β-lactamase activity. To investigate the potential of targeting disulfide bond formation as a strategy to overcome resistance mechanisms in challenging pathogens, we chose to primarily explore β-lactamases that are produced by bacteria intimately associated with CF lung infections. DsbA dependence has been previously shown for a handful of such enzymes (Furniss et al., 2022), like the chromosomally encoded class B3 metallo-β-lactamase L1-1 from S. maltophilia (Supplementary file 6, supplementary table 1), which contributes significantly to AMR in this organism (Brooke, 2012), as well as β-lactamases from the GES and OXA families, which are broadly disseminated, but commonly found in P. aeruginosa (Hammoudi Halat and Ayoub Moubareck, 2022; Yoon and Jeong, 2021). Here, we selected six clinically important β-lactamases from different Ambler classes (classes A, B, and D) that are exclusively encoded either by P. aeruginosa or by the Burkholderia complex. The P. aeruginosa enzymes (BEL-1, CARB-2, AIM-1, and OXA-50) are all phylogenetically distinct, while the Burkholderia β-lactamases (BPS-1m and BPS-6) belong to the same phylogenetic class (Supplementary file 1). Class A, C, and D β-lactamases, like the BPS-6 (class A) and OXA-50 (class D) enzymes investigated here, are serine-dependent hydrolases. Serine β-lactamases are structurally related to penicillin binding proteins, which have a major role in the synthesis of the peptidoglycan (Ambler, 1980). By contrast, class B enzymes are evolutionarily distinct and rely on one or two Zn²⁺ ions for catalytic activity (Tooke et al., 2019; Hong et al., 2015). In addition to belonging to different phylogenetic classes, the selected enzymes have different numbers of cysteines, display varied hydrolytic activities, can be both resident on the chromosome or on mobile genetic elements, and have diverse inhibitor susceptibility profiles (Supplementary file 6, supplementary table 1).

We expressed all six β-lactamases in the Escherichia coli K-12 strain MC1000 and its isogenic dsbA deletion mutant. This strain background was selected because it has been traditionally used in oxidative protein folding studies (Landeta et al., 2015; Mössner et al., 1999; Rietsch et al., 1997; Stewart et al., 1999) and it lacks endogenous β-lactamase enzymes or any other mechanisms that could contribute to antibiotic resistance. We recorded β-lactam minimum inhibitory concentration (MIC) values for each enzyme in both strain backgrounds. We found that expression of all test enzymes in the dsbA mutant background resulted in markedly reduced MICs for at least one β-lactam antibiotic (Figure 1 and Supplementary file 2A), compared to the MICs recorded in the wild-type E. coli strain; only differences larger than twofold were considered. These results indicate that the presence of DsbA is important for the function of all tested resistance proteins.

To ensure that effects shown in Figure 1 are not due to factors that are not specific to the interaction of DsbA with the tested β-lactamases, we also performed a series of control experiments. We have previously shown that deletion of dsbA does not affect the aerobic growth of E. coli MC1000, or the permeability of its outer and inner membranes (Furniss et al., 2022). Furthermore, here we observed no changes in MIC values for the aminoglycoside antibiotic gentamicin, which is not degraded by β-lactamases, or between the parental E. coli strain and its dsbA mutant harboring only the empty vector (Figure 1 and Supplementary file 2A). In addition, E. coli strains expressing either of two disulfide-free enzymes, the class A β-lactamases L2-1 and LUT-1 from S. maltophilia and Pseudomonas luteola, respectively, did not exhibit decreased MICs in the absence of dsbA (Figure 1 and Supplementary file 2A). These proteins were selected because they both contain two or more cysteine residues but lack disulfide bonds due to the fact that they are transported to the periplasm, pre-folded, by the Twin-arginine translocation (Tat) pathway, rather than by the Sec system. In the case of L2-1, Tat-dependent transport has been experimentally confirmed (Pradel et al., 2009), whilst LUT-1 contains a predicted Tat signal sequence (SignalP 5.0 (Almagro Armenteros et al., 2019) likelihood scores: Sec/SPI = 0.0572, Tat/SPI = 0.9312, Sec/SPII (lipoprotein)=0.0087, other = 0.0029). Finally, the specific interaction between DsbA and our selected test enzymes was further supported by the fact that complementation of dsbA generally restores MICs to near wild-type values for the latest generation β-lactam that each β-lactamase can hydrolyze (Figure 1—figure supplement 1); we only achieve partial complementation for the dsbA mutant expressing BPS-1m, which we attribute to the fact that expression of this enzyme in E. coli is sub-optimal.

Taken together, our data show that DsbA-mediated disulfide bond formation is important for the function of all tested, species-specific β-lactamases. Of these, the most affected enzymes (largest MIC value decreases; Figure 1 and Supplementary file 2A) are the class A extended-spectrum-β-lactamases (ESBLs) from Burkholderia (BPS-1m and BPS-6) and the class B3 metallo-β-lactamase AIM-1, which, like all other class B enzymes (Ju et al., 2018), is resistant to inhibition by classical β-lactamase inhibitor compounds (Supplementary file 6, supplementary table 1; Tooke et al., 2019).

β-Lactamase abundance and folding. To gain insight into how impairment of disulfide bond formation impacts the production or activity of the tested enzymes (Figure 1), we first performed immunoblotting for all phylogenetically distinct β-lactamases (AIM-1, BEL-1, OXA-50, CARB-2, and BPS-1m) to assess their protein levels in the presence and absence of dsbA. For four of the five tested β-lactamases (AIM-1, BEL-1, OXA-50, and CARB-2), deletion of dsbA resulted in drastically reduced protein levels compared to the levels of the control enzyme L2-1, which remained largely unaffected (Figure 2A). This shows that without their disulfide bonds, these proteins are unstable and are ultimately degraded by other cell envelope proteostasis components (Goemans et al., 2014). This was further corroborated by the fact that lysates from dsbA mutants expressing these four enzymes showed significantly reduced hydrolytic activity towards the chromogenic β-lactamase substrate nitrocefin (Figure 2B). In the case of BPS-1m, enzyme levels were unchanged in the absence of dsbA (Figure 2A). However, without its disulfide bond, this protein was significantly less able to hydrolyze nitrocefin (Figure 2B), suggesting a folding defect that results in loss of function. The latter is consistent with the reduced MICs conferred by BPS-1m (and its sister enzyme BPS-6) in the absence of dsbA (Figure 1). The data presented so far (Figures 1 and 2) demonstrate that disulfide bond formation is essential for the biogenesis (stability and/or protein folding) and, in turn, activity of an expanded set of clinically important β-lactamases, including enzymes that currently lack inhibitor options.

