Molecular mechanisms of vancomycin resistance

Abstract

Vancomycin and related glycopeptides are drugs of last resort for the treatment of severe infections caused by Gram‐positive bacteria such as Enterococcus species, Staphylococcus aureus, and Clostridium difficile. Vancomycin was long considered immune to resistance due to its bactericidal activity based on binding to the bacterial cell envelope rather than to a protein target as is the case for most antibiotics. However, two types of complex resistance mechanisms, each comprised of a multi‐enzyme pathway, emerged and are now widely disseminated in pathogenic species, thus threatening the clinical efficiency of vancomycin. Vancomycin forms an intricate network of hydrogen bonds with the d‐Ala‐d‐Ala region of Lipid II, interfering with the peptidoglycan layer maturation process. Resistance to vancomycin involves degradation of this natural precursor and its replacement with d‐Ala‐d‐lac or d‐Ala‐d‐Ser alternatives to which vancomycin has low affinity. Through extensive research over 30 years after the initial discovery of vancomycin resistance, remarkable progress has been made in molecular understanding of the enzymatic cascades responsible. Progress has been driven by structural studies of the key components of the resistance mechanisms which provided important molecular understanding such as, for example, the ability of this cascade to discriminate between vancomycin sensitive and resistant peptidoglycan precursors. Important structural insights have been also made into the molecular evolution of vancomycin resistance enzymes. Altogether this molecular data can accelerate inhibitor discovery and optimization efforts to reverse vancomycin resistance. Here, we overview our current understanding of this complex resistance mechanism with a focus on the structural and molecular aspects.

1. INTRODUCTION

Drug‐resistant gram‐positive bacteria such as Staphylococcus aureus, Streptococcus pneumoniae, Clostridium difficile, Enterococcus faecium are considered Urgent or Serious Threats by the Centers for Disease Control and the World Health Organization.1, 2 Treatment options for infections by these species include glycopeptide antibiotics. The prototypical glycopeptide vancomycin is a natural product produced by the Actinobacteria Amycolatopsis orientalis discovered in the 1950s by researchers at Eli Lilly (Indianapolis) and has been in clinical use since 1958.3 Another natural product glycopeptide antibiotic, teicoplanin, is produced by Actinoplanes teichomyceticus and was approved for use in 1998. Industrial semi‐synthetic efforts have led to development of next‐generation glycopeptides or their lipidated analogs, lipoglycopeptides, including telavancin, dalbavancin, and oritavancin.4, 5, 6

The mode of action of vancomycin and other glycopeptide antibiotics is in binding to the terminal d‐Ala‐d‐Ala moiety of un‐crosslinked Lipid II (undecaprenyl‐diphospho‐N‐acetylmuramoyl‐[N‐acetylglucosamine]‐l‐alanyl‐γ‐d‐glutamyl‐l‐lysyl‐d‐alanyl‐d‐alanine), an intermediate in the peptidoglycan layer maturation process4, 7 (Figure 1a). This binding obstructs the activity of penicillin‐binding proteins (PBPs) to cross‐link Lipid II into mature peptidoglycan and thus compromises the integrity of the cell envelope, leading to osmotic stress and bursting of the cell (Figure 1b).

Figure 1.

(a) Crystal structure of vancomycin•di‐acetyl‐Lys‐d‐Ala‐d‐Ala (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=1FVM 8). The hydrogen bond lost if d‐Ala‐d‐Ala is modified to d‐Ala‐d‐lac is indicated by a pink star and pink dashes, while the region of steric clash if d‐Ala‐d‐Ala is modified to d‐Ala‐d‐Ser is indicated with a blue star. (b) Representative peptidoglycan synthesis pathway showing reactions catalyzed by housekeeping enzymes labeled by enzyme name in black text, while those catalyzed by vancomycin resistance enzymes labeled by enzyme name in red text. VanA and VanC ligases shown as representative for d‐Ala‐d‐lac and d‐Ala‐d‐Ser ligases, respectively. Modification of Lipid II to terminate in d‐lac or d‐Ser lowers affinity toward vancomycin. Adapted from9, 10

High‐level vancomycin resistance (MIC 64 to 1,000 mg/L) was first discovered in 1986, more than 30 years after the first clinical use of vancomycin, when vancomycin resistant enterococci (VRE) were isolated.11 The resistance determinants in these strains were encoded in a transposon on a plasmid, raising the specter of dissemination of vancomycin resistance among Gram‐positive species via horizontal gene transfer. Indeed, since this initial discovery, acquired vancomycin resistance has been detected in various Enterococci species (E. faecium, E. faecalis, E. gallinarum, plus others), as well as Staphylococcus aureus (VRSA), Clostridium difficile, and Streptococcus bovis; furthermore, intrinsic (chromosome‐encoded) vancomycin resistance has also been discovered in many species.9, 12

The pioneering work of many research groups9, 13 over three decades led to characterization of the genetic and molecular framework of vancomycin resistance. As a result, a unique resistance mechanism was revealed, which involves the concerted action of multiple enzymes altering the basic structure of the cell envelope in order to overcome the action of the bactericidal molecule. Recent advances in structural and molecular characterization of the key vancomycin resistance pathway proteins have opened a new phase in research of this important topic. With much remaining to be uncovered, we review here our recent advances in understanding of the molecular structures and mechanisms of vancomycin resistance cascade proteins.

2. VANCOMYCIN RESISTANCE MECHANISMS: TWO MAIN ROUTES FOR MODIFICATION OF PEPTIDOGLYCAN

In this review, we focus on the glycopeptide resistance mechanisms which rely on the concerted action of multiple enzymes to modify peptidoglycan; these confer either high‐level (i.e., MIC >64 mg/L vancomycin) or low‐level (i.e., MIC 4 to 32 mg/L vancomycin) resistance depending on the type of modification of the terminal d‐amino acid in Lipid II. These types of glycopeptide resistance determinants have been reviewed previously.5, 9, 14, 15, 16 Thus, we only briefly introduce these types and gene cassettes below to serve as a foundation for the further discussion of the structural and molecular details of individual resistance enzymes.

