Discovery of small-molecule inhibitors of multidrug-resistance plasmid maintenance using a high-throughput screening approach

Significance

Carbapenem-resistant Enterobacteriaceae (CRE) are multidrug-resistant bacterial pathogens designated as urgent threats by the US Centers for Disease Control and Prevention. Resistance to carbapenems and other antibiotics is usually carried on diverse large, low-copy-number plasmids. Interfering with the replication of such resistance plasmids could potentially restore antibiotic susceptibility. Therefore, we designed a high-throughput screen to identify compounds with anti-CRE plasmid activity. We identified several compounds which by blocking plasmid replication and/or evicting a representative IncFIA CRE plasmid potentiated carbapenem activity, thereby providing proof of principle for our approach. Our findings underscore the potential of antiplasmid therapeutics to restore treatment options for highly drug-resistant pathogens and validate a high-throughput screening approach to identify these agents.

Abstract

Carbapenem-resistant Enterobacteriaceae (CRE) are multidrug-resistant pathogens for which new treatments are desperately needed. Carbapenemases and other types of antibiotic resistance genes are carried almost exclusively on large, low-copy-number plasmids (pCRE). Accordingly, small molecules that efficiently evict pCRE plasmids should restore much-needed treatment options. We therefore designed a high-throughput screen to identify such compounds. A synthetic plasmid was constructed containing the plasmid replication machinery from a representative Escherichia coli CRE isolate as well as a fluorescent reporter gene to easily monitor plasmid maintenance. The synthetic plasmid was then introduced into an E. coli K12 tolC host. We used this screening strain to test a library of over 12,000 known bioactive agents for molecules that selectively reduce plasmid levels relative to effects on bacterial growth. From 366 screen hits we further validated the antiplasmid activity of kasugamycin, an aminoglycoside; CGS 15943, a nucleoside analog; and Ro 90-7501, a bibenzimidazole. All three compounds exhibited significant antiplasmid activity including up to complete suppression of plasmid replication and/or plasmid eviction in multiple orthogonal readouts and potentiated activity of the carbapenem, meropenem, against a strain carrying the large, pCRE plasmid from which we constructed the synthetic screening plasmid. Additionally, we found kasugamycin and CGS 15943 blocked plasmid replication, respectively, by inhibiting expression or function of the plasmid replication initiation protein, RepE. In summary, we validated our approach to identify compounds that alter plasmid maintenance, confer resensitization to antimicrobials, and have specific mechanisms of action.

The US Centers for Disease Control and Prevention lists carbapenem-resistant Enterobacteriaceae (CRE) in its most urgent antimicrobial resistant threat category (1). CRE strains are resistant to carbapenems, a class of last-resort β-lactam antibiotics. CRE strains are also largely resistant to multiple other classes of antibiotics as well. As a result, CRE infections are often incredibly difficult, if not impossible, to treat.

Carbapenemase and other antibiotic resistance genes are typically carried on diverse, large, low-copy-number plasmids (2, 3). CRE resistance plasmids (pCRE) are maintained at only 1 to 10 copies per cell and can be as large as 300 kilobases (kb) in size. These large, low-copy-number plasmids utilize a combination of precise replication machinery and maintenance strategies to ensure their persistence, including specific plasmid replication genes, partitioning mechanisms that segregate plasmids to progeny bacteria, resolvases to disentangle plasmid multimers, conjugation systems, and plasmid addiction systems (4). Plasmid addiction systems encode a stable toxin and unstable antidote. Bacterial host cells that lose their plasmids can no longer make antidote and are thus killed by the persisting toxin.

Previously, a limited number of compounds with antiplasmid activity were identified. They were found to intercalate into DNA, cause DNA strand breakage, affect supercoiling, inhibit conjugation, or cause plasmid-dependent fitness defects (5). Almost all, where investigated, acted nonspecifically and were highly toxic to the host bacterial cell, for example chemotherapeutic intercalators. Of interest, however, DeNap et al. (6) and Thomas et al. (7) identified several positively charged aminoglycosides that affected the RNA-regulated plasmid replication machinery of the IncB incompatibility group plasmid, pMU2403. These findings supported the possibility of identifying antiplasmid agents with greater mechanistic specificity.

As pCRE plasmids encode carbapenemase genes, we hypothesized that interference with plasmid maintenance would restore carbapenem susceptibility and offer new therapeutic options. We therefore designed a high-throughput screening approach to identify antiplasmid agents that did not presuppose a particular mechanism of action. For proof of principle, we screened known bioactive compounds that might more quickly advance our understanding of antiplasmid mechanisms and would be available in sufficient quantities for follow-up studies. To do so, we constructed a synthetic screening plasmid (pScreening) containing the IncFIA replication and maintenance machinery from a representative pCRE plasmid as well as a fluorescent reporter that would allow us to easily monitor plasmid loss upon exposure to small-molecule inhibitors. Here, we describe the results of a screen of over 12,000 known-bioactive compounds for antiplasmid activity in which we identify several potent antiplasmid small molecules. We further characterize their activity and find support for the predicted ability to restore antibiotic susceptibility to CRE strains.

Results and Discussion

High-Throughput Screening Assay to Identify Antiplasmid Agents.

To identify prospective anti-pCRE agents, we designed a high-throughput screening approach to identify small molecules that disrupt bacterial plasmid maintenance. Our screening strategy made use of a synthetic screening plasmid (pScreening) that contains the maintenance, replication, and resolvase machinery of a carbapenemase-encoding plasmid from a representative Escherichia coli CRE isolate, BIDMC20a (8). The CRE plasmid in BIDMC20a belongs to the IncF incompatibility group, a leading incompatibility group associated with multidrug resistance (9–12). Specifically, we chose the IncFIA replicon present on pCRE BIDMC20a as the target of the screen (9–12). IncFIA replicons have high sequence conservation and are present on a wide array of diverse plasmids (13). The genetic material incorporated from pCRE BIDMC20a encoded the origin of replication, the RepE replication initiation protein, the IncC incompatibility region, the ResV resolvase, and the SopABC plasmid partitioning system (Fig. 1A). Importantly, we elected to exclude the plasmid addiction system, CcdAB, as we predicted that its presence would hinder our ability to distinguish between antiplasmid and antibacterial effects. The synthetic plasmid reduced the pCRE BIDMC20a plasmid size from 140 kb to 8 kb, focused our screen on plasmid maintenance mechanisms conserved in many CRE plasmids, and in the future will enable testing machinery of other incompatibility groups through simple substitution.

Fig. 1.

