Elucidation of critical chemical moieties of metallo-β-lactamase inhibitors and prioritisation of target metallo-β-lactamases

Abstract

The urgent demand for effective countermeasures against metallo-β-lactamases (MBLs) necessitates development of novel metallo-β-lactamase inhibitors (MBLIs). This study is dedicated to identifying critical chemical moieties within previously developed MBLIs, and critical MBLs should serve as the target in MBLI evaluations. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), a systematic literature analysis was conducted, and the NCBI RefSeq genome database was exploited to access the abundance profile and taxonomic distribution of MBLs and their variant types. Through the implementation of two distinct systematic approaches, we elucidated critical chemical moieties of MBLIs, providing pivotal information for rational drug design. We also prioritised MBLs and their variant types, highlighting the imperative need for comprehensive testing to ensure the potency and efficacy of the newly developed MBLIs. This approach contributes valuable information to advance the field of antimicrobial drug discovery.

Graphical Abstract

Keywords:

• Metallo-β-lactamase inhibitor
• chemical moiety
• metallo-β-lactamase variant type

Introduction

Antibiotic resistance has been recognised as a pressing health concern, posing substantial threats characterised by diminishing efficacy and potency of antibiotics against bacterial infectious diseases¹. The challenge to antibiotic efficacy has persisted since their first discovery, exacerbated by the continuous emergence and rapid dissemination of antibiotic resistance genes (ARGs) across One Health sectors². Among β-lactam antibiotics, including penicillins, cephalosporins, monobactam, and carbapenems, the latter is crucially employed as one of the last-resort antibiotics against infectious diseases caused by Gram-negative bacteria³. Carbapenem resistance is primarily caused by the presence of carbapenemases, responsible for enzymatic inactivation of the carbapenems⁴. Carbapenemases consist of some of serine β-lactamases (SBLs) of Ambler classes A and D, and all of class B β-lactamases categorised as Zn (II)-dependent metalloenzymes⁵. Class B β-lactamases represent a class of zinc-binding proteins where the hydroxide ion, attacking nucleophile, is activated and positioned by Zn (II) cations in the active site. This mechanism is clearly distinguished from that of SBLs⁴. Thus, all metallo-β-lactamases (MBLs) belong to the class B β-lactamases. Zinc ion(s) coordination in the active site is crucial for their activity, though zinc coordination and residues involved in the active sites vary among MBL subclasses (B1, B2, and B3)⁵. Among these subclasses, the rapid dissemination of subclass B1 MBLs, including New Delhi MBL (NDM), Verona integron-encoded MBL (VIM), and imipenemase (IMP), coupled with the continuous emergence of novel variant types of these MBLs, has intensified the menace posed by MBLs⁶. Moreover, the absence of clinically approved inhibitors for MBLs is a notable challenge. In contrast, while some serine β-lactamase inhibitors (SBLIs) are used in clinical settings, their unsuitability as MBL inhibitors (MBLIs) is due to distinct mechanisms and active sites between SBLs and MBLs^6,7.

In the past few decades, substantial efforts have been dedicated to developing effective MBLIs as a countermeasure for carbapenemase-producing pathogens^4,7. Notably, three candidates including taniborbactam (VNRX-5133), are currently undergoing clinical trials^7–10. Recently comprehensive review articles have excellently summarised the characteristics of MBLs, challenges in MBLI development, breakthroughs in drug development, and the classification of MBLIs along with their inhibition spectra^4,7,11,12. However, there is a knowledge gap concerning MBLIs at the level of chemical moieties, indicating urgent need for comprehension of the correlation between chemical moieties and their corresponding inhibited MBLs. This insight is crucial for discerning critical chemical moieties, facilitating rational drug design. Furthermore, understandings on target MBLs are essential to mitigate the risk of encountering unexpected variants that may not be effectively inhibited¹³.

This study aims to provide essential information for the ongoing development and evaluation of MBLIs. To achieve this objective, we employed two distinct methodologies: a systematic literature analysis using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and a genome database analysis using the NCBI RefSeq database. The systematic literature analysis facilitated a comprehensive understanding of critical chemical moieties of developed MBLIs and potent candidates. This analysis also delineated the inhibition spectra of these MBLIs against various MBL families and their variant types. Concurrently, the genome analysis provided abundance profiles and taxonomic distribution of MBLs at the variant type level. The approaches revealed pivotal information guiding the selection of critical chemical moieties in drug development and provided a list of MBLs which should be prioritised for testing in the evaluation of MBLIs.

