CRISPR-Cas9-Mediated Carbapenemase Gene and Plasmid Curing in Carbapenem-Resistant Enterobacteriaceae

Combating plasmid-mediated carbapenem resistance is essential to control and prevent the dissemination of carbapenem-resistant Enterobacteriaceae (CRE). Here, we conducted a proof-of-concept study to demonstrate that CRISPR-Cas9-mediated resistance gene and plasmid curing can effectively resensitize CRE to carbapenems. A novel CRISPR-Cas9-mediated plasmid-curing system (pCasCure) was developed and electrotransferred into various clinical CRE isolates. The results showed that pCasCure can effectively cure bla_KPC, bla_NDM, and bla_OXA-48 in various Enterobacteriaceae species of Klebsiella pneumoniae, Escherichia coli, Enterobacter hormaechei, Enterobacter xiangfangensis, and Serratia marcescens clinical isolates, with a >94% curing efficiency.

ABSTRACT

Combating plasmid-mediated carbapenem resistance is essential to control and prevent the dissemination of carbapenem-resistant Enterobacteriaceae (CRE). Here, we conducted a proof-of-concept study to demonstrate that CRISPR-Cas9-mediated resistance gene and plasmid curing can effectively resensitize CRE to carbapenems. A novel CRISPR-Cas9-mediated plasmid-curing system (pCasCure) was developed and electrotransferred into various clinical CRE isolates. The results showed that pCasCure can effectively cure bla_KPC, bla_NDM, and bla_OXA-48 in various Enterobacteriaceae species of Klebsiella pneumoniae, Escherichia coli, Enterobacter hormaechei, Enterobacter xiangfangensis, and Serratia marcescens clinical isolates, with a >94% curing efficiency. In addition, we also demonstrated that pCasCure can efficiently eliminate several epidemic carbapenem-resistant plasmids, including the bla_KPC-harboring IncFIIK-pKpQIL and IncN pKp58_N plasmids, the bla_OXA-48-harboring pOXA-48-like plasmid, and the bla_NDM-harboring IncX3 plasmid, by targeting their replication and partitioning (parA in pKpQIL) genes. However, curing the bla_OXA-48 gene failed to eliminate its corresponding pOXA-48-like plasmid in clinical K. pneumoniae isolate 49210, while further next-generation sequencing revealed that it was due to IS1R-mediated recombination outside the CRISPR-Cas9 cleavage site resulting in bla_OXA-48 truncation and, therefore, escaped plasmid curing. Nevertheless, the curing of carbapenemase genes or plasmids, including the truncation of bla_OXA-48 in 49210, successfully restore their susceptibility to carbapenems, with a >8-fold reduction of MIC values in all tested isolates. Taken together, our study confirmed the concept of using CRISPR-Cas9-mediated carbapenemase gene and plasmid curing to resensitize CRE to carbapenems. Further work is needed to integrate pCasCure in an optimal delivery system to make it applicable for clinical intervention.

INTRODUCTION

The emergence and spread of carbapenem-resistant Enterobacteriaceae (CRE) have created a significant global public health threat (1). Carbapenem resistance in Enterobacteriaceae is largely mediated by numerous plasmid-borne carbapenemase genes. The following three classes of carbapenemases are commonly associated with the global spread of CRE: KPC (Ambler class A), metallo-β-lactamase (MBL) (Ambler class B, e.g., NDM), and OXA-48-like carbapenemase (Ambler class D) (2). These carbapenemase genes are generally carried by large conjugative plasmids that can horizontally transfer between different species and strains. Additionally, these plasmids frequently carry other resistance genes, resulting in multidrug or extensively drug-resistant phenotypes. Combating plasmid-mediated antimicrobial resistance is, therefore, of paramount importance in controlling and preventing the dissemination of carbapenem-resistant Enterobacteriaceae (3).