Sensitization of multidrug-resistant P. aeruginosa clinical isolates. The efficacy of commonly used treatment options against P. aeruginosa in CF lung infections, namely piperacillin-tazobactam and cephalosporin-avibactam combinations, as well as more advanced drugs like aztreonam or carbapenems (George et al., 2009; Kunz Coyne et al., 2022), is increasingly threatened by an array of β-lactamases, encompassing both broadly disseminated enzymes and species-specific ones (George et al., 2009; Kunz Coyne et al., 2022; Haines et al., 2022). To determine whether the effects on β-lactam MICs observed in our inducible system (Figure 1 and Furniss et al., 2022) can be reproduced in the presence of other resistance determinants in a natural context with endogenous enzyme expression levels, we deleted the principal dsbA gene, dsbA1, in several multidrug-resistant (MDR) P. aeruginosa clinical strains (Supplementary file 6, supplementary table 2). Pathogenic bacteria often encode multiple DsbA analogues (Heras et al., 2009; Landeta et al., 2018) and P. aeruginosa is no exception. It encodes two DsbAs, but DsbA1 has been found to catalyze the vast majority of the oxidative protein folding reactions taking place in its cell envelope (Arts et al., 2013).

We first tested two clinical isolates (strains G4R7 and G6R7; Supplementary file 6, supplementary table 2) expressing the class B3 metallo-β-lactamase AIM-1, for which we recorded reduced activity in an E. coli dsbA background (Figures 1 and 2). This enzyme confers high-level resistance to piperacillin-tazobactam and the third-generation cephalosporin ceftazidime, both anti-pseudomonal β-lactams that are used in the treatment of critically ill patients (Yong et al., 2012). Notably, while specific to the P. aeruginosa genome, aim-1 is flanked by two ISCR15 elements suggesting that it remains mobilizable (Yong et al., 2012; Supplementary file 6, supplementary table 1). MICs for piperacillin-tazobactam and ceftazidime were determined for both AIM-1-positive P. aeruginosa isolates and their dsbA1 mutants (Figure 3AB). Deletion of dsbA1 from P. aeruginosa G4R7 resulted in a substantial decrease in its piperacillin-tazobactam MIC value by 192 µg/mL and sensitization to ceftazidime (Figure 3A), while the dsbA1 mutant of P. aeruginosa G6R7 became susceptible to both antibiotic treatments (Figure 3B). Despite the fact that P. aeruginosa G4R7 dsbA1 was not sensitized for piperacillin-tazobactam, possibly due to the high level of piperacillin-tazobactam resistance of the parent clinical strain, our results across these two isolates show promise for DsbA as a target against β-lactam resistance in P. aeruginosa. To further test our approach in an infection context, we performed in vivo survival assays using the wax moth model Galleria mellonella (Figure 3C), an informative non-vertebrate system for the study of new antimicrobial approaches against P. aeruginosa (Hill et al., 2014). Larvae were infected with P. aeruginosa G6R7 or its dsbA1 mutant, and infections were treated once with piperacillin at a final concentration below the EUCAST breakpoint, as appropriate. No larvae survived beyond 20 hr post-infection when infected with P. aeruginosa G6R7 or its dsbA1 mutant without antibiotic treatment (Figure 3C; blue and light blue survival curves). Despite this clinical strain being resistant to piperacillin in vitro (Figure 3B), treatment with piperacillin in vivo increased larval survival (52.5% survival at 28 hr post-infection) compared to the untreated conditions (Figure 3C; blue and light blue survival curves) possibly due to in vivo ceftazidime MIC values being discrepant to the value recorded in vitro. Nonetheless, treatment of P. aeruginosa G6R7 dsbA1 with piperacillin resulted in a significant improvement in survival (77.5% survival at 28 hr post-infection), highlighting increased relative susceptibility compared to the treated wild-type condition (Figure 3C; compare the red and pink survival curves).

Next, we tested two P. aeruginosa clinical isolates (strains CDC #769 and CDC #773; Supplementary file 6, supplementary table 2), each expressing two class A enzymes from the GES family (GES-19/GES-26 or GES-19/GES-20), for which we have previously demonstrated DsbA dependence (Furniss et al., 2022). The GES family comprises 59 distinct ESBLs (Supplementary file 1), which are globally disseminated and commonly found in P. aeruginosa, as well as other critical Gram-negative pathogens (for example Klebsiella pneumoniae and Enterobacter cloacae) (Weldhagen, 2006). Deletion of dsbA1 in these clinical strains resulted in sensitization to piperacillin-tazobactam and aztreonam for P. aeruginosa CDC #769 (Figure 3D), and to representative compounds of all classes of anti-pseudomonal β-lactam drugs (piperacillin-tazobactam, aztreonam, and ceftazidime) for P. aeruginosa CDC #773 (Figure 3E). P. aeruginosa CDC #773 and its dsbA1 mutant were further tested in a G. mellonella infection model using ceftazidime treatment (Figure 3F). In this case, no larvae survived 24 hr post-infection (Figure 3F; blue, light blue, and red survival curves), except for insects infected with P. aeruginosa CDC #773 dsbA1 and treated with ceftazidime at a final concentration below the EUCAST breakpoint, whereby 96.7% survival was recorded (Figure 3F; pink survival curves).

