Inhibition of Pseudomonas aeruginosa and Mycobacterium tuberculosis disulfide bond forming enzymes

Abstract

In bacteria, disulfide bonds confer stability on many proteins exported to the cell envelope or beyond, including bacterial virulence factors. Thus, proteins involved in disulfide bond formation represent good targets for the development of inhibitors that can act as antibiotics or anti-virulence agents, resulting in the simultaneous inactivation of several types of virulence factors. Here, we present evidence that the disulfide bond forming enzymes, DsbB and VKOR, are required for P. aeruginosa pathogenicity and M. tuberculosis survival, respectively. We also report the results of a HTS of 216,767 compounds tested against P. aeruginosa DsbB1 and M. tuberculosis VKOR using Escherichia coli cells. Since both P. aeruginosa DsbB1 and M. tuberculosis VKOR complement an E. coli dsbB knockout, we screened simultaneously for inhibitors of each complemented E. coli strain expressing a disulfide-bond sensitive β-galactosidase reported previously. The properties of several inhibitors obtained from these screens suggest they are a starting point for chemical modifications with potential for future antibacterial development.

Graphical Abstract

Introduction

Disulfide bonds confer activity to many proteins by stabilizing them in their active conformations. The formation of disulfide bonds in proteins is an oxidative process that generates a covalent bond linking the sulfur atoms of two cysteine residues. In bacteria, structural disulfide bonds are found in proteins of the cell envelope and they play a role in folding and stability of proteins involved in important cellular processes such as cell division, motility, virulence of pathogenic bacteria, transport of virulence proteins into the host cells, response to environmental threats and assembly of the outer membrane of gram-negative bacteria (Landeta et al., 2018).

The formation of disulfide bonds in E. coli utilizes the thioredoxin-family protein DsbA, which catalyzes through its CXXC active site the formation of sulfur-sulfur bonds between pairs of cysteines in substrate proteins (Bardwell et al., 1991). In the process, DsbA becomes reduced thus necessitating a second enzyme DsbB to regenerate DsbA’s activity by transferring DsbA’s electrons to quinone (Bardwell et al., 1993). While the DsbAB system is widespread in gram-negative bacteria, Actinobacteria and Cyanobacteria use the protein VKOR (vitamin K epoxide reductase) instead of DsbB to oxidize DsbA (Dutton et al., 2008). Bacterial VKOR is a homologue of the vertebrate protein VKOR, an endoplasmic reticular enzyme which maintains the reduced form of vitamin K, a cofactor in the post-translational modification of proteins involved in blood coagulation (Kurosu and Begari, 2010). Bacterial VKOR is the target of the anticoagulant drug warfarin (Dutton et al., 2010). DsbB and VKOR share no protein sequence homology but they exhibit similar structural features and contain a quinone cofactor to generate a disulfide bond de novo (Inaba et al., 2009; Li et al., 2010). In fact, the interactions between the redox-active cysteines of DsbA and VKOR proceed by the same steps seen between DsbA and DsbB (Wang et al., 2011). Furthermore, VKORs from other bacteria and one eukaryote can complement an E. coli with a dsbB deletion (Dutton et al., 2008; Hatahet et al., 2015).

Given that disulfide bond formation contributes to the pathogenicity of bacteria, proteins involved in disulfide bond formation represent good targets for the development of inhibitors that can act as antibiotics or anti-virulence agents, since inhibition of this process would result in the simultaneous inactivation of several types of virulence factors. We have previously reported that a highly sensitive assay for detecting defects in disulfide bond formation allows the identification of compounds potentially useful in the development of antibiotics (Landeta et al., 2015). This assay detects inhibition of disulfide bond forming enzymes in growing E. coli cells using a version of E. coli β-Galactosidase (β-Gal^dbs) that is only active when the functionality of the disulfide bond formation pathway is inactivated. The screening of compounds is performed in parallel for inhibition of the non-homologous enzymes, DsbB and VKOR from another bacteria (Figure 1). This approach allows us to search for inhibitors for one or the other protein. The parallel screening thus provides reciprocal controls that eliminate inhibitors that are influencing β-galactosidase activity by acting directly on DsbA or affecting membrane protein assembly, since those molecules would appear as hits in screens for both proteins. Specific inhibitors would only register as hits against one strain or the other (Landeta et al., 2015). While it is formally possible that a particular compound could inhibit both DsbB and VKOR, the sequence dissimilarity between the two would make it unlikely.

Figure 1.

Outline of screening and inhibitor selection. The first step “Targets” represents the disulfide bond forming systems of two human pathogens, P. aeruginosa and M. tuberculosis. The second step “E. coli system” indicates the expression of the targets, PaDsbB1 (CL523 strain) and MtVKOR (CL382 strain), in E. coli reporter strain expressing β-Gal^dbs. The third step “HTS” indicates the high throughput screening of 216,767 compounds and shows an example of strong, medium and weak hits on agar minimal media plates. In the fourth step “Re-test”, the compounds were cherry picked from the chemical library and tested against the strains used in screening plus two additional E. coli counter-screening strains expressing low levels of EcDsbB and MtVKOR (DHB7935 and DHB7657 strains, respectively) and one additional strain expressing RnVKOR (FSH250 strain) to discard molecules with unwanted anticoagulant side effect. The next step “Characterization in E. coli” involved resupplying the molecules selected in the previous stage, making serial dilutions on agar plates and testing them against a collection of E. coli strains expressing VKOR proteins and various DsbB proteins from gram-negative pathogens in order to determine their MBCs (minimal concentration of give a pale blue color indicating inhibition of DsbB or VKOR in E. coli cells expressing β-Gal^dbs). Finally, “In vivo and in vitro characterization” involved determination of Ki and IC50 for all compounds in a ubiquinone reduction assay with purified enzymes. For Pseudomonas inhibitors, the in vivo redox states of PaDsbA1 (the substrate of the two PaDsbB proteins) was also determined in the presence of compound as well as the activity of elastase (a substrate of PaDsbA1) using a fluorescent assay. On the other hand, Mycobacterium inhibitors were tested in a growth assay using wild type and a vkor knockdown (KD) strain. See text for further details.

Using this methodology in a previous screen of 50,000 molecules, we found a family of compounds structurally related to pyridazinones that inhibit DsbB enzymes from several gram-negative pathogenic bacteria (Landeta et al., 2015). Structure-activity relationship (SAR) analyses yielded more effective DsbB pyridazinone-related inhibitors, such as compounds 12 and 36 (Landeta et al., 2015; Landeta et al., 2017) (Figure 3). This class of drugs inhibits DsbB by binding covalently to the second cysteine of DsbB (Cys44) thus preventing the recycling of DsbB by quinones (Landeta et al., 2015; Landeta et al., 2017). These molecules do not inhibit cysteine-containing enzymes such as DsbA, human protein disulfide isomerase (PDI) or VKOR, but they do react with reduced glutathione, which can potentially lead to unfavorable toxicological outcomes (Potashman and Duggan, 2009).

Figure 3.

Structures of inhibitors found in HTS. Control compounds include 12 and 36, which are inhibitors of DsbB enzymes, while Bromindione is a known anticoagulant that targets MtVKOR and RnVKOR enzymes. The structures of six PaDsbB1 inhibitors and 22 MtVKOR inhibitors found to inhibit the respective enzyme in the HTS using E. coli reporter strain (CL523 and CL382, respectively). The MBCs are reported below for each of the molecules except the last three molecules that were not resupplied from commercial libraries. The MBC is the minimal concentration required to produce a pale blue color in E. coli cells expressing β-Gal^dbs indicating inhibition of DsbB or VKOR. The results represent the average of at least two independent experiments.

In this work we focused on two important human pathogens, namely P. aeruginosa and M. tuberculosis. P. aeruginosa is one of six pathogens for which new antibacterial agents are most desperately needed (Rice, 2010). It is a gram-negative organism often associated with extremely difficult-to-treat infections that resist antibiotic treatments and can cause a variety of infections in a wide range of tissue types where mortality rates of more than 60% have been reported (Page and Heim, 2009). Infections include acute pneumonia, ulcerative keratitis, bacteremia, urinary tract, intra-abdominal, chronic airway and wound infections (Lyczak et al., 2000; Page and Heim, 2009). This organism frequently emerges as a threat to neutropenic, immunosuppressed patients undergoing treatment for cancer wherein the spread of antibiotic-resistant organisms from gastrointestinal sites into the blood stream or in the setting of cystic fibrosis has also been observed (Maschmeyer and Braveny, 2000; Fujitani et al., 2011; Folkesson et al., 2012). P. aeruginosa harbors two dsbA and two dsbB homologs. Disulfide bond formation is mainly driven by DsbA1 which is reoxidized by both DsbB1 and DsbB2 (Ha et al., 2003; Arts et al., 2013), while DsbA2 is unable to complement dsbA1 mutant and its function remains unknown (Arts et al., 2013) (See Results below). Some Pseudomonas virulence factors known to contain disulfide bonds are the assembly of pilin protein (PilA) involved in twitching motility (Ha et al., 2003; Harvey et al., 2009), the elastolytic metalloprotease, elastase (LasB) and exotoxin A (Madshus and Collier, 1989; Braun et al., 2001; Urban et al., 2001). Furthermore, for certain virulence factors the disulfide bond formation pathway is required for their proper secretion, including the two type-III effectors ExoT and ExoU (Ha et al., 2003) as well as other exoproteins such as the chitin-binding protein (CbpD), an immunomodulating metalloprotease IMPa, and proteases PaAP and PrpL (Arts et al., 2013). In addition, the P.aeruginosa dsbA1 mutant was detected in a screen for virulence factors in a D. melanogaster infection model, and was confirmed to be avirulent in a peritoneal mouse infection model (Kim et al., 2008). We hypothesized that the failure to find DsbB mutants in such a screen was due to the redundancy of DsbB. On the other hand, M. tuberculosis is a gram-positive organism that causes tuberculosis, one of the leading causes of infectious disease deaths worldwide (World Health Organization, 2014). The pathway that introduces disulfide bonds into proteins in M. tuberculosis utilizes a VKOR instead of DsbB to oxidize mycobacterial DsbA (Dutton et al., 2008; Ke et al., 2018). The mechanism of electron transfer between DsbA, VKOR and quinones is analogous to that observed between DsbA, DsbB and quinones (Dutton et al., 2008; Wang et al., 2011; Ke et al., 2018). Additionally, VKOR has been shown to be essential for growth of M. tuberculosis and M. smegmatis (Sassetti and Rubin, 2003; Dutton et al., 2010), likely because one or more essential proteins require disulfide bonds to be active. In contrast, the Dsb pathway in E. coli is only essential under anaerobic growth (B.M. Meehan et al., 2017).

