Metallo-β-Lactamases: Structure, Function, Epidemiology, Treatment Options, and the Development Pipeline

Modern medicine is threatened by the global rise of antibiotic resistance, especially among Gram-negative bacteria. Metallo-β-lactamase (MBL) enzymes are a particular concern and are increasingly disseminated worldwide, though particularly in Asia. Many MBL producers have multiple further drug resistances, leaving few obvious treatment options. Nonetheless, and more encouragingly, MBLs may be less effective agents of carbapenem resistance in vivo, under zinc limitation, than in vitro.

ABSTRACT

Modern medicine is threatened by the global rise of antibiotic resistance, especially among Gram-negative bacteria. Metallo-β-lactamase (MBL) enzymes are a particular concern and are increasingly disseminated worldwide, though particularly in Asia. Many MBL producers have multiple further drug resistances, leaving few obvious treatment options. Nonetheless, and more encouragingly, MBLs may be less effective agents of carbapenem resistance in vivo, under zinc limitation, than in vitro. Owing to their unique structure and function and their diversity, MBLs pose a particular challenge for drug development. They evade all recently licensed β-lactam–β-lactamase inhibitor combinations, although several stable agents and inhibitor combinations are at various stages in the development pipeline. These potential therapies, along with the epidemiology of producers and current treatment options, are the focus of this review.

INTRODUCTION

Antimicrobial therapy is threatened by the global rise of resistance, especially in Gram-negative bacteria (1), where resistance to β-lactams is largely mediated by β-lactamases (2). Carbapenems evade most β-lactamases but are hydrolyzed by metallo-β-lactamases (MBLs) as well as by a few active-site serine β-lactamases (SBLs), notably members of the KPC and OXA-48-like groups. MBLs are chromosomal and ubiquitous in some nonfermenters, including Stenotrophomonas maltophilia, Aeromonas spp., and Chryseobacterium spp., which are of modest clinical concern. A minority of Bacteroides fragilis strains have a chromosomally encoded MBL, CfiA or CcrA, but this is uncommon and is expressed strongly only if an upstream insertion sequence provides an efficient promoter (3). More important are the acquired MBLs that are spreading among members of the Enterobacterales and Pseudomonas aeruginosa (4); these are associated with extremely drug-resistant (XDR) phenotypes, with the producers generally also being resistant to multiple aminoglycosides, fluoroquinolones, and other agents as well as to β-lactams.

CLASSIFICATION AND DIVERSITY OF METALLO-β-LACTAMASES

β-Lactamases are classified by two major systems. The first is based on substrate profiles and vulnerability to inhibitors (5) and places MBLs in its group 3, whereas groups 1 and 2 comprise SBLs. The second classifies β-lactamases according to their amino acid sequences, recognizing four enzyme classes (6). MBLs form class B, while SBLs are divided among classes A, C, and D (7). The MBLs are structurally and mechanistically dissimilar from SBLs, suggesting a separate evolutionary origin.

Class B enzymes are further divided into three subclasses—B1, B2, and B3—based on differences in amino acid sequence at the active site, zinc ligands, zinc stoichiometry, loop architecture, and substrate profiles (8). The important acquired MBLs, comprising the IMP, NDM, and VIM types, fall into subclass B1. They hydrolyze all currently available β-lactam antibiotics except monobactams (e.g., aztreonam) (9), as do most or all other subclass B1 or B3 enzymes. In contrast, the CphA (subclass B2) MBLs of Aeromonas spp. have narrow-spectrum activity directed exclusively against carbapenems. Irrespective of subclass, MBLs are not inhibited by clavulanic acid, sulbactam, tazobactam, or avibactam or by developmental penicillanic acid sulfones and diazabicyclooctanes.

The important acquired subgroup B1 MBLs (Table 1) are mostly named based on where they were first described; thus, for example, Verona integron-encoded metallo-β-lactamase (VIM) and New Delhi metallo-β-lactamase (NDM). The first acquired MBL (imipenemase; IMP-1), was reported from clinical isolates of P. aeruginosa and Serratia marcescens in Japan in the 1990s (10), and its family now includes over 85 sequence variants (11). The first VIM enzyme was found in P. aeruginosa in 1997 (12), with over 69 variants since described (11). NDM—now the most prevalent MBL in Enterobacterales and Acinetobacter baumannii—was first identified in 2008 in Klebsiella pneumoniae and Escherichia coli isolates from a patient who had travelled to Sweden from New Delhi, India (13). Twenty-nine NDM variants have since been described, (11).

TABLE 1.

Examples of chromosomal and plasmid-associated MBLs (11)

MBL type	Species	Enzyme(s)	Subclass
Chromosomal	Bacillus cereus	BcII	B1
	Chryseobacterium indologenes	IND	B1
	Elizabethkingia meningoseptica	BlaB	B1
	Myroides odoratimimus	MUS and MYO	B1
	Bacteroides fragilisa	CfiA/CcrA	B1
	Aeromonas spp.	CphA	B2
	Stenotrophomonas maltophilia	L1	B3
	Elizabethkingia meningoseptica	GOB	B3
Plasmid associated		Verona integron-encoded metallo-β-lactamase (VIM)	B1
		New Delhi metallo-β-lactamase (NDM)	B1
		Imipenemase (IMP)	B1
		Sao Paulo metallo-β-lactamase (SPM)	B1
		German imipenemase (GIM)	B1
		KHM	B1
		Dutch imipenemase (DIM)	B1
		Serratia metallo-β-lactamase (SMB)	B3
		Adelaide imipenemase (AIM)	B3

^aUnlike most other chromosomal MBLs, the Bacteroides fragilis enzyme is rare in the species.