We have demonstrated the specific interaction of DsbA with the tested β-lactamase enzymes in our E. coli K-12 inducible system using gentamicin controls (Figure 1 and Supplementary file 2A) and gene complementation (Figure 1—figure supplement 1). To confirm the specificity of this interaction in P. aeruginosa, we performed representative control experiments in one of our clinical strains, P. aeruginosa CDC #769. We first tested the general ability of P. aeruginosa CDC #769 dsbA1 to resist antibiotic stress by recording MIC values against gentamicin and found it unchanged compared to its parent (Figure 3—figure supplement 1A). Gene complementation in clinical isolates is especially challenging and rarely attempted due to the high levels of resistance and lack of genetic tractability in these strains. Despite these challenges, to further ensure the specificity of the interaction of DsbA with tested β-lactamases in P. aeruginosa, we have complemented dsbA1 from P. aeruginosa PAO1 into P. aeruginosa CDC #769 dsbA1. We found that complementation of dsbA1 restores MICs to wild-type values for both tested β-lactam compounds (Figure 3—figure supplement 1B) further demonstrating that our results in P. aeruginosa clinical strains are not confounded by off-target effects.

Our data on the sensitization of AIM- and GES-expressing P. aeruginosa clinical isolates to commonly used anti-pseudomonal β-lactam drugs, combined with our previous results on strains producing β-lactamases from the OXA family (Furniss et al., 2022), show that our approach holds promise towards inactivating numerous clinically important Pseudomonas-specific enzymes. These include resistance determinants that cannot be currently targeted by classical β-lactamase inhibitor compounds (for example, enzymes from the OXA and AIM families Tooke et al., 2019) and, therefore, limit our treatment options.

New treatment options for extremely drug-resistant S. maltophilia clinical isolates. We have previously used our inducible E. coli K-12 experimental system to demonstrate that the function of the inhibitor-resistant class B3 metallo-β-lactamase L1-1 from S. maltophilia is dependent on DsbA (Furniss et al., 2022). By contrast, the second β-lactamase encoded on the chromosome of this species, L2-1, which we use as a negative control in this study (Figures 1 and 2), is not DsbA dependent. The hydrolytic spectra of these β-lactamases are exquisitely complementary (Brooke, 2012; Calvopiña et al., 2017), making this bacterium resistant to most β-lactam compounds commonly used for CF patients. Considering that L1 enzymes are the sole drivers of ceftazidime resistance, we wanted to investigate the DsbA dependency of L1-1 in its natural context to determine whether inhibition of oxidative protein folding potentiates the activity of complex cephalosporins against this pathogen.

We compromised disulfide bond formation in two clinical isolates of S. maltophilia (strains AMM and GUE; Supplementary file 6, supplementary table 2), by deleting the main dsbA gene cluster (directly adjacent dsbA and dsbL genes, with DsbL predicted to be a DsbA analogue Landeta et al., 2018) and recorded a drastic decrease of ceftazidime MIC values for both mutant strains (Figure 4A and B). Since S. maltophilia cannot be treated with ceftazidime, there is no EUCAST breakpoint available for this organism. That said, for both tested dsbA dsbL mutant strains, the recorded ceftazidime MIC values were lower than the ceftazidime EUCAST breakpoint for the related major pathogen P. aeruginosa (Calza et al., 2003).

In addition to being resistant to β-lactams, S. maltophilia is usually intrinsically resistant to colistin (Amin et al., 2020), which precludes the use of yet another broad class of antibiotics. Bioinformatic analysis on 106 complete Stenotrophomonas genomes revealed that most strains of this organism carry two chromosomally encoded MCR analogues that cluster with clinical MCR-5 and MCR-8 proteins (Supplementary file 4). We have previously found the activity of all clinical MCR enzymes to be dependent on the presence of DsbA (Furniss et al., 2022), thus we compared the colistin MIC value of the S. maltophilia AMM dsbA dsbL strain to that of its parent. We found that impairment of disulfide bond formation in this strain resulted in a decrease of its colistin MIC value from 32 μg/mL to 0.75 μg/mL (Figure 4C). Once more, there is no colistin EUCAST breakpoint available for S. maltophilia, but a comparison with the colistin breakpoint for P. aeruginosa (4 μg/mL) demonstrates the magnitude of the effects that we observe.

Since the dsbA and dsbL are organized in a gene cluster in S. maltophilia, we wanted to ensure that our results reported above were exclusively due to disruption of disulfide bond formation in this organism. First, we recorded gentamicin MIC values for S. maltophilia AMM dsbA dsbL and found them to be unchanged compared to the gentamicin MICs of the parent strain (Figure 3—figure supplement 1C). This confirms that disruption of disulfide bond formation does not compromise the general ability of this organism to resist antibiotic stress. Next, we complemented S. maltophilia AMM dsbA dsbL. The specific oxidative roles and exact regulation of DsbA and DsbL in S. maltophilia remain unknown. For this reason and considering that genetic manipulation of extremely drug-resistant organisms is challenging, we used our genetic construct optimized for complementing P. aeruginosa CDC #769 dsbA1 with dsbA1 from P. aeruginosa PAO1 (Figure 3—figure supplement 1B) to also complement S. maltophilia AMM dsbA dsbL. We based this approach on the fact that DsbA proteins from one species have been commonly shown to be functional in other species (Ng et al., 1997; Paxman et al., 2009; Santos-Martin et al., 2021; Urban et al., 2001). Indeed, we found that complementation of S. maltophilia AMM dsbA dsbL with dsbA1 from P. aeruginosa PAO1 restores MICs to wild-type values for both ceftazidime and colistin (Figure 3—figure supplement 1D), conclusively demonstrating that our results in S. maltophilia are not confounded by off-target effects.