Thus, here we sought to determine the impact of disulfide bond formation in the virulence of P. aeruginosa dsb mutants and the redundancy in the Dsb system using lung and corneal infections in murine model, two models that best mimic P. aeruginosa human infections (Preston et al., 1995; Mizgerd and Skerrett, 2008; Zhao et al., 2016). We also constructed a conditionally lethal strain of M. tuberculosis using a tunable CRISPR interference platform (Rock et al., 2017) to knock down vkor and used it as a tool to determine VKOR inhibition in M. tuberculosis. We then performed a High Throughput Screen (HTS) of 216,767 compounds tested against the P. aeruginosa DsbB1 (PaDsbB1) and M. tuberculosis VKOR (MtVKOR) enzymes using E. coli cells expressing β-Gal^dbs in order to find inhibitors for antibacterial development against these two human pathogens. The properties of several inhibitors obtained from these screens suggest they are candidates with potential for future antibacterial development and a starting point for chemical modifications in order to improve their potency.

Results

P. aeruginosa strains defective in disulfide bond formation display attenuated virulence in lung and corneal infections.

In order to determine the effect of disulfide bond formation in P. aeruginosa virulence we constructed the clean deletions of dsbA1, dsbA2, dsbB1 and dsbB2 genes as well as the double deletions of dsbA1 and dsbA2 (ΔdsbA1A2) and of dsbB1 and dsbB2 (ΔdsbB1B2) in P. aeruginosa UCBP-PA14 strain. Before assessing virulence in vivo, we tested two virulence-related phenotypes that dsb mutants display, namely the lack of twitching motility and elastase production (Urban et al., 2001; Ha et al., 2003; Arts et al., 2013). Similar to previous results (Arts et al., 2013), the ΔdsbA1 as well as the ΔdsbB1B2 mutant are unable to produce elastase (Figure 2A) or twitch (Supplementary Figure 1). We also determined complementation of the mutant strains by expressing the dsb genes from a plasmid. To this end, we cloned each of the four dsb genes in a plasmid (pJN105) under the control of their own promoters except for dsbA2, for which we used dsbA1’s promoter since it is unclear what conditions allow for expression of this gene (Arts et al., 2013). We transformed these plasmids into the double deletion mutants and tested complementation of the resulting strains by looking at twitching motility and elastase production. Both dsbB1 and dsbB2 complemented motility and elastase of the ΔdsbB1B2 mutant, consistent with previous reports (Arts et al., 2013). However, only dsbA1 but not dsbA2 complemented the ΔdsbA1A2 mutant (Data not shown).

Figure 2.

Mutations in genes encoding disulfide bond forming enzymes in P. aeruginosa cause severe defects in virulence. A) ΔdsbA1 and ΔdsbB1B2 mutants have little or no elastase activity (<0.1 μg mL⁻¹ A₆₀₀⁻¹). Activity was measured in the supernatant of cell cultures grown in 96-well plates for 15 h at 37ºC in M63 minimal medium supplemented with 0.2% glucose and 0.2% casamino acids. Values represent the average ±SD of at least three independent experiments. B) Comparison of growth in corneal tissue between P. aeruginosa PA14 (CL418), ΔdsbA1A2 (CL541) and ΔdsbB1B2 (CL356) strains using C57BL/6 female mice. Groups of five mice were scratch-injured in the eyes and challenged with 5 μL of P. aeruginosa strains (∼10⁷ CFU/eye). The severity of corneal disease was scored by a masked observer after 48 h on a scale of 0 to 4: 0, eye macroscopically identical with the uninfected contralateral control eye; 1, faint opacity partially covering the pupil; 2, dense opacity covering the pupil; 3, dense opacity covering the entire anterior segment; 4, perforation of the cornea, phthisis bulbi (shrinkage of the globe after inflammatory disease), or both. Mice were euthanatized and the corneas excised to determine extracellular and internalized P. aeruginosa CFU. Significance between mutants vs wild type strain was determined by Kruskal Wallis test for pathology score and one-way ANOVA followed by Dunnett’s test for CFU counts. Data represent individual values with the geometric mean and 95% confidence intervals. C) Comparison of growth in corneal tissue between P. aeruginosa ΔdsbA1A2 (CL541) and ΔdsbB1B2 (CL356) mutants and these strains complemented with dsbA1 or dsbB1 genes expressed from a plasmid (pJN105). Groups of mice were treated as described in B. Significance between mutant vs complemented strain was determined by Kruskal Wallis test for pathology score and one-way ANOVA followed by Dunnett’s test for CFU counts. Data represent individual values with the geometric mean and 95% confidence intervals. D) Comparison of growth in lung tissue between P. aeruginosa PA14 (CL418), ΔdsbA1 (CL384) and ΔdsbB1B2 (CL356) strains using neutropenic ICR mice. Groups of 20 animals were rendered immune suppressed by two intraperitoneal injections of cyclophosphamide and were challenged intranasally (0.02 mL/lung) with bacteria (∼10⁴ CFU/mouse). Groups of five animals were sacrificed at 2 h after inoculation (Supplementary Figure 2E), the rest (15 animals) were sacrificed at 26 h after inoculation and the lung tissues were harvested to enumerate bacteria. Significant difference (p<0.05) compared to the parental strain was determined by one-way ANOVA followed by Dunnett’s test. Data represent individual values with the geometric mean and 95% confidence intervals. *, p < 0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001.

We then used two murine models to test the virulence of dsb mutants, keratitis and pneumonia models. First, using a keratitis model we tested the virulence of the double dsb mutants and two of these complemented counterparts in C57BL/6 mice. We tested the two double knockouts of ΔdsbA1A2 and ΔdsbB1B2 in corneal infection by scratching the eyes with 10⁷ CFU as reported previously (Preston et al., 1995). We then scored the eye pathology (Figure 2B, left) and determined the extracellular (Figure 2B, center) and intracellular bacterial counts (Figure 2B, right) of the corneas. Our results show that infection with ΔdsbB1B2 mutant displayed a 60% reduction in the pathology score (p<0.05), while ΔdsbA1A2 mutant showed 87% reduction (p<0.001). Additionally, the ΔdsbB1B2 mutant has a 99% and 98% reduction in extracellular and intracellular bacterial counts, respectively, when compared to wild type (p<0.0001). While the ΔdsbA1A2 mutant yielded no bacteria in both cases (p<0.0001).

We then determined the effect on virulence using two of the complemented strains in the keratitis model, the ΔdsbA1A2 strain complemented with pJN105-dsbA1 and the ΔdsbB1B2 strain complemented with pJN105-dsbB1 (Figure 2C). The plasmids expressing dsbA1 and dsbB1 genes from a plasmid restored the virulence to the ΔdsbA1A2 and ΔdsbB1B2 mutants, respectively, as observed by the higher eye pathology scores (Figure 2C, left) or extracellular and intracellular bacterial counts (Figure 2C, center and right). Thus, ΔdsbA1A2 and ΔdsbB1B2 mutants showed reduced virulence in the murine keratitis model and the virulence is restored when the mutants are complemented with the dsbA1 or dsbB1 genes, respectively.

In addition to the keratitis model, we also determined the virulence of dsb mutants in a pneumonia model. We first conducted an experiment to determine the optimal bacterial density to infect the lungs of immunosuppressed ICR mice. We tested ten-fold dilutions ranging from 10² to 10⁷ CFU/mouse of the P. aeruginosa wild type strain and determined the bacterial counts in the lungs at 2 and 26h after infection (Supplementary Figure 2A-B and Supplementary Tables 2–1 to 2–6). We observed that 10⁴ CFU/mouse resulted in an approximate 3-log increase in bacterial counts in a 24 h infection period and only one of six mice died at 26 h. Higher inoculum densities resulted in over 50% mortality at 26 h and lowering the inoculum yielded insufficient bacterial growth in lung tissues. Therefore, we used 10⁴ CFU/mouse to evaluate the virulence of dsb mutants and increased the population size to 20 animals, from which 5 were sacrificed at 2 h and 15 were sacrificed at 26 h after infection (Figure 2D, Supplementary Figure 2E and Supplementary Tables 4–1 to 4–5). We only used in this model the double ΔdsbB1B2 and the single ΔdsbA1 mutant given that dsbA2 gene did not complement two phenotypes when expressed from a dsbA1 promoter. Our results show that infection with ΔdsbB1B2 double mutant yielded an 87% reduction in bacterial counts in lung tissue compared to wild type (p<0.05, Figure 2D), while ΔdsbA1 yielded 98% reduction in bacterial counts compared to wild type (p<0.05, Figure 2D) at 26 h after infection. Therefore, both ΔdsbA1 and ΔdsbB1B2 mutants displayed attenuated virulence in the murine pneumonia model.

We also attempted to determine the virulence of dsb mutants using immune competent ICR mice, however this was not feasible due to poor growth and excessive variability of the wild type strain in the pneumonia model (Supplementary Figure 2C-D and Supplementary Tables 3–1 to 3–7).

