macheboeuf et al. 2006; Sauvage et al. 2008). β-lactams are capable of inhibiting the transpeptidation step through the irreversible acylation of the active site serine residue (Yao et al. 2012). The heavy use of β-lactams in the clinics since the 1940s have resulted in widespread bacterial resistance, of which one major resistance mechanism involves the production of enzymes – the β-lactamases – able to hydrolyse this class of antibiotic (Bush and Fisher 2011; Zaman et al. 2017). β-lactamases can be divided into four different classes A, B, C and D based on sequence homology, of which three of them (A, C and D) share a common serine residue in their active site acting as a nucleophile for ring-opening of the β-lactam. Class B β-lactamases are metal-dependent and have either one or two Zn²⁺ ions in their active site (Ambler et al. 1980).

In the mid-1970s, Marr and co-workers published their seminal work on the derivatisation of β-lactam antibiotics such as penicillins and cephalosporins with ferrocenly moieties (Edwards et al. 1975, 1976a,b, 1979). In order to escape the two-dimensional profile of these semi-synthetic antibiotics, the authors rationalised that replacement of the conventional phenyl and heteroatomic groups with a ferrocenly moiety would increase their “third dimension” and allow the installation of a metal atom in close proximity of the β-lactam ring. While precise MIC values of the newly prepared compounds have not been determined, Marr and co-workers stated that some of their new organometallic derivatives exhibited antibiotic activity against various strains of S. aureus (Edwards et al. 1975, 1976a,b, 1979). Nevertheless, it is questionable whether these compounds could overcome resistance associated with β-lactamases. Similarly to this work by Edwards, Epton and Marr, Simionescu and colleagues reported ferrocenyl conjugates of 6-aminopenicillanic acid (6-APA), 7-aminocephalosporanic acid (7-ACA) and 7-amino-3(2-methyl-1,3,4-thiadiazol-5-thio-methyl)cephalosporanic acid (7-ADCA), but unfortunately the authors only state that these conjugates seem to be active against Gram-positive and Gram-negative bacteria without reporting the corresponding MIC values (Scutaru et al. 1991a,b; Simionescu et al. 1983, 1985). Only much later, in 2010, Long et al. described in more detail the significant antibacterial activities of ferrocenyl-containing penems against both Gram-positive (including methicillin-resistant S. aureus, MRSA) and Gram-negative bacteria (Long et al. 2010). Various other metal-containing β-lactam derivatives have been more extensively discussed in a recent review by Sierra, Casarrubios and de la Torre (Sierra et al. 2019).

Since 2012 the research group of Kowalski reported on several ferrocenly- and ruthenocenyl-conjugates of β-lactams antibiotics and the use of structural data to guide and understand compound interactions and inhibition of their target. Their first publication described ferrocenyl conjugates of ampicillin and 6-APA and the antibacterial activity of these derivatives against a panel of Gram-positive bacteria. The ferrocenly 6-APA (Fc-6-APA) conjugate showed significant antibacterial activity, and was able to inhibit bacterial DD-carboxypeptidase 64–575 in the nanomolar regime (Figure 2A) (Skiba et al. 2012).

Figure 2:

Different classes of β-lactam antibiotics, their organometallic counter parts and respective X-ray crystal structures.

(A) Structure of 6-aminopenicillanic (6-APA) and ferrocenyl 6-aminopenicillanic derivative (Fc-6-APA). (B) Structure of ruthenocenyl 6-aminopenicillanic derivative (Ru-6-APA, 1) and its hydrolysed decarboxylated form (2). (C) Active site interaction of the hydrolysed and decarboxylated form of Ru-6-APA with CTX-M-14 E166A (PDB 4XXR). (D) Structure of 7-aminodesacetoxycephalosporanic acid (7-ADCA) and ruthenocenyl 7-aminodesacetoxycephalosporanic acid (Ru-7-ADCA, 3). (E) Covalent acyl-enzyme complex of 3 (Ru-7-ADCA) with CTX-M-14 E166A mutant, covalent linkage with Ser70 and active site interactions (PDB 5UJO). (F) Structure of 7-aminocephalosporanic acid (7-ACA) and ruthenocenyl 7-aminocephalosporanic acid (Ru-7-ACA, 4). (G) Covalent acyl-enzyme complex of 4 (Ru-7-ACA) with CTX-M-14 E166A mutant, covalent linkage with Ser70 and active site interactions (PDB 6VNU).