High-throughput screen identified antiplasmid compounds. (A) Genetic content of 8.1-kb high-throughput screening plasmid (pScreening) includes ori2, origin of replication; repE, encoding replication initiation protein; IncC, incompatibility region; sopABC, partioning system; and resV, resolvase, cloned from a carbapenemase-encoding plasmid (pCRE BIDMC20a) from a representative CRE strain. In addition, pScreening also encodes a mNeptune2 fluorescent reporter under control of the constitutive proD promoter, which allows real-time monitoring of plasmid levels, and natR, a selectable marker conferring nourseothricin resistance. (B) Plasmid levels of pCRE BIDMC20a and pScreening were comparable by qPCR. Data were normalized to the chromosomal gene dxs and plotted relative to chromosome number. P = 0.82 by Student’s t test. Structures of selected screening hits: (C) kasugamycin, (D) CGS 15943, and (E) Ro 90-7501.

Finally, to monitor plasmid abundance in bacterial cells we also incorporated an mNeptune2 fluorescence reporter gene (14). The mNeptune2 gene was driven by the constitutive, insulated proD promoter (15) and was linked to a nourseothricin selectable marker as previously described (16). Importantly, although we drastically reduced the size of our synthetic construct by over 10-fold compared with the parent plasmid, pCRE BIDMC20a, the synthetic plasmid was maintained at comparable levels based on qPCR analysis (Fig. 1B), suggesting appropriate function of plasmid maintenance mechanisms.

For high-throughput screening purposes, pScreening was introduced into an E. coli K-12 tolC mutant. We reasoned that inactivating major classes of TolC-dependent efflux pumps would allow small molecules to more easily access the bacterial cytoplasm and thereby increase sensitivity of the screening assay (17, 18). Medicinal chemistry approaches could then later be used to develop analogs with more favorable gram-negative membrane penetration characteristics (19, 20).

To perform the high-throughput screen, the screening strain was inoculated into white, clear-bottomed, 384-well plates at a low inoculum. Compounds were robotically transferred to individual test wells. Testing of each compound was performed in duplicate on separate screening plates to reduce the false discovery rate (21). Subsequent incubation at 37 °C allowed ∼12 generations of bacterial growth and concomitant replication of pScreening. After incubation, the screening wells were assayed for both mNeptune2 fluorescence (relative fluorescence units, RFU) to monitor plasmid levels and for absorbance at 600 nm (A₆₀₀) to monitor bacterial growth. Z′ for both measurements routinely exceeded 0.7, supporting robust assay performance (22). We screened a library of 12,301 compounds from the known bioactive collection at the ICCB-Longwood Screening Facility (Harvard Medical School, Boston, MA). Compounds in this collection include United States Food and Drug Administration–approved drugs, and a variety of compounds with at least partially known biological effects including active biolipids, kinase inhibitors, ion channel inhibitors, and G protein-coupled receptor inhibitors.

To distinguish between compounds that affect plasmid maintenance directly rather than indirectly through inhibiting growth of the host cell, we imposed two constraints in selecting positive screening hits: 1) the RFU must be reduced greater than two times more than any reduction in the A₆₀₀ in both replicate plates and 2) screening hits must not be overtly antibacterial, that is, the A₆₀₀ was not reduced by more than 80% compared to the median of all treatment wells in both replicate plates. Using these criteria, we identified 366 compounds with antiplasmid activity (Dataset S1) and 1,205 compounds with antibacterial activity (Dataset S2). The latter was not unexpected as the known bioactive collection contains over 250 known antimicrobials, which are often represented several times in the libraries (23, 24).

Validation of Antiplasmid Activity.

We chose to characterize three antiplasmid compounds based on potency and known activities: kasugamycin (Fig. 1C), CGS 15943 (Fig. 1D), and Ro 90-7501 (Fig. 1E). Kasugamycin is an aminoglycoside antibiotic. Previously, apramycin, another aminoglycoside, was found to block replication of an IncB incompatibility-group plasmid leading to plasmid curing, although the ability to potentiate antimicrobial effects was not specifically examined (6). CGS 15943 is an adenosine nucleoside analog and adenosine receptor antagonist with pharmacological effects in mammals similar to caffeine (25, 26). Ro 90-7501 has several previously identified mammalian cell activities: activating immune pattern recognition receptor signaling, inhibiting protein phosphatase 5, and interfering with amyloid β42 fibril assembly (27–29). CGS 15943 and Ro 90-7501 (Fig. 1 D and E and Dataset S1) were notable for their especially potent antiplasmid activity. We chose not to pursue further analysis of antiplasmid compounds that were known DNA intercalator chemotherapeutics due to their toxicity profiles.

As a first step in hit confirmation, we performed a dose–response assessment with the same readouts as the primary screen: RFU and A₆₀₀ (Fig. 2). CGS 15943 and Ro 90-7501 reduced fluorescence in a dose-dependent manner but had no effect on bacterial growth, confirming highly selective antiplasmid activity (Fig. 2 B and C). Although kasugamycin had antibacterial properties at higher concentrations, it was nevertheless relatively selective for antiplasmid activity (Fig. 2A).

Fig. 2.

Dose–response and selectivity of screening hits. Primary high-throughput screening hits were tested for antiplasmid and antibacterial activity against the screening strain in dose–response experiments. After a 24-h treatment, A₆₀₀ (absorbance to monitor bacterial growth, demarcated with a red line) and RFU (mNeptune2 fluorescence, a marker for plasmid abundance, demarcated with a blue line) were determined. Data were normalized to no treatment controls. (A) Kasugamycin, (B) CGS 15943, and (C) Ro 90-7501.

As an orthogonal confirmation of antiplasmid activity, we directly quantified pScreening plasmid levels using real-time qPCR. To control for antibacterial effects, data were normalized to qPCR quantification of the chromosomal metabolic gene, dxs, in order to determine plasmid level relative to genomic content (30). By qPCR, all compounds significantly reduced pScreening levels in a dose-dependent manner (Fig. 3). Notably, after treatment with either CGS 15943 or Ro 90-7501, complete clearance of pScreening was observed within the assay detection limit (Fig. 3 B and C).

Fig. 3.

Screening hits quantitatively reduced plasmid level by orthogonal qPCR assay. DNA was isolated from bacteria treated with doubling dilutions of (A) kasugamycin, (B) CGS 15943, and (C) Ro 90-7501. Plasmid level, normalized to the single copy chromosomal gene, dxs, was determined by qPCR. Data are plotted as a percentage of plasmid level in no treatment controls.

