Whole-genome recombination and dynamic accessory genomes drive the phenotypic diversity of Mycobacterium abscessus subspecies

Abstract

Background

Mycobacterium abscessus (Mab) is a multidrug-resistant bacterial pathogen capable of causing widespread infections, often with a poor prognosis in susceptible populations. Mab comprises three distinct subspecies that exhibit phenotypic diversity and genetic heterogeneity.

Methods

We performed whole-genome sequencing and phenotypic antimicrobial susceptibility testing on 109 Mab isolates collected at zhongshan hospital from 2018 to 2023.

Results

The results indicate that recombination, especially distributed conjugation transfer, promotes the formation and sustained diversity of Mab subspecies. Through pangenome analysis, the synergistic gain/loss of accessory genes was found to contribute to different metabolic profiles and the ability to adapt to oxidative stress, facilitating strain adaptation to host environments. We conducted phenotypic antimicrobial susceptibility testing, revealing resistance to macrolide antibiotics differed among subspecies. We identified 24 genes whose gain or loss may increase the likelihood of macrolide resistance, including those involved in biofilm formation, the stress response, virulence, biotin synthesis, and fatty acid metabolism. Genomic variations within Mab species may have significant implications for disease epidemiology, infection pathogenesis, and host interactions.

Conclusions

Our findings provide a valuable genetic basis for the success of Mab as a highly adaptive and drug-resistant pathogen, informing current efforts to control and treat Mab infections, including strategies targeting specific sequence types or lineages.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12941-025-00804-9.

Introduction

Mycobacterium abscessus (Mab) is a rapidly growing pathogenic nontuberculous mycobacterium that primarily causes pulmonary and extrapulmonary infections. Mab infection predominantly affects immunocompromised individuals or those with underlying lung diseases, with global infection rates steadily increasing in recent years [1]. There are three Mab subspecies: Mycobacterium abscessus subsp. abscessus, Mycobacterium abscessus subsp. massiliense, and Mycobacterium abscessus subsp. bolletii. M.a. abscessus and M.a. massiliense are the main pathogenic subspecies. The subspecies exhibit phenotypic heterogeneity and distinct epidemiological characteristics, including varied prevalence rates and environmentally mediated clustered infections associated with specific dominant circulating clones [2]. Epidemiological studies indicate that the infection rate of M.a. abscessus infection is higher than that of M.a. massiliense [3]. However, in outbreak events among cystic fibrosis (CF) and non-CF patients, most isolated strains are classified as M.a. massiliense [4, 5]. Treatment outcomes and predictive factors differ between M.a. abscessus and M.a. massiliense. A recent meta-analysis revealed an overall treatment success rate of 45.6% for nontuberculous mycobacterial pulmonary disease (NTM-PD), with success rates of 33.0% for abscessus infections and 56.7% for massiliense infections [6]. Differences in adaptability, toxicity, and multidrug resistance among subspecies are closely related to prognosis. Mab colonies are smooth or rough in morphology, with rough colonies appearing to have greater virulence, although the genetic basis for these differences remains to be clarified [7]. Additionally, while M.a. massiliense has a truncated erm41 gene and relatively low clarithromycin resistance, the infection cure rate remains low. Therefore, the genetic determinants of pathogenic adaptation among the different Mab subspecies warrant further investigation.

Unlike the conserved genome of Mycobacterium tuberculosis, the genomic evolution of Mab appears to be governed primarily by intraspecies and interspecies recombination events [8, 9]. During these recombination processes, bacteria can exchange genes related to virulence, antibiotic resistance, and environmental adaptation to confront new challenges such as phage predation, antibiotic treatment, or environmental disturbances [10]. This recombination may facilitate speciation and the ongoing diversification of Mab.

De novo assembled pangenome analysis allows the exploration of the entire genetic repertoire of the species. Identifying core and strain-specific Mab genes is expected to elucidate the role of genetic variation in the pathogen’s virulence and pathogenesis. The genome acquires or loses specific functional genes in a coordinated manner while adapting to environmental or host conditions, thereby driving the diversity of subspecies [11]. Additionally, the gain and loss of genes suggest significant variability in the genetic backgrounds within the species. The evidence indicates that different genetic backgrounds in natural populations can alter the propensity of bacteria to adapt to specific point mutations, which are known to contribute to antibiotic resistance [12].

Long-term and continuous population genomic studies of Mab offer insights into its speciation, diversification, and potential genetic mechanisms underlying host adaptation. In this study, we performed whole-genome sequencing of Mab collected from various infection sites within our center’s Mab infection cohort from 2018 to 2023. Our findings elucidate the bacterial determinants of infection across different Mab subspecies, primarily linked to extensive recombination events, dynamic accessory genomes, and genetic backgrounds contributing the evolution of antibiotic resistance.

