Past and Present Perspectives on β-Lactamases

One of the most studied enzyme families is the group of enzymes known as β-lactamases, with more than 28,900 citations in Medline (https://www.ncbi.nlm.nih.gov/pubmed/). These enzymes, whose most obvious role is to inactivate β-lactam antibiotics, have driven research in academic laboratories since the early 1940s (1, 2). Perhaps more importantly, they have been responsible for scores of pharmaceutical research programs that have attempted to find ways to protect effective antibiotics from destruction. β-Lactamases provide intellectual challenges for academic investigators with their deceptively simple mechanisms of action, as enzymes whose hydrolysis activity approaches that of a “fully efficient enzyme” with diffusion-limited reaction rates (3). Attempts to provide economically viable anti-infective agents that can circumvent the action of these enzymes have yielded substantial economic rewards to successful pharmaceutical companies. As a result, the safe and effective β-lactam antibiotics have become one of the most widely prescribed classes of antibacterial agents (4). Many review articles have been published about various aspects of these intriguing enzymes (5–11), but they have tended to focus on narrowly targeted sets of β-lactamases. In this minireview, a short history of the β-lactamases is presented, emphasizing key points that have driven both academic and pharmaceutical science over the past 75 years. High points in their history will be discussed, based on the “Table of Firsts” shown as Table 1. This history is not meant to be a comprehensive review of all the current literature on β-lactamases but is meant to tell a story about where these enzymes came from, how they have driven antibiotic discovery programs, and what challenges they pose for today.

β-Lactamases are versatile enzymes with a limited range of molecular structures found in a diversity of bacterial sources. Their commonality is the ability to hydrolyze chemical compounds containing a β-lactam ring. As shown in Fig. 1, they have been classified biochemically into two broad divisions according to the mechanism by which they perform hydrolysis, either through the formation of an acyl enzyme with an active-site serine (Fig. 2) (12) or via a hydrolytic reaction facilitated by one or two essential zinc ions in the active sites of metallo-β-lactamases (MBLs) (13). After sequence analyses became available for key β-lactamases, four molecular classes, A, B, C, and D, were eventually assigned based on molecular size and homology between active-site amino acid motifs (14–16). However, biochemical differences between penicillinases and cephalosporinases had been recognized well before sequences were available (17), and known β-lactamases were distinguished based on functional capabilities related to substrate and inhibitor profiles (Table 2). Early classification schemes relied upon relative hydrolysis rates of penicillins and early cephalosporins, together with the enzymatic response to protein-modifying agents (18, 19). As additional substrates and inhibitors were introduced into clinical practice and gene sequencing became inexpensive and routine, both molecular and functional characteristics were combined into a more comprehensive classification scheme (20, 21). Today, at least 17 functional groups associated with the four molecular classes have been distinguished (20), with the major groups shown in Fig. 1. Enzymes are still differentiated with respect to the relative hydrolysis of the β-lactam substrates, penicillins, cephalosporins, carbapenems, and monobactams. Further differentiation is possible based on reactions with the class A β-lactamase inhibitor clavulanic acid (22), the broad-spectrum serine β-lactamase inhibitor avibactam (23), and the metal ion chelator EDTA to identify MBLs (24).

The antibacterial effect of a β-lactam results in inhibition of the growth of replicating bacteria by acylating an active-site serine in essential penicillin-binding proteins (PBPs) (25, 26). During the terminal stages of cell wall biosynthesis, these enzymes are thus prevented from effecting the cross-linking of peptide chains to form peptidoglycan, resulting in cell death. Tipper and Strominger proposed similarities between the structures of penicillin and d-Ala-d-Ala, the terminal amino acids of the nascent acetylmuramyl-pentapeptide fragment, to explain the antimicrobial activity of penicillin (27). Because the majority of β-lactamases contain an active-site serine that also can be acylated by β-lactam molecules, it is not unexpected to find mechanistic and structural similarities between the two sets of enzymes (28, 29). Fisher and Mobashery have cogently outlined a history of studies demonstrating the biochemical commonality between the two acyl enzymes (30). Molecular modeling of various serine β-lactamases and PBP structures has demonstrated three-dimensional similarities with conserved folding patterns and preservation of topology at the active site, in spite of low amino acid identities (28, 31).

