Genetic analysis of Mycobacterium abscessus reveals genomic diversity linked to global plasmid distribution

Summary

Genetic epidemiological analysis of mobile genetic elements, such as plasmids, has rarely been carried out in Mycobacterium abscessus, regardless of its usefulness in speculating the historical contact between bacteria. In this study, through whole genome sequencing analysis of M. abscessus isolated from the same patient, we identified three plasmids (pMAB625-1, pMAB625-2, and pMAB625-3) that were not found in the type strain ATCC 19977. We conducted an in silico analysis to investigate the distribution of previously identified plasmids, including the pMAB625 plasmids, in M. abscessus isolated from 462 clinical specimens worldwide. pMAB625-3 was the most prevalent plasmid, detected in 11.5% (53/462) of isolates. Furthermore, phylogenetic tree analysis revealed that these plasmids transferred beyond regions and subspecies and acquired unique mutations. These results indicate that the transmission of plasmids increases the genomic diversity in M. abscessus, and plasmid epidemiology is useful for estimating the historical contact between bacteria.

Graphical abstract

Highlights

• •Three pMAB625 plasmids were identified in five Mycobacterium abscessus isolates
• •pMAB625-3 is the most prevalent of the analyzed plasmids in M. abscessus globally
• •Host genome phylogeny reveals horizontal transfer of the pMAB625 plasmids
• •M. abscessus plasmid epidemiology helps elucidate global microbial diversity

Phylogenetics; Microbial genomics; Genomic analysis; Sequence homology

Introduction

In recent years, the threat of antimicrobial-resistant bacteria has been increasing worldwide, and they could be responsible for 1.91 million deaths per year by 2050.¹ Hence, research into the mechanisms and routes of transmission is becoming increasingly important to develop effective countermeasures against antimicrobial-resistant bacteria. The global incidence of non-tuberculous mycobacterial infections has been on the rise, presenting a significant challenge due to their increasing prevalence and innate drug resistance, particularly in Mycobacterium abscessus infections.² M. abscessus is one of the rapidly growing mycobacteria (RGM) and is not limited to pulmonary infections but also serves as the causative agent for various extrapulmonary infections, including bloodstream, skin, soft tissue, surgical site, and peritoneal dialysis-related infections.³ Furthermore, it demonstrates resistance to a wide range of antimicrobial agents.

M. abscessus is divided into three subspecies: M. abscessus subsp. abscessus (ABS), M. abscessus subsp. bolletii (BOL), and M. abscessus subsp. massiliense (MAS) (Figure S1). These subspecies are very closely related genetically and share similar clinical features, including multidrug resistance; among them, ABS and BOL exhibit intrinsic and/or inducible resistance to macrolides, the key drugs for treatment, making therapy particularly challenging. We have collected RGM isolates from patients across Japan, including M. abscessus, as part of a nationwide surveillance of RGM species and their antimicrobial susceptibilities.⁴ Among these, M. abscessus accounted for approximately 60% of RGM, and among the three subspecies of M. abscessus, ABS was the most common, accounting for about 57%.

The route of infection of M. abscessus remains controversial. It has been assumed that M. abscessus is transmitted from the environment, such as soil and water systems, and person-to-person transmission does not occur. However, a few cases of suspected person-to-person transmission have been reported.^5,6,7 In addition, the existence of “dominant clones,” which were isolated from almost all regions and phylogenetically close,^7,8 makes the source and transmission routes of the bacteria mysterious.⁹

Whole genome sequencing (WGS) analysis and phylogenetic tree analysis of the core genome, which is the genomic region shared by all isolates, are often conducted to predict the origin and transmission route of M.abscessus.¹⁰ However, the accessory genome, which is shared by a certain percentage but not all of the isolates of M. abscessus, such as plasmids, prophage, and genomic islands, is made up of approximately 30% of the typical number of all genes of M. abscessus.¹¹ Therefore, the pan-genome, including both the core and accessory genomes, was analyzed to compare the genetic relationships among all gene sets of M. abscessus.^9,12,13 Although pan-genome analysis has the advantage of comparing whole bacterial genomes, it has the disadvantage of treating accessory genomes as gene units, thereby losing information about the units of mobile genetic elements. Therefore, epidemiological analysis of mobile genetic element units, such as individual plasmids, may more accurately predict the origin and route of infection of bacteria.

In this study, we performed WGS of 13 M. abscessus isolates obtained from patients and three type strains: ATCC 19977 (ABS), JCM 15297 (BOL), and JCM 15300 (MAS). These included five isolates (Abs1 through Abs5) that were derived from a single case of disseminated infection in which ABS was detected over a six-month period during prolonged treatment. In isolates Abs1-5, we identified three plasmids (pMAB625-1, pMAB625-2, and pMAB625-3) that were absent from ATCC 19977. Moreover, Abs5 exhibited a higher minimum inhibitory concentration (MIC) for imipenem (IPM) compared with that of ATCC 19977 and Abs1. In contrast, M. abscessus isolates other than Abs1-5, as well as other RGM species used as controls (two M. fortuitum, one M. septicum, and one M. iranicum), did not simultaneously possess all three plasmids. Therefore, we analyzed the distribution of plasmids identified in M. abscessus before, including the pMAB625 plasmids, and found that pMAB625-3 was the most widely spreading plasmid among M. abscessus isolated from clinical specimens and registered in the database worldwide. Thus, we investigated whether predicting the past contact between isolates is possible by analyzing single-nucleotide variants (SNVs) and structural variants (SVs) of the plasmids.

Results

Abs5 exhibited a higher minimum inhibitory concentration to imipenem compared with ATCC 19977

We encountered a patient with ABS osteomyelitis of the lower extremity associated with bacteremia. During a six-month period, a total of five ABS strains (Abs1-5) were isolated from the patient’s specimens (Table 1). We first compared the growth rate and antimicrobial susceptibility of Abs1 and Abs5 with the type strain ATCC 19977. The growth rates of Abs1 and Abs5 were significantly slower than ATCC 19977 (Figure S2). Antimicrobial susceptibility testing showed that the MIC of Abs5 against IPM, which binds the penicillin-binding proteins (PBPs) and inhibits the bacterial cell wall synthesis, was over four times higher than that of ATCC 19977 (Table 1). When applying the breakpoints defined in CLSI M24 3rd edition¹⁴ (8–16 μg/mL as intermediate and ≥32 μg/mL as resistant), IPM susceptibility shifted from intermediate to resistant. For the other antibiotics, there were no differences in MIC values between ATCC 19977 and the two isolates (Abs1 and Abs5) that would alter the susceptibility classification (susceptible, intermediate, or resistant) according to CLSI breakpoints.

Table 1.

The source and isolation time point of Abs1-5 and antimicrobial susceptibility

Isolates	Source	Period from the first isolation (month)	MICa (μg/mL)
			AMK	TOB	IPM	FRPM	LVFX	MFLX	AZM	CAM 4days	CAM 14days	ST	DOXY	MEPM	LZD	CLF	STFX
ATCC 19977	–	–	8	8	16	>64	32	>8	64	2	>64	>152/8	>16	64	32	1	2
Abs1	catheter	0	8	8	16b	>64	16	>8	64	2	>64	>152/8	>16	>64	32	1	2
Abs2	catheter	4	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Abs3	bone biopsy	4	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Abs4	heel	6	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Abs5	malleolus	6	8	8	>64c	>64	32	>8	64	2	>64	>152/8	>16	>64	32	1	2

^aMICs were determined as the minimum concentration at which the bacteria did not grow. No antimicrobial susceptibility testing was conducted on the isolates other than Abs1 and Abs5, representing hyphens in the table. The test was repeated three times, and all MICs were confirmed to be the same between the experiments, with the exception of imipenem for the isolates. AMK, amikacin; TOB, tobramycin; IPM, imipenem; FRPM, faropenem; LVFX, levofloxacin; MFLX, moxifloxacin; AZM, azithromycin; CAM, clarithromycin; ST, trimethoprim/sulfamethoxazole; DOXY, doxycycline; MEPM, meropenem; LNZ, linezolid; CLF, clofazimine; STFX, sitafloxacin.

^bFor Abs1, the MICs against IPM in the three experiments were 64, 16, and 16 μg/mL, respectively.

^cFor Abs5, the MICs against IPM in the three experiments were >64, 16, and >64 μg/mL, respectively.

Whole-genome sequencing of Abs1-5

To explore the factors determining the differences in growth rate and antimicrobial susceptibility between ATCC 19977 and the two isolates (Abs1 and Abs5), we performed WGS using a short-read sequencer and analyzed SNVs in the genome (Table S1). We detected 34 missense mutations and nine frameshift mutations in the Abs1 genome compared with ATCC 19977. Furthermore, to analyze the difference in the genome sequence between Abs1-5, we analyzed SNVs among Abs1-5 (Figure S3). As a result, we detected seven SNVs (one in Abs3 and six in Abs5). It was reported that several mutations correlated with the development of IPM resistance in the outbreak strains of MAS.¹⁵ The mutations detected in Abs5 did not match those reported to be associated with IPM resistance. We also detected a few SNVs in several transporters in Abs1-5 compared with ATCC 19977; however, we could not identify any SNVs that clearly explain the phenotypic differences, such as growth rate and antimicrobial susceptibility.

De novo genome assembly of Abs1-5

Plasmids could change host characteristics such as AMR in various bacteria.¹⁶ However, it is difficult to find novel plasmids through WGS using a short-read sequencer. This is because de novo assembly cannot construct circular structures based solely on short-read sequences. Therefore, we confirmed the presence of plasmids in Abs1-5 by long-read sequencing using the MinION Nanopore long-read sequencer. Furthermore, to facilitate the circularization of contigs, we performed two steps. First, we removed reads shorter than one or five kilobases from the long-read data. Raw long-reads and filtered long-reads data metrics are indicated in Table 2. Second, we corrected the long-read sequence data using short-read data acquired using the Ion Proton sequencer. We then performed de novo genome assembly using the corrected long-read sequence data. Two circularized contigs (contig1, 5 Mbp; contig2, 25 Kbp) and two linearized contigs (contig3, 202 Kbp; contig4, 140 Kbp) were obtained from each of Abs1-5 (Table S2). Since the genome size of ATCC 19977 is 5 Mbp, we concluded that contig1 was the genome, while the others were plasmids. As a result, we named these three plasmids, contig3, 4, and 2, as pMAB625-1, pMAB625-2, and pMAB625-3, respectively. The length, GC content, and number of coding sequences (CDS), rRNA, tRNA, and tmRNA of the genome or each plasmid are shown in Table 2. The CRISPR sequence was not detected in the genome.

