Morphological plasticity facilitates lung adaptation and dynamic immune remodeling in Mycobacterium abscessus infection

Summary

Mycobacterium abscessus pulmonary disease (Mab-PD) shows rising global incidence and low cure rates. Mab exhibits marked phenotypic heterogeneity, with infection outcomes likely dependent on strain-specific host interactions. Through in vivo morphotype switching, we obtained a rough strain (Mab-R) that showed higher virulence, bacterial loads, and lung pathology than the smooth morphotype. Transcriptomic analysis revealed that Mab-R adapts to macrophage functional states by modulating virulence and metabolic genes, promoting intracellular survival. Single-cell RNA sequencing uncovered dynamic macrophage reprogramming during infection. In acute infection, alveolar macrophages and C1qa+ interstitial macrophages cleared Mab-R via programmed cell death, while newly recruited monocytes showed strong phagocytic capacity but incomplete bacterial clearance. During chronic infection, enhanced T cell immunity and resident-like macrophage dominance facilitated inflammation resolution. These findings reveal host-pathogen interplay driving Mab-R persistence and identify bacterial metabolic adaptation and macrophage reprogramming as potential therapeutic targets for chronic Mab-PD.

Graphical abstract

Highlights

• •In vivo morphotype switching generated rough M. abscessus with increased virulence
• •Mab-R promotes intracellular persistence through virulence and metabolic adaptation
• •Single-cell transcriptomics reveals macrophage reprogramming during Mab infection
• •Bacterial metabolism and host macrophage remodeling are therapeutic targets in chronic Mab-PD

Immunology; Microbiology; Transcriptomics

Introduction

Mycobacterium abscessus (Mab) frequently causes pulmonary infections in patients with chronic lung diseases, with a rising incidence globally.^1,2 Mab is notoriously difficult to treat due to its intrinsic multidrug resistance and biofilm-forming ability, leading to persistently low cure rates.¹

Genome sequencing has revealed considerable plasticity in Mab, with evidence of horizontal gene transfer and genetic mutations arising during infection that may enhance virulence and adaptability.^2,3 Longitudinal studies in cystic fibrosis patients have shown that the Mab can accumulate adaptive mutations during chronic infection that favor persistent colonization.⁴ Notably, Mab is capable of transitioning between two colony morphotypes—smooth (S) and rough (R)—which exhibit markedly different interactions with host cells.⁵ The S morphotype synthesizes glycopeptidolipids (GPLs) that form a glycolipid-rich outer layer, promoting adhesion and colonization in the airway.⁶ In contrast, the R morphotype lacks GPL expression, enabling the formation of characteristic serpentine cords,^6,7 which enhance resistance to phagocytosis, trigger stronger immune responses, and are closely associated with more severe tissue damage in mouse and zebrafish infection models.⁵ However, it is not fully elucidated how genomic adaptations acquired during chronic infection influence bacterial gene expression and function, or how this pathogen interacts with host immune cells, especially in immunocompetent individuals.⁴

The clinical course of Mab pulmonary disease is largely determined by the host-pathogen interplay, particularly the host’s immune response in the lungs. Following inhalation, Mab is primarily phagocytosed by alveolar macrophages (AMs), which serve as the first line of immune defense in the pulmonary environment. Studies have shown that depletion of these cells results in uncontrolled bacterial proliferation within the lungs.⁸ Differential activation of NF-κB and HIF-1α between alveolar and monocyte-derived macrophages contributes to distinct inflammatory responses following infection.⁸ In parallel, Mab can escape from the phagosome into the cytosol by disrupting the phagosomal membrane, thereby activating immune pathways that promote IL-1β and type I interferon (IFN-I) production, which in turn facilitates intracellular survival and cell-to-cell dissemination.⁹

The outcome of this initial immune engagement is critical in determining whether the infection is cleared or progresses. If not promptly controlled, pattern recognition receptors (PRRs), such as Toll-like receptors, on macrophages and other immune cells detect mycobacterial components and initiate proinflammatory signaling cascades.⁵ As the infection transitions from an acute to a chronic phase, both the composition and activation state of pulmonary immune cells undergo dynamic changes. During the early stage, inflammatory “M1-like” macrophages, predominantly derived from Ly6C⁺ monocytes recruited to the lungs, are activated by IFN-γ and TNF-α to produce high levels of proinflammatory mediators. These macrophages, together with neutrophils, exert antimicrobial effects and contribute to granuloma formation.⁵ In the later stages of infection, inflammatory cells decline, and alternatively activated, tissue-repairing macrophage subsets increase, mediating inflammation resolution and tissue remodeling. This shift reflects a functional transition from proinflammatory to homeostasis-maintaining macrophage responses.¹⁰