Follow-on work focused on the replacement of the ferrocenly portion in 6-APA derivatives with a ruthenocenyl moiety. The most promising Ru-6-APA derivative had superior antibacterial activity compared to its Fc-6-APA analogue with the highest activities observed against S. aureus with an MIC value of 2.0 μg/mL⁻¹ (0.5 μg/mL⁻¹ ampicillin) and Staphylococcus epidermidis with an MIC value of 4.0 μg/mL⁻¹ (5.0 μg/mL⁻¹ ampicillin). In 2015, Lewandowski et al. revealed for the very first time a high-resolution X-ray crystal structure at 1.18 Å resolution of a novel synthetic ruthenocenyl-6-aminopenicillinic acid in complex with CTX-M-14 E166A β-lactamase (PDB 4XXR, Figure 2C). The authors chose to use the CTX-M-14 E166A β-lactamase mutant due to its catalytic site similarities to PBPs with the E166A mutation expected to prevent the deacylation and possibly allowing to trap the hydrolysed β-lactam product. Instead of Ru-6-APA (1), Lewandowski and colleagues observed 2, a hydrolysed product of 1 where the carboxylate group from the ring-opened β-lactam was missing, but which was still binding in the active site forming many favourable interactions with active site residues (Figure 2B) (Lewandowski et al. 2015). Kowalski, Chen and co-workers reported on metallocenyl 7-aminodesacetoxycephalosporanic acid (7-ADCA) bioconjugates of which some Ru-7-ADCA derivatives had several-fold lower IC₅₀ values against dd-carboxypeptidase 64–575 compared to their Fe-7-ADCA counterparts and the control penicillin G (Figure 2D) (Lewandowski et al. 2017). They were also able to trap the first organometallic Ru-7-ADCA derivative 3 with CTX-M-14 E166A β-lactamase as a covalent acyl-enzyme complex (PDB 5UJO, Figure 2E). The most recent work by Lewandowski et al. describes a ruthenocenyl-7-aminocephalosporanic acid (Ru-7-ACA, 4) bioconjugate with superior antibacterial activity against several S. aureus strains compared to its ferrocenyl analogue and a lack of host toxicity (Figure 2F). The X-ray crystal structure of 4 trapped in an acyl-enzyme state in complex with the CTX-M-14 E166A β-lactamase mutant confirms its binding to the active site in a similar fashion to previously reported ruthenocenyl β-lactam bioconjugates (Figure 2G) (Lewandowski et al. 2020).

In order to conclude if the reduced antibacterial activity of several ferrocenyl β-lactam bioconjugates was due to their reduced interactions with target proteins DD-carboxypeptidases, DD-transpeptidases or PBPs, additional structural information would be required to compare the binding of ferrocenyl- and ruthenocenyl-β-lactam analogues. In case X-ray crystallography cannot answer these questions, alternative biophysical techniques could be considered to understand the binding of these metallocenyl β-lactam analogues.

Metallo-fragment libraries as alternative starting points for inhibitor development

Various attempts have been made to develop metal-containing antimicrobials by starting from a known organic molecular scaffold and replacing some parts with metallorganic moieties. In comparison, there are only few studies where metal complexes constitute the starting entity to be further developed towards specific binders. A notable study by Morrison et al. focussed on the development of a metallofragment library as the starting point for a fragment-based drug discovery approach (Morrison et al. 2020). The authors created a small library of 71 metal-containing compounds that are mainly three-dimensional in shape: the library broadly populates the topological 3D space (77% of the compounds are considered 3D), while conventional fragment libraries mainly consist of planar or linear compounds (∽25% considered 3D). Higher shape diversity in screening libraries has been correlated to broader biological activity, and diversifying fragment libraries with metal-containing compounds to increase shape diversity is therefore expected to improve the outcome of primary screens (Sauer and Schwarz 2003).

Morrison and coworkers run primary biochemical screens with their 71-compound metallofragment library on three therapeutically relevant targets, including the New Delhi metallo-β-lactamase 1 (NDM1-1), the PA N-terminal (PA_N) endonuclease domain of the RNA-dependent RNA polymerase complex of the influenza A virus and the N-terminal domain of heat shock protein 90-α (Hsp90). High hit rates (15–40%) were observed as compared to purely organic fragment libraries (3–30%), and while a series of fragments showed unspecific activity, several fragments specifically inhibited only one target. Dose-response curves of a subset of compounds confirmed that hits with IC₅₀<25 µM were found for each target (Morrison et al. 2020). These results demonstrate the suitability of a small metallofragment library to find initial fragment hits for protein targets of interest and indicate that metals are a powerful means to augment shape diversity in fragment libraries, which might further improve screening outcomes.