Methods

Patient data and bacterial isolate collection

We performed whole-genome sequencing on 109 clinical isolates of Mab collected from patients hospitalized with Mab infections at Zhongshan Hospital, Fudan University, from 2018 to 2023. All the isolates were cultured on Middlebrook 7H10 solid media (BD, France) supplemented with 10% (vol/vol) oleic acid-albumin-dextrose-catalase (Thermo Fisher Scientific, USA) and incubated at 37 °C for 7 days, with sequencing performed on multiple colonies collected. Only the first isolate from each patient was included (duplicates were excluded), and we collected information on the patients’ infection sites and isolate sources. This retrospective study was approved by the Ethical Committee of Zhongshan Hospital, Fudan University (approval number: B2022-513R).

Phenotypic antimicrobial susceptibility testing

The isolates were subjected to culture-based drug susceptibility testing (DST) for their response to the following 15 antimicrobial agents: amikacin (AMK), ciprofloxacin (CIP), moxifloxacin (MXF), trimethoprim-sulfamethoxazole (SXT), linezolid (LZD), ceftriaxone (AXO), cefepime (FEP), cefoxitin (FOX), tobramycin (TOB), tigecycline (TGC), minocycline (MIN), doxycycline (DOX), amoxicillin-clavulanate (AMC), imipenem (IMI), and clarithromycin (CLA) [1]. The minimum inhibitory concentrations (MICs) of these 15 agents were determined for 109 Mab isolates via Sensititre RapmycoI MIC plate (Thermo Fisher Scientific) and broth microdilution methods per the Clinical and Laboratory Standards Institute (CLSI) guidelines M24 and CLSI supplement M62 [13].

Whole-genome sequencing

Mab samples were obtained from clinical specimen isolates and subsequently subcultured on solid medium to generate multiple colonies for sequencing. The quality-checked DNA samples were prepared with inserts of approximately 400 bp and sequenced on the Illumina HiSeq ×10 platform. Paired-end (PE150) sequencing was conducted, achieving a minimum coverage depth of 100× for each sample.

Phylogenetic tree

The raw sequencing data were preprocessed and quality controlled using Fastp [14]. Core genome single nucleotide polymorphisms (SNPs) were mapped to the Mycobacterium abscessus ATCC19977 reference genome (GenBank accession number GCA_000069185) using Snippy (https://github.com/tseemann/Snippy), and recombination sites were inferred using Gubbins [15]. Phylogenetic trees were constructed with RAxML [16] using the GTR-GAMMA model to generate maximum likelihood trees, which were subsequently visualized and annotated using iTOL [17].

De novo assembly and genomic nucleotide identity analysis

Draft genomes were assembled using SPAdes in careful mode, with read correction, automatic k-mer sizing, and mismatch correction [18]. The quality of the assembled genomes was assessed using QUAST [19] and CheckM [20]. Genomes with completeness below 90% or contamination above 5% were excluded from downstream analyses. Pairwise genomic average nucleotide identities (ANIs) were calculated using fastANI with default parameters on the basis of the assembled genomes [21].

Multilocus sequence typing (MLST) determination

We performed MLST typing of the 109 Mab isolates in this study based on the seven-housekeeping-gene scheme proposed by Song Yee Kim et al. [22]. The scheme includes the genes cya, gnd, murC, pta, purH, argH, and rpoB. Allelic profiles and sequence types (STs) were assigned using the PubMLST database (https://pubmlst.org/saureus/).

Pangenomic construction and whole-genome association analysis

Draft genomes were annotated using Prokka [23]. The pangenomes of Mab and its subspecies were independently reconstructed using Panaroo in clean mode, set to moderate [24]. Scoary was used to screen accessory pangenome genes associated with specific bacterial phenotypes. The presence-absence of accessory genes and their association with subspecies were analyzed using Scoary v1.6.16 [25], with gene presence-absence matrices generated by Panaroo and binary matrices of subspecies classifications. Pairwise comparisons were considered significant when the Benjamini–Hochberg adjusted p-value was less than 0.05. DeCoTUR was employed to detect co-evolving genes in large bacterial genome datasets [26]. Gene association/dissociation analyses were performed for M. abscessus subsp. abscessus and subsp. massiliense using DeCoTUR. Pairwise distance matrices were calculated using Mashtree (v2.1) to obtain Mash distances [27]. Gene presence-absence matrices and distance matrices were input into DeCoTUR to generate co-evolution scores. Strains with scores greater than 0.35 were used to construct gene-gene co-evolution networks, which were visualized and clustered using Gephi v0.9.2 (https://github.com/gephi/gephi).

Determining rates of gene gain and gene loss in the lineage pangenomes

We used Panstripe v0.3.0 to estimate gene gains and losses in the pangenome of each subspecies [28]. The core genome SNP phylogeny and the gene presence-absence matrix generated by Panaroo were used as inputs. Ancestral state reconstruction of gene gain and loss events on each branch was performed using maximum parsimony implemented in Panstripe. Panstripe then compared the branch lengths in the phylogeny to the number of gene gain and loss events inferred for each branch using a generalized linear model, with 1,000 bootstrap replicates conducted.