During chronic infection, the host immune response undergoes complex and coordinated dynamic changes. Therefore, macrophage function should be systematically evaluated based on their functional states and temporal transitions. In this study, we employed an immunocompetent mouse model of chronic Mab pulmonary infection to observe the bacterium’s adaptation and the dynamic trajectory of the host immune response. By combining longitudinal bacterial analyses (including colony morphotype, genomics, and transcriptomics of Mab during infection) with high-dimensional immune cell profiling (using single-cell RNA sequencing at early and late stages of infection), we provide an integrated view of host-pathogen co-evolution in chronic infection. In the early stage of infection, AMs rapidly mediate Mab killing through regulation of cell death pathways, while monocyte-derived macrophages and neutrophils cooperate to restrict bacterial spread. Subsequently, T cell-mediated antimicrobial responses are enhanced, and resident-like macrophages become predominant, playing key roles in limiting excessive inflammation and promoting tissue repair. This study also reveals how Mab-R enhances intracellular survival by modulating its gene expression and employing multiple strategies to adapt to macrophage metabolic reprogramming, offering new pathogen-centered insights into persistent clinical infections.

Results

In vivo selection of a rough morphotype with a GPL gene mutation

To model chronic Mab lung infection, immunocompetent mice were infected via an agar-bead method with the smooth colony morphotype reference strain ATCC 19977 (Mab-S).¹¹ This approach prevented rapid bacterial clearance and enabled persistent infection. Six weeks post-infection, bacterial colonies isolated from serially collected lung homogenates of infected mice exhibited a rough morphotype (Figure S1A), whereas the original inoculum was uniformly smooth. The emergence of rough colonies in vivo suggested that adaptive genetic changes occurred during infection. Whole-genome sequencing of the rough variant (Mab-R) revealed a limited number of genomic variations (including single nucleotide polymorphisms (SNPs), insertions, and deletions) at a frequency of 0.024%, with only a minor fraction representing nonsynonymous mutations (0.001%). Notably, within the GPL gene locus, a single nonsynonymous mutation (Asp145Asn) was exclusively identified in MAB_4105c (Figure S1B). Structural modeling predicted that this mutation would likely reduce enzyme stability and ligand-binding affinity, thereby impairing its function in GPL synthesis (Figure S1C). MAB_4105c is believed to mediate O-methylation of GPL fatty acyl chains. Thus, loss of its function would eliminate this modification and diminish GPL production. Consistent with this, mutations in GPL biosynthetic genes are known to trigger a switch from smooth to rough colony morphology.¹²

Global transcriptomic reprogramming of the adapted strain reveals metabolic and virulence shifts

To investigate how the S-to-R phenotypic transition affects Mab physiology, we performed RNA-seq on log-phase cultures of Mab-R and Mab-S under identical conditions. RNA-seq was performed on log-phase cultures, and differentially expressed genes (DEGs) were identified using a threshold of |log₂FC|>1, FDR <0.05. A total of 1,047 DEGs were detected, with 521 genes upregulated and 526 downregulated in Mab-R relative to Mab-S. These DEGs were distributed across the chromosome without obvious clustering (Figure 1A), indicating a genome-wide transcriptional overhaul in the rough variant. This extensive transcriptional reprogramming likely reflects global effects caused by GPL locus dysfunction, involving cell wall structural alterations and impaired methylation activity, analogous to DpnM-mediated changes.⁴

Figure 1.

Transcriptomic differences between the rough variant (Mab-R) and the parental smooth strain (Mab-S) in vitro (n = 5)

(A) Circular genome map depicting genome-wide distribution of differentially expressed genes (DEGs, |log₂Foldchange| > 1, FDR <0.05) in Mab-R versus Mab-S. Red bars indicate upregulated genes; blue bars, downregulated genes. Genome coordinates (Mb) are marked.

(B) KEGG pathway enrichment of DEGs. Bar length (–log₁₀p value) represents enrichment significance; numeric labels indicate DEG counts per pathway. The pink bars represent upregulation in Mab-R, while the blue bars represent downregulation in Mab-R.

(C) Heatmap of selected lipid metabolism and cell envelope DEGs.

(D) Heatmap of virulence, iron acquisition, and antibiotic resistance DEGs.