In future, the crucial step to elaborate metallofragment hits to actual lead compounds with high binding affinity and specificity will need to be approached to demonstrate that these hits can be the starting point of successful lead development campaigns. Fragment-based drug discovery starting from metallofragments will likely help increasing the accessible scope of 3D shape in drug discovery, which might help improving clinical success (Lovering et al. 2009; Prosser et al. 2020). Likely, this approach, which puts the metal-containing moiety at the centre of the inhibitor development process, will also allow for the rational development of bioactive compounds that rely on metal-specific modes of action.

Metal complexes as photosensitisers for antimicrobial photodynamic therapy

A common approach in the development of new antibacterials is to develop inhibitors targeting proteins that are specific to bacteria and not present in mammals. This minimises interference with the metabolism of the host and possible side effects of antibacterial treatment. There are however some approaches that allow for the selective targeting of bacteria that do not rely on specificity for a target protein. Such an alternative approach, however not covered in this review, is the use of light activated carbon monoxide-realising molecules (PhotoCORMs) as antibacterials (Betts et al. 2017; Rana et al. 2017; Tinajero-Trejo et al. 2016).

Antimicrobial Photodynamic Therapy (aPDT), also called Antimicrobial Photodynamic Inactivation (APDI), is an alternative non-antibiotic approach to combat bacterial infection. This non-invasive technique relies on the accumulation of a photosensitiser (PS) in target bacteria allowing for site-specific production of reactive oxygen species (ROS) with light (Hamblin and Hasan 2004; Plaetzer et al. 2009).

The process to generate toxic ROS requires two stages and is initiated when a preferential non-toxic PS is irradiated with light of a specific wavelength. This leads to the excitation of the PS into an electronic excited singlet state S₁. This S₁ state is short-lived and can either decay back to the ground state (S₀) via the emission of light (fluorescence) or through the release of heat (internal conversion). The S₁ state can also undergo a process known as intersystem crossing (ISC) to form a more stable excited triplet state (T₁). This T₁ state can again decay to the singlet ground state (S₀) via a non-radiative decay (phosphorescence) or react with its environment by two processes. In a type I mechanism, the PS in its T₁ state can undergo an electron transfer reaction, leading to the production of cytotoxic species such as superoxide, hydrogen peroxide and hydroxyl radicals. Alternatively, the T₁ state can directly react with molecular oxygen (³O₂) by an energy transfer reaction generating toxic singlet oxygen (¹O₂) (type II mechanism) (Dolmans et al. 2003). The generated reactive oxygen species (ROS) can damage the bacterial cell, resulting in functional inactivation of vital cellular processes and ultimately bacterial death (Figure 3). Hamblin and Hasan proposed that generated ROS cause bacterial killing and bacterial cell damage via three mechanisms: (1) damage of the cell membrane, (2) inactivation of proteins, and/or (3) damage of DNA (Figure 3A) (Kashef and Hamblin 2017).

Figure 3:

Schematic representation of the photophysical processes and mechanism of action of aPDT with representative examples of metal complexes acting as photosensitisers.

(A) Adapted Jablonski diagram to illustrate the photochemical and photophysical principals of antimicrobial Photodynamic Therapy (aPDT). S₀: ground singlet state of PS molecule, S₁: excited singlet state of PS molecule, T₁: triplet excited state of PS molecule, ³O₂: ground state oxygen, ¹O₂: singlet oxygen, O₂^−.: superoxide anion, HO^.: hydroxyl radical, H₂O₂: hydrogen peroxide. (B) Structures of the Ru(II)-based photosensitisers TDL-1411 and TDL-1433. (C) Structures of the Ru(II) polypyridyl complexes 5 and 6. (D) Structures of the Re(I) complexes 7–9.

In 2013, Lei et al. first reported the photoinactivation of Gram-negative E. coli bacteria by testing a series of [Ru(byp)₂(dppn)]²⁺ (byp = 2,2′-bipyridine, dppn = 4,5,9,16-tetraazadibenzo[a,c]naphthacene) complexes. Reduction in colony-forming units (CFU) of up to 70% at 0.1 μM was observed after an irradiation period of 5 min by visible light (>450 nm) (Lei et al. 2011).

