Strategies to combat antimicrobial resistance: anti-plasmid and plasmid curing

1One of the major threats facing society is the rise in number of antimicrobial-resistant (AMR) bacteria (O’Neill ). Antimicrobials underpin modern medicine; they are used to treat infections, to prevent infections (prophylaxis) during medical procedures (e.g. surgery) and they are crucial for patients with compromised immune function (Holmes et al.; Laxminarayan et al.). Between 2000 and 2010, global human use of antibiotics increased by 36%, and the use of two last-resort antibiotics, carbapenems and polymyxins, increased by 45% and 13%, respectively (Van Boeckel et al.). Antimicrobials have many non-human uses including in animals for growth promotion, veterinary treatment and aquaculture (Cabello; Meek, Vyas and Piddock; Van Boeckel et al.). In 2013, an estimated 131 109 tons of antimicrobials were used globally in food animals; by 2030 this is expected to increase to 200 235 tons (Van Boeckel et al.). However, there is a growing trend to improve antimicrobial stewardship in many countries. For example, in Switzerland veterinary antimicrobial sales increased between 2006 and 2008, but then steadily decreased, reaching a 26.2% reduction in 2013 (Carmo et al.). In addition to human and animal use, many cleaning and personal hygiene products contain biocides, such as triclosan, which can select for mutants resistant to biocides, and in some cases to antibiotics used in medicine (Meek, Vyas and Piddock; Webber et al., ).

2A key factor that has led to the rise and global dissemination of multidrug-resistant (MDR) bacteria are mobile antimicrobial resistance genes (ARGs). These are frequently located on plasmids, which are pieces of usually circular, self-replicating DNA which can code for a variety of different functional gene groups. Aspects of plasmid biology have been extensively reviewed elsewhere, but, in brief, plasmids often include partitioning systems, toxin–antitoxin (TA) systems and conjugative/transmission systems (Van Melderen and Saavedra De Bast; Pinto, Pappas and Winans; Carattoli; Baxter and Funnell; Goessweiner-Mohr et al.; Kado; MacLean and San Millan; Ruiz-Maso et al.; Cabezon et al.; Ilangovan, Connery and Waksman; Chan, Espinosa and Yeo; Banuelos-Vazquez, Torres Tejerizo and Brom; Hall et al.; Hulter et al.). Conjugation is mediated by type IV secretion coupled with a relaxosome complex to mediate DNA movement from one cell to another (Ilangovan, Connery and Waksman ).

3Plasmids are frequently categorised based on incompatibility groups (Inc), defined as the inability of two related plasmids to be propagated stably in the same cell and may be due to competition for the same replication or segregation sites, or caused by repression of replication initiation (Novick; Carattoli ). Reviews on incompatibility groups and plasmid classification can be found elsewhere (Novick; Carattoli; Shintani, Sanchez and Kimbara; Orlek et al.). Plasmids that share the same mechanisms for replication or partitioning are placed in the same incompatibility groups. Plasmid incompatibility has been used to follow the movement and evolution of plasmids conferring AMR (Carattoli et al.).

4ARGs that pose a serious threat to human medicine are typically found in Gram-negative bacteria. These include genes coding for extended spectrum β-lactamases (ESBL) (e.g. CTX-M), carbapenemases (e.g. KPC, NDM and OXA-58) (Holmes et al.) and colistin resistance (e.g. MCR-1) (Liu et al.). The issues surrounding AMR plasmids are derived in part by their substantial complexity. Plasmids often display a high degree of plasticity, with frequent insertions, deletions and rearrangements of DNA including changes to specific ARGs (Kado ). For example, the blaCTX-M gene is highly variable, and the CTX-M family of ESBLs are commonly coded for by multiple different plasmids, such as pCT (Fig. 1A) (Cottell et al., Bevan, Jones and Hawkey ). According to the Beta-Lactamase DataBase, 207 variants of blaCTX-M have been identified (accessed on 11 May 2018) (Naas et al.). Another example of a plasmid-mediated ARG is the mcr-1 gene, first identified on a transmissible plasmid, pHNSHP45, in 2016 (Fig. 1b) (Liu et al.). Since then mcr-1 and variants of this gene have been identified on multiple plasmid backbones and host strains. Of concern are isolates carrying colistin and carbapenem ARGs, as few treatment options would remain for infections caused by such bacteria (Lai et al.; Wang et al.; Zhou et al.). In addition to these examples, plasmids can carry a variety of other resistance genes, including qnr variants, aac(6΄)-lb-cr and plasmid-mediated efflux pump genes such as oqxAB and qepA, which confer low levels of resistance to quinolone antimicrobials (Jacoby, Strahilevitz and Hooper ). Increasingly, research should focus on ARGs which are frequently mobilised and transmit between bacteria (Crofts, Gasparrini and Dantas ).

5In the European Union, resistance to carbapenem antibiotics in invasive Klebsiella pneumoniae isolates ranges from 66.9% (Greece), 33.9% (Italy), 2.1% (Spain) to <5% (Northern Europe) (ECDC ). For invasive E. coli infections, resistance to third-generation cephalosporins ranges from 5% in Iceland to 50% in Italy, Slovakia and Bulgaria, while carbapenem resistance in E. coli is <1% for most of the EU and between 1–5% for Romania (ECDC ). A study of travellers returning to the Netherlands found 30.5% of participants had ESBLs in their bacterial flora, while only 8.6% had ESBLs before their trip (Paltansing et al.). A large prospective study of 2001 Dutch travellers found 34.7% with no ESBL producing Enterobacteriaceae prior to international travel returned with ESBL producing strains (Arcilla et al.). A similar study of 188 Swedish travellers found 32% returned from regions associated with high levels of ESBL producing Enterobacteriaceae carrying these antibiotic-resistant bacteria (Vading et al.). One isolate contained both blaCTX-M and mcr-1 (Vading et al.). Indeed, mcr-1 was detected by metagenomics in 4.9% of faecal samples from 122 healthy Dutch travellers upon return from travel to South/East Asia and/or Southern Africa undertaken between 2011 and 2012 (von Wintersdorff et al.). However, in this study little is known about the index isolate in which the mcr-1 gene originated, including the isolate's susceptibility profiles. Therefore, it is possible that the isolates were susceptible to other antimicrobials. In the majority of studies travellers who obtained ESBL-producing bacteria eventually lost the ESBL genes upon return. Of 15 Swiss volunteers, 3 were colonised by ESBL-resistant Enterobacteriaceae before their trip, all were colonised upon return and 6 were still colonised 6 months post-travel (Pires et al.). Of the resistant isolates 80% contained IncF family plasmids, and in some of the participants who were colonised 6 months after travel, the plasmids had moved into new host bacteria (Pires et al.). blaCTX-M-15 was the most prevalent ESBL, comprising 92% of the ESBL producers immediately after travel (Pires et al.). Together, this highlights the need to reduce the prevalence of ARGs on a global scale.