Table 2.

Genome assembly and the annotation statics of the Abs1

Isolate	Isolateion#	Number of raw readsa	Total bases of raw readsa (bp)	Number of filtered readsb	Total bases of filtered readsb (bp)	Number of contig	N50 (bp)	Completenessc (%)
Abs1	57625	79,281	844,070,568	36,027	500,001,589	4	5,072,346	98.9
Abs2	57629	46,661	574,017,773	36,370	500,001,832	4	5,072,394	98.6
Abs3	57630	69,286	612,660,927	45,838	500,009,487	4	5,072,447	98.3
Abs4	57596	72,529	811,169,367	34,097	500,002,362	4	5,072,548	98.5
Abs5	57626	126,008	1,414,800,180	28,559	500,005,196	4	5,079,558	98.6

Contig	Form	Length (bp)	GC contents (%)	Number of CDS	Number of rRNA	Number of tRNA	Number of tmRNA
Genome	circular	5,072,346	64.1	5,034	3	48	1
pMAB625-1	linear	202,398	64.7	247	0	29	0
pMAB625-2	linear	142,689	62.2	171	0	0	0
pMAB625-3	circular	24,988	65.6	30	0	0	0

^aRaw reads were basecalled using Guppy basecaller.

^bFiltered reads were raw reads trimmed using Porechop and filtered by Filtlong to remove shorter reads. Total bases of the filtered reads were equalized to 100X of the genome size of ATCC 19977 (5 Mbp).

^cCompleteness were automatically calculated using CheckM based on the marker set of Mycobacterium abscessus during gene annotation using DFAST.

Next, we analyzed SVs among the de novo assembled genome sequences of Abs1-5 and ATCC 19977. As a result, we found that two large insertions (13 kbp and 8 kbp) and a deletion (16 kbp) existed in the sequence of Abs1 compared with ATCC 19977 (Figures S4A and S4B). The coding sequences of phage integrase and transposase were present near the three large SVs (Figure S4C), indicating that these regions were prophage and transposon. Additionally, phylogenetic analysis showed that Abs1-5 were genetically similar to ATCC 19977 (Figure S5). We found that 20 genes were acquired and 25 genes were lost in the Abs1 genome compared with the ATCC 19977 genome via SVs (Table S1); however, we did not identify the causal genes explaining the phenotypic difference between ATCC 19977 and Abs1 with clear evidence. We next analyzed SVs between Abs1-5, identifying two SVs (one insertion and one deletion in Abs5) (Figure S6). In the Abs5 genome, nine genes were acquired, and one gene was lost via SVs compared with Abs1 (Table S1). However, we could not find the causal genes responsible for the difference in the sensitivity to IPM between Abs1 and Abs5 through homology search in the CARD AMR gene database.¹⁷

Abs1-5 had three plasmids that encode various virulence factors

WGS analysis also revealed that Abs1-5 harbored three plasmids that were not found in ATCC 19977, designated as pMAB625-1, pMAB625-2, and pMAB625-3. Dedrick et al. reported on their analysis of the prophages and plasmids in 82 clinical isolates of M. abscessus that they had collected.¹⁸ In their report, it was shown that about half of the analyzed isolates did not have any plasmids, 26 isolates had one, 11 isolates had two, and only one isolate had three plasmids. Therefore, Abs1-5, which have three plasmids, are a very rare case. Thus, we next focused on the pMAB625 plasmids, and compared the pMAB625 plasmid sequences with other plasmids registered in PLSDB¹⁹ previously (Table 3). A taxonomic keyword search for “abscessus” returned 20 plasmids. To investigate the homology between these plasmids, we calculated ANI between them and the pMAB625 plasmids. No plasmid had a high ANI for pMAB625-1; however, the ANI between pMAB625-2 and pGD21-2 was 97.4% with some SVs. Furthermore, the ANI for pMAB625-3 against pGD69A-1 and pGD42-1 was 100%. These results indicated that pMAB625-2 and pMAB625-3 were retained in M. abscessus, which were isolated by other groups previously. We next examined the differences in the pMAB625 plasmid copy numbers to verify whether the difference in sensitivity against IPM between Abs1 and Abs5 was due to plasmid copy numbers. We performed qPCR using primers specific to each plasmid to verify the copy numbers of the pMAB625 plasmids in Abs1 and Abs5 (Figure S7). However, there were no significant differences in the copy numbers of any pMAB625 plasmids between Abs1 and Abs5. This analysis also showed that the copy numbers of pMAB625-1 and pMAB625-2 were less than one per bacterium, and that of pMAB625-3 was more than five per bacterium in Abs1 and Abs5. Therefore, Abs1 and Abs5 were considered to be heterogeneous populations in terms of plasmid possession. Dedrick et al. classified the plasmids based on sequence homology and reported the copy numbers of each cluster.¹⁸ They analyzed the copy numbers by calculating the fold difference in depth of coverage between the plasmid and the genome. According to their findings, the average copy number of the pB cluster, to which both pGD69A-1 and pGD42-1 (which are almost identical to pMAB625-3) belong, was 2.3. Meanwhile, the copy number of pGD21-2, which is similar to pMAB625-2, was found to be 1.3. Our results differed slightly from theirs, likely due to the differences in the analytical method and the strains analyzed.

Table 3.

Homology search result between plasmids identified in Mycobacterium abscessus

Plasmid	Accession no.	Topology	Length (bp)	Homologous plasmid (ANI(%)>99)	Homologous plasmid (97< ANI(%) <99)	Use for “Reference”
pMAB625-1	AP028614.1	linear	202,398	–	–	TRUE
pMAB625-2	AP028615.1	linear	142,689	–	pGD21-2b	TRUE
pMAB625-3	AP028616.1	circular	24,988	pGD69A-1, pGD42-1	–	TRUE
pMAB23	NC_010394.1	circular	23,319	pMabS_GZ002a, FDAARGOS_1604_plasmid	–	TRUE
pMabS_GZ002	GenBank: NZ_CP034180.1	linear	15,203	pMAB23, FDAARGOS_1604_plasmid	–	TRUE
pGD19	GenBank: NZ_CP063329.1	circular	18,605	–	–	TRUE
pGD21-1	GenBank: NZ_CP065285.1	circular	112,633	–	–	TRUE
pGD21-2	GenBank: NZ_CP065286.1	linear	155,609	–	pMAB625-2b	TRUE
pGD22-1	Genbank: NZ_CP063325.1	circular	19,694	–	–	TRUE
pGD22-2	GenBank: NZ_CP063326.1	circular	18,117	pGD42-2a	–	TRUE
pGD25-1	GenBank: NZ_CP063321.1	circular	31,413	–	–	TRUE
pGD25-2	GenBank: NZ_CP063322.1	circular	27,424	–	–	TRUE
pGD25-3	GenBank: NZ_CP063323.1	circular	23,599	–	–	TRUE
pGD42-1	GenBank: NZ_CP065281.1	circular	24,993	pMAB625-3, pGD69A-1	–	FALSE
pGD42-2	GenBank: NZ_CP065282.1	circular	9,547	pGD22-2	pGD69A-2	TRUE
pGD69A-1	GenBank: NZ_CP065270.1	circular	25,000	pMAB625-3, pGD42-1	–	FALSE
pGD69A-2	GenBank: NZ_CP065271.1	circular	18,611	–	pGD42-2a	TRUE
pJCM30620_1	GenBank: NZ_AP022622.1	linear	119,389	50594_plasmid2a	–	TRUE
pJCM30620_2	GenBank: NZ_AP022623.1	linear	38,763	50594_plasmid1b	–	TRUE
BRA100	NC_017908.2	circular	56,265	–	–	TRUE
FDAARGOS_1604_plasmid	GenBank: NZ_CP085920.1	circular	23,319	pMAB23, pMabS_GZ002a	–	FALSE
50594_plasmid1	NC_021278.1	circular	172,814	pJCM30620_2a	–	TRUE
50594_plasmid2	NC_021279.1	circular	97,240	pJCM30620_1b	–	TRUE

^aOne structural variant (SV) exists in the sequence.

^bMore than two SVs exist in the sequence.

We then investigated the functions of the genes encoded by the pMAB625 plasmids using a homology search. Plasmids encode various factors that are able to influence host characteristics. Therefore, we listed the lost and acquired genes in Abs1 relative to ATCC 19977 via SVs and plasmid acquisition to investigate factors affecting the phenotypical differences (Table S3). Of note, 480 candidate open reading frames (ORFs) were acquired, and 37 candidate ORFs were lost in Abs1 compared with ATCC 19977. Then, we searched for the AMR genes in several databases using Blast+ based on the amino acid sequences of the acquired candidate ORFs. However, we did not detect any genes with sufficiently high identity and coverage (over 35% coverage and over 33% identity).