As an additional quantitative measure of antiplasmid activity, we plated serial dilutions of treated bacteria on media with and without nourseothricin. Since pScreening expresses a nourseothricin resistance marker, bacterial colonies should only form if host cells retain the pScreening plasmid. Using this approach, we found that kasugamycin reduced plasmid levels by about 90% (Fig. 4). In contrast, CGS 15943 and Ro 90-7501 were considerably more effective. Plasmid levels were reduced over 99.9% after treatment with 20 μg/mL of CGS 15943 (Fig. 4). The absolute level of plasmid after the 24-h incubation was similar to the plasmid level in the starting inoculum, indicating that CGS 15943 completely halted plasmid replication and/or passage of the plasmids to daughter cells. When bacteria were treated with Ro 90-7501 at concentrations greater than 10 µg/mL, plasmid levels were reduced over 99.99%. The number of bacteria retaining plasmid was well below the number of bacteria in the starting inoculum, to the extent that very few, if any, bacteria retained plasmid (Fig. 4). This result suggested Ro 90-7501 must go beyond simply interfering with plasmid replication or partitioning and also cause eviction or destruction of remaining plasmid DNA.

Fig. 4.

Compounds reduced plasmid level as determined by colony-forming unit enumeration. The screening strain was treated with doubling dilutions of kasugamycin, CGS 15943, and Ro 90-7501. After 24 h of treatment, bacteria were serially diluted and plated on media with or without nourseothricin. The pScreening plasmid encodes nourseothricin resistance. Therefore, the ratio of colony-forming unit counts with and without nourseothricin selection indicates the percentage of bacteria still retaining plasmid. The horizontal black line corresponds to the plasmid content in the initial bacteria inoculum prior to incubation. Therefore, reduction of plasmid to this level during compound treatment suggests complete block of plasmid replication, and reduction below this level suggests additional eviction of plasmid.

As a tolC mutant was utilized in the initial high-throughput screen and confirmatory assays to increase compound permeability, we tested the effect of the screening hits in wild-type E. coli K-12. Kasugamycin demonstrated antiplasmid activity, as indicated by RFU reduction, although activity was less pronounced than in the tolC background (SI Appendix, Fig. S1 A and B). In contrast, CGS 15943 and Ro 90-7501 showed no antiplasmid activity, suggesting that these compounds are not able to efficiently penetrate the gram-negative outer membrane barrier (SI Appendix, Fig. S1 D and F).

Ro 90-7501 Has Antiplasmid Activity Distinct from Other DNA Intercalators.

As some DNA intercalators are known to cure plasmids (5), we next tested the ability of the compounds to intercalate into DNA. To do so, we employed a previously described acridine orange intercalator displacement assay (31). Acridine orange fluoresces strongly when bound to DNA. If a compound has DNA intercalating activity, it will displace acridine orange resulting in a reduced acridine orange fluorescent signal. We found that Ro 90-7501 reduced acridine orange fluorescence in this assay to a greater extent than mitoxantrone, a known intercalator control (32), suggesting that Ro 90-7501 is also a DNA intercalator (Fig. 5A). The absence of acridine orange displacement for other compounds suggested that DNA intercalation does not account for their antiplasmid effects.

Fig. 5.

Ro 90-7501 is a DNA intercalator with unique activity. (A) Acridine orange displacement assay. The DNA intercalator acridine orange was added to wells containing DNA and doubling dilutions of compounds were subsequently added. After incubation, the RFU of acridine orange bound to DNA was determined. Data points are normalized to negative controls. (B–D) Comparison of activity of Ro 90-7501 and known DNA intercalators on pScreening. After 24 h, the (B) RFU, (C) A₆₀₀, and (D) percentage plasmid retention by qPCR were determined as described earlier. (E) After 24 h of treatment with intercalators, screening strain cultures were plated on media with and without nourseothricin as described in Fig. 4 to quantify percent plasmid retention. (F and G) The DNA intercalator acridine orange was added to wells containing either pScreening DNA or tolC genomic DNA and doubling dilutions of (F) Ro 90-7501 or (G) EtBr as indicated. After incubation, the RFU of acridine orange bound to DNA was determined. Data points are normalized to negative controls. ***P < 0.0001.

As Ro 90-7501 treatment resulted in almost complete plasmid eviction (Figs. 2–4), we examined whether such extreme antiplasmid activity is a common feature of other DNA intercalators. We compared the antiplasmid activity of Ro 90-7501 to a panel of known antiplasmid DNA intercalators including ethidium bromide, acridine orange, and 9-aminoacridine. Based on RFU values, all comparators produced a dose-dependent reduction in plasmid levels (Fig. 5B). However, all also showed significant antibacterial effects at concentrations ranging from 5 µg/mL to 80 µg/mL (Fig. 5C), that is, in the same concentration range as their antiplasmid effects. In contrast, Ro 90-7501 demonstrated only minimal antibacterial effects at 80 µg/mL, substantially above the 20 to 40 µg/mL associated with maximal plasmid eviction, indicating a considerably higher antiplasmid selectivity.

We then quantified plasmid abundance after intercalator treatment by qPCR and by plating on nourseothricin-selective media as described above (Fig. 5 D and E). Although ethidium bromide, acridine orange, and 9-aminoacride all demonstrated antiplasmid activity by these measures, these intercalators were not so effective as Ro 90-7501 (Fig. 5 D and E) and did not reduce plasmid levels below quantities present in the starting inoculum. These data indicate that Ro-90-7501 is unique among DNA intercalators examined in both potency and selectivity.

One potential explanation for the unique antiplasmid activity is that Ro 90-7501 intercalates more selectively into plasmid DNA than other intercalators. To investigate this hypothesis, we performed an acridine orange displacement assay using either pScreening plasmid DNA or E. coli tolC genomic DNA as an intercalation substrate. We found that Ro 90-7501 was significantly more effective at displacing acridine orange from plasmid DNA than genomic DNA in contrast to ethidium bromide, whose ability to displace acridine orange from plasmid and genomic DNA was indistinguishable (Fig. 5 F and G). Potential reasons for the selectivity of Ro 90-7501 for plasmid DNA are not clear but may relate to preferred DNA binding under supercoiling torsion. Interestingly, Ro 90-7501 was previously shown to inhibit λ phage propagation (33). This activity may be partially explained by observations of Ro 90-7501 treatment leading to forced genome ejection from λ phage particles, leaving empty phage heads (34). Although the genome ejection mechanism is not understood, this forced ejection is reminiscent of Ro 90-7501’s ability to evict pScreening plasmid from E. coli.

Ro 90-7501 was also previously found to have no toxic effects on mammalian cells at over 700 µg/mL (28), a concentration that is over 30 times higher than the concentration where Ro 90-7501 exerts maximal antiplasmid effect. Therefore, although our initial supposition was that all types of intercalators should a priori be excluded as antiplasmid therapeutic candidates, it is possible that some of these compounds may still have compelling, selective properties. Precedent for this can be found in commonly used, well-tolerated, nonchemotherapeutic intercalators such as riboflavin (35).