6Plasmid curing is the process by which plasmids are removed from bacterial populations. This is an attractive strategy to combat AMR as it has the potential to remove ARGs from a population while leaving the bacterial community intact. This means, for example, that the structure of the gastrointestinal microbiome of a chicken treated with a plasmid curing agent might remain largely unchanged, but potentially pathogenic bacteria which may unfortunately be transmitted into the food chain would be susceptible to antibiotics. Alternatively, a plasmid curing agent could be given to a patient prior to surgery, to reduce the likelihood of a resistant hospital acquired infection. Plasmid curing agents could also be taken by international travellers to reduce the global spread of AMR. Unfortunately, at the moment no such treatment options are in use. In fact, there are very few curing mechanisms that have been tested in vivo, even in experimental models. Therefore, research in this area is urgently needed. Recently, it was shown that 24% of non-antibacterial drugs impact growth of members of the human microbiome (Maier et al.). Studies such as this would be important for determining any impact of anti-plasmid compounds on the microbiome.

7The ‘One Health’ approach to tacking AMR is based around the notion that AMR does not abide by human, animal or political boundaries, and therefore a multisectoral and multifaceted approach is required. Likewise, anti-plasmid strategies should also adopt a One Health strategy, and not be focused on human medicine alone. Indeed some anti-plasmid strategies are unsuitable or unviable for human use. Furthermore, anti-plasmid strategies alone will never ‘solve’ AMR; nonetheless, they could play an important role in reducing global resistance levels. Removing drug-resistance plasmids is a strategy for all sectors to reduce the overall burden of AMR. For example, plasmid curing could be used to remove ARGs from bacteria in sewage before release into the environment. Human and animal waste is often recycled and used to fertilise agricultural land; this can contain high concentrations and varieties of ARGs which can be passed on to people (Meek, Vyas and Piddock; Rahube et al.). One study performed in Canada found in the first year vegetables grown above, on and below the surface of soil treated with sewage contained significantly more ARGs than non-treated soil (Rahube et al.). An abundance of ARGs were detected in plasmid metagenome libraries constructed from the influent, activated sludge and digested sludge from two wastewater treatment plants in Hong Kong, demonstrating that these were important reservoirs of ARGs (Li, Li and Zhang ). River samples taken upstream and downstream of a tertiary waste water treatment plant in the UK in 2009 and 2011 were examined for third-generation cephalosporin-resistant Enterobacteriaceae (Amos et al.). Significantly higher amounts of blaCTX-M-15 were found downstream of the plant, and 10 novel genetic contexts were identified (Amos et al.). The plasmids containing blaCTX-M-15 were conjugative, and were in pathogens such as the highly successful extraintestinal E. coli ST131 (Stoesser et al.), and other species never before reported to carry blaCTX-M-15 (Amos et al.). IncP-1ε plasmids were detected in manure and arable soil in Germany, and a correlation was found between the presence of IncP-1ε plasmids and antibiotic use (Heuer et al.). A waste water treatment plant in Brazil found 34% of E. coli and 27% of K. pneumoniae were resistant to cephalosporins and/or quinolones, and 5.4% of Klebsiella species were carbapenem resistant in raw as well as treated water (Conte et al.). Analysis of these ARGs showed a high prevalence of blaCTX-M and blaSHV (Conte et al.). Recent work from our group examined wastewater used for irrigation of urban agriculture plots in Burkina Faso. This wastewater contained multiple ARGs including ESBLs, 10 different Enterobacteriaceae-associated plasmid incompatibility groups and 30 Gram-positive replicons associated with ARGs (Bougnom et al., submitted). Together, these studies demonstrate that a treatment such as plasmid curing agents to remove ARGs from manure, sewage and waste water are needed.

8The search for plasmid curing compounds began decades ago, and gained momentum in the 1970s (Table 1). The number of publications peaked in the 1980s (based on searches for publications relating to plasmid curing performed on NCBI PubMed). However, most compounds were toxic, and would produce adverse or unwanted side effects and thus had little use in human medicine. This was followed by a decline in interest and publications. Generally, plasmid curing properties have been evaluated by culturing strains in the presence of a compound or extract at subgrowth inhibitory concentrations. Curing effects are then confirmed by the reversal of plasmid-mediated antibiotic resistance and/or by physical loss of the plasmid(s). Therefore, many of the older publications only refer to the loss of an AMR phenotype.

9The rise in AMR, specifically plasmid-mediated resistance, combined with the dwindling pipeline of new drugs in development has resulted in a resurgence of interest in plasmid curing. Strategies of plasmid curing vary greatly, such as the use of chemicals, drugs, natural products, phage therapies, other plasmids and even CRISPR/Cas. A recent study demonstrated that inhibiting plasmid conjugation was an effective means to remove a plasmid from a bacterial population over time (Lopatkin et al.). The authors concluded that strategies to prevent plasmid conjugation should be explored as a means to reduce AMR plasmid prevalence (Lopatkin et al.). Plasmid curing of a population can also occur when plasmid replication is prevented or reduced, or if plasmid segregation is disrupted, resulting in gradual reduction in plasmid carrying cells. Plasmid curing can also be achieved by increasing the fitness cost associated with plasmid carriage. We anticipate over the next decade that these mechanisms will be studied, streamlined and new practical ways to reduce global AMR plasmid carriage, and hence presence of ARGs, will be developed.

10Many compounds have shown some plasmid curing activity. These include detergents, biocides, DNA intercalating agents, antibiotics (e.g. aminocoumarins, quinolones, rifampicin), ascorbic acid, psychotropic drugs (e.g. chlorpromazine) and plant-derived compounds (Table 1). The effectiveness of these compounds varies greatly and depends on bacterial strain, plasmid and growth conditions. Plasmid curing compounds can act through different mechanisms. In many cases, the compound disrupts plasmid replication by integrating into the DNA (e.g. intercalating agents and chlorpromazine), causing breaks in DNA (e.g. ascorbic acid) or by influencing plasmid supercoiling (e.g. aminocoumarins and quinolones). Plasmid curing compounds can also act by preventing conjugation (e.g. unsaturated fatty acids and TraE inhibitors). Each of these can result in reduced plasmid prevalence within the population over time. The mechanism of action of some curing agents remains to be fully elucidated. One could hypothesise that plasmid curing compounds could also target plasmid segregation, by preventing equal distribution among daughter cells, or increase the fitness burden associated with plasmid carriage.