Novel strategies to combat antimicrobial resistance, especially carbapenem resistance, are urgently needed. Resistance plasmid curing, interruption of plasmid movement, and turning off resistance gene expression are all antiplasmid approaches that have been considered to combat multiple drug-resistant (MDR) bacteria and ideally remove resistance genes from the population without affecting the bacterial community (4). The CRISPR-Cas9 (clustered regularly interspaced short palindromic repeats–CRISPR-associated protein-9 nuclease) system, originally discovered as a bacterial immune system, has been engineered to create site-specific double-strand breaks (DSBs) for genome editing in a wide range of organisms, including mammalian cells, plants, fungi, and bacteria (4, 5). Accordingly, the CRISPR-Cas9 system provides a new tool and approach to eradicate carbapenem-resistant genes and plasmids. The RNA-guided DNA nuclease can mimic the natural bacterial defense against genetic targets and specifically cleave bacterial resistance and/or plasmid genes. Recently, a few Cas9-single guide RNA (sgRNA) systems have been developed for specific plasmid curing in bacterial hosts carrying multiple plasmids (6–8). These results showed that antibiotic-resistant bacteria can revert to antibiotic-susceptible bacteria by specifically cleaving resistance genes or plasmids using the CRISPR-Cas9 system. In this study, we developed a novel CRISPR-Cas9-mediated pCasCure plasmid-curing system as a proof-of-concept to resensitize CRE to carbapenems. The electrotransformation of pCasCure and arabinose-induced Cas9 expression demonstrated high efficiency to eliminate carbapenemase genes or plasmids in clinical Enterobacteriaceae isolates.

RESULTS

Construct of pCasCure.

The newly constructed pCasCure vector contains a pMB1/pUC ori (9), belongs to the ColE plasmid incompatibility family, and can effectively replicate in numerous Enterobacteriaceae species. The pCasCure also harbors the sucrose-sensitive self-curing element (sacB) and sgRNA scaffold in combination with the Cas9 enzyme-coding gene (Fig. 1A). Expression of Cas9 is regulated by the arabinose-inducing promoter pBAD. The original kanamycin-resistant marker on pSGKp-km was replaced by rifampin- (arr-3) or apramycin-resistant genes (aac(3)-IV), allowing for selection and application in clinical multidrug-resistant isolates.

FIG 1.

Plasmid map of pCasCure-Rif and S1-PFGE result. (A) pCasCure-Rif plasmid containing the Cas9 gene with an l-arabinose inducible promoter pBAD, the sgRNA with the synthetic J23119 promoter, and the sacB gene for plasmid self-curing. F and R denote the locations of primers used for the sgRNA construct. (B) S1-PFGE analysis of K. pneumoniae strains 13001, Kp97_58, 53433, and 49210 and their corresponding target-cured strains. M, marker; Salmonella enterica serotype Braenderup H9812. Lanes 1 and 6, strain 13001; lane 2, 13001 IncFIIK repA cured; lane 3, 13001 IncFIB(pQIL) repB cured; lane 4, 13001 parA cured; lanes 5 and 7, 13001 bla_KPC cured; lane 8, strain Kp97_58; lane 9, Kp97_58 bla_KPC cured; lane 10, strain 53433; lane 11, 53433 bla_NDM cured; lane 12, strain 49210; lane 13, 49210 IncL repA cured; lane 14, 49210 bla_OXA-48 cured. The arrows show the locations of the IncFIIK pKpQIL-like (∼110 kb, lanes 1 and 6), IncN (∼70 kb, lane 8), IncX3 (∼46 kb, lane 10), and IncL pOXA-48-like (∼63 kb, lanes 12 and 14) plasmids.

Curing time and arabinose concentration.

We initially investigated bla_KPC target-curing efficiency in Klebsiella pneumoniae 13001 under different arabinose induction concentrations (0.01%, 0.1%, and 1%). The results showed that after 6 h of incubation, 89.0%, 98.1%, and 99.3% of bacterial populations, respectively, lost the carbapenem resistance phenotype as a result of bla_KPC gene curing (Fig. 2A). In addition, we examined the curing efficiency along with different incubation times. The results showed that under 0.1% arabinose treatment, over 60% of bacteria lost bla_KPC after 2 h, while the curing frequency reached 70.5%, 98.2%, and 100% at 4, 6, and 16 h, respectively (Fig. 2B). Taking both incubation time and curing frequency into consideration, in the following experiments, we used 0.1% arabinose and 6 h of induction time as the curing conditions.