2.1. d‐Ala‐d‐lac based resistance (VanA, VanB, VanD, VanF, and VanM types)

The replacement of d‐Ala‐d‐Ala with d‐Ala‐d‐lac as the terminal amino acids in Lipid II results in a 1,000‐fold decrease in binding constant between vancomycin and peptidoglycan due to the loss of a single hydrogen bond,17 thus conferring high levels of resistance (MIC >64 mg/L) (Figure 1a). The “core” resistance cassette able to support this conversion is composed of five genes.9 The vanHAX cluster encodes for three enzymes involved in peptidoglycan modification and is accompanied upstream by the vanR and vanS genes which comprise a canonical two‐component regulation system.18 This composition of vancomycin resistance genes was called “VanA”‐type resistance while in subsequently identified vancomycin resistance cassettes the genes encoding the VanA homologs were designated as vanB, vanD, vanF, and vanM. Accordingly, these designations were also applied to the whole corresponding resistance module. To simplify the nomenclature of homologues VanA‐type resistance components, in this review we will add a subscript to the protein names referring to the cassette type these proteins belong to, that is, VanH_A will stand for the VanH variant from VanA‐type cassette and so forth (Figure 2).

Figure 2.

Vancomycin resistance operons organized by d‐Ala‐d‐lac and d‐Ala‐d‐Ser mechanisms, and then by the cassette name. Gene sizes not to scale. Key: black outline for gene arrow = insoluble protein in expression attempts by CSGID and no evidence in literature for soluble protein; tan green fill = purified by CSGID and/or in the literature; dark green fill = crystal structure solved (PDB codes indicated under gene arrow)

In addition to the “core” genes, the cassettes supporting the d‐Ala‐d‐lac‐based vancomycin resistance also contain the “accessory” genes vanY and vanZ, which are not strictly required for conferring vancomycin resistance, as will be discussed in respective sections below.19, 20, 21, 22 A d‐Ala‐d‐lac resistance operon providing high level resistance to vancomycin is present in various Actinobacteria, with the Streptomyces coelicolor operon serving as a model system for studies of glycopeptide resistance in this genus.23

The structure and function of the five core proteins encoded by the d‐Ala‐d‐lac based vancomycin resistance cassettes are discussed below.

2.1.1. VanH dehydrogenase

This gene encodes a dehydrogenase (VanH_A) that catalyzes conversion of pyruvate to d‐lactate, providing the substrate for synthesis of the alternative d‐Ala‐d‐lac depsipeptide. VanH_A is specific for production of the d‐isomer of lactate and can utilize NADH or NADPH as the reductant, although the VRE enzyme shows a slight preference for NADPH.17, 24 VanH_A is also capable of reducing other small α‐keto acids however all tested VanH homologues show preferable activity toward pyruvate.17, 24

The molecular mechanism by which this enzyme is capable of discriminating for production of the d‐isomer of lactate, and how it is specific for lactate rather than other α‐keto acids, are not fully understood. No molecular structure of a representative of this enzyme family has been determined. The purification and preliminary crystallization of a VanH fusion with thioredoxin has been communicated but no subsequent crystal structure has been published.25

VanH_A shows moderate sequence similarity (33–37% identity) to d‐lactate dehydrogenase enzymes from the Gram‐positive genuses Lactobacillus and Leuconostoc.26, 27 Accordingly, molecular structure and enzymological analyses of these enzymes could be used to derive a molecular model for VanH_A function. Comparative sequence analysis indicates that the three residues essential for lactate dehydrogenase catalysis26 are conserved in VanH and correspond to residues Arg‐231, Glu‐260, and His‐292 in VanH_A. Structure‐based analysis of lactate dehydrogenase also suggests that a cluster of hydrophobic residues in the active site is responsible for the preferential binding of the hydrophobic sidechain of d‐isomers of the substrate.26, 28, 29, 30 Whether this model can be directly applied to VanH enzymes necessitates further structural and functional studies.

2.1.2. VanA, VanB, VanD, VanF, and VanM d‐Ala‐d‐Lacligases

The d‐alanyl‐d‐lactate ligases catalyze the formation of the alternative precursor depsipeptide and share significant similarity (between 40 and 60% identity) with the d‐alanyl‐d‐alanine ligases DdlA and DdlB, the components of the regular peptidoglycan biosynthetic pathway. Compared to the Ddl ligases which possess narrow specificity for synthesis of d‐Ala‐d‐Ala,31 Van ligases demonstrate gain of function capabilities supporting not only the formation of a peptide bond between d‐Ala with various d‐amino acids but also the formation of an ester bond between the carboxyl terminus of d‐Ala and d‐lactate or α‐hydroxy acids.17, 32 d‐Ala‐d‐lac is further utilized by MurF,10 which catalyzes the formation of UDP‐MurNAc‐depsipeptide[d‐lac], thereby replacing the native peptidoglycan precursor with the vancomycin‐resistant variant. ATP is utilized as a co‐substrate by the d‐Ala‐d‐Ala and d‐Ala‐d‐lac ligases in a two‐step reaction whereby the γ‐phosphate of ATP is transferred to the carboxyl group of d‐Ala, followed by the second step of condensation of the amino group of d‐Ala or the hydroxyl group of d‐lac with the acyl group of the activated d‐Ala, liberating free phosphate and the product d‐Ala‐d‐Ala or d‐Ala‐d‐lac.12