High throughput screening to find inhibitors of PaDsbB1 and MtVKOR in E. coli

Given the importance of DsbB proteins in the virulence of P. aeruginosa (above) and of VKOR in the growth of M. tuberculosis (Sassetti and Rubin, 2003; Dutton et al., 2010), we screened for inhibitors of both proteins simultaneously using an approach described previously (Landeta et al., 2015). That the ΔdsbA1 mutant displayed a slightly more severe virulence phenotype in P. aeruginosa than the DsbB double mutant might suggest that DsbA1 would also be a useful target. However, DsbA belongs to the widespread thioredoxin family and inhibitors of DsbA may also inhibit one of the many other thioredoxin family members that in humans contribute to the reducing environment of the cytoplasm and to protein folding in the endoplasmic reticulum. On the other hand, DsbB has no known homologs in eukaryotes, thus making it more suitable for developing virulence inhibitors. Being that DsbB and VKOR do not share protein sequence homology they can be used to counter screen molecules in order to find more specific inhibitors for each. Since both the P. aeruginosa DsbB homologs and the M. tuberculosis VKOR complement E. coli dsbB null mutants, we can screen simultaneously for inhibitors of each complemented strain in 384-well plates seeking compounds that inhibit one protein but not the other. Inhibition is scored visually through the appearance of blue color in a well due to increased activity of the disulfide-bond sensitive β-galactosidase (β-Gal^dbs) (Figure 1) (Landeta et al., 2015). In these strains, strong inhibition of disulfide bond formation should lead to a substantial increase in β-galactosidase activity (Bardwell et al., 1991; Tian et al., 2000). For the screening of Pseudomonas DsbB, we expressed the DsbB1 homolog in E. coli from an IPTG-inducible promoter in a pDSW206 plasmid (CL523 strain). For the Mycobacterial VKOR screen, we expressed the vkor gene in E. coli from a strong IPTG-inducible promoter in a pTrc99a plasmid (CL382 strain). Our previous screening strain expressed vkor from an IPTG-inducible promoter in the chromosome and this strain did show a very pale blue color, indicating that the MtVKOR did not completely restore levels of disulfide bond formation in the E. coli dsbB deletion (Dutton et al., 2010). The higher copy number of the plasmid led to better complementation of the re-oxidation of EcDsbA (Supplementary Figure 3) and hence to lower β-galactosidase background.

We carried out a HTS of 216,767 compounds, tested in duplicate, against the P. aeruginosa DsbB1 (PaDsbB1) and M. tuberculosis VKOR (MtVKOR) enzymes using E. coli cells (Figure 1). The libraries were part of the collection of the Institute of Chemistry and Cell Biology (ICCB) at Harvard Medical School (Supplementary Table 1A). In addition to the E. coli ΔdsbB strain as negative control, we used two molecules as positive and negative controls for each strain that were placed in column 2 of agar plates. One is the anticoagulant bromindione, which inhibits MtVKOR as well as human VKOR but not PaDsbB1 and the other molecule is compound 12 a dichloropyridazinone, which inhibits PaDsbB1 but not MtVKOR (Figure 3 and Supplementary Figure 4) (Landeta et al., 2015). The hits found in HTS were classified by the intensity of the blue signal on agar plates, as strong, medium, weak or very weak (Figure 1) (Landeta et al., 2015). We found 22 hits of PaDsbB1 and 124 hits of MtVKOR that appeared in both replica plates from the 216,767 compounds tested (Supplementary Table 1B). Testing the effects of compounds on the two strains in parallel allowed us to pick candidate inhibitors that are specific to either PaDsbB1 or MtVKOR. Therefore, we eliminated from further consideration compounds giving positive signals with both strains since these are most likely inhibiting EcDsbA or less likely cytoplasmic membrane protein assembly (Landeta et al., 2015). From all the compounds tested, we only found 3 hits that inhibited both strains (Supplementary Figure 5).

We re-tested all the hits against the two strains used for HTS as well as their respective counter-screening strains, i.e., strains that show a very pale blue background due to lower expression of either DsbB or VKOR (Figure 1). These strains were used in order to detect weak inhibitors that are not specific to each strain, that is for PaDsbB1 inhibitors, a strain expressing low levels of MtVKOR (DHB7657 strain) while for MtVKOR inhibitors, a strain expressing low levels of EcDsbB (DHB7935 strain). Given the importance of the human VKOR homolog in the activation of blood clotting factors, we also screened hits with an additional E. coli mutant strain expressing a rat (Rattus novergicus) VKOR homolog (RnVKOR, FSH250 strain) which conferred upon the strain the ability to oxidize EcDsbA (Hatahet et al., 2015). This was done in order to discard compounds that may have an unwanted anticoagulant side effect. We found 9 potential inhibitors of PaDsbB1 that share similarity to pyridazinone molecules and had weaker activity than the compounds we found and studied previously (Landeta et al., 2015; Landeta et al., 2017). These were not re-tested. For the remaining compounds, the retest yielded a repeat rate of 92% (12/13) for PaDsbB1 inhibitors and 30% (37/124) for MtVKOR (Supplementary Table 1B). The low rate observed for MtVKOR was due to the low repeat rate among weak and very weak hits (Supplementary Table 1B).

In the re-test we also found that 4 PaDsbB1 hits inhibited the counter-screening MtVKOR strain (DHB7657 strain) and 2 hits inhibited the RnVKOR strain (FSH250 strain). Thus, from the 12 PaDsbB1 hits that repeated only 6 hits were candidates to follow up. Among the 37 MtVKOR hits that repeated, 5 of them had the same structure thus leaving only 33 unique hits. From these, 3 inhibited our counter-screening EcDsbB strain (DHB7935 strain) and 8 inhibited the RnVKOR strain, which left us with 22 candidates to follow up. We purchased all 6 candidate inhibitors of PaDsbB1 and 19 of MtVKOR (Figure 3 and Material and Methods) and tested the resupplied molecules again in our agar assay using a collection of E. coli strains expressing different DsbB or VKOR homologs (Landeta et al., 2015) in order to determine the MBC, i.e., the minimal concentration to give a pale blue color in agar plates indicative of inhibition of the disulfide bond formation machinery (Figure 1). We then ranked the inhibitors by potency for each one of the strains (Figure 4), and observed that most of the P. aeruginosa DsbB1 inhibitors also inhibited the E. coli DsbB as well as H. influenza, F. tularensis and A. baumannii DsbB homologs but not the rest of the homologs used in our previous study (Landeta et al., 2015). Compound PA4, which is similar in structure to PA5 and to the anticoagulant warfarin, also inhibited RnVKOR. On the other hand, some VKOR inhibitors found using M. tuberculosis VKOR also inhibited RnVKOR and S. typhymurium DsbI. A few of them inhibited F. tularensis and A. baumannii DsbB (Figure 4).

Figure 4.

In vivo inhibition of DsbB enzymes from gram-negative bacteria expressed in E. coli. E. coli dsbB mutant strains expressing β-Gal^dbs and dsbB genes from Salmonella enterica sv. Typhimurium (St), Klebsiella pneumoniae (Kp), Vibrio cholerae (Vc), Haemophilus influenzae (Hi), Pseudomonas aeruginosa (Pa), Acinetobacter baumannii (Ab), Francisella tularensis (Ft) as well as two DsbB-homologs of P. aeruginosa (Pa) and S. enterica sv. Typhimurium (dsbI) and two vkor genes from Mycobacterium tuberculosis (Mt) and Rattus norvegicus (Rn) were tested against resupplied compounds found in HTS as candidate inhibitors. Inhibition range from strong to weak is relative to each DsbB/VKOR-expressing strain and was obtained by dividing the MBC of each compound between the lowest MBC observed for each particular strain (See Supplementary Table 5 for raw data). Results are the average of three independent experiments. Compounds that did not inhibit at the highest concentration tested are shown as black. Strains used in screening are in the first columns of both tables. In order to have a more sensitive counter screening in each case, we used for PaDsbB1 hits (top) a strain expressing low levels of MtVKOR (DHB7657), whilst for MtVKOR hits (bottom) a strain expressing low levels of EcDsbB was used (DHB7935).

Inhibition of PaDsbB1 and PaDsbB2 in P. aeruginosa

We tested the effect of PaDsbB1 inhibitors obtained in the HTS with P. aeruginosa. In previous work we found that the E. coli disulfide bond formation enzymes are dispensable under aerobic growth but essential for anaerobic growth. This property allowed us to validate EcDsbB inhibitors under anaerobic growth (Landeta et al., 2015; B.M. Meehan et al., 2017). Thus, we determined whether Dsb proteins are also essential anaerobically in P. aeruginosa. The two known essential proteins that are substrates of the DsbAB pathway in E. coli are the cell division protein, FtsN and lipopolysaccharide transport protein, LptD (B. M Meehan et al., 2017; B.M. Meehan et al., 2017). Even though P. aeruginosa lacks FtsN homologs, it has an LptD homolog with six cysteines and four of them are conserved with EcLptD, presumably connected by disulfide bonds. We thus investigated the essentiality of the disulfide bond formation in P. aeruginosa growing ΔdsbA1A2 and ΔdsbB1B2 mutants under anaerobic conditions (Material and Methods). Unlike E. coli, both mutants were able to grow anaerobically after several passages in M63 minimal media supplemented with 0.2% glucose and 0.2% casamino acids, which precluded our ability to test PaDsbB1 inhibitors using this assay.

We then assayed the activity of the virulence factor elastase since ΔdsbA1 and ΔdsbB1B2 mutants secrete an inactive elastase (Braun et al., 2001; Urban et al., 2001; Arts et al., 2013). We measured elastase (LasB) activity in the supernatant of cells grown in minimal media using a protease substrate that gives a low fluorescence signal. After cleavage of the peptide by LasB the fluorescent donor group cannot transfer the energy to the quenching acceptor group resulting in a high fluorescence signal, which is directly related to the enzymatic activity (Rust et al., 1994; Cathcart et al., 2009). Given that we observed low or no elastase activity in the supernatant of both ΔdsbA1 and ΔdsbB1B2 mutants (Figure 2A), we used this assay to determine DsbB inhibition in P. aeruginosa by scaling down the assay to 96 well plates and testing the PaDsbB1 inhibitors against three P. aeruginosa strains: wild type, ΔdsbB1 and ΔdsbB2 mutants. The strains were grown for 15 h in M63 minimal media in the presence of inhibitors and the supernatants of those cultures were used to determine the amount of active elastase being produced. The IC₅₀s were then calculated comparing the activity observed when the cells were grown in the absence of inhibitor. In addition to the six hits found in HTS, we also tested two pyridazinone-related molecules, compound 12 and 36, found in our previous work since these molecules showed potency in inhibiting several DsbB homologs from gram-negatives including PaDsbBs (Landeta et al., 2015; Landeta et al., 2017). We found that all compounds except PA2 and PA5 decreased elastase activity of the ΔdsbB1 and ΔdsbB2 mutants (Table 1, Columns 3 and 4). Overall these compounds showed inhibitory activity at lower inhibitory concentrations for the ΔdsbB1 mutant than the ΔdsbB2 mutant. The fact that all the compounds except PA2 and PA5 showed activity in the ΔdsbB1 mutant indicates that the compounds target DsbB2 of P. aeruginosa, which is the only dsbB present in the mutant strain. The same was observed for the ΔdsbB2 mutant but higher concentrations of the compounds were required to see inhibitory activity, especially for compounds PA1 and PA4. Similarly, this indicates that the compounds also inhibited DsbB1 of P. aeruginosa, which is the only DsbB enzyme present in the ΔdsbB2 mutant and the enzyme used to find these inhibitors by HTS. The inhibition of elastase activity in wild type cells, required even higher concentrations of compounds than those required for either mutant strain, presumably due to the presence of two DsbB proteins (Table 1, Column 2).

