key: cord-0268213-gbnea49e
authors: Mikati, Marwa O.; Miller, Justin J.; Osbourn, Damon M.; Ghebremichael, Naomi; Shah, Ishaan T.; Burnham, Carey-Ann D.; Heidel, Kenneth M.; Yan, Victoria C.; Muller, Florian L.; Dowd, Cynthia S.; Edwards, Rachel L.; Odom John, Audrey R.
title: Antimicrobial prodrug activation by the staphylococcal glyoxalase GloB
date: 2020-07-23
journal: bioRxiv
DOI: 10.1101/2020.07.23.214460
sha: e6233bde8d8bb74949f4ef0cfdfe3ae9653e98c7
doc_id: 268213
cord_uid: gbnea49e

With the rising prevalence of multidrug-resistance, there is an urgent need to develop novel antibiotics. Many putative antibiotics demonstrate promising in vitro potency but fail in vivo due to poor drug-like qualities (e.g. serum half-life, oral absorption, solubility, toxicity). These drug-like properties can be modified through the addition of chemical protecting groups, creating “prodrugs” that are activated prior to target inhibition. Lipophilic prodrugging techniques, including the attachment of a pivaloyloxymethyl group, have garnered attention for their ability to increase cellular permeability by masking charged residues and the relative ease of the chemical prodrugging process. Unfortunately, pivaloyloxymethyl prodrugs are rapidly activated by human sera, rendering any membrane permeability qualities absent during clinical treatment. Identification of the bacterial prodrug activation pathway(s) will allow for the development of host-stable and microbe-targeted prodrug therapies. Here, we use two zoonotic staphylococcal species, S. schleiferi and S. pseudintermedius, to establish the mechanism of carboxy ester prodrug activation. Using a forward genetic screen, we identify a conserved locus in both species encoding the enzyme hydroxyacylglutathione hydrolase (GloB), whose loss-of-function confers resistance to carboxy ester prodrugs. We enzymatically characterize GloB and demonstrate that it is a functional glyoxalase II enzyme, which has the capacity to activate carboxy ester prodrugs. As GloB homologs are both widespread and diverse in sequence, our findings suggest that GloB may be a useful mechanism for developing species-or genus-level prodrug targeting strategies.

INTRODUCTION absorption or solubility. For example, the third-generation cephalosporin, cefditoren, is poorly

Selection of prodrug-resistant staphylococci. 10

In our previous study, we identified phosphonate antibiotics with activity against zoonotic 11 staphylococci (S. schleiferi and S. pseudintermedius) (27) . Lipophilic carboxy ester prodrug 12 modification of these phosphonates dramatically increases antistaphylococcal potency, 13 We find that POM-ERJ-resistant staphylococci remain equally sensitive to non-prodrugged 5 compounds (such as ERJ analogues) and the third-generation cephalosporin cefditoren. In 6 contrast, POM-ERJ-resistant staphylococci exhibit significantly increased MICs to multiple 7 classes of lipophilic ester prodrugs, exhibiting cross-resistance to both cefditoren pivoxil (cell 8 wall inhibitor) and POM-HEX (inhibitor of enolase) (Fig. 4, Table S2 ). Thus, POM-ERJ-resistant 9 staphylococci are cross-resistant to other POM-prodrug inhibitors, regardless of the intracellular 10 target. Our data suggest that POM-prodrugs follow a common and conserved activation 11 mechanism that has been disrupted in our POM-ERJ-resistant isolates. 12 13 Figure 4 . Cross-resistance to lipophilic ester prodrugs in POM-ERJ-resistant S. schleiferi. WT and POM-ERJ resistant S. schleiferi were treated with the compounds displayed in Figure 3 . Compounds are grouped by mechanism of action and color coded to indicate whether a given compound is a carboxy ester prodrug. Displayed are the mean values of the fold change (resistant isolate/WT) of three independent experiments performed in technical duplicate. * indicates compounds whose MIC values were too high to measure. Numerical data additionally provided in Table S2. ester bond, we selected an additional highly lipophilic antibiotic, mupirocin, which inhibits protein 9 biosynthesis (Fig. 3) . POM-ERJ-resistant staphylococci were not cross-resistant to mupirocin, 10 further supporting that prodrug resistance in these strains is specific to the carboxy ester bond 11 of the prodrug (Fig. 4) . 12 13 POM-ERJ resistant staphylococci are enriched in mutations in the GloB gene. 14 To characterize the genetic changes associated with carboxy ester prodrug resistance, we 15 performed whole genome sequencing of prodrug resistant isolates of both S. schleiferi and S. 16 pseudintermedius. The whole genomes of each isolate were compared to the respective 17 parental genome and candidate genetic changes were verified by Sanger sequencing. We 18 prioritized nonsynonymous genetic changes that were represented in more than one strain. A 19 complete list of identified mutations is found in Table S3 . 20