To gain further insight into the effect of gene acquisition, we searched for the homology of the acquired candidate ORFs using the mycobacterial gene database Mycobrowser.²⁰ This analysis detected 37 mycobacterial proteins with amino acid homology to candidate ORFs of Abs1 (identity >30%, e-value < 1e-50) (Table S4). Notably, genes associated with the ESX secretion system were highly enriched (10/37, 27%). ESX secretion systems are encoded in various bacteria on their genomes and plasmids. It has been reported that the ESX secretion system (also known as the type VII secretion system) is involved in bacterial virulence through several mechanisms, including the evasion of the host immune system,^21,22 and biofilm formation.²³ We further searched whether other ORFs of the ESX secretion system components were encoded near the loci of the detected ORFs of the same system using blast+. Several components of the ESX secretion system in pMAB625-1 and pMAB625-2 were detected (Figure 1A). Core genes that construct membrane pores (eccB, eccC, eccD, mycP, and eccE) were present in these gene clusters. ATCC 19977 and Abs1-5 have ESX-3 and ESX-4 loci on their genomes; therefore, Abs1-5 acquired two additional ESX loci by plasmid acquisition. Next, to predict the functions of these genes, we performed phylogenetic tree analysis using known amino acid sequences of ESX secretion system components from genomes and plasmids. It was previously reported that plasmid-encoded ESX loci can be classified into four major clusters.²⁵ Importantly, our analysis revealed that plasmid-encoded ESX loci were classified into five major clusters (ESX-P cluster 1–5) (Figure 1B). Notably, the newly classified ESX-P cluster 5 included the ESX loci of pMAB625-1 and pMAB625-2. As shown in the plasmid homology search results (Table 3), the ESX loci of pGD21-2 and pMAB625-2 were genetically close. The ESX loci of pMAB625-1 were genetically close to those of pMFLV01, which was harbored by Mycobacterium gilvum PYR GCK. It was previously reported that pMFLV01 was enriched in biofilm formed in household water purifiers through metagenomic analysis.²⁶ Biofilm formation is one of the drug resistance mechanisms in Mycobacterium and acts as a barrier against antibiotics. Additionally, we detected the toxin-antitoxin (TA) system components, YefM/YoeB and VapB5/VapC5, in pMAB625-1. The TA system is present in various bacteria and archaea and encoded in their genomes and plasmids. This system is known to help maintain plasmids, inhibit bacteriophage propagation, and increase antimicrobial tolerance by inducing dormancy.²⁷ Therefore, these results indicated that Abs1-5 acquired factors that could contribute to bacterial virulence through plasmid transfer.

Figure 1.

Phylogenetic and gene expression analysis of the genes related to the ESX secretion system

(A) The loci of the ESX secretion system are located on pMAB625-1 and pMAB625-2. CDSs of the ESX secretion system were searched using Blast+. The core components that construct the membrane pore (eccB, eccC, eccD, mycP, and eccE) are contained in this operon.

(B) Phylogenetic tree analysis of various ESX loci located on genomes and plasmids. The amino acid sequence of the genes that construct the membrane pore was concatenated, and the phylogenetic tree was created by Ngphylogeny.fr²⁴ using the PhyML+SMS inference method. The classification of each cluster was followed as previously reported.²⁵

(C) RNA-seq analysis of ATCC 19977 and Abs strains. The x axis represents the mean log expression levels, and the y axis of the top plot represents the log fold change. The bottom dot plot and boxplot represent the distribution of the expression levels of each gene. The green dots represent the data of genes that Abs1-5 acquired by SVs and plasmid acquisition. The red dots represent the data of ESX secretion system genes located on the plasmids.

Global gene expression analysis revealed that genes on the plasmid were expressed in Abs1-5

We found that the various genes associated with bacterial virulence were located on the pMAB625 plasmids. However, it was unclear whether the genes located on the pMAB625 plasmids were expressed in Abs1-5. Thus, we investigated global gene expression in ATCC 19977 and Abs1-5 through RNA-seq analysis (Figure 1C and Table S5). The total number of differentially expressed genes (DEGs) was 390, which accounted for 7.0% (390/5563) of all genes in the reference sequence. Additionally, 83.1% (324/390) of the DEGs were on the pMAB625 plasmids. The expression levels of 92.6% (300/324) of DEGs on the plasmids were relatively low (below the 25th percentile of the distribution of expression levels of all genes that were expressed in Abs1-5 and ATCC 19977). Notably, the expression levels of 37.0% of the genes associated with the ESX secretion system (10 out of 27), which were located on pMAB625-1 and pMAB625-2, showed relatively moderate (below the 75th percentile and above the 25th percentile) (Figure 1C). In addition, the genes associated with the TA system (YoeB-YefM and VapB5-VapC5) and the mycobacterial membrane protein large (MMPL) family transporter, which were also located on pMAB625-1 and pMAB625-3, respectively, were expressed at relatively low levels. These results indicated that the numerous genes associated with bacterial virulence located on the plasmids were expressed at relatively low to moderate levels in Abs1-5, despite the low copy numbers per bacterium of pMAB625-1 and pMAB625-2. Overall, these results indicated that the pMAB625 plasmids were major causes of the difference in the transcriptome of Abs1-5 compared with ATCC 19977.

pMAB625 plasmids were globally distributed in Mycobacterium abscessus

We found that Abs1-5 carried three plasmids that were absent from ATCC 19977, and that these plasmids had genes potentially contributing to bacterial virulence, which were also expressed in Abs1-5. However, we analyzed ABS isolated from only one patient, and it was uncertain whether the plasmid acquisition occurred extensively in Mycobacterium. Therefore, we investigated whether the pMAB625 plasmids were present in isolates registered in public databases along with other isolates of Mycobacterium we collected. We first performed PCR using the genomic DNA from 11 isolates collected in Japan,⁴ as well as Abs1 and ATCC 19977, with primers specific to each pMAB625 plasmid. Apart from Abs1, which was identified with all three plasmids, we detected pMAB625-3 without the presence of pMAB625-1 and pMAB625-2 in the other two isolates (Isolate57605 and Isolate57578) (Figure 2A). Both the isolates in which pMAB625-3 was detected were M. abscessus. We then confirmed the plasmid presence by mapping the read data from the short-read sequencer to the pMAB625 plasmid sequences. We used 116 read data of M. abscessus, which were isolated from Japan⁸ and obtained from the NCBI database, and nine data of M.abscessus collected by us (including Abs1, Isolate57590, Isolate57600, Isolate58243, Isolate57605, Isolate57550, Isolate57578, Isolate57778, which were used for PCR indicated in Figure 2A, and Isolate57551). Out of 125 isolates, pMAB625-2 was detected in four isolates (3.2%), and pMAB625-3 was detected in 16 isolates (12.8%) (Figure 2B). These isolates were collected from the four prefectures in Japan (Tokyo, Shizuoka, Aomori, and Okinawa) (Figure 2C). Next, to verify whether the pMAB625 plasmids are specific to M. abscessus, we examined the distribution of the pMAB625 plasmids in other Mycobacterium species using data registered in public databases. We mapped the read data of three of Mycobacterium iranicum, 20 of Mycobacterium fortuitum, and 25 of Mycobacterium chelonae; however, no pMAB625 plasmid was detected in these isolates (data not shown).

Figure 2.

The distribution of the pMAB625 plasmids in mycobacteria isolated from Japan

(A) PCR analysis of the possession of the pMAB625 plasmid in mycobacteria, which were isolated from Japan by our group. PCR was performed using the primers specific to each plasmid and HrpA. The color bar represents the species of each isolate. The primer sets of HrpA were different depending on the species.

(B) Mapping the read data of Mycobacterium abscessus isolated from Japan to the Abs1 genome sequence. The bar plot represents the depth of coverage to genome and each plasmid. The dot plot represents the percentage of the covered genome region.

(C) Geographical map of Japan. The prefectures in which the pMAB625 plasmids were detected, are represented by color.

Furthermore, to investigate whether M. abscessus in other areas of East Asia possesses the pMAB625 plasmids, we mapped the available read data of M. abscessus from Taiwan⁸ and China.²⁸ pMAB625-2 was detected in the isolates from China (2/69, 2.9%) and pMAB625-3 was detected in the isolates from Taiwan (8/98, 8.2%), and China (5/69, 7.2%) (Figure S8). pMAB625-2 and pMAB625-3 were detected in ABS and MAS but not in BOL. In addition, we analyzed M. abscessus isolated from Spain, the UK, and the US to determine whether they carried the pMAB625 plasmids (Table 4). pMAB625-2 was detected in isolates from the UK (2/30, 6.7%), and the US (2/110, 1.8%). pMAB625-3 was detected in the isolates from Spain (9/30, 30%), the UK (1/30, 3.3%), and the US (14/110, 12.7%). Overall, pMAB625-1 was detected in only Abs1-5 (1/462, 0.2%), pMAB625-2 was detected in 10 isolates (10/462, 2.2%), and pMAB625-3 was detected in 53 isolates (53/462, 11.5%).

Table 4.

pMAB625 and known plasmids distribution in Mycobacterium abscessus worldwide

Number (%) of isolates retaining plasmid
Region	Unknown (Type strain)	Japan	China	Taiwan	US	Spain	UK	totala
Analyzed isolates	3	125	69	98	110	30	30	462
pMAB625-1	0	1 (0.8)	0	0	0	0	0	1 (0.2)
pMAB625-2	0	4 (3.2)	2 (2.9)	0	2 (1.8)	0	2 (6.7)	10 (2.2)
pMAB625-3	0	16 (12.8)	5 (7.2)	8 (8.2)	14 (12.7)	9 (30.0)	1 (3.3)	53 (11.5)
pMAB23	1 (33.3)	0	0	0	1 (0.9)	0	0	1 (0.2)
pMabS_GZ002	1 (33.3)	0	0	0	1 (0.9)	0	0	1 (0.2)
pGD69A-2	0	0	0	1 (1.0)	9 (8.2)b	2 (6.7)	3 (10.0)	15(3.2)
pGD19	0	0	4 (5.8)	3 (3.1)	0	1 (3.3)	1 (3.3)	9(1.9)
pJCM30620_1	0	0	1 (1.4)	0	0	0	1 (3.3)	2(0.4)
50594_plasmid2	0	0	1 (1.4)	0	0	0	1 (3.3)	2(0.4)
pGD22-2	0	0	0	0	6 (5.5)	4 (13.3)c	1 (3.3)	11(2.4)
pGD42-2	0	0	0	1 (1.0)	12 (10.9)d	4 (13.3)b	1 (3.3)	18(3.9)
pGD25-1	0	0	0	0	3 (2.7)	0	4 (13.3)c	7(1.5)
pGD25-2	0	0	0	0	1 (0.9)	0	2 (6.7)	3(0.6)

Fisher’s exact tests were performed to verify regional differences in the frequency of appearance of each plasmid. Corrections for the multiple comparisons were performed using Holm’s method.