FIG 2.

bla_KPC curing efficiency with different arabinose concentrations and incubation times. (A) The curing efficiency of bla_KPC at three different levels of arabinose (0.01%, 0.1%, 1.0%) at 6 h incubation. (B) The curing efficiency at induction time 0, 2, 4, 6, and 16 h with 0.1% arabinose induction. Each experiment was performed in triplicate. Data points represent the mean value of three biological replicates with error bars showing standard deviation (SD).

Elimination of carbapenemase genes and plasmids by CRISPR-Cas9 system.

We first examined whether plasmid curing can be achieved by targeting different plasmid genes. The pKpQIL-harboring strain 13001 was used as the test strain, and four sgRNAs, including the IncFIIK replication gene repA (pKpQIL_p007), and the pKpQIL IncFIB replication gene repB (pKpQIL_p051), plasmid partition gene parA (pKpQIL_p050), and the carbapenemase gene bla_KPC (pKpQIL_p016), (Table 1) were targeted. Under 0.1% arabinose induction and 6 h incubation, 98.6%, 100%, 99.3%, and 98.6% of the bacterial population lost their corresponding target genes. In addition, S1 pulsed-field gel electrophoresis (PFGE) analysis of selected cured strains showed that the removal of these target genes also resulted in the complete curing of the pKpQIL plasmid from K. pneumoniae 13001 (Fig. 1B). The results indicated that plasmid curing can be obtained through CRISPR-Cas9-mediated cleavage at different genes on the same plasmid.

TABLE 1.

Carbapenemase gene and plasmid curing

Strain	ST	Carbapenemase gene and/or plasmid	Reference or source	Target gene(s)	Target sequence(s) (N20)	Curing efficiency (%)b
Strains containing plasmid and resistance genes
K. pneumoniae 13001	258	blaKPC-2, IncFIIK-pKpQIL	25	IncFIIK repA	GGTTTTTAACCTGTGAATAG	98.6 ± 2.4
				IncFIB(pQIL) repB	ACTGACTTCAAGGTGTGGGT	100 ± 0
				parA	TAAGGTATTTCAACGTGGAA	99.3 ± 1.2
				blaKPC	CAATTTGTTGCTGAAGGAGT	98.6 ± 1.2
K. pneumoniae Kp97_58	111	blaKPC-2, IncN	10	IncN repA	GAGCAAGACTTTCATAAAGG	97.9 ± 2.1
				blaKPC	CAATTTGTTGCTGAAGGAGT	96.5 ± 2.4
E. coli 53433	167	blaNDM-5, IncX3	This study	IncX3 repA	AATGTTAGATGATGATATTT	99.3 ± 1.2
				blaNDM	CCCAACGGTGATATTGTCAC	97.2 ± 4.8
K. pneumoniae 49210	23	blaOXA-48, IncL- pOXA48	This study	IncL repA	ATATCATGCGTTTCTAAGCG	100 ± 0
				blaOXA-48-like	CACCAAGTCTTTAAGTGGGA	99.3 ± 1.2
				IncL repA and blaOXA-48-likec	ATATCATGCGTTTCTAAGCG and CACCAAGTCTTTAAGTGGGA	100 ± 0c
Strains containing resistance genes
E. hormaechei 34978	171	blaKPC-3	26	blaKPC	CAATTTGTTGCTGAAGGAGT	95.8 ± 2.1
E. xiangfangensis 34399	114	blaKPC-3	26	blaKPC	CAATTTGTTGCTGAAGGAGT	95.1 ± 2.4
E. coli 28009	131	blaKPC-2	27	blaKPC	CAATTTGTTGCTGAAGGAGT	97.2 ± 3.2
S. marcescens SmN01	a	blaNDM-7	28	blaNDM	CCCAACGGTGATATTGTCAC	94.4 ± 3.2
K. pneumoniae 51933	307	blaOXA-48	29	blaOXA-48-like	CACCAAGTCTTTAAGTGGGA	98.6 ± 1.2

^aNot examined.