In both independent genetic screens, we found that prodrug resistant staphylococci were 22 enriched in mutations in an evolutionarily conserved locus. We identified multiple isolates (3/16 23 S. schleiferi, 14/18 S. pseudintermedius) with sequence modifications in the locus annotated as 24 hydroxyacylglutathione hydrolase, gloB (LH95_06060 in S. schleiferi, SPSE_1252 in S. 25 pseudintermedius, Table S3 ). Most genetic changes in gloB were nonsynonymous single nucleotide polymorphisms, though two nonsense alleles that would truncate approximately 50% 1 of the protein were also identified (Fig. 5 , Table S3 ). In several strains, the only genetic variation 2 that distinguished WT and resistant genomes was within the gloB locus. 3 4 Of the 17 identified GloB mutations, 12 unique alleles were identified in prodrug-resistant 5 staphylococci. Using PROVEAN, an algorithm which quantifies the predicted impact of amino 6 acid substitutions on protein function, each of these 12 alleles is predicted to have deleterious 7 effects on protein function (below the threshold score of -2.5) (Fig. 5 ) (38). S. schleiferi and S. 8

pseudintermedius are non-model organisms that possess endogenous CRISPR-Cas9 systems 9 and transformation of these organisms has not yet been described (39). Attempts to ectopically 10 complement gloB mutant strains with WT GloB (>90 independent transformation attempts using 11 Figure 5 . POM-ERJ resistant staphylococci are enriched for mutations in the locus encoding hydroxyacylglutathione hydrolase (GloB). (A) Locations and identities of GloB mutations discovered by whole-genome sequencing and independently verified by Sanger sequencing. Line coloring represents predicted impact of a given mutation on GloB function, scores below -2.5 are predicted to be deleterious. (B, C) Homology models of S. schleiferi (B) and S. pseudintermedius GloB generated using SWISS-MODEL. Residues found to be mutated in POM-ERJ resistant staphylococci explicitly shown in blue. established methods for S. aureus, S. epidermidis, and B. subtilis) were unsuccessful in 1 recovering transformed colonies, despite preparing plasmid from the S. aureus restriction 2 deficient cloning intermediate, RN4220, and the cytosine methyltransferase negative E. coli 3 mutant, DC10B (40-46). However, the independent selection of 12 unique loss-of-function 4 alleles in two different species strongly suggests that loss of GloB function is responsible for 5 prodrug resistance in S. schleiferi and S. pseudintermedius. 6 7 Structural basis of GloB loss-of-function. 8

As prodrug-resistance mutations in GloB map along its entire linear sequence, we next 9 examined the structural basis for GloB loss-of-function. We generated homology models of both 10

SsGloB and SpGloB using SWISS-MODEL (47). The resulting staphylococcal model is based 11 on the sequence-similar metallo-β-lactamase superfamily member from Thermus thermophilus 12 (PDB 2ZWR) (48). This hit had a global model quality estimate (GNQE) of 0.71 and 0.70 for S. 13 schleiferi and S. pseudintermedius GloB homologs, respectively, suggesting the built models 14 are reliable and accurate. In both protein models, we find that POM-ERJ-resistance mutations 15 are primarily located towards the interior of the protein, occupying the same cavity as the well 16 conserved glyoxalase II metal binding motif (THxHxDH) (49). This modeling thus indicates that 17 these prodrug-resistance alleles impair the GloB active site (Fig. 5) . 18

GloB is a functioning type II glyoxalase, not a β-lactamase. 20

GloB is predicted to be a type II glyoxalase and a member of the large metallo-β-lactamase 21 protein superfamily (INTERPRO IPR001279). Members of this superfamily hydrolyze thioester, 22 sulfuric ester, and phosphodiester bonds, such as the ester linkage present in POM-ERJ (49-23 52). Type II glyoxalases catalyze the second step in the glyoxalase pathway that is responsible 24 for the conversion of methylglyoxal (a toxic byproduct endogenously produced during 3 To determine whether SsGloB encodes a functional type II glyoxalase, we evaluated whether 4