Recombination analysis

We utilized fastGEAR to infer genomic recombination based on core genome alignments, with a default clustering number of 15 [29]. FastGEAR employs the subspecies results from the phylogenetic tree to define the “optimal” number of clusters. It can detect both ancestral recombination affecting all isolates within a lineage and recent recombination affecting only a subset of isolates. By default, the larger lineage is assumed to be the donor.

Additionally, we investigated recombination across the whole Mab genome using Gubbins. Whole-genome sequences of individual strains were aligned to the reference genome (GCA_000069185) using MAFFT, creating a comprehensive alignment based on the reference genome [30]. Gubbins was then employed to identify recombination events using default parameters.

Identification of coincident resistance genes

We constructed a database of resistance genes for Mab based on previously reported resistance genes [31] (details in the supplementary materials). The nucleotide sequences were retrieved using BLAST v2.14.1 with default parameters (identity threshold of 90%) [32]. Multiple sequence alignments were generated using MAFFT with the -auto parameter. The alignments were parsed using R. For the erm41 gene, we examined each position (including gaps) for potential point mutations that might confer resistance to macrolides.

Epistatic interactions between gene gain/loss and point mutations in drug-resistant genes

We inferred the ancestral states (presence or absence) of each gene family at every internal node of the phylogenetic tree using PastML v1.9.33 [33]. The JOINT method with default parameters was employed to reconstruct the states using maximum likelihood. From the ancestral state matrix calculated by PastML, we inferred gene gain and loss events on all branches by subtracting the gene content of parent nodes from their descendants. We used Evo-Scope to analyze the types and strengths of interactions between these gained/lost genes and resistance mutations [34].

Statistical analysis

Unless mentioned otherwise, all the statistical analyses were performed using R (v4.1.0).

Results

Diversity of Mab clinical isolates

We performed whole-genome sequencing on 109 clinical Mab isolates, with an average genome length of 5.05 Mb and an average of 5,013 protein-coding genes. Among these isolates, 83 were identified as M.a. abscessus, and 26 were identified as M.a. massiliense, with no isolates of M.a. bolletii detected. The clinical data revealed that 65/83 (78.3%) of the M.a. abscessus isolates came from respiratory samples, whereas 18/83 (21.7%) came from extrapulmonary samples. In contrast, 24/26 (92.3%) of the M.a. massiliense isolates were from respiratory specimens, and 2/26 (7.7%) from extrapulmonary specimens. Comparative genomic analysis was conducted with an additional 28 publicly available Mab genomes (Table S1). After removing recombinant regions, a maximum likelihood phylogenetic tree was generated (Fig. 1A). All Mab isolates showed average nucleotide identity (ANI) values of at least 97%, consistent with the genomic species definition for microorganisms. Pairwise ANI within subspecies was greater than 98%, while between subspecies it was less than 98% (Fig. 1B). The MLST scheme identified 28 distinct STs among the M.a. abscessus isolates, with ST101 (n = 10/83, 12.0%), ST5 (n = 8/83, 9.6%), and ST107 (n = 8/83, 9.6%) being the predominant types (Fig. 1C). In M.a. massiliense, 7 STs were detected, with ST37 (n = 10/26, 38.5%) and ST46 (n = 4/26, 15.4%) being the most prevalent (Fig. 1C). Detailed ST distributions are presented in the Table S2.

Fig. 1.

Diversity of Mycobacterium abscessus. (A) Core genome-based maximum-likelihood phylogeny of the 139 isolates. (B) Pairwise genomic average nucleotide identity (ANI) within and between the Mycobacterium abscessus species. (C) Sequence type (ST) profiles of Mab subspecies

Recombination drives the diversity of the Mab population structure

Further evaluation of recombination events in the core and whole genomes of different Mab subspecies was conducted. The fastGEAR algorithm was employed to analyze the core genome alignment results. Ancestral recombination was infrequent in M.a. abscessus and M.a. massiliense, occurring primarily in M.a. bolletii, which involved an average of 1.21 kb (0.04%) of the core genome, contributing to its evolutionary origin (Figure S1A). Widespread recent recombination, occurring after species formation, has led to a highly mosaic genome in Mab (Figure S1B), with recombination fragments comprising 0–14% of the total genome across different strains (Fig. 2A). Most recombinant fragments presented high homology (≥ 99%) with the genomic regions of closely related subspecies, suggesting that these sequences originated from closely related subspecies. A few remaining fragments with relatively lower homology are thought to have resulted from recombination with other species closely related to Mab.

Fig. 2.