The general reaction mechanism for both enzymes is shown in Fig. 2. It is commonly assumed that PBPs were the precursors of the β-lactamases, with the k₃ rate for deacylation increased dramatically for β-lactamases, compared to PBPs that exhibit a fast acylation step (k₂) compared to a slow deacylation step. Formation of the PBP-acyl enzyme complex is considered by many to be an irreversible reaction compared to the lifetime of a bacterial cell, with half-lives of acyl enzymes ranging from a relatively short 9 to 15 min for Pseudomonas aeruginosa PBP2 and PBP3 with amdinocillin or aztreonam (32, 33) to a half-life of more than 24 h for nitrocefin with PBP2a from Staphylococcus aureus (34). Some PBPs may have evolved to function as weak β-lactamases with a slow turnover of the β-lactam substrate (28); for example, cefotaxime deacylation rates are 70- to 80-fold higher for PBP2x variants than that for the wild-type enzyme in Streptococcus pneumoniae (35).

β-Lactamases are ancient enzymes that existed even in the absence of the pressure of therapeutic antibiotics. In 1979, Hamilton-Miller claimed that “penicillinase was born on December 28, 1940” (36), the date Abraham and Chain reported that a bacterial enzyme was capable of destroying penicillin (1). Little did he know that modern phylogenetic analyses have estimated the age of serine β-lactamases to be more than 2 billion years old (37), with plasmid-encoded β-lactamases appearing millions of years ago (38). Recent explorations of remote environmental niches have identified functional β-lactamase-like activity in locations, such as a region of the Lechuguilla Cave in New Mexico (USA) estimated to have been isolated for over 4 million years (39). Sampling of the cave resulted in the growth of 93 bacterial strains, with as many as 62% capable of hydrolyzing β-lactams. However, no sequences of known β-lactamase sequences were identified. The same group also examined metagenomic DNA from 30,000-year-old permafrost sediments east of Dawson City, Yukon (Canada) (40). In these samples, deduced amino acid sequences with 53% to 84% identities to the ubiquitous TEM β-lactamases clustered into the following two enzyme groups: (i) β-lactamases from various streptomycetes, and (ii) uncharacterized β-lactamase-like hydrolytic proteins. In a slightly younger sample of cold-seep sediments of Edison seamount (Papua, New Guinea) estimated to be 10,000 years old, 30 clones produced genes encoding β-lactamases primarily belonging to enzymes in the TEM family (41). Additional metagenomic analyses of ancient and naive environmental samples identified a novel metallo-β-lactamase (MBL), isolated from a 14th century bone sample, with an unusual substrate profile unlike that of other known MBLs; its sensitivity to clavulanic acid and to carbapenems suggested that it may have emerged to provide resistance to currently unknown β-lactam-containing molecules (42). In addition to these β-lactamase-type enzymes in historical samples, β-lactamase production has been identified in multiple remote pristine Antarctic soil samples (43), in ice cores in glaciers outside Antarctica (44), and in remote South American populations that have rarely, or never, been treated with commercial β-lactam antibiotics (45, 46). Curiously, an IMP MBL, a somewhat infrequent contributor to β-lactam resistance in contemporary clinical isolates (47), was found in approximately half of the glacial samples but not from Greenland or Antarctica (44). β-Lactamases have also been identified in fungi, where they are suspected of serving to inactivate plant or microbial naturally occurring β-lactam compounds (48). However, there are still organisms with no confirmed β-lactamases, notably S. pneumoniae, Streptococcus pyogenes, Helicobacter pylori, Mycoplasma spp. and the Chlamydiae, possibly due to the production of homologous enzymes with the ability to inactivate β-lactams (49).