SsGloB hydrolyzes S-lactoylglutathione using an assay in which hydrolysis of S-5 lactoylglutathione is linked to a change in absorbance (Fig. 6A) . We purified recombinant WT 6

SsGloB protein and its catalytically inactive variant, SsGloB H54N , in which the histidine of the 7 canonical metal binding motif (THxHxDH) has been altered to an asparagine ( Fig. S1 ) (49, 51, 8 52). We find that SsGloB, but not SsGloB H54N , hydrolyzes S-lactoylglutathione with a specific 9 activity of 0.493 μmol*min -1 mg -1 (Fig. S2, Fig. 6B ,C). This activity is similar to other 10 characterized microbial type II glyoxalases (Saccharomyces cerevisiae, 1.34 μmol*min -1 mg -1 ; 11

Trypanosoma brucei, ~8 μmol*min -1 mg -1 ), but is much lower than that of previously 12 characterized type II glyoxalases from plants and mammals (20-2000 μmol*min -1 mg -1 ) (53-61). 13 glyoxalase in manganese, cobalt, calcium, and zinc, with a modest preference noted towards 2 magnesium (Fig. S3) . 3 4 As some members of the metallo-β-lactamase protein superfamily mediate hydrolysis of β-5 lactam antibiotics, we considered whether GloB also had β-lactamase activity. Because gloB 6 mutant strains are not cross-resistant to the β-lactam-containing antibiotics (except for the 7 prodrugged cephalosporin, cefditoren pivoxil) (Fig. 4, Table S2 ), we predicted that GloB was not 8 a functional metallo-β-lactamase. As expected, we find that SsGloB does not hydrolyze the β-9 lactamase ring of nitrocefin (a canonical β-lactamase substrate), in contrast to the active B. to other ester prodrugs. Because GloB does not mediate resistance to ERJ or other 3 phosphonates, our data suggested that GloB might directly catalyze the conversion of POM-4 ERJ to ERJ. To determine whether GloB de-esterifies POM-ERJ, we developed a liquid 5 chromatography-mass spectrometry (LC-MS)-based assay to quantify POM-ERJ 6 concentrations. Incubation of purified recombinant SsGloB protein, but not its inactive variant 7 (SsGloB H54N ), with POM-ERJ results in rapid loss of POM-ERJ, consistent with SsGloB-8 mediated cleavage (Fig. 7A ). To determine whether prodrug activation activity is conserved 9 among staphylococcal GloB homologs, we also purified recombinant GloB from the human 10 pathogen S. aureus (Fig. S1 ). We find that SaGloB also directly hydrolyzes POM-ERJ (Fig. 7A) . 11

To determine whether GloB mediates intracellular prodrug activation, we evaluated the 13 intracellular concentrations of POM-ERJ in drug-treated WT and gloB mutant staphylococci. We 14 prepared staphylococcal cultures treated with POM-ERJ and quenched the reaction at several 15 timepoints to monitor the course of intracellular prodrug depletion. As expected, we find that 16 POM-ERJ is rapidly depleted in WT S. schleiferi, consistent with enzymatic activation. In 17 contrast, POM-ERJ concentrations do not decrease over time in gloB mutant strains, in which 18 the sole genetic change in each strain compared to WT is in the gloB locus (Fig. 7B ). This 19

suggests that the initial step in carboxy ester prodrug activation in staphylococci lacks functional 20 redundancy and is exclusively dependent on GloB. Because staphylococcal GloB mediates de-esterification of ester prodrugs, we sought to 25 evaluate the feasibility of using these enzymes to design prodrugs specifically targeted for activation in staphylococci. We constructed a phylogenetic tree of GloB homologs across 1 diverse microbial genomes, as well as in humans and mice (Fig. S4A) , specifically including 2 sequences of previously characterized GloB homologs. We find that considerable sequence 3 variation exists within GloB homologs, with no clear clustering by phylogeny except for those 4 GloB homologs originating in plants and mammals. This contrasts with a phylogenetic tree 5 generated using the DNA-directed RNA polymerase subunit beta (rpoB), which generally follows 6 the traditional tree of life (Fig. S4B) . 7 8 While sequence differences between staphylococcal GloB and human GloB suggest that there 9 may be substrate utilization differences between humans and staphylococci, ultimately 10 differences within the active site are likely to drive substrate specificity. Using pymol, we aligned 11 our homology model of SsGloB with the glutathione bound GloB from humans (PDB ID: 1qh5) 12 (62, 63). The two structures align well with a root-mean-square deviation (RMSD) of 1.528Å, 13