Genomic recombination and its contribution to the formation and diversification of Mab subspecies. (A) Recent recombination events in the core genome inferred by fastGEAR. Each line represents the genomic structure of a recently recombined isolate (depicted as colored bars), with colors indicating different species. (B) Whole-genome recombination events inferred from Gubbins showing the length distributions of the recombination fragments across the three subspecies. (C) Length distribution of gene sequences involved in recombination events among the three subspecies. Boxes indicate the median and interquartile range (IQR), with whiskers extending to a maximum of 1.5× IQR. (D) Genes associated with recombination fragments and their locations in the reference genome

The core genome alignment consists of concatenated sequence fragments from each strain, which may not fully represent the recombination characteristics, such as the genomic distribution and length of recombination fragments. Therefore, based on whole genome alignments of the three subspecies, Gubbins analysis was used to further explore recent recombination events. Recombination was uniformly distributed across the genomes of all three subspecies (Figure S2), with fragment lengths ranging from a few base pairs to a maximum of 176.1 kb, potentially indicating distributed conjugative transfer (DCT) recombination, a form of horizontal gene transfer in mycobacteria. The recombination fragments in M.a. massiliense were generally longer than those in M.a. abscessus and M.a. bolletii. Additionally, the number of sequences acquired through recombination in the M.a. massiliense genome exceeded that of the other two subspecies (Fig. 2B and C). We compared the lengths of recombinant genomic fragments among strains belonging to the five major STs. The results indicated that the extent of recombination is associated with ST classification and exhibits substantial heterogeneity within subspecies. In M.a. massiliense, ST37 harbored longer recombinant fragments compared to ST46. Similarly, within M.a. abscessus, ST107 showed longer recombination regions than ST5 and ST101 (Figure S3). We further identified the genes associated with the recombination fragments of M.a. massiliense and M.a. abscessus. The recombination fragments in M.a. abscessus included 488 genes, comprising gpl-related genes (MAB_4098c and MAB_4099c) and genes associated with the ESX-4 type VII secretion system (Fig. 2D). In contrast, the recombination fragments in M.a. massiliense involved 1,246 coding genes, including gpl-related genes (MAB_4098c) and genes linked to the MCE operon (MCE3 and MCE5) (Fig. 2D).

Pangenome analysis of Mab species and subspecies reveals extensive genetic diversity

We used Panaroo to calculate the pangenome at the species level for all Mab strains, identifying the core and accessory genomes. The Mab pangenome comprised 3993 core genes (≥ 99% genomes), 85 soft-core genes (< 99% to ≥ 95% genomes), 1160 shell genes (< 95% to ≥ 15% genomes), and 9083 cloud genes (< 15% genomes) (Fig. 3A). The M.a. abscessus contains 3,970 core genes, while M.a. massiliense has 4,099 core genes, with 3,886 core genes shared between the two subspecies (Fig. 3B). Gene distribution across strain genomes and sharing between strains were illustrated using a gene presence-absence matrix and pairwise Jaccard index (Fig. 3C and D). We further analyzed gene sharing among strains belonging to the five major STs. The results revealed substantial genetic heterogeneity within Mab, with gene sharing predominantly occurring among phylogenetically closely related strains, particularly those within the same ST (Figure S4).

Fig. 3.

Pangenome composition and distribution of 109 Mab strains. (A) A histogram depicting the distribution of core, soft-core, shell, and cloud genes within the genus pangenome, complemented by a pie chart showing the number of genes in each category, with percentages indicated in brackets. (B) A Venn diagram illustrates the distribution of 14,316 genus pangenome genes across subspecies, with total gene counts outside parentheses and proportions inside, highlighting that two subspecies share 3,886 core genes (genes detected in ≥ 99% of the genomes in the subset). (C) Presence–absence patterns of genes detected in the pangenome are shown, with sidebars indicating subspecies and the type of infection associated with each genome isolation. (D) The pairwise Jaccard index between genomes reflects the proportion of shared genes, where a value of 1.0 indicates complete overlap and lower values signify reduced sharing. Hierarchical clustering of the genomes and accompanying sidebars align with the information presented in Figure C

Throughout the evolutionary history of the lineage, highly variable accessory genes have been gained or lost at different rates. We aimed to determine the dynamics of gene acquisition and loss along the phylogenetic branches of different subspecies. Using core SNP phylogeny and gene presence/absence matrices, we inferred the gene gain and loss rates for M.a. abscessus and M.a. massiliense. The gene gain and loss rates were slightly greater in M.a. abscessus than in M.a. massiliense, though the difference was not statistically significant (p = 0.367) (Figure S5A and S5B). Additionally, gene gain and loss events were concentrated in more recent phylogenetic branches for both subspecies, indicating frequent and rapid gene turnover (Figure S5C and S5D).