Important molecular insights into the activity of d‐alanyl‐d‐lactate ligases were gained from structure‐based biochemical analysis of the canonical VanA representative of this enzyme family.33, 34, 35 The crystal structure of VanA from a VRE determined in complex with ADP and a phosphinate inhibitor (a transition state mimic)34 revealed a two‐domain composition similar to that of DdlB, featuring the PreATP‐grasp (PF01820,36) and ATP grasp (PF07478) domains at the N‐ and C‐termini, respectively. The ATP binding site is formed between the two lobes of the ATP grasp domain while both N‐ and C‐terminal domains are involved in positioning of the ligands (Figure 3). Both DdlB and VanA form two subsites for binding of d‐amino acid/d‐lactate ligands. One of the subsites engages d‐Ala for phosphorylation at its carboxyl end and the other positions the second d‐Ala/d‐lac for nucleophilic attack on the phosphoryl‐activated D‐Alag, leading to formation of d‐Ala‐d‐ala. In case of VanA, the specific amino acid composition of the second subsite was shown to be responsible for the altered substrate specificity.33, 34 In both DdlB and VanA this subsite is defined by a secondary structure element called the ω‐loop (Figure 3b). The systematic mutational analysis of this loop in VanA implicated His244 as the residue critical for activity on d‐lac (Figure 3a). The mutation of this residue to alanine increases the enzyme's K _m for d‐lac by more than 75‐fold while the K _m for d‐Ala only increased two‐fold.33, 34 At neutral pH, VanA His244 is suggested to abstract the d‐lactate‐OH proton. Notably, in DdlB the same position at the ω‐loop is occupied by a tyrosine. Other deviations in VanA's ω‐loop composition relative to DdlB could play a role in stabilizing the proper conformation of His244. In support of this model the ω‐loop sequence including the abovementioned histidine residue is conserved in VanB and VanD, indicative of a common enzymatic mechanism.

Figure 3.

Crystal structures of vancomycin resistance ligases. (a) Structure of VRE VanA•Mg²⁺•ADP•phosphinate inhibitor complex (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=1E4E 34). The ω‐loop that participates in subsite 2 for binding of the carboxyl‐terminal d‐Ala is labeled and zoom in shows details of residues, particularly His244, thought to confer specificity against d‐Ala‐d‐lac. In zoom, double‐ended arrow indicates potential clash between His244 and the ester bond of d‐Ala‐d‐lac should that compound be present in the active site. (b) Structure of VanG•ADP complex (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=4FU0 37). The ω‐loop of this enzyme is shown but is in a conformation away from subsite 2

A recent study by Wright and colleagues discovered a repertoire of resistance genes in 30,000 year‐old antibiotic‐naïve permafrost sediments (Yukon, Canada) including a vanHAX gene cluster variant and confirming the notion that antibiotic resistance has predated their clinical use.38 Purified “ancient” VanA, which shared 61% primary sequence identity with VanA from recently isolated VRE, showed robust d‐Ala‐d‐lac ligase activity. The crystal structure of this enzyme determined in complex with ATP was also highly similar to that of VRE VanA and superimposed with RMSD 0.6 Å over 275 matching Cα atoms. While the ω‐loop was largely disordered in the “ancient” VanA structure, likely due to the absence of corresponding substrate, it is highly similar in primary sequence to that in VRE VanA, suggesting that the evolution of d‐lac discrimination mechanism was an ancient event. Interestingly, this ancient VanA variant is most similar to d‐Ala‐d‐lac enzymes encoded by a vanHAX gene clusters found in Actinobacteria species (i.e., 94% identical to a sequence WP_020390273.1 from Kribella catacumbae). Since many glycopeptide producers belong to Actinobacteria,39 this observation is consistent with the hypothesis that antibiotic production and resistance are linked and that vanHAX cluster in pathogenic species could have been acquired from producing organisms.40, 41

High‐level vancomycin resistance is intrinsic to lactic acid bacteria, including those from the genuses Leuconostoc and Lactobacillus that are known to produce the d‐Ala‐d‐lac‐containing peptidoglycan precursors.42 This functionality is dependent on activity of d‐Ala‐d‐lac ligases such as LmDdl2 which shows a relative ratio of catalytic efficiency of 3:1 for formation of d‐Ala‐d‐lac versus d‐Ala‐d‐ala.31 The crystal structure of LmDdl2 determined in complex with ADP and a phosphinate d‐Ala‐d‐Ala analog43 revealed the specifics of d‐amino acid binding, with a phenylalanine residue occupying the critical substrate binding position in subsite two of the ω‐loop. By analogy with VanA, this and other observed deviations compared to the canonical d‐Ala‐d‐ala ligase such as DdlB are proposed to be responsible for LmDdl2 selectivity for d‐lac over d‐Ala.

2.1.3. VanX dipeptidase

The VanX enzyme belongs to the M15D subfamily of Zn²⁺‐ dependent peptidases that cleaves d‐Ala‐d‐Ala into individual d‐Ala residues, thus depleting the cytosolic pool of this dipeptide from the peptidoglycan synthesis pathway. VanX_A demonstrates much higher activity against d‐Ala‐d‐Ala compared to d‐Ala‐d‐Ser substrates. More importantly, this enzyme showed no detectable activity against d‐Ala‐d‐lac,44 thus allowing this compound to be incorporated into peptidoglycan.

VanX is an aminopeptidase, demonstrating higher specificity for the dipeptide's amino‐terminal amino acid. The crystal structure of VanX_A has been determined in complex with d‐Ala‐d‐Ala substrate and d‐Ala product,45 providing an explanation for the high specificity of this enzyme towards the d‐Ala‐d‐Ala substrate. The VanX_A protein features a central five stranded β‐sheet that contains two histidine residues and a glutamate residue that chelate the Zn²⁺ ion at the base of the pocket that accommodates the d‐Ala‐d‐Ala substrate (Figure 4). The Zn²⁺ ion engages the carbonyl group of the dipeptide, in line with its proposed role in stabilizing the hydrolysis transition state after nucleophilic attack by a water molecule activated by the catalytic base glutamate (Glu181 in VanX_A).44, 46 VanX's specificity toward a dipeptide substrate has been explained by the constraints of the relatively small substrate binding pocket. The position of the amino‐terminal d‐Ala in the VanX‐d‐Ala‐d‐Ala complex structure deep in the active site while the carboxy‐terminal d‐Ala faces outward provided the molecular explanation for aminopeptidase specificity. The peptide nitrogen of d‐Ala‐d‐Ala substrate in the complex structure is positioned close to the backbone carbonyl atoms of VanX residues Tyr109 and Val110 and to the sidechain of Glu181 (Figure 4). Docking of the d‐Ala‐d‐lac molecule into this position indicated that accommodation of this compound in the VanX active site would be energetically unfavorable due to the incompatibility of the ester oxygen of d‐Ala‐d‐lac with this local interaction network.