11The detergents bile and sodium dodecyl sulphate (SDS) are able to cure some plasmids from some bacterial strains (Table 1). Four notable examples include a study where bile salts dose-dependently caused the loss of the Salmonella enterica serovar Typhimurium virulence plasmid, pSLT (García-Quintanilla et al.). However, the level of bile required was 10%–15%, which is significantly higher than that found normally within the small intestine (0.2%–2%) (García-Quintanilla et al.; Kristoffersen et al.). The Salmonella virulence plasmid can be transmitted to new hosts in the mouse intestine, but transmission is unlikely to occur in areas with high levels of bile (García-Quintanilla, Ramos-Morales and Casadesús ). Bile (>1%) reduced expression of conjugative pilus genes pilV and pilT, and decreased conjugation of S. enterica Infantis mega plasmid pESI (280 kb), encoding resistance to tetracycline, sulfamethoxazole and trimethoprim as well as virulence traits (Aviv et al.; Aviv, Rahav and Gal-mor ). The relevance of bile-mediated plasmid curing during human Salmonella infections remains unclear. In addition, the levels of bile required for plasmid curing or to reduce plasmid transmission may result in diarrhoea, and therefore bile is unlikely to be used as a treatment.

12SDS-based plasmid curing methods have been used as a laboratory tool for decades. In 1968, SDS was shown to reduce carriage of fertility and resistance factors (F and R factors/plasmids) (Tomoeda et al.). Over the years, SDS has been used to cure plasmids from E. coli (Rosas et al.; Bharathi and Polasa; Keyhani et al.; Zaman, Pasha and Akhter ), K. pneumoniae (El-Mansi et al.), Pseudomonas aeruginosa (Raja and Selvam ), Lactobacillus species (Lavanya et al.) and Staphylococcus aureus (Jetten and Vogels ) (Table 1). SDS also had other effects on bacteria; these included changes in the peptidoglycan layer, bacterial cell size, septation and loss of outer membrane components (Rosas et al.).

13In summary, detergents are unlikely to be used in humans or animals to reduce AMR plasmids, mainly due to the high concentrations needed, and the associated unwanted gastrointestinal side effects, such as SDS-induced colitis. However, detergents continue to be used in the laboratory setting as a tool to study plasmid biology.

14Recently, it was shown that concentrations well below the MIC of triclosan (also called irgasan) increased the loss of a GFP reporter plasmid pMIB4 from E. coli (Riber et al.). A key finding of this paper was that triclosan embedded in interpenetrating polymer networks of silicone hydrogels was effective at reducing plasmid carriage. The use of such technology as a drug delivery system is appealing, especially for items such as indwelling medical devices (e.g. catheters). However, exposure to triclosan can select for MDR bacterial mutants, largely due to overexpression of bacterial efflux pumps (Chuanchuen et al.; Webber et al.,,; Hernandez et al.; Fernando et al.; Rensch et al.; Gantzhorn, Olsen and Thomsen ). Therefore, caution should be used implementing such a strategy.

15The DNA intercalating agents acridine orange, ethidium bromide and acriflavine also have plasmid curing properties. Acridine orange cured E. coli (Keyhani et al.; Zaman, Pasha and Akhter ), Vibrio parahaemolyticus (Letchumanan et al.), Lactobacillus plantarum (Adeyemo and Onilude ), S. aureus (Jetten and Vogels; Ersfeld-Dressen, Sahl and Brandis ), Bacteroides fragilis and B. thetaiotaomicron (Rotimi, Duerden and Hafiz ) (Table 1). In the late 1960s and early 1970s, ethidium bromide was found to eliminate plasmids from various strains of S. aureus and Escherichia coli (Table 1) (Bouanchaud, Scavizzi and Chabbert; Rubin and Rosenblum ). More recently, it has been used to cure other plasmids from E. coli (Rosas et al.), Bacillus cereus (Borah and RNS ), clinical Enterobacter aerogenes isolates of a blaTEM-1 and blaKPC-3 pKpQIL-like plasmid (Pulcrano et al.), and cured plasmids from two avian Salmonella strains (Poppe and Gyles ) (Table 1).

16Acriflavine cured some resistance plasmids from Salmonella Oranienburg, S. Panama and E. coli K12 in vitro and in a murine in vivo model (Bouanchaud and Chabbert ). Acriflavine also cured E. coli of haemolysin production (Mitchell and Kenworthy ). However, it is more commonly associated with curing of Gram-positive bacteria. It cured plasmids from Group A Streptococci (Nakae, Inoue and Mitsuhashi ), Lactobacillus casei (Chassy, Gibson and Guiffrida ), L. reuteri (Axelsson et al.) and Oenococcus oeni (used in wine production) (Mesas, Rodriguez and Alegre ) (Table 1). Acriflavine was effective at curing resistance from antibiotic-resistant Enterococcus faecium and E. faecalis (Coleri et al.).

17Acriflavine, ethidium bromide and acridine orange caused loss of a plasmid-encoded staphylococcin production in Staphylococcus species (Jetten and Vogels; Ersfeld-Dressen, Sahl and Brandis ); however, as strains became resistant to acriflavine they also became resistant to its curing effects (Jetten and Vogels ). Acriflavine, acridine orange and ethidium bromide cured resistance to antimicrobials from both donor and transconjugants B. fragilis and B. thetaiotaomicron (Rotimi, Duerden and Hafiz ).

18The practical applications of DNA intercalating agents are few, due to their activity as powerful mutagens, associated with significant toxicity and the carcinogenic nature of these molecules. The harm of using such compounds vastly outweighs any potential benefit derived from plasmid curing. In addition, as many intercalating agents are substrates of bacterial efflux pumps, the use of such compounds could select for overexpression of efflux pumps which can lead to MDR (Piddock ). However, these compounds can still be useful in a laboratory setting to cure strains of plasmids (Coleri et al.; Mesas, Rodriguez and Alegre; Chin et al.; Raja and Selvam; Zaman, Pasha and Akhter; Adeyemo and Onilude; Pulcrano et al.).

19Many well-studied plant extracts come from traditional medicine. Plumbagin (5-hydroxy-2-methyl-1,4-naphthoquinone) is a yellow dye derived from the root of the tropical/subtropical Plumbago species (Patwardhan et al.). Plumbagin is reported to have anticancer, antifungal and antimicrobial activity (Padhye et al.; Tyagi and Menghani ). In E. coli, plumbagin effectively eliminated a conjugative, MDR plasmid (Lakhmi, Padma and Polasa ) and the RP4 plasmid (Bharathi and Polasa ). Plumbagin eliminated plasmids from E. coli, by decreasing plasmid copy number and reducing the toxic effect of plasmid loss (Lakshmi and Thomas ) (Table 1).

20Subinhibitory concentrations of Plumbago zeylanica root extract were tested on MDR clinical isolates of S. Paratyphi, S. aureus, E. coli and Shigella dysenteriae, as well as E. coli containing pUK651, but the extract only cured 14% of E. coli of pUK651 (Beg and Ahmad ). Subinhibitory concentrations of P. auriculata root extracts cured drug-resistance plasmids from P. aeruginosa, E. coli, Proteus vulgaris and K. pneumoniae, which were slightly higher than pure plumbagin (Patwardhan et al.) (Table 1).