Materials and methods

Systematic literature analysis

The systematic literature analysis of MBLIs adhered to the PRISMA guidelines¹⁴. Comprehensive searches were conducted across scientific literature databases, including PubMed, MEDLINE, and Embase, using the keywords “metallo-β-lactamase(s)” and “inhibitor(s)”. The searches were confined to titles and abstracts of literatures published since 2010. Initially, a total of 1,059 non-redundant literatures were identified. Subsequently, 613 records including review articles, articles without full-text, articles without β-lactamase inhibitor data and those deficient of useful information, were excluded during the screening procedures. After the screening, 446 articles remained, of which 220 articles that did not report antibacterial therapy results and focused on the application of previously developed SBLIs to MBLs or test of MBLIs against SBLs were eliminated. This resulted in a dataset of 226 scientific literatures for the systematic analysis (Figure 1).

Figure 1. PRISMA (Preferred Reporting Items for Systematic Review and Meta-Analysis) flow diagram illustrating the study selection process for MBLs and MBLIs.

Previous studies have undertaken the classification of MBLIs based on critical chemotypes for the essential moieties involved in MBL binding and inhibition, such as carboxylate, nitrogen-containing cyclic, sulphur-containing, nitrogen/sulphur-containing cyclic, phosphonate, and boronate moieties^4,11,12,15. Referring to these classifications, our investigation focused on identifying essential chemical moieties based on the most potent and effective MBLIs in the selected 226 literatures (Table S1). Other chemical structures conferring distinct inhibition mechanisms, such as chelating, metal replacement, covalent modification, and allosteric inhibition, were also surveyed in the dataset (Table S1). Furthermore, MBLIs were categorised into five types based on the mechanism of inhibition^4,16. MBLIs of 138 studies out of 226 studies were categorised based on the type classification according to the previously report (Table S1)⁷. MBLs that were tested for evaluation of MBLIs in these studies, were surveyed and among those, only MBLs inhibited by MBLIs were regarded as target MBLs (Table S1). The analyses postulated critical chemical moieties in MBLIs, inhibition types of MBLIs, and inhibited target MBLs. The complete sets of chemical moieties of MBLI and MBL pairs, encompassing all instances of MBL inhibition in the literatures, were counted and used for comparison.

Genome database analysis

Protein sequences of MBLs were extracted from the Comprehensive Antibiotic Resistance Database (CARD, https://card.mcmaster.ca/)¹⁷. The information included MBL subclass, family, and variant type number. All these sequences were used to construct the Maximum-Likelihood phylogenetic tree of MBLs, employing JTT substitution model and bootstrapping, implemented in MEGA software¹⁸. The clustered group of protein sequences in the tree was assigned to a specific MBL clade, with the majority aligning with MBL family classification. To search for MBLs in the NCBI RefSeq genome database (release 221, November 2023), a representative protein sequence for each phylogenetic clade (or family), demonstrating the highest average similarity to all other sequences within the clade (or family), was selected as a query protein sequence for Tblastn (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=tblastn&PAGE_TYPE=BlastSearch&BLAST_SPEC=&LINK_LOC=blasttab&LAST_PAGE=tblastn) search against the database. The sequence identity criterion for MBL family was determined by establishing a cut-off identity which was the lowest identity value in comparison of all of variant types within family. A sequence coverage criterion of 90% was fixed to exclude short fragment MBLs. Signal sequences of NDM, VIM and IMP, were also excluded. Extracted MBLs from the database were identified using variant type sequence information in the CARD. Only identical sequence was assigned to a specific MBL variant type.

Information for genus-level taxonomic origin of assigned MBLs was obtained using the NCBI RefSeq genome database annotation. Variant types were found in the genome database within each family were used to construct a Maximum-Likelihood phylogenetic tree¹⁸. To calculate the frequency of the variation in each amino acid residue of MBL, the Shannon index for each residue was calculated and displayed¹⁹. Interactive Tree of Life (iTOL) tool was used to visualise multiple panel figures, including MBL phylogenetic tree, taxonomic clustering based on the NCBI taxonomy, taxonomic distribution heatmap (presence or absence), and abundance bar chart²⁰.