It is easy to be dismissive of the chromosomal subclass B2 and B3 MBLs, but recent reports highlight Stenotrophomonas maltophilia as a multidrug-resistant pathogen in immunocompromised hosts (14). S. maltophilia carries a subclass B3 MBL (L1 enzyme), which is unique among MBLs in having four identical subunits (15), in addition to a chromosomally mediated SBL (L2 enzyme). This combination confers resistance to almost all β-lactams, although MICs vary with methodology, because media affect the expression and/or function of these enzymes (16). Elizabethkingia meningoseptica has two chromosomal MBLs, a B1 enzyme (BlaB) and a B3 type (GOB), with the former predominantly contributing to resistance (17).

GENETIC SUPPORT OF ACQUIRED MBLS

Acquired IMP and VIM enzymes generally are encoded by gene cassettes within class 1 or class 3 integrons. These may be embedded within transposons, allowing insertion into the bacterial chromosome or plasmids (18). In contrast, the bla_NDM gene is not integron associated and has been observed on narrow-host-range plasmids belonging to incompatibility group IncF, in addition to wide-host-range plasmids belonging to IncA/C, IncL/M, IncH, and IncN (19–22). K. pneumoniae and E. coli are the frequent hosts of these plasmids, and there are particular associations with K. pneumoniae sequence type 11 (ST11), ST14, ST15, and ST147 and E. coli ST167, ST410, or ST617 (23). These should not, however, be seen as global epidemic strains along the lines of K. pneumoniae ST258 variants with KPC carbapenemases, for many are common STs without carbapenemases. In A. baumannii, the bla_NDM-1 gene is generally located within the composite transposon Tn125 and embedded between two copies of a strong promoter gene ISAba125 (24, 25); it is much less prevalent in this genus than are OXA carbapenemases (class D).

B2 and B3 MBLs are generally chromosomally encoded, ubiquitous in their host species, and not transmissible. However, exceptions exist, with horizontal transfer having been observed. Thus, the AIM-1 MBL (B3) was initially reported, in 2012, to be encoded by a gene inserted in (and atypical of) the chromosome of a P. aeruginosa isolate; subsequently, in 2019, it was reported from K. pneumoniae (26). The bla_LMB-1 gene, encoding another subclass B3 enzyme, was reported to be located on a plasmid in Rheinheimera pacifica, where it was flanked by ISCR mobilization sequences, implying transfer from some other (unknown) source organism. (27). Mobilization of bla_SMB-1, encoding a third subclass B3 enzyme, has occurred similarly (28).

STRUCTURE AND CATALYTIC FUNCTION OF MBLS

Irrespective of subgroup, MBLs contain the αβ/βα fold typical of the metallo-hydrolase/oxidoreductase superfamily (29). The S. maltophilia enzyme has four identical subunits (15), whereas other MBLs are monomeric.

B1 and B3 MBLs have a shallow active-site groove containing 1 or 2 catalytically functional divalent zinc ions, flanked by flexible loops (29). In contrast, the B2 enzymes have an active site that is less accessible and flanked by a helix (30). Except for these consistencies, MBLs are highly divergent even within subclasses and have as little as 20% sequence identity between subclasses (7). Figure 1 illustrates the amino acid residues that bind zinc at the active sites of B1, B2, and B3 MBLs (8).

FIG 1.

Amino acid residues that bind zinc at the active sites of B1, B2, and B3 MBLs. (A) Crystal structures of B1 enzymes, including IMP, VIM, NDM, and B. fragilis CcrA, reveal two zinc-binding sites (Zn1 and Zn2). The Zn1 site contains three histidine residues (His116, His118, and His196), whereas the ligands for the Zn2 site are aspartic acid (Asp120), cysteine (Cys221), and histidine (His263). (B and C) There is only one zinc ion in the active site of the Aeromonas hydrophila enzyme (subclass B2) (B) and two in the active site S. maltophilia enzyme (subclass B3) (C). (Republished with permission from reference 8.)

Mechanistically, the zinc ion(s) activates a water molecule, which acts to open the β-lactam ring (31). There is no covalent intermediate, as with SBL-mediated catalysis. Anionic intermediates have been characterized when MBLs hydrolyze carbapenems (32), but not when NDM-1 enzymes hydrolyze penicillins or cephalosporins (33). In general, imipenem and meropenem are similarly good substrates for MBLs: for example, NDM-1 displays similar catalytic activity, reflected in values of the k_cat/K_m ratio, for imipenem (0.09 μM⁻¹ s⁻¹) and meropenem (0.06 μM⁻¹ s⁻¹) (34); biapenem is a weaker substrate, owing to high K_m values, but seems unsuitable for high-dose development (35).

Differences in assay methodology between workers make it difficult to compare hydrolytic efficiencies for different MBLs. Variation within, e.g., the VIM, IMP, Sao Paulo metallo-β-lactamase (SPM), and German imipenemase (GIM) family appears largely inconsequential (36). Nevertheless, subtle but important evolution may be ongoing, as illustrated in the NDM family. Here, experimental data do not define major differences in the catalytic efficiencies among NDM-1, -3, -4, -5, -6, -7, and -8 enzymes (37) under standard conditions, but differences are seen under zinc deprivation. Thus, studies comparing NDM-1, NDM-4 (Met154Leu), and NDM-12 (Met154Leu, Gly222Asp) demonstrate that the Met154Leu substitution, present in 50% of clinical NDM variants in some locales, enhances the ability to confer resistance at low Zn²⁺ concentrations (38, 39). This is potentially important because, as discussed below, zinc is restricted in infection (40) and its scarcity may impede the ability of classical NDM-1 enzyme to confer clinical resistance. NDM variants that have increased affinity for zinc (up to ∼10-fold-decreased dissociation constant for zinc [K_{d, Zn2}]) display selective advantages in experiments that mimic zinc scarcity imposed by the host immune system (41). Perhaps driven by similar pressures, the NDM-15 variant has evolved to function efficiently as a mono- rather than a bi-zinc enzyme (41). In addition, there are suggestions that NDM enzymes are evolving to develop greater thermodynamic stability (37).

