A dual-plasmid CRISPR/Cas9-based method for rapid and efficient genetic disruption in Mycobacterium abscessus

ABSTRACT

Mycobacterium abscessus is increasingly recognized for causing infections that are notoriously difficult to treat, owing to its large arsenal of intrinsic antibiotic resistance mechanisms. Tools for the genetic manipulation of the pathogen are critical for enabling a better understanding of M. abscessus biology, pathogenesis, and antibiotic resistance mechanisms. However, existing methods are largely recombination-based, which are relatively inefficient. Meanwhile, CRISPR/Cas9 has revolutionized the field of genome editing including its recent adaptation for use in mycobacteria. In this study, we report a streamlined and efficient method for rapid genetic disruptions in M. abscessus. Harnessing the CRISPR1 loci from Streptococcus thermophilus, we have developed a dual-plasmid workflow that introduces Cas9 and sgRNA cassettes in separate steps but requires no other additional factors to engineer mutations in single genes or multiple genes simultaneously or sequentially using multiple targeting sgRNAs. Importantly, the efficiency of mutant generation is several orders of magnitude higher than reported for homologous recombination-based methods. This work, thus, reports the first application of CRISPR/Cas9 for gene editing in M. abscessus and is an important tool in the arsenal for the genetic manipulation of this human pathogen.

IMPORTANCE

Mycobacterium abscessus is an opportunistic pathogen of increasing clinical importance due to its poor clinical outcomes and limited treatment options. Drug discovery and development in this highly antibiotic-resistant species will require further understanding of M. abscessus biology, pathogenesis, and antibiotic resistance mechanisms. However, existing methods for facile genetic engineering are relatively inefficient. This study reports on the first application of CRISPR/Cas9 for gene editing in M. abscessus using a dual-plasmid workflow. We establish that our method is easily programmable, efficient, and versatile for genetic disruptions in M. abscessus. This is a critical advancement to facilitating targeted gene function studies in this emerging pathogen.

INTRODUCTION

Reports of nontuberculous mycobacteria infections have risen over the past few decades (1–3). In particular, the prevalence of Mycobacterium abscessus infections has been increasing. Such infections were previously mainly reported in cystic fibrosis patients but are now increasingly reported in patients without any risk factors, manifesting as either pulmonary or extrapulmonary disease (2–7). Found ubiquitously in the environment, M. abscessus can cause a wide range of infections and is also notoriously difficult to treat, underscored by low treatment success rates (20%–50%) despite long-term combination therapy. This failure is due to its large intrinsic arsenal of antibiotic resistance mechanisms including multiple efflux pumps and an array of target- and drug-modifying enzymes (4–8). Taken together, M. abscessus has thus been frequently described as an “incurable nightmare” (7, 8), emphasizing its clinical relevance as an emerging pathogen.

Tools for genetic manipulation of M. abscessus are needed to gain a better understanding of M. abscessus biology, pathogenesis, and antibiotic resistance mechanisms. While transposon-sequencing is a viable approach for genome-wide studies, homologous recombination-based methods have been the primary tool for producing genetic deletion mutants in many bacterial species including M. abscessus (9–14). In mycobacteria, homologous recombination using allelic exchange mutagenesis or recombineering has been used extensively to generate gene deletions, although the efficiency is suboptimal with the frequency of obtaining a desired mutation ranging from 10^–6 to 10^–4 (11, 12). Moreover, in M. abscessus, high rates of spontaneous background resistance to the antibiotic selection marker, in combination with the low success rates of homologous recombination, result in a cumbersome process requiring the screening of a lot of clones to find the rare, desired correct mutant (9–12). These technical challenges have motivated the development of improvements or alternate methods. For instance, an approach incorporating tdTomato as a fluorescent marker along with an antibiotic resistance gene for selection was used to accelerate the identification of successful recombinants over background spontaneous antibiotic resistance (15). The use of negative selection markers can further facilitate two-step allelic exchange for the generation of nonmarked gene deletions. However, sacB, which is often used to confer sucrose sensitivity, does not work in M. abscessus (10). Instead, alternate markers such as Mycobacterium tuberculosis katG or Escherichia coli galK genes have been used to confer susceptibility to isoniazid or 2-deoxygalactose, respectively, for counterselection (16, 17). Nonetheless, these recombination-based techniques are often laborious and have suboptimal efficiencies, necessitating extensive screening of mutants to identify successful recombination events.

Recently, clustered regularly interspaced palindromic repeats (CRISPR)/CRISPR-associated protein (Cas) technologies have rapidly emerged as a revolutionary tool for highly programmable genome editing and have been extensively developed, especially for eukaryotes (18). The expression of Cas9 and a targeting sgRNA leads to double-strand breaks (DSBs) in the genome. In the absence of DNA repair, cell death occurs, while in the presence of nontemplated repair mechanisms, indels are formed at the DSB site, which can affect the function of the gene product (Fig. 1A). In cases where a frameshift occurs, the formation of a premature stop codon early enough in the gene leads to a loss of function. This form of genetic disruption is advantageous as it is less prone to polar effects such as those that can arise from the introduction of large selectable markers at the target locus in some homologous recombination strategies.

Fig 1.