and are well conserved in the overall structure as well as the characteristic Zn binding motif, 14 THxHxDH (Fig. S5A,B) . Notably, however, HsGloB has a significant C-terminal extension which 15 is not present in SsGloB. This C-terminal extension forms an α-helix which borders the active 16 site and contains two residues, K252 and R249, which appear to be involved in coordinating the 17 co-crystallized glutathione substrate (Fig. S5C ). The absence of this C-terminal extension in our 18

SsGloB homology model suggests that HsGloB and SsGloB have distinct active site chemistry 19 that may be exploited to drive prodrug activation selectively by SsGloB vs HsGloB. effects. To achieve cell-targeted prodrug activation, knowledge of the activation mechanisms in 20 sera, as well the target cell, are essential. While prodrug targeting has been achieved for liver 21 therapies, this strategy has yet to be employed for bacterial antibiotics that employ ester 22 prodrug moieties (33). 23

In this work, we have identified a new mechanism for the de-esterification and activation of 25 lipophilic ester prodrugs though a conserved staphylococcal esterase in the metallo-β-lactamase superfamily. Loss-of-function of GloB confers resistance to lipophilic carboxy ester 1 prodrugs in two zoonotic pathogens, S. schleiferi and S. pseudintermedius (Fig. 1D, Table S3 ). 2

Purified recombinant GloB from S. schleiferi and the related human pathogen S. aureus directly 3 catalyzes pro-drug de-esterification in vitro (Fig. 7A) . Because gloB mutant staphylococci are 4 cross-resistant to other POM-containing prodrugs that differ in "warhead" and intracellular 5 targets (Fig. 4) , we propose that substrate-specificity of GloB appears driven by recognition of 6 the lipophilic promoiety, rather than the target inhibitory portion of each compound. 7 8 Bacterial prodrug ester activation through GloB hijacks a conserved bacterial protective 9 mechanism in bacteria, as hydroxyacylglutathione hydrolase represents the second enzyme of 10 the two-step glyoxalase pathway. During normal metabolism, the glycolytic intermediates 11 glyceraldehyde-3-phosphate (GAP) and dihydroxyacetone phosphate (DHAP) undergo 12 nonenzymatic decomposition to methylglyoxal, a toxic metabolite. GloB is required for 13 methylglyoxal detoxification, as methylglyoxal is highly reactive and irreversibly glycates 14 proteins and nucleic acids (73-75). In S. aureus, methylglyoxal accumulation potentiates 15 antibiotic susceptibility (76). In addition, methylglyoxal is itself directly antibacterial and 16 postulated to be the primary antistaphylococcal ingredient in Manduka honey (used on chronic 17 wounds) (76-79). Our studies suggest that strains of S. schleiferi and S. pseudintermedius 18 lacking GloB have preserved axenic growth in rich media, which raises concern for the ease of 19 resistance development when GloB-targeted prodrugs are used as anti-infectives. However, the 20 known toxicity of methylglyoxal in a host infection setting suggests that reduced methylglyoxal 21 detoxification as the result of GloB loss-of-function would not be well tolerated in vivo. 22

Identification of GloB as a prodrug activating enzyme in staphylococci is a major step forward 24 for highly selective microbial targeting of compounds. Though GloB homologs are widespread in 25 microbes and are present in humans, significant sequence variation exists in GloB sequences, which results in a variety of GloB substrate preferences (Fig. S4) . For example, human GloB 1 has an additional α-helix along the active site that introduces two additional residues, K252 and 2 R249 to the substrate binding pocket (Fig. S5) (63) . These residues, and this α-helix, are 3 notably absent in microbial GloBs, suggesting that there are underlying substrate differences 4 between human and microbial GloB enzymes. Furthermore, there is substantial sequence 5 variation in GloB orthologs across all microbes, suggesting that GloB substrate specificities may 6 discern between individual clades of bacteria. We expect that development of prodrugs specific 7 to GloB would result in a narrow-spectrum antibiotic which would reduce off-target effects on the 8 microbiome and decrease the broad pressure to evolve resistance. 9 carried out over a range of GloB concentrations to ensure that the reaction rates are linear over 20 the period of the assay. Assays were performed using Zn +2 , Mn +2 , Mg +2 , Co +2 , and Ca +2 . 21

containing 25 mM Tris HCl (pH 7.5), 250 mM NaCl, 10% glycerol, 1 mM MnCl2, and 1 mM POM-24 ERJ were pre-warmed to 37°C before addition of WT GloB, catalytically inactive GloB (H54N) , 25 boiled GloB, or an equal amount of protein storage buffer to a final concentration of 1 μM.