Identification and Characterization of CGS 15943-Resistant Mutation.

Since CGS 15943 was potent, selective, and lacking potentially undesirable DNA intercalation activity, we sought to identify its mechanism of action. To do so, we passaged the screening strain in 10 μg/mL CGS 15943 while simultaneously selecting for maintenance of pScreening with nourseothricin. After four sequential passages we isolated a single clone (CGS^R) resistant to antiplasmid effects of 20 μg/mL CGS 15943, the highest concentration tested (see Fig. 7A). In contrast, CGS^R remained completely susceptible to antiplasmid effects of both Ro 90-7501 and kasugamycin (Fig. 6 B and C).

Fig. 7.

repE levels are altered by CGS 15943 and aminoglycosides. (A) β-Galactosidase activity of the lacZ fusions was assayed to characterize repE regulatory circuits. Two types of fusions were analyzed: transcriptional fusions with expression of the lacZ gene driven by wild-type repE or CGS^R repE promoters and translational fusions of repE with lacZ genes driven by wild-type or CGS^R repE promoters. Miller units were determined after 4 h of growth. (B) Effect of 4-h CGS 15943 treatment on lacZ fusion activity. β-Galactosidase activity was expressed as a percentage of activity in no treatment controls. (C) Effect of CGS 15943 on pScreening levels in the presence of a second plasmid exogenously expressing repE from a pBAD regulated promoter in the presence or absence of arabinose inducer or on pScreening with the monomeric mutant repE54 substituted for repE. DNA was isolated after 24 h of treatment and quantified by qPCR. Data were normalized to the chromosomal dxs gene and expressed as a percentage of pScreening levels in untreated controls. (D) Kasugamycin and (E) apramycin effects on lacZ fusion expression. (F) Effect of kasugamycin on pScreening levels in the presence of repE54 or exogenously expressed repE. (*P < 0.05, **P < 0.001, ***P < 0.0001; n.s., not statistically significant).

Fig. 6.

Characterization of CGS 15943-resistant mutant. The original screening strain (pScreening), the CGS 15943 resistant mutant (designated CGS^R), a treatment-naive tolC strain transformed with the CGS 15943 resistant plasmid (designated tolC+pCGS^R), CGS^R cured of its plasmid and transformed with a pScreening plasmid unexposed to CGS 15943 (designated CGS^R + pScreening), and a tolC strain transformed with a pScreening engineered to contain an insertion of an additional thymidine (7xT) in its promoter region identical to the mutation observed in pCGS^R (designated pScreening + 7xT) were treated with doubling dilutions of (A) CGS 15943, (B) Ro 90-7501, and (C) kasugamycin. Plasmid levels were determined, normalized to the single-copy chromosomal dxs gene, by qPCR and are expressed as a percentage of untreated controls. (D) The CGS 15943 resistance mutation was identified as an insertion of an extra thymidine in the poly-T tract (6xT -> 7xT) located between the predicted −35 and −10 regions of the repE promoter in pScreening, notated in red text in the figure. Through site-direct mutagenesis, a cytosine was also inserted at either end of the 6 poly-thymidine repeat as indicated. Then the effect of (E) CGS 15943 and (F) kasugamycin on levels of plasmid with these insertions compared with unmutated pScreening was determined by qPCR, normalized to the chromosomal dxs gene, and expressed as a percentage of plasmid level in untreated controls. (G) repE mRNA was quantified by qPCR after RNA isolation from bacterial cultures and reverse transcription. Values were normalized to the housekeeping gene dxs and expressed relative to levels of pScreening control strain. (H) Relative plasmid levels were quantified by qPCR normalized to chromosomal dxs and plotted relative to pScreening levels.

To test whether the resistance mutation was localized to pScreening or the chromosome, we isolated the pScreening plasmid (pCGS^R) from CGS^R and transferred it into a drug-naive tolC strain. In parallel, we cured the pCGS^R from the resistant isolate using Ro 90-7501 treatment and transformed a copy of drug-naive pScreening plasmid into this host strain background. Only the former strain containing pCGS^R was resistant to CGS 15943, indicating that the resistance phenotype tracks with the pCGS^R plasmid (Fig. 6A).

Therefore, in order to identify the resistance mutation, we sequenced pCGS^R. The only mutation present was a single base pair insertion in the promoter of the plasmid replication gene, repE (Fig. 6D). This insertion was located between the −35 and −10 promoter elements, increasing the length of an existing thymidine nucleotide repeat from six to seven thymidine residues (Fig. 6D). To confirm that this single base alteration was responsible for resistance, we recreated this mutation in pScreening using site-directed mutagenesis (pScreening-7xT) and again observed complete resistance to CGS 15943 (Fig. 6A).

Fortuitously, the pScreening plasmid replicon is essentially identical to that of the F plasmid, which was used as a tool in early genetic mapping of E. coli and therefore has been well studied (36). From this earlier investigation we know that RepE is a critical regulator of plasmid copy number. RepE binds to iteron repeats in the origin of replication where it cooperates with DnaA to melt the origin and initiate plasmid replication.

The sequence of the repE promoter region and the −35/−10 spacing of exactly 16 base pairs is highly conserved across very diverse IncFIA plasmids (by BLAST analysis), in contrast to the generally predominant 17-base pair spacing in E. coli promoters (37). Therefore, we next investigated whether it was the specific mutation in the promoter of repE or the increased spacing between −35 and −10 elements that was responsible for CGS 15943 resistance. We created two additional mutations in the repE promoter: one with a cytosine inserted on the −35 side of the stretch of six thymidines and one with a cytosine inserted on the −10 side (Fig. 6D). Both conferred complete resistance to CGS 15943 (Fig. 6E), suggesting that increased promoter spacing was responsible for resistance. As expected, neither mutation affected the antiplasmid activity of kasugamycin (Fig. 6F). In summary, it appears that expansion of the distance between the −35 and −10 promoter sequences is critical to CGS 15943 resistance, rather than the specific base and site of insertion, at least based on the limited number of insertions that were tested (Fig. 6). These observations further suggest that RepE is either the direct or indirect target of CGS 15943 and that conserved promoter spacing is integral to its effects.

CGS 15943 Mechanism of Action.

As the mutations conferring resistance were located in the repE promoter, we tested whether the mutations altered repE transcription. We first isolated RNA from the screening strain and CGS 15943-resistant strains and quantified repE mRNA by RT-qPCR. An approximately fourfold increase in repE transcript in CGS 15943-resistant strains was observed (Fig. 6G). There was also a corresponding ∼50% increase in plasmid levels (Fig. 6H).