Table 1.

Analysis of PaDsbB1 inhibitors

ID	Elastase in vivo assay IC50 (μM) *			In vitro assay**
	WT	PaDsbB1 (ΔdsbB2)	PaDsbB2 (ΔdsbB1)	PaDsbB1		PaDsbB2		EcDsbB		PDI *** IC50 (μM)
				IC50 (μM)	Ki (μM)	IC50 (μM)	Ki (μM)	IC50 (μM)	Ki (μM)
12	228.8 (170.6–306.8)	8.6 (5.48–13.63)	0.01 (0.006–0.02)	4.1±0.6	3±0.2	6±1	4±0.2	0.014 ± 0.0001	0.005 ± 0.001	N
36	89 (60.4–131.3)	2.2 (1.2–3.89)	0.025 (0.019–0.033)	4±0.37	4±0.4	3±0.5	2.7±0.1	0.01 ± 0.001	0.002 ± 0.0005	N
PA1	279.8 (96.8–808.3)	120.7 (83.7–173.9)	29.5 (23.6–36.9)	530±60	ND	600±50	ND	150±50	100±30	N
PA2	N	N	N	83±17	ND	83±13	ND	37.5±5	30±7	98.34 (65.45–147.8)
PA3	646.4 (436.7.7–956.9)	9 (5.5–14.83)	1.26 (0.64–2.47)	6±0.6	2.8±0.2	3±0.11	7.5±0.8	17±2	13±1.5	98.02 (45.03–213.4)
PA4	144.8 (103.1–203.5)	101.2 (86.04–119.1)	2.1 (1.6–2.8)	15±1	17±1.2	23±5	16±1.5	17±1	10±1.2	82.03 (35.99–186.9)
PA5	N	N	N	9±1	4.25±0.5	18±3	9±1.5	6±0.5	6±1	N
PA6	139.8 (92.1–212.2)	16.5 (12.5–21.8)	5 (4.2–6.03)	26±3.5	12±1.4	25±4	18±2	3±0.1	5±1	59.6 (31.55–112.6)

^*In vivo inhibition was determined as elastase activity in the supernatant of cell cultures after 15 h incubation at 37ºC in M63 minimal medium supplemented with 0.2% glucose, 0.2% casamino acids and serial dilutions of drug dissolved in DMSO.

^**In vitro inhibition was measured by following the reduction of ubiquinone with purified His-tagged enzymes. The assay used 10 nM of PaDsbB1/PaDsbB2 or EcDsbB with 20 μM reduced-PaDsbA1 or EcDsbA and 25 μM Ubiquinone-1. For K_i determination 1–100 μM of Ubiquinone-1 was used.

^***PDI activity was measured using the insulin reduction assay.

ND: Not determined. N: No inhibition observed up to 100 μM. Data represent the average of at least two independent experiments with 95% confidence intervals in parenthesis or ±SD.

In addition, we assessed the in vivo redox state of DsbA1 and LasB in P. aeruginosa grown in the presence of compound 12 in order to determine whether the inhibition of LasB activity was due to accumulation of reduced DsbA (the direct substrate of DsbB1 and DsbB2) and reduced LasB (the direct substrate of DsbA1). We grew cells in M63 minimal media, precipitated the proteins and alkylated them with MalPEG2k and NEM, respectively to distinguish the reduced and oxidized forms of each protein. Both DsbB1 and DsbB2 are able to re-oxidize DsbA1 since ΔdsbB1 and ΔdsbB2 mutants show DsbA1 in oxidized form (Figure 5A, lane 3 and 4). However, in the absence of both dsbB1 and dsbB2, DsbA1 accumulates in the reduced state (Figure 5A, lane 6). These results are in agreement with previous studies (Arts et al., 2013). Similarly, when wild type cells are incubated with 25 and 50 μM of compound 12, they also show accumulation of DsbA1 in the reduced state (Figure 5B), which is consistent with compound inhibition of both DsbB1 and DsbB2. When we investigated the redox state of elastase in the supernatant of cell cultures, most of pre-elastase (PE) in the ΔdsbB1 and ΔdsbB2 mutants existed in the oxidized state with a small amount in the mature form (Figure 5C, lanes 5 and 6), while both ΔdsbA1 and ΔdsbB1B2 show pre-elastase in the reduced state and with very little mature elastase (Figure 5C, lanes 3 and 4). This is also consistent with previous findings of the ΔdsbA1 mutant (Braun et al., 2001). When wild type cells are incubated in the presence of 50 μM compound 12, we observe accumulation of pre-elastase in both reduced and oxidized states and similar levels to wild type of mature elastase. There is also a prominent band that runs below pre-elastase and does not correspond to any other reported forms, which may be a degradation product of pre-elastase. This band is more prominent when DsbB1 is mutated or when the cells are treated with compound 12 (Figure 5C, lane 5 and 2). Altogether these results indicate that there is accumulation of reduced DsbA1 and reduced pre-elastase when wild type cells are incubated with compound 12, a PaDsbB1 and PaDsbB2 inhibitor.

Figure 5.

In vivo accumulation of reduced PaDsbA1 (PaDsbB1 and PaDsbB2 substrate) and PaLasB (PaDsbA1 substrate) caused by compound 12. A) In vivo redox states of PaDsbA1 in wild type and mutants. Cells were grown aerobically in M63 minimal media and TCA-precipitated proteins were alkylated with 12.5 mM of MalPEG-2kDa. Samples were run by SDS-PAGE and immunoblotted with α-PaDsbA1 antibody. Dithiothreitol (DTT) was used for reducing disulfide bonds in controls. B) In vivo redox states of PaDsbA1 in wild type cells grown in the presence of compound 12. Samples were processed as above. C). In vivo redox states of PaLasB in wild type cells grown in the presence of compound 12 and in mutants. Cells were grown aerobically in M63 minimal media and precipitated proteins from the supernatant were alkylated with 20 mM of NEM. Samples were analyzed by SDS-PAGE and immunoblotted with α-PaLasB antibody. ox: oxidized, refers to the position of the oxidized protein which is the same as that of the reduced protein with no alkylating agent present; red: reduced, refers to bands where the positions of the protein with reduced cysteines are detected because of the alkylation which adds to the molecular weight; PreLasB: pre-elastase; LasB: elastase; *: degraded Pre-LasB. Blots show representative results from at least two independent experiments.

In order to assess in vitro inhibition of the Pseudomonas DsbBs, we purified the enzymes DsbB1, DsbB2 as well as DsbA1 enzymes from P. aeruginosa. We then measured the DsbB’s activity in oxidizing DsbA using a quinone reduction assay (Regeimbal et al., 2003), calculating IC₅₀ and K_i for each inhibitor (Table 1, columns 5–8). We also tested these inhibitors in the same assay but using E. coli DsbB and DsbA (Table 1, columns 9–10). Overall, the IC₅₀s of P. aeruginosa DsbB1 and DsbB2 were similar. With the exception of PA1, all the compounds were effective against both DsbB1 and DsbB2. All compounds showed activity against E. coli DsbB as well, but the potency of these compounds was not as high as compounds 12 and 36. Note that these two pyridazinone-related molecules were found after several rounds of SAR studies and thus inhibitors found in the HTS reported here were expected to be less active than these two compounds (Landeta et al., 2015; Landeta et al., 2017).

In order to determine whether the Pseudomonas inhibitors obtained in the HTS could have off-target effects on enzymes with thiol-redox activity, we assayed for inhibition of the thiol-redox enzyme, human protein disulfide isomerase (PDI). The assay measures PDI activity in vitro by its ability to reduce insulin. We found that compounds PA2, PA3, PA4 and PA6 inhibited PDI’s activity (Table 1, Column 11). These four compounds together with PA4, which inhibited the RnVKOR expressed in E. coli (Figure 4), allowed us to select PA1 as a candidate to follow up in inhibitor development since PA1 does not target human PDI or RnVKOR and it is active in P. aeruginosa.

Inhibition of MtVKOR in M. tuberculosis

We assessed the effect of MtVKOR inhibitors obtained in the HTS on M. tuberculosis growth using the H37Rv strain as previous results showed that VKOR is essential in this strain (Sassetti and Rubin, 2003; Dutton et al., 2010). We determined MIC for all our 19 candidates and bromindione. With the exception of compounds MT11, MT12, MT15 and MT16, all inhibited growth of wild type M. tuberculosis in the low to high micromolar range (Table 2, Column 2).

Table 2.

Analysis of MtVKOR inhibitors

ID	In vivo assay (MIC, μM)&	Fold change knockdown*	In vitro assay**		PDI*** IC50 (μM)
			IC50 (μM)	Ki (μM)
Bromindione	250 ± 0	1	5 ± 0.73	1 ± 0.09	N
MT1	50.5 ± 0	6.66	ND	ND	N
MT2	26.04 ± 9	32.01	5 ± 0.5	4 ± 0.3	N
MT3	21.68 ± 0	2	4 ± 0.25	2.5 ± 0.1	233.1 (160.1–339.4)
MT4	575.2 ± 0	0.49	ND	ND	68.12 (30.13–154.0)
MT5	125 ± 0	1	ND	ND	N
MT6	138 ± 0	1	ND	ND	N
MT7	41 ± 0	3.46	ND	ND	N
MT8	114.5 ± 0	1	20 ± 2.5	9 ± 1	N
MT9	>979 ± 0	-	3 ± 0.5	2 ± 0.05	N
MT10	25.83 ± 11.2	1	ND	ND	N
MT11	>309.6 ± 0	-	ND	ND	N
MT12	>166 ± 0	-	ND	ND	7.61 (3.397 to 17.08)
MT13	30.25 ± 0	1	ND	ND	N
MT14	169 ± 0	1	ND	ND	N
MT15	>523 ± 0	-	ND	ND	339.5 (82.64 to 1394)
MT16	>521.6 ± 0	-	ND	ND	N
MT17	7.46 ± 0	2	11 ± 1.1	5 ± 0.25	N
MT18	0.4 ± 0.17	0.005	ND	ND	N
MT19	0.5 ± 0	0.125	9 ± 1	3 ± 0.5	291.6 (114.4–743.6)

^&In vivo inhibition was done by growth inhibition using an alamar blue assay.