Results

Systematic literature analysis

MBLI types and profiles of their chemical moieties

A total of 226 scientific literatures were selected and employed for a comprehensive exploration of the chemical moieties and inhibition mechanisms of MBLIs and their target MBLs. These records encompass the identification of novel MBLI candidates and the validation of the candidates. Various approaches, such as virtual screening^21–25, high-throughput screening of chemical libraries^26–28, fragment-based drug discovery^29–32, natural compound screening^33,34, structure-activity relationship analysis^35,36, and repurposing approved drugs^37–39, have been employed to elucidate novel MBLI scaffolds and their improved derivatives. Microbiological assessment, enzyme kinetic analyses, molecular docking studies, and investigations using animal models were conducted to access the potency of MBLIs. Recent efforts have endeavoured to categorise MBLIs based on their essential chemical structures^4,11,12. In accordance with the categorizations, the chemical moieties in the selected scientific literatures were delineated as follows: those generally implicated in competitive inhibition, such as carboxylate, nitrogen-containing cyclic, sulphur-containing, nitrogen/sulphur-containing cyclic, boronate, and phosphonate moieties, were surveyed. The chemical structures of chemical moieties incorporated in MBLIs are displayed in Figure 2. Additionally, other chemical structures conferring distinct inhibition mechanisms, including chelating, metal replacement, covalent modification, and allosteric inhibition, were explored.

Figure 2. The chemical structures of the chemical moieties incorporated in MBLIs.

Moreover, a recent publication classified MBLIs into five types based on their binding modes in the active site and zinc coordination⁷. Type I MBLIs involve chemicals in coordination and/or bridging between zinc ion(s) as zinc-binding pharmacophores. Type II, III, and IV MBLIs mimic the binding modes between MBLs and β-lactams as the initial binding, tetrahedral intermediate complex, and product binding in MBL-hydrolysis reactions, respectively. Type V MBLIs stand out, as they degrade or modify the metal ion binding active sites, incorporating features such as zinc chelation, covalent modification of metal-binding site residues, and zinc replacement⁷.

In our investigation, we delved into the classification of MBLI types within the literatures (Figure 3(A)). Notably, type V MBLIs emerged as the most frequently reported, attributed to frequent chelator studies, as they displayed broad-spectrum inhibition activity across various MBLs⁴. A fungal metabolite aspergillomarasmine A (AMA), Ca-EDTA, 1,4,7-triazacyclononane-N,N′,N″-triacetic acid (NOTA), and dipicolinic acid have been developed and validated as chelator MBLIs^40–52. Following type V, type I MBLIs was identified as the most prevalent. Within type I MBLIs, characterised by zinc-binding pharmacophores, a diverse array of chemical moieties was encompassed within this category. Type III MBLIs, which include bicyclic boronate inhibitors such as taniborbactam (VNRX-5133), xeruborbactam (QPX7728), and QPX7831, have received frequent attention, because these represent the only three MBLIs currently undergoing clinical trials^{8–10,53–60}. Type II and IV have received comparatively less investigation in the selected literatures^{16,22,36, 61–64} (Figure 3(A)).

Figure 3. Chemical moieties of MBLIs and their target MBLs in the systematic literature analysis. (A) Composition of MBLIs categorised by the classification of inhibitor types and chemical moieties, and composition of MBLs according to subclasses (B1, B2, and B3) and families such as NDM; (B) Pairs of chemical moieties and inhibited MBL variant types in the literatures were counted as cases and their counted amounts were displayed as a heatmap and bar charts, based on MBL variant types used for targets of efficacious MBLIs and chemical moieties of effective MBLIs.

We scrutinised chemical characteristics of competitive MBLIs in the literatures at the chemical moiety-level. Nitrogen-containing cyclic, sulphur-containing, and carboxylate moieties emerged as the most prevalent (Figure 3(A)). Among nitrogen-containing cyclic moieties, triazole, pyridine, indole, and pyrrolidine were the most frequently found in MBLIs. Triazole was predominantly found in type I MBLIs such as 1,2,4-triazole-3-thione and azolylthioacetamide compounds^65–69, while it was also identified in type II inhibitors such as NH-1,2,3-triazole⁶⁰. Pyridine moiety was identified in type I, III, and V MBLIs^70–78. Indole moiety serves as an essential component in indole 2-carboxylate (type II)¹⁶. Pyrrolidine moiety was commonly found in type I MBLIs together with thiol and carboxylate moieties^79–84. Carboxylate moiety was identified across all types, with predominant presence in type I as zinc coordination moiety and in type III as part of bicyclic boronate drugs (taniborbactam and xeruborbactam). Dicarboxylates were detected in maleic acid derivative (ME1071) and phthalic acid derivatives^85–89. Concerning sulphur-containing moieties, the majorities were found in type I MBLIs, suggesting their general involvement as zinc-binding pharmacophores. Thiol emerged the most frequently utilised sulphur-containing moiety for MBLIs exemplified by captopril, thienolate, and mercaptoacetate^81,82,84,90. Similar to thiol moiety, thioether were found in type I MBLIs and some of them were also present in type II and III MBLIs^62,66,91. Sulphonamide was frequently found as a constituent of type IV inhibitors, exemplified by ANT431 and ANT2681 (Figure 3(A))^{34,35, 92,93}.