EPIDEMIOLOGY AND DISTRIBUTION OF ACQUIRED MBLs

Bacteria with IMP, VIM, and NDM enzymes have been identified in a range of community, hospital, and environmental settings (42). Their prevalence and their importance relative to serine carbapenemases vary greatly by country. Figure 2 illustrates the global distribution of acquired MBLs.

FIG 2.

Global distribution of acquired MBLs.

Indian subcontinent, Asia, and Russia.

The greatest burden of acquired, plasmid-mediated MBLs lies in South and Southeast Asia (43), where NDM types are prevalent. As already noted, bla_NDM-1 was first identified in bacteria isolated in 2008 from a patient who had travelled to Sweden from India (44). NDM variants have subsequently been spread worldwide via patient transfers and travel (45). Epidemiological surveillance has confirmed that NDM-1 and its variants are widely disseminated throughout India, Pakistan, and Bangladesh (46, 47); moreover, a review of 39 carbapenem-resistant Enterobacterales (CRE) collected in India in 2006-2007 by the SENTRY Antimicrobial Surveillance Program found that 15 harbored bla_NDM-1 (48), indicating that it was circulating prior to its “discovery” in 2008. Enterobacterales with bla_NDM were isolated from public tap water in India (49) and in river systems around pilgrimage sites (42), demonstrating that the gene has become established beyond health care environments.

In India, there is frequent cocarriage with other carbapenemases in Enterobacterales (50); thus, in 2012, of 113 nonclonal CRE isolates at a Mumbai hospital, 106 produced NDM enzymes, and 21 of these also had a second carbapenemase, most often an OXA-48-like (n = 17) or VIM-type (n = 4) enzyme. Surprisingly, given that most international reports of NDM enzymes relate to Enterobacterales, P. aeruginosa was the most common MBL host (24%) among 3,414 carbapenem-resistant Gram-negative bacteria collected from community and hospital settings in North India (51), with bla_NDM-1 (36%) being the most prevalent carbapenemase gene, followed by bla_VIM (18.4%).

Although KPC is the principal carbapenemase among Enterobacterales (CPE) in China, a survey across 25 provinces showed that 32% of phenotypic carbapenem resistance in Enterobacterales was linked to bla_NDM-1 (52), while a study (2012 to 2016) of clinical Enterobacter cloacae across three tertiary hospitals found bla_NDM-1 to be the most common carbapenemase gene (80%), followed by bla_IMP-26 (8%) and bla_IMP-4 (6%) (53). The importance of IMP MBLs, particularly IMP-4, in China has been underscored by others; thus, multiple Enterobacterales species carrying a plasmid encoding IMP-4 enzyme were identified from patients with epidemiological links to China (54), and surveillance at a Beijing hospital highlighted both IMP-4 and NDM-1 in K. pneumoniae (55). Colocalization of bla_NDM-9 and the plasmid-mediated colistin resistance gene mcr-1 was seen in an E. coli strain recovered from retail chicken meat in Guangzhou, China (56). Having been recognized 30 years ago in Japan, IMP-type enzymes are now endemic there, though not highly prevalent (57).

NDM MBLs are the second most prevalent carbapenemases after OXA-48 in the Middle East, excepting Israel (58, 59). This probably reflects extensive interactions with the Indian subcontinent. As in India, there is significant penetration of bla_NDM into P. aeruginosa, for which a much greater proportion of carbapenem resistance appears to be carbapenemase mediated in the Middle East than in Europe or the United States. Thus, in the Gulf Cooperation Council countries, bla_VIM was found in 39% of carbapenem-resistant P. aeruginosa isolates (60), with most hosts belonging to internationally disseminated high risk clones, including ST235, ST111, ST233, ST654, and ST357 (60). These lineages seem unusually adept at acquiring extrinsic resistance genes. In Dubai, 32% of resistant P. aeruginosa isolates produced VIM-type MBLs (61), though a larger proportion had outer membrane impermeability.

The proportion of carbapenem-resistant P. aeruginosa harboring MBLs in Russia rose from 4.5% between 2002 and 2004 to 28.7% between 2008 and 2010 (62), largely reflecting the spread of an XDR bla_VIM-2-positive ST235 high-risk clone, also present in Belarus and Kazakhstan (62). NDM is reported as the predominant carbapenemase among Enterobacterales in St. Petersburg (63, 64), whereas OXA-48 is predominant in Moscow (65).

Europe.

Although Italy had earlier reported both IMP and VIM enzymes (66), Greece was the first European country to report extensive dissemination of Enterobacterales with MBLs. Specifically, K. pneumoniae with VIM carbapenemases were reported from multiple hospitals in 2003 to 2007, and multilocus sequence typing identified three major clonal complexes (CCs)—CC147, CC18, and CC14—among the producers (67). By 2006, 20% of K. pneumoniae isolates collected from hospital wards and 50% of those from intensive care units (ICUs) monitored by the Greek System for the Surveillance of Antimicrobial Resistance were carbapenem resistant, largely owing to the spread of the bla_VIM-1 cassette (68). By 2010, KPC had displaced VIM to become the dominant carbapenemase in Greece, largely through the spread of a K. pneumoniae ST258 variant (69). Nonetheless, VIM types remained scattered, and they may now be reemerging due to suppression of the KPC carbapenemases via the use of ceftazidime-avibactam (70).

Elsewhere in Europe, concern about carbapenemases grew following a flurry of press interest in NDM enzymes from 2008 to 2010 and with the spread of K. pneumoniae ST258/512 lineages with KPC carbapenemases in Italy from 2010. The United Kingdom, taken as an example, recorded a few P. aeruginosa and Enterobacterales strains with IMP and VIM MBLs before 2008. Thereafter, the number of Enterobacterales with NDM enzymes increased (46). Most early cases were imports via patients who had travelled to (and often been hospitalized in) the Indian subcontinent. Multiple NDM variants have subsequently been reported in the United Kingdom, with NDM-1 being the most frequent among Enterobacterales, followed by NDM-5 and NDM-7 (71). In contrast, VIM variants account for 91% of the (uncommon) MBLs in P. aeruginosa, again associated with international high-risk clones, including ST235, ST111, ST233, and ST357 (72).