^*Fold change was determined using M. tuberculosis JR1740 strain in which CRISPRi is induced with ATc. The MIC was determined by triplicate using alamar blue assay and the fold change was determined by dividing the MIC of the strain grown without ATc (non-induced) to the MIC of the strain grown with 500 ng mL⁻¹ of ATc (induced). Fold change was not calculated for compounds that did not inhibit growth at the highest concentration tested.

^**In vitro inhibition was measured by following the reduction of ubiquinone with purified His-tagged enzymes. The assay used 10 nM of MtVKOR with 20 μM reduced-EcDsbA and 25 μM Ubiquinone-1. For K_I determination 1–100 μM of Ubiquinone-1 was used.

^***PDI activity was measured using the insulin reduction assay.

ND: Not determined. N: No inhibition observed up to 100 μM. Data represent the average of at least two independent experiments with 95% confidence intervals in parenthesis or ±SD.

In order to obtain more evidence that the remaining MT inhibitors were acting in M. tuberculosis directly on VKOR, we constructed a conditionally lethal strain. To this end we used a CRISPRi (CRISPR interference) platform to knock down vkor in M. tuberculosis (Rock et al., 2017). The system allows transcriptional repression using a Streptococcus thermophilus nuclease-dead Cas9 (dCas9_Sth1) under the control of a Tet promoter and a ∼20 bp targeting region into the single guide RNA (sgRNA) scaffold. One of the advantages of this methodology is that it is tunable either by varying the targeted PAM (Protospacer Adjacent Motif within the target DNA sequence) ‘strength’ or by varying the concentration of the dCas9_Sth1 inducer, ATc (Rock et al., 2017). Such a system allowed us to express the minimal amount of vkor necessary for growth, thus sensitizing the strain to potential VKOR inhibitors. We constructed several CRISPRi vectors with different PAM sequences and transformed them into M. tuberculosis. We then tested the growth of the resulting strains under inducing and non-inducing conditions of dCas9_Sth1 (Supplementary Figure 6). The vkor-silenced strains were defective in growth in 7H9 media when CRISPRi is induced and the growth defect correlated with the PAM strength and the ATc concentration (Rock et al., 2017). We selected the JR1740 strain, which displays low PAM strength side with high concentrations of dCas9_Sth1 inducer (Supplementary Figure 6) and partial growth. We determined the protein levels of VKOR in M. tuberculosis JR1740 cell lysates by immunoblotting using an anti-VKOR antibody (Figure 6). This strain has a ∼4-fold decrease in VKOR levels compared to a wild type or non-induced parental strain and a two-fold decrease in growth as measured by absorbance at 600 nm (A₆₀₀), which is indicative of the essentiality of this gene in M. tuberculosis (Figure 6). We then used this knockdown strain to test our MtVKOR inhibitors. If any of the candidate molecules inhibit VKOR, they should display lower MICs in a strain that produces less VKOR (induced with ATc) than a strain on which VKOR levels are comparable to wild type levels (non-induced). Five inhibitors did indeed show lower MICs (MT1, MT2, MT3, MT7 and MT17) than those observed for the non-induced strain (Table 2, column 3). For reasons that are unclear the VKOR knockdown strain displayed more resistance to compounds MT4, MT18 and MT19, opposite to what we expected. Even though more experiments are needed to explain these results, one could speculate these drugs may be targeting another essential process that is not relevant to the cell when growing slowly, such as when VKOR levels are decreased.

Figure 6.

Knocking down disulfide bond forming enzyme VKOR in M. tuberculosis causes a growth defect. A. M. tuberculosis knockdown JR1740 strain shows reduced levels of VKOR when CRISPRi is induced with ATc. M. tuberculosis wild type and JR1740 strains were diluted to an A₆₀₀ of 0.005 and grown in 7H9 media without or with 500 ng mL⁻¹ of ATc at 37ºC for seven days. Cells were lysed and extracts were filtered twice, then proteins were TCA precipitated and quantified by Pierce-BCA method. 50 μg of total protein were run in 12% acrylamide gels and transferred to PVDF membranes to be immunoblotted against anti-MtVKOR antibody or anti-RpoA. B. M. tuberculosis knockdown JR1740 strain shows a decrease in growth and reduced levels of VKOR when CRISPRi is induced with ATc. M. tuberculosis wild type and JR1740 strains were diluted to an A₆₀₀ of 0.005 and grown in 7H9 media without or with 500 ng mL⁻¹ of ATc at 37ºC for seven days. The relative amount of M. tuberculosis growth was calculated comparing the final A₆₀₀ of JR1740 strain ± ATc to the A₆₀₀ of wild type strain. The relative intensity of MtVKOR bands was estimated by ChemiDoc-XRS using wild type as a reference band. Data represent the average of at least three independent experiments ±S.D.

We chose 6 MtVKOR inhibitors to test their effects in vitro with purified MtVKOR and EcDsbA using a ubiquinone reduction assay (Table 2, Column 4 and 5). We selected compounds that were in the low (MT17 and MT19), medium (MT2 and MT3) and high (MT8) micromolar range of M. tuberculosis growth inhibition and one that did not inhibit at the highest concentration tested (MT9) (Table 2, Column 2). All these compounds showed an IC₅₀ in the range of 3 to 20 μM when MtVKOR oxidizes reduced EcDsbA using ubiquinone-1. In this assay MT9 displayed an IC₅₀ of 3 ± 0.5 μM but has no effect on growth of M. tuberculosis at concentrations up to and including 979 μM, suggesting that the drug is not targeting MtVKOR in M. tuberculosis while it targets MtVKOR when expressed in E. coli, in which the MIC is 0.44 μM (Figure 3). This may be due to the drug being metabolized by M. tuberculosis as observed with other molecules (Awasthi and Freundlich, 2017) or the drug may target MtVKOR when partnered with EcDsbA rather than with MtDsbA. Unfortunately, our attempts to use MtDsbA in the ubiquinone reduction assay with MtVKOR showed very low activity, precluding our ability to measure inhibition using both mycobacterial proteins.

Similarly to PaDsbB1 inhibitors, we used PDI in vitro assay to determine off-target effects. We found that compounds MT3, MT4, MT12, MT15 and MT19 inhibited PDI (Table 2, Column 6), while compounds MT1, MT2, MT3, MT5, MT7, MT10, MT13 and MT19 inhibited the RnVKOR expressed in E. coli (Figure 4). Thus, MT17 is the only compound that did not inhibit human PDI nor RnVKOR, is active in M. tuberculosis and displays a two-fold change in the MIC of the M. tuberculosis knockdown strain, making this compound a candidate to follow up in inhibitor development.

Discussion

The animal studies presented here provide evidence that the dsbB1B2 double mutant and dsbA1 mutant of P. aeruginosa are defective in virulence in two mice infection models, pneumonia and keratitis. Animal experiments with the dsbA1 mutant showed slightly larger defects in the lungs than the dsbB1B2 double mutant, while in the eye model both mutant strains were similarly affected. This is not surprising as it is well known that DsbA mutants confer greater defects in disulfide bond formation than DsbB mutants, presumably because some degree of oxidation of DsbA takes place in the presence of oxygen (B.M. Meehan et al., 2017). These animal studies, along with the availability of a sensitive assay for screens, provide additional reason to consider that the search for inhibitors of bacterial DsbBs may yet yield candidates for antibacterial development.

In an attempt to demonstrate inhibition of M. tuberculosis VKOR we have used a CRISPR-dCas9 system to generate knockdown strains of the essential gene vkor (Rock et al., 2017). Similar to previous studies, we observed that the silencing of VKOR affects the growth of M. tuberculosis (Sassetti and Rubin, 2003; Dutton et al., 2010). We anticipate that these knockdown strains will allow the study of disulfide bond formation in M. tuberculosis and may allow for identification of the essential substrates of the pathway.

Our HTS presented in this work allowed detection of inhibitors of the DsbB1 enzyme of P. aeruginosa and VKOR enzyme of M. tuberculosis when each of them is expressed in an E. coli dsbB mutant strain. The HTS based on a sensitive assay, has yielded inhibitors that then were shown to also inhibit the enzymes purified from their native organism, and to interfere with disulfide bond formation in P. aeruginosa cultures and with the growth of M. tuberculosis. In the case of PaDsbB1 inhibitors, we have used an indirect assay to measure DsbB inhibition by measuring elastase activity in the supernatant of cell cultures. Since elastase is a substrate of the disulfide bond formation pathway in P. aeruginosa, its activity is decreased in dsbA1 and dsbB1B2 mutants. For MtVKOR inhibitors, we have made use of one of the knockdown strains to determine the synergy of the growth defect with MtVKOR inhibitors as one way to demonstrate that the small molecules target MtVKOR. While there are many factors affecting enzyme inhibition, we have observed in 6 out of 19 molecules a 2 to 32-fold reduction in the MIC using the knockdown strain, thus helping us prioritize the molecules for further study.

A potential strength of our screen is its selectivity. We observed an unusually low hit rate in the PaDsbB1 screen with a large sample of molecules (∼200,000 compounds) similarly to what we observed in our previous screen with EcDsbB (∼50,000 compounds) (Landeta et al., 2015). The hit rates of DsbB proteins are in the range of 0.01% (Supplementary Table 1) to 0.02% (Landeta et al., 2015). For the MtVKOR screens, the hit rate dropped 5-fold when we used a strain that more efficiently complements the E. coli dsbB mutant (0.057% vs 0.29%, Supplementary Table 1) (Landeta et al., 2015). This observation is consistent with our hypothesis that the much higher rate of hits is explained by the lower effectiveness of oxidation of DsbA, which we overcame significantly by overexpressing MtVKOR (Supplementary Figure 4).