^aTotal number excludes the number of the type strains.

^bp < 0.05, compared with Japan.

^cp < 0.05, compared with Japan and Taiwan.

^dp < 0.05, compared with Japan, China, and Taiwan.

Since previous analyses have shown that the pMAB625 plasmids are retained in M. abscessus isolated worldwide, we investigated how extensively other plasmids, except for the pMAB625 plasmids, are retained in the same set of isolates indicated in Table 4. We removed three plasmids from the plasmid list indicated in Table 3 (pGD42-1, pGD69A-1, and FDAARGOS_1640_plasmid) because they were identical to other plasmids with no SV. We selected 20 plasmids shown in Table 3 as “TRUE” for “Use for Reference,” along with three pMAB625 plasmids. Then, we mapped the read data of M. abscessus, which were the same strains used for the validation of the presence of the pMAB625 plasmids, to these plasmid sequences. Except for the pMAB625 plasmids, three plasmids (pGD69A-2, pGD22-2, and pGD42-2) were detected in more than 10 isolates (Table 4). Seven plasmids (pGD21-1, pGD21-2, pGD22-1, pGD25-3, pJCM30620_2, BRA100, and 50594_plasmid1) were not detected in the analyzed dataset. Statistical analysis showed regional variations in the frequency of appearance of these plasmids (Table 4). pGD69A-2 was detected more frequently in the US (9/110, 8.2%) than in Japan (0/125, p = 0.014). pGD22-2 was detected more frequently in Spain (3/40, 13.3%) than in Japan (0/125, p = 0.018) and Taiwan (0/98, p = 0.036). pGD42-2 was detected more frequently in Spain (4/30, 13.3%) than in Japan (0/125, p = 0.017). pGD42-2 was also detected more frequently in the US (12/110, 10.9%) than in Japan (0/125, p = 0.001), China (0/69, p = 0.045), and Taiwan (1/98, 1.0%, p = 0.041). pGD25-1 was detected more frequently in the UK (4/30, 13.3%) than in Japan (0/125, p = 0.018) and in Taiwan (0/98, p = 0.036). Moreover, we performed a phylogenetic tree analysis of the genomes to determine whether these plasmids are being transferred horizontally or whether the same strains harboring the plasmids are spreading (Figure 3). The phylogenetic tree analysis showed that pGD22-2 and pGD42-2 were only detected in the isolates belonging to Clade I. In contrast, pMAB625-3 was detected mainly in the isolates belonging to the four clades (Clade II, IV, V, and VI). pGD25-1 and pGD69A-2 were detected mainly in the isolates belonging to Clade II and Clade V. pMAB625-2 and pGD19 were also detected in the isolates belonging to several clades. These results indicated that some plasmids (pMAB625-2, pMAB625-3, pGD19, pGD25-1, and pGD69A-2) were spread across clades, and others (pGD22-2 and pGD42-2) were limited to the specific clade.

Figure 3.

The relationship between the phylogenetic distance of host bacterial genomes and plasmid distribution in Mycobacterium abscessus worldwide

The phylogenetic tree was created by Gubbins²⁹ using genome sequences and visualized by ggtree.³⁰ The read data of Mycobacterium abscessus collected worldwide were mapped to the pMAB625 plasmids and known plasmid sequences to verify the plasmid distribution. Tree-scale represents the number of SNVs per genome.

Phylogenetic and structural variant analysis of the pMAB625 plasmids enabled the detailed classification of isolates and the prediction of historical contact between bacteria

We acquired the sequence information for pMAB625-3, which is retained in 53 different isolates of M. abscessus, and for pMAB625-2, which is retained in 10 different isolates of M. abscessus. Therefore, we examined the genetic relationship between the plasmids and investigated whether it could predict the route of transmission and classify the host bacteria in more detail. At first, we performed a phylogenetic tree analysis of 53 sequences of pMAB625-3, which were detected in Figure 3, and 11 plasmids (pGD18, pGD23, pGD36, pGD42-1, pGD47, pGD69A-1, pGD72, pGD87, pGD95, pGD108A, and pMabs-09-13 [ERR4326576]), which were reported to be identical or closely related to pMAB625-3 (Figure 4A).^18,31 Forty-one consensus sequences (covered by a yellow box in Figure 4A) were identical to the sequence of pMAB625-3 in Abs1. However, 23 sequences belonged to different clades. Next, we utilized a tanglegram wherein inter-tree edges connect the corresponding terminal taxa of the two topologies to analyze the relationship between the phylogenetic trees of pMAB625-3 and the host genome (Figure 4B). The complete genome sequences of GD36, GD47, GD108A, GD95, GD18, GD23, GD72, and GD87, which were the host of pGD36, pGD47, pGD108A, pGD95, pGD18, pGD23, pGD72, and pGD87, respectively, were not available. The tanglegram analysis revealed three findings. First, pMAB625-3, whose sequences were identical to the sequence in Abs1, exhibited the widespread distribution (the taxa connected by orange lines). Second, hosts belonging to Clades II and IV retained the pMAB625-3 unique to these clades (the taxa connected by green lines). SRR18969525 (US), which consisted of a distinct clade, also had a unique pMAB625-3. The host genome clades were supported by bootstrap values of more than 95% (Figure S9A). The clades of pMAB625-3, which were retained in Clades II and IV host, were also supported by bootstrap values of more than 87% (Figure S9B). Thirdly, four isolates (ERR3198481, UK; SRR7800511, US; GD69A, US; and DRR317383, Japan) each harbored a distinct pMAB625-3 from those harbored by other members of the clades to which these three isolates belonged (the taxa connected by blue lines).

Figure 4.

Phylogenetic association of pMAB625-3 and the host genome

(A) Phylogenetic tree of the pMAB625-3. The 53 sequences of pMAB625-3 detected in Figure 3 and 11 plasmids, which were previously reported to be identical to or similar to pMAB625-3, were used to create the phylogenetic tree using iQ-TREE.³² The consensus sequences belonging to the clade covered by the yellow box were identical to the sequence in Abs1, although the sequence of SRR18969469 was heterogeneous.

(B) The tanglegram of the pMAB625-3 and the host genome phylogenetic trees. Phylogenetic tree of the host genome was created using Gubbins based on the complete genome. The clades in the phylogenetic tree of the host genome were supported by the clades in Figure 3 pMAB625-3, which were connected by orange lines, exhibited widespread distribution. The taxa, which were connected to pMAB625-3 with green lines, retained pMAB625-3 that were unique to each clade. The isolates, which were connected by blue lines, retained the distinct pMAB625-3 that differed from pMAB625-3 harbored by other members of the clades to which the isolates belonged.

In addition, the verification of the mapping results also showed that pMAB625-3 sequences in the six isolates (DRR317373, Japan; SRR18969469, China; SRR7800518, US; SRR7800703, US; SRR7800409, US; SRR7800511, US) were heterogeneous because the unmutated reads and mutated ones were mixed (Figure S10). Thus, we ran freebays with the parameter “-ploidy 2” to detect SNVs in these mixed reads. Interestingly, the variant hotspot region was located at the CDS of the MMPL family transporter, which was the second most reliably highly expressed gene in Abs1-5 in the RNA-seq analysis. The variant pattern in the MMPL family transporter was classified into three types (Figure S10). Six sequences (DRR317373, Japan; SRR7800518, US; SRR7800703, US; SRR7800409, US; pGD36, Canada; pGD47, US) were classified as Type A, and one sequence each was classified as Type B (SRR18969469, China) and Type C (SRR7800511, US); the other 56 sequences had no mutation. There were three missense mutations and 14 or 15 silent mutations in Type A, 20 missense mutations and 43 silent mutations in Type B, and 47 missense mutations and 126 silent mutations in Type C (Table S6). Surprisingly, although many mutations were identified, no frameshift or nonsense mutations were observed. It was therefore speculated that the mutant MMPL family transporters were being translated into full-length proteins in isolates, which had pMAB625-3 with the mutant MMPL family transporter. Thus, these data indicated that pMAB625-3 was spread across strains and regions, each acquiring unique mutations, and some clinical isolates retaining pMAB625-3 expressed mutated MMPL family transporters.

We then classified pMAB625-2 sequences in terms of SVs. The sequences of pMAB625-2 were classified into four types (Figure 5A). Assuming that Type I, which was the sequence of pMAB625-2 in Abs1, was the base sequence, Type II and III had the large SVs at Region B, and Type II but not Type III had the large SVs at Region C. In contrast, Type IV had large SVs at Region A and C but not at Region B. Type I, Type III, and Type IV were detected in ABS, and Type II was detected in MAS, except for ERR3198514. In addition, Type III was detected only in ABS collected in China, and Type IV was detected only in ABS collected in the US. All pMAB625-2 sequences retained the ESX locus. These results indicated that pMAB625-2 had been horizontally transferred between subspecies and had acquired the SVs unique to each subspecies and region. Next, we analyzed the relationship between the phylogenetic distance of host genome sequences and the pMAB625-2 type (Figure 5B). The distribution of the pMAB625-2 type was almost consistent with the phylogenetic relationship of the host genomes. However, the pMAB625-2 type of Abs1 and ERR3198415 were different from those of the other isolates belonging to the same clade of the phylogenetic tree of the host genomes. Interestingly, Type II was detected in four MAS but also in one ABS (ERR3198415). This result indicated the possibility of direct or indirect contact between ERR3198415 and the other four MAS in the past, especially with ERR3198440 isolated from the same region.

Figure 5.

Classification of pMAB625-2 by structural variants and relationship with the phylogenetic distance of host bacterial genomes

(A) The mapping results of the read data from each isolate to pMAB625-2 were visualized using IGV. Each sequence was classified into 4 types by the pattern of structural variants (surrounded by black line squares).