Overview of the dual-plasmid CRISPR/Cas9 workflow. (A) Sth1Cas9 mediates DSBs at the target site, which leads to either bacterial death in the absence of DNA repair or inaccurate indels upon nontemplated DNA repair. Created with BioRender. (B) M. abscessus was first transformed with the inducible Cas9 on an integrative plasmid pDN-Sth1Cas9 containing the tetracycline repressor TetR and resistance selection marker Kan^R. This Cas9-background strain was then transformed with the sgRNA cassette on a mCherry-expressing plasmid pDN-Cherry-sgRNA containing the resistance selection marker Hyg^R. The induction of the CRISPR system leads to gene disruption and loss of function. Finally, curing the mCherry-sgRNA plasmid creates the edited strain still carrying the inducible Cas9 to allow the subsequent introduction of additional sgRNA cassettes.

Various CRISPR/Cas-based systems (Francisella novicida Cas12a FnCpf1, Streptococcus thermophilus Sth1Cas9) have been studied for gene editing in mycobacteria including Mycobacterium smegmatis, Mycobacterium marinum, and M. tuberculosis (19–21). Some of these tools have also been repurposed for CRISPR interference in mycobacterium (22–24). However, these methods for genetic disruption have not been shown to be translatable across all mycobacterial species and in some cases require the co-expression of additional proteins to aid DNA repair. Hence, it remains unknown if inter-species differences in endogenous repair machinery would impair the efficacy and usability of CRISPR/Cas-genetic disruption in M. abscessus.

In this study, we show for the first time that M. abscessus does possess sufficient endogenous repair functions to allow genetic disruption using Sth1Cas9 without the requirement of any additional factors. We describe a streamlined and efficient two-step approach to introduce the inducible Cas9 and guide RNAs (sgRNA) separately (Fig. 1B) to enable efficient generation and recovery of desired mutants. By targeting known antibiotic resistance genes, we show the generation of frameshift indels that lead to functional gene knockout with 10²–10⁴ higher efficiency than reported for homologous recombination-based methods. Moreover, we demonstrate that this system can generate multiple gene disruptions by introducing two sgRNAs simultaneously or single sgRNAs sequentially. Finally, introducing two sgRNAs in proximity also facilitated the deletion of larger genomic regions.

RESULTS

Design of a dual-plasmid CRISPR/Cas9 strategy for M. abscessus

We designed a strategy that separates the Cas9 and sgRNA cassettes onto two separate plasmids to minimize any toxicity associated with DSBs that can occur from the leaky expression of a targeting guide with Cas9 in a single integrative plasmid system (20, 21) (Fig. 1A). We placed the sgRNA cassette on an episomal mCherry-expressing plasmid to facilitate the rapid visual identification of both successful sgRNA/mCherry-containing transformants as well as strains that have been cured of the sgRNA plasmid after genetic knockouts have been generated (Fig. 1B). We then encoded active Cas9 under control of the tetracycline repressor (TetR), rendering it inducible by the addition of anhydrotetracycline (AHT), on an integrating plasmid (pDN-Sth1Cas9) that also encodes a kanamycin (KAN) resistance marker (Kan^R). This plasmid also encodes an L5 integrase protein for site-specific integration into the genomic attB site as a single copy. Transforming this plasmid into M. abscessus created a Cas9-background strain, which was selected on KAN and verified using colony polymerase chain reaction (PCR) (Fig. S1A). Of note, only 50% of picked colonies (6 out of 12) were successful transformants while the rest were mutants that had acquired KAN resistance without possessing the Cas9 construct (Fig. S1B and C). The high rate of spontaneous resistance observed with the Cas9 integrating plasmid further reinforced our desire to utilize two marker genes on the second sgRNA plasmid to aid the identification of correct sgRNA-containing clones in the second step (Fig. S2).

This second episomal plasmid (pDN-Cherry-sgRNA) encodes a hygromycin (HYG) resistance marker (Hyg^R), a constitutively expressed mCherry under the control of the strong mycobacterial promoter P_smyc (25), and the AHT-inducible sgRNA cassette. The original Golden Gate cloning site from pJR962 was also cloned together with the sgRNA cassette to allow for multiple sgRNAs to be cloned onto the same plasmid (23, 26). Guides for sequences of interest are cloned into the sgRNA/mCherry-plasmid and introduced into the M. abscessus-Cas9 strain using HYG selection with successful transformants easily identifiable as pink colonies (Fig. 1B; Fig. S2).

Based on the previously reported expanded set of 15 protospacer adjacent motifs (PAMs) defined for Sth1Cas9 (23, 26), we identified 112,869 targetable sites in the genome of M. abscessus ATCC 19977 (or ~1 in every 45 bp) (Table S3). As Cas9-mediated DSBs will occur regardless of targeting the template or nontemplate strand, this number of sites translates to a very high targeting coverage of 99.9% [4,917 out of 4,920 open reading frames (ORFs)], with 94.8%–98.8% of all ORFs (4,662 and 4,862 ORFs, respectively) containing a targetable site within the first 25%–50% of their length (Table S4). Targeting sites in these regions is more optimal for generating frameshifts that result in premature stop codons leading to loss of function (null mutations) as opposed to mutations in more distal regions of the gene, which can lead to the expression of truncated proteins that still retain some function (hypomorphic mutations). The remaining outstanding genes could theoretically also be knocked out through the introduction of two or more sgRNAs to mediate the deletion of a larger gene region.