The DSB proteins have been shown to play a central role in bacterial virulence, and in this context, they have been proposed as promising targets against bacterial pathogenesis (Heras et al., 2009; Landeta et al., 2018; Heras et al., 2015). As a result, several laboratory compounds against both DsbA (Duprez et al., 2015; Totsika et al., 2018) and its partner protein DsbB (Landeta et al., 2015), which maintains DsbA in a catalytically active state (Kadokura et al., 2003), have been developed. We have successfully used one of these inhibitors, 4,5-dichloro-2-(2-chlorobenzyl)pyridazin-3-one, termed ‘compound 12’ in Yong et al., 2012, to achieve sensitization of clinical strains of Enterobacteria to β-lactam and colistin antibiotics (Furniss et al., 2022). Here, we used a derivative compound, 4,5-dibromo-2-(2-chlorobenzyl)pyridazin-3(2 H)-one, termed ‘compound 36’ in Landeta et al., 2017, which is an improved analog of compound 12 and has been shown to target several DsbB proteins from Gram-negative pathogens that share 20–80% in protein identity. Compound 36 was previously shown to inhibit disulfide bond formation in P. aeruginosa via covalently binding onto one of the four essential cysteine residues of DsbB in the DsbA-DsbB complex (Landeta et al., 2017). Since S. maltophilia DsbB shares ~28% protein sequence identity with analogues from P. aeruginosa, we reasoned that this pathogen could be a good candidate for testing DSB system inhibition. Exposure of S. maltophilia AMM to the DSB inhibitor lowered its ceftazidime MIC value by at least 16–20 μg/mL and decreased its colistin MIC value from 32 μg/mL to 2 μg/mL (Figure 4D); this decrease in the colistin MIC is commensurate with the results we obtained for the S. maltophilia AMM dsbA dsbL strain (Figure 4C). The activity of compound 36 is specific to inhibition of disulfide bond formation since the gentamicin MIC values of S. maltophilia AMM remain unchanged in the presence of the inhibitor and treatment of S. maltophilia AMM dsbA dsbL with the compound does not affect its colistin MIC value (Figure 3—figure supplement 1E). Considering that this inhibitor has not been specifically optimized for S. maltophilia strains, the recorded drops in MIC values (Figure 4D) are encouraging and suggest that the DSB system proteins are tractable targets against species-specific resistance determinants in this pathogen.

Currently, the best clinical strategy against S. maltophilia is to reduce the likelihood of infection (Gibb and Wong, 2021); therefore, novel treatment strategies against this organism are desperately needed. Overall, our results on targeting oxidative protein folding in this organism show promise for the generation of therapeutic avenues that are compatible with mainstream antibiotics (β-lactams and polymyxins), which are commonly used for the treatment of other pathogens, for example P. aeruginosa, in CF lung infections.

Inhibition of cross-resistance in S. maltophilia - P. aeruginosa mixed communities. The antibiotic resistance mechanisms of S. maltophilia impact the antibiotic tolerance profiles of other organisms that are found in the same infection environment. S. maltophilia hydrolyzes all β-lactam drugs through the action of its L1 and L2 β-lactamases (Brooke, 2012; Calvopiña et al., 2017). In doing so, it has been experimentally shown to protect other pathogens that are, in principle, susceptible to treatment, such as P. aeruginosa (Bottery et al., 2022). This protection, in turn, allows active growth of otherwise treatable P. aeruginosa in the presence of complex β-lactams, like imipenem (Bottery et al., 2022), and, at least in some conditions, increases the rate of resistance evolution of P. aeruginosa against these antibiotics (Quinn et al., 2022).

We wanted to investigate whether our approach would be useful in abrogating interspecies interactions that are relevant to CF infections. We posited that ceftazidime resistance in S. maltophilia is largely driven by L1-1, an enzyme that we can incapacitate by targeting disulfide bond formation (Furniss et al., 2022; Figure 4A, B and D). As such, impairment of oxidative protein folding in S. maltophilia should allow treatment of this organism with ceftazidime, and at the same time eliminate any protective effects that benefit susceptible strains of co-occurring organisms. With ceftazidime being a standard anti-pseudomonal drug, and in view of the interactions reported between P. aeruginosa and S. maltophilia (Bottery et al., 2022; Quinn et al., 2022; Law et al., 2022), we chose to test this hypothesis using S. maltophilia AMM and a P. aeruginosa strain that is sensitive to β-lactam antibiotics, P. aeruginosa PA14. We followed established co-culture protocols for these organisms (Bottery et al., 2022) and first monitored the survival and growth of P. aeruginosa under ceftazidime pressure in monoculture, or in the presence of S. maltophilia strains. Due to the naturally different growth rates of these two species (S. maltophilia grows much slower than P. aeruginosa), especially in laboratory conditions, the protocol we followed (Bottery et al., 2022) requires S. maltophilia to be grown for 6 hr prior to co-culturing it with P. aeruginosa. To ensure that at this point in the experiment, our two S. maltophilia strains, with and without dsbA, had grown comparatively to each other, we determined their cell densities (Figure 4—figure supplement 1A). We found that S. maltophilia AMM dsbA dsbL had grown at a similar level as the wild-type strain, and both were at a higher cell density (~10⁷ colony forming units (CFUs)) compared to the P. aeruginosa PA14 inoculum (5x10⁴ CFUs).

P. aeruginosa PA14 monoculture cannot grow in the presence of more than 4 μg/mL of ceftazidime (Figure 4E; white bars). However, the same strain can actively grow in concentrations of ceftazidime up to 512 μg/mL in the presence of S. maltophilia AMM (Figure 4E; dark pink bars), showing that the protective effects previously observed with imipenem (Bottery et al., 2022) are applicable to other clinically relevant β-lactam antibiotics. Cross-resistance effects are most striking at concentrations of ceftazidime above 64 μg/ml; for amounts between 16 and 64 μg/ml, P. aeruginosa survives in the presence of S. maltophilia, but does not actively grow. This is in agreement with previous observations showing that the expression of L1-1 is induced by the presence of complex β-lactams (Okazaki and Avison, 2008). In this case, the likely increased expression of L1-1 in S. maltophilia grown in concentrations of ceftazidime equal to or higher than 128 μg/ml promotes ceftazidime hydrolysis and decrease of the active antibiotic concentration, in turn, shielding the susceptible P. aeruginosa strain. By contrast, protective effects are almost entirely absent when P. aeruginosa PA14 is co-cultured with S. maltophilia AMM dsbA dsbL, which cannot hydrolyze ceftazidime efficiently because L1-1 activity is impaired (Furniss et al., 2022; Figure 4A, B and D). In fact, in these conditions, P. aeruginosa PA14 only survives in concentrations of ceftazidime up to 8 μg/mL (Figure 4E; light pink bars), 64-fold lower than what it can endure in the presence of S. maltophilia AMM (Figure 4E; dark pink bars).