Reactions were placed at 37°C and sampled at 0, 15, 30, 60, 90, and 120 min. A 50 µL sample 1 was withdrawn from each reaction at the times indicated, and the sample reaction was 2 quenched by the addition of 200 µL acetonitrile containing 100 ng/μL enalapril as an internal 3 standard. The samples were immediately frozen on dry ice and stored at -80°C until analysis. 4

The quenched reaction mixtures were centrifuged at 3200 rpm for 5 min, and 2 μL of the 6 supernatant was diluted to 500 µL with water containing 100 ng/mL enalapril as an internal 7 standard. Samples were analyzed by LC-MS/MS using an Applied Biosystems-Sciex API 4000. 8

Analyte/internal standard peak area ratios were used to determine concentration and evaluate 9 stability. Standards were evaluated over the range of 1 ng/mL to 1000 ng/mL. The MRM 10 transitions for enalapril and POM-ERJ were m/z: 376.9 > 91.2 and 424.0 > 364.0, respectively. 11 A Phenomenex Luna Omega polar C18 column (2.1 × 50 mm, 5 μm) was used for 12 chromatographic separation. Mobile phases were 0.1% formic acid in water and acetonitrile with 13 a flow rate of 0.5 mL/min. The starting phase was 1% acetonitrile increased to 100% acetonitrile 14 over 0.9 min. Peak areas were integrated using Analyst Software (AB Sciex, Foster City, CA). 

Measurement of transmembrane 6 potentials in phospholipid vesicles

The continuing challenge of 8 metallo-β-lactamase inhibition: mechanism matters

Lipophilic prodrugs of FR900098 are antimicrobial against 10

Francisella novicida in vivo and in vitro and show GlpT independent efficacy

Activity of nitrogen-containing and non-nitrogen-containing 13 bisphosphonates on tumor cell lines

Synthesis of a phosphoantigen prodrug that potently activates 15

Prodrugs of phosphonates and phosphates: crossing the 17 membrane barrier

Prodrugs of phosphates and phosphonates

Phosphonate prodrugs: an overview and recent advances

Synthesis and oral activity of ME 1207, a new orally active 23 cephalosporin

Discovery and synthesis of a phosphoramidate prodrug of a 25 pyrrolo[2,1-f][triazin-4-amino] adenine C-nucleoside (GS-5734) for the treatment of

Purification and cloning of glyoxalase II from 14 rat liver

Purification and characterization of S-2-hydroxyacylglutathione hydrolase 16 (glyoxalase II) from human liver

Schrödinger The PyMOL molecular graphics system

glyoxalase II and its complex with a glutathione thiolester substrate analogue

A genetic resource for rapid and comprehensive phenotype

Diarylquinolines target subunit c of mycobacterial ATP synthase

Selectivity of TMC207 towards Mycobacterial ATP synthase 14 compared with that towards the Eukaryotic homologue

A diarylquinoline drug active on the ATP synthase of Mycobacterium 17 tuberculosis

Molecular 19 modification of glycolytic enzymes and heat shock proteins

Formation of strand breaks and interstrand 21 cross-links in DNA by methylglyoxal

Genotoxicity of methylglyoxal: cytogenetic damage in human 23 lymphocytes in vitro and in intesting cells of mice

Manuka honey and methylglyoxal increase the sensitivity of

The variant call format and VCFtools

A program for annotating and predicting the effects of single 15 nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster 16 strain w1118; iso-2; iso-3

A facile method for high-throughput co-expression of protein 18 pairs

Novel method for detection of 20 β-lactamases by using a chromogenic cephalosporin substrate

Human gut microbiome viewed across age and geography

The EMBL-EBI search and sequence analysis tools APIs in

Interactive Tree Of Life (iTOL) v4: recent updates and new 1 developments

The authors are grateful to Joe Jez for ongoing support and helpful discussions. Financial 25 support provided by NIH AI123433 to CSD and the GWU Department of Chemistry. A.O.J. is