Contrary to what we observed with pyridazinone-related molecules such as compounds 12 and 36 (Landeta et al., 2015; Landeta et al., 2017), the PaDsbB1 inhibitors found in the HTS did not inhibit the DsbBs from Salmonella enterica sv Typhimurium, Klebsiella pneumonia or Vibrio cholerae. Similarly, most of the MtVKOR inhibitors targeted only RnVKOR but not DsbB enzymes in E. coli, with the exception of S. enterica sv typhimurium DsbI, an atypical DsbB homologue that together with DsbL (a DsbA homologue) is dedicated to oxidize arylsulfate-sulfotransferase in S. enterica (Lin et al., 2009; Heras et al., 2010). Thus, overall these molecules show selectivity towards either DsbB or VKOR enzymes. We also found no in vitro inhibition of PDI for 2 out of 6 PaDsbB1 inhibitors and 16 out of 20 MtVKOR inhibitors, which also shows some degree of selectivity among thiol-redox enzymes. The promiscuity of pyridazinone-related molecules toward DsbB enzymes but not toward other thiol-redox enzymes with catalytic cysteines (such as VKOR, DsbA or PDI) is explained by the particular mechanism of action of this class of compounds. We previously demonstrated that pyridazinones displace quinone from its binding site on DsbB and react covalently to a conserved catalytic cysteine in the enzyme, thus preventing catalytic turnover. These newly discovered inhibitors of some DsbB enzymes may therefore speak to a shared mechanism of inhibition and the conservation of those residues/structural features with which these compounds interact.

Our studies suggest that compounds PA1 and MT17 are candidates to follow up in the development of more effective inhibitors of PaDsbB and MtVKOR enzymes, respectively. For instance, further SAR studies of PA1 and MT17 may improve their effectiveness and help in the understanding of the mechanism of inhibition as we have demonstrated for pyridazinone-related compounds 12 and 36 (Landeta et al., 2015; Landeta et al., 2017). While 8 out of 19 MtVKOR inhibitors also inhibited RnVKOR, we think that testing analog structures of these compounds may give more candidates that are selective towards MtVKOR but not RnVKOR. Of particular interest are those molecules that inhibited growth of M. tuberculosis and showed a change in MIC with the VKOR-silenced strain, such as compounds MT2 and MT7. In agreement with this, we have observed that molecules structurally related such as compound MT8, PA4 and PA5 (which are also structurally related to the anticoagulant warfarin) display different activity against RnVKOR and MtVKOR. This suggests that modifying groups at the core structure, in this case coumarin, gives selectivity towards one enzyme or the other. Overall, these molecules represent a starting point for chemical modifications in order to improve their potency and selectivity as well as to unravel their mechanisms of inhibition.

Material and Methods

Bacterial strains and growth conditions

The strains and plasmids used in this study are listed in Table 3. Strains and growth conditions used for screening were done as described previously (Landeta et al., 2015). The E. coli screening strain of PaDsbB1 was constructed by site directed mutagenesis of the −10 box of Trc promoter in pCL25 plasmid (Landeta et al., 2015) in order to convert the plasmid from pDSW204 to pDSW206. The resulting plasmid (pLEM6) was transformed into HK325 and this strain (CL523) was used in X-Gal plates without the addition of IPTG. The E. coli screening strain of MtVKOR was constructed by transforming pRD33 into HK325 (CL382), since our original VKOR screening strain (DHB7657) (Landeta et al., 2015) did not completely restore DsbB activity, giving rise to colonies with a pale blue color on X-Gal plates. The conditions used for CL382 strain in X-Gal plates required the addition of 25 μM of IPTG. FSH250 strain was used with 300 μM IPTG and 60 μg mL⁻¹ of X-Gal instead of the 120 μg mL⁻¹ used for all the other strains (Hatahet et al., 2015). P. aeruginosa clean deletion mutants were constructed using overlap extension PCR and homologous recombination as described before (Horton, 1993). The deletion of the targeted genes was confirmed by PCR amplification and sequencing. Plasmids used to complement P. aeruginosa mutants were constructed cloning the gene and the upstream region containing the promoter (500 bp upstream) using EcoRI-XmaI sites for dsbB1 (pCL84) and XbaI-SacI sites for dsbB2 (pCL85). For dsbA genes the dsbA1 promoter (500bp upstream dsbA1) was cloned in NheI-EcoRI sites and then EcoRI-XmaI sites were used to clone dsbA1 (pCL119) or dsbA2 (pCL120) genes. The resulting plasmids were verified by sequencing and transformed in P. aeruginosa mutants and strains were tested in twitching motility to determine restoration of the mutant phenotype. All E. coli strains were grown in NZ or M63 0.2% glucose either liquid or agar media at 30ºC when indicated. All P. aeruginosa strains were grown in LB or M63 0.2% glucose at 37ºC. The antibiotic concentrations used were: ampicillin 100 μg mL⁻¹, kanamycin 40 μg mL⁻¹ and chloramphenicol 10 μg mL⁻¹, gentamicin 15 μg mL⁻¹ for E. coli or 75 μg mL⁻¹ for P. aeruginosa.

Table 3.

Strains and plasmids

Strain	Genotype	Ref.
E. coli strains
BL21 (DE3)	huA2 [lon] ompT gal (λ DE3) [dcm] ΔhsdS λ DE3 = λ sBamHIo ΔEcoRI-B int::(lacI::PlacUV5::T7 gene1) i21 Δnin5	NEB
JR6 (FSH94)	C43 (DE3) ΔdsbB (Kmr)	(Regeimbal et al., 2003)
NT	C43 (DE3) ΔdsbB (Kmr) pLysS	This study
HK295	MC1000 Δara714 leu+	(Kadokura and Beckwith, 2002)
HK320	HK295 ΔdsbB	(Kadokura and Beckwith, 2002)
HK325	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr)	Kadokura H.
DHB7935	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) ϕ80::pDSW206dsbB (Ampr)	(Landeta et al., 2015)
DHB7657	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pTrc99aMtVKOR (Ampr) recA::Cmr	(Landeta et al., 2015)
CL320	HK295 λatt::malF-lacZ102 (Kmr) ΔdsbB::Trc204promoter-KpdsbB (Ampr)	(Landeta et al., 2015)
CL377	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204PadsbB2 (Ampr)	(Landeta et al., 2015)
CL378	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204AbdsbB (Ampr)	(Landeta et al., 2015)
CL369	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204StdsbB (Ampr)	(Landeta et al., 2015)
CL368	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204StdsbI (Ampr)	(Landeta et al., 2015)
CL373	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204VcdsbB (Ampr)	(Landeta et al., 2015)
CL370	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204FtdsbB (Ampr)	(Landeta et al., 2015)
CL371	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204HidsbB (Ampr)	(Landeta et al., 2015)
CL379	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW204 (Ampr)	(Landeta et al., 2015)
FSH250	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) ΔmetL::FRT HslVC160Y YidCT362I ΔtnaA::CAT (Cmr) pTrc99a-RnVKORc1 ΔA31AR (pJBL420.1, Ampr)	(Hatahet et al., 2015)
CL382	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pTrc99a-6XHisMtVKOR (Ampr)	This study
CL523	HK295 ΔdsbB λatt::malF-lacZ102 (Kmr) pDSW206PadsbB1 (Ampr)	This study
P. aeruginosa strains
PA14	P. aeruginosa UCBP-PA14	Lory Lab collection
CL384	P. aeruginosa ΔdsbA1 (PA14_72450)	This study
CL339	P. aeruginosa ΔdsbB1 (PA14_07000)	This study
CL355	P. aeruginosa ΔdsbB2 (PA14_69400)	This study
CL356	P. aeruginosa ΔdsbB1 ΔdsbB2	This study
CL540	P. aeruginosa ΔdsbA2 (PA14_59960)	This study
CL541	P. aeruginosa ΔdsbA1 ΔdsbA2	This study
CL725	P. aeruginosa ΔdsbA1 ΔdsbA2 (CL541) pJN105-PdsbA1-PadsbA1 (pCL119, Gmr)	This study
CL726	P. aeruginosa ΔdsbB1 ΔdsbB2 (CL356) pJN105-PdsbB1-PadsbB1 (pCL84, Gmr)	This study
M. tuberculosis strains
WT	M. tuberculosis H37Rv	Rubin Lab collection
JR1740	M. tuberculosis pJR1740 inserted into L5 attachment site	This study
Plasmids
pTrc99a	Expression vector, Ampr	(Amann et al., 1988)
pDSW206	Promoter down mutation in −10 and −35 of pTrc99A	(Weiss et al., 1999)
pDSW204	Promoter down mutation in −35 of pTrc99A	(Weiss et al., 1999)
pJN105	Broad-host-range cloning vector with araC-PBAD (Gmr)	(Newman and Fuqua, 1999)
pET28a	Expression vector, T7lac promoter, N-terminal and C-terminal His tag, thrombin cleavage site, pBR322 origin, Kmr	EMD
pET23a	Expression vector, T7lac promoter, C-terminal His tag, thrombin cleavage site, pBR322 origin, Ampr	EMD
pLysS	pACYC184-PTetT7 lysozyme gene (at BamHI), p15A origin (Cmr)	Novagen
pRD33	pTrc99a-6XHis-Mtvkor (Ampr)	(Dutton et al., 2008)
pCL25	pDSW204PadsbB1 (Ampr)	(Landeta et al., 2015)
pLEM6	pDSW206PadsbB1 (Ampr)	This study
pCL84	pJN105-PdsbB1-PadsbB1 (Gmr)	This study
pCL119	pJN105-PdsbA1-PadsbA1 (Gmr)	This study
pCL85	pJN105-PdsbB2-PadsbB2 (Gmr)	This study
pCL120	pJN105-PdsbA1-PadsbA2 (Gmr)	This study
pJR1740	Mtvkor (Rv2968c) sgRNA under constitutive promoter, dCas9Sth1 under Tet promoter (Kmr)	This study
pFL39	pET28a-EcDsbA (N’ 6His tag, cloned at NdeI-XhoI)	(Landeta et al., 2015)
pWM76	pQE70-EcDsbBC8A,C49V-6His (Ampr)	(Regeimbal et al., 2003)
pCL96	pET23a-MtVKOR (C’ 6His tag, cloned at NdeI-XhoI)	This study
pCL97	pET23a-MtDsbA1–255 (C’ 6His tag, cloned at NdeI-XhoI)	This study
pCL98	pET23a-ssTorT1–18-MtDsbA47–255 (C’ 6His tag, cloned at NdeI-XhoI). DsbA’s transmembrane segment has been replaced by TorT signal sequence),	This study
pCL99	pET28a-MtDsbA47–255 (N’ 6His-tag, cloned at NdeI-XhoI, deletion of transmembrane segment)	This study
pNT1	pET23a-PaDsbA121–211 (C’ 6His tag, cloned at NdeI-XhoI)	This study
pNT2	pET23a-PaDsbB1C154Y (C’ 6His tag, cloned at NdeI-XhoI, non-catalytic cysteine mutated)	This study
pNT3	pET23a-PaDsbB2C17A,C53A (C’ 6His tag, cloned at NdeI-XhoI, non-catalytic cysteines mutated)	This study