^bThe values shown are the mean ± SD of three independent experiments.

^cThe double sgRNAs were generated by inserting the IncL sgRNA in pCasCure-OXA48-Apr at the PstI and XmaI restriction sites. The curing result was obtained at 4 h of incubation with 0.1% arabinose treatment.

We then examined the plasmid-curing efficiency by targeting plasmid replication genes from several CRE plasmids. Besides the two replicon genes in pKpQIL, we also designed the sgRNAs against plasmid replication genes in IncN, IncX3, and IncL plasmids. We examined the IncN replicon-curing efficiency in K. pneumoniae sequence type 111 (ST111) strain Kp97_58, which harbors bla_KPC on the IncN plasmid pKp58_N (10). For the IncX3 replicon, we selected a bla_NDM-5-harboring ST167 Escherichia coli strain 53433, where the bla_NDM-5 is harbored on an IncX3 plasmid (data not shown). For the IncL replicon, we used the K. pneumoniae ST23 strain 49210, which harbors bla_OXA-48 on an IncL pOXA48-like plasmid. To test the system, specific sgRNAs against IncN, IncX3, and IncL plasmid replicons were designed and used to generate pCasCure-N-apr, pCasCure-X3-apr, and pCasCure-L-apr, respectively. The results showed that these pCasCure plasmids had a high frequency (>97%) of curing the three replicon genes (Table 1). S1-PFGE analysis showed that the curing of plasmid replicons resulted in the complete loss of the corresponding plasmids (data not shown).

We then examined the curing efficiency of the three most common carbapenemase genes, bla_KPC, bla_NDM, and bla_OXA-48-like. In addition to the bla_KPC sgRNA described above, we designed the sgRNAs against bla_NDM and bla_OXA-48-like genes. The specific sgRNA sequences were 100% conserved among all reported bla_KPC and bla_NDM variant sequences in the NCBI reference gene catalog (https://www.ncbi.nlm.nih.gov/pathogens/isolates#/refgene/#/refgene/), as well as common bla_OXA-48-like class D β-lactamase gene variants (bla_OXA-48, bla_OXA-162, bla_OXA-163, bla_OXA-181, bla_OXA-232, bla_OXA-244, bla_OXA-247, etc.). In this case, the three sgRNAs can be respectively used to eliminate bla_KPC, bla_NDM, and bla_OXA-48-like gene variants.

In this study, we tested the carbapenemase gene-curing efficiency for bla_KPC-2, bla_KPC-3, bla_NDM-5, bla_NDM-7, and bla_OXA-48 in various Enterobacteriaceae species of K. pneumoniae, E. coli, Enterobacter hormaechei, Enterobacter xiangfangensis, and Serratia marcescens. Among them, isolates were from certain clinically high-risk clones, e.g., K. pneumoniae ST258 and ST307, E. coli ST131, E. hormaechei ST171, commonly associated with the spread of carbapenem resistance in diverse clinical settings. The results showed that our pCasCure platform can effectively eliminate bla_KPC, bla_NDM, and bla_OXA-48-like gene variants in these clinical strains with a >94% curing frequency (Table 1).

S1-PFGE, next-generation sequencing, and susceptibility testing.

The plasmid-curing effects were examined in selected isolates, including K. pneumoniae Kp97_58, E. coli 53433, and K. pneumoniae 49210, and their carbapenemase gene-cured strains using S1-PFGE (Fig. 1B). Similar to the curing result in K. pneumoniae strain 13001, curing of bla_KPC and bla_NDM in Kp97_58 and 53433, respectively, caused the loss of the corresponding bla_KPC- or bla_NDM-harboring plasmids (Fig. 1B). However, curing of bla_OXA-48 in 49210 did not result in the loss of the IncL pOXA-48-like plasmid (Fig. 1B, lane 14), and a similar size of plasmid band was present in the bla_OXA-48-cured strain, despite the real-time PCR results being negative for bla_OXA-48.