Moieties containing both nitrogen and sulphur within cyclic structure are less commonly found in MBLIs. However, thiazole and thiazolidine were identified in certain MBLIs, such as thiazole in ANT431 (type IV) and thiazolidine in bisthiazolidine and 2-mercaptomethyl-thiazolidines^{34,35,92,94–96}. Boronate and phosphonate moieties play a role in the binding mode mimicking the tetrahedral transition state^4,7. Given the active researches on type III MBLIs, such as taniborbactam and xeruborbactam, bicyclic boronate as their core chemical structure showed a relatively high frequency in the literatures. The presence of the phosphonate moiety was less common and was found in phospholactam and mercaptophosphonate MBLIs^97,98. Other chemical structures conferring distinct inhibition mechanisms, such as chelator, metal replacement, covalent, and allosteric inhibition, generally belong to type V MBLIs, and they exhibit diverse chemical structures rather than the presence of conserved chemical moiety (Figure 3(A))^4,7.

Tested target MBLs for evaluation of MBLIs

Considering the diverse families of MBLs and numerous variant types within each family, an extensive array of MBL variant types exists, each displaying distinct β-lactam resistance spectra and inhibition patterns by MBLIs⁵. This underscores the critical importance of selecting the appropriate MBLs for MBLI evaluation to ensure drug efficacy. In our analyses, we identified MBLs successfully inhibited by MBLIs (Figure 3(A)). Subclass B1 MBLs emerged as the predominant inhibition targets. Within this subclass, NDM, VIM, and IMP collectively accounted for over 93% of subclass B1 MBLs tested (Figure 3(A)). This suggests that these MBLs, known for their horizontal gene transfer⁶, are suitably employed for evaluating MBLIs. In subclass B2, ImiS and CphA MBLs were frequently employed, while L1 MBLs constituted the primary target in subclass B3 (Figure 3(B)). Generally, these B2 and B3 MBLs were used to elucidate inhibition spectra of developed NDM, VIM, and IMP inhibitors for other subclasses. However, it is noteworthy that among the numerous variant types of NDM, VIM, and IMP, only a limited number of variants have been used for MBLI evaluation. Among subclass B1 MBL variant types, only NDM-1, VIM-2, VIM-1, VIM-4, and IMP-1 were frequently tested, while the others showed very low frequency in the literatures (Figure 3(B)).

MBLs and MBLI moieties

Pairs of chemical moieties and inhibited MBL variant types systematically tallied to elucidate the correlation between chemical moieties and MBL variant types. The analyses yielded insights into the involvement of each chemical moiety in inhibition spectra, revealing moieties (notably carboxylate and bicyclic boronate) that were involved in the broad spectra (Figure 3(B)). Most chemical moieties demonstrated efficacy to specific MBL families or types. For instance, thiol and triazole moieties exhibited effective inhibition against VIM variant types but were limitedly effective against NDM-1 among NDM variant types (Figure 3(B)). Regarding MBL variant types, inhibition was observed across various chemical moieties for NDM-1, VIM-2, VIM-1, and IMP-1, while inhibition of other variants was not consistently verified. As an example, NDM-9 variant exhibiting the reduced inhibition by taniborbactam was shown to be inhibited in only three studies, employing an azetidine moiety and two phytochemicals (magnolol and fisetin) (Figure 3(B))^99–101.