The history of β-lactamases can hardly be written without referring to the history of β-lactam-containing antibacterial agents, including the penicillins, cephalosporins, carbapenems, and monobactams, all of which contain a 4-membered 2-azetidinone ring (50). Although most of the β-lactam-containing agents in use today are synthetic or semisynthetic molecules, β-lactams as therapeutic agents originate from naturally occurring sources, identified initially because of their antibacterial activities in agar-based assays. Penicillin (penicillin G or benzylpenicillin) was first recognized because it prevented the growth of staphylococci on a petri dish contaminated with the common mold Penicillium notatum (51). Cephalosporin C, the starting point for the cephalosporins, was isolated from sewage sludge in Sardinia (52). Carbapenems were found in various β-lactam-producing actinomycetes from soil samples (53), and monobactams were isolated from bacteria in New Jersey soil and water samples (54). All these iterations of unmodified β-lactams existed in natural environments and placed selective pressure on the neighboring bacteria to evade a milieu of deleterious agents. When β-lactam-producing organisms compete with nonproducing bacteria in the same environmental niche, survival strategies are quickly developed. Because bacteria, possibly the first living organisms, have existed for 2.5 to 4 billion years (55, 56), there has been a long incubation period for naturally occurring β-lactam biosynthesis to be perfected and for neighboring bacteria to learn how to survive in the presence of these molecules. Interestingly, mammalian enzymes, including serum proteins (57) and renal dehydropeptidase (dipeptidase) (58), are also capable of hydrolyzing β-lactams.

Resistance to β-lactams can be mediated by multiple mechanisms. Because it is in the best interest of the bacteria to maintain active PBPs, the first line of defense would appear to be for PBPs to alter their affinities for β-lactams while maintaining their physiological function (28). This approach has been utilized most effectively by Gram-positive bacteria, where a primary resistance mechanism has been acquisition of PBPs with decreased affinity for common β-lactams (59). However, it took over 20 years of penicillin usage before clinical isolates of penicillin-resistant S. pneumoniae (60, 61) and penicillin/methicillin-resistant S. aureus (MRSA) were reported as a result of production of low-affinity PBP variants (62–64). Instead, penicillinases, or β-lactamases that hydrolyze penicillin, emerged as the initial resistance mechanism in S. aureus to challenge the therapeutic use of this agent (65). The drug was used more extensively and promiscuously to treat nosocomial streptococcal infections previously associated with high mortality (66), but, as collateral damage, staphylococci initially fully susceptible to penicillin rapidly developed resistance as a result of penicillinase production. In a single English hospital, the percentage of penicillinase-producing staphylococci increased from 14% in 1946 to 59% in 1948 (67) and up to 80% in 1953 (66). More recently, a 2013 study of U.S. clinical isolates showed that 86.5% of all S. aureus isolates (75.6% of methicillin-susceptible isolates) contained the penicillinase gene blaZ (68), indicating that a subpopulation of methicillin-susceptible staphylococci has survived in the absence of either a low-affinity PBP or the production of penicillinase. A transferable penicillinase was also reported in a few clinical isolates of Enterococcus faecalis in the 1980s (69), with identical or closely homologous amino acid sequences as staphylococcal penicillinases (70). A genomic analysis in 2005 demonstrated at least two divergent origins (71). However, penicillinase-producing enterococci have not been observed for the past decade and appear to have disappeared from these pathogens (72). Another analysis of recent genomic data noted similarities among a set of 88 class A β-lactamases from Gram-positive bacteria and two enzymes from Gram-negative bacteria, the ROB-1 penicillinase from Haemophilus influenzae, and the ACl-1 enzyme from Acidaminococcus fermentans (73), possibly due to “trans-Gram transfer” of β-lactamase genes between Gram-positive and Gram-negative bacteria.