To examine effects of CGS 15943 on repE levels, we generated two sets of β-galactosidase reporter fusions: a transcriptional fusion of lacZ directly under control of the repE promoter and a translational fusion of the entire repE promoter and gene to lacZ. All constructs were driven by either wild-type or 7xT repE promoters and cloned into a pUC19 plasmid backbone. pUC19 is resistant to the antiplasmid activity of CGS 15943 (SI Appendix, Fig. S2), allowing us to examine the effects of CGS 15943 without loss of the plasmid constructs. In both types of lacZ fusions, the mutated promoter increased β-galactosidase activity two- to threefold (Fig. 7A), consistent with RT-qPCR observations (Fig. 6G).

CGS 15943 treatment significantly reduced β-galactosidase activity of the wild-type repE promoter transcriptional fusion by ∼33% (Fig. 7B). In contrast, the corresponding 7xT promoter fusion was unaffected. (Fig. 7B). Similarly, a reduction in the translational fusion expression driven by the wild-type repE promoter was also observed upon CGS 15943 treatment, although it did not reach statistical significance (P = 0.1). Again, the 7xT translational fusion was unaffected by CGS 15943 treatment. Taken together, these data initially suggested that CGS 15943 disrupted plasmid maintenance through reduced repE expression and that the 7xT mutation might negate this effect.

To investigate this possibility we placed repE under the control of the arabinose inducible pBAD promoter on a pBR322-based plasmid resistant to the antiplasmid effects of CGS 15943 (SI Appendix, Fig. S2). If the effects of CGS 15943 were solely due to repression of repE expression, then arabinose induction of exogenous repE expression should impart CGS 15943 resistance. However, the effect of exogenous repE expression was not consistent with the complete abrogation of CGS 15943 activity in the promoter mutants (Figs. 6A and 7C). Therefore, the simple model for CGS antiplasmid activity, that CGS 15943 decreases repE expression and the 7xT mutation restores this expression, could not fully account for our observations.

The monomer to dimer ratio of RepE is a key regulator of plasmid replication. Monomers bind to ori and melt the DNA in concert with DnaA to allow binding of the DNA replication complex. Dimers bind to the repE operator to inhibit its own transcription (38) and also participate in DNA looping between iteron repeats in ori and incC, which locks the DNA in a conformation in which replication cannot occur (39, 40). In a general sense, monomer predominance favors plasmid replication and dimer predominance the converse. We therefore considered whether alteration of the monomer:dimer equilibrium might account for the effects of CGS 15943 on plasmid replication. The RepE mutant, RepE54, exists only in monomeric form (41, 42). When repE54 was substituted for repE in pScreening, we observed a ∼1,000-fold reduction in CGS 15943 antiplasmid effects by qPCR; however, it is noteworthy that complete resistance to CGS 15943 comparable to the 7xT mutation was not conferred (Fig. 7C). These results suggest that CGS 15943 may not be acting solely through shifting equilibrium of RepE to the dimeric form.

Our results are consistent with several potential CGS 15943 antiplasmid mechanisms in the context of RepE regulatory circuits currently understood for the IncFIA replicon (diagrammed in SI Appendix, Fig. S3). The critical nature of internal repE promoter spacing for the effects of CGS 15943 is intriguing and suggests possibilities involving disruption of the alignment of replicon machinery, effects on open plasmid replication complex formation, effects on interactions that may be dependent on DNA twist, or effects on binding of yet unknown factors in the repE promoter region. Such possibilities will be the basis of future studies.

Kasugamycin Mechanism of Action.

Aminoglycosides such as apramycin have been shown previously to disrupt plasmid replication of an IncB incompatibility group plasmid by interfering with expression of the replication protein, RepA (28). The translation of RepA is controlled by an antisense RNA that binds to the 5′ untranslated region of the repA mRNA to stabilize a messenger RNA (mRNA) secondary structure that blocks access of the ribosome to its Shine–Dalgarno sequence. Apramycin is presumptively a natural mimetic of this inhibitory RNA based on its site of mRNA binding and activity. In contrast, repE expression is thought to be primarily regulated through binding of a RepE dimer to an operator sequence in its own promoter. To determine if kasugamycin blocks expression of the functionally analogous RepE, we tested the effect of kasugamycin on the repE-lacZ transcriptional and translation reporter fusions described above. Importantly, kasugamycin treatment significantly reduced the β-galactosidase activity of the repE translational fusion but had no effect on the repE transcriptional fusion (Fig. 7D). In contrast to observations for CGS 15943, the 7xT mutation did not inhibit kasugamycin’s suppression of repE translation, and exogenous expression of repE from an arabinose-inducible, pBAD promoter dramatically reduced kasugamycin antiplasmid activity (Fig. 7 D and F). Additionally, substitution of the monomeric repE54 for repE in pScreening did not reduce kasugamycin activity. Therefore, kasugamycin’s antiplasmid activity appears to result from inhibition of repE translation.

We also examined the effect of apramycin on repE expression. Similar to kasugamycin, apramycin significantly reduced repE translation but not transcription, and activity was unaffected by the 7xT resistance mutation (Fig. 7E). Therefore, the antiplasmid effects of aminoglycosides on IncFIA and IncB plasmids are analogous (Fig. 7), and the antiplasmid effects of aminoglycosides and CGS 15943 on IncFIA are distinct.

Compounds Potentiate Meropenem Activity.

We next tested whether the compounds have antiplasmid activity against the parent pCRE plasmid from BIDMC20A. To perform these experiments, we transformed the ∼140-kb pCRE BIDMC20a plasmid into the E. coli tolC strain, treated that strain with antiplasmid compounds, quantified plasmid levels by qPCR, and normalized data to bacterial chromosomal DNA as described above. Unexpectedly, however, none of the compounds decreased the relative plasmid levels of pCRE BIDMC20a normalized to the chromosomal dxs gene, as determined by qPCR, with the exception of a 20% reduction in relative pCRE BIDMC20a levels observed for kasugamycin (SI Appendix, Figs. S4 and S8A). However, for CGS 15943 and Ro 90-7501, absolute levels of plasmid and genome copies were dramatically reduced (SI Appendix, Fig. S4B), suggesting a potential link between antiplasmid activity and loss of viability of host cells containing pCRE BIDMC20a.