Validation of CRISPR/Cas9 for genetic disruption in M. abscessus

To test the use of Cas9 in M. abscessus, we introduced three different classes of sgRNA guides: (i) a nontargeting control guide (Ctrl, NT) with a sequence not found on the M. abscessus genome, (ii) a guide targeting the essential (ES) gene encoding RNA polymerase B (rpoB, MAB_3869c), and (iii) guides targeting nonessential (NE) genes known to mediate antibiotic resistance. These latter resistance genes were chosen to provide a clear phenotype, wherein genetic disruption would confer antibiotic sensitization. The beta-lactamase (encoded by bla, MAB_2875) mediates resistance against beta-lactams such as ampicillin (27), the rifamycin ADP-ribosyl transferase (encoded by arr, MAB_0591) mediates resistance against rifampicin (28), and the previously reported Eis (enhanced intracellular survival) acetyltransferase homolog (encoded by eis2, MAB_4532c) confers intrinsic resistance to aminoglycosides such as capreomycin and amikacin (17).

Bacterial survival following the induction of Cas9 and sgRNA differed depending on the class of sgRNA expressed. For example, induction of the nontargeting control sgRNA guide resulted in no observable bacterial death compared to uninduced conditions (Fig. 2A). In contrast, as expected, a significant reduction in survival was observed when disrupting an essential gene with the rpoB sgRNA [>99.95% or >3-log fold colony-forming unit (CFU) reduction]. This finding is consistent with bacterial death resulting from unrepaired Cas9 DSBs or nontemplated DNA repair in the rpoB gene resulting in indels that are not tolerated in essential genes (18, 20, 21). Meanwhile, disruption of nonessential genes that mediate antibiotic resistance also resulted in reduced survival (~90%–98% or 1- to 2-log fold reduction), albeit to a lesser extent than rpoB (1- to 2-log improved survival compared to rpoB), suggesting that indels resulting from nontemplated DNA repair in nonessential genes are better tolerated. Overall, the reduced bacterial survival seen with the expression of both the Cas9 and targeting sgRNA cassette further justifies our rationale to separate the sgRNA on a second plasmid to facilitate subsequent removal after the induction to prevent any toxicity from the leaky expression of the targeting sgRNA.

Fig 2.

Validation of CRISPR/Cas9 in M. abscessus. (A) Bacterial survival upon induction of Cas9 in strains containing nontargeting control sgRNA Ctrl (NT) and targeting sgRNAs against the NE genes bla, arr, and eis2 or against the ES gene rpoB. Survival (%) was calculated relative to CFUs in uninduced bacteria. Bars represent the mean values and standard deviations from three technical replicates. (B) Outcomes of DSB repair at the target sites following Cas9 induction. Colonies from strains with bla, arr, and eis2-targeting sgRNAs were picked for targeted PCR and sequencing after Cas9 induction. The numbers of colonies picked for each strain are stated in parenthesis (n = 8 for bla and 6 for arr and eis2-sgRNA strains). Bars represent the frequency of outcomes relative to the number of colonies picked.

We analyzed six to eight surviving mutants resulting from sgRNAs targeting each of the resistance genes by PCR amplifying targeted regions and sequencing (Fig. 2B). All analyzed clones across the three genes contained some form of editing indicating that Cas9-mediated DSBs had occurred in each mutant. Additionally, a high frequency (67%–88%) of analyzed mutants was found to possess frameshift mutations resulting in early termination with the formation of premature stop codons. For the bla sgRNA, seven out of eight clones (88%) contained frameshift deletions while the final clone contained an in-frame deletion. For the arr and eis2 sgRNAs, frameshift indels were found in five (83%) and four (67%) clones, respectively, out of six analyzed mutants. One clone from the eis2 sgRNA contained a large deletion >100 bp, while the remaining clones from both arr and eis2 sgRNA-containing strains could not be PCR-amplified, suggesting the possibility that larger deletions had occurred leading to PCR failure. These long deletions may disrupt other regions upstream/downstream of the gene of interest and are, therefore, undesirable.

To confirm that loss of function of the respective resistance gene products confers antibiotic hypersensitivity, we determined the minimum inhibitory concentrations (MICs) for select mutants against ampicillin, rifampicin, and capreomycin (Table 1; Fig. S3). Wild-type (WT) M. abscessus is completely resistant to ampicillin [MIC₉₀ >50 µg/mL (highest concentration tested)], but antibacterial activity was observed in strains harboring frameshift indels in bla (MIC₉₀ = 3.13–6.25 µg/mL), thus phenotypically confirming the genotype determined by sequencing. Similarly, strains harboring frameshift indels in arr were sensitized to rifampicin [MIC₉₀ ≤0.049 µg/mL (lowest concentration tested)] compared to WT (MIC₉₀ = 3.13 µg/mL), and strains with frameshift indels in eis2 were sensitized to capreomycin (MIC₉₀ = 3.13 µg/mL) compared to WT (MIC₉₀ = 50 µg/mL).

TABLE 1.

Editing outcomes and MIC values for representative M. abscessus strains with single gene disruptions

sgRNA	Gene length (base pairs)	Predicted cut site positiona	Strain	Editing outcome	WGS	Antibiotic	MICb (μg/mL)
bla (MAB_2875)	870	28/29	Wild-type	N/A	Yes	Ampicillin	>50c
			Clone 1	2 bp del	Yes		6.25
			Clone 2	1 bp del	Yes		3.13
			Clone 4	11 bp del	No		3.13
arr (MAB_0591)	426	54/55	WT	N/A	Yes	Rifampicin	3.13
			Clone 2	2 bp del	Yes		≤0.049d
			Clone 3	1 bp ins	Yes		≤0.049d
			Clone 4	1 bp ins	No		≤0.049d
eis2 (MAB_4532 c)	1,236	87/88	WT	N/A	Yes	Capreomycin	50
			Clone 1	22 bp del	No		3.13
			Clone 2	5 bp del	Yes		3.13
			Clone 5	14 bp del	Yes		3.13

^a The predicted cut site position numbered based on each gene of interest with 1 being the first nucleotide on the ORF.