(B) Relationship between the phylogenetic distance of the host bacterium and pMAB625-2 distribution.

Finally, we investigated the distribution of the plasmids identified in other NTMs except for M. abscessus in the analyzed strains in Table 4 to verify whether plasmids were transferred between M. abscessus and other NTM species. Recently, Wetzstein et al. reported on clinical and genomic features of Mycobacterium avium complex (MAC) in Europe,³³ which is the most frequently detected NTM globally.^34,35 In their report, they investigated the distribution of the 152 known plasmids registered in PLSDB and showed that plasmids identified in M. abscessus, including pGD69B-1 (the same as pGD69A-1) and pGD42-1, which are identical to pMAB625-3, were not detected in their dataset. Therefore, we investigated the distribution of these 152 plasmids indicated in Table S7 using same methods in this study. As a result, we detected the same plasmids indicated in Table 3 but did not find any other plasmids, which were previously identified in other NTM species, except for pMN23 (accession number: NC_010604.1) and pMFLV03 (accession number: NC_009341.1). pMN23 and pMFLV03 were present in two and one isolates, respectively (Table S8). pMN23 was almost identical to pMAB23 (ANI: 99.9%), so pMN23 was detected redundantly in ATCC 19977 and SRR7800408, which had pMAB23. Interestingly, pMFLV03 was detected in ERR3198408 (covered genome region: 78.7%, depth of coverage: 73.1x). pMN23 was originally identified in Mycobacterium marinum,³⁶ and pMFLV03 was identified in M. gilvum PYR-GCK (according to the GenBank registry information), indicating that these plasmids could be present across NTM species. These results suggested that plasmid distribution is strictly different from M. abscessus and other NTMs except for very few cases.

Discussion

Plasmids are mobile genetic elements that can transfer between different bacterial strains and modify host traits, such as virulence and antimicrobial susceptibility. Plasmids are horizontally transferred among bacteria via transformation and conjugation. Therefore, plasmid epidemiology, combined with core-genome analysis, can help infer past interactions between bacteria. In this study, we demonstrated that the multidrug-resistant ABS isolated from a patient with osteomyelitis had three plasmids (pMAB625-1, pMAB625-2, and pMAB625-3) that ATCC 19977 did not harbor. Additionally, analysis of M. abscessus identified worldwide indicated that the pMAB625 plasmids were transferred horizontally between regions and subspecies. It is also sometimes possible to predict historical contact by tracing the phylogenetic relationship of these plasmids. Dedrick et al. reported that pGD69-1 (pGD69A-1), which is identical to pMAB625-3, is the same as the other five plasmids.¹⁸ Lewin et al. identified pMabs-09-13 through WGS of the clinical isolates of M. abscessus from patients with cystic fibrosis and reported that pMabs-09-13 shares an 99.9% identity with pGD42-1, pGD69A-1, and pGD69B-1.³¹ These reports suggested that pMAB625-3 is spreading among many other M. abscessus isolates, but the detailed distribution of the pMAB625 plasmids worldwide and among other subspecies remained unclear. Our study shows that pMAB625-3 is the plasmid spreading across regions and subspecies in the highest number of clinical isolates of M. abscessus. This allowed us to explore the possibility of speculating on the precise classification of bacteria and understanding historical contact between bacterial strains. The tanglegram analysis revealed that the host-clade-specific clustering of pMAB625-3, regarding Clades II and IV (Figure 4B). Clustering of plasmids into host-clade-specific groups may arise from several processes, such as vertical transmission within host lineages, restricted horizontal transfer among related hosts, ecological or geographical co-distribution, or co-evolution.^37,38 However, more analysis is required for confirmation. The tanglegram analysis also revealed that four isolates (ERR3198481, UK; SRR7800511, US; GD69A, US; and DRR317383, Japan) each harbored a distinct pMAB625-3 from those harbored by other members of the clades to which these three isolates belonged (the taxa connected by blue lines indicated in Figure 4B). This result suggests that plasmid sequence analysis enables the detailed classification of isolates within the same clade of the host genome phylogeny. It also indicates that it is possible to predict the bacterial contact using the distribution of the unique variant, such as SRR7800511. For pMAB625-2, it is actually indicated that the past contact between the different subspecies is predictable based on the relationship between the phylogenetic tree of the host genome and the distribution of pMAB625-2 type.

As important findings, we found that 2.2% (10/462) and 11.5% (53/462) of the clinical isolates worldwide had pMAB625-2 and pMAB625-3, respectively (Table 4). This result provides an epidemiological insight into the spread of plasmids in M. abscessus beyond regions and strains. Other than person-to-person transmission of M.abscessus, this global spread of pMAB625 plasmids may be mainly due to increased global logistical activity and human mobility with associated interference of human and microbial habitats.³⁹ The habitat of M. abscessus is soil and water systems, and it is also possible that increased flooding due to recent climate change may have increased the chance of contact between the different strains. Analysis of plasmid distribution revealed that some plasmids (pMAB625-2, pMAB625-3, pGD19, pGD25-1, and pGD69A-2) are widely spread across strains, regions and subspecies, while others (pGD22-2 and pGD42-2) are limited to specific strains and areas (Figure 3). Therefore, it is speculated that pGD22-2 and pGD42-2, the identical isolates retaining these plasmids, had been spread within or between the regions, whereas pMAB625-2, pMAB625-3, pGD19, pGD25-1, and p69A-2, plasmids themselves, had been spread between different strains, subspecies, and regions through horizontal transfer. It is unknown what factors determine these differences in plasmid distribution. It is possible that the factors encoded by the plasmid or the periods from the emergence were different among the plasmids. In addition, we also showed that plasmid distribution was different from M. abscessus and MAC (Table S8), indicating that incompatibility also exists in the plasmids present in NTM, strictly distinguishing M. abscessus from other NTMs except for very few cases, such as pMN23 and pMFLV03.

We identified various factors associated with bacterial virulence and multidrug resistance on the pMAB625 plasmids. The components of the ESX secretion system, the MMPL family transporter, and the components of the TA system were encoded by the plasmids, and these genes were highly expressed in Abs1-5 compared with ATCC 19977. The ESX secretion system is involved in the export of various pathogenic proteins and metal ions associated with proliferation, survival in the cytosol of macrophages, and conjugation. Recent findings in M. tuberculosis also highlight that ESX-linked pathways can drive pathogenic evolution.⁴⁰ It has also been reported that ESX-1 components are required for sliding motility and biofilm formation in M. avium.²³ Therefore, given the close genetic distance between ESX-1 and ESX-P cluster 5, which includes the components of the ESX secretion system on the pMAB625-1 and pMAB625-2 (Figure 1B), the components of the ESX-P cluster 5 may be involved in biofilm formation. In addition, MMPL family transporters can transport antibiotics across the cell membrane. It has been reported that the increased expression of MMPL5 is associated with decreased susceptibility to bedaquiline and clofazimine in M. tuberculosis,^41,42 bedaquiline and clofazimine in M. intracellulare,⁴³ and thiactazone derivative in M. abscessus.⁴⁴ The function of the MMPL family transporter encoded by pMAB625-3 is unknown; however, the presence of mutational hotspots (Figure S10) indicates that this MMPL transporter may be compatible with a variety of substrates, including antimicrobial agents. Therefore, our results suggest that Abs strains enhanced defense mechanisms against antibiotics through the acquisition of the ESX secretion system and transporters, accompanied by changes in biofilm barriers and transport of antibiotics. However, further studies will be essential to determine whether the pMAB625 plasmids could affect AMR.

In conclusion, we reported the distribution of plasmids in M. abscessus and found that pMAB625-3 is the most widely distributed plasmid in M. abscessus worldwide. In addition, phylogenetic tree analysis and mutational analysis of the pMAB625 plasmids partially predict past bacterial contact. These findings provide a new perspective on the acquisition of genomic diversity, the origin, and transmission route of M. abscessus. These findings will also lead to the development of effective countermeasures to identify pathways of bacterial transmission and prevent serious bacterial outbreaks.

Limitations of the study

Although it is assumed that phylogenetically close plasmids with similar SNV and SV patterns were spread via bacterial contact, it cannot be excluded that they may have acquired identical mutation patterns in their respective clinical isolates without bacterial contact. To clarify this, it is necessary to calculate the mutation rate of the plasmid and compare the mutation acquisition period with the interval of appearance in the different strains. In addition, future studies are required to verify whether the pMAB625 plasmids induce multidrug resistance in M. abscessus by plasmid transfection. We also identified several SNVs in the Abs5 genome compared with Abs1 and ATCC 19977. Among them, the S314A mutation in PpbA, the target of carbapenem antibiotics, was suspected to be responsible for the differences in MIC against IPM. However, S314A was not located in any active motifs of PbpA (SXXK or SXN). Further experiments are required to verify whether the S314A mutation affects the MIC.