Figure 4.

Crystal structure of dipeptidase VanX_A•Zn₂₊•d‐Ala‐d‐Ala complex (personal communication from C.H. Park, also PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=1R44 = apoenzyme structure45). Left = cartoon + sticks representation, zoom shows the details of d‐Ala‐d‐Ala binding and H‐bond between carbonyl oxygen of Tyr109 and the amide bond of d‐Ala‐d‐Ala. Right = surface representation showing small size of d‐Ala‐d‐Ala binding cleft

2.1.4. VanS sensor kinase

This histidine kinase contains an N‐terminal membrane‐embedded domain, with some parts of it expected to protrude into the extracellular space, and a C‐terminal cytoplasmic kinase domain. In the presence of a specific signal, the identity of which remains debated as will be discussed below, VanS catalyzes auto‐phosphorylation of the His164 residue localized on the cytoplasmic side of the protein and the phosphoryl‐group is then transferred to Asp53 of the VanR regulator, triggering transcriptional activation of the van regulon.47, 48, 49 In the absence of the signal, VanS catalyzes dephosphorylation of VanR suggesting that this protein also possesses phosphatase activity.18, 49 Direct interaction between VanS_A and VanR_A supposedly through the cytosolic domain of VanS_A has been reported.50

The earliest model of VanS_A activity was that it senses the presence of vancomycin through a direct interaction between this molecule at the protein's extracellular face. Recognition of vancomycin would trigger VanS_A dimerization, autophosphorylation, and transfer of the phosphoryl group to VanR_A. Experimental evidence for this model has been provided in a study that used biophysical approaches such as analytical ultracentrifugation and near‐UV circular dichroism (CD) to monitor the changes in oligomerization and conformation of detergent‐solubilized VanS_A in the presence of vancomycin.51 Another study further supported this model demonstrating that vancomycin induced significant thermostability and conformational changes in purified VanS_A, while other potential signals such as the peptidoglycan components Ala‐d‐γ‐Glu‐Lys‐d‐Ala‐d‐Ala, N‐acetylmuramic acid or d‐Ala‐d‐Ala had no such effect.52 However, other studies contested this model, reporting the lack of a VanS_A‐vancomycin complex even at high concentrations of vancomycin and suggesting that the enzymatic activities of detergent or amphipol‐solubilized VanS_A are not altered in the presence of vancomycin.50 A possible explanation of the contradictory results of these latter studies can be due to the VanS_A solubilization using detergents, which are known to affect the function of sensor kinase membrane proteins.50, 53 Furthermore, the in vitro experimental setup in these studies also lacks the target of vancomycin – the d‐Ala‐d‐Ala group of lipid II, and thus do not necessarily reflect the natural environment of VanS wherein the protein could form a proposed VanS_A‐lipid II‐vancomycin complex.54

In an attempt to minimize the effect of detergent, the S. coelicolor VanS (VanSSc) was recombinantly expressed and purified from S. coelicolor and E. coli membranes and shown to bind a photo‐labeled vancomycin derivative.55 While this study did not explicitly rule out the presence of residual peptidoglycan/lipid II in the membrane fraction that could mediate VanSSc‐vancomycin interactions, the fact that VanSSc derived from both native and recombinant expression systems demonstrate the ability to bind vancomycin is consistent with formation of a direct binary VanSSc‐vancomycin complex. It should be noted that VanS proteins from Actinobacteriaceae show significant divergence in sequence from VanS_A and thus this type of analysis must be performed on other VanS variants in order to determine the nature of their signal molecule.

An alternative model for VanS activation suggests that it recognizes a cellular component that accumulates as a consequence of vancomycin binding to lipid II, or, that VanS senses a general change in membrane dynamics induced by antibiotic binding.56, 57, 58, 59 A hybrid model also has been proposed whereby activation of VanSSc requires binding of vancomycin to lipid II, resulting in formation of a VanSSc‐vancomycin‐d‐Ala‐d‐Ala complex.54 This last model has been supported by data showing that induction of the van resistance cascade by VanSSc is correlated with the level of d‐Ala‐d‐Alacontaining peptidoglycan in engineered S. coelicolor cells with tunable d‐Ala‐d‐Ala versus d‐Ala‐d‐lac peptidoglycan.54

To summarize, several important aspects of VanS activity remain unresolved, primarily, the precise nature of the signal that triggers activation of VanS. As described above, the identity of the ligand remains enigmatic and the aggregate data also suggests that it may vary across the vancomycin resistance types. Other important questions that require further investigation are how the recognition of this signal is transformed into VanS kinase activity and what is the molecular mechanism enabling the switch between kinase and phosphatase activity of VanS. Determination of the molecular structures of VanS family representatives would greatly facilitate these studies.

2.1.5. VanR transcription regulator

VanR is a canonical transcriptional regulator comprised of a N‐terminal response regulator receiver (REC) domain and a C‐terminal winged helix‐turn‐helix DNA binding domain. More specifically, VanR falls into the OmpR‐PhoB subclass of response regulators.18 As mentioned above, VanR_A is phosphorylated on Asp53 by activated phospho‐VanS_A and this phosphorylated form is capable of binding to the promoters P _res and P _reg which drive the transcription of the vanHAX operon and the vanR _A and vanS _A genes, respectively.18, 49, 60, 61 Notably, VanR_A can also be phosphorylated in the absence of phospho‐VanS_A by acetyl phosphate or by other promiscuous sensor histidine kinases.60 Therefore, it is thought that the vanR _A and vanS _A genes are constitutively expressed in the absence of vancomycin but VanR_A is dephosphorylated by VanS_A, preventing significant expression of the VanA‐type resistance cascade. The presence of vancomycin activates VanS_A and increases the pool of phospho‐VanR_A, thereby increasing expression at P _res and P _reg.60