218-epidiosbulbin E acetate is isolated from the bulbs of Dioscorea bulbifera, a plant known in Ayurvedic alternative medicine (Shriram et al.). 8-epidiosbulbin E acetate belongs to the clerodane class of diterpenes. Its antibacterial and curing activity was evaluated, and it cured reference strains of E. coli, B. subtilis, P. aeruginosa, and clinical isolates of E. coli, E. faecalis and S. sonnei with an average efficiency of 34% (Shriram et al.) (Table 1).

22The curing activity of the crude extract of Alpinia galanga (L.) Swartz, a medicinal plant indigenous to Southeast Asian countries, was tested (Latha et al.). The bioactive fraction containing 1΄-acetoxychavicol acetate was tested on nine bacterial reference strains carrying antibiotic-resistance plasmids. A subinhibitory concentration of crude extract cured plasmids from S. Typhi, E. coli and E. faecalis. Purified 1΄-acetoxychavicol acetate cured MDR plasmids from S. Typhi, P. aeruginosa, E. faecalis, E. coli and B. cereus (Latha et al.) (Table 1).

23Taken together, plant-derived compounds can be effective at curing plasmids in vitro; however, more research is needed to confirm spectrum of activity, identify the active components and to determine any toxicity and in vivo efficacy.

24Work from de la Cruz and colleagues has focused on performing high-throughput screens of compounds to search for inhibitors of conjugation (Fernandez-Lopez et al.; Getino et al.,; Ripoll-Rozada et al.). Their high-throughput screening method used a lux reporter under the control of the lac promoter, on a simple conjugative plasmid derived from R388 in E. coli (Fernandez-Lopez et al.). The donor carried the lacI repressor; thus, luminescence was only produced after conjugation (Fernandez-Lopez et al.). They tested a library of microbial extracts, and showed that unsaturated fatty acids, including dehydrocrypenynic acid, linoleic acid and oleic acid, inhibited conjugation (Fernandez-Lopez et al.). Recently, Lopatkin et al. used linoleic acid to determine the impact of reduced conjugation on plasmid persistence within a population. Indeed, 3.5 μM linoleic acid was sufficient to destabilise a plasmid with low conjugation efficiency from a population; however, it was ineffective for plasmids with higher conjugation efficiencies or which carried a fitness benefit (Lopatkin et al.).

25A study of synthetic fatty acids demonstrated that 2-alynoic fatty acids inhibited conjugation; of these, 2-hexadecynoic acid was the most potent, followed by 2-octadecynoic acid (Getino et al.). At concentrations of 0.4 mM, 2-hexadecynoic acid reduced conjugation frequencies of IncW, IncH and IncF plasmids by 100 times, while concentrations of 1 mM were required to reduce conjugation of IncI, IncL/M and IncX plasmids. Conjugation of IncP and IncN plasmids was not affected by 2-hexadecynoic acid. Using molecules with similar structures, they determined that the carboxylic group, a 16-carbon chain and one unsaturated bond were optimal for conjugation inhibition. They showed that 2-hexadecanoic acid acted on the donor, and inhibited conjugation in E. coli, S. enterica, P. putida and Acinetobacter baumannii (Getino et al.).

26Four unsaturated fatty acids (linoleic, oleic, 2-hexadecynoic and 2-ocatadecynoic acid) inhibited the activity of the plasmid encoded TrwD ATPase (VirB11 homologue) (Ripoll-Rozada et al.). TrwD acts as a traffic ATPase, regulating switching between pilus biogenesis and DNA translocation through the conjugation machinery (Ripoll-Rozada et al.). Fatty acids which did not inhibit conjugation had no impact on TrwD activity (Ripoll-Rozada et al.). The authors suggested that the mechanism for the conjugation inhibiting activity of unsaturated fatty acids was due to their binding to the N-terminal domain and linker region of TrwD, inhibiting the movement of the N-terminal domain over the C-terminal domain, thus preventing ATPase activity of the enzyme (Ripoll-Rozada et al.).

27One of the concerns about any clinical use of synthetic fatty acids, such as 2-hexadecanoic acid, is toxicity in people or animals. Recent work focused on finding less-toxic molecules by screening a natural compound library produced by aquatic microbes (Getino et al.). Tanzawaic acid A and B, polyketides produced by Penicillium species, were identified as effective conjugation inhibitors of IncW and IncFII plasmids. Tanzawaic acid B (0.4 mM) reduced conjugation by 100-fold for IncW and IncFII, as compared to untreated controls. However, they were only moderately effective on IncFI, IncI, IncL/M, IncX and IncH plasmids, reducing conjugation by between 10% and 50% compared to untreated cells. In addition, they did not inhibit conjugation of IncN and IncP plasmids (Getino et al.). Importantly, oleic acid, linoleic acid and tanzawaic acids A and B were less toxic on bacteria, fungi and tissue culture cells than 2-hexadecynoic and 2-oxydecynoic acid (Getino et al.).

28Unsaturated fatty acids have been shown to be effective conjugation inhibitors in many laboratory settings, and on a variety of plasmids. Furthermore, they are associated with reduced toxicity on tissue culture cells. Further studies are needed to determine the in vivo safety and efficacy of unsaturated fatty acids, but they are promising candidates for future plasmid curing work.

29Using a targeted approach, Baron and colleagues have identified small molecules which bind to and inhibit the dimerisation of TraE, an essential component of the type IV secretion system involved in a variety of functions including conjugation of pKM101 (Paschos et al.; Casu et al., ). Structural studies of the pKM101 encoded TraE dimerisation (VirB8 homologue) were used as a basis for uncovering small molecules which inhibited dimerisation, four of which (molecules B8I-16, BAR-072, BAR-073 and UM-024) also inhibited transmission of pKM101 (Casu et al.). None of these molecules impacted upon transmission of RP4, highlighting their specificity for pKM101 TraE (Casu et al.). In a follow-up study, Casu et al. screened a fragment library for compounds which bound to TraE. They used this information to design two molecules which bound with high affinity to TraE and were able to reduce transmission of pKM101 (molecules 105055 and 239852) (Casu et al.). Together, this work demonstrates the feasibility and specificity of structure-based design of anti-plasmid compounds.