Genome database analysis

Abundance profiles and taxonomic distribution of MBLs

Based on the literature analysis, it is evident that a limited number of MBLs have been used for MBLI evaluation. Considering the evolving inhibition spectra resulting from mutations in MBLs⁵, the current approaches, which primarily focused on representative type(s) of each MBL family (typically centred on NDM-1 for NDMs, VIM-2 and VIM-1 for VIMs, and IMP-1 for IMPs), requires reconsideration. While the literature analysis is valuable for understanding the important chemical moieties in MBLIs and their inhibition capacity, it may not provide the actual MBLs that should be urgently studied as targets. To identify critical and urgent MBL targets for MBLI development, elucidation of critical MBL families and their variant types through understanding their abundance profiles and taxonomic distribution in the NCBI RefSeq genome database was conducted.

As of November 2023 (v3.2.8), a total of 440 MBLs were deposited in the CARD, and they were clustered into 91 distinct clades, which predominantly aligned with family classification. A phylogenetic tree exhibited the clear distinctions among subclasses, with the length of collapsed branches indicating the number of types within each family (Figure 4(A)). Initially, we profiled the total abundance of MBL family proteins deposited in the NCBI RefSeq database. NDM emerged as the most abundant, followed by Bla2/BcII clade, IMP, VIM, CphA, BJP, BlaB, GOB, IND/CGB clade, CfiA, CAR, and L1 (each exceeding 100 proteins in the database). With the exception of CphA (B2), BJP (B3), GOB (B3), CAR (B3), and L1 (B3), the majority belonged to subclass B1. Notably, Bla2/BcII, CphA, BJP, BlaB, GOB, CfiA, CAR, and L1, while abundant, exhibited the presence only in a single genus, Bacillus, Vibrio, Aeromonas, Bradyrhizobium, Elizabethkingia, Elizabethkingia, Bacteroides, Pectobacterium, and Stenotrophomonas, respectively, suggesting intrinsic resistance without dissemination (Figure 4(B)). In the case of proteins belonging to IND/CGB clade, they were distributed in two genera (Chryseobacterium and Elizabethkingia), but most of the them were found in Chryseobacterium (98%) (Figure 4(B)).

Figure 4. Abundance profiles of MBL families deposited in the NCBI RefSeq database according to their respective host genera. (A) The taxonomic distribution and abundance of MBLs at the genus and family levels. A total of 81 genera were observed as carriers of MBLs, and their families were displayed above the genus names. The phylogenetic tree of MBLs was constructed based on all MBL protein sequences retrieved from Comprehensive Antibiotic Resistance Database using the Maximum-Likelihood method with 500 bootstrap repetitions. The clustered group of protein sequences in the tree was assigned to a specific MBL clade, with the majority aligning with the classification of MBL families. The horizontal length of an isosceles triangle symbol reflects the proportion of the collapsed variant types’ number. Each subclass of MBLs is differentiated by a specific colour label (red, subclass B1; yellow, subclass B2; blue, subclass B3). Bar charts represent the abundance of each MBL family. Major MBL families (more than 100 proteins) are highlighted with red labels; (B) Taxonomic compositions of the MBL families.

NDM, VIM, and IMP belonging to subclass B1, displayed high abundance and a broad taxonomic distribution across 29, 27, and 24 genera belonging to diverse families, respectively. NDM was dominantly present in Escherichia (35%) and Klebsiella (33%), Enterobacter (12%), and Acinetobacter (9%). In the case of VIM, Pseudomonas occupied 47% of total VIM proteins, and followed by Klebsiella (20%) and Enterobacter (14%). IMP are distributed mainly in Enterobacter (30%), Klebsiella (25%), and Pseudomonas (18%). These distributions indicate critical importance of NDM, VIM, and IPM as target MBLs for MBLI development. These findings align with the literature analyses (Figures 3 and 4).

Apart from these three MBLs, others were found in only 1–3 genera which are taxonomically close, indicating that these MBLs are presumed as not mobile ARGs in a current status. Remarkably, AFM demonstrated a broader distribution across 7 genera, despite having only 41 proteins in the database. Among these proteins, 34 proteins were found in Pseudomonas (family Pseudomonadaceae), while others were found in Commamonas (family Comamonadaceae, 2 protein), Stutzerimonas (family Pseudomonadaceae, 1 protein), Achromobacter (family Alcaligenaceae, 1 protein), Bordetella (family Alcaligenaceae, 1 protein), Aeromonas (family Aeromonadaceae, 1 protein), and Stenotrophomonas (family Xanthomonadaceae, 1 protein), revealing the presence of the ARG across broad taxa. It indicates that AFM should be considered a crucial target MBL for MBLI development, alongside NDM, VIM, and IMP (Figure 4). As of November 2023, only four variant types have been reported for AFMs in BLDB (http://bldb.eu/)¹⁰². Among these variants, only AFM-1 exhibited the distribution across the seven genera whereas the others were exclusively found in Pseudomonas, indicating that AFM-1 represents a disseminated variant type within this family.