Similarly, McFarland, Lilge and colleagues evaluated the photodynamic antibacterial effect of two Ru(II)-based photosensitisers (TDL-1411 and TDL-1433) against a strain of S. aureus (ATCC 25923) and a methicillin-resistant strain of S. aureus (MRSA, ATCC 33592) (Figure 3B). Interestingly, TDL-1433 has entered clinical phase II trials in 2019 as novel PS for the treatment of non-muscle invasive bladder cancer (Theralase Inc 2019). Both compounds were investigated in the dark and after exposure to 530 nm light (90 J/cm²) under normoxic (ambient atmosphere) and hypoxic (0.5% O₂) conditions. While both TDL-1411 and TDL-1433 lead to a reduction in CFU against both bacterial strains under normoxic condition, TDL-1433 was even more active in an environment that had less oxygen. The researchers concluded that apart from the ability of following a type II mechanism under normal oxygen levels, TDL-1433 is able to switch to a type I oxygen-independent mechanism (Arenas et al. 2013). Gasser and co-workers reported two Ru(II) polypyridyl complexes, 5 and 6, which were tested against both Gram-positive S. aureus and Gram-negative Escherichia coli bacterial strains after 420 nm light irradiation (Figure 3C). Both complexes let to a >6 log₁₀ reduction in bacterial viability against S. aureus at a concentration of 50 μM and a light dose of 8 J/cm². However, only complex 6 was able to affect the viability of the Gram-negative E. coli strain (99.99% reduced viability) at the same concentration and light dose (Frei et al. 2014). Recently, Blaskovich, Frei and co-workers reported a series of three rhenium bisquinoline complexes (7–9) where they combined the ability of such rhenium complexes to undergo a dual mechanism of action (Figure 3D) (Frei et al. 2020a). Previous work by Metzler-Nolte on tri-metallic antimicrobial peptides highlighted the importance of the rhenium tricarbonyl core for the antibacterial activity (Patra et al. 2014). Complementarily, Leonidova et al. investigated the use of such type of rhenium tricarbonyl complexes as novel chemotherapeutic PS and their ability to generate singlet oxygen (Leonidova et al. 2014). By combining the reported antibacterial activity of the Re(Co)₃ core with its ability to generate singlet oxygen, the researchers validated their hypothesis by testing 7–9 against both Gram-positive (S. aureus ATCC 25923) and Gram-negative strains (E. coli ATCC 25922), along with methicillin-resistant S. aureus (MRSA) and colistine-resistant E. coli in the dark and after UV light irradiation (365 nm, 3 J/cm²) for 1 h. Overall, all three complexes had a 4- to 16-fold lower MIC with light than without against E.coli and complex 7 was also effective on resistant bacterial strains (Frei et al. 2020a).

aPDT is an attractive alternative approach to design non-classical antibiotics that act on and interfere with bacterial viability through different light-activated pathways. The generation of lethal ROS through light activation is however limited by the fact that the majority of PSs have only a marginal absorption in the biological optical window (600–900 nm). As light of lower wavelength has a lower tissue penetration depth, PSs outside of this window may only be utilised for topical application (Heinemann et al. 2017). Furthermore, questions remain to be answered regarding the underlying mechanisms that ultimately result in bacterial cell death, viable strategies to increase the accumulation of PSs in pathogens and the possible development of tolerance to aPDT.

Recent reports on the development of bacterial tolerance to aPDT have sparked the interest of several research groups, and a number of divergent articles have covered the controversial question if and how bacteria can develop tolerance to a process that is thought to affect a variety of different cellular targets (Bartolomeu et al. 2016; Giuliani et al. 2010; Kashef and Hamblin 2017; Rapacka-Zdonczyk et al. 2019). In a recent study, Snell et al. have identified mechanisms that could contribute to aPDT tolerance in S. aureus (Snell et al. 2021). Tolerance of bacteria to aPDT remains debated and several aspects need to be addressed in the future.

Metabolome-dependent catalytic antibacterial prodrug activation with metal complexes

While aPDT relies on the ability of metal complexes to mediate photocatalytic reactions and the selective accumulation of these complexes in target bacteria, there are many imaginable ways how the catalytic propensities of metal-containing compounds can be harnessed. A recent study by Weng et al. proposes a new way that could be used to achieve selectivity for some bacteria by relying on differences in their metabolome (Weng et al. 2020).

In many bacteria, formate is a central metabolite present at mM levels, while its concentration in mammals is orders of magnitude lower (Leonhartsberger et al. 2002; Pietzke et al. 2020). Weng et al. posited that they could rely on the higher formate levels in some pathogenic bacteria as an internal hydride source for an intracellular transfer hydrogenation reaction mediated by ruthenium complexes. This could for example be used to generate an antibiotic compound from an inactive prodrug specifically within a bacterial target.