KEGG pathway enrichment analysis revealed that upregulated DEGs in Mab-R were significantly enriched in fatty acid degradation, diverse amino acid metabolism (including both biosynthesis and catabolism), siderophore biosynthesis, alternative carbon metabolism (e.g., glyoxylate cycle), and xenobiotic degradation pathways. In contrast, downregulated genes were primarily concentrated in core energy metabolism pathways (TCA cycle and oxidative phosphorylation), sulfur metabolism, and short-chain fatty acid metabolism (Figure 1B). This metabolic reprogramming suggests a shift from the high-efficiency energy production mode dependent on TCA cycle and oxidative phosphorylation to an adaptive metabolic strategy focused on acquiring exogenous nutrients (e.g., fatty acids and amino acids), while enhancing xenobiotic degradation capacity and iron acquisition systems to cope with environmental stresses in the host. Transcriptional changes in Mab-R were not limited to metabolism; key virulence-associated gene clusters were differentially expressed as well. A striking finding was the upregulation of multiple mammalian cell entry (MCE) operons in Mab-R (Figure 1C). Four distinct MCE loci (mce1, mce3, mce4, and mce6 operons) showed significantly higher transcript levels in Mab-R compared to Mab-S. MCE proteins form part of lipid transporter complexes in the mycobacterial cell envelope, enabling the uptake of fatty acids and cholesterol from the host.¹³

In addition to metabolic rewiring, the Mab-R transcriptome revealed enhancements in virulence and stress resistance pathways (Figure 1D). Specifically, two type VII secretion systems, ESX-3 and ESX-4, were markedly upregulated in Mab-R. ESX-3 is a conserved secretion system mediating iron acquisition through mycobactin, and its enhanced expression likely improves iron uptake efficiency in iron-limited host environments.¹⁴ ESX-4, though less characterized, has been linked to Mab virulence.¹⁵ Its upregulation may enhance bacterial secretion of effector proteins, inhibiting phagosome maturation, evading immune killing, and facilitating intercellular dissemination within host tissues.^15,16 Furthermore, Mab-R showed increased expression of multiple antibiotic resistance genes (Figure 1D). Compared with Mab-S, Mab-R exhibited increased expression of resistance genes to multiple major drug classes, including aminoglycosides (e.g., amikacin), macrolides, β-lactams, and rifamycins, with the exception of downregulated expression of the bedaquiline resistance gene. In summary, under nutrient-rich in vitro conditions, Mab-R exhibits enhanced metabolic regulation and shows higher activation of virulence factors and antibiotic resistance genes.

Genomic diversity of GPL locus genes in clinical isolates

We sequenced 109 clinical isolates of Mab to assess genetic variation within the glycopeptidolipid (GPL) biosynthesis locus and its association with subspecies and dominant circulating clones (DCC) (Figure 2). Among these, 83 isolates belonged to M. a. abscessus and 26 to M. a. massiliense, with no M. a. bolletii identified. Within M. a. massiliense, 14 isolates were classified as DCC3, 7 as DCC6, 1 as DCC7, and 4 as non-DCC. Among M. a. abscessus isolates, 11 were DCC1, 8 DCC2, 11 DCC4, 20 DCC5, 31 Abs1, and 2 non-DCC.

Figure 2.

Genomic diversity of the GPL biosynthesis locus in 109 clinical Mab isolates

Each row represents a clinical isolate, categorized by subspecies (abscessus or massiliense) and dominant circulating clone (DCC or non-DCC). Columns correspond to 21 genes spanning the GPL locus (MAB_4097c-MAB_4117c). Heatmap intensity denotes the number of non-synonymous single-nucleotide polymorphisms (nsSNPs) per gene per isolate (0–32). All isolates retain the full GPL locus. The O-methyltransferase gene MAB_4105c shows the highest mutation frequency across isolates. Clustering patterns reflect associations between nsSNP burdens, subspecies identity, and DCC classification.

The GPL gene cluster (MAB_4097c-MAB_4117c, 21 genes) encodes enzymes and transporters involved in GPL biosynthesis and modification, which determine colony morphotypes: smooth (GPL-expressing), rough (GPL-deficient), or mixed.¹⁷ All isolates retained a complete GPL locus, indicating its conservation in clinical strains and highlighting the functional importance of non-synonymous mutations for bacterial survival and cell envelope phenotypes.

We performed SNP analysis across GPL genes, quantifying non-synonymous mutations per gene per isolate and visualizing their distribution as a heatmap. GPL gene variation was widespread, with most strains harboring at least one non-synonymous SNP across multiple genes. Mutation patterns correlated with subspecies and DCC, with M. a. massiliense exhibiting a higher burden of non-synonymous mutations than M. a. abscessus (Figure 2). In M. a. abscessus, mutations were predominantly found in GPL-modifying enzymes rather than core biosynthetic genes such as MAB_4098c and MAB_4099c, and the majority of DCC1 isolates (8/11, 72.7%) showed no mutations.