^b Representative MIC curves are shown in Fig. S3.

^c Highest concentration tested.

^d Lowest concentration tested.

Using whole genome sequencing (WGS) in select strains, we found no additional indels suggesting that Cas9 induction did not lead to undesirable off-target editing (Table 1). Overall, this validates the ability of using CRISPR-Cas9 to generate frameshift indels leading to loss of gene function in M. abscessus in a highly specific manner. Accounting for reduced bacterial survival upon induction of Cas9 and a targeting sgRNA, we calculated the frequency of desired outcomes (bacterial survival × frameshift frequency) to be between 1.1% and 9.3% (Table 2). This high frequency greatly reduces the number of required starting bacteria and resulting surviving colonies needed to identify desired mutants, justifying the use of CRISPR/Cas9 for efficient genetic disruption in M. abscessus.

TABLE 2.

Comparison of efficiency of generating frameshift mutations from each category of gene disruption

Category	Strain + sgRNA	Bacterial survival (%)a	Clones analyzed	Frameshift frequency (%)	Overall frequency of desired outcomes (%)b
Single gene disruption	WT + bla sgRNA	10.6	8	87.5	9.3
	WT + arr sgRNA	9.1	6	83.3	7.6
	WT + eis2 sgRNA	1.6	6	66.7	1.1
Simultaneous gene disruption	WT + (bla + arr) double sgRNA	0.19	12	58.3c	0.11
Sequential gene disruption	Δbla-1 + arr sgRNA	5.7	6	33.3	1.9
	Δbla-2 + arr sgRNA	4.8	6	66.7	3.2
	Δbla Δarr + eis2 sgRNA	0.56	6	66.7	0.32

^a Bacterial survival (%) calculated from CFUs from Cas9-induced bacteria relative to CFUs in uninduced bacteria. Mean of three technical replicates indicated.

^b Overall frequency of desired frameshift outcomes calculated as (bacterial survival × frameshift frequency).

^c Frameshift frequency of both genes.

Disruption of multiple genes via simultaneous or sequential introduction of multiple sgRNAs

There is increasing interest in generating more complex mutants containing multiple gene disruptions to study gene–gene interactions. With the high efficiency of obtaining single gene disruptions, we thus sought to establish two complementary workflows for generating multiple gene disruptions through introducing guides either simultaneously (Fig. 3A) or sequentially (Fig. 3B).

Fig 3.

Multiple gene disruptions through a simultaneous or a sequential editing strategy. (A) In the simultaneous editing strategy, a mCherry-sgRNA plasmid containing multiple sgRNAs is introduced into the WT-Cas9 strain to disrupt multiple genes concurrently. (B) In the sequential editing strategy, successive rounds of transformation of single sgRNAs, Cas9 induction, and plasmid curing are performed to achieve the desired multiple gene knockout mutant. (C and D) Outcomes of DSB repair at the target sites following Cas9 induction of two sgRNAs in the simultaneous editing strategy. Twelve colonies were picked for targeted PCR and sequencing for both bla and arr. Genetic outcomes for each clone represented at each individual gene (C), as well as overall at both target genes (frameshift in both genes, frameshift in only one gene, and long deletions in either gene) (D). Bars represent the frequency of outcomes relative to the number of colonies picked. (E and F) Outcomes of DSB repair at the target sites following Cas9 induction of single sgRNAs in the sequential editing strategy. Two Δbla clones were induced with an arr-targeting sgRNA (E), while one Δbla Δarr clone was induced with an eis2-targeting sgRNA (F). Six colonies each were picked for targeted PCR and sequencing. Bars represent the frequency of outcomes relative to the number of colonies picked.

First, using the Golden Gate handles from the original sgRNA cassette, we cloned a plasmid containing sgRNAs targeting both bla and arr and introduced this into the WT Cas9-integrated background strain to generate a double knockout in a single step (Fig. 3A). As expected with the presence of two sgRNAs, Cas9 induction led to an almost 2-log lower survival than seen with either guide alone with CFU reductions of 99.8% due to unrepaired DSBs (Fig. S4; Fig. 2A). Twelve surviving colonies were picked for targeted colony PCR and sequencing of each gene (Fig. 3C). Ninety-one percent (11/12) of selected clones contained frameshift mutations in bla, while 67% (8/12) of clones contained frameshift mutations in arr. These frameshift rates were similar to those seen in strains generated from each sgRNA individually, suggesting that DSB and repair at each target site are largely independent of each other. Overall, 58% of clones (7/12) contained frameshift mutations in both target genes, 17% (2/12) had frameshift mutations in only one gene, and 25% (3/12) had PCR failures in arr suggestive of potentially undesirable long deletions beyond the target site (Fig. 3D).

Selected clones containing mutations in both genes were phenotypically confirmed to contain loss of function in both bla and arr gene products as evidenced by increased sensitivity to both ampicillin and rifampicin (Table 3; Fig. S5A and B). Hence, the double knockout mutant Δbla Δarr could be easily generated through the introduction of two sgRNAs concurrently on a single plasmid.

TABLE 3.