While referral of CPE isolates to the national reference laboratory has increased 100-fold since 2008, many producers are from screening rather than clinical samples. OXA-48 is now the fastest-spreading carbapenemase, but isolates with NDM enzymes account for 20 to 25% of CPE submitted. A growing minority of these, particularly E. coli, have both NDM- and OXA-48-like enzymes (71, 73).

In 2012, the European Centre for Disease Prevention and Control launched its European Survey of Carbapenemase-Producing Enterobacteriaceae (EuSCAPE) project. The geographic distribution of enzyme types was estimated by national experts across 38 European countries in 2015 (74). A random sample of carbapenem-susceptible and -nonsusceptible K. pneumoniae and E. coli strains subsequently were collected prospectively to determine the occurrence of carbapenemases (75). The results, published in 2017, revealed that SBLs (KPC or OXA-48 enzymes) were more prevalent than MBLs in most countries but that MBLs were widely scattered and were the most prevalent carbapenemases among Enterobacterales in a few countries. Thus, VIM enzymes were the dominant carbapenemases in Hungary, and NDM enzymes were dominant in Serbia and Montenegro. The prevalence of NDM enzymes in the latter countries tallies with early descriptions of producers linked to these Balkan states. It is unclear whether these originated as imports from India or as independent local gene escapes from the unknown source organism (76).

North America.

Infections due to Enterobacterales carrying bla_VIM-2, bla_VIM-7, bla_IMP-4, and bla_IMP-18 genes were recorded in the United States prior to 2005 but, in general, MBLs remained extremely rare (1, 77). In 2010, Enterobacterales harboring NDM-1 were isolated from three patients in different states (78) and, as with many contemporaneous cases in the United Kingdom and elsewhere, the source patients had all recently been in India or Pakistan (21). Subsequent expansion of NDM enzymes in the United States has been less marked than in the United Kingdom, with KPC carbapenemases becoming considerably more prevalent. Nevertheless, until December 2017, 379 CPE with NDM carbapenemases were reported to the CDC from 34 states, with just under a third (i.e., 109) from Illinois (79), where an outbreak was associated with contaminated endoscopes.

Enterobacterales with NDM enzymes have been increasing in Canada since 2008, and these MBLs are now the second most common carbapenemases in the country, with a higher prevalence in the western provinces (80). Surveillance conducted between 2007 and 2015 in Toronto revealed that, among 291 clinical CPE, 51% had NDM enzymes, and 24% of the patients had never received health care abroad or travelled to high-risk areas (81), suggesting that the enzymes are established locally. In 2019, a novel MBL, bla_CAM-1, was identified from isolates that were collected in 2007 (82). No subsequent isolates harboring this gene have been reported.

Africa.

Because of the paucity of data, the prevalence of CPE carrying MBLs in Africa is difficult to estimate. Apparent infrequency may reflect true rarity, limited sampling, or a lack of infrastructure for accurate detection. CPE with VIM MBLs nonetheless have been identified in Nigeria, Morocco, Algeria, Tunisia, Tanzania, and South Africa, and CPE with NDM enzymes have been found in Kenya, Nigeria, Morocco, Algeria, Tunisia, Tanzania, and South Africa (83, 84). Infections caused by Enterobacterales producing MBLs are reported from both imported and local cases, raising concerns regarding emerging endemicity (85). Those with IMP-type enzymes have been identified in small numbers in Morocco, Tunisia, and Tanzania and appear genuinely uncommon (84). An outbreak caused by Klebsiella spp. carrying bla_NDM-5 was reported from a neonatal unit in Nigeria (86). A concern is that African patients are strongly represented in medical tourism to India, which is a risk factor for colonization with Enterobacterales producing MBLs (87).

Rest of the world.

KPC enzymes dominate among carbapenemases from Enterobacterales in Latin America, with (unusually) some penetration also into P. aeruginosa. Nonetheless, Enterobacterales with NDM enzymes are endemic in Brazil, with several outbreaks reported (88). Early case reports of MBL-producing Enterobacterales in Latin America often concerned members of the Proteeae, including Providencia spp. and Morganella (89, 90), which are infrequent hosts of bla_NDM elsewhere. This creates a treatment issue, since these genera are inherently resistant to polymyxins and newer tetracyclines, which remain options against other MBL-producing Enterobacterales (below).

Unique to South America is the wide distribution in Brazil of P. aeruginosa with SPM-1 MBL (91), principally associated with an ST277 clone. Outcomes of severe infections with this clone are often poor, reflecting a lack of good treatment options (92).

Carbapenemases are rare in Australasia, but there is spread of bla_IMP-4 among Enterobacterales (93), as in parts of China. E. cloacae is a major host, with dissemination mediated by an IncHI2 plasmid (94). Production of IMP-4 has also been recorded in Salmonella spp. from domestic pets (95) and seagulls (96), but the significance of this is uncertain.

MBL FUNCTION IN RESISTANCE, IN VITRO AND IN VIVO

For many years MBLs were perceived as clinically unimportant chromosomally encoded enzymes from nonpathogenic organisms, notably Bacillus cereus (97, 98). This perception changed with the recognition that MBLs confer much of the resistance seen in Chryseobacterium spp. and E. meningoseptica (99) and with heightened awareness of the morbidity and mortality associated with S. maltophilia bacteremia (100, 101). Interest then escalated with the discovery and proliferation of acquired MBLs, especially NDM-1, which drew extensive press coverage in 2010.