To ensure that ceftazidime treatment leads to eradication of both P. aeruginosa and S. maltophilia when disulfide bond formation is impaired in S. maltophilia, we monitored the abundance of both strains in each synthetic community for select antibiotic concentrations (Figure 4—figure supplement 1B). In this experiment, we largely observed the same trends as in Figure 4E. At low antibiotic concentrations, for example 4 μg/mL of ceftazidime, S. maltophilia AMM is fully resistant and thrives, thus outcompeting P. aeruginosa PA14 (dark pink and dark blue bars in Figure 4—figure supplement 1B). The same can also be seen in Figure 4E, whereby decreased P. aeruginosa PA14 CFUs are recorded. By contrast, S. maltophilia AMM dsbA dsbL already displays decreased growth at 4 μg/mL of ceftazidime because of its non-functional L1-1 enzyme, allowing comparatively higher growth of P. aeruginosa (light pink and light blue bars in Figure 4—figure supplement 1B). Despite the competition between the two strains, P. aeruginosa PA14 benefits from S. maltophilia AMM’s high hydrolytic activity against ceftazidime, which allows it to survive and grow in high antibiotic concentrations even though it is not resistant (see 128 μg/mL; dark pink and dark blue bars in Figure 4—figure supplement 1B). In stark opposition, without its disulfide bond in S. maltophilia AMM dsbA dsbL, L1-1 cannot confer resistance to ceftazidime, resulting in killing of S. maltophilia AMM dsbA dsbL and, consequently, also of P. aeruginosa PA14 (see 128 μg/mL; light pink and light blue bars in Figure 4—figure supplement 1B).

The data presented here show that, at least under laboratory conditions, targeting protein homeostasis pathways in specific recalcitrant pathogens has the potential not only to alter their own antibiotic resistance profiles (Figures 3 and 4), but also to influence the antibiotic susceptibility profiles of other bacteria that co-occur in the same conditions (Figure 5). Admittedly, the conditions in a living host are too complex to draw direct conclusions from this experiment. That said, our results show promise for infections, where pathogen interactions affect treatment outcomes, and whereby their inhibition might facilitate treatment.

Unless otherwise stated, chemicals and reagents were acquired from Sigma-Aldrich or Thermo Fisher Scientific, growth media were purchased from Oxoid and antibiotics were obtained from Melford Laboratories. Lysogeny broth (LB) (10 g/L NaCl) and agar (1.5% w/v) were used for routine growth of all organisms at 37 °C with shaking at 220 RPM, as appropriate. Mueller-Hinton (MH) broth and agar (1.5% w/v) were used for Minimum Inhibitory Concentration (MIC) assays. Growth media were supplemented with the following, as required: 0.25 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG), 50 μg/mL kanamycin, 100 μg/mL ampicillin, 33 μg/mL chloramphenicol, 33 μg/mL gentamicin (for cloning purposes), 400–600 μg/mL gentamicin (for genetic manipulation of P. aeruginosa and S. maltophilia clinical isolates), 12.5 μg/mL tetracycline (for cloning purposes), 100–400 μg/mL tetracycline (for genetic manipulation of P. aeruginosa clinical isolates), 50 μg/mL streptomycin (for cloning purposes), 2000–5000 μg/mL streptomycin (for genetic manipulation of P. aeruginosa clinical isolates), and 6000 μg/mL streptomycin (for genetic manipulation of S. maltophilia clinical isolates).

Bacterial strains, plasmids, and oligonucleotides used in this study are listed in Supplementary file 6, supplementary tables 2–4, respectively. DNA manipulations were conducted using standard methods. KOD Hot Start DNA polymerase (Merck) was used for all PCR reactions according to the manufacturer’s instructions, oligonucleotides were synthesized by Sigma-Aldrich and restriction enzymes were purchased from New England Biolabs. All constructs were DNA sequenced and confirmed to be correct before use.

Genes for β-lactamase enzymes were amplified from genomic DNA extracted from clinical isolates (Supplementary file 6, supplementary table 5) with the exception of bps-1m, bps-6, carb-2, ftu-1, and smb-1, which were synthesized by GeneArt Gene Synthesis (Thermo Fisher Scientific). β-lactamase genes were cloned into the IPTG-inducible plasmid pDM1 using primers P1-P16. All StrepII-tag fusions of β-lactamase enzymes (constructed using primers P3, P5, P7, P9, P11, P13, P15, and P17-23) have a C-terminal StrepII tag (GSAWSHPQFEK).