Editing outcomes and MIC values for representative M. abscessus strains with more complex editing

sgRNA	Strain background	Strain	Gene	Editing outcome	Antibiotic	MICa (μg/mL)
bla + arr (simultaneous)	WT	WT	bla	N/A	Ampicillin	>50b
			arr	2 bp del	Rifampicin	2.5
		Clone 5	bla	1 bp del	Ampicillin	12.5
			arr	N/A	Rifampicin	≤0.078c
		Clone 7	bla	2 bp del	Ampicillin	12.5
			arr	1 bp ins	Rifampicin	≤0.078c
arr (sequential)	Δbla	WT-arr	arr	N/A	Rifampicin	3.13
	Clone 1	Clone 2		1 bp ins		≤0.049c
		Clone 6		1 bp ins		≤0.049c
eis2 (sequential)	Δbla Δarr	WT-eis2	eis2	N/A	Capreomycin	50
		Clone 2		16 bp del		3.13
		Clone 6		8 bp del		3.13
arr (two sgRNAs 324 bp apart)	WT	WT	arr	N/A	Rifampicin	3.13
		Clone 5		328 bp del		0.02
		Clone 9		326 bp del		0.02

^a Representative MIC curves are shown in Fig. S5.

^b Highest concentration tested.

^c Lowest concentration tested.

The overall frequency of obtaining double Δbla Δarr mutants in this case was 0.11%, which was, as expected, at least an order of magnitude lower than seen previously with single guides alone (Table 2). As survival decreases with the induction of multiple guides, this may limit the success of obtaining and successfully identifying mutants with desired mutations, particularly if disruption of more than two genes is targeted. We, thus, sought to demonstrate a sequential workflow to generate iterative genetic knockouts to increase the success rate in obtaining strains with disruption of three or more genes (Fig. 3B).

We passaged previously generated Δbla strains on media lacking HYG to visually identify mutants that had lost their pink coloration and, thus, the bla sgRNA mCherry encoding plasmid (Fig. 1B). After plating cultures in the absence of selection pressure, white colonies accounted for approximately 60%–90% of all colonies, indicating that curing occurs at a reasonable frequency (unpublished data). We then performed a new transformation with the mCherry plasmid now carrying an arr-targeting sgRNA before inducing Cas9-sgRNA expression (Fig. 3B). We analyzed six resulting colonies for each of two Δbla strains (Fig. 3E) and found that half of all examined clones contained frameshift indels in arr (2/6 for Δbla-1 and 4/6 for Δbla-2), while in-frame indels and large deletions were observed in the remaining clones. Selected colonies previously verified to be ampicillin-sensitive were now verified to have increased sensitivity to rifampicin compared to the corresponding parent arr-wild-type strains (Table 3; Fig. S5C).

To demonstrate the efficient ability to disrupt three genes, we performed an additional round of curing and transformation with the ampicillin- and rifampin-sensitive mutants (Δbla Δarr) to generate sensitivity to capreomycin using an eis2-targeting sgRNA. Targeted PCR and sequencing confirmed frameshifts in 4/6 clones (Fig. 3F). MICs of these strains showed increased sensitivity to capreomycin (Table 3; Fig. S5D), while also retaining sensitivity to ampicillin and rifampicin, confirming the formation of a triple knockout mutant (Δbla Δarr Δeis2).

While, in principle, the simultaneous one-step editing strategy may require a shorter time to obtain mutants with multiple gene disruptions than the sequential editing strategy, this advantage is offset by a decrease in the frequency of recovering desired mutants if multiple loci are targeted with different sgRNAs on a single plasmid. For example, simultaneously disrupting both bla and arr was at least an order of magnitude less efficient than the sequential strategy (Table 2). Thus, when targeting three or more loci, the sequential strategy is preferred to increase the frequency of obtaining desired mutants.

Generation of large deletions via introduction of two adjacent sgRNAs

Finally, we sought to investigate if introducing two sgRNAs spaced within some reasonable proximity could mediate a targeted deletion of a larger gene region (Fig. 4). A plasmid containing two sgRNAs targeting arr spaced 324 bp apart was introduced into the WT Cas9-background strain. We characterized 12 independent mutants following Cas9 induction. While a precise deletion of the 324 bp region between the predicted cut sites was not observed among the 12 clones, we did observe longer deletions in 91% of cases (11/12) with nucleotides deleted beyond the target sites (Fig. 4). Five of these clones (325–424 bp deleted) also had frameshifts resulting in premature stop codons, which were phenotypically confirmed to have increased sensitivity to rifampicin (Table 3; Fig. S5E). Half of all clones (6/12) had more extensive deletions greater than 100 bp outside the target sites (>424 bp deleted). Interestingly, an inversion event was also observed in the final clone where the region between the two target sites was religated in the opposite orientation with small deletions (Fig. 4).

Fig 4.

Generation of larger targeted deletions through concurrent Cas9 induction with two adjacent targeting sgRNAs. (A) Two sgRNAs spaced 324 bp apart on the arr gene were introduced to generate a targeted deletion of a larger region between the two DSB sites. Twelve colonies were picked for targeted PCR and sequencing with the number of clones for each observed outcome listed in parenthesis in (B–D). (B) Accurate deletion of 324 bp (gray region and purple PAM 2 site) would be theoretically expected but was not observed in any of the 12 colonies. (C) Longer deletions (>324 bp) occurred with the removal of additional nucleotides beyond the two DBS sites in 11 of 12 colonies. Six of these contained more extensive deletions ≥100 nucleotides beyond the predicted cut sites. (D) An inversion event was also observed in one clone with the targeted region being excised and religated in the opposite orientation (purple PAM 2 site and gray region now in reversed orientation as red PAM 1 site). Created with BioRender.