Many MBL producers are broadly resistant in vitro and, on this basis, real concern exists about lack of treatments. On the other hand, there is evidence that in vivo resistance to carbapenems may be less than it appears in vitro, because susceptibility tests are conventionally done in media (e.g., cation-adjusted Mueller-Hinton broth) with high zinc concentrations (102), whereas the host immune system imposes a state of zinc deprivation in infection (40, 103). This lack of zinc may not only impede the catalytic function of MBLs but may also interfere with their protein folding (102) and may promote degradation of the enzyme in the periplasm (104).

Several preclinical studies suggest a disconnect between high-level in vitro resistance to carbapenems associated with NDM-1 enzymes and a weak ability to protect against carbapenems in standard murine infection models (105). Moreover, NDM enzymes appear to be less effective than other carbapenemases in causing resistance to carbapenems in patients (106, 107). Thus, mortality in severe infections due to Enterobacterales with bla_NDM appears to be relatively low, ranging from 13% (108) to 55% (109), compared to that seen with bacteria expressing other MBLs (18% to 67%) (13) or KPC carbapenemases (41% to 65%) (110, 111). Good clinical outcomes have been reported despite treatment with agents to which NDM enzymes confer resistance in vitro (106, 107, 112). As yet, there are no studies that confirm or refute whether the higher-numbered NDM alleles, encoding variants with greater affinity for zinc (above), are better able to cause clinical resistance than NDM-1 (39, 41).

Finally, it should be underscored that while these indications that NDM MBLs are less potent in vivo are intriguing, they should be approached with caution. Double-blinded randomized controlled trials have not been conducted, and existing outcome data are subject to various biases (113, 114). For VIM MBLs, clinical outcomes correlate with carbapenem MICs, implying little or no such in vitro-in vivo discordance (115).

CURRENT TREATMENT OPTIONS

Limited data exist to inform clinicians on the optimal treatment for infections caused by MBL-producing Gram-negative bacteria (106). Cotrimoxazole remains the standard of care for infections due to S. maltophilia, but most Enterobacterales with acquired MBLs also have sul and dfr genes, conferring resistance. Genes encoding resistance to fluoroquinolones and aminoglycosides are often present alongside genes encoding acquired MBLs. In particular, bla_NDM genes are often linked to the genes encoding ArmA or RmtB methyltransferases, which modify ribosomes to block binding of aminoglycosides, including plazomicin; bla_IMP and bla_VIM generally occur within integrons that often also carry aac(6′), encoding an acetyltransferase that compromises amikacin and tobramycin, though not gentamicin or plazomicin (116). A thorough review of treatment options for MDR and XDR Enterobacterales is available (117). This highlights observational studies comparing monotherapy to combination therapy for bloodstream infection (BSI) involving CRE, although few of these were specifically identified as having MBLs (118, 119).

Colistin.

Colistin is the current mainstay of treatment for infections due to MBL producers. A multinational survey of MBL-producing Enterobacterales and P. aeruginosa conducted from 2012 to 2014 found >97% susceptibility among MBL-producing P. aeruginosa strains (variously with IMP, VIM, and NDM enzymes) and >85% for MBL-producing Enterobacterales (>86.1% NDM type and >88.9% IMP type) (83). Exceptions are Proteeae and Serratia spp., which have intrinsic polymyxin resistance.

For bacteria harboring KPC and OXA-48 carbapenemases, colistin has recently been shown to be less effective than microbiologically active β-lactamase inhibitor combinations (120), making it plausible that an active β-lactam likewise would be more efficacious than colistin against MBL producers. Of note, the emergence of colistin resistance during treatment, with secondary transmission of resistant variants, is a concern (121, 122).

Tigecycline, omadacycline, and eravacycline.

The tetracyclines tigecycline, omadacycline, and eravacycline have strong in vitro activity against many MBL-producing Enterobacterales, except Proteeae, although not against P. aeruginosa. During November 2018, 275 unique Enterobacterales isolates carrying bla_NDM collected by the U.S. Centers for Disease Control and Prevention were tested with tigecycline (86.5% susceptible, based on a ≤2-μg/ml FDA breakpoint), eravacycline (66.2% susceptible, based on a ≤0.5-μg/ml FDA breakpoint), and omadacycline (59.6% susceptible, based on a ≤4-μg/ml breakpoint) (123). The higher susceptibility rate for tigecycline than eravacycline reflects the higher FDA breakpoint for Enterobacterales; in Europe, both agents have an identical 0.5-μg/ml breakpoint, and eravacycline is the more active on a simple gravimetric basis, though it is unclear whether this confers a clinical advantage (124). Merits of omadacycline are its minimal known drug interactions and its ability to be administered orally (125); however, it has the least relevant license (for community-acquired bacterial pneumonia and acute bacterial skin and skin structure infections) in relation to the clinical burden of MBL producers.

While the in vitro activity of these tetracyclines is encouraging, there are multiple caveats. First, tigecycline carries an FDA “black box” warning of increased mortality when the drug was used as monotherapy (126); second, both tigecycline and eravacycline have failed to achieve noninferiority to comparators in one or more clinical trials (ventilator-associated pneumonia and diabetic foot infection for tigecycline, complicated urinary tract infection [cUTI] for eravacycline); third, there is little provenance for tetracyclines as monotherapy in the severely ill patients who commonly develop infections due to MBL-producing opportunists; fourth, particularly for tigecycline, the disparity between EUCAST (≤0.5 μg/ml) and FDA (≤2 μg/ml) susceptibility breakpoints creates categorization uncertainty; last, the lack of anti-Proteeae activity is important in Latin America, where Providencia spp. are frequent hosts of bla_NDM (127). Given these uncertainties, the best advice is to consider using these tetracyclines in combination against MBL producers, not as monotherapy.

Aztreonam.