A pangenome-wide association study was conducted to investigate genomic differences between M.a. abscessus and M.a. massiliense. A total of 294 genes were identified as unique to one of the two subspecies. Functional module and pathway enrichment analyses were performed on 54 of these genes (excluding unannotated genes) (Table 1). Unique genes in M.a. abscessus were associated with fatty acid biosynthesis, pantothenate and CoA biosynthesis, and quorum sensing, while M.a. massiliense had unique genes involved in fluorobenzoate degradation, toluene degradation, chloroalkane and chloroalkene degradation, and naphthalene degradation (Fig. 4). Coevolution analysis of the functionally annotated genes from each subspecies was conducted, constructing gene-gene coevolution networks for significant gene pairs with scores above 0.35 and clustering them accordingly (Figure S6A and S6B). The subspecies-specific accessory genes formed distinct clusters in the coevolutionary networks, indicating that these genes were acquired or lost in a coordinated manner.

Table 1.

Accessory genes specific to the Mab subspecies

K number	group	gene	function
K01911	abscessus	menE_2;menE_4;;menE_5	2-succinylbenzoate–CoA ligase; hypothetical protein
K00604	abscessus	fmt_1	Methionyl-tRNA formyltransferase
K00077	abscessus	panE	2-dehydropantoate 2-reductase
K18326	abscessus	mdtD_3;jefA	Putative multidrug resistance protein MdtD; Drug efflux pump JefA
K00648	abscessus	fabH1	3-oxoacyl-[acyl-carrier-protein] synthase 3 protein 1
K12440	abscessus	ppsA	Phthiocerol/phenolphthiocerol synthesis polyketide synthase type I PpsA
K07003	abscessus	mmpL5	Siderophore exporter MmpL5
K15657	abscessus	srfAD	hypothetical protein; Surfactin synthase thioesterase subunit
K01563	abscessus	dhaA_2;ephA_2	Haloalkane dehalogenase; Epoxide hydrolase A
K27108	abscessus	dhaA_2;ephA_2	Haloalkane dehalogenase; Epoxide hydrolase A
K14256	abscessus	oxyS_3	hypothetical protein;12-dehydrotetracycline 5-monooxygenase/anhydrotetracycline 6-monooxygenase
K00060	abscessus	adhA; tdh_3	putative alcohol dehydrogenase AdhA; L-threonine 3-dehydrogenase
K00359	abscessus	qorA_6	Quinone oxidoreductase 1
K07749	abscessus	frc_3;frc_1	hypothetical protein; Formyl-CoA: oxalate CoA-transferase
K02483	abscessus	tcrA	Transcriptional regulatory protein TcrA
K02484	abscessus	tcrY	putative sensor histidine kinase TcrY
K01897	abscessus	fadD3_1;menE_7	3-[(3aS4S7aS)-7a-methyl-15-dioxo-octahydro-1 H-inden-4-yl]propanoyl: CoA ligase;2-succinylbenzoate–CoA ligase
K20470	abscessus	mmpL1_1;mmpL5_1;mmpL1_3	putative transport protein MmpL1;Siderophore exporter MmpL5;Siderophore exporter MmpL4
K02069	abscessus	fetB	hypothetical protein; putative iron export permease protein FetB
K00156	abscessus	ydaP	Putative thiamine pyrophosphate-containing protein YdaP
K18687	abscessus	fadD3_1;menE_7	3-[(3aS4S7aS)-7a-methyl-15-dioxo-octahydro-1 H-inden-4-yl]propanoyl: CoAligase;2-succinylbenzoate–CoA ligase
K21271	abscessus	auaH	hypothetical protein; Aurachin B dehydrogenase
K01596	abscessus	pckG_2;;pckG_1	Phosphoenolpyruvate carboxykinase [GTP]; hypothetical protein
K22108	abscessus	kstR2_1	HTH-type transcriptional repressor KstR2
K02433	abscessus	QRSL1	Glutamyl-tRNA(Gln) amidotransferase subunit A chloroplastic/mitochondrial
K10536	abscessus	aguA_1	Agmatine deiminase
K00060	abscessus	tdh_3	L-threonine 3-dehydrogenase
K09681	abscessus	gltC_1;gltC_4	HTH-type transcriptional regulator GltC
K07047	abscessus	nfdA_2;ade	N-substituted formamide deformylase; Adenine deaminase; hypothetical protein
K01278	abscessus	dap4	Dipeptidyl aminopeptidase 4;hypothetical protein
K06359	abscessus	rapA	RNA polymerase-associated protein RapA
K03087	massiliense	rpoS	RNA polymerase sigma factor RpoS
K08162	massiliense	mdtH_1;mdtH_2	Multidrug resistance protein MdtH
K14956	massiliense	esxG_1	ESAT-6-like protein EsxG
K01673	massiliense	cynT_3;cynT_2;mtcA2_1	Carbonic anhydrase; Carbonic anhydrase 2
K14956	massiliense	esxH_3;;esxH_1	ESAT-6-like protein EsxH; hypothetical protein
K00951	massiliense	relA_2	hypothetical protein; Bifunctional (p)ppGpp synthase/hydrolase RelA
K03367	massiliense	dltA_3;dltA_4	hypothetical protein; D-alanine–D-alanyl carrier protein ligase
K27097	massiliense	espR_1;espR_4;espR_3	Nucleoid-associated protein EspR
K01669	massiliense	phrA	Deoxyribodipyrimidine photo-lyase
K00574	massiliense	ufaA1	hypothetical protein; Tuberculostearic acid methyltransferase UfaA1
K03088	massiliense	rskA	Anti-sigma-K factor RskA
K05770	massiliense	crtK-2_2	Tryptophan-rich protein TspO
K26597	massiliense	cmaA1_2;cmaA1_1	Cyclopropane mycolic acid synthase 1
K07695	massiliense	devR_1;devR_2	DNA-binding transcriptional activator DevR/DosR
K27105	massiliense	dosT_1;devS; devS_2;dosT_2	hypothetical protein; Oxygen sensor histidine kinase response regulator DosT; Oxygen sensor histidine kinase response regulator DevS/DosS
K01621	massiliense	xpkA_1;xfp_2;xpkA	Xylulose-5-phosphate phosphoketolase; Xylulose-5-phosphate/fructose-6-phosphate phosphoketolase
K22473	massiliense	adhA; tdh_3	putative alcohol dehydrogenase AdhA; L-threonine 3-dehydrogenase
K06720	massiliense	ectC_1;ectC_2	L-ectoine synthase
K07636	massiliense	phoR; sasA_4	Alkaline phosphatase synthesis sensor protein PhoR; Adaptive-response sensory-kinase SasA
K01061	massiliense	clcD	Carboxymethylenebutenolidase
K06127	massiliense	COQ5_2	2-methoxy-6-polyprenyl-14-benzoquinol methylase mitochondrial
K13953	massiliense	adhT	Alcohol dehydrogenase;2-haloacrylate reductase; hypothetical protein
K24967	massiliense	nagR_2;mngR; nagR_1	HTH-type transcriptional repressor NagR; Mannosyl-D-glycerate transport/metabolism system repressor MngR