2.1.6. VanY pentapeptidase

Representatives of this protein family play an “accessory” role in the Van cascade and are not strictly required for vancomycin resistance, however their expression is induced by vancomycin and it does contribute to higher levels of resistance.19, 20 VanY enzymes contain a single N‐terminal transmembrane helix and a cytosol‐oriented catalytic domain that possesses d,d‐carboxypeptidase activity against UDP‐MurNAc‐pentapeptide[d‐Ala] and Lipid II[d‐Ala] but not against Lipid II[d‐Ala‐d‐Lac].19, 20, 62 Therefore, VanY_A's contribution to vancomycin resistance is in lowering the availability of regular Lipid II[d‐Ala] precursor and as a result enriches peptidoglycan with vancomycin‐resistant peptidoglycan precursors terminating in d‐Lac. Interestingly, most VanY enzymes share no significant sequence similarity with penicillin‐binding proteins with d,d‐carboxypeptidase activity except for VanY_D, which is 30 to 35% identical in sequence with penicillin‐binding proteins such as PBP5 and PBP6b.63, 64 Accordingly, this VanY variant is the only representative of this family which demonstrates sensitivity to penicillin G.62, 65, 66

Despite sharing Zn²⁺‐dependent metalloprotease activity, VanY shows little sequence similarity to VanX enzymes described above beyond the conservation of key catalytic residues. Furthermore, unlike VanX, VanY is a carboxypeptidase and belongs to the distinct M15B subfamily.67 Accordingly, prior to the availability of structural information about VanY, key insights into molecular mechanism of this enzyme activity were gained from comparative analysis to the VanXY bifunctional dipeptidase/pentapeptidase enzymes from the VanC and VanG‐type resistance types.67 The molecular structure and the details of the bifunctional activity of VanXY enzymes will be discussed later in this review, but the key finding from this study relevant for VanY enzymes was that the VanY/VanXY enzymes share a common structural element controlling entry to the active site that is distinct from the one in VanX enzymes even though the residues making up the active centers in these proteins are conserved. This major structural distinction results in a different position of the d‐Ala‐d‐Ala groups in the active site, thus presenting distinct VanY active site residues for recognition of the sidechains of d‐Ala‐d‐Ala and allowing for the carboxypeptidase specificity of VanY and VanXY. Another major distinction between VanX and VanXY relevant for VanY enzymes is that the overall active site cleft of VanXY is much larger than for VanX, thereby providing enough space to accommodate UDP‐MurNAc‐pentapeptide[d‐Ala] or lipid II[d‐Ala]. A similar larger active site cleft nature was predicted for VanY. More recently, the crystal structure of the VanY_B representative was determined in apo form as well as in complex with the d‐Ala‐d‐Ala and d‐Ala ligands.68 As predicted, the VanY_B structures revealed a large active site exposed to the solvent with the entry path to the Zn²⁺ active center distinct from that established for VanX (Figure 5a). VanY_B was shown to have a binding affinity (K _D) of 246 μM for l‐Ala‐d‐γ‐Glu‐l‐Lys‐d‐Ala‐d‐Ala pentapeptide while no binding was detected for d‐Ala‐d‐ala dipeptide68; this observation is consistent with a model in which the VanY active site is better adapted to accommodate the larger pentapeptide substrate. Determination of the structure of VanY_B in complex with d‐Ala‐d‐Ala was achieved by substituting Cu²⁺ for Zn²⁺ in the active site. Such substitution inhibited the enzyme activity and may have allowed trapping of the d‐Ala‐d‐Ala compound in the active site. Given the substrate preference of VanY_B, a crystal structure of a VanY representative in complex with pentapeptide[d‐Ala] or UDP‐MurNAc‐pentapeptide[d‐Ala] would delineate additional important interactions between the enzyme and its ligands.

Figure 5.

(a) Crystal structures of pentapeptidase VanY_B in complex with Cu²⁺ and d‐Ala‐d‐Ala (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=5ZHW 68), shown in the same orientation as VanX_A in Figure 4. Left = two views are shown, rotated 120°. d‐Ala‐d‐Ala shown in black sticks. Right = bottom view in surface representation, showing larger size of active site cleft. (b) Structure of the VanXY_C•Cu²⁺•d‐Ala‐d‐Ala complex (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=4OAK 67), shown in the same orientation as VanX_A in Figure 4. The bisubstrate selectivity loop and Leu113 in this loop, which interact with d‐Ala‐d‐Ala, are shown in red

2.1.7. VanZ

The VanA‐type resistance gene cluster contains another gene downstream from vanY that was called vanZ. The corresponding protein contains five predicted transmembrane helices and shows no sequence similarity to known domains/proteins. This gene appears dispensable for vancomycin resistance but is required for conferring resistance to teicoplanin.21, 22 VanZ_A does not affect the population of UDP‐MurNAc‐pentapeptide[d‐Ala] or UDP‐MurNAc‐pentapeptide[d‐lac] and its role in the Van cascade is currently not known.21, 22 VanZ_A was recently purified to homogeneity and its localization to the membrane validated.69

Of note, a second vanZ was discovered in a VanF‐type (d‐Ala‐d‐Lac) resistance cassette in Paenibacillus papillae.70 The corresponding VanZ_F protein shared 21% primary sequence identity with VanZ_A but the significance of this protein in the resistance cascade remains unknown. In Clostridiodes difficile a VanZ homologue CD1240 (38% identity to VanZ_A) is encoded on the chromosome outside the van regulon and was shown to be required for resistance to teicoplanin but not to vancomycin.71 These three VanZ orthologs belong to the “VanZ superfamily” (PF04892 in Pfam), which comprises 7,428 sequences present in Firmucutes (Bacilli, Clostridia, Proteobacteria and Bacteroidetes) and Actinobacteria. While the majority of these proteins are comprised of VanZ sequence exclusively, some representatives combine the VanZ sequence with RDD domain (PF06271), which is a domain also of unknown function.