Aztreonam is stable to MBLs, though activity is lost against organisms that coproduce extended-spectrum β-lactamases (ESBLs) or AmpC enzymes (128), which are common in MBL-producing Enterobacterales. Clinical experience with aztreonam as monotherapy is lacking for MBL producers, although some success was recorded when aztreonam was used in combination with ceftazidime-avibactam (129, 130), with avibactam serving to inhibit ESBLs. Six of 10 patients survived following treatment with this combination during an outbreak of K. pneumoniae with NDM-1, OXA-48, and CTX-15 β-lactamases in Barcelona (129). Although no adverse events were reported, the safety is unclear, and it is difficult to match the regimen of aztreonam-avibactam (1.5 g + 0.5 g every six hours [q6h]) that is presently being developed (see below).

Fosfomycin.

Fosfomycin commonly retains full in vitro activity against MBL-producing Enterobacterales and has been successful in trials, very recently, as an i.v. agent in cUTI (131). It may be an option against MBL producers—particularly E. coli, which is more susceptible than other Enterobacterales—but it is mainly advocated for use in combination due to concerns about emergence of resistance, particularly in Klebsiella spp. (132). Fosfomycin has little direct antipseudomonal activity, with typical MICs above breakpoints. However, in vitro synergy is seen when fosfomycin is combined with meropenem against MBL-producing P. aeruginosa strains (133), suggesting a need for in vivo exploration.

DEVELOPMENT PIPELINE

The development pipeline represents four main strategies against MBL producers: (i) protection of MBL-stable-monobactams from other coproduced β-lactamases, as, e.g., with aztreonam-avibactam; (ii) development of β-lactams stable to MBLs as well as SBLs, as with, e.g., cefiderocol and BOS-228; (iii) combinations of cephalosporins and carbapenems with triple-action diazabicyclooctanes (DBOs); and (iv) direct inhibition of MBLs with cyclic boronates, thiols, chelators, dicarboxylic acids, and other agents.

Aztreonam-avibactam.

Aztreonam-avibactam is the first antibiotic to be developed under a public-private partnership agreement (134, 135), with partial financing from the European Union’s Innovative Medicine’s Initiative and, latterly, also the U.S. Biomedical Advanced Research and Developmental Authority (BARDA). A prospective randomized phase 3 study (NCT03580044) is scheduled to begin in 2020 to determine efficacy, safety, and tolerability versus best available therapy (BAT) for hospitalized adults with complicated intra-abdominal infections (cIAI), nosocomial pneumonia (NP), cUTI, or BSI due to MBL-producing Gram-negative bacteria (135).

Aztreonam evades hydrolysis by MBLs (128) but is compromised by the ESBL and AmpC enzymes that are coproduced by many MBL-positive CPE. These SBLs are inhibited by avibactam, a diazabicyclooctane (DBO) (136, 137), and consequently, MBL-producing Enterobacterales that also carry ESBLs or AmpC are susceptible to aztreonam-avibactam in vitro (138) and in vivo (139). The combination is less reliably active against MBL-producing P. aeruginosa (140), because aztreonam has weak antipseudomonal activity.

Considerable interest in this approach exists, because the safety and efficacy of aztreonam are well established and because avibactam was established to be effective at inactivating ESBLs and AmpC enzymes during trials with ceftazidime. Moreover, case reports suggest success against infections caused by MBL producers when aztreonam was coadministered with ceftazidime-avibactam (see “Aztreonam” above) (129, 130).

MBL-stable β-lactams.

(i) Cefiderocol (S-649266). Cefiderocol (S-649266) is a novel parenteral siderophore cephalosporin designed by Shionogi & Co., Ltd., with a catechol linked to its 3-position side chain. It is licensed in the United States for cUTI and in the European Union and United Kingdom for “treatment of infections due to aerobic Gram-negative organisms in adults with limited treatment options” (141). It is retained among developmental agents here, rather than being included in the established treatments, because there is little published experience with MBL producers to date (142, 143).

Critically, the catechol moiety forms a chelation complex with ferric iron, and this complex is actively accumulated by Gram-negative bacteria, which are forced to scavenge this essential element (144). Cefiderocol has good activity in vitro under iron starvation against Gram-negative bacteria, including CPE, P. aeruginosa, and A. baumannii (145). It is relatively stable to both SBLs and MBLs (144); however, the MICs for Enterobacterales and nonfermenters with NDM carbapenemases tend to be slightly higher than those for isolates of the same species with other carbapenemase types (146). Cefiderocol proved effective against carbapenem-resistant P. aeruginosa (expressing IMP-1 enzymes), A. baumannii (expressing OXA-51-like enzymes), and K. pneumoniae (expressing NDM-1 enzymes) in immunocompetent rat respiratory tract infection models, achieving a ≥3-log reduction in the number of viable bacteria in the lungs when given over 4 days so as to recreate the human exposures of an infusion regimen involving 2 g q8h given i.v. for 3 h (147). Efficacy decreased when the infusion time was reduced to 1 h, owing to a lower percentage of the dosing interval during which free-drug concentrations were above the MIC (T_f >MIC) (147). Interestingly, the mean T_f >MIC required for a 1-log₁₀ reduction was 18 to 24% greater for A. baumannii isolates (expressing OXA-23 or OXA-24) in the murine lung infection model than for Enterobacterales expressing NDM-1, NDM-4, or KPC-2 enzymes and for P. aeruginosa isolates expressing IMP-1 or VIM-10 MBLs (148).

In humans, the infusion regimen of 2 g q8h i.v. over 3 h provided >90% probability of target attainment (PTA) with 75% T_f >MIC for MICs of $\le$4 μg/ml for patients with normal renal function (149). A phase 3 trial (NCT03032380) has shown noninferiority to meropenem in nosocomial pneumonia (150). Less encouragingly, another trial (NCT02714595) found excessive deaths in the cefiderocol arm, compared with best available therapy, for patients with severe infections caused by carbapenem-resistant Gram-negative pathogens (151). Full analysis is awaited but, notably, deaths were mostly associated with Acinetobacter infections (152), not Enterobacterales.