To infer the possible origin of pMAB625-3, we examined the temporal signal by assessing the relationship between genetic distance (root-to-tip distance) in the phylogenetic tree and sampling time using TempEst,⁴⁵ to confirm whether a time-scaled phylogenetic analysis using BEAST⁴⁶ was feasible. We analyzed the correlation between the genetic distance in the phylogenetic trees and sampling time, regarding the host genome and pMAB625-3. However, the regression analysis using TempEst showed almost no correlation (host genome, R² = 0.07; pMAB625–3, R² = 0.01), indicating a lack of temporal structure in the data. Consequently, it was difficult to estimate the origin and root position of pMAB625-3 using BEAST, which assumes that nucleotide substitutions accumulate at a constant rate over time. Based on the isolation year of the host strains, pMAB625-3, retained in ERR3198389, which was isolated from Spain in 2009, appeared to be the oldest among the analyzed samples, and pMAB625-2, retained in ERR3198415, which was isolated from the UK in 2009, appears to be the oldest.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Masashi Toyoda (mtoyoda@tmig.or.jp).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •All raw sequence data generated in this study have been deposited in the DDBJ database at https://www.ddbj.nig.ac.jp/ under the BioProject accession number BioProject: PRJDB16220. The complete genome sequences of the isolates are available in the DDBJ database. They can be accessed using the accession numbers, GenBank: AP028613, GenBank: AP028614, GenBank: AP028615, and GenBank: AP028616 for Abs1 (registered as Isolate57625); GenBank: AP028617, GenBank: AP028618, GenBank: AP028619, and GenBank: AP028620 for Abs2 (registered as Isolate57629); GenBank: AP028621, GenBank: AP028622, GenBank: AP028623, and GenBank: AP028624 for Abs3 (registered as Isolate57630); GenBank: AP028625, GenBank: AP028626, Genbank: AP028627, and GenBank: AP028628 for Abs4 (registered as Isolate57596); GenBank: AP028629, GenBank: AP028630, GenBank: AP028631, and GenBank: AP028632 for Abs5 (registered as Isolate57626), for the genome and three plasmids in each isolate.
• •The analyzed RNA-seq data have been deposited in the Genomic Expression Archive (GEA) at https://www.ddbj.nig.ac.jp/gea/ under the accession number E-GEAD-1137.
• •Raw data of Figures 2, S3, and S6, and Tables 1 and S2 were deposited on Mendeley at https://doi.org/10.17632/sgchhsycv7.1.
• •All data reported in this article will be shared by the lead contact upon request.
• •All codes used in this study are available at https://github.com/kensukon/M.abscessus_plasmid_analysis.
• •Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments

The authors thank the staff of the microbiology laboratory of Dokkyo Medical University Hospital for performing identification and drug susceptibility tests, and Dr. Yoshishige Masuda and Dr. Takashi Inamatsu at TMIG for supervising in clinical practice. Computations were partially performed on the NIG supercomputer at ROIS National Institute of Genetics. This work was supported by Japan Society for the Promotion of Science (JSPS) KAKENHI Grant Number JP19K08938.

Author contributions

Conceptualization, K.O. and A.Y.; formal analysis, K.O.; methodology, K.O., A.Y., and H. Koganemaru; data curation, K.O.; investigation, K.O., K. Kamada, and K. Kikuchi; validation, K.O.; writing – original draft, K.O. and A.Y.; writing – review and editing, K.O., Y.A. and M.T.; funding acquisition, A.Y. and K. Kikuchi; patient care and resources, H. Kitazawa, Y.I., T.S., and K.W.; supervision, M.T. All authors have read and approved the final article.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used DeepL and Grammarly to write English. After using this tool/service, the authors reviewed and edited the content as needed and took full responsibility for the content of the published article.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
M. abscessus subsp. abscessus type strain, ATCC 19977	ATCC	19977
M. abscessus subsp. massiliense type strain, JCM 15300	Japan Collection of Microorganisms (JCM), RIKEN BRC	15300
M. abscessus subsp. bolletii type strain, JCM 15297	JCM, RIKEN BRC	15297
Abs1, Abs2, Abs3, Abs4, and Abs5	This study	N/A
Chemicals, peptides, and recombinant proteins
Trypticase soy agar with 5% sheep blood	Beckton Dickinson	Cat.# 251239
BBL Mueller Hinton II Broth (Cation-Adjusted)	Beckton Dickinson	Cat.# 212322
Middlebrook 7H9 Medium	Sigma-Adlrich	Cat.# M0178
2% OGAWA medium	Kyokuto Pharmaceutical Industrial	Cat.# 551-08026
Lysozyme	Sigma-Aldrich	Cat.# L4919
Critical commercial assays
BrothMIC RGM	Kyokuto Pharmaceutical Industrial	Cat.# 551-10087
EZ-Beads	Promega	Cat.# AMR76813M
Wizard HMW DNA Extraction Kit	Promega	Cat.# A2920
Qubit BR DNA Quantitation Kit	Thermo Fisher Scientific	Cat.# Q32850
Ion Xpress Plus Fragment Library Kit	Thermo Fisher Scientific	Cat.# 4471252
Ion Xpress Barcode Adapters 1-16 kit	Thermo Fisher Scientific	Cat.# 4471250
TaqMan Quantitation Kit	Thermo Fisher Scientific	Cat.# 4468802
Ion PI Chip kit v3	Thermo Fisher Scientific	Cat.# A26711
Ultrapure Dnase/RNase-free Distilled Water	Thermo Fisher Scientific	Cat.# 10977015
SUPERase-In RNase Inhibitor	Thermo Fisher Scientific	Cat.# AM2696
mirVana miRNA Isolation Kit	Thermo Fisher Scientific	Cat.# AM1561
IonXpress RNA-Seq Barcode 1-16 Kit	Thermo Fisher Scientific	Cat.# 4475485
Ion PI Hi-Q OT2 200 Kit v2	Thermo Fisher Scientific	Cat.# A26434
Ion Total RNA-Seq Kit v2	Thermo Fisher Scientific	Cat.# 4479789
PowerUp SYBR Green Master Mix	Thermo Fisher Scientific	Cat.# A25741
KOD One PCR Master Mix	TOYOBO	Cat.#LMM-101
Short Read Eliminator Kit	PacBio	Cat.# 102-208-300
Ligation Sequencing Kit	Oxford Nanopore Technologies	Cat.# SQK-LSK109
Native Barcoding Expansion 1-12	Oxford Nanopore Technologies	Cat.# EXP-NBD104
Flow cell (R9.4.1)	Oxford Nanopore Technologies	Cat.# FLO-MIN106
NEBNext rRNA Depletion Kit (Bacteria)	NEB	Cat.# E7850
Deposited data
All raw sequence data generated in this study	DDBJ	BioProject: PRJDB16220
Complete genome sequences (genome, pMAB625-1, pMAB625-2 and pMAB625-3) of the Abs1 (registered as Isolate57625)	DDBJ	GenBank: AP028613 GenBank: AP028614 GenBank: AP028615 GenBank: AP028616
The analyzed RNA-seq data	Genomic Expression Archive (GEA)	E-GEAD-1137
Raw data of Figures 2, S3, and S6, and Tables 1 and S2	Mendeley Data	https://doi.org/10.17632/sgchhsycv7.1
Oligonucleotides
Forward primer for validation of large deletion in Figure S2 for Deletion_1, CAAGAACTGATGGCTGAAATCCTAATG	This study	N/A
Reverse primer for validation of large deletion in Figure S2 for Deletion_1, AAAGTGATGCTTGCTCTGCCAGTC	This study	N/A
Forward primer for validation of large deletion in Figure S2 for Insertion_1, GTTAGATATGTGCAGGAGTTCCGAGTG	This study	N/A
Reverse primer for validation of large deletion in Figure S2 for Insertion_1, ACCAATACTGCCGAATTCTCTAAAAGC	This study	N/A
Forward primer for validation of large deletion in Figure S2 for Insertion_2, GTGACATGATCTTGCTGCTCAAAAG	This study	N/A
Reverse primer for validation of large deletion in Figure S2 for Insertion_2, GCTATCTGTCGCAGTACTTCGTGACC	This study	N/A
Forward Primers for plasmid copy number analysis in Figure S6 for HrpA (MAB), GGATCTGGTCCACAAGACCTACAG	This study	N/A
Reverse Primers for plasmid copy number analysis in Figure S6 for HrpA (MAB), AAAGTGATGCTTGCTCTGCCAGTC	This study	N/A
Forward Primers for plasmid copy number analysis in Figure S6 for HrpA (no MAB), ACCCTGGTGAAGTTGATGACCGAC	This study	N/A
Reverse Primers for plasmid copy number analysis in Figure S6 for HrpA (no MAB), AGGAGGAAGTCGATGTTGAGGCTG	This study	N/A
Forward Primers for plasmid copy number analysis in Figure S6 for pMAB625-1(LOCUS_52450), TCTACTACGCAGACCCTGCACATATC	This study	N/A
Reverse Primers for plasmid copy number analysis in Figure S6 for pMAB625-1(LOCUS_52450), AACTGCTGCCACATCGCTTTCTG	This study	N/A
Forward Primers for plasmid copy number analysis in Figure S6 for pMAB625-2(LOCUS_53740), GACCGATCACGTCATCATTCCTAATC	This study	N/A
Reverse Primers for plasmid copy number analysis in Figure S6 for pMAB625-2(LOCUS_53740), GAACCTCAGCCTCAATATGCTCCTC	This study	N/A
Forward Primers for plasmid copy number analysis in Figure S6 for pMAB625-3(LOCUS_54770), CGACCAGGTACACATCCTGACCAC	This study	N/A
Reverse Primers for plasmid copy number analysis in Figure S6 for pMAB625-3(LOCUS_54770), GAGTTGGCTGATTCCGTCTTGGAG	This study	N/A
Software and algorithms
R (ver. 4.3.1)	R Core Team (2024)	https://www.R-project.org/
MinKNOW (ver.21.06.0)	Oxford Nanopore Technologies	https://nanoporetech.com/
Guppy (ver. 5.0.7)	Oxford Nanopore Technologies	https://nanoporetech.com/
NanoPlot (ver. 1.38.0)	De Coster et al.47	https://github.com/wdecoster/NanoPlot
Porechop (ver. 0.2.4)	Wick et al.48	https://github.com/rrwick/Porechop
Filtlong (ver. 0.2.1)	Wick	https://github.com/rrwick/Filtlong
Fmlrc (ver. 1.0.0)	Wang et al.49	https://github.com/holtjma/fmlrc
Flye (ver. 2.9-b1768)	Kolmogorov et al.50	https://github.com/mikolmogorov/Flye
Pilon (ver. 1.24.0)	Walker et al.51	https://github.com/broadinstitute/pilon
QUAST (ver. 5.0.2)	Gurevich et al.52	https://quast.sourceforge.net/
Bandage (ver. 0.8.1)	Wick et al.53	https://rrwick.github.io/Bandage/
DFAST (ver.1.5.0)	Tanizawa et al.54,55	https://dfast.ddbj.nig.ac.jp/
FastANI (ver.1.34)	Jain et al.56	https://github.com/ParBLiSS/FastANI
CheckM (ver. 1.2.4)	Park et al.57	https://github.com/Ecogenomics/CheckM
Mauve (ver. 2.4.0)	Darling et al.58	https://darlinglab.org/mauve/mauve.html
Abricate (ver.1.0.1)	Seemann	https://github.com/tseemann/abricate
BWA-MEM (ver. 0.7.17-r1188)	Li59	https://github.com/lh3/bwa
Samtools (ver.1.12)	Danecek et al.60	https://github.com/samtools/samtools
Freebayes (ver. 1.3.5)	Garrison et al.61	https://github.com/freebayes/freebayes
SnpEff (ver. 5.2)	Cingolani et al.62	https://pcingola.github.io/SnpEff/
IGV (ver.2.17.4)	Robinson et al.63	https://igv.org/
Minimap2 (ver. 2.21-r1071)	Li64	https://github.com/lh3/minimap2
Bowtie2 (ver. 2.5.4)	Langmead and Salzberg65	https://github.com/BenLangmead/bowtie2
HTseq (ver. 1.99.2)	Anders et al.66	https://github.com/htseq/htseq
R package, TCC (ver. 1.26.0)	Sun et al.67	https://bioconductor.org/packages/release/bioc/html/TCC.html
Snippy (ver. 4.6.0)	Seemann	https://github.com/tseemann/snippy
IQ-TREE (ver.2)	Minh et al.32	https://www.iqtree.org
iTOL (ver.6.7)	Letunic and Bork68	https://itol.embl.de
NGphylogeny.fr	Lemoine et al.24	https://ngphylogeny.fr
Fastp (ver. 0.23.4)	Chen69	https://github.com/OpenGene/fastp
Seqkit (ver.2.4.0)	Shen et al.70	https://github.com/shenwei356/seqkit
Bedtools (ver. 2.30.0)	Quinlan and Hall71	https://github.com/arq5x/bedtools2
Gubbins (ver.3.3.1)	Croucher29	https://github.com/nickjcroucher/gubbins
MAFFT (ver.7)	Katoh et al.72	https://mafft.cbrc.jp/alignment/software/
SplitsTree (ver.6.6.1)	Huson et al.73	https://github.com/husonlab/splitstree6
TempEst (ver.1.5.3)	Rambaut et al.45	https://tree.bio.ed.ac.uk/software/tempest/
R package, ggtree (ver 3.10.1)	Yu et al.30	https://bioconductor.org/packages/release/bioc/html/ggtree.html
R package, RVAideMemoire (ver. 0.9-83-7)	Herve	https://github.com/cran/RVAideMemoire
Torrent Suite v5.0.5 software	Thermo Fisher Scientific	https://www.thermofisher.com/