Fig. 4.

Subspecies-associated genes and functional enrichment. The Scoary algorithm was used to assess the statistical associations of genes with each subspecies. Subspecies-specific genes were subjected to functional annotation and pathway enrichment analysis

In summary, distinct metabolic pathways and their associated genes were found to be the most relevant, suggesting that the coevolution of metabolic genes plays a crucial role in shaping the evolution of Mab subspecies and their adaptation to hosts.

Antimicrobial resistance and distribution of resistance genes

We further investigated the distribution of 19 previously reported resistance genes across 109 clinical strains with known antibiotic susceptibility profiles (Table S3), and all the clinical genomes contained the complete set of resistance genes. Antibiotic susceptibility testing for 15 antimicrobial agents was conducted according to guidelines (see Methods). The results indicated that the MIC distributions were similar between the M.a. abscessus and M.a. massiliense for nearly all antibiotics, except for the 14-day macrolide MIC (Fig. 5). A higher proportion of M.a. abscessus exhibited resistance compared to M.a. massiliense. Previous studies have shown that the 14-day CLA MIC is typically associated with the functionality of the erm41 gene. The M.a. massiliense typically carries a truncated erm41 gene, leading to enzyme inactivation, which does not affect the action of macrolides on the rrl gene. We further analyzed variations in the erm41 gene, including point mutations and deletions (Fig. 6). All M.a. massiliense strains had a truncated erm41 gene, while some M.a. abscessus strains exhibited a T28 mutation, forming clusters on the phylogenetic tree. The distribution of resistance mutations across the Mab population was uneven, with certain clades harboring more mutations than others (Fig. 6). Finally, we examined the relationship between resistance-associated gene mutations, antimicrobial resistance phenotypes, and ST classifications. Our analysis revealed that strains belonging to the same ST exhibited highly consistent mutation patterns in the erm41 gene. In M.a. massiliense, both ST46 and ST37 strains carried a truncated form of the erm41 gene, rendering them susceptible to macrolide antibiotics. In M.a. abscessus, ST107 strains consistently harbored the T28C mutation, which was associated with macrolide susceptibility and the absence of inducible resistance. In contrast, ST101 strains retained the wild-type erm41 gene (T28 allele) and predominantly displayed inducible macrolide resistance. Intriguingly, although ST5 strains also carried the wild-type T28 erm41 allele with no additional mutations, they did not show inducible resistance to macrolides. This observation suggests that ST5 strains may possess alternative molecular mechanisms that suppress inducible macrolide resistance (Fig. 6). Overall, the clustered distribution of resistance and the frequent occurrence of multiple erm41 mutations suggest that these mutations accumulate nonrandomly.

Fig. 5.