(ii) BOS-228 (formerly LYS228). BOS-228 is a monobactam and, like aztreonam, is stable to MBLs (153). Unlike aztreonam, it is also stable to many potent SBLs, including carbapenemases, ESBLs, and AmpC types (154); BOS-228 binds to penicillin-binding protein 3 (PBP3) similarly to aztreonam, in addition to weak binding to PBP1a and PBP1b of Enterobacterales (155). BOS-228 had an MIC₉₀ of 2 μg/ml for a clinical panel of 88 Enterobacterales isolates expressing ESBLs, KPCs, and MBLs (153), and no single-step mutants were selected from 12 β-lactamase-expressing Enterobacterales exposed to the drug at 8× MIC, though mutants were selected from 2/12 strains, neither of which expressed MBLs, at 4× MIC (155).

A randomized evaluator-blinded multicenter phase 2 trial (NCT03354754) to evaluate pharmacokinetics, clinical responses, safety, and tolerability of BOS-228 in cIAI commenced in 2018. The drug is being administered as i.v. monotherapy (without metronidazole) q6h for at least 5 days and compared to standard of care, with outcomes evaluated at day 28. A randomized controlled evaluator-blinded multicenter trial (NCT03377426) in cUTI has also been initiated.

Cephalosporins or carbapenems combined with triple-action DBOs—zidebactam and nacubactam.

Unlike with cyclic boronates (see below), it has not been possible to discover DBOs that directly inhibit MBLs. However, nacubactam and zidebactam are DBO analogs that combine inhibition of SBLs with direct antibacterial activity by inhibiting PBP2 (156). When combined with PBP3-targetted β-lactams, this attack on PBP2 leads to an “enhancer” effect, with further β-lactamase inhibition-independent synergy observed (156, 157). Consequently, cefepime-zidebactam and cefepime- or meropenem-nacubactam combinations are active in vitro against >75% of MBL-producing Enterobacterales and, in cefepime-zidebactam’s case, also against many MBL-producing P. aeruginosa strains (158).

Although the direct antibacterial activity of nacubactam and zidebactam is readily lost via mutations compensating for inhibition of PBP2 (159), the enhancer effect is retained, with many of the mutants consequently remaining susceptible to, e.g., cefepime-zidebactam or meropenem-nacubactam at low concentrations (156, 157). Cefepime-zidebactam is currently the most advanced of these combinations, with a phase 3 trial due to commence (160).

Direct inhibitors of MBLs.

(i) Cyclic boronates—VNRX-5133 (taniborbactam) and QPX7228. Inhibitors that target both SBLs and MBLs are of great interest but have proved difficult to obtain owing to structural and functional differences between and among these enzymes. This combination of inhibitory activities nonetheless has recently been achieved with several cyclic boronates, notably taniborbactam and QPX7228. These mimic the tetrahedral anionic intermediate common to SBL and MBL catalysis (161) and additionally inhibit some penicillin-binding proteins (e.g., PBP 5, which is nonessential) by the same mechanism (162). They represent a considerable expansion in spectrum over vaborbactam, their progenitor, which inhibits only few class A β-lactamases, notably KPC types (163).

Taniborbactam (VenatoRx) is the more advanced of these two boronates and is in phase III trials combined with cefepime (164). It irreversibly inhibits class A, C, and D SBLs and is a reversible competitive inhibitor of VIM and NDM MBLs, though not of IMP types (165). Safety has been established in healthy volunteers (NCT02955459), and the FDA has allowed cefepime-taniborbactam to proceed via a fast-track pathway for the clinical indications of cUTI and cIAI. QPX7728 (QPEX) likewise inhibits both SBLs and MBLs: 50% inhibitory concentrations (IC₅₀) for KPC enzymes are around 2.9 ± 0.4 nM, compared with 22 ± 8 nM for the class C cephalosporinase of E. cloacae P99, 55 ± 25 nM for the NDM-1 MBL, and 14 ± 4 nM for VIM-1. As with taniborbactam, the IC₅₀ for IMP-1 is considerably higher, at 610 ± 70 nM (166). An i.v. combination of QPX7728 with meropenem is being explored. This significantly lowered bacterial counts in murine thigh and lung infection models with carbapenem-resistant K. pneumoniae, P. aeruginosa, and A. baumannii compared to meropenem alone, although strain genotypes were not reported. Unlike taniborbactam, QPX7228 is orally bioavailable, and combinations with ceftibuten and tebipenem were evaluated in vitro against CPE, including those with MBLs (167).

(ii) Thiol-containing MBL inhibitors and chelating agents. Small molecules that bind and/or chelate zinc ions include thiols, dicarboxylates, hydroxamates, and tetrazoles; these are widely reported to inhibit MBLs, but human metalloproteases are vulnerable too, so toxicity may preclude clinical development.

Thiol-containing compounds inhibit all MBL subtypes (B1, B2, and B3) (168), with strong competitive inhibition of IMP-1 by thioester derivatives first being reported in 1999 (169). The dipeptide l-captopril deserves mention in this context. It is used as an ACE (angiotensin-converting enzyme) inhibitor in the treatment of hypertension and is reported also to inhibit MBLs by chelating the active-site zinc ions via its thiol group (170); the corresponding d-stereoisomer is a more potent inhibitor and can potentiate meropenem against strains with VIM-2 MBLs (170). Both captopril isomers act via zinc chelation, and repurposing is attractive, given the known safety of the l-isomer at its licensed dose; however, the economic model for development is yet to be established, and safety issues for the d-isomer need exploration. Other thio-carbonyl compounds, such as thiomandelic acid, exhibit synergy with meropenem against Enterobacterales with VIM, NDM, and IMP enzymes (171).