In order to resolve the discrepancy between the PCR detection and S1-PFGE, the parental and a selected cured strain were sequenced using Illumina HiSeq. In the parental strain 49210, bla_OXA-48 is located in a Tn1999.2 element (Fig. 3). In comparison to the prototype bla_OXA-48-harboring Tn1999.1 transposon, Tn1999.2 contained a 767-bp IS1R that inserted and truncated the IS1999 element located upstream of bla_OXA-48, with the insertion site sequence AAGAATGTTA. Our designed N20 sequences were located at nucleotides (nt) 267 to 287 downstream of the start codon in bla_OXA-48. The sequence comparison revealed that an 816-bp element, from the IS1R downstream direct repeated sequences (AAGAATGTTG) to 5′ 672-bp bla_OXA-48 sequences, was deleted in the cured strain. The deleted element encompassed the CRISPR-Cas9 cleavage site and the real-time PCR target region, which explained our inability to cure the pOXA-48-like plasmid by bla_OXA-48-targeted pCasCure and the negative PCR screening result. In contrast, curing of the IncL replicon gene successfully removed the IncL pOXA-48-like plasmid (Fig. 1B, lane 13). In addition, we selected 8 more bla_OXA-48-cured 49210 colonies and sequenced the regions using primers spanning IS1R and lysR. The Sanger sequencing results showed that 5 of them carried the same 816-bp deletion as that described above, while 3 demonstrated another 836-bp deletion, which also removed a region of the bla_OXA-48 gene, the CRISPR-Cas9 cleavage site, and the real-time PCR target region (Fig. 3).

FIG 3.

Genetic comparison of the Tn1999.2 elements in K. pneumoniae 49210 and its bla_OXA-48-cured strain. The bla_OXA-48, IS1999, and IS1R genes are presented as red, green, and yellow arrows, while the rest of the genes are shown as orange arrows. Gray shading denotes shared regions of homology. The Cas9-mediated DNA cleavage site was marked by a scissor sign, with the N20 and protospacer adjacent motif (PAM, in red) sequences showing above the intact bla_OXA-48 sequences. The IS1R insertion sites were highlighted in blue and underlined, while the IS1R invert repeats were shown in blue italic and underlined. Small black arrowheads above bla_OXA-48 in the parental strain sequences denoted the locations of primers used for real-time PCR detection.

The above results suggested that IS1R-mediated recombination outside the CRISPR-Cas9 cleavage site resulted in bla_OXA-48 truncation and escaped plasmid curing in strain 49210. To examine this hypothesis, we then conducted bla_OXA-48 curing in K. pneumoniae strain 49202, of which the bla_OXA-48 was located within Tn1999.1 on a pOXA-48-like plasmid. IncL replicon and bla_OXA-48 PCR detection of the cured colonies showed negative results for both the IncL replicon and bla_OXA-48, indicating that the curing of bla_OXA-48 in 49202 simultaneously removed the pOXA-48-like plasmid. These results also suggested that transposase-mediated recombination may potentially interfere with CRISPR-Cas9-mediated plasmid curing.

We then examined the MIC of imipenem and meropenem against the parental CRE strains and their targeted cured strains (see Table S1 in the supplemental material). The results showed that the MICs of imipenem and meropenem in the cured isolates reduced ≥8-fold (from >16 μg/ml to <0.25 μg/ml), including the partial bla_OXA-48-cured strain. The results indicated that carbapenemase gene curing effectively restored the carbapenem susceptibility among these strains.

DISCUSSION

In this study, we developed a CRISPR-Cas9-based system to eliminate carbapenemase genes and plasmids in clinical Enterobacteriaceae isolates. Using this system, curing of an antibiotic resistance gene or plasmid can be achieved by transferring a pCasCure plasmid into a clinical CRE isolate, followed by arabinose induction. Confirmation of the carbapenemase gene removal can be examined by the loss of carbapenem resistance. We demonstrated the high curing efficiency and the wide applicability of our system by experimentally testing the system in clinical carbapenem-resistant K. pneumoniae, E. coli, E. hormaechei, E. xiangfangensis, and S. marcescens, including isolates from selected high-risk clones. Carbapenemase genes, bla_KPC, bla_NDM, and bla_OXA-48 gene variants, can be efficiently cured. Curing of the resistance genes correlated with the significant decrease of carbapenem MICs from >16 μg/ml to lower than 0.25 μg/ml, leading to resensitization of the antibiotic-resistant bacteria. Our results suggested that resistance gene- or plasmid-targeted curing, e.g., by the pCasCure system, provides a novel direction in resensitizing antibiotic susceptibility in clinically multidrug-resistant CRE strains.