A previously developed coordination-driven three-components-assembly was used to generate a library of 768 Ru–Arene Schiff-base complexes, which they screened for a fluorogenic transfer hydrogenation of azide-coumarin to the corresponding amine with formate as the hydride source (Chow et al. 2014; Weng et al. 2020). They selected a catalytically active Ru-complex for further work and showed in fluorescence microscopy studies that they could selectively generate the fluorescent coumarin amine from its azide precursor in a fluorogenic reaction inside S. aureus cells. They also treated different bacterial strains with a combination of an azide-caged prodrug of the antibacterial sulfanilamide and the Ru-complex. In E. coli and S. aureus, which have high intracellular formate levels, the combination of prodrug and ruthenium catalyst resulted in an 8-fold increase in bacterial growth inhibition as compared to treatment with the azide prodrug alone (IC₅₀ values decreased from 2583 to 323 µM for E. coli and from 1292 to 161 µM for S. aureus). In M. smegmatis, which has low intracellular formate levels, the combination of prodrug and catalyst was only 2-fold more active than the prodrug alone (IC₅₀ values decreased from 161 to 81 µM). These results indicate a putative dependency of prodrug activation on intracellular formate levels.

The concept of relying on metabolites unique to bacteria to summon bactericidal effects without affecting host organisms opens diverse new possibilities for which the catalytic propensity of metal-containing compounds is prone to play a central role. More work remains to be done to confirm the viability of the approach set out in this proof-of-concept study. Likely, future focus will lie on optimising the approach harnessing bacterial formate in terms of activity gain and specificity, but also on exploring similar approaches relying on other bacteria-specific metabolites.

Future perspectives

In a recent report on antibiotic resistance, the WHO states that “[w]hile there are some new antibiotics in development, none of them are expected to be effective against the most dangerous forms of antibiotic-resistant bacteria” (WHO|Antibiotic resistance 2020). It is thus clear that research towards antibacterial drugs with novel modes of action needs to be intensified. Ideally, this will lead to a series of shelf-ready new antibiotics that will only be used as a last resort in very rare cases. From an economic point of view, this means that expected revenues from novel antibacterial drugs are in no relation to the necessary investment. Therefore, it will be crucial to find solutions to incentivise and sustain their development and translation to the clinics, for example by establishing new revenue models (Årdal et al. 2020).

Importantly, recent reviews of current antibiotics development programmes have concluded that while the clinical pipeline predominantly consists of derivatives of well-established antibiotic classes, the preclinical pipeline is more diverse, with a considerable number of projects focussing on new targets, classes, and modes of action (Theuretzbacher et al. 2020a,b). This difference between the pre-clinical and the clinical pipeline indicates that it is difficult to translate drug candidates with novel modes of action to the clinical stage, as tapping into new territory likely brings new challenges that have not been addressed before.

This also needs to be considered for the development of metal-containing antibiotics, as only a few examples have yet reached clinical trials. Clearly, significant collective efforts as well as resources will need to be put in translating initial discoveries to shelf-ready antibiotics. Yet, several metal-based pharmaceuticals against other diseases are on the market, and the pre-clinical and clinical development of metal-containing antibiotics will likely benefit from lessons learned with other metal-based drug candidates (Boros et al. 2020). As for these other diseases, it is expected that metals – with the breadth of new modes of actions that they can bring about as components of antibacterial agents – have the potential to make a difference in the search for next-generation therapeutics to cure bacterial infections.

Future focus in the field of metal-based antibacterials research will lie on getting an in-depth understanding of the pharmacokinetics and modes of action of the various compounds that are under development. Thorough studies of stability and aqueous chemistry of metal-based drug candidates will allow pinpointing their active species as well as potentially harmful products and metabolites they could form. A clear picture of the interaction of the compounds with biological systems – the bacterial pathogen as well as the host – will allow to understand how their drug properties can be fine-tuned.

The growing knowledge of the chemistry and biology of metal-containing compounds will fuel the recently described paradigm shift in metal-based drug discovery from the initially followed route of first identifying active compounds, then investigating possible mode of actions, towards an approach that is more rationally driven i.e. by relying on a specific mechanism of action (Boros et al. 2020). A rationalisation of the discovery process will likely streamline the early development pipeline and propel the advancement of metal-containing antibiotics. As demonstrated with different examples in this minireview, the manifold opportunities metals provide to tackle problems in medicinal chemistry constitute a diverse foundation to draw from to address the need for new generations of antibiotics.