Abundance profiles and taxonomic distribution of variant types of NDM, VIM, and IMP

NDM, VIM, and IMP were shown to be the most abundant and prevalent MBLs in the database (Figure 4), thus their variant types were scrutinised to postulate critical variant types of them. In the case NDM, NDM-1 emerged as the most frequently detected variant, and NDM-5 also exhibited high abundance in the database. Mutations in NDM-5 (V88L, M154L, and A233V) compared to NDM-1, were conserved in many other variants, highlighting them as major polymorphic residues (Figure 5(A,B)). NDM-1 and NDM-5 were found in 29 and 12 genera, emphasising their significance as mobile ARGs. The total number of NDM-1 and NDM-5 proteins accounted for more than 93% of detected NDMs in the database. These findings suggest that NDM-5 should be recognised as a critical NDM variant together with NDM-1. Among the less abundant types, NDM-7 (2.1%), NDM-9 (1.6%), NDM-4 (0.9%), and NDM-14 (0.5%) were found frequently, albeit at lower levels compared to two major types. Unique alterations such as D130N, E152K, and D130G were identified in NDM-7, NDM-9, and NDM-14, respectively (Figure 5(A,B)). Notably, NDM-14 was exclusively found in Klebsiella, whereas NDM-7, NDM-9, and NDM-4 were distributed across 5, 4, and 3 genera, respectively. Their proportions in genera were depicted as a composition of taxonomic origin, revealing that NDM-1, NDM-5, NDM-7, NDM-9, and NDM-4 are mobile ARGs of priority (Figure 5(C)). Considering their abundance, unique variations, and taxonomic distribution, these five NDM variants are suggested as critical MBLs for use in MBLI development. Notably, among critical NDM variant types, only NDM-1 and NDM-9 were found in Acinetobacter, indicating that NDM-9 is a critical NDM variant that should be concerned although with relatively low prevalence (Figure 5(C)).

Figure 5. Abundance profiles, phylogeny, polymorphisms, and taxonomical compositions of NDM variant types deposited in the NCBI RefSeq database. (A) Abundance profiles, phylogeny, and polymorphisms of NDM variant types. The binary chart displays distribution of polymorphic residues across all NDM variant types. Heat map exhibits the taxonomic distribution of NDM proteins at the genus levels. Abundance of each variant type was displayed using a bar chart. The phylogenetic tree of NDM proteins was constructed based on all of NDM protein sequences retrieved from Comprehensive Antibiotic Resistance Database using the Maximum-Likelihood method with 500 bootstrap repetitions. NDM-1, the most abundant and most prevalent NDM variant type, was selected as the reference protein for comparison of polymorphic residues. Conserved and mutated polymorphic residues of other NDM proteins except for signal sequence were displayed as empty and filled circle, respectively. The actual amino acid residue of mutated site in each NDM family protein was represented in the circles with white letters. A total of 29 genera were observed to contain NDM variant types, and were utilised for assess their distribution across the genera in heat map. Bar charts represent the abundance of each NDM variant types. Major NDM families (more than 40 proteins) are highlighted with red labels; (B) Shannon diversity indices of each amino acid residue of all NDM variant types deposited in the NCBI RefSeq database. (C) Taxonomic compositions of the major NDM variant types.

The phylogenetic tree of VIM variant types clearly demonstrates two distinct clades, each represented by VIM-1 and VIM-2 (Figure 6(A)). They accounted for more than 78% of total VIMs, and the majority of polymorphisms of VIM was attributed to their sequence difference (Figure 6(A,B)). VIM-1 and VIM-2 were found in 18 and 11 genera across various families. Beside these two major types, VIM-4 and VIM-5 were also frequently detected, distributed across 15 and 6 taxa, respectively. In contrast, the other minor types were found in 1 or 2 genera. VIM-4 and VIM-5 are variants of VIM-1, and VIM-5 exhibited unique sequence variations, such as A128K, E202A, and K257T (Figure 6(A,B)). Analyses of the composition of taxonomic origin revelled that VIM-2, VIM-1, VIM-4, and VIM-5 are mobile ARGs (Figure 6(C)). Regarding their abundance, the presence of unique variations, and broad taxonomic distribution, these four VIM variants are suggested as critical MBLs that should be used as targets in the MBLI development. For instance, VIM-5 was frequently found in Acinetobacter (55%), compared to other critical VIM variant types, revealing the importance of this variant type (Figure 6(C)).