30DNA gyrase is essential in bacteria as it introduces supercoiling into DNA molecules; it is comprised of two GyrA and two GyrB monomers (Andriole ). Multiple antibiotics target DNA gyrase. Aminocoumarin antibiotics, such as novobiocin and coumermycin A, inhibit GyrB (Gellert et al.). These and the related compounds clorobiocin and isobutyryl novenamine were effective at plasmid curing (Hooper et al.). The GyrB inhibiting activities of aminocoumarins are responsible for their plasmid curing properties (Taylor and Levine ), and the E. coli gyrase B subunit is required for plasmid maintenance, and curing activity of coumermycin A1 (Wolfson et al.). Novobiocin interfered with plasmid maintenance, rather than selecting plasmid-free isolates (Hooper et al.). Furthermore, bacteria with a mutation in gyrB conferring resistance to coumermycin required higher levels of the antibiotic to produce the curing effect (Hooper et al.).

31Novobiocin was effective at curing plasmids from many Gram-positive bacteria including L. plantarum, Lactobacillus strains isolated from chickens, L. acidophilus isolated from molasses, E. faecalis, clinical isolates of enterococci, B. subtilis and S. aureus (Table 2) (McHugh and Swartz; Ruiz-Barba, Piard and Jiménez-Díaz; Chin et al.; Karthikeyan and Santosh ). Escherichia coli and other Gram-negative Enterobacteriaceae were cured of a variety of plasmids by novobiocin (Michel-briand et al.). Novobiocin eliminated the Salmonella virulence plasmid from S. Typhimurium, resistance plasmids from Serratia marcescens and a cryptic plasmid from Chlamydia muridarum (Gulig and Curtiss III; Llanes et al.; O’Connell and Nicks ). Coumermycin eliminated some plasmids from E. coli, but not RP4 (Danilevskaya and Gragerov; Wolfson et al.; Bharathi and Polasa ).

32Quinolone antimicrobials also target DNA gyrase. There have been numerous reports of plasmids cured from various bacterial species by different quinolone antibiotics (Table 2). The majority of studies have been done using E. coli. For example, five fluoroquinolones and two quinolones cured four plasmids (Weisser and Wiedemann ), and subinhibitory levels of quinolones cured E. coli of various plasmids including large clinical plasmids (Table 2) (Oliva et al.; Platt and Black; Courtright, Turowski and Sonstein; Selan et al.). However, quinolones have variable curing activity on some plasmids (Weisser and Wiedemann, ). For example, quinolones resulted in incomplete curing and reduced copy number of several plasmids (Table 2) (Phillips and Towner; Brandi, Falconi and Ripa ), and were ineffective at curing E. coli of other plasmids (pBP1, R391, R27 or three small, high-copy plasmids from a clinical E. coli isolate) (Weisser and Wiedemann; Platt and Black ). In line with this, one study demonstrated in E. coli that quinolones cured pORF2 with high efficiency, three plasmids were poorly cured and three plasmids were unaffected (Table 2) (Fu et al.). Interestingly, this study also examined quinolone efficacy at curing pORF2 from E. coli in vivo. They found quinolone treatment of mice infected with E. coli/pORF2 led to significant reduction in plasmid carriage (Fu et al.).

33In a large study, 12 quinolones were tested for their ability to cure 11 plasmids of different incompatibility groups from E. coli, and virulence plasmids in five other species of Enterobacteriaceae (Table 2) (Michel-briand et al.). The authors concluded that non-fluorinated quinolones had slightly higher curing activity, but that novobiocin cured better than quinolones (Michel-briand et al.). Other studies examining a range of bacteria showed subinhibitory concentrations of quinolones reduced resistance and virulence plasmids in S. aureus, S. Typhimurium, E. coli, P. aeruginosa and Yersinia pseudotuberculosis (Hahn and Ciak; Sonstein and Burnham ).

34In summary, aminocoumarin-mediated curing appears to be more effective on Gram-positive bacteria than Gram-negative bacteria. Quinolone-mediated plasmid curing is effective on some plasmids in Gram-negative bacteria such as E. coli. However, this is complicated by the presence of plasmid-mediated quinolone-resistance genes, such as qnr, aac(6΄)-1b-cr, qepA and oqxAB (Jacoby, Strahilevitz and Hooper; Rodriguez-Martinez et al.). Attempting to use quinolones to cure plasmids carrying quinolone-resistance genes could provide a fitness advantage to plasmid-containing cells, and would therefore select for plasmid maintenance. Furthermore, plasmid-mediated quinolone-resistance genes are frequently coded for by plasmids which carry other resistance genes conferring resistance to antimicrobials including beta-lactams, extended spectrum beta-lactams, carbapenems, aminoglycosides, trimethoprim and chloramphenicol (Rodriguez-Martinez et al.). Taken together, it is unlikely that antibiotics will be used to cure AMR plasmids in humans, animals or the environment as this will provide selection pressure for resistance to arise or be maintained within bacteria. Therefore, aminocoumarin and quinolone antibiotics are an effective laboratory tool, but are unlikely to be used elsewhere for plasmid curing.

35The antibiotic rifampicin inhibits RNA polymerase and is used to treat tuberculosis. Subinhibitory concentrations of rifampicin cured a penicillin-resistance plasmid from S. aureus (Johnston and Richmond ) and the F’lac plasmid from E. coli (Bazzicalupo and Tocchini-Valentini ). However, a rifampicin-resistant strain was not susceptible to curing, suggesting that the mechanism of curing was dependent upon the interaction of rifampicin with RNA polymerase (Bazzicalupo and Tocchini-Valentini ). In E. coli, haemolysin production was effectively cured by rifampicin (Mitchell and Kenworthy ). A gentamicin-resistance plasmid was not cured from two S. aureus strains using rifampicin, but it did cure one strain of a penicillin-resistance plasmid (Wood, Carter and Best ). Multiple studies have found rifampicin to be less effective than other curing agents (Rubin and Rosenblum; Poppe and Gyles ), and given the importance of rifampicin in treating infections such as tuberculosis, it is unlikely to be used as a general plasmid curing agent.

36Research on the bioactive compound ascorbic acid (vitamin C) dates back to the first half of the 20th century. In aerobic conditions, ascorbic acid converts circular covalently closed DNA into open circular DNA (Morgan, Cone and Elgert ). To investigate the mechanism of action, fragments of pBR322 with radio-labelled 3΄ ends were used to demonstrate that efficient cleavage occurred preferentially at purine-rich regions (Chiou et al.). Studies on DNA extracted from E. coli demonstrated that ascorbic acid specificity was linked to negative torsion of the DNA, and this was influenced by ionic strength, salt concentration and pH (Wang and Ness ). In vitro studies on plasmid pBR322 DNA showed that ascorbic acid increased the damaging effects of dimethylarsinous acid and human liver ferritin (Ahmad, Kitchin and Cullen ). Synthesised ascorbic acid variants with protected (non-reactive) hydroxyl groups were tested for their ability to relax pUC19, which demonstrated that the hydroxyl groups at position C2 and C3 were essential for DNA damage (Liu et al.).