P.aeruginosa growth under anaerobiosis

To prepare anaerobic media, M63 supplemented with 0.2% glucose, 0.2% casamino acids and 40 mM KNO₃ was boiled and incubated for two days in a Coy anaerobic chamber containing 85% nitrogen, 10% carbon dioxide, and 5% hydrogen (B. M Meehan et al., 2017). Then, 5 mL of this media were dispensed in glass tubes sealed with plastic caps to maintain anaerobiosis outside the anaerobic chamber. P. aeruginosa strains ΔdsbA1 (CL384), ΔdsbA1A2 (CL541), ΔdsbB1B2 (CL356) and wild type were grown aerobically overnight in LB, washed with PBS buffer and injected with a syringe into the 5 mL of anaerobic M63 media to an A₆₀₀ of 0.0001 (∼10⁵ CFU mL⁻¹, bacteria were diluted and enumerated on LB plates). Cultures were incubated for 24 h at 37ºC yielding ∼10⁸ CFU mL^-1. 50 μL of these anaerobic grown cultures were then diluted (∼10⁶ CFU mL⁻¹) into 5 mL of fresh anaerobic minimal media and grown for 24 h at 37ºC, again reaching ∼10⁸ CFU mL^-1. Two additional passages were done with the same results. Therefore, no obvious growth defect was observed after a total of approximately 28 doublings.

High Throughput Screen of Compound Libraries

Compound collections were supplied by the Institute of Chemistry and Cell Biology (ICCB) at Harvard Medical School. Libraries different from the previous tested libraries (Landeta et al., 2015) were selected. The screen included 216,767 compounds from several commercial small molecule libraries (Supplementary Table 1). Agar screening plates and 100 nL-compound transferred into the agar were done as reported before (Landeta et al., 2015). Final concentration of compounds in screening ranged from 16–25 μM. The compounds that appeared as hits in the first round of screening were confirmed following cherry-picking and re-testing (at final concentration of 15 and 50 μM) using Hewlett Packard D300 dispenser and HPD Digital Dispenser 3.1.2 Premier software (USA). For PaDsbB1 hits two additional strains were tested in the cherry-pick assay, weakened MtVKOR (DHB7657) and RnVKOR (FSH250) to discard molecules that were not targeting specifically PaDsbB1 or can have an anticoagulant side effect. On the other hand for MtVKOR hits, weakened EcDsbB (DHB7935) and RnVKOR (FSH250) were used instead. This allowed us to narrow down the molecules to resupply.

Compound resupply and retest of small molecule inhibitors

Compound 36 was obtained by custom synthesis from Sundia MediTech Company Ltd China, its purity was over 95% when analyzed by NMR and LCMS. Bromindione (sc-396742) was purchased from ChemCruz (USA). Compound 12 (EN300–173996, purity 95%), PA1 (Z56868922), PA2 (Z55331918), PA3 (EN300–43376, purity 95%), PA6 (EN300–03922) MT1 (Z56871780), MT2 (Z56801377), MT5 (EN300–220348) were purchased from Enamine (Ukraine). Compound MT9 (A2385) and MT10 (PHR1686–1G) were purchased from Sigma (USA). Compounds PA4 (JFD02470SC) and PA5 (RJC03767SC) were purchased from Maybridge, Ltd (UK). Compound MT8 (5162526) was purchased from Hit2Lead (USA). Compounds MT15 (G751–2182), MT16 (L466–0343), MT17 (L438–0032), MT18 (L454–0207) and MT19 (D268–0470) were purchased from ChemDiv, Inc (USA). Compounds MT6 (OSSK_695834), MT13 (OSSL_944753) and MT14 (OSSK_748556) were purchased from Princeton BioMolecular Research, Inc (USA). Compound MT3 (19063) was purchased from MolMall Sarl (Switzerland), Compound MT4 (ST014023) from TimTec (USA), MT7 (AG-690/10252054) from Specs (USA), MT11 (MD-0230) from Key Organics, Ltd (UK), MT12 (BML-GR343–0010) from EnzoLife Sciences (USA). All purchased compounds were analyzed by mass spectrometry (LCMS) to verify their molecular weights and to confirm their purity (over 90%). All resupplied molecules were dissolved in DMSO using a wider range of concentrations (often from 50 μM to 0.5 μM) for pin-transfer similar to the process of HTS and tested against the collection of E. coli strains to determine MBCs (minimal concentration that causes a pale blue color in the well) as described before (Landeta et al., 2015).

Protein purification and in vitro assays

Inhibitors were evaluated in vitro by a ubiquinone-1 (UQ-1) reduction assay at 275 nm. His-tagged DsbB or VKOR enzymes were produced in E.coli using fermentation and purification as described before (Regeimbal et al., 2003; Landeta et al., 2015). Purity of proteins was over 90% as determined by SDS-PAGE. Reduced PaDsbA1 or EcDsbA was used as a source of electrons for PaDsbB1/B2 or MtVKOR/EcDsbB, respectively. For determination of inhibition constants (IC₅₀ and K_I), various amounts of inhibitors were mixed with 10 nM DsbB in 50 mM sodium phosphate buffer (pH 6) containing 300 mM NaCl, 0.2% DDM (Affymetrix Inc.; Santa Clara, CA, USA) and UbQ-1 (Sigma, 1–100 μM). For VKOR, 10% glycerol was added to the buffer and 0.3% DDM was used instead of 0.2%. Reactions were started at room temperature by the addition of small amounts of highly concentrated DsbA solution to give a final concentration of 20 μM. The following enzyme kinetics were determined for EcDsbB: K_m(UQ-1) of 2.1 μM, K_m(EcDsbA) of 4.6 μM and k_cat of 4.1. For PaDsbB1: K_m(UQ-1) of 3 μM, K_m(PaDsbA1) of 3 μM and k_cat of 3. For PaDsbB2: K_m(UQ-1) of 2.9 μM, K_m(PaDsbA1) of 2.9 μM and k_cat of 2.1. For MtVKOR: K_m(UQ-1) of 5.5 μM, K_m(EcDsbA) of 7 μM and k_cat of 1.1.

PDI insulin reduction assay

PDI enzymatic activity was evaluated using PROTEOSTAT^® PDI assay kit (Enzo Life Sciences, Inc., USA) as per manufacturer’s instructions. 1 μL of compounds dissolved in DMSO were added in the final concentrations: 0.5, 10, 50 and 100 μM. 1 μL of DMSO was used instead to determine 100% PDI activity. Bacitracin was used as a positive control and insulin only or insulin with compound were used as negative controls. To follow the kinetics of the reaction, the PROTEOSTAT^® PDI stop reagent was substituted by buffer. Plates were incubated in a microplate reader (VERSAmax) for 1 h at room temperature while reading the kinetics of the reaction every 47 sec at λ_ex 500 nm/λ_em 603 nm. For compounds that interfered with PROTEOSTAT^® PDI detection reagent, absorbance was used instead substituting the dye by buffer and A₆₅₀ was read every minute during 1 h. Slopes were calculated with SoftMax®Pro (Molecular Devices, LLC) and used to determine inhibition using PDI with no drug activity as 100% activity. IC₅₀ was determined using non-linear log inhibitor vs normalized response with Prism 6 software (GraphPad, La Jolla USA).

Determination of secreted elastase activity in P. aeruginosa cultures

P. aeruginosa strains were grown in LB overnight, washed with PBS buffer and diluted in M63 containing 0.2% glucose 0.2% casamino acids media to an A₆₀₀ of 0.02. 200 μL aliquots of diluted cultures were placed in 96-well plates with flat-bottom (Greiner Bio-one, Austria) and different concentrations of compounds dissolved in DMSO were added to a final concentration of 0.2% DMSO. Serial two-fold dilutions ranging from 0.195 to 100 μM of drug were tested. Plates were incubated at 37ºC and 800 rpms 80% humidity in an orbital shaker (Multitron, ATR). After 15 h of growth, 10 μL of cells were diluted in 90 μL of PBS to read A₆₀₀ in a microplate reader (VERSAmax). All strains grew to a similar A₆₀₀, see growth curve in Supplementary Figure 7. Then, the rest of cells were transferred to v-shape 96-well plates (Greiner Bio-one) to centrifuge for 10 min at 3,000 rpms. Aliquots of 10 μL of supernatant were placed into 96-well black plates with clear flat bottom (Corning, Costar) and diluted with 85 μL of 50 mM Tris-HCl pH 7 with 2.5 mM CaCl₂ buffer and these dilutions were used to determine elastase activity. The reaction was started with the addition of 5 μL of 5 mM metalloendopeptidase substrate (2-aminobenzoyl-Ala-Gly-Leu-Ala-4-nitrobenzylamide, #SC00185, Sigma USA) to reach a final concentration of 250 μM. Plates were incubated in a microplate reader (VERSAmax) for 1 h at 37ºC while reading the kinetics of the reaction every minute at λ_ex 320 nm/λ_em 430 nm. Slopes were calculated with SoftMax®Pro software (Molecular Devices, LLC) and divided by A₆₀₀ to correlate with growth, these were then used to determine the extent of inhibition using wild type strain (or ΔdsbB1 or ΔdsbB2) without inhibitor (only 0.2% DMSO), which activity was considered as 100%. ΔdsbA and ΔdsbB1B2 mutants were used as negative controls since little or no activity (below level of detection, <0.1 μg/mL/A₆₀₀) was detected in their supernatants under these conditions. Each concentration of drug was tested by triplicate and used to calculate the IC₅₀ (concentration of drug where 50% inhibition of elastase activity is observed compared to no-drug condition) and their 95% confidence intervals using non-linear log inhibitor vs normalized response with Prism 6 software (GraphPad, La Jolla USA). Amount of elastase in the supernatant of wild type was estimated to be around 127.3 ± 19 μg/mL/A₆₀₀ and it was calculated by generating a concentration curve with a solution of purified elastase (EMD Millipore, Calbiochem #324676) dissolved in the same assay buffer.