Figure 6. Abundance profiles, phylogeny, polymorphisms, and taxonomical compositions of VIM variant types deposited in the NCBI RefSeq database. (A) Abundance profiles, phylogeny, and polymorphisms of VIM variant types. The binary chart displays distribution of polymorphic residues across all VIM variant types. Heat map exhibits the taxonomic distribution of VIM proteins at the genus levels. Abundance of each variant type was displayed using a bar chart. The phylogenetic tree of VIM proteins was constructed based on all of VIM protein sequences retrieved from Comprehensive Antibiotic Resistance Database using the Maximum-Likelihood method with 500 bootstrap repetitions. VIM-2, the most abundant and most prevalent VIM variant type, was selected as the reference protein for comparison of polymorphic residues. Conserved and mutated polymorphic residues of other VIM proteins except for signal sequence were displayed as empty and filled circle, respectively. The actual amino acid residue of mutated site in each VIM family protein was represented in the circles with white letters. A total of 27 genera were observed to contain VIM variant types, and were utilised for assess their distribution across the genera in heat map. Bar charts represent the abundance of each VIM variant types. Major VIM families (more than 38 proteins) are highlighted with red labels; (B) Shannon diversity indices of each amino acid residue of all VIM variant types deposited in the NCBI RefSeq database. (C) Taxonomic compositions of the major VIM variant types.

Compared to NDM and VIM, IMP displayed less conserveness throughout protein sequences (Figure 7(A)). Numerous polymorphic residues were present, and many of them conferred high values of Shannon index (Figure 7(B)). IMP-4 and IMP-1 were the most abundant types, collectively occupying more than 68% of total IMP proteins, and present in 12 and 18 genera, respectively (Figure 7(A)). In addition to two major types, there were four types (IMP-27, IMP-6, IMP-8, and IMP-26) demonstrating relatively high abundance and a broad taxonomic distribution, each represented by more than 30 proteins across more than 4 genera (Figure 7(C)). Along with IMP-4 and IMP-I, these IMP variant types also should be considered target MBLs in MBLI development.

Figure 7. Abundance profiles, phylogeny, polymorphisms, and taxonomical compositions of IMP variant types deposited in the NCBI RefSeq database. (A) Abundance profiles, phylogeny, and polymorphisms of IMP variant types. The binary chart displays distribution of polymorphic residues across all IMP variant types. Heat map exhibits the taxonomic distribution of IMP proteins at the genus levels. Abundance of each variant type was displayed using a bar chart. The phylogenetic tree of IMP proteins was constructed based on all of IMP protein sequences retrieved from Comprehensive Antibiotic Resistance Database using the Maximum-Likelihood method with 500 bootstrap repetitions. IMP-1, the most abundant and most prevalent IMP variant type, was selected as the reference protein for comparison of polymorphic residues. Conserved and mutated polymorphic residues of other IMP proteins except for signal sequence were displayed as empty and filled circle, respectively. The actual amino acid residue of mutated site in each IMP family protein was represented in the circles with white letters. A total of 24 genera were observed to contain IMP variant types, and were utilised for assess their distribution across the genera in heat map. Bar charts represent the abundance of each IMP variant types. Major IMP families (more than 50 proteins) are highlighted with red labels; (B) Shannon diversity indices of each amino acid residue of all IMP variant types deposited in the NCBI RefSeq database. (C) Taxonomic compositions of five major IMP variant types and IMP-26.

Discussion

In this study, we employed two distinct approaches to decipher critical chemical moieties required for the development of effective MBLIs and to identify MBLs which should be prioritised as target MBLs for the MBLI evaluation. Assessments of the prevalence of chemical moieties in the literatures were based on the previous classification of chemotypes of MBLIs, usually concerned with the essential moieties involved in MBL binding and inhibition^4,11,12,15. The analyses unveiled both narrow and broad-spectrum moieties, providing pivotal information crucial for rational drug design (Figure 3).