DISCUSSION

Current methods for genetically engineering M. abscessus have been largely limited to suboptimal and tedious recombination-based methods due to the high rates of spontaneous background antibiotic resistance. These methods are inefficient, necessitating the selection of numerous colonies to identify the correct mutants, or ineffective, where the desired clone cannot be obtained (9–12). Meanwhile, CRISPR/Cas9 has revolutionized the field of genome editing but has only recently been adapted for use in mycobacteria for this purpose (19–21). This adaptation has been made more challenging because of different DNA repair mechanisms in the different mycobacterial species, preventing the facile translation of any particular gene editing system across all species. We herein demonstrate the use of Sth1 CRISPR/Cas9 as a genome editing tool in M. abscessus.

Considering the toxicity associated with inducing both Cas9 and targeting sgRNAs resulting in DSBs (Fig. 2A; Fig. S4) (20, 21), we separated each component into two different plasmids and introduced them in separate steps. One of the major challenges of genetic editing in M. abscessus has been the high rates of spontaneous resistance to antibiotics used for selection after transformation. Indeed, during the initial transformation of the first integrative Cas9 plasmid into the parental WT strain, spontaneous KAN resistance occurred in at least 50% of isolated mutants, which did not contain the desired integration (Fig. S1). To more efficiently identify the desired transformants with the sgRNA plasmid, we utilized the fluorescent mCherry reporter to visually identify successful plasmid uptake. Successful transformants containing the desired sgRNA guide in the inducible Cas9 background were thus easily selected as HYG-resistant pink colonies. Importantly, mCherry could also be used as a surrogate reporter for successful plasmid loss during curing of the mCherry-sgRNA plasmid once the target gene had been disrupted. This process removes the original targeting guide while preserving the integrated Cas9 for additional sequential editing. Finally, it is possible to excise Cas9 from strains with the desired mutation by transforming them with an additional attB-integrating vector (21, 29).

By targeting genes in M. abscessus, we have verified by targeted sequencing that the induction of Cas9 leads to DSBs that can be repaired, albeit imprecisely. We showed that Cas9 can create frameshift mutations in antibiotic resistance genes resulting in increased susceptibility to the respective antibiotic. Furthermore, no additional indels were observed upon comparing whole genome sequencing results between select edited strains and parental wild-type strains, confirming that Sth1Cas9 is highly specific and off-target editing is rare (20, 21). With our dual-plasmid workflow, we were able to efficiently generate disruptions in multiple genes through an iterative process of curing and reintroducing new sgRNAs sequentially. Alternatively, introducing multiple guides on the same plasmid allowed us to generate double gene knockouts in a single step. While this simultaneous editing strategy accelerated mutant generation, multiple sgRNAs also increased toxicity (Fig. 2A; Fig. S4) compared to single sgRNAs, with the frequency of identifying frameshifts in both genes lower than in a single gene. Hence, the sequential editing strategy may provide greater success with generating complex mutants disrupting more than two genes despite the longer timeframe required.

The highly programmable nature of CRISPR allows for easy design and preparation of sgRNA sequences, greatly enabling simple genetic disruption. Bioinformatic analysis of the M. abscessus genome indicated that almost 95% of genes can be targeted for disruption to achieve frameshifts early in the reading frame. In genes lacking upstream targetable sites or where the frequency of obtaining frameshift indels is low, we show the possibility of genetic disruption through introducing multiple adjacent sgRNAs to mediate deletions of longer gene regions.

The induction of both Sth1Cas9 and a targeting sgRNA led to frameshifts in 33%–91% of clones selected (excluding strains bearing two adjacent sgRNAs), with an overall frameshift frequency of 73% (45/62 analyzed clones across all genes studied). In turn, this enables the screening of only small numbers of clones to obtain the desired mutants. Indeed, we report an efficiency rate of 1.1%–9.3% using this two-plasmid system for the disruption of a single gene, which is at least 10²–10⁴ more efficient than rates reported for homologous recombination (9–12). Compared to traditional approaches where success is limited due to the occurrence of spontaneous background resistance and low efficiency of homologous recombination, the high frameshift frequency suggests that CRISPR/Cas9 is an important addition to the toolbox for genetic manipulation in M. abscessus. Importantly, we show that the expression of Cas9 and targeting sgRNA are sufficient to mediate genetic disruption with the native repair machinery in M. abscessus. Thus, additional components are not required to promote DNA repair, in contrast to M. marinum and M. tuberculosis (19–21). It is also noteworthy that while we would expect indels of random lengths to occur upon non-homologous end joining (NHEJ) DNA repair, the breadth of lesions we observed following the repair of Cas9-mediated double-strand breaks directed by the same sgRNA was less than one might expect. For instance, editing of the bla gene with the same guide only led to six different mutations, three of which accounted for 85% of all analyzed clones, and identical 1 bp deletions were observed in 60% of analyzed clones (cumulative total of 20 clones analyzed). While work on nontemplated DNA repair pathways in mycobacteria has largely focused on NHEJ pathways, these editing events in M. abscessus suggest a possible role of alternate end-joining where DNA repair is instead mediated by microhomologies around the DSB site (18). The exact contributions of these pathways in M. abscessus merit deeper investigation.