Among the 109 isolates, 89 (82%) carried mutations in the MAB_4105c gene. As a GPL-modifying gene previously observed to mutate in our mouse infection model, MAB_4105c is also highly representative in clinical isolates.

We further analyzed the relationship between antibiotic resistance and mutations in the GPL locus, as well as the association between GPL mutations and sputum culture conversion in patients (Figures S2A and SB). The results showed that strains harboring more GPL mutations appeared to have lower rates of clarithromycin resistance. However, we consider that this observed difference in drug resistance is more likely attributable to biological differences between subspecies rather than being directly driven by GPL mutations themselves. In addition, among 64 patients with at least 1 year of microbiological follow-up, we compared GPL mutation burden between those who achieved sputum culture conversion within 1 year and those who remained persistently culture-positive, and found no statistically significant difference between the two groups. These findings suggest that although GPL locus mutations are widespread among clinical isolates, their direct association with microbiological outcomes remains limited.

Mab-R exhibits enhanced virulence in mice compared to Mab-S

To directly compare the virulence of Mab-S and Mab-R variants in vivo, we established a high-dose intratracheal infection model (Figure 3A). Mice were randomly assigned to receive ∼1 × 10⁷ colony-forming units (CFU) of Mab-S or Mab-R (or sterile PBS as control) to minimize in vivo morphotype switching. Acute-phase assessments were performed at day 7 and day 14 post-infection. At each time point, a subset of mice (n = 5 per group) was euthanized to determine lung bacterial burden and for additional analyses, while the remaining animals were monitored for body weight and survival over the following 3 weeks. Body weight monitoring showed that Mab infection initially caused weight loss in both infected groups, with Mab-R-infected mice exhibiting a more pronounced and prolonged weight reduction (Figure 3B). For the chronic-phase assessment, the remaining mice were euthanized at week 6 to measure lung bacterial burden, and lung tissues were collected from the Mab-R group and the control group for single-cell transcriptomic analysis.

Figure 3.

Virulence comparison of Mab-S and Mab-R in a murine lung infection model

(A) Experimental timeline: mice were infected intratracheally with 1 × 10^7 CFU of Mab-S, Mab-R, or PBS (control). Lung samples were harvested at days 7 and 14; survival and body weight were monitored up to day 21.

(B) Body weight changes post-infection. Mab-R-infected mice showed greater initial weight loss and delayed recovery compared to Mab-S and controls (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, two-way repeated-measures ANOVA).

(C) Pulmonary bacterial loads at weeks 1, 2, and 6 post-infection. The initial inoculum (day 0) represents the administered bacterial dose. At the chronic stage (week 6), bacterial burden was substantially reduced, and colony counts inevitably fell below the recommended 30–300 range for reliable quantification; however, this had limited impact on the statistical outcomes. Mab-R-infected lungs consistently exhibited significantly higher bacterial loads than Mab-S-infected lungs at all time points (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, two-way repeated-measures ANOVA). Control lungs remained sterile. Bars represent medians. CFU per gram of lung tissue was calculated based on the weight ratio of the processed lung segment to the total lung weight.

(D) Kaplan-Meier survival curves. Mab-R–infected mice had significantly lower survival compared to Mab-S-infected mice (log rank test, p = 0.00056). Controls showed 100% survival. (n = 30/group).

(E) Lung histopathology (H&E) at day 7 post-infection. Control lungs were normal. Scale bars: 2 mm (low magnification), 100 μm (high magnification).

Consistent with previous studies, pulmonary bacterial burden in Mab-R-infected mice was markedly higher than in Mab-S-infected mice (Figure 3C). At 7 days post-infection, the median lung CFU in the Mab-R group was approximately 1–2 log_10 greater than in the Mab-S group. By 14 days, lung CFU had declined in both groups, indicating partial immune clearance; nevertheless, Mab-R-infected lungs still harbored significantly more bacteria than Mab-S lungs (two-way repeated-measures ANOVA, p < 0.05). These data demonstrate that the rough variant establishes a higher and more persistent bacterial load in the host, correlating with a more significant impact on the host’s body weight and overall health status. Correspondingly, outcome monitoring over 3 weeks showed striking differences in survival between the groups (Figure 3D). Kaplan-Meier survival analysis confirmed that Mab-R infection led to significantly worse survival outcomes than Mab-S infection (log rank test, p = 0.045). This heightened lethality of the rough morphotype is in agreement with previous reports that rough variants of Mab are hypervirulent and can cause accelerated mortality in mice.⁷ Histopathological analysis at 7 days post-infection revealed that the rough variant Mab-R induced severe neutrophil-driven acute pneumonia with focal abscess formation and marked tissue destruction. In contrast, the smooth variant Mab-S elicited milder inflammation characterized predominantly by lymphocytic infiltration, preserving overall lung architecture, indicating substantial differences in pathogenicity between these two morphotypes (Figure 3E).