It is noteworthy that pCRE BIDMC20a, but not pScreening, contains the CcdAB plasmid addiction system. Therefore, based on our findings, we further considered whether CcdAB-mediated host cell death might account at least in part for the difference in observations between pScreening (relative plasmid loss) and native pCRE BIDMC20a plasmid (absolute plasmid loss). Notably, by colony-forming unit analysis, treatment with CGS 15943 and Ro 90-7501 significantly reduced survival of bacteria containing pCRE BIDMC20a to the extent that no viable bacteria were recovered during treatment with higher concentrations of Ro 90-7501 (Fig. 8B), but not survival of bacteria containing pScreening which lacks the ccdAB genes (Fig. 2). Therefore, antibacterial properties of CGS 15943 and Ro 90-7501 were specific to pCRE BIDMC20a and likely resulted from the toxic effects of retained, stable CcdB toxin in the absence of unstable CcdA antitoxin during compound-induced plasmid curing. Therefore, removal of the CcdAB plasmid addiction system from pScreening proved prescient, as we were able to identify screening hits that would have otherwise been misclassified as antibacterial compounds and not further studied.

Fig. 8.

Compounds exhibit synergy with meropenem. (A) Levels of pScreening and pCRE BIDMC20a were quantified by qPCR after kasugamycin treatment, normalized to chromosomal dxs, and expressed as a percentage of untreated controls. (B) pCRE BIDMC20a containing tolC bacteria were treated with indicated concentrations of CGS 15943 or Ro 90-7501 for 24 h and plated for viability on solid media without antibiotics. Data are expressed as a percentage of viable colonies in untreated controls. # indicates no viable bacteria detected (limit of detection 15 cfu per well). In pCRE BIDMC20a-containing bacteria, (C) kasugamycin, (D) Ro 90-7501, and (E) CGS 15943 significantly potentiated activity of meropenem up to 32-fold. In these assays, an A₆₀₀ of 0.05 is equivalent to the reference MIC, that is, complete visible inhibition of bacterial growth (49), based on our prior validation studies (33), and corresponds to media control without bacteria. (*P < 0.05, **P < 0.001, ***P < 0.0001).

The pCRE from BIDMC20A encodes a Klebsiella pneumoniae carbapenemase (KPC) gene conferring resistance to meropenem. Therefore, we predicted that the compounds might potentiate the activity of meropenem due to their antiplasmid activity. Notably, kasugamycin, Ro 90-7501, and CGS 15943 all potentiated meropenem activity against a tolC strain containing pCRE BIDMC20a, decreasing the minimal inhibitory concentration (MIC) by 2- to 32-fold depending on the concentration of inhibitor tested (Fig. 8 C–E). The current understanding of carbapenem therapeutic efficacy, codified in rules used in clinical laboratories to interpret isolates as carbapenem susceptible or resistant for clinicians (43), is that the MIC alone should be used to predict therapeutic efficacy without regard to underlying strain genotype. The clinical susceptibility breakpoint is a meropenem MIC ≤1 μg/mL, and several concentrations of each active compound were able to push the meropenem MIC to or below this threshold, albeit in a tolC background. Therefore, our results demonstrate that an antiplasmid compound can render a CRE strain susceptible to carbapenems and thereby restore activity of a critical, drug-of-last-resort therapeutic.

Compounds Show Differential Activity against IncFII.

As pCRE plasmids often contain multiple incompatibility groups, we sequenced pCRE BIDMC20a using PacBio technology [GenBank accession no. MW057772 (50)] to definitively link together plasmid sequence present on multiple contigs in previously available Illumina sequence data (GenBank accession no. GCA_000522145.1). PacBio sequencing allowed full assembly of the 142.8-kb pCRE BIDMC 20a plasmid and identified the presence of a second plasmid replicon from incompatibility group IncFII. To test whether the IncFII replicon was functional, we substituted this putative IncFII replication machinery for the IncFIA replicon in pScreening to create pIncFII. Interestingly, pIncFII replicated in a stable fashion with a plasmid copy number approximately four times higher than pScreening (IncFIA) and pCRE BIDMC 20a. (Fig. 9A). Therefore, as compounds were active against pCRE BIDMC 20a containing both replicons, we considered whether they would also inhibit pIncFII. We found that kasugamycin demonstrated equivalent antiplasmid activity against pIncFII and pScreening (IncFIA) by qPCR (Fig. 9 B–D). However, CGS 15943 and Ro 90-7501 only reduced pIncFII plasmid levels by a modest 25% at most compared with their reduction of pScreening (IncFIA) levels below the limit of detection (Fig. 4). Similar results were observed when plating treated bacteria on nourseothricin to quantify the number of bacteria maintaining nourseothricin resistance-expressing pIncFII (Fig. 9E).

Fig. 9.

Compounds have differing activity against IncFII. (A) The relative abundance of pCRE BIDMC20a, IncFIA (pScreening), and IncFII plasmids in a tolC strain was determined by qPCR and is shown normalized to pCRE BIDMC20a copy number. After treatment with (B) kasugamycin, (C) CGS 15943, or (D) Ro 90-7501 for 24 h, plasmids levels were quantified by qPCR relative to the chromosomal dxs gene and are expressed as a percentage of plasmid level in untreated controls. (E) The tolC strain containing the IncFII plasmid was treated with doubling dilutions of kasugamycin, CGS 15943, and Ro 90-7501. After 24 h of treatment, bacteria were serially diluted and plated on media with or without nourseothricin. The IncFII plasmid encodes nourseothricin resistance. Therefore, the ratio of colony-forming unit counts with and without nourseothricin selection indicates the percentage of bacteria still retaining plasmid.

Based on these results, it is striking that pCRE BIDMC20a was susceptible to CGS 15943 and Ro 90-7501, as these compounds only have minimal activity against the second IncFII replicon on pCRE BIDMC20a. IncFII replicons are regulated by an antisense transcriptional control of the RepA replication protein rather than by iteron-based mechanisms found in IncFIA. Our results suggest that the mechanistically distinct IncFIA replicon is dominant over IncFII in pCRE BIDMC20a, as both pCRE BIDMC20a’s plasmid copy number and susceptibility to CGS 15943 and Ro 90-7501 are reflected in effects on the IncFIA replicon individually. Taken together, our data suggest that compounds that specifically target the IncFIA incompatibility group can successfully inhibit replication of a large carbapenemase-encoding plasmid, despite the presence of an additional, distinct, functional replicon.