Experimental model and study participant details

Microbe strains

Species and strains

The primary focus of this study was M. abscessus. Regarding Abs1, Abs2, Abs3, Abs4, and Abs5, specimens were obtained from a patient with osteomyelitis at the five time points in six months indicated in Table 1. Each strain was isolated from the specimen at Shizuoka Children’s Hospital for diagnosis. The other clinical isolates used for PCR or WGS were stored at Tokyo Metropolitan Institute for Geriatrics and Gerontology or Tokyo Women’s Medical University or Dokkyo Medical University.⁴

Type strains and sources

The type strain of ABS, ATCC 19977, was purchased from ATCC (Manassas, VA, US). The type strain of MAS, JCM 15300, and the type strain of BOL, JCM 15297, were provided by Japan Collection of Microorganisms, RIKEN BRC which is participating in the National BioResource Project of the MEXT, Japan.

Human study participants

Sample origin

The clinical isolates Abs1–Abs5 were derived from specimens obtained from a single human patient. The influence of sex and gender on the study results could not be determined due to the single-case nature of the primary clinical source.

Ethical oversight and allocation

Ethical approval for this study was obtained from the Ethics Committee of Dokkyo Medical University (approval number 28045). Each strain was isolated for diagnostic purposes at Shizuoka Children’s Hospital. Since this study focused on the longitudinal analysis of isolates from a single clinical case, allocation to experimental groups was not applicable.

Other models

This study did not employ any animal models, primary cell cultures, or established cell lines.

Method details

Growth curve

ATCC 19977 and Abs1-5 were inoculated onto trypticase soy agar with 5% sheep blood (Beckton Dickinson, Franklin Lakes, NJ, US) and incubated at 30 °C for five days. Next, some colonies were inoculated into BBL Mueller Hinton II Broth (Cation-Adjusted) (Beckton Dickinson) and incubated overnight at 30°C. These culture media were measured for absorbance (OD₆₀₀), from which they were prepared to 0.1 by medium. Seventy microliter of the adjusted culture solution was inoculated into 7 mL of BBL Mueller Hinton II Broth (Cation-Adjusted) and incubated at 30 °C for eight days with agitation (200 rpm). OD₆₀₀ was measured on days one to eight after inoculation. Statistical analysis was performed using R ver. 4.3.1.

Antimicrobial susceptibility testing

ATCC 19977, Abs1 and Abs5 were subcultured on 2% OGAWA medium (Kyokuto Pharmaceutical Industrial, Tokyo, Japan) at 30 °C for five days. Antimicrobial susceptibility testing was performed using BrothMIC RGM (Kyokuto Pharmaceutical Industrial) following CLSI M24 3rd ed.¹⁴ MICs were determined as the minimum concentration at which bacteria had not grown on day four (all antibiotics) and day 14 (clarithromycin). The test was repeated three times and all MICs were confirmed to be the same between the experiments, except for imipenem (IPM) for the isolates. For IPM, the MICs of the type strain were confirmed to be the same between all three experiments, while for the MICs of Abs1 and Abs5, they were confirmed to be the same between two out of three experiments, as indicated in Table 1.

Mycobacterial DNA isolation and whole-genome sequencing

ATCC 19977 and the isolates of M. abscessus were stored in frozen stock containing skim milk and 5% (v/v) glycerol. Each strain was inoculated from frozen stocks into Middlebrook 7H9 Medium (Sigma-Aldrich, St. Louis, MO, US) containing 10% (v/v) oleic acid-albumin-dextrose-catalase and 0.2% (v/v) glycerol and cultured at 30 °C for six days with agitation (200 rpm). Bacterial genomic DNA for short-read sequencing was extracted as previously reported.⁷⁴ The quality of the extracted DNA was confirmed by agarose gel electrophoresis, and the concentration was measured using Qubit BR DNA Quantitation Kit (Thermo Fisher Scientific, Waltham, MA, US). Genomic DNA was fragmented, and the sequencing library was prepared using the Ion Xpress Plus Fragment Library Kit and the Ion Xpress Barcode Adapters 1-16 kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The quantity of sequencing library was assessed using the Ion Library TaqMan Quantitation Kit (Thermo Fisher Scientific). Emulsion PCR for sequence template synthesis was performed with the Ion PI Hi-Q OT2 200 Kit v2 using the Ion One Touch II System and the Ion One Touch II ES (Thermo Fisher Scientific). Sequencing was performed using the Ion PI Chip kit v3 and the Ion Proton System (Thermo Fisher Scientific). Data were collected using the Torrent Suite v5.0.5 software (Thermo Fisher Scientific).

De novo genome assembly

High-molecular-weight genomic DNA for long-read sequencing was extracted as follows. The bacterial pellet was collected by centrifugation and homogenized using EZ-beads (Promega, Madison, WI, US) and Shakeman 6 (Bio Medical Science, Tokyo, Japan). The homogenized pellet was lysed with Lysozyme (Sigma-Aldrich) at 37 °C for one hour. Next, DNA was extracted using Wizard HMW DNA Extraction Kit (Promega) following the manufacturer’s instructions. Fragmented DNA in the extracted DNA was removed using the Short Read Eliminator Kit (PacBio, Menlo Park, CA, US). The sequencing library for long-read sequencing was prepared using the Ligation Sequencing Kit (SQK-LSK109, Oxford Nanopore Technologies, Oxford, UK) and the Native Barcoding Expansion 1-12 (EXP-NBD104, Oxford Nanopore Technologies). In brief, the barcode sequences for multiplex sequencing were attached to the ends of the HMW genomic DNA. In addition, the sequencing adapters were ligated to the end of HMW genomic DNA. The prepared DNA was loaded into a flow cell (FLO-MIN106, Oxford Nanopore Technologies) attached to MinION Mk1B (Oxford Nanopore Technologies). Sequencing data were acquired using MinKNOW ver.21.06.0 (Oxford Nanopore Technologies), and base calling was performed using Guppy ver. 5.0.7 (Oxford Nanopore Technologies) with high accuracy mode using the National Institute of Genomics supercomputer system (NIG, Shizuoka, Japan). Data quality was confirmed using NanoPlot ver. 1.38.0⁴⁷ and adapter sequences were trimmed using Porechop ver. 0.2.4.⁴⁸ Raw reads shorter than 1000 bp (Abs1, Abs2, Abs3, Abs4) or 5000 bp (Abs5) were removed using Filtlong ver. 0.2.1 (https://github.com/rrwick/Filtlong). Filtered reads were corrected using Fmlrc ver. 1.0.0⁴⁹ with the data of short-read sequencing acquired by the Ion Proton Sequencer. De novo assembly was performed using Flye ver. 2.9-b1768⁵⁰ with corrected long-reads. Assembled genome sequences were polished using Pilon ver. 1.24.0,⁵¹ evaluated using QUAST ver. 5.0.2,⁵² and visualized using Bandage ver. 0.8.1.⁵³ Genome annotation was performed using DFAST ver.1.5.0.^54,55 Genome completeness was automatically calculated using CheckM ver. 1.2.4⁵⁷ based on the marker set of M. abscessus during annotation.