In the case of MBLs used in the literature, the survey revealed that only a limited number of MBLs have been used in evaluating MBLIs. Out of tremendous variant types of major subclass B1 MBLs (NDM, VIM, and IMP), only five variant types (NDM-1, VIM-2, VIM-1, VIM-4, and IMP-1) have been mainly used for evaluation of the MBLIs. Although some studies focused on mutations or variant types and their inhibition by MBLIs^{42,53,103,104}, most of studies employed only major variants to evaluate MBLIs. This observation may account for instances where certain MBL variant types, such as NDM-9, were not tested in the development process of taniborbactam. A recent study corroborated this, demonstrating that the pre-existing NDM variant, NDM-9, is not inhibited by taniborbactam due to unique sequence alteration (E152K)¹³. Furthermore, a follow-up study revealed that NDM-9 was successfully inhibited by a different bicyclic boronate MBLI, xeruborbactam¹⁰⁵. These results indicate that critical variant types should be elucidated and applied to the MBLI tests, indicating that proper criteria and approach are demanded for selection. Based on the insights gained from literature analyses, we hypothesise that proposing critical MBLs and their variants for MBLI evaluation could enhance the efficacy of novel MBLIs and ensure their effectiveness. Leveraging the genome database analyses regarding prevalence and taxonomic distribution of MBLs, we have compiled a comprehensive list of critical MBLs and their variant types, including some that were previously underestimated in the literatures (Figures 4–7). These MBL variant types are designated as targets for the extended-spectrum MBLI development.

Considering overall landscape of MBLs and their variant types, variants exhibiting the high prevalence and taxonomic distribution must be classified as potentially significant variants which should be used for MBLI evaluations (Figure 5). This underscores the importance of considering variant types of NDM, VIM and IMP suggested in this study as target for MBLI evaluations.

Furthermore, AFM family was initially reported in Alcaligenes faecalis¹⁰⁶, and now they are distributed across various genera, suggesting that it should be under strict surveillance. AFM-1 was shown to be a major disseminated variant type within this family. However, polymorphic residues in other variant types were located in the signal sequence region¹⁰⁷, indicating that there may not be the distinct resistance spectra against AFM variant types and the different inhibition patterns by MBLIs among the variant types. As a study in the future, it would be better (i) to conduct analyses that explore the relationship between chemical moieties and MBL families (or variant types) through enzyme-ligand docking analyses, (ii) to investigate the relative abundance of the different MBL variant types in the clinical isolates, and (iii) to reveal the impact of the chemical moieties of developed MBLIs when targeting Enterobacterales, or non-fermentors such as Pseudomonas aeruginosa or Acinetobacter baumannii.

Conclusion

Based on two systematic approaches, we propose critical chemical moieties essential for MBLIs, providing pivotal information for further rational drug design. Additionally, we identified critical MBLs and their variant types, which must undergo testing to ensure the efficacy and potent of novel MBLIs. Together with previously examined NDM, VIM, and IMP families, AFM family is also suggested as a critical MBL family to be used in the extended-spectrum MBLI evaluation. In the case of variant types for the evaluation of extended-spectrum MBLIs, NDM-5, NDM-7, NDM-9, NDM-4, VIM-4, VIM-5, IMP-4, IMP-27, IMP-6, IMP-8, IMP-26, and AFM-1 were newly selected as target MBL variant types, along with previously used variants NDM-1, VIM-2, VIM-1, and IMP-1.

Author contributions statement

JHL, S-GK and D-WK contributed to the genome database survey and analysis. JHL, S-GK, K-MJ, KS, HJ, D-WK, BCJ and SHL performed the literature search and article writing (original draft preparation). SHL contributed to the funding acquisition and supervision. All authors have read and agreed to the published version of the manuscript.

Acknowledgements

We thank our colleagues, both past and present, at the National Leading Research Laboratory of Drug Resistance Proteomics, Myongji University, for sharing their insights into the concepts presented here.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

None.

Additional information

Funding

This work was supported by research grants from the Bio & Medical Technology Development Program of the National Research Foundation of Korea (NRF) funded by the Ministry of Science, Information and Communications Technology (MSIT; grant No. NRF-2017M3A9E4078014); and from the NRF funded by the MSIT (grant Nos. NRF-2021R1A2C3004826 and NRF-2022R1F1A1074068).