P. aeruginosa dsbA1 mutants and S. maltophilia dsbA dsbL mutants were constructed by allelic exchange, as previously described (Vasseur et al., 2005). Briefly, the dsbA1 gene area of P. aeruginosa strains (including the dsbA1 gene and ~600 bp on either side of this gene) was amplified (primers P24/P25) and the obtained DNA was sequenced to allow for accurate primer design for the ensuing cloning step. The pKNG101-dsbA1 plasmid was then used for deletion of the dsbA1 gene in P. aeruginosa G4R7 and P. aeruginosa G4R7, as before (Furniss et al., 2022). For the deletion of dsbA1 in P. aeruginosa CDC #769 and P. aeruginosa CDC #773, ~500 bp DNA fragments upstream and downstream of the dsbA1 gene were amplified using P. aeruginosa CDC #769 or P. aeruginosa CDC #773 genomic DNA (primers P28/P29 (upstream) and P30/P31 (downstream)). Fragments containing both regions were then obtained by overlapping PCR (primers P28/P31) and inserted into the XbaI/BamHI sites of pKNG102, resulting in plasmids pKNG102-dsbA1-769 and pKNG102-dsbA1-773. For S. maltophilia strains, the dsbA dsbL gene area (including the dsbA dsbL genes and ~1000 bp on either side of these genes) was amplified (primers P26/P27), and the obtained DNA was sequenced to allow for accurate primer design for the ensuing cloning step. Subsequently, ~700 bp DNA fragments upstream and downstream of the dsbA dsbL genes were amplified using S. maltophilia AMM or S. maltophilia GUE genomic DNA (primers P32/P33 (upstream) and P34/P35 (downstream)). Fragments containing both of these regions were then obtained by overlapping PCR (primers P32/35) and inserted into the XbaI/BamHI sites of pKNG101, resulting in plasmids pKNG101-dsbA dsbL-AMM and pKNG101-dsbA dsbL-GUE. The suicide vector pKNG101 (Kaniga et al., 1991) and its derivative pKNG102 are not replicative in P. aeruginosa or S. maltophilia; both vectors are maintained in E. coli CC118λpir and mobilized into P. aeruginosa and S. maltophilia strains by triparental conjugation. For P. aeruginosa, integrants were selected on Vogel Bonner Minimal medium supplemented with streptomycin (for P. aeruginosa G4R7 and P. aeruginosa G6R7) or tetracycline (for P. aeruginosa CDC #769 and P. aeruginosa CDC #773). For S. maltophilia, integrants were selected on MH agar supplemented with streptomycin and ampicillin. Successful integrants were confirmed using PCR, and mutants were resolved by exposure to 20% sucrose. Gene deletions were confirmed via colony PCR and DNA sequencing (primers P24/P25).

P. aeruginosa PA14, S. maltophilia AMM, and S. maltophilia AMM dsbA dsbL were labeled with a gentamicin resistance marker using mini-Tn7 delivery transposon-based vectors adapted from Zobel et al., 2015. The non-replicative vectors pTn7-M (labeling with gentamicin resistance only (accC gene), for P. aeruginosa PA14) and pBG42 (labeling with gentamicin resistance and msfGFP, for S. maltophilia strains) were mobilized into the respective recipients using conjugation, in the presence of a pTNS2 plasmid expressing the TnsABC + D-specific transposition pathway. Correct insertion of the transposon into the attTn7 site was confirmed via colony PCR and DNA sequencing (primers P44/P45 for P. aeruginosa, primers P46/P47 for S. maltophilia).

P. aeruginosa CDC #769 dsbA1 and S. maltophilia AMM dsbA dsbL were complemented with DsbA1 from P. aeruginosa PAO1 using a mini-Tn7 delivery transposon-based vector adapted from Zobel et al., 2015. Briefly, the msfGFP gene of pBG42 was replaced with the dsbA1 gene of P. aeruginosa PAO1 by HiFi DNA assembly according to the manufacturer’s instructions (NEBuilder HiFi DNA Assembly, New England Biolabs). The vector was linearized with primers P36/P37 and the dsbA1 gene of P. aeruginosa PAO1 was amplified from genomic DNA using primers P38/P39. msfGFP amplified from pBG42 with primers P40/P41 was reintroduced onto the vector under the PEM7 promoter between the HindIII and BamHI sites of pBG42 (García-Gutiérrez et al., 2020) resulting in plasmid pBG42-PAO1dsbA1. Correct assembly of pBG42-PAO1dsbA1 was confirmed by colony PCR (primers P42/P43) and DNA sequencing. pBG42-PAO1dsbA1 was mobilized into the recipient strains using conjugation, in the presence of a pTNS2 plasmid expressing the TnsABC + D-specific transposition pathway. GFP positive colonies were screened using colony PCR and correct insertion of the transposon into the attTn7 site of clinical strains was confirmed via DNA sequencing (primers P44/P45 for P. aeruginosa, primers P46/P47 for S. maltophilia).

Due to the evolving nomenclature and classification of L1 and L2 enzymes at the time of this study’s publication, the designations L1-1 and L2-1 are used here in a generic sense to refer to enzymes belonging to the L1 and L2 β-lactamase families that are variants of L1-1 and L2-1, respectively.

Unless otherwise stated, antibiotic MIC assays were carried out in accordance with the EUCAST recommendations (Kadeřábková et al., 2024) using ETEST strips (BioMérieux). Briefly, overnight cultures of each strain to be tested were standardized to OD₆₀₀ 0.063 in 0.85% NaCl (equivalent to McFarland standard 0.5) and distributed evenly across the surface of MH agar plates. ETEST strips were placed on the surface of the plates, evenly spaced, and the plates were incubated for 18–24 hr at 37 °C. MICs were read according to the manufacturer’s instructions. MICs were also determined using the Broth Microdilution (BMD) method in accordance with the EUCAST recommendations (Kadeřábková et al., 2024) for specific β-lactams, as required, and for colistin sulfate (Acros Organics). Briefly, a series of antibiotic concentrations was prepared by twofold serial dilution in MH broth in a clear-bottomed 96-well microtiter plate (Corning). The strain to be tested was added to the wells at approximately 5x10⁵ CFUs per well, and plates were incubated for 18–24 hr at 37 °C. The MIC was defined as the lowest antibiotic concentration with no visible bacterial growth in the wells. When used for MIC assays, tazobactam was included at a fixed concentration of 4 μg/mL, in accordance with the EUCAST guidelines. All S. maltophilia MICs were performed in synthetic CF sputum medium (SCFM) as described in Palmer et al., 2007, using ETEST strips (for β-lactam antibiotics) or the BMD method (for colistin). For S. maltophilia GUE, imipenem at a final concentration of 5 µg/mL was added to the overnight cultures to induce β-lactamase production.