Conclusions

Carbapenems are a last-resort antibiotic that provides safe, effective therapy against most otherwise β-lactam–resistant, gram-negative pathogens. Unfortunately, carbapenemase genes, carried on large plasmids, hydrolyze these antibiotics, rendering them ineffective. Therefore, we hypothesized that small-molecule–induced eviction of these plasmids would restore susceptibility to carbapenems. As many additional resistance genes are carried on these plasmids, the strategy presumably would restore susceptibility to many other antibiotics as well. We therefore designed a high-throughput screening assay to identify antiplasmid agents that did not presuppose a particular mechanism of action. For proof of principle, we screened known bioactive compounds that might more quickly advance our understanding of antiplasmid mechanism and would be immediately available in sufficient quantities for follow-up studies. We discuss validation of the antiplasmid activity of three identified compounds, all of which were active against the native pCRE BIDMC20a plasmid. Furthermore, for two of these compounds we identified the plasmid replication protein, RepE, as the putative target. Our work also demonstrated that by targeting pCRE BIDMC 20a, carbapenem susceptibility can be restored. Although the antiplasmid compounds identified in this study have been validated to work on an IncFIA replicon, further study will be needed to determine if these compounds are active against other common incompatibility groups. Importantly, the ability to inhibit plasmid maintenance in a clinical setting would allow us to treat previously resistant CRE infections. In addition to more direct clinical implications, antiplasmid compounds will also provide powerful tools to learn more about the basic biology of large, low-copy, multidrug-resistance plasmids and thereby identify additional therapeutic targets in the future. Our work demonstrates the feasibility of a high-throughput screening approach to identify antiplasmid agents with specific mechanisms of action.

Materials and Methods

Strain Backgrounds.

In this study we utilized the E. coli CRE clinical isolate BIDMC 20A, the sequence of which is available in multiple contigs (GenBank accession no. GCA_000522145.1), the E.coli K12 strain BW25113, and a tolC mutant, JW5503-KanS. The latter two were from the Keio collection (44) and obtained from the Coli Genetic Stock Center (Yale University, New Haven, CT). BIDMC20A was propagated on blood agar plates; all other strains were cultured on lysogeny broth (LB) agar or in LB media (BD). Where indicated, nourseothricin (GoldBio) at 50 µg/mL was used for selection.

Construction of the Incompatibility Group Plasmids.

Primers 1 and 2 were used to amplify a previously described ProD driven mNeptune2 reporter gene linked in an operon format with a nourseothricin resistance cassette (16). Primers also contained sequences homologous to pCRE of BIDMC20A in order to facilitate downstream cloning steps. The replication and maintenance machinery of pCRE BIDMC20a was amplified in two PCR fragments: one was amplified with primer 3 and 4 and the other with primers 5 and 6. The two pCRE fragments along with the mNeptune2 fragment were assembled using the HiFi system (New England Biolabs). To remove the toxin CcdB from this construct, the plasmid was again amplified in two fragments, one with primers 7 and 8 and the other with primers 9 and 10, and the two fragments were reassembled using the HiFi system. The final pScreening plasmid was confirmed to be error-free by sequencing. Primer sequences are shown in Dataset S3. Once constructed, pScreening was transformed into the E. coli K12 tolC mutant by electroporation.

The IncFII plasmid was amplified from pCRE BIDMC using primers F IncFII and R IncFII (Dataset S3). The IncFII fragment was assembled with the mNeptune2 fragment using the HiFi system. The final plasmid was confirmed to be error-free by sequencing, and IncFII was transformed into the E. coli K12 tolC mutant by electroporation.

High-Throughput Screen.

Cultures were started in LB media containing 50 µg/mL nourseothricin the afternoon prior to screening. The next morning cultures were diluted 10-fold with fresh medium and allowed to grow for 2 to 3 h until reaching midlog phase. Bacteria were then seeded into white, clear-bottomed, 384-well plates (Grenier Bio-One) at 10⁴ colony-forming units (cfu) per well in 30 μL of LB media without antibiotic utilizing a Multidrop Combi Reagent Dispenser (Thermo Fisher Scientific). Plates were then centrifuged at 150 × g for 1 s, and 300 nL of each test compound was added to screening wells in duplicate screening plates using a pin transfer robot. Plates were incubated for 48 h at 37 °C. Twenty-four positive and negative controls wells were included per microplate to monitor Z′ on an ongoing basis and ensure continued robustness of assay during screening. Z′ was determined as described previously (23). Positive control wells contained 20 μg/mL of meropenem (to prevent bacterial growth and plasmid replication) (Ark Pharm). Negative control wells were not seeded with compound.

A₆₀₀ and RFU of the mNeptune2 reporter (excitation 600 nm and emission 650 nm) were then quantified using an EnVision Plate Reader (PerkinElmer). Compounds were scored as antibacterial if they reduced the A₆₀₀ in both replicate wells by more than 80% compared to the median of all treatment wells. If not antibacterial, compounds were scored as antiplasmid hits if the reduction in RFU was greater than two times the reduction in A₆₀₀ in both replicate wells compared to the median RFU and A₆₀₀ of all treatment wells.

Antiplasmid Hit Confirmation.

Bacteria at midlog phase were seeded at 10⁴ cfu per well in 50 μL of LB media in 384-well plates. Doubling dilutions of compounds were dispensed into the 384-well plates using a D300 Digital Dispensing System (TECAN). Compounds tested include kasugamycin (Sigma-Aldrich), CGS 15943 (Alfa Aesar), Ro 90-7501 (Tocris Bioscience), and apramycin (Alfa Aesar). Total dimethyl sulfoxide (DMSO) concentration in the wells was kept below 1% (43). Plates were incubated at 37 °C for 24 h. Although assays were evaluated after a 48-h incubation during the high-throughput screen, for convenience a 24-h incubation was used for all confirmatory and exploratory investigation, as all hits showed essentially equivalent readouts at both time points. Bacterial growth and RFU were quantified as described above except measurements were made using a TECAN M1000 plate reader. Percent growth and plasmid abundance was normalized to no treatment controls. Antiplasmid activity of known DNA intercalators including ethidium bromide (Life Technologies), 9-aminoacridine (TCI America), and acridine orange (Sigma-Aldrich) was quantified in an identical manner to hit compounds in this assay and additional assays described below.

Real-Time qPCR.

Midlog phase cultures were seeded into 384-well plates in 50 μL of media containing doubling dilutions of compounds at a concentration of 10⁴ cfu per well. After 24 h incubation, bacteria were pelleted, resuspended in 50 mM NaOH, and heated at 98 °C for 10 min to lyse the bacterial cells. Then, 0.5 M Tris⋅HCl was added to neutralize the sample, and DNA was diluted 1:5 with deionized water to a concentration suitable for downstream applications. To quantify pScreening and pCRE BIDMC20a, we utilized primers that bind to the partitioning factor gene, sopA. For pUC19 and pBR322 plasmids quantified in this study, we utilized primers that bind to ampR. Real-time qPCR was completed using 3 μL of DNA template in triplicate technical replicates using DyNAmo HS SYBR Green qPCR Kit (Thermo Fisher). Transcripts were normalized to the single-copy, chromosomal gene, dxs. Using qPCR we are able to detect a decrease in plasmid levels up to 99.95%. Primer sequences were as follows: sopA (GGA​CTG​AGA​GCC​ATT​ACT​ATT​GCT​GT and GAC​TAC​ACC​TCC​GCA​CTG​C), ampR (GAC​AGT​AAG​AGA​ATT​ATG​CAG​TGC​TGC and GGC​TTC​ATT​CAG​CTC​CGG​TTC​C), IncFII (GGT​CTT​TTT​CTG​CAT​CAC​TGG​GC) and (GCG​TAA​CGA​CCT​ATG​AGG​ACG), and dxs (CGC​TTC​ATC​AAG​CGG​TTT​CAC and GCG​AGA​AAC​TGG​CGA​TCC​TTA​AC).