The redox states of P. aeruginosa DsbA and LasB

P. aeruginosa strains were grown in LB overnight, washed with PBS buffer and diluted in 2 mL of M63 0.2% glucose media to an A₆₀₀ of 0.01. Different concentrations of compound 12 dissolved in DMSO were added to a final concentration of 0.2% DMSO and incubated for 15h at 37ºC and shaken in a roller drum. Cells were then precipitated with TCA and alkylated with 12.5 mM of MalPEG-2kDa (α-[3-(3-Maleimido-1-oxopropyl)amino]propyl-ω-methoxy, polyoxyethylene, NOF corporation) as described previously (Landeta et al., 2015). For reduced controls, TCA-precipitated samples were treated with 100 mM Dithiothreitol (DTT, Sigma) for 30 min at room temperature. A 10 μg-aliquot of the total protein was diluted with non-reducing Laemmli loading buffer and subjected to SDS-PAGE (12% acrylamide, Biorad), semi-dry transfer and western blotting with 0.5 μg mL⁻¹ of α-PaDsbA antibody, which was raised using the whole protein without its signal sequence (PaDsbA_Δ1–22, GenScript, USA). For elastase alkylation we processed samples as described previously (Braun et al., 2001). Briefly, P. aeruginosa strains were grown in LB overnight, the bacteria were washed with PBS buffer and diluted in 2 mL of M63 0.2% glucose media to an A₆₀₀ of 0.01. Compound 12 was added to a final concentration of 50 μM in a final concentration of 0.2% DMSO and incubated for 22 h at 37ºC and shaken in a roller drum. Cultures were centrifuged for 5 min at 13,000 rpm in a microcentrifuge. The recovered supernatant was used for alkylation with 20 mM NEM (N-Ethylmaleimide, Thermo Scientific). Proteins were precipitated as described above and a 7.5 μg-aliquot of the total protein was subjected to non-reducing SDS-PAGE (12% acrylamide, Biorad), semi-dry transfer and western blotting with α-PaLasB antibody using a 1:6,000 dilution (Braun et al., 2001).

Animals

For lung infection studies, female ICR mice weighing 22 ± 2 g were provided by BioLasco Taiwan (under Charles River Laboratories Licensee). Animals were acclimated for 3 days prior to use and were confirmed to be in good health. The space allocation for animals was 30 × 19 × 13 cm. All animals were maintained in a hygienic environment with controlled temperature (20 – 24ºC), humidity (30% - 70%) and 12 hours light/dark cycles. Free access to sterilized standard lab diet [MFG (Oriental Yeast Co., Ltd., Japan)] and autoclaved tap water were granted for the study duration. All aspects of this work, including housing, experimentation, and disposal of animals were performed in general accordance with the Guide for the Care and Use of Laboratory Animals (National Academy Press, Washington, D. C., 2011). The study was performed in AAALAC accredited ABSL2 laboratory under the supervision of the veterinarians and the animal care and use protocol was approved by the IACUC at Eurofins Panlabs Taiwan.

For eye infection studies, the experimental protocols were approved by the Institutional Animal Care and Use Committee of the Harvard Medical Area, Office for Research Subject Protection and were consistent with the ARVO Animal Statement for the Use of Animals in Ophthalmic and Vision Research.

Lung infection studies with neutropenic mice (Third-party study by Eurofins)

Two pilot infection experiments (Supplementary Figure 2 and Supplementary Tables 2–1 to 2–6) were conducted with the parental wild type PA14 strain to identify an optimal inoculum density. Groups of 6 female specific-pathogen-free ICR mice weighing 22 ± 2 g were used. Animals were immunosuppressed by two intraperitoneal injections of cyclophosphamide, the first at 150 mg/kg four days before infection (day –4) and the second at 100 mg/kg, one day before infection (day –1). On day 0, animals were inoculated intranasally (0.02 mL/mouse) with target inoculation densities of 10², 10³, 10⁴, 10⁵, 10⁶, and 10⁷ CFU per mouse of P. aeruginosa PA14 wild type strain. Animals were first anesthetized with intravenous administration of etomidate-lipuro emulsion, 20 mg/kg, prior to inoculation. The actual inoculum densities were 2.02×10², 2.04×10³, 1.86×10⁴, 1.95×10⁵, 1.85×10⁶, and 1.80×10⁷ CFU/mouse. Animals were sacrificed at 2 or at 26 h after inoculation with CO₂ asphyxiation and the lung tissues were harvested and weighted from each of the test animals. The removed lungs were homogenized in 1 ml of sterile PBS, pH 7.4, and then 10-fold serial dilutions were generated. Aliquots of the dilutions (100 μL) were separately plated onto MacConkey II agar plates. The bacterial counts (CFU/g) in lung tissues were calculated. 10⁴-inoculum was selected for the main study to compare virulence between strains. To evaluate strain virulence, groups of 20 neutropenic mice were intranasally infected with a target density of 10⁴ CFU/mouse of each strain. Bacterial counts were measured in lung tissue of 5 animals sacrificed at 2 h after infection and 15 animals sacrificed at 26 h after infection.

Eye infection studies in mice (Harvard Medical School)

Scratch-injured eyes of anesthetized C57BL/6 female mice (N= 5 per group; 6 weeks old) were challenged in vivo with the P. aeruginosa PA14 or mutants as described before(Preston et al., 1995). Briefly, the infecting inoculum was delivered in a 5 μL volume with ∼10⁷ CFU/eye. The actual bacterial counts for experiment shown in Figure 2B were WT: 1×10⁷, ΔdsbB1B2: 1.4×10⁷, and ΔdsbA1A2: 1.1×10⁷ CFU/eye. The actual bacterial counts for experiment shown in Figure 2C were ΔdsbB1B2: 1×10⁷, ΔdsbA1A2: 1×10⁷, ΔdsbB1B2pJN105-dsbB1: 1.2×10⁷, and ΔdsbA1A2pJN105dsbA1: 1.6×10⁷ CFU/eye. The severity of corneal disease was scored by a masked observer on a scale of 0 to 4: 0, eye macroscopically identical with the uninfected contralateral control eye; 1, faint opacity partially covering the pupil; 2, dense opacity covering the pupil; 3, dense opacity covering the entire anterior segment; 4, perforation of the cornea, phthisis bulbi (shrinkage of the globe after inflammatory disease), or both. The disease severity scores observed at 48 hours after infection were determined, the mice were euthanatized by CO₂ and the extracellular and internalized Colony Forming Units (CFU) of P. aeruginosa were determined by serial dilution and plating. The excised corneas were placed in DMEM with 1% fetal bovine serum (FBS) and mixed with a Vortex mixer for 1 minute. The extracellular bacterial counts were determined from the solution obtained after vortexing. Each cornea was removed from the solution and placed in a solution of DMEM with 500 μg mL⁻¹ gentamicin and 1% FBS and incubated for 1 hour at 37°C, to kill any remaining extracellular bacteria. After exposure to gentamicin, the cornea was removed from the gentamicin solution and washed three times in 5 mL solution of DMEM 1% FBS. Next, the cornea was homogenized with 0.5% Triton X-100 in trypticase soy broth, serial dilutions were made in DMEM 1% FBS, and bacteria were plated for enumeration of the intracellular or invading bacterial population.

VKOR knockdown in M. tuberculosis using CRISPRi-dCas9_Sth1

In order to silence VKOR expression, we used a dCas9_Sth1 system reported elsewhere (Rock et al., 2017). The system uses a vector that expresses dcas9 under the control of a Tet promoter (TetR) and PAM sequences with different degrees of strength in silencing genes (Rock et al., 2017). The ∼20 bp vkor (Supplementary Figure 6) targeting region was cloned into the sgRNA scaffold under the control of a constitutive promoter. All vectors were transformed into M. tuberculosis and selected for kanamycin resistance. Induction of dcas9 with ATc caused poor growth of M. tuberculosis strains with strongest PAM sequences (Supplementary Figure 6). Expression of VKOR was determined in the strains that yielded enough cell lysates and total protein (quantified by Pierce BCA kit, Thermo Scientific) to be analyzed by SDS-PAGE and western blotting using 2 μg mL⁻¹ of α-MtVKOR antibody (raised in rabbit using the PMRENRGSQERVGAR peptide as immunogen) and anti-EcRpoA (4RA2; BioLegend, USA). MtVKOR band intensity was estimated using ChemiDoc-XRS choosing wild type as a reference band. JR1740 strain was selected due to its lower VKOR levels and partial growth.

M. tuberculosis H37Rv growth inhibition

A culture of M. tuberculosis was grown to stationary phase and diluted to A₆₀₀ of 0.005. Compounds (in DMSO, final concentration 1%) were dissolved in 100 μL of 7H9 medium (±ATc) and subsequent serial two-fold dilutions were performed in 96-well plates. A bacterial inoculum of 100 μL was added to each well, yielding a final volume of 200 μL/well. Each compound was tested in triplicate and repeated with bacteria grown in Middlebrook 7H9 medium (Difco) supplemented with 0.2% glycerol, 0.05% Tween 80 (Sigma) and OADC (Becton Dickinson). The plates were sealed with breathable membranes and incubated in a shaker at 37°C. After seven days, 10 μL of Alamar Blue reagent (Biosource) was added to each well and the plates were incubated for 48 h at 37°C. The MIC was defined as the lowest drug concentration that prevented a color change from blue to pink.

Abbreviations

β-Gal^dbs

Disulfide-bond sensitive β-Galactosidase (MalF-LacZ102). LacZ fused to the membrane protein MalF, localizes LacZ in the periplasm making it sensitive to disulfide bond formation

X-Gal

5-Bromo-4-chloro-3-indolyl galactopyranoside

CFU

Colony forming units

IPTG

Isopropyl β-D-1-thiogalactopyranoside

MIC

Minimal Inhibitory Concentration

MBC

Minimal Blue Concentration, the minimal concentration required to see a pale blue color indicating inhibition of DsbB or VKOR in E. coli cells expressing β-Gal^dbs

HTS

High Throughput Screening

NEM

N-Ethylmaleimide

Mal-PEG

α-[3-(3-Maleimido-1-oxopropyl)amino]propyl-ω-methoxy, polyoxyethylene

ATc

Anhydrotetracycline

SAR

Structure-activity relationship

PDI

Protein disulfide isomerase