The covalent DsbB inhibitor 4,5-dibromo-2-(2-chlorobenzyl)pyridazin-3(2 H)-one (Landeta et al., 2017) was used to chemically impair the function of the DSB system in S. maltophilia strains. Inactivation of DsbB results in abrogation of DsbA function (Kishigami et al., 1995) only in media free of small-molecule oxidants (Dailey and Berg, 1993). Therefore, MIC assays involving chemical inhibition of the DSB system were performed using SCFM media prepared as described in Palmer et al., 2007, except that L-cysteine was omitted. Either DMSO (vehicle control) or the covalent DsbB inhibitor 4,5-dibromo-2-(2-chlorobenzyl)pyridazin-3(2 H)-one (Landeta et al., 2017; Bioduro-Sundia; ¹H-NMR and LCMS spectra are provided in Supplementary file 5), at a final concentration of 50 μM, were added to the cysteine-free SCFM medium, as required.

Samples for immunoblotting were prepared as follows. Strains to be tested were grown on LB agar plates as lawns in the same manner as for MIC assays described above. Bacteria were collected using an inoculating loop and resuspended in LB to OD₆₀₀ 2.0. The cell suspensions were centrifuged at 10,000 x g for 10 min and bacterial pellets were lysed by addition of BugBuster Master Mix (Merck Millipore) for 25 min at room temperature with gentle agitation. Subsequently, lysates were centrifuged at 10,000 x g for 10 min at 4 °C and the supernatant was added to 4 x Laemmli buffer. Samples were boiled for 5 min before separation by SDS-PAGE.

SDS-PAGE analysis was carried out using 10% BisTris NuPAGE gels (Thermo Fisher Scientific) and MES/SDS running buffer prepared according to the manufacturer’s instructions; pre-stained protein markers (SeeBlue Plus 2, Thermo Fisher Scientific) were included. Proteins were transferred to Amersham Protran nitrocellulose membranes (0.45 µm pore size, GE Life Sciences) using a Trans-Blot Turbo transfer system (Bio-Rad) before blocking in 3% w/v Bovine Serum Albumin (BSA)/TBS-T (0.1 % v/v Tween 20) or 5% w/v skimmed milk/TBS-T and addition of primary and secondary antibodies. The following primary antibodies were used in this study: Strep-Tactin-AP conjugate (Iba Lifesciences; dilution 1:3000 in 3 w/v % BSA/TBS-T), and mouse anti-DnaK 8E2/2 antibody (Enzo Life Sciences; dilution 1:10,000 in 5% w/v skimmed milk/TBS-T). Goat anti-mouse IgG-AP conjugate (Sigma-Aldrich; dilution 1:6000 in 5% w/v skimmed milk/TBS-T) was used as a secondary antibody in this study. Membranes were washed three times for 5 min with TBS-T prior to development. Development for AP conjugates was carried out using SigmaFast BCIP/NBT tablets.

Immunoblot samples were also analyzed for total protein content. SDS-PAGE analysis was carried out using 10% BisTris NuPAGE gels (Thermo Fisher Scientific) and MES/SDS running buffer prepared according to the manufacturer’s instructions; pre-stained protein markers (SeeBlue Plus 2, Thermo Fisher Scientific) were included. Gels were stained for total protein with SimplyBlue SafeStain (Thermo Fisher Scientific) according to the manufacturer’s instructions.

β-Lactam hydrolysis measurements were carried out using the chromogenic β-lactam nitrocefin (Abcam). Briefly, overnight cultures of strains to be tested were centrifuged, pellets were weighed and resuspended in 150 μL of 100 mM sodium phosphate buffer (pH 7.0) per 1 mg of wet-cell pellet, and cells were lysed by sonication. Lysates were transferred into clear-bottomed 96-well microtiter plates (Corning) at volumes that corresponded to the following weights of bacterial cell pellets: strains harboring pDM1, pDM1-bla_L2-1, and pDM1-bla_OXA-50 (0.34 mg of cell pellet); strains harboring pDM1-bla_BEL-1, pDM1-bla_AIM-1, and pDM1-bla_SMB-1 (0.17 mg of cell pellet); strains harboring pDM1-bla_POM-1 (0.07 mg of cell pellet); strains harboring pDM1-bla_BPS-1m (0.07 mg of cell pellet); strains harboring pDM1-bla_CARB-2 (0.03 mg of cell pellet). In all cases, nitrocefin was added at a final concentration of 400 μM and the final reaction volume was made up to 100 μL using 100 mM sodium phosphate buffer (pH 7.0). Nitrocefin hydrolysis was monitored at 25 °C by recording absorbance at 490 nm at 60 s intervals for 15 minutes using an Infinite M200 Pro microplate reader (Tecan). The amount of nitrocefin hydrolyzed by each lysate in 15 min was calculated using a standard curve generated by acid hydrolysis of nitrocefin standards.

The wax moth model G. mellonella was used for in vivo survival assays (McCarthy et al., 2017). Individual G. mellonella larvae were randomly allocated to experimental groups; no masking was used. Overnight cultures of all the strains to be tested were standardized to OD₆₀₀ 1.0, suspensions were centrifuged, and the pellets were washed three times in PBS and serially diluted. For experiments with P. aeruginosa G6R7, 10 μL of the 1:10,000 dilution of each bacterial suspension was injected into the last right abdominal proleg of 40 G. mellonella larvae per condition. One hour after infection, larvae were injected with 2.75 μL of piperacillin to a final concentration of 5 μg/mL in the last left abdominal proleg. For experiments with P. aeruginosa CDC #773, 10 μL of the 1:1000 dilution of each bacterial suspension was injected into the last right abdominal proleg of 30 G. mellonella larvae per condition. Immediately after the injection with the inoculum, the larvae were injected with 4.5 μl of ceftazidime to a final concentration of 6.5 μg/mL in the last left abdominal proleg. All larvae were incubated at 37 °C and their mortality was monitored for 30 hr. Death was recorded when larvae turned black due to melanization and did not respond to physical stimulation. For each experiment, an additional ten larvae were injected with PBS as negative control, and experiments were discontinued and discounted if mortality was greater than 10% in the PBS control.