Plasmid sequence analysis

Known plasmid sequences were downloaded from PLSDB¹⁹ by the taxonomic search for “abscessus”. Average nucleotide identity (ANI) was calculated using FastANI ver.1.34⁵⁶ and structural variants (SVs) were analyzed using Mauve ver. 2.4.0.⁵⁸

Homology search was performed by blast+⁷⁵ using the amino acid sequence as the query or by Abricate ver.1.0.1 (http://github.com/tseemann/abricate) using the nucleotide sequences as the query. NCBI AMRFinderPlus,⁷⁶ CARD,¹⁷ Resfinder,⁷⁷ ARG-ANNOT,⁷⁸ VFDB,⁷⁹ PlasmidFinder,⁸⁰ EcOH,⁸¹ and MEGARes 2.00⁸² were used as the database for the search of AMR genes. Mycobrowser²⁰ was used as the database for the homology search of the Mycobacterium genes.

SNVs and SVs analysis

Short-read data acquired by the Ion Proton Sequencer were mapped using BWA-MEM ver. 0.7.17-r1188⁵⁹ against the reference sequence (CU458896.1, CU458745.1). Output BAM files were sorted using samtools ver.1.12⁶⁰ Valiant calling was performed using Freebayes ver. 1.3.5,⁶¹ and variants were annotated using SnpEff ver. 5.2⁶² with annotated genome sequence. The positions of single nucleotide variants (SNVs) were visualized using IGV⁶³ (Broad Institute, Cambridge, MA, US). Assembled genome sequences were aligned using Mauve ver. 2.4.0⁵⁸ for SVs analysis. In this study, we define SV as genomic alterations that are larger than one kilobase. Mapping the short- and long-read to the assembled genome sequence using minimap2 ver. 2.21-r1071⁶⁴ was performed for the validation of assembly. We visually checked the mapping results using IGV for abnormally high mapping coverage and whether reads were mapped in reverse direction from middle of read. We also confirmed the integrity of the de novo assembly by mapping long reads to the sequences of SVs.

PCR for the validation of SVs was performed using KOD One PCR Master Mix (TOYOBO, Osaka, Japan) and the primers indicated in Table S1. Amplified DNA was confirmed using agarose gel electrophoresis following staining with SYBR Safe Gel Stain (Thermo Fisher Scientific).

RNA-seq analysis

Total RNA was extracted as follows: The bacterial pellet was collected by centrifugation and suspended in UltraPure DNase/RNase-free Distilled Water (Thermo Fisher Scientific) and homogenized using EZ-beads (Promega) and Shakeman 6 (Bio Medical Science). SUPERase-In RNase Inhibitor (Thermo Fisher Scientific) was added to the homogenized pellet. Next, the homogenized pellet was lysed by Lysozyme (Sigma-Aldrich) at 37 °C for an hour. Total RNA was purified by mirVana miRNA Isolation Kit (Thermo Fisher Scientific). The size and quality of total RNA were assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, US) and the RNA 6000 Nano Kit (Agilent Technologies). rRNA was depleted using the NEBNext rRNA Depletion Kit (Bacteria) (NEB, Ipswich, MA, US) from 500 ng of total RNA. The whole transcriptome library was prepared using the Ion Total RNA-Seq Kit v2 (Thermo Fisher Scientific) as follows: First, the rRNA-depleted RNA was fragmented by RNase III at 37°C for 10 minutes and purified using the Magnetic Beads Cleanup Module. Next, adapter sequences were ligated and reverse transcription was performed. The whole transcriptome library was amplified and barcoded using the IonXpress RNA-Seq Barcode 1-16 Kit (Thermo Fisher Scientific). The quantity of the amplified whole transcriptome library was assessed using the Ion Library TaqMan Quantitation Kit (Thermo Fisher Scientific). Emulsion PCR for sequence template synthesis was performed using the Ion PI Hi-Q OT2 200 Kit v2 with the Ion One Touch II System and the Ion One Touch II ES (Thermo Fisher Scientific). Sequencing was performed using the Ion Proton System with Ion PI Chip kit v3 (Thermo Fisher Scientific). Data was collected using the Torrent Suite ver. 5.0.5 software (Thermo Fisher Scientific). Raw read data were filtered using fastp ver. 0.23.4⁶⁹ with “--length_requrired 35” to remove the reads potentially mapped to the multiple regions of the reference sequence. Sequence reads were mapped using Bowtie2 ver. 2.5.4⁶⁵ against the assembled genome sequence of Abs1. Reads with “XS”, which were potentially mapped to the multiple regions of the reference sequence were removed from the output sam files using “grep -v “XS:” command. Mapped reads were counted using HTseq ver. 1.99.2.⁶⁶ Merged transcript count data were normalized using TCC ver. 1.26.0⁶⁷ and differentially expressed genes (DEGs) were also identified using TCC package. To control for multiple testing, we applied a false discovery rate (FDR) < 0.01 as the threshold, as TCC detects DEGs through statistical testing with adjustment for false positives rather than relying solely on fold-change cutoffs. For reference, the output includes m.value, corresponding to the log2(Abs strains/ATCC 19977), and a.value, defined as (log2[Abs strains] + log2[ATCC 19977])/2, which represents the average expression level on the log scale (i.e., the A-axis in MA-plots). TCC ver.1.26.0 used log2(Abs strains/0.88) as m.value when normalized value of ATCC 19977 was zero. We used the log2(Abs strains) as a.value (represented as a.value.re in Table S5) when normalized value of ATCC 19977 was zero, and used log2(ATCC19977) as a.value (represented as a.value.re in Table S5) when normalized value of Abs strains was zero to visualize the expression levels of genes on plasmid, whose expression levels in ATCC 19977 were almost zero. These values are summarized in Table S5.

Phylogenetic tree analysis

The genome sequences of the various isolates shown in Figure S5 for phylogenetic tree analysis were obtained from the NCBI database and are listed in Table S7. Core-genome SNVs were detected using Snippy ver. 4.6.0 (https://github.com/tseemann/snippy) with each assembled genome sequence and the type strain of ABS (ATCC 19977, CU458896.1) as the reference. The concatenated core-genome SNVs were also aligned using the snippy-core script. The phylogenetic tree was constructed using IQ-TREE ver.2³² using the maximum likelihood criterion and rendered using iTOL ver.6.7.⁶⁸ The phylogenetic trees of pMAB625 sequences and host genome sequences in Figure 4 were created as follows. Complete genome sequences for GD42, GD69, as well as complete plasmid sequences for pGD36, pGD47, pGD108A, pGD95, pGD18, pGD23, pGD72, and pGD87, were downloaded from NCBI data repository. The other genome or plasmid sequences were prepared using Snippy ver. 4.6.0 (https://github.com/tseemann/snippy). Multiple alignment were conducted usinf MAFFT ver.7.⁷² Phylogenetic trees were constructed using Gubbins²⁹ to remove recombinant regions. For the phylogenetic tree analysis regarding the components of the ESX secretion system, the amino acid sequences of ESX loci (eccB, eccC, eccD, mycP, and eccE) were concatenated and used as input for NGphylogeny.fr²⁴ using the PhyML+SMS inference method. Tangleglam were created using SplitsTree ver.6.6.1.⁷³ Regression analysis to assess the correlation between the sampling time and the genetic distance regarding the phylogenetic trees of host genome and pMAB625-3 were conducted using TempEst ver.1.5.3.⁴⁵

PCR for plasmid detection and copy number analysis

PCR was performed using KOD One PCR Master Mix (TOYOBO) and the primers indicated in Table S1. Amplified DNA was confirmed using agarose gel electrophoresis. The plasmid copy number was analyzed by qPCR using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific). qPCR was performed using the QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific). Five nanogram of genomic DNA was used for a 20 μL reaction volume of qPCR. The plasmid copy number is shown as a relative value normalized using the copy number of HrpA, which is present in one copy on the genome.

Mapping the read data of various clinical isolates to the genome sequence of Abs1

Raw read data of M. abscessus isolated from Japan, Taiwan,⁸ China,²⁸ the US, the United Kingdom (UK) and Spain were obtained from the NCBI sequence reads archive (SRA). Raw read data were trimmed using fastp ver. 0.23.4⁶⁹ and reduced using seqkit ver.2.4.0⁷⁰ with “-p N –rand-seed 11” options. N were determined by each data volume to 200 Mb for the Ion Proton single-end data and 100 Mb for the one of the Illumina pair-end data (total 200 Mb with paring) to equalize the data volume with the aim of making it easier to evaluate the mapping coverage between isolates. The integrity of the read pairs is maintained while the read data is randomly sampled using “--rand-seed 11“ option. Reads were mapped using BWA-MEM ver. 0.7.17-r1188 against the genome and plasmid sequences of Abs1 and other plasmids which were hit using PLSDB by the taxonomic searching of “abscessus” with removing the redundant plasmid sequences (ANI [%]>97, without large SV [>8 kbp]). The depth of coverage was calculated using Qualimap ver. 2.2.2-dev.⁸³ The covered genome region was calculated using bedtools ver. 2.30.0.⁷¹ We considered the isolate whose depth of coverage was higher than 10x and the covered genome region was higher than 75% as the plasmid-harboring isolate. The 152 sequences for the analysis of the distribution of plasmids identified in other NTM and previously reported³³ were listed in Table S2. For the phylogenetic tree analysis, the aligned assumed genome sequences using Snippy-core script were used as input for Gubbins ver.3.3.1.²⁹ The phylogenetic tree was rendered using ggtree ver 3.10.1.³⁰

Quantification and statistical analysis

Statistical analysis in this study were performed using R ver. 4.3.1. Statistical significance in Figure S2 were calculated using one-way ANOVA followed by Dunnett’s test. Statistical significance shown in Figure S7 was calculated using Mann-Whitney’s U test. The data indicated in Figures S2 and S7 represent the mean of three independent experiments, and the error bars represent the standard deviation. Statistical analysis of the number of isolates retaining plasmid in Table 4 was performed using Fisher’s exact test with Holm’s method for correction of multiple comparisons using RVAideMemoire ver. 0.9-83-7 (https://github.com/cran/RVAideMemoire).

Published: February 26, 2026