2.2. d‐Ala‐d‐Ser based resistance (VanC, VanE, VanG, VanL, and VanN types)

A distinct mechanism of vancomycin resistance that was named the VanC resistance type was identified in Enterococcus gallinarum, Enterococcus cassiflavus, and Enterococcus flavescens.72, 73 Subsequently, resistance cassettes with a similar core set of genes were identified in Enterococcus faecalis and were classified as VanE, VanG, VanL, and VanN resistance types.74, 75, 76, 77 More recently, Clostridium difficile species carrying VanG‐type resistance were also discovered.78

In contrast to the VanA resistance type, the VanC to VanN cassettes enable the synthesis of alternative vancomycin‐resistant peptidoglycan precursors terminating in d‐Ala‐d‐Ser rather than d‐Ala‐d‐Lac. The additional hydroxyl group of d‐Ser relative to d‐Ala introduces steric hindrance for interaction with vancomycin which lowers the drug's affinity by 6‐fold79 thus conferring lower levels of resistance (MIC 4 to 32 mg/L) than the d‐Ala‐d‐Lac based mechanism. These cassettes (Figure 2) contain the vanR and vanS regulator/sensor kinase pair, similar to those described for the VanA resistance type, a d‐Ala‐d‐Ser ligase encoded depending on the specific type of cassette by vanC, vanE, vanG, vanL or vanN, and two distinct genes not found in d‐Ala‐d‐lac‐based resistance cassettes: a serine racemase encoded by vanT and a bifunctional dipeptidase/pentapeptidase encoded by vanXY. Below, we discuss the individual structural and enzymatic features of the Van proteins representative of this resistance mechanism.

2.2.1. VanC, VanE, VanG, VanL, and VanN d‐Ala‐d‐Ser ligases

These enzymes catalyze synthesis of d‐Ala‐d‐Ser from the corresponding amino acids. The VanG member of this family has also been shown to catalyze formation of other d‐Ala‐containing dipeptides.37 These enzymes show significant deviation in primary sequence from the d‐Ala‐d‐lac and d‐Ala‐d‐Ala ligases.37 However, structural characterization and functional analysis of VanG37 suggested that the three subfamilies of d‐Ala‐d‐X ligases share a common overall structure, including such molecular features as binding of ATP and coordination of the N‐terminal d‐Ala residue (subsite 1), but dramatically differ in the sequence composition and as a result in structural features of their ω‐loops and subsite 2 (Figure 3b). In line with this hypothesis, replacement of the ω‐loop sequence of VanA (d‐Ala‐d‐Lac ligase) with that of VanC2 (d‐Ala‐d‐Ser ligase) changed the former enzyme's specificity toward synthesis of d‐Ala‐d‐Ala and d‐Ala‐d‐Ser rather than d‐Ala‐d‐Lac.33 This data along with site‐directed mutagenesis probes into the role of individual the ω‐loop residues of VanC2 (i.e., Lys255, which is expected to interact with the phosphoryl transition state) for d‐Ala‐d‐Ser ligase activity33 suggested that specific residues in this loop control d‐Ala‐d‐Ser specificity. Validation of this model would be possible from further structural analysis of d‐Ala‐d‐Ser ligase family representatives in complex with their amino acid substrates.

2.2.2. VanT racemases

The vanT genes found in the VanC, VanE, VanG, and VanN resistance cassettes encode a membrane‐bound serine racemase.77, 80, 81, 82, 83, 84 In the VanL cassette the membrane and cytoplasmic catalytic domains of VanT are encoded by separate genes, vanT _mL and vanT _rL.76

The membrane‐embedded portion of VanT containing 10 predicted transmembrane helices belongs to the PF01757 domain (Pfam database) and possesses the l‐serine transporter activity that is required for vancomycin resistance.85 It has been proposed that this domain stimulates the racemase activity of this protein by increasing the concentration of its substrate, however, no structural information is available for this domain or other representatives of the Membrane acyl transferase superfamily (Pfam clan CL0316).

The VanT racemase domain is a type II pyridoxal‐phosphate (PLP)‐dependent enzyme that catalyzes conversion of l‐Ser to d‐Ser, the essential substrate for the d‐Ala‐d‐Ser‐based resistance mechanism. VanT enzymes are sequence‐related to d‐Alanine racemases, known as Alr enzymes, and also to broad‐spectrum amino acid racemases.86, 87, 88 Accordingly, VanT enzymes demonstrate l‐Ala racemase activity but it is significantly lower as compared to that of Alr enzymes.80, 81, 86 The crystal structure of the VanT_G catalytic domain revealed two subdomains containing TIM barrel and β‐barrel folds, respectively.86 The VanT_G catalytic domain forms a dimer with a “head to tail” arrangement of subunits, where the C‐terminal subdomain contributes to the active site formed by the N‐terminal subdomain from the partner chain (Figure 6). Accordingly, both subunits in the VanT_G dimer provide residues to each d‐amino acid binding site. More specifically, the C‐terminal β‐barrel subdomain of VanT_G carries the Asn696 residue which participates in substrate amino acid positioning and interacts with the Tyr543 and Ser567 residues from the TIM barrel subdomain of the partner subunit (Figure 6). Mutation of Asn696 to a tyrosine which occupies the corresponding position in the Alr enzyme from Bacillus stearothermophilus (BsAlr) dramatically increases the activity of VanT_G for l‐Ala. Accordingly, the reverse mutation in BsAlr (Y354N) was shown to increase serine racemase activity 62‐fold.90 Furthermore, the VanT_G variant with three active site residues substitutions designed to mimic the active site of BsAlr (543A/S567P/N696Y) demonstrated an activity ratio against l‐Ala vs. l‐Ser substrates comparable to that of BsAlr. Since the combination of these three residues is fully conserved only in VanT enzymes compared to the other d‐amino acid racemases,81 it was proposed that VanT enzymes evolved from a putative Alr‐like ancestor through gain‐of‐function thereby expanding on the pre‐existing “housekeeping” racemase activity.86

Figure 6.