The protection assay was based on the approach described in Bottery et al., 2022. Briefly, 75 μL of double-strength SCFM medium were transferred into clear-bottomed 96-well microtiter plates (VWR) and inoculated with S. maltophilia AMM or its dsbA dsbL mutant that had been grown in SCFM medium at 37 °C overnight; S. maltophilia strains were inoculated at approximately 5x10⁴, as appropriate. Plates were incubated at 37 °C for 6 hr. Double-strength solutions of ceftazidime at decreasing concentrations were prepared by two-fold serial dilution in sterile ultra-pure H₂O and were added to the wells, as required. P. aeruginosa PA14 was immediately added to all the wells at approximately 5x10⁴ CFUs, and the plates were incubated for 20 hr at 37 °C.

To enumerate P. aeruginosa in this experiment, the P. aeruginosa PA14 attTn7::accC strain was used. Following the 20 hr incubation step, serial dilutions of the content of each well were performed in MH broth down to a 10^–7 dilution, plated on MH agar supplemented with gentamicin (S. maltophilia AMM strains are sensitive to gentamicin, whereas P. aeruginosa PA14 attTn7::accC harbors a gentamicin resistance gene on its Tn7 site) and incubated at 37 °C overnight. CFUs were enumerated the following day. S. maltophilia was enumerated in a separate experiment, whereby S. maltophilia AMM attTn7::accC msfgfp or its dsbA dsbL mutant were used. Following the 20 hr incubation step, serial dilutions of the content of each well were performed in MH broth down to a 10^–7 dilution, plated on MH agar supplemented with gentamicin (S. maltophilia AMM strains harbor a gentamicin resistance gene on their Tn7 site, whereas P. aeruginosa PA14 is sensitive to gentamicin) and incubated at 37 °C overnight. CFUs were enumerated the following day.

The total number of performed biological experiments and technical repeats is mentioned in the figure legend of each display item. Biological replication refers to completely independent repetition of an experiment using different biological and chemical materials. Technical replication refers to independent data recordings using the same biological sample.

Antibiotic MIC values were determined in biological triplicate, except for MIC values recorded for dsbA complementation experiments in our E. coli K-12 inducible system that were carried out in duplicate. All ETEST MICs were determined as a single technical replicate, and all BMD MICs were determined in technical triplicate. All recorded MIC values are displayed in the relevant graphs; for MIC assays where three or more biological experiments were performed, the bars indicate the median value, while for assays where two biological experiments were performed, the bars indicate the most conservative of the two values (i.e. for increasing trends, the value representing the smallest increase and for decreasing trends, the value representing the smallest decrease). We note that in line with recommended practice, our MIC results were not averaged. This should be avoided because of the quantized nature of MIC assays, which only inform on bacterial survival for specific antibiotic concentrations and do not provide information for antibiotic concentrations that lie in-between the tested values.

For all other assays, statistical analysis was performed in GraphPad Prism v8.3.1 using either an unpaired T-test with Welch’s correction or a Mantel-Cox log-rank test, as appropriate. Statistical significance was defined as p<0.05. Outliers were defined as any technical repeat >2 SD away from the average of the other technical repeats within the same biological experiment. Such data were excluded, and all remaining data were included in the analysis. Detailed information for each figure is also provided below:

Figure 2B: unpaired T-test with Welch’s correction; n=3; 3.417 degrees of freedom, t-value=0.3927, p=0.7178 (non-significance; for pDM1 strains); 2.933 degrees of freedom, t-value=0.3296, p=0.7639 (non-significance; for pDM1-bla_L2-1 strains); 2.021 degrees of freedom, t-value=7.549, p=0.0166 (significance; for pDM1-bla_BEL-1 strains); 2.146 degrees of freedom, t-value=9.153, p=0.0093 (significance; for pDM1-bla_CARB-2 strains); 2.320 degrees of freedom, t-value=5.668, p=0.0210 (significance; for pDM1-bla_AIM-1 strains); 3.316 degrees of freedom, t-value=4.353, p=0.0182 (significance; for pDM1-bla_OXA-50 strains); 3.416 degrees of freedom, t-value=13.⁶⁸, p=0.0004 (significance; for pDM1-bla_BPS-1m strains).

Figure 3C: Mantel-Cox test; n=40; p=0.3173 (non-significance; P. aeruginosa vs P. aeruginosa dsbA1), p<0.0001 (significance; P. aeruginosa vs P. aeruginosa treated with piperacillin), p<0.0001 (significance; P. aeruginosa dsbA1 vs P. aeruginosa treated with piperacillin), p=0.0147 (significance; P. aeruginosa treated with piperacillin vs P. aeruginosa dsbA1 treated with piperacillin).

Figure 3F: Mantel-Cox test; n=30; p<0.0001 (significance; P. aeruginosa vs P. aeruginosa dsbA1), p>0.9999 (non-significance; P. aeruginosa vs P. aeruginosa treated with ceftazidime), p<0.0001 (significance; P. aeruginosa dsbA1 vs P. aeruginosa treated with ceftazidime), p<0.0001 (significance; P. aeruginosa treated with ceftazidime vs P. aeruginosa dsbA1 treated with ceftazidime).

Figure 5—figure supplement 1C: unpaired T-test with Welch’s correction; n=3; 3.417 degrees of freedom, t-value=0.3927, p=0.7178 (non-significance; for pDM1 strains); 2.933 degrees of freedom, t-value=0.3296, p=0.7639 (non-significance; for pDM1-bla_L2-1 strains); 3.998 degrees of freedom, t-value=4.100, p=0.0149 (significance; for pDM1-bla_POM-1 strains); 2.345 degrees of freedom, t-value=15.02, p=0.0022 (significance; for pDM1-bla_SMB-1 strains).

The following bioinformatics analyses were performed in this study. Short scripts and pipelines were written in Perl (version 5.18.2) and executed on macOS Sierra 10.12.5.

All materials generated in this study are available from the corresponding author upon reasonable request.