Distribution of results from phenotypic drug susceptibility testing (pDST) of Mab isolates

AMK, amikacin; CIP, ciprofloxacin; MXF, moxifloxacin; SXT, trimethoprim-sulfamethoxazole; LZD, linezolid; AXO, ceftriaxone; FEP, cefepime; FOX, cefoxitin; TOB, tobramycin; TGC, tigecycline; MIN, minocycline; DOX, doxycycline; AMC, amoxicillin-clavulanate; IMI, imipenem; CLA, clarithromycin

Fig. 6.

Distribution of deletion or mutation sites in the erm41 gene across the strains’ genomes. The left panel displays the phylogenetic tree of 109 Mab isolates along with their subspecies designations and ST information. The middle heatmap illustrates the distribution of deletion or mutation sites for the erm41 gene. On the right, the results from phenotypic drug susceptibility testing (pDST) for clarithromycin on days 5 and 14 are shown, with vertical lines in the pDST panel representing the breakpoints defined by the CLSI

Epistatic effects of gene gain/loss and antibiotic resistance point mutations

Frequent and extensive recombination, along with accessory genome diversity in Mab, has established distinct genetic backgrounds that influence the evolution of antibiotic resistance. We sought to identify genes that were frequently gained or lost before the emergence of mutations associated with macrolide resistance. First, we reconstructed the ancestral states of each gene family and resistance mutations in the pangenome (see Methods) to infer the timing of gene family gain/loss and resistance mutation acquisition. Using Evo-Scope to compare the chronology of pangenome gene family gains and losses with the acquisition of antibiotic resistance, we found that 8 gene losses and 16 gene gains (excluding unannotated genes) were closely associated with the emergence of the T28 mutation, facilitating the acquisition of resistance (Table 2). These genes are primarily involved in biofilm formation, stress response, virulence, biotin synthesis, and fatty acid metabolism. MspB is a homolog of the channel-forming protein MspA. The loss of the MspB gene may reduce the cell wall permeability to hydrophilic antibiotics and decrease glucose uptake [35]. The low permeability of the Mycobacterium cell wall to nutrients contributes to its slow growth, creating conditions conducive to the development of multidrug resistance. Tam, an O-methyltransferase, transfers a methyl group from SAM to the O-carboxyl group of malonyl-ACP, producing malonyl-ACP methyl ester, a dedicated primer for biotin biosynthesis [36]. This enzyme plays a crucial role in biotin synthesis in mycobacteria, and biotin is essential for mycobacterial growth and the establishment of chronic infections.

Table 2.

Genes consistently gained or lost prior to acquiring macrolide antibiotic resistance

name_e1	name_e2	Pvalue	lambda	state
leuO	T28	0.00061	499.95	gain
rhaS_1	T28	0.00617	32.47	gain
car_2…car_1…car_3	T28	0.00633	183.43	loss
gltA2	T28	0.00523	254.47	gain
tetR_1…tetR_3…tetR_2	T28	0.00523	254.47	gain
rutD_4…rutD_2	T28	0.00523	254.47	gain
mdtH	T28	0.00523	254.47	gain
caiD_3…dpgD_2	T28	0.00523	254.47	gain
nfsB	T28	0.00523	254.47	gain
rclC	T28	0.00523	254.47	gain
bphC2	T28	0.04508	Inf	loss
mmpL4_15	T28	0.00523	254.47	gain
esxE_1	T28	0.00523	254.47	gain
whiB_5…whiB_3	T28	0.00523	254.47	gain
xerC_5…xerC_1…xerC_3	T28	0.0001	58.69	gain
thcD	T28	0.00523	254.47	loss
mspB_2	T28	0.00523	254.47	loss
recD2	T28	0.00523	254.47	gain
Dml	T28	0.00523	254.47	loss
xerC_7…xerC_3…xerC_5	T28	0.00523	254.47	gain
ybaQ	T28	0.00523	254.47	loss
iolS…yajO	T28	0.00163	488.79	loss
pcaF	T28	2.48E-05	127.02	loss
tam_4	T28	0.00094	131.38	gain

Discussion

In this study, we conducted a population genomic analysis of Mab strains from China to investigate the relationships among population diversity, host adaptation, evolution, and antibiotic resistance mutations. We explored the genomic recombination patterns among different Mab subspecies and their associations with strain prevalence. Ancestral core genome recombination primarily occurred in M.a.bolletii, with segments from M.a.abscessus and M.a.massiliense contributing to its formation, which may explain the lower global prevalence of M.a.bolletii. Whole-genome recombination analysis revealed that, compared with M.a. abscessus, M.a. massiliense has longer recombination segments, a higher recombination frequency, and more functional genes. Notably, regions of extensive recombination in Mab include the gpl biosynthesis gene cluster (mps1 and mps2), which may explain the observed R and S phenotypic diversity among Mab strains and the S-to-R transitions within hosts [37, 38]. These morphotypes have distinct impacts on bacterial adhesion and host interactions: the smooth morphotype is considered non-invasive, while the rough morphotype is virulent and associated with disease progression [39]. In addition, recombination in M. abscessus involves the EsxU and EsxT genes from the ESX-4 system, a key secretion system in Mycobacterium that aids bacterial survival and mediates virulence in host cells. The EsxU/EsxT heterodimer in M. abscessus is involved in phagosomal membrane damage in macrophages, making EsxU/EsxT crucial for inducing membrane permeability, which is beneficial during the early stages of infection [40]. Widespread recombination events among Mab subspecies, including large segment transfers likely driven by DCT, provide strong evidence that recombination, particularly DCT, is a key mechanism in subspecies formation and phenotypic variation.