37In S. aureus 1 mM ascorbic acid resulted in loss of penicillin and aminoglycoside resistance encoding plasmids (Table 1) (Amábile Cuevas; Amábile-Cuevas, Piña-Zentella and Wah-Laborde ). Two plasmids, pI258 (penicillin resistance) and pT181 (tetracycline resistance), were not cured by ascorbic acid. However, there was a significant decrease in the MIC of tetracycline, which the authors hypothesised was due to reduction in plasmid copy number (Amábile-Cuevas, Piña-Zentella and Wah-Laborde ).

38Ascorbic acid (1 mM) cured the lactic acid bacterium Pediococcus acidilactici of a plasmid coding for the production of pediocin, a metabolite which inhibits growth of some pathogenic bacteria, thus minimising food spoilage (Ramesh, Halami and Chandrashekar ). Ascorbic acid is non-toxic and is associated with human health benefits. This makes it an attractive curing agent, although it seems to be more effective at curing plasmids from Gram-positive rather than Gram-negative bacteria. Furthermore, after vitamin C supplementation concentrations of ascorbic acid in the plasma are relatively low (0.07 mM), but concentrations in lymphocytes can be much higher (3.5 mM), and concentrations in duodenal biopsies were around 1.2 mmol/kg (Levine et al.; Waring et al.). Conversely, Maier et al. estimated ascorbic acid concentrations in the intestine to be around 0.379 mM. Together this shows that while plasma concentrations after supplementation would not reach sufficient levels to have anti-plasmid activity, ascorbic acid is concentrated in the intestine, where it could potentially affect plasmids within intestinal bacteria. However this remains to be demonstrated in vivo.

39The phenothiazines have been widely used in human medicine, originally as anti-helminthics, but now this class of drugs comprises the largest of five classes of anti-psychotic drugs (Ohlow and Moosmann ). The impact of these molecules on bacteria has been reviewed elsewhere (Amaral, Viveiros and Molnar; Spengler et al.; Varga et al.). Plasmid curing properties have also been attributed to phenothiazines (Table 1) (Amaral, Viveiros and Molnar; Spengler et al.; Dastidar et al.). In addition, a recent study found that chlorpromazine significantly impacted the growth of diverse members of the human microbiome, including Akkermansia muciniphila, Bacteroides uniformis, B. vulgatus, Clostridium perfringens, Parabacteroides distasonis and P. merdae (Maier et al.). Phenothiazines, including chlorpromazine, cured plasmids from E. coli (Table 1) (Mandi et al.; Molnar, Mandi and Kiraly ), and the curing activity was enhanced by methylene blue (Molnar et al.). Thioridazine cured the AMR phenotype from E. coli, S. flexneri and V. cholerae isolates, but not from S. aureus (Table 1) (Radhakrishnan et al.), while promethazine and trifluoperazine were tested on clinical isolates of E. coli, Citrobacter freundii and E. cloacae, but only one E. coli isolate was cured, despite E. coli K12 being readily cured of a lac-reporter plasmid (Spengler et al.). However, trifluoroketone 18 or trifluoromethyl-ketone 14 (proton pump inhibitors) enhanced curing activity of the phenothiazines, suggesting the compounds may be effluxed (Spengler et al.; Wolfart et al.). In mixed cultures of E. coli, B. cereus and S. epidermidis, promethazine cured F’lac from E. coli (Molnár, Amaral and Molnár ). Chlorpromazine cured the MRSA Iberian clone strain HPV107 of a plasmid encoding the QacA efflux pump (Costa et al.).

40Together, this shows phenothiazines have in vitro curing activity on some bacteria and plasmid combinations. However, their in vivo efficacy as plasmid curing compounds remains unclear. Any potential connection between anti-plasmid and the anti-commensal activity of chlorpromazine remains to be elucidated. In patients being treated for psychosis with chlorpromazine, serum concentrations are around 0.1–0.3 μg/mL, and toxic side effects occur at 0.75 μg/mL (Sanofi-Aventis ). However, Maier et al. estimate intestinal concentration of chlorpromazine to be around 46 μM (14.67 μg/mL). The concentrations used for plasmid curing are generally around 10–100 μg/mL (Mandi et al.; Spengler et al.). Therefore, concentrations resulting in curing may be reached in the intestines of individuals being treated with chlorpromazine. Novel approaches involving targeted drug delivery or preventing uptake of orally administered phenothiazines may help to improve curing efficacy and reduce toxicity. Until such obstacles are overcome, the use of phenothiazines for in vivo plasmid curing is unlikely.

41Curing based upon the principle of plasmid incompatibility is an alternative method to chemical or drug-based strategies to remove plasmids from bacteria. Plasmid curing using an incompatible plasmid vector has been widely used in plasmid characterisation of Gram-positive and Gram-negative species. Introducing a smaller high-copy-number plasmid from the same incompatibility group may specifically eliminate a resident plasmid (Bringel, Frey and Hubert ). Incompatibility-based curing has been useful for investigating incompatibility mechanisms, plasmid–host interactions and for the construction of gene transfer systems (Uraji, Suzuki and Yoshida ). The main advantage of this method is the reduced risk of chromosomal mutations and toxicity sometimes associated with chemical curing agents (Hovi et al.; Poppe and Gyles ). In addition, incompatibility-based curing is specific to plasmids of the targeted incompatibility group. One major drawback of incompatibility-based curing methods is the extensive cloning required for set up, and the detailed knowledge of the target plasmid. Ni et al. reported the main difficulty in constructing incompatibility plasmids for curing is the replication control and/or partition region of the plasmid must be identified before curing (Ni et al.). Additional plasmid genes (e.g. antitoxin from a TA system) may need to be included (Ni et al.; Hale et al.).

42Incompatibility-based curing has been used in a variety of bacteria and plasmids (Table 3). In particular, when chemical curing methods have proven less effective, e.g. Lactobacillus, and Y. pestis (Ruiz-Barba, Piard and Jiménez-Díaz; Chin et al.; Ni et al.; Karthikeyan and Santosh ). Incompatibility-based curing systems were designed and used in L. acidophilus, L. plantarum and L. pentosus (Table 3) (Bringel, Frey and Hubert; Posno et al.). Incompatibility has been used to study the contribution of plasmids to bacterial pathogenesis, including a systematic investigation of the role of plasmids in Y. pestis pathogenesis (Table 3) (Ni et al.). Incompatibility was used to cure vaccine and wild-type strains of B. anthracis of two large pathogenicity-related plasmids (Table 3), allowing study of their contribution to capsule and anthrax toxin production (Wang et al.; Liu et al.). Incompatibility has been used not only in human pathogens, but also to remove tumour inducing (Ti) plasmids from Agrobacterium tumefaciens, a dicotyledonous plant pathogen, in which Ti plasmids are responsible for inducing vegetable tumours (Table 3) (Uraji, Suzuki and Yoshida ).