Bisthiazolidines are carboxylate-containing bicyclic compounds, considered penicillin analogs that inhibit MBLs through a zinc-bridging thiol group and a carboxylate that interacts with K224 (172). The orientations of the carboxylate and thiol moieties create diverse binding that is observed on X-ray crystal structures and has been shown to inhibit all MBL types (173). The bisthiazolidine scaffold inhibits NDM-1 enzymes in vitro, with K_i values in the low micromolar range (from 7 ± 1 to 19 ± 3 μM); they restore imipenem activity against E. coli strains producing NDM-1 (172).

ANT2681 (Antabio) is the lead compound from a series of thiazole carboxylate derivatives which inhibit NDM and VIM through interaction with the dinuclear zinc ion cluster present at the active site (189, 190). It binds in a noncovalent competitive manner with respect to the substrate and is in late-stage preclinical development.

The divalent cation chelator EDTA has raised interest, too, both as an inhibitor of MBLs and because it disrupts the Gram-negative outer membrane and neutralizes various bacterial enzymes and toxins (174, 175). It is widely used in identification tests for MBLs. Sodium calcium EDTA, which is licensed for use for treatment of lead poisoning, reportedly restored imipenem’s activity in vivo against P. aeruginosa strains producing IMP and VIM enzymes and against E. coli strains producing NDM-1 enzyme (176, 177), raising the issue of whether it might be used to potentiate carbapenems in human infections. Elores, which is marketed in India, combines ceftriaxone, sulbactam, and EDTA (178, 179) and reportedly achieved cures of infections due to MBL producers in multiple patients, with no serious adverse events (178). However, prospective and controlled studies are lacking, the dose of EDTA is low, and there remains uncertainty (see above) about the function of NDM-1 in vivo. More negatively, the FDA has placed strict limits on the amount of EDTA permissible even in food (180), and sodium calcium EDTA is capable of producing toxic effects that can be fatal (181). High concentrations of EDTA are likely to strip divalent cations from human metalloenzymes, including matrix metalloproteinases, carbonic anhydrase, and carboxypeptidases, thus limiting clinical applicability.

Aspergillomarasmine A (AMA) is a fungal natural product discovered in the 1960s (182) and reevaluated in the 1980s as an inhibitor of the human metalloproteinase angiotensin-converting enzyme (ACE). AMA inhibits MBLs via a metal ion sequestration mechanism and displays rapid and potent inhibition of NDM-1 and VIM-2 enzymes in vitro (183). It restored the activity of meropenem against a K. pneumoniae strain expressing NDM-1 enzyme in an intraperitoneal murine infection model (184). Again, the hazard of inhibiting human metalloenzymes requires careful investigation.

CHALLENGES FOR THE DEVELOPMENT OF INHIBITORS OF MBLS

One of the biggest challenges in designing MBL inhibitors is the diversity among these enzymes, which share less than one-third sequence identity at their active sites. Thus, for example, taniborbactam and QPX7728 target NDM and VIM enzymes, but not IMP types (185). Development of inhibitors that bind remotely from the active site might overcome this limitation, but possible target areas also vary within class B1 and seem even better able to tolerate mutations than the active site (29). Another challenge is the shallow binding site in B1 enzymes, meaning that inhibitors can carry out only limited interactions (29). Specificity for bacterial MBLs is a further recurring challenge; interactions with human metalloenzymes and contingent toxicity are major concerns. Molecules that solely inhibit MBLs are limited by the fact that many MBL producers also coproduce SBLs, including carbapenemases, meaning that the partner β-lactam must evade these enzymes, that the inhibitor must inactivate both MBLs and SBLs, or that a second inhibitor is required.

Preclinical development is challenging, too, because it is difficult to establish reliable animal models in which MBL-mediated resistance is expressed, perhaps owing to the already-mentioned lack of essential zinc at infection sites. Moreover, bacteria are prone to lose MBL-encoding plasmids, or fail to reliably express them, in murine models, resulting in pharmacodynamic data that suggest meropenem susceptibility (186, 187). Consequently, it is difficult to establish the efficacy of candidate MBL-stable drugs or inhibitor combinations. It is unclear if the same phenomena occur in patients (188), and this requires further research. Irrespective of this aspect, it is also challenging to find and recruit the required number of patients with MBL-producing pathogens to clinical trials. Rapid diagnostics should help, but their use is complicated by cost and the need to deploy them to all trial sites, including in countries where they are not licensed or are licensed only to inform infection control, not treatment.

CONCLUSION

MBLs are disseminating internationally, particularly in Asia, and often are produced by Gram-negative bacteria with extremely broad spectra of in vitro resistance. Unlike for KPC and OXA-48-like carbapenemases, producers are typically not susceptible to recently licensed β-lactamase inhibitor combinations such as ceftazidime-avibactam, meropenem-vaborbactam, imipenem-relebactam, although cefiderocol may be a potential answer. The ability of MBLs to confer resistance to carbapenems may not be as great in vivo as in vitro, though this is uncertain and may vary by enzyme type even within MBL subclasses.

Inhibitors are known, and the developmental compounds boronates, taniborbactam, and QPX7728 are of particular interest. Nonetheless, the quest for effective inhibitors is complicated by differences in active-site structure and zinc ligand interactions among MBLs and by difficulties in the design of appropriate preclinical and clinical trials. Nonboronate inhibitors face toxicity issues, particularly if they interact with other metalloenzymes or are general chelators. Other approaches to overcoming MBLs include avibactam-protected aztreonam; stable β-lactams, notably BOS-228 as well as cefiderocol; and combinations of β-lactams with triple-action DBOs, notably cefepime-zidebactam and meropenem-nacubactam.

And that is the positive aspect on which to close: there is now a diverse and exciting pipeline of potential agents for the treatment of infections caused by bacteria that produce MBLs. It remains to be seen which will be the most effective of these agents.