It should also be noted that disruption of a single nonessential gene using CRISPR/Cas9 led to variability in bacterial survival (1.6%–10.6%) and hence editing efficiency (1.1%–9.3%). The sgRNA sequences targeted PAM sites with similar strength and had similar length and GC content, suggesting the possibility of differences in the native DSB repair efficiency across the genome. Analyzing a broader set of target genes and a larger set of surviving mutants would be beneficial to better understand target-specific factors that affect editing efficiency.

While this method is more efficient and programmable than existing homologous recombination-based tools, the repair of Cas9-mediated DSBs is imprecise, resulting in the formation of indels of various lengths. This may be sufficient for the purpose of generating null mutations to study the overall loss of function of a target gene. However, imprecise DNA repair can also lead to more extensive deletions (e.g., >100 bp deletions) that can cause undesired off-target (polar) effects on neighboring genes. Here, we noted that extensive deletions occurred in 22% of all editing events analyzed (16 of 72). However, this method may be less prone to polar effects than other methods, for example, the insertion of a large antibiotic resistance cassette or CRISPRi, which may result in the downregulation of an entire operon rather than a single gene within an operon. Here, prolonged transcriptional termination of an operon is unlikely since Cas9 is only transiently induced to generate DSBs for the purposes of genetic disruption.

We also recognize that there are situations where specific point mutations, insertions, or deletions are preferred to address more mechanistic questions rather than gross loss of function. However, generating more precise mutations of these types requires inefficient homologous recombination or newer CRISPR-based tools such as prime editing or base editing (18, 30, 31). The latter approach was recently demonstrated for M. tuberculosis for targeted G:C to A:T base pair conversion (31). While this approach enabled the engineering of precise, targeted point mutations in M. tuberculosis, the optimization of several additional factors was required prior to its successful application in this species, including the fused Cas9 nickase-cytidine deaminase-uracil DNA glycosylase inhibitor for base editing, as well as RecX and NucS from M. smegmatis to suppress the native DNA repair pathways that compete with base editing (31). The extent to which additional factors may be required to develop these methodologies in M. abscessus remains unclear.

In summary, we validate the use of CRISPR/Cas9 in M. abscessus as a programmable, efficient, and versatile tool for genetic disruptions in single as well as multiple genes. CRISPR/Cas9 will, thus, greatly facilitate targeted gene function studies to interrogate the biology, pathogenesis, and antibiotic resistance mechanisms of M. abscessus. The efficiency with which desired mutants can be obtained and the simplicity of this system, where no additional factors are required, will make this an important new tool for studying this human pathogen, which heretofore has been difficult to genetically manipulate.

MATERIALS AND METHODS

Assembly of CRISPR/Cas9 plasmids

Our designed CRISPR/Cas9 methodology required the Cas9 and sgRNA cassettes on two separate plasmids (Fig. 1B). Primers used for the construction and verification of these plasmids are listed in Table S1. Plasmids generated through this study are listed in Table S2. All enzymes used for cloning were purchased from NEB.

pJR962 (Addgene plasmid #115162; http://n2t.net/addgene:115162; RRID: Addgene_115162; gift from Jeremy Rock) was used as the template for the construction of the Cas9-containing plasmid (23). Site-directed mutagenesis was performed with the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs) to reintroduce A9D and A599H on the dCas9 allele. These two mutations restore the nuclease activity of the Cas9 allele to facilitate the formation of DSBs. The resulting plasmid pDN-Sth1Cas9 contains genes for the AHT-inducible Sth1Cas9, L5 integrase for integration into the attB site, tetracycline repressor TetR, and resistance selection marker Kan^R.

pCHERRY3 (Addgene plasmid #24659; http://n2t.net/addgene:24659; RRID: Addgene_24659; gift from Tanya Parish) was used as the template for the construction of the sgRNA-containing plasmid (25). The original mCherry promoter was replaced with the constitutive strong mycobacterial promoter P_smyc amplified from pJR962 to create pDN-Cherry. Next, the sgRNA cassette and Golden Gate cloning site from pJR962 were amplified and cloned into pDN-Cherry while also removing the native BsmBI site on pCHERRY3, which would interfere with subsequent cloning of targeting sgRNA sequences. The resulting plasmid pDN-Cherry-sgRNA contains genes for the AHT-inducible sgRNA cassette, Golden Gate sites for cloning multiple sgRNA guides, constitutively expressing mCherry, and resistance selection marker Hyg^R.

All targetable sites on the M. abscessus genome (NCBI Reference Sequence: NC_010397.1) were extracted based on the previously reported 15 PAMs for Sth1Cas9 (23, 26) (Table S3). The predicted cut site (between the third and fourth nucleotides from the PAM) was used to assign target sites to their corresponding gene regions (Table S4). Selected targeting guides were purchased as complementary oligos (IDT) and assembled into BsmBI-linearized pDN-Cherry-sgRNA according to previously reported protocols (23, 26) (Table S5). For plasmids containing multiple sgRNA sequences on the same vector, SapI-based Golden Gate cloning was performed.

Plasmids were transformed and amplified into NEB 5-alpha competent E. coli (New England Biolabs), isolated with the QIAprep Spin Miniprep Kit (Qiagen), and verified by sequencing prior to further cloning or transformation.