In addition to the observed penicillinase production in Gram-positive clinical isolates, enzymes with β-lactam-hydrolyzing properties from numerous environmental bacilli were used by theoretical enzymologists as prototypical biochemical examples of well-behaved enzymes (74, 75). Early studies with β-lactamases from Bacillus cereus (76), Bacillus anthracis (77), and Bacillus licheniformis (78) provided tools on which future biochemical studies of β-lactamases were based. Assay development, enzyme purification, induction properties, and identification of a metal-dependent β-lactamase from these bacilli were activities critical to the establishment of β-lactamases as important enzymological entities (75, 79, 80). It is notable that the classical structural classification of β-lactamases initially proposed by Ambler in 1980 was based on five known amino acid sequences, four of which were from enzymes originating in Gram-positive bacteria. These included three class A penicillinases, PC1 from S. aureus, penicillinase from B. licheniformis 749C, and the B. cereus 569/H β-lactamase I. The only class B metallo-β-lactamases (MBL) that had been sequenced and studied in any detail at the time was the zinc-containing B. cereus 569/H β-lactamase II (14). The first report of β-lactamases other than class A and B enzymes in Gram-positive bacilli described a set of oxacillin-hydrolyzing (class D) penicillinases from a set of environmental strains in 2016 (81).

With the variety of unique β-lactamases that have been identified in natural isolates now exceeding 2,770 (Table 3 and Fig. 3), it is important to have reliable and easily understandable nomenclature to refer to these enzymes. Beginning with Sawai and colleagues, who classified β-lactamases as penicillinases or cephalosporinases according to substrate profiles (18), other classification schemes arose (Table 2) based traditionally on the functional characteristics of these enzymes (19). As nucleotide and amino acid sequences became available, molecular relatedness was added as a defining characteristic (14–16). One of the most cited classification schemes using functional group designations was proposed by Bush, initially in 1988 (187), and then in collaboration with Jacoby and Medeiros (21), as discussed above. Their updated scheme described in 2010 (20) has been further expanded, as shown in Fig. 1, with the addition of avibactam as a differentiating inhibitor to separate serine carbapenemases from MBLs, potentially a useful diagnostic characteristic in whole-cell phenotypic assays (188). Fortuitously, assignments to functional groups generally aligned with molecular classes, although exceptions were noted, particularly with the intrinsically diverse OXA enzymes (189).

Numbering of variants within β-lactamase families has been a challenge, beginning with the emergence of numerous ESBLs resulting from point mutations in both nucleotide and amino acid sequences (5). As a result, in 1996, Jacoby offered to serve as the curator for assignments of new allele numbers for naturally occurring plasmid-encoded ESBLs, with the establishment of the website at http://www.lahey.org/Studies/. The site was later expanded to include assignments for natural alleles in all major β-lactamase families containing more than three known variants. Functional assignments were made for the TEM, SHV, and OXA enzymes if substrate and inhibitor profiles were available. In 2015, the task of assigning new numbers for novel β-lactamases was transferred to NCBI (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA313047), with a request for phenotypic information and genotype data. Tracking of the LEN, OXY, and OKP enzyme families in Klebsiella spp. is being curated by investigators at the Institut Pasteur (http://bigsdb.web.pasteur.fr/klebsiella/klebsiella.html). Bonomo and colleagues have established numbering systems for the Acinetobacter-derived cephalosporinases (ADCs) (190), followed by similar numbering of Pseudomonas-derived cephalosporinases (PDCs) (136, 191), families whose numbers have increased greatly in the past 5 years as a result of whole-genome sequencing projects (185) (Table 3 and Fig. 3).