43Incompatibility-based plasmids called pCURE were constructed for curing pO157 (a typical F-like plasmid), other F-like and IncP-1α plasmids from E. coli (Table 3) (Hale et al.). To create the pCURE constructs, elements expected to interfere with specific functions were chosen, such as repressing vital components (e.g. transcriptional repressor, antisense RNA or other translational regulators) and competition for vital steps (e.g. replication origin) (Hale et al.). To control the TA system, either the putative antitoxin or antisense RNA repressor was included (Hale et al.).

44In a recent study, ‘interference plasmids’ were designed which combined an antitoxin gene and replicon genes to cure blaIMP-4 and blaCMY-2 encoding plasmids both in vitro and in vivo (Table 3) (Kamruzzaman et al.). In the presence of the antibiotic selecting for the interference plasmid, target plasmids were effectively removed from E. coli, K. pneumoniae, C. freundii and Morganella morganii in vitro, and from E. coli colonising the mouse intestine. Interference plasmids were lost from the mouse intestine after cessation of antibiotic treatment.

45One targeted approach sought small molecules which mimic the incompatibility system of IncB plasmids (Denap et al.). They found the aminoglycoside apramycin binds to the SLI region of the RepA mRNA, preventing translation of RepA, which is necessary for plasmid replication (Denap et al.). Treatment of E. coli harbouring pMU2403 (IncB) with apramycin resulted in almost complete plasmid elimination (Denap et al.).

46An important question regarding use outside the laboratory of incompatibility-based curing systems is how to apply the curing plasmids to people, animals or the environment. Plasmids could be delivered via bacteria or phage. However, the potential requirement for antibiotic treatment to select for the curing plasmids (Kamruzzaman et al.) would be a significant drawback. Another concern regarding curing plasmids is the potential for acquisition of ARG(s) onto the curing backbone. More research is needed in increasingly complex plasmid systems to study the dynamics between curing plasmids and AMR plasmids, including research focused on minimising the need for antibiotic selection.

47For the past 50 years, bacteriophages which specifically target the pili of plasmid conjugation systems have been studied (Caro and Schnös ). More recently, this has been studied in the context of AMR plasmids. Phages which target the conjugation pilus preferentially kill bacteria with high pilus expression (Dionisio ). Low pilus expression results in reduced susceptibility to phage, but also reduced conjugation rates. Therefore, diversity in pilus expression within a bacterial population improves the chances of plasmid survival (Dionisio ). Another example of bacteriophages specifically targeting AMR plasmids involved the phage PRD1 which targeted the mating pair complex of plasmids RP4 and RN3 (Jalasvuori et al.). PRD1 reduced plasmid carriage within E. coli and Salmonella populations from 100% to 5% after 10 days. Furthermore, the 5% which retained plasmid had lost the ability to conjugate (Jalasvuori et al.). PRD1 significantly reduced the number of E. coli K12 RP4 transconjugants, and even reduced transconjugants when single, sub-MIC antibiotic selection was applied (Ojala, Laitalainen and Jalasvuori ). However, when double selection for transconjugants was applied phage-resistant mutants arose, but 65% had lost the ability to conjugate (Ojala, Laitalainen and Jalasvuori ). Together, this demonstrates the use of phage to produce an evolutionary pressure which results in either plasmid loss or evolution of a non-conjugative plasmid. This fits with the other research focused on using phage-mediated directed evolution to select for antibiotic sensitive bacteria (Chan et al.).

48The M13 filamentous phage minor coat protein g3p was necessary and sufficient to inhibit F-plasmid conjugation in E. coli (Lin et al.). Another study modelled the dynamics of the F-plasmid and M13 phage in E. coli (Wan and Goddard ). They found M13 infection reduced cell growth rate, and the conjugation rate was only one order of magnitude faster than the rate of phage infection. This implies that a high concentration of phage would be required to effectively prevent conjugation, and they showed that conjugation continues even with phage (Wan and Goddard ). Recently, the evolutionary and ecological implications of lytic bacteriophage predation on plasmid maintenance in a population of P. fluorescens were examined (Harrison et al.). They concluded that phage accelerates plasmid loss in the absence of selective pressure (Harrison et al.).

49In summary, these studies show that phage can be a highly effective tool for reducing plasmid prevalence within a population. Another advantage of bacteriophage approaches is their status as ‘generally regarded as safe’, which streamlines downstream applications such as use of phage to decolonise surfaces, as a probiotic or use on farms. However, unclear regulatory pathways for use of phage as medication still pose a problem. Another problem associated with phage therapy is bacterial evolution of resistance to phage. By understanding the evolutionary pressures applied to bacteria by phage predation, this evolution can be harnessed to increase susceptibility to antibiotics (Jalasvuori et al.; Ojala, Laitalainen and Jalasvuori; Chan et al.). Future research is needed to further our understanding of the phage-plasmid-host dynamics, to improve upon evolution-optimised approaches and to test these approaches in increasingly complex models.

50CRISRP/Cas is a bacterial ‘adaptive immune system’ which allows recognition, degradation and memory of foreign DNA sequences. CRISPR/Cas works as a result of spacer DNA segments coded by the bacteria that are transcribed into crRNA. The crRNA is bound by the Cas protein complex which cleaves nucleic acid sequences matching the crRNA, resulting in double-stranded breaks (Sternberg and Doudna ). DNA repair mechanisms can be used to insert a desired sequence into the break (Sternberg and Doudna ). In bacteria, double-stranded breaks are often fatal, but combination with traditional recombineering systems such as λ-red can allow for effective genome editing (Peters et al.). The highly specific CRISPR/Cas system has been extensively described and reviewed elsewhere (Jiang and Doudna; Sternberg and Doudna; Wright, Nuñez and Doudna ). In a seminal paper, Garneau et al. showed that Streptococcus thermophiles isolates which had lost the plasmid pNT1 had acquired new spacer sequences which targeted pNT1. This work demonstrated that CRISPR/Cas acted to remove plasmid DNA from bacteria.

51Recently, the CRISPR/Cas system has been explored as a method for plasmid curing. Firstly, it can be designed to target specific plasmid genes, including ARGs. The double-stranded breaks introduced in the process can reduce the stability of the plasmid, and in some cases result in plasmid loss (Fig. 2a) (Kim et al.; Lin et al.). Plasmids in isolates from man, animals or the environment frequently carry TA systems. TA systems, sometimes called addiction systems, are comprised of a toxin and an antitoxin gene (Van Melderen and Saavedra De Bast; Chan, Espinosa and Yeo ). Generally, the activity of the stable toxin is mitigated by a less stable antitoxin. Therefore, as long as the antitoxin is produced, the toxin cannot act (Van Melderen and Saavedra De Bast; Chan, Espinosa and Yeo ). When encoded on plasmids, the TA system functions by killing daughter cells which do not contain a copy of the plasmid coding for the antitoxin gene, a process termed postsegregational killing (Chan, Espinosa and Yeo ). Therefore, targeting plasmids with TA systems resulted in bacterial cell death (Fig. 2a) (Citorik, Mimee and Lu ). Toxin-mediated cell death could be complemented by antitoxin-encoding phage (Citorik, Mimee and Lu ). Specific ARGs can also be targeted by CRISPR/Cas systems (Citorik, Mimee and Lu; Kim et al.). For example, homologous regions in TEM and SHV beta-lactamases were targeted (Kim et al.). CRISPR/Cas systems can also target plasmid backbone genes such as replicase genes (Cao et al.). CRISPR/Cas systems targeted and removed multiple AMR plasmids simultaneously (Yosef et al.).