These carbapenemase genes were mostly plasmid borne, which can horizontally transfer to different species and strains. Combating the spread of plasmid-mediated antibiotic resistance has become a priority and a challenge in controlling antimicrobial resistance. Besides directly curing the resistance genes, our pCasCure system also demonstrated the ability to specifically inactivate these epidemic plasmids. In this study, we showed that our pCasCure system can effectively remove several carbapenemase gene-harboring epidemic plasmids, including the bla_KPC-harboring IncFIIK-pKpQIL and IncN, bla_OXA-48-harboring pOXA-48-like, and bla_NDM-harboring IncX3 plasmids, by targeting their replication (repA and repB) and partitioning (parA in pKpQIL) genes. In addition, the cleavage of carbapenemase genes bla_KPC and bla_NDM in these plasmids can also remove the corresponding plasmids, presumably because the DSB caused by the highly efficient pCasCure cleavage was not able to be repaired.

However, the curing of bla_OXA-48 in K. pneumoniae 49210 failed to remove the IncL pOXA-48-like plasmid despite successfully deactivating OXA-48 by gene truncation. Our results suggested that the inability to cure pOXA-48-like plasmids is associated with the IS1R element located upstream of the bla_OXA-48 cleavage site. As most of these clinical multidrug-resistant plasmids are highly promiscuous and frequently contain multiple transposons, insertion sequence (IS) elements, and repeated regions, single nonessential gene-targeted curing may not be enough for effective plasmid curing. In this case, a more efficient approach would be to target its essential genes or to create multiple cleavages sites using a CRISPR array or more than one sgRNA (6, 11). In this study, the curing of the IncL replicon successfully removed the pOXA-48-like plasmid (Fig. 1B). In addition, we combined both the IncL replicon and bla_OXA-48 sgRNAs in the same pCasCure plasmid (by inserting the IncL sgRNA in pCasCure-OXA48-Apr at the PstI and XmaI restriction sites, Fig. 1A), and the result showed that at approximately 4 h, the pOXA-48-like plasmids were 100% cured, suggesting improved plasmid curing efficiency (Table 1).

In order to combat current plasmid-mediated resistance, multidrug-resistant plasmid curing is a new strategy that may curtail the spread of antibiotic resistance by interrupting the dissemination of resistance horizontal transfer in different populations of Enterobacteriaceae pathogens. However, we recognize that the ultimate challenge for plasmid or resistance gene curing in the clinical setting is the lack of an optimal deliver strategy to introduce the CRISPR-Cas9 system into target bacterial populations. Some studies have tried to use conjugation to deliver the CRISPR-Cas9 plasmid into the recipient bacteria (6, 8); however, a low transfer efficiency (usually <10⁻¹) (6) limits the utility of this strategy to neutralize resistance at a population level. Other studies have used phages as delivery vectors (5, 12), which could specifically eliminate a target bacterial genotype in a mixed population, both in vitro and in vivo (5, 12, 13). However, the wide application of a phage system is limited by its narrow host range and the emergence of phage resistance. Although we demonstrated that our pCasCure system is highly efficient in curing carbapenemase genes and plasmids in clinical CRE isolates, the delivery of pCasCure was artificially conducted through electroporation. Further work is needed to integrate pCasCure in an optimal delivery system to make it applicable for clinical intervention.

Taken together, we developed a novel CRISPR-Cas9-mediated pCasCure system for the curing of carbapenemase genes and plasmids. This platform is highly efficient in removing carbapenemase genes and plasmids in CRE isolates, thereby resensitizing CRE to carbapenems. With the integration of a safe and effective delivery system, CRISPR-Cas9-mediated carbapenemase gene and plasmid curing may serve as a potential tool to control the dissemination of carbapenem resistance in clinical pathogens.