Crystal structure of VanT_G racemase (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=4ECL 86). Two chains of the homodimer are shown in thin line and cartoon representation. PLP‐d‐Ser is from the structure of alanine racemase (http://bioinformatics.org/firstglance/fgij//fg.htm?mol=1L6F 89) that was superimposed on VanT_G. Asn696, which is in position to interact with d‐Ser, along with nearby residues, are highlighted in the zoom‐in detail

2.2.3. VanXY bifunctional dipeptidase/pentapeptidase

The vanXY gene is found in all the d‐Ala‐d‐Ser‐based resistance operons and codes for an enzyme with hydrolytic activity against d‐Ala‐d‐Ala and UDP‐MurNAc‐pentapeptide[d‐Ala]. VanXY also demonstrates a basal activity against d‐Ala‐d‐Ser but not against UDP‐MurNAc‐pentapeptide[d‐Ser].91, 92 Key molecular insights into the unique bifunctional specificity of VanXY were uncovered in the structural characterization of the VanXY_C and VanXY_G enzymes.67

These crystal structures of VanXY enzymes revealed an overall fold typical of M15 Zn²⁺‐dependent peptidase as seen in the structures of VanX and VanY described earlier. However, detailed analysis of the VanXY active site suggested that the path to the active center as well as the binding position of d‐Ala‐d‐Ala was distinct from that seen in VanX (Figure 5b). In VanXY, the carboxylate of d‐Ala‐d‐Ala is positioned deep into the active site in line with the d,d‐carboxypeptidase specificity of VanXY and VanY enzymes. Furthermore, two tightly packed hydrophobic pockets accommodating the sidechains of d‐Ala dipeptide would introduce steric clashes in case of d‐Ser at the Cterminus of the dipeptide, thus providing a possible explanation for the discrimination against the vancomycin‐resistant d‐Ser‐terminating peptidoglycan fragments. Mutagenesis analysis highlighted the role of a prominent loop element in the VanXY structures, which partially covered the large active site. This loop, called the bi‐substrate selectivity loop (Figure 5b), was proposed to be involved in binding both the d‐Ala‐d‐Ala dipeptide or UDP‐MurNAc‐pentapeptide[d‐Ala], as specific residues of this loop were critical for VanXY activity against each of the two substrates. Based on the structure of VanXY_C in complex with d‐Ala‐d‐Ala, a functional model was proposed whereby the bi‐substrate selectivity loop in a “closed” conformation contributes to binding interactions with the smaller dipeptide substrate, while in an “open” one allows for binding of the larger UDP‐MurNAc‐pentapeptide[d‐Ala] substrate. Interestingly, the subsequently determined structure of VanY_B 68 showed that this enzyme lacks the bisubstrate selectivity loop, consistent with the inability of VanY_B to hydrolyze d‐Ala‐d‐Ala. The precise role of the bi‐substrate selectivity loop remains to be elucidated through structure–function studies of VanXY interactions with its pentapeptide[d‐Ala] or UDP‐MurNAc‐pentapeptide[d‐Ala] substrates.

2.3. Additional vancomycin resistance genes

2.3.1. VanU_G transcription regulator

The VanG resistance cassette contains the vanU _G gene upstream from vanR _G and these three genes are co‐transcribed from the same promoter P _UG.84 vanU _G codes for a 75‐residue protein belonging to the XRE transcription regulator family. The structure of VanU_G determined in two crystal forms (PDB codes http://bioinformatics.org/firstglance/fgij//fg.htm?mol=3TYR and http://bioinformatics.org/firstglance/fgij//fg.htm?mol=3TYS) reveals a tight dimer with each chain adopting a canonical helix‐turn‐helix motif featuring a recognition helix rich in basic residues (Figure 7). The role of this protein in regulation of the vancomycin resistance was not understood until a recent study showing that VanU_G negatively regulates the expression from both P _UG and P _YG (the resistance) promoters thereby repressing transcription of all genes in this Van cassette.93 Notably, phospho‐VanR_G was shown to not bind to P _UG region but instead to compete with VanU_G for binding to P _YG.93 Thus, it was suggested that VanU_G fine‐tunes the expression levels of the vancomycin resistance genes by dampening the activation by phospho‐VanR_G.93 This model requires further investigation, such as a comparison of the in vitro affinity of VanU_G vs. VanR_G for P _YG, studies on the oligomeric assembly of VanU_G with DNA, or crystal structure determination of VanU_G in complex with DNA.

Figure 7.

Crystal structure of VanU_G (PDB http://bioinformatics.org/firstglance/fgij//fg.htm?mol=3TYS). Two chains in the homodimer and the helix‐turn‐helix are indicated. Putative DNA binding residues in recognition helix are labeled

2.3.2. VanW

The VanB and VanG resistance cassettes contain vanW genes which code for 275 and 281‐residue long proteins that share 52% primary sequence identity.84, 94 The molecular function of VanW and its role in vancomycin resistance remain unknown since these proteins show no significant similarity to characterized proteins outside the VanW superfamily (Pfam PF04294). This family includes over 3,000 representatives widespread in Firmucutes (Bacilli, Clostridia) and Actinobacteria. “VanW‐like” domains are found in proteins such as penicillin‐insensitive l,d‐transpeptidase from E. faecium, that also contain peptidoglycan‐binding domains. Notably the gene encoding this enzyme is proximal to a gene coding for l‐lactate dehydrogenase; with these two activities reminiscent of VanY and VanH, respectively. The function of VanW awaits experimental characterization.

2.3.3. S. coelicolor proteins VanJ, VanK

S. coelicolor contains an unusual vancomycin resistance cluster comprised of seven genes vanSRJKHAX.23 The vanJ gene codes for a 330‐residue protein containing 3 predicted transmembrane helices and a C‐terminal domain belonging to the PF03372 domain (Pfam database) featuring enzymes with endonuclease, exonuclease or phosphatase activities. VanJ was shown to be localized to the membrane with its C‐terminal domain oriented extracellularly.

It has been reported that deletion of vanJ led to increased susceptibility to vancomycin,23 however a subsequent study also reported that this gene does not contribute to resistance to vancomycin but rather is essential for resistance to teicoplanin and to teicoplanin‐like glycopeptides.95 Furthermore, vanJ gene variants were identified in many Actinomycetes in proximity to the putative vanRS genes and biosynthetic gene clusters, suggesting a possible role for VanJ in self‐resistance. While several possible functions for VanJ have been proposed including undecaprenol pyrophosphate recycling or Lipid II[d‐Ala] degradation95 the specific function of this protein remains enigmatic and needs further investigation.