Plasmid Quantification by Colony-Forming Unit Enumeration.

Bacteria at midlog phase were seeded as above into 384-well plates containing compounds at indicated concentrations. After a 24-h incubation, bacteria were serially diluted, plated in 10-μL drops onto plates with or without 50 μg/mL nourseothricin and enumerated by the drop plate method (45). Data were normalized to growth on plates without antibiotic.

Acridine Orange Displacement Assay.

The acridine orange displacement assay was conducted as previously described (31) in 384-well plate format. Briefly, the DNA substrate (salmon sperm DNA, pScreening plasmid DNA, or tolC genomic DNA) was diluted to 6 μg/mL in HENS buffer (10 mM Hepes, pH 7.5, 1 mM EDTA, and 100 mM NaCl), and 30 μL of the DNA solution was added to each well with acridine orange (Sigma-Aldrich) at 50 nM final concentration. Compounds were added in doubling series with mitoxantrone (Selleck Chemical) used as a positive intercalator control. Plates were incubated at room temperature for 20 min, and acridine orange fluorescence was read on a TECAN M1000 plate reader (excitation 480 nm, emission 535 nm). Data were normalized to control wells without test compound.

Meropenem Synergy Assay.

Synergy with meropenem (Ark Pharm) was examined using a checkerboard array. Bacteria at midlog phase were seeded at 10⁴ cfu per well in 50 μL of LB media in 384-well plates. Serial twofold dilutions of meropenem and test compounds were dispensed into wells in orthogonal titrations using the D300 Digital Dispensing System. Plates were incubated for 24 h and combinatorial MICs determined based on A₆₀₀ measure using the TECAN M1000 as previously described (46, 47).

Isolation of CGS 15943-Resistant Strain.

To isolate CGS 15943-resistant mutants, the screening strain was cultured in 10 μg/mL CGS 15943 and 50 µg/mL nourseothricin. Every 24 h for 4 d the culture was diluted to an A₆₀₀ of 0.5 and allowed to regrow at 37° for 24 h. The culture was then plated on LB agar medium with 50 µg/mL nourseothricin. Over 150 isolated colonies were screened for resistance to pScreening plasmid eviction by CGS 15943 and retained sensitivity to plasmid eviction by Ro 90-7501, based on mNeptune2 fluorescence. One such isolate was identified: CGS^R. The plasmid was isolated from this strain and sequenced to identify the resistance mutation (pCGS^R). To cure the plasmid from the resistant isolate, the strain was cultured overnight in 20 µg/mL Ro 90-7501 and then plated without antibiotic selection. Colonies were tested on plates with and without nourseothricin to identify candidates cured of pScreening.

Site-Directed Mutagenesis.

Mutations were made in the promoter region of repE on pScreening using overlap extension PCR site-directed mutagenesis. Site directed mutagenesis was also utilized to generate pScreening-repE54 using pScreening as a template. Primer sequences are provided in Dataset S4. After PCR, DpnI (New England Biolabs) was added to degrade unmutated, methylated template prior to electroporation to isolate pScreening mutant candidates. Mutations were confirmed by Sanger sequencing.

Arabinose-Inducible repE.

To create a pBAD-regulated repE, repE was amplified from pScreening and joined to the linearized pBAD/HisB vector using a HiFi cloning kit (New England Biolabs). Primers are listed in Dataset S5. The pBAD-RepE plasmids were transformed into a tolC mutant strain containing pScreening. RepE transcription was induced using 0.001% arabinose for 4 h before adding varying concentrations of CGS 15943 or kasugamycin. DNA was also isolated and plasmid levels were quantified by qPCR as described above.

RNA Isolation.

Triplicate cultures were grown overnight in 50 μg/mL nourseothricin. The next day cultures were diluted 1:200 and grown at 37 °C until reaching an A₆₀₀ of 0.6. RNA was extracted from the samples using Direct-zol RNA Miniprep Plus (Zymo Research). Complementary DNA (cDNA) was synthesized using the iScript cDNA Synthesis Kit (Bio-Rad). Real-time qPCR was completed using 25 ng of cDNA template in triplicate technical replicates using the DyNAmo SYBR Green qPCR kit (Thermo Fisher). Transcripts were normalized to the housekeeping gene dxs (primer sequences as above). The sequences of the repE primers were GAG​CCA​TCC​GGC​TTA​CGA​TAC and CCC​ATA​TCT​CAT​TCC​CTT​CTT​TAT​CGG​GTT​A.

LacZ Fusions and β-Galactosidase Assay.

To create LacZ fusions, lacZ was amplified from MG1655 and joined to a fragment containing either the promoter of repE or the promoter and repE gene amplified from pScreening or pCGS^R. In the lacZ translational fusions, lacZ was fused to the C terminus of repE. The respective fragments were assembled with the pUC19 vector using HiFi (NEB). Primer sequences are provided in Dataset S6.

To quantify β-galactosidase activity of indicated strain constructs, triplicate cultures of relevant strains were grown overnight in 100 μg/mL ampicillin. The next day the cultures were diluted 1:200 and grown with 20 μg/mL CGS 15943, 6 μg/mL kasugamycin, 2 μg/mL apramycin, or DMSO control for 4 h, at which time the A₆₀₀ of the cultures was determined, and β-galactosidase activity was assayed using the single-step method described by Schaefer et al. (48). Briefly, 80 μL of each sample and 120 μL of β-galactosidase mix were transferred to a 96-well plate. The optical density at 420 nm was read every minute for 1 h, and Miller units were quantified.

pCRE BIDMC 20a Plasmid Sequencing.

pCRE BIDMC20a was isolated using a Qiagen Large Construct Kit. The plasmid was sequenced by PacBio long read sequencing at Genewiz. The sequence was deposited at GenBank and assigned the accession number MW057772 (50).

Statistical Analysis.

All experiments were conducted with at least three biological replicates and repeated at least three independent times. Statistical comparisons were performed using Prism 8 for macOS (Graphpad Software, LLC) using Student's t test or ANOVA with a Holm–Sidak post hoc test to correct for multiple comparisons. A P < 0.05 was considered significant.

Data Availability.

All study data are included in the paper, SI Appendix, Datasets S1–S6, and GenBank accession number MW057772.