MATERIALS AND METHODS

Bacterial strains and susceptibility testing.

The CRE isolates used for plasmid or gene curing are listed in Table 1. Escherichia coli DH10B was used as the bacterial host for the construction of the pCasCure plasmid. Bacterial strains were grown in lysogeny broth (LB) with shaking at 200 rpm or on LB agar plate at 37°C. When appropriate, antibiotics were added to the growth medium at the following concentrations: rifampin (Rif) at 100 μg/ml and apramycin (Apr) at 30 μg/ml.

Imipenem and meropenem susceptibility testing (AST) was performed using a standard broth microdilution method following CLSI guideline (14). The testing was performed in duplicate on two different days. Quality control (QC) strains Pseudomonas aeruginosa ATCC 27853 and Escherichia coli ATCC 25922 were used in the testing.

Construction of pCasCure plasmid.

The pCasCure plasmid was constructed as follows. The plasmid pSGKp-km (9), containing a single artificial chimeric guide RNA (sgRNA) under the control of the synthetic constitutive J23119 promoter, a sucrose-sensitive sacB gene, and a kanamycin resistance marker, was used as the backbone to construct pCasCure. Initially, the kanamycin resistance gene was replaced by rifampin resistance gene arr-3, amplified from a clinical isolate, Kp202 (15), to generate plasmid pSGKp-rif. The gene encoding the Cas9 nuclease was amplified from plasmid pCasKp (9) and cloned into NotI and XbaI sites in pSGKp-rif, resulting in the plasmid pSGKp-Cas-rif. Then the araC gene and arabinose-inducible promoter P_BAD were amplified from pBAD24 (16) and subsequently inserted into a NotI-digested plasmid pSGKp-Cas-rif, generating the plasmid pCasCure-rif. In order to make this plasmid-curing system applicable for clinical rifampin-resistant isolates, we also replaced the arr-3 gene in pCasCure-rif with the apramycin-resistant gene aac(3)-IV using the In-Fusion cloning kit (TaKaRa Bio USA, Inc.), creating pCasCure-apr. The sacB gene in the pCasCure plasmids encodes a levansucrase, which hydrolyzes sucrose and glucose to fructose, resulting in the production of a toxic fructan that kills the recombinants containing the suicide plasmid (17). Under the selection of sucrose, the pCasCure plasmid can be easily removed for scarless genetic modification. The pCasCure-rif vector is schematically shown in Fig. 1A.

sgRNA design and cloning.

The 20-nt base-pairing region (N20) of an sgRNA was designed through an online web server (https://eu.idtdna.com/site/order/designtool/index/CRISPR_CUSTOM), followed by off-target specificity examination in Geneious 11.5 (18) against RefSeq representative genomes (https://www.ncbi.nlm.nih.gov/assembly/) of Klebsiella (n = 7), Enterobacter (n = 18), Escherichia (n = 5), Shigella (n = 4), Citrobacter (n = 14), and Salmonella (n = 2) species. The sgRNA fragment, flanked by SpeI and XbaI restriction sites, along with the N20 (Table 1), were amplified with forward primer AATACTAGT-N20-GTTTTAGAGCTAGAAATAGC (SpeI restriction site is underlined) and reverse primer CTGGTT TCTAGA ACTAGTGGA (XbaI restriction site is underlined) (Fig. 1A, primer F and R), using the pCasCure-rif plasmid as the template. The PCR product was subsequently inserted into the XbaI- and SpeI-digested pCasCure-rif or pCasCure-apr plasmid, generating the final pCasCure plasmids with targeted sgRNA.

Curing condition—incubation time and arabinose concentrations.