The S. coelicolor vanK gene encodes a 397‐residue protein belonging to the Fem family of proteins (Pfam PF02388) which are non‐ribosomal peptidyltransferases that catalyze branching of the pentapeptide precursors of peptidoglycan. This gene was shown to be essential for vancomycin resistance in this species.23 Based on sequence similarity with Fem‐family proteins from S. aureus that catalyze the cross‐bridge between pentapeptide precursors,96 it has been shown that VanK catalyzes formation of the glycine cross‐bridge between UDP‐MurNAc‐pentapeptide[d‐Lac] precursors and the ΔvanK strain is non‐viable in the presence of vancomycin.23, 97 This contrasts with the product of the femX gene in S. coelicolor which is unable to catalyze the glycine cross‐bridge between UDP‐MurNAc‐pentapeptide[d‐Lac] precursors and is restricted to UDP‐MurNAc‐pentapeptide[d‐Ala] precursors; therefore, VanK is expected to possess molecular recognition of the d‐Lac moiety while FemX is not able to accommodate this modification.

2.4. Outlook and open questions

After more than 30 years of studies of the genetic and molecular basis of vancomycin resistance, we have made significant progress in visualizing the molecular framework supporting most of the steps in the vancomycin resistance cascade. Recent advances in structural characterization of Van proteins including the d‐Ala‐d‐Ala ligases, the VanX dipeptidase and VanT racemase provided molecular images and insights into the biochemical activity of these essential resistance enzymes. In conjunction with previous analysis the new data have highlighted the unifying tendency that despite their ability for discriminating between vancomycin sensitive and resistance peptidoglycan precursors, Van enzymes share significant sequence similarity with housekeeping enzymes involved in bacterial cell envelope maintenance. This observation is in line with the hypothesis of evolution of resistance enzymes from non‐resistance ancestry without major changes of general fold but rather through key specific changes in active site composition and architecture.98

Despite these recent advances, gaps still remain in our understanding of some molecular aspects of this resistance mechanism particularly in membrane‐bound components of this cascade. Structural biology methods are capable of shedding light on these gaps and here we suggest such research avenues.

The role of the VanT membrane domain as a l‐Ser transporter remain poorly understood. Along the same lines, further molecular data on VanT interactions with d‐Ser could expand our understanding of specific aspects of racemase activity of this protein.86 For VanH, a molecular structure would shed light on whether it is homologous to d‐lactate dehydrogenase enzymes and contains amino acids that favor the conversion of pyruvate to lactate.

For d‐Ala‐d‐Ala ligases, the structure of VanA in complex with d‐Ala‐d‐Ala would inform on conformational differences between binding of this natural ligand versus the phosphinate d‐Ala‐d‐Ala analog the interactions of which with VanA has been visualized. Structures of VanB and VanD would provide a molecular framework for the role of their ω‐loops and the nearby residues in subsite 2 for d‐Lac discrimination. For d‐Ala‐d‐Ser ligases, a structure with an intact ω‐loop and in complex with ligands is still lacking and would help to understand how these enzymes accommodate d‐Ser in subsite 2.

Structures of VanY and VanXY in complex with larger peptidoglycan fragments such as pentapeptide[d‐Ala], UDP‐MurNAc‐pentapeptide[d‐Ala] or Lipid II[d‐Ala] would reveal the key interactions between the enzyme and substrate's amino acids beyond the d‐Ala‐d‐Ala termini. Such structures would also validate our hypothesis that the VanXY bisubstrate selectivity loop changes conformation away from the active center to accommodate these larger ligands; based on our mutational analysis67 we suggested that the loop actively participates in interactions with the ligands.

One of the key unclarified aspects of glycopeptide resistance remains the nature of the activating ligand and its ability to activate the VanS receptor kinase. Up to now, most research has focused on the VanS_A and VanSSc representatives of this family with the consensus notion that both these receptors interact directly with vancomycin. Further studies should shed light on the molecular structure of these membrane proteins and on the conformational changes triggered by interactions with the ligand. The possibility that more than one signal molecule may be involved in activation of other vancomycin resistance cascade also needs further investigation; the activating molecule may differ across the VanS variants as suggested by significant sequence deviation. Studies should focus on obtaining molecular structural information about these membrane proteins in the presence and in the absence of their corresponding signal molecule. Another open research avenue is investigations into conformational changes of VanS upon its recognition with its ligand, that is, how this protein is able to alternate between the phosphatase and kinase activities.

The understanding of downstream activation of expression of van gene clusters is also incomplete due to the lack of molecular data on activation of signal transduction between VanS and VanR. Structures of VanR and phospho‐VanR, as well as its complex with operator DNA, would illuminate how phosphorylation changes its DNA binding capacity, ostensibly through conformational changes as seen for other OmpR‐family regulators.99 Furthermore, the role of VanU as an additional regulator of transcription of van genes and its interactions with VanR needs further analysis. A structure of a VanU_G‐DNA complex would shed light on how this protein competes with VanR_G for binding to the same promoter region.

Perhaps the greatest benefit from future studies would arise from molecular characterization of the vancomycin resistance proteins that are less well studied: VanW, VanZ, VanJ and VanK. The ability of VanK to catalyze the cross‐bridge between UDP‐MurNAc‐pentapeptide[d‐Lac] that is lacking in its ortholog FemK, perhaps through molecular recognition of d‐Lac by VanK, deserves investigation and should be explained through structural analysis. Notably, VanZ and VanJ are membrane‐bond proteins. Given recent advancement of molecular and structural methodologies to study this type of proteins we anticipate that the next breakthrough in our understanding of vancomycin resistance mechanism would come from molecular characterization of membrane‐associated components of this complex cascade.

Stogios PJ, Savchenko A. Molecular mechanisms of vancomycin resistance. Protein Science. 2020;29:654–669. 10.1002/pro.3819