Gene gain or loss events are commonly associated with pathogenic evolution and host adaptation in species [41]. Pangenome analysis revealed that the Mab accessory genome is large, diverse, and subspecies-specific. Within a subspecies, strains of the same sequence type (ST) share a higher proportion of genes. Pangenome-wide association studies indicate that accessory genes specific to M.a. abscessus include those involved in fatty acid biosynthesis, pantothenate and CoA biosynthesis, and quorum sensing. Pathogenic mycobacteria require host fatty acids (FA) and cholesterol for energy production, building their unique lipid-rich cell walls, and producing lipid virulence factors. Pantothenate (vitamin B5) is a crucial precursor of coenzyme A (CoA) and acyl carrier proteins and is essential for various intracellular processes, such as fatty acid metabolism and cellular signaling [42]. In contrast, accessory genes specific to M.a. massiliense strains are involved in fluorobenzoate degradation, toluene degradation, chloroalkane and chloroalkene degradation, and naphthalene degradation, indicating a stronger ability to withstand diverse environmental stresses potentially contributing to its role for outbreak infections. The highly variable accessory genome drives the evolution of the M.a. abscessus and M.a. massiliense, resulting in distinct metabolic patterns and varying capacities to adapt to oxidative stress, leading to different epidemiological profiles and characteristics.

Given the inherent and acquired antibiotic resistance of Mab and its poor clinical outcomes, understanding the development of antibiotic resistance in strains within the host is crucial. We obtained antibiotic susceptibility profiles for 109 clinical isolates, which revealed significant differences between subspecies only in the MIC distribution of CLA on day 14. CLA is a key antibiotic in combination therapy against Mab. Certain subspecies and specific sequence types (STs) can serve as molecular markers for corresponding drug resistance mutation patterns and antimicrobial susceptibility profiles. This association facilitates rapid clinical prediction of resistance through molecular diagnostics. Extensive genome-wide recombination and accessory genome diversity have established distinct genetic backgrounds among Mab subspecies, with epistatic interactions possibly contributing to why some lineages more readily develop antibiotic resistance. Specifically, we propose that gene acquisition/loss events are closely associated with the development of resistance mutations.

These results suggest that many functions frequently acquired before macrolide resistance mutations may enhance bacterial tolerance or persistence within the host. For example, the Tam gene, which is essential for biotin synthesis, supports fatty acid remodeling, increases cell membrane fluidity, and promotes Mab infection [36]. Transporter-related genes may enhance the nutrient uptake capacity and growth rate of strains [35]. Metabolic adaptations acquired through gene gain or loss contribute to bacterial survival and the further development of multidrug resistance.

Overall, Mab is difficult to eradicate from patients, with multidrug resistance being one of the key factors contributing to this challenge. The high adaptability of the Mab genome provides more strategies for resisting antibiotics and host immune clearance.

Our findings reveal significant differences in the structure and evolution of the accessory genome among the major subspecies of Mab and their associated STs. This genomic variation within the species may have important implications for disease epidemiology, infection pathogenesis, and interactions with hosts. Our results provide valuable insights into the potential genetic basis for the success of Mab as a highly adaptable and drug-resistant pathogen, which will inform current efforts to control and treat Mab infections, including strategies targeting specific Mab clones or lineages.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

YC, BJH, and QM designed the study; RB, NL, TTF, XYY, and LQ acquired data; YC conducted data analysis; YC, BJH, and QM written the manuscript.

Funding

This study was funded by Zhongshan Hospital, Fudan University (2024ZSFZ39), Shanghai Hospital Development Center Foundation (SHDC22024315) and Natural Science Foundation of Fujian Province (2024J08352). The funders had no role in the design of the study, collection and analysis of data, decision to publish or preparation of the manuscript.

Data availability

Strains analysed during the current study are available from the corresponding author on reasonable request. Genomic sequencing data from public databases are available at the National Center for Biotechnology Information using the accession numbers given in Supplementary Table S1

Declarations

Ethics approval and consent to participate

The samples used in the study were passaged bacterial isolates, not patient specimens. Information regarding the infection sites and sources of the isolates was approved by the Ethics Committee of Zhongshan Hospital, Fudan University (Approval No.: B2022-513R).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.