We then evaluated the target curing efficiency under different arabinose induction concentrations and incubation time periods. K. pneumoniae strain 13001, harboring bla_KPC-2 in a pKpQIL-like plasmid, was used as the test strain to examine the curing efficiency of bla_KPC (19). An sgRNA specific to bla_KPC was generated as described above (Table 1) and used to construct the plasmid pCasCure-apr-KPC for bla_KPC curing. pCasCure-apr-KPC was electroporated into K. pneumoniae strain 13001, followed by selection on agar plates supplemented with 30 μg/ml apramycin and PCR confirmation. Overnight culture of pCasCure-apr-KPC-positive K. pneumoniae 13001 was diluted 100-fold in 5 ml LB containing 0.01%, 0.1%, or 1% arabinose in combination with 30 μg/ml apramycin. The cultures were incubated for 6 h at 37°C with shaking at ∼200 rpm. After 6 h, the cultures were diluted and simultaneously plated on LB agar plates containing 1 μg/ml doripenem plus 30 μg/ml apramycin or 30 μg/ml apramycin only to determine the number of CFU. The curing efficiency was calculated as follows: (1 − colonies grown on doripenem and apramycin/colonies on apramycin plate) × 100%. The experiments were conducted using three randomly picked pCasCure-apr-KPC-positive K. pneumoniae 13001 colonies and repeated on two different days with two different operators. A bla_KPC real-time PCR (20) was conducted to confirm the colony counting results.

In addition, target curing efficiency under different arabinose incubation time was evaluated. Similarly, the overnight cultures of pCasCure-apr-KPC-harboring 13001 were diluted 100-fold in 5 ml LB containing 0.1% arabinose and 30 μg/ml apramycin, followed by incubation at 37°C with shaking for 16 h. At 0, 2, 4, 6, and 16 h, 100 μl of culture was removed, diluted, and plated out for colony counting. The curing efficiency was calculated as above.

Carbapenemase gene- and plasmid-curing experiment.

The pCasCure plasmid with different target sgRNA (pCasCure-gRNA) was generated as described above and introduced into a chosen strain by electroporation, followed by selection on agar containing rifampin (100 μg/ml) or apramycin (30 μg/ml). Single colony transformants were selected and checked for the presence of the pCasCure-sgRNA by PCR using primers ara405-F (TGTGGAATTGTGAGCGGATA) and ara405-R (CGCCAGCAGTTAGGGATTAG) (targeting the gene araC on pCasCure). Positive colonies were selected and inoculated in 5 ml LB containing rifampin (100 μg/ml) or apramycin (30 μg/ml) and 0.1% arabinose. After incubation for 6 h at 37°C with shaking (∼200 rpm), the culture was diluted and simultaneously plated on LB agar containing rifampin (or apramycin) with or without 1 μg/ml doripenem. In addition, ∼48 colonies grown on rifampin (or apramycin) plates were subject to PCR detection to confirm the loss of targeted carbapenem-resistant genes or plasmids. The curing efficiency was calculated based on the PCR detection results. For self-curing the pCasCure plasmid, single colonies of the resistant gene- or plasmid-cured strain were streaked onto an LB agar plate containing 5% sucrose and incubated at 37°C overnight, followed by colony PCR using primers ara405-F and ara405-R to confirm the successful self-curing of pCasCure-sgRNA.

S1 pulsed-field gel electrophoresis.

S1-PFGE was used to examine the effect of plasmid curing on selected strains. In brief, the parental and cured strains were digested with S1 nuclease after being embedded in 1% SeaKem gold agarose gels, followed by plasmid DNA separation by PFGE as described previously (21). Salmonella enterica serotype Braenderup H9812 was used as a size marker.

Next-generation sequencing.

Selected parental and cured strains were subject to next-generation sequencing (NGS) using Illumina HiSeq. Sequencing data were trimmed using Trimmomatic v0.36 (22) followed by de novo assembly using SPAdes v3.11.1 (23). The resulting contigs were examined using Mauve (24) to compare the variations between parental and cured genomes.

Accession number(s).

The complete nucleotide sequences of pCasCure-Apr and pCasCure-Rif were deposited in GenBank under accession numbers MT262892 and MT262893, respectively. The raw sequences of strains 49202, 49210, 53433, and bla_OXA-48-cured 49210 were deposited in the NCBI Sequence Read Archive under BioSample accessions SAMN15146842, SAMN15146843, SAMN15146844, and SAMN15149925, respectively.