The agreement within the β-lactamase community to provide a unique designation for each unique (natural) allele has met with criticism from others in the antibiotic resistance field who advocate that there should be at least a 2% change in nucleotide or predicted amino acid sequence before a new enzyme or gene name is assigned (192). However, a single amino acid substitution in a TEM or SHV β-lactamase can alter its biochemical properties, such that the substrate spectrum is dramatically altered to provide for an ESBL phenotype (5), or resistance to β-lactamase inhibitors can occur (118). These enzymes clearly have different functional characteristics and as such warrant a unique designation (193). A recent attempt to lump ESBLs and carbapenemases into a single ESBL category (194) was challenged by many in the β-lactamase community (195). Although the original intention was to simplify β-lactamase nomenclature for the practicing physician, many thought that this would be confusing, especially in situations in which carbapenems, which are routinely used to treat traditional ESBL-producing pathogens (196), would not be effective against infections caused by carbapenemase-producing organisms if carbapenemases were also named ESBLs.

More recent classification schemes are based on the association between 3-dimensional structures and functional information, particularly for the class A/group 2 β-lactamases (197–200). As crystallographic analyses become available for additional β-lactamases, determination of structure-function relationships will become more common. However, for this approach to be successful, high-quality phenotypic data must be provided for novel β-lactamases. The numbers of β-lactamases continue to increase almost exponentially (see Fig. 3 and reference 201), partly because of easy access to inexpensive and rapid gene sequencing. Laboratories now have the ability to identify dozens of “new” β-lactamase sequences (185), with no functional information provided. Unfortunately, whole-genome sequencing provides more data than we perhaps need to know. Some of the genes that are identified as β-lactamase-encoding genes are incomplete, some are misannotated, and some are not expressed, so that assignments for new alleles are not aligned with functionality (202). In addition, “β-lactamase” may be included in the annotation for a new protein sequence that has no β-lactamase hydrolytic capability. An example is the NCBI annotation for sequences in the metallo-β-lactamase superfamily, a family of more than 6,000 enzymes that includes oxidoreductases, as well as enzymes which hydrolyze thiol-ester, phosphodiester, and sulfuric ester bonds (203) but <300 verified MBLs (Fig. 3). Naive investigators may incorrectly assume that this commonly identified set of sequences signals the presence of an MBL, potentially triggering aggressive medical treatment in the hospital laboratory. Even for legitimate enumerations of new β-lactamases based only on sequence data, a lack of functional data minimizes the usefulness of the information. Attempts by structural biologists to align structures with function will suffer due to the lack of phenotypic information to correlate with novel sequences. Although minimally, it is recommended that any novel β-lactamase gene be transferred to a non-β-lactamase-producing strain and its resistance phenotype confirmed, it is preferable that a novel enzyme be purified and its biochemical properties determined and then correlated with microbiological properties in the producing organism (20, 193). Ideally, both genomic and biochemical information should be available before classifying enzymes into families (20).

Selective pressure from both naturally occurring β-lactams and clinically overused β-lactam-containing drugs has created an environment in which new β-lactamases readily emerge, together with maintenance of some of the older, more fit, enzymes. Recent laboratory studies have demonstrated that conjugal plasmids with antibiotic resistance genes are stable in multispecies communities over long periods of time, even in the absence of antibiotic pressure (204). This partially explains the continued presence of antibiotic resistance genes in apparently naive environments (45, 46) and the perpetuation of β-lactamase genes encoding common TEM, SHV, and OXA enzymes. It is notable that TEM-1 and, to a lesser extent, OXA-1, two of the first known plasmid-encoded β-lactamases, have not disappeared and are frequently identified in contemporary, multidrug-resistant, clinical, and environmental isolates (132, 205, 206), perhaps because of their exquisite ability to hydrolyze inexpensive penicillins that are still used therapeutically in community settings. However, there is great concern about environmental contamination in public water supplies by β-lactamase-producing antibiotic-resistant pathogens. A recent study in Hyderabad, India, in both urban and rural areas, found that 100% of all water samples collected from bulk drug manufacturing facilities and multiple water sources, including those contaminated by sewage treatment plants, were positive for ESBL genes, and 95% of them contained carbapenemase genes, often with multiple genes per sample (e.g., bla_OXA-48, bla_KPC, and bla_NDM) (178). Several studies have identified environmental sources of carbapenem-resistant organisms carrying primarily bla_KPC and bla_VIM genes in hospital drains and sinks, thereby providing sources for these pathogens to enter water supplies outside health care centers (153, 207).