Bacterial strains and growth conditions

Mycobacterium abscessus ATCC 19977 (WT) was used as the parental wild-type strain for generating intermediate and mutant strains. M. abscessus was grown in Middlebrook 7H9 broth (BD) or Middlebrook 7H10 agar (BD) supplemented with 10% Middlebrook oleic albumin dextrose catalase growth supplement (VWR), 0.5% glycerol (VWR), and 0.05% Tween-80 (Alfa Aesar) at 37°C.

For strains transformed with pDN-Sth1Cas9, growth media was supplemented with 50 µg/mL (liquid) or 500 µg/mL (solid) KAN (Sigma-Aldrich). For strains transformed with pDN-Cherry-sgRNA, growth media was supplemented with 0.5 mg/mL (liquid) or 1 mg/mL (solid) HYG (Life Technologies). For the induction of CRISPR/Cas9, growth media was supplemented with 100 ng/mL AHT (Sigma-Aldrich).

For plasmid amplification, NEB 5-alpha competent E. coli (New England Biolabs) was used and grown in LB broth or agar supplemented with 50 µg/mL KAN or 150 µg/mL HYG as required.

Generation of M. abscessus CRISPR/Cas9 strains

Electroporation of M. abscessus strains was performed according to previously published protocol using the BioRad Gene Pulser Xcell (10), recovered in 7H9 broth, and plated on 7H10 agar plates supplemented with antibiotics as required. All M. abscessus strains generated through this study are listed in Table S6.

For transformation with pDN-Sth1Cas9, colonies were selected from 7H10 plates containing KAN after 5 days and grown in 7H9 with KAN for 2 days. Five microliters of each culture was mixed vigorously with 100 µL of UltraPure DNase/RNase-free distilled water (ThermoFisher Scientific) before incubation at 100°C for 6 min. After cooling to room temperature, 5 µL of this heat-killed suspension was used as a template for colony PCR. Primers targeting the integration sites (attL and attR) as well as the Sth1Cas9 cassette were used to verify the successful integration of pDN-Sth1Cas9 (Table S1). PCR products were separated and visualized on 2% (wt/vol) agarose gel (Fig. S1).

For the transformation of the Cas9-integrated background strain with pDN-Cherry-sgRNA, colonies were selected from 7H10 plates with KAN and HYG after 5 days. Successful transformants that had taken up pDN-Cherry-sgRNA were identified visually as pink colonies (Fig. S2).

For curing of pDN-Cherry-sgRNA, strains were grown in 7H9 broth with KAN only (with no HYG) and serially passaged (100-fold dilution) for approximately 2–3 days and plated on 7H10 agar plates with KAN only. Mutants that had successfully cured the plasmid were conversely identified visually as white colonies.

Induction of CRISPR/Cas9 to generate edited M. abscessus strains

Tenfold dilutions of each strain were prepared and plated into six-well 7H10 plates containing necessary antibiotics in the presence or absence of AHT in triplicate wells for CFU determination. Bacterial survival upon Cas9/sgRNA induction was calculated by dividing the average CFU with induction by the average CFU without induction and plotted using GraphPad Prism 9.4.0. Colonies from plates containing AHT were picked for colony PCR as described above. PCR products were separated and visualized on 2% (wt/vol) agarose gel and purified by the QIAquick PCR Purification Kit (Qiagen) for sequencing (Genewiz) to determine the genetic outcomes of DNA repair after Cas9 induction. The frequency of outcomes from each sgRNA was plotted using GraphPad Prism 9.4.0. The overall efficiency of gene editing was calculated as the product of bacterial survival and frameshift frequency following Cas9/sgRNA induction. The strains generated are listed in Table S6. Editing outcomes for each mutant clone are listed in Table S7.

Determination of MICs

Select edited M. abscessus strains were further phenotypically verified for loss of function by determining MICs against ampicillin sodium (EMD Millipore), rifampicin (Sigma-Aldrich), and capreomycin sulfate (Sigma-Aldrich). Strains were grown to optical density at wavelength 600 nm (OD₆₀₀) ~0.8–1.0 in media supplemented with KAN and HYG. Cultures were diluted to 100 µL of serially diluted antibiotics in 96-well clear bottom plates (Corning) (final OD₆₀₀ = 0.0025) in four technical replicates. One percent dimethylsulfoxide (DMSO) (Sigma-Aldrich) was used as the negative untreated control, and 25 µg/mL ciprofloxacin hydrochloride (MP Biomedicals) was used as the positive control. Plates were incubated at 37°C without shaking for approximately 4 days before OD₆₀₀ measurements. Growth (%) was calculated by normalizing OD₆₀₀ from treatment wells with OD₆₀₀ from DMSO control wells and plotted using GraphPad Prism 9.4.0. MIC values were taken as the minimum concentration needed to inhibit growth by 90%.

Genomic DNA extraction for whole genome sequencing

Genomic DNA from select edited strains and corresponding parental wild-type strains were extracted using the cetyltrimethylammonium bromide method (32). Sequencing libraries were prepared using the Nextera XT DNA Library Preparation Kit (Illumina) and sequenced on a NextSeq platform. Sequencing reads were aligned to ATCC 19977 reference genome (NC_010397.1) and analyzed using the Pilon pipeline to identify off-target mutations upon Cas9 induction (33).

DATA AVAILABILITY

The plasmids pDN-Sth1Cas9 and pDN-Cherry-sgRNA are available at Addgene or on request. The custom python script to identify PAM sites on the M. abscessus genome is provided in the Supplementary Information.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jb.00335-23.

Supplemental material. jb.00335-23-s0001.docx.

Figures S1 to S5, Tables S1 and S2, and supplemental code.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.