52Secondly, CRISPR/Cas is an attractive strategy because it can be used to prevent plasmid transmission by ‘vaccination’ (Fig. 2b). Methicillin-sensitive S. aureus was vaccinated against pUSA02, the plasmid responsible for methicillin resistance in the epidemic MRSA strain USA300 (Bikard et al.). Likewise, E. coli containing CRISPR/Cas targeting blaCTX-M-15 and blaNDM-1 were less efficiently transformed with an AMR plasmid carrying these genes (Yosef et al.). In E. faecalis the CRISPR3-Cas locus was deleted, resulting in significantly higher acquisition of the pAD1 plasmid (target sequence located in CRISPR3), while acquisition of pCF10 was unaffected, as it is not targeted by CRISPR3 (Price et al.). In line with this, carbapenem-resistant K. pneumoniae were less likely to have active CRISPR/Cas systems than carbapenem-sensitive strains (Lin et al.). Similarly, type I-F CRISPR/Cas systems were more common in E. coli isolates that were antimicrobial sensitive (Aydin et al.). Some of these CRISPR spacers aligned to sequences commonly found in IncFII and IncI1 plasmids, which are associated with clinical resistance (Aydin et al.). This strongly suggests that antimicrobial-sensitive isolates can use CRISPR/Cas systems to degrade incoming AMR plasmids.

53The benefits of using CRISPR/Cas systems to cure bacterial plasmids are clear. However, there are significant drawbacks associated with this strategy. One of the primary problems is delivery. A variety of delivery methods including plasmid transformation, conjugation, phagemid and bacteriophages have been used predominantly in vitro, with limited studies using in vivo models (Bikard et al.; Yosef et al.; Kim et al.). Despite this, so far the practical use of these systems is limited. One study used bacteria sensitised to antibiotics by a lysogenic CRISPR phage which also carried resistance to a lytic phage, thus combining use of the lytic and lysogenic phages to put evolutionary pressure on bacteria to become drug sensitive (Fig. 2c) (Yosef et al.). A recently devised system termed GOTraP enhances DNA transduction to a variety of bacterial species, including E. coli, S. sonnei and K. pneumoniae (Yosef et al.). Such a strategy could be employed to improve the delivery of anti-ARG or anti-plasmid CRISPR/Cas systems. In another study, a phagemid was effective for treating S. aureus skin lesions in mice (Bikard et al.). CRISPR/Cas may be an effective treatment for curing plasmids from surface wounds and burns, but advances in systemic treatment are still required.

54The specificity of CRISPR/Cas is both a benefit and a drawback. For example, the lack of common sequences among variants of blaOXA and blaCTX-M beta-lactamases restricted CRISPR design (Kim et al.). Frequently, plasmid-mediated antibiotic resistance genes have multiple sequence variants, so if CRISPR/Cas sequences were used to kill bacteria there could be selective pressure for mutations in the regions targeted by CRISPR to give a CRISPR-resistant plasmid (Gomaa et al.). Similarly, in a phagemid system, resistance occurred due to deletions of the cas9 gene on the phagemid (Bikard et al.). Future research must consider the evolutionary pressure, and as some have done (Yosef et al.), design strategies to reduce development of resistance.

55We and others anticipate that future research will continue in this area, driven in large part by the need to prevent and treat resistant infections (Getino and de la Cruz ). Methods to effectively and safely cure plasmids have the potential to diminish the severity of the impact of drug-resistant infections. Currently, few studies have examined curing methods in vivo. These studies along with future in vivo plasmid curing studies will be crucial in developing methods to sensitise bacteria to existing antibiotics. In the future, it may be that doctors prescribe a plasmid curing agent to help ensure that the antibiotics taken by the patient are effective. Alternatively, a plasmid curing agent could be taken by an individual (e.g. on return from an area where plasmid-mediated drug-resistance is common) as a way of restoring drug-susceptible bacteria to the gastrointestinal microbiome.

56The use of plasmid curing strategies in settings other than in humans and animals should not be under appreciated. Reducing the global burden of AMR will require a multifaceted One-Health approach, and curing AMR plasmids from ARG hot spots such as waste water, manure and downstream of pharmaceutical (antibiotic) factories is a viable strategy. Some of the approaches described may be more suited to an environmental or agricultural setting and not for human use. For instance, another potential use of curing strategies could be on farms where livestock are often exposed to antibiotics, and harbour multiple MDR plasmids. Soil, waste water treatment and aquaculture could all be treated with plasmid curing agents to reduce drug resistance.

57One concern surrounding plasmid curing is the potential for the development of resistance to anti-plasmid approaches, and the impact of these approaches on the bacterial community structure. Bacteria are constantly evolving, which makes developing ‘evolution-proof’ drugs extremely challenging, but by striving for this gold standard it may be possible to delay resistance (Bell and MacLean ). While these ideas apply to antimicrobials, similar principles can applied to plasmid curing strategies. If an anti-plasmid approach kills or produces a fitness defect in the bacteria, it seems likely that resistance will occur. For example, some of the CRISPR/Cas approaches selectively kill resistant bacteria by targeting specific sequences. Mutations in these sequences would then result in survival of the mutants, and expansion due to the resistance phenotypes. Thus, methods to provide selective advantages for the sensitive strains or to guide evolution towards antimicrobial susceptible strains can be utilised to help overcome these challenges. Antibiotics which have curing properties (e.g. quinolones and rifampicin) are obvious examples where resistance would be selected, and therefore should not be used as curing compounds outside the laboratory. Another concern is cell permeability. If plasmids provide a beneficial trait, mutations which reduce permeability or increase efflux may be selected by plasmid curing compounds. Indeed bacteria may increase expression of efflux to remove potentially detrimental plasmid curing compounds from the cells. Such mutations could also produce reduced susceptibility to clinically important antimicrobials. Research should consider and address these concerns.

58Altogether, plasmid curing has come a long way, from the use of toxic compounds to novel designer curing methods based on incompatibility or CRISPR/Cas. Further research is now needed to uncover safe and effective means to cure plasmids, particularly in the face of the global AMR crisis.