One of the most worrisome threats described by the Centers for Disease Control and Prevention (CDC) is carbapenemase production in recent Gram-negative clinical isolates, especially those enzymes with genes carried on mobile genetic elements (208). Fortunately, serine carbapenemases are readily inhibited by new β-lactamase inhibitors, including the diazabicyclooctanone (DBO) inhibitors, such as avibactam (209), and the boronic acid derivative vaborbactam (210), both of which have been recently approved for marketing in combination with ceftazidime or meropenem, respectively. These inhibitor combinations allow relatively effective clinical treatment of infections caused by organisms producing serine carbapenemases (211), thus lowering the risk of clinical failures. A combination of avibactam with the MBL-stable monobactam aztreonam is currently in phase 3 clinical trials to treat Gram-negative bacterial infections, including those with MBL-producing pathogens (https://clinicaltrials.gov/ct2/show/NCT03329092?term=avibactam&draw=2&rank=19). Although resistance has been reported in patients treated with avibactam-ceftazidime, the numbers are currently small (212). As can be expected, however, resistance will increase as the new agents are used. At this time, a β-lactam-containing agent inhibiting a broad spectrum of pathogens, including nonfermentative and anaerobic pathogens, is still to be identified. This is due in part to the increase in multidrug-resistant organisms producing MBLs and/or OXA-48 carbapenemases in combination with β-lactamases of different classes and the lack of an effective inhibitor of all the enzymes (213). Thus, it is possible that MBLs rather than serine β-lactamases may become the predominant carbapenemases in multidrug-resistant bacteria. It is also possible that β-lactamase-stable β-lactams, such as cefiderocol (S-649266), or non-β-lactam-containing agents, such as plazomicin, omadacycline, or eravacycline, will serve as reliable and safe therapeutic alternatives against the most resistant carbapenem-resistant pathogens (214).

In conclusion, β-lactamases, some of our oldest enzymes, have emerged as perhaps the most studied, and most troublesome, of the antibiotic resistance determinants. The fact that many β-lactamase-encoding genes travel together on mobile elements with transmissible resistance factors for other antibiotic classes means that resistance can arise due to a multiplicity of selection pressures. One can anticipate that additional unique β-lactamases with unusual properties will be identified due to the widespread existence of these genes in both environmental and clinical sources and to continued pressure from the use of β-lactam antibiotics. It will be up to the scientific community to monitor their presence with detailed structural and functional studies to provide a scientific basis for the design of new antimicrobial agents that may evade these emerging enzymes, at least for the immediate future.

This article is an updated version of the 2014 ICAAC Lecture “β-Lactamases: Ubiquitous and Formidable.” It is dedicated to the many investigators who set the stage for those of us who have continued to study β-lactam resistance, especially as it drives the discovery and development of new antibacterial agents.

I particularly thank my many collaborators and β-lactamase friends with whom I have worked in this area, including those named below (in alphabetical order): Robert Bonomo, Patricia Bradford, Jed Fisher, Jean-Marie Frère, Janet Hindler, Romney Humphries, George Jacoby, Steve Jenkins, David Livermore, Antone Medeiros, Shahriar Mobashery, Malcolm Page, Timothy Palzkill, Anne Marie Queenan, the late John Quinn, Gian Maria Rossolini, Christine Sanders, Richard Sykes, the late Frank Tally, Fred Tenover, Carl Urban, and Youjun Yang.

There was no financial support associated with the writing of this work.