In summary, Mab-R caused more severe murine infection, manifested by reduced survival rates, higher bacterial loads, decreased body weight, and severe acute pulmonary pathology.

Global transcriptional changes in Mab-R during mouse lung infection

To delineate the in vivo adaptive mechanisms of Mab-R, we conducted bacterial RNA-seq on infected murine lung tissues (high-dose infection, 1 week post-infection). Differential expression analysis (adjusted p < 0.05, |log₂ fold-change| > 1) identified 245 genes with significant expression changes in vivo. Strikingly, 225 of these genes were upregulated in Mab-R relative to the smooth variant (Mab-S), whereas only 20 were downregulated (Figure 4A). Functional enrichment analysis revealed that downregulated genes were predominantly associated with the protein homeostasis system (Figure 4B). The decreased expression of chaperones and folding catalysts suggests that Mab-R may reduce quality control of nascent proteins and shift toward alternative stress response mechanisms. Concurrently, multiple carbon metabolism pathways were activated, including glyoxylate and dicarboxylate metabolism, the TCA cycle, fatty acid degradation, butanoate metabolism, and ketone body synthesis and degradation (Figure 4B). In vivo, Mab-R also exhibited upregulation of amino acid metabolic pathways, including alanine, aspartate, and glutamate metabolism, along with increased expression of bacterial secretion systems (Figure 4B). Notably, the activation of carbon fixation and porphyrin metabolism suggests a potential adaptive response to host-specific microenvironments, such as elevated CO₂ levels and iron limitation at sites of inflammation. We further assessed the in vivo expression of Mce operons, ESX-3 and ESX-4 type VII secretion systems, and antibiotic resistance genes in Mab-R compared to Mab-S. Most genes remained upregulated in Mab-R, though the magnitude of induction was lower than that observed under in vitro conditions (Figure 4C).

Figure 4.

In vivo transcriptomic differences between Mab-R and Mab-S in the mouse lung

(A) Total number of differentially expressed genes (DEGs, |log₂FC| > 1, FDR <0.05) identified between Mab-R and Mab-S at 1 week post-infection.

(B) Functional enrichment analysis of Mab-R upregulated genes. Bars represent enrichment significance (–log₁₀(p value)) for selected metabolic and stress-response pathways. The value associated with each bar represents the number of differentially expressed genes enriched in the corresponding pathway.

(C) Heatmaps illustrating selected DEGs associated with cell envelope remodeling (lipid transport), virulence (ESX-3, ESX-4 secretion systems), and drug resistance/mycobactin biosynthesis.

In vivo vs. in vitro differential gene expression and pathway enrichment

To quantify the transcriptional plasticity of Mab genes under distinct in vivo and in vitro conditions, we calculated the difference in gene expression fold-changes (Δlog₂FC) by subtracting the log₂ fold-change (Mab-R/Mab-S) obtained in vitro from that obtained in vivo (Δlog₂FC = [log₂FC in vivo] − [log₂FC in vitro]). The histogram of Δlog₂FC values showed a pronounced right-skewed distribution (Figure 5A), indicating that most genes in Mab-R underwent modest transcriptional changes in vivo compared to in vitro (|Δlog₂FC| < 2), while only a subset exhibited substantial regulation, reflecting targeted transcriptional reprogramming in response to the host environment. Using a threshold of |Δlog₂FC| > 2 (>4-fold difference between environments), we identified a set of genes with markedly divergent regulation amplitudes in vivo versus in vitro.

Figure 5.

In vivo vs. in vitro transcriptional differences and pathway enrichment for Mab-R

(A) Histogram of Δlog₂ fold-change values comparing Mab-R gene expression in vivo (mouse lung) versus in vitro. Δlog₂FC was calculated as (Mab-R vs. Mab-S in vivo)—(Mab-R vs. Mab-S in vitro). Red dashed lines indicate ±2 cutoff. The distribution is skewed toward positive values, highlighting many genes with stronger upregulation in vivo.

(B) KEGG pathway enrichment analysis for genes with |Δlog₂FC| ≥ 2 (adjusted p < 0.05).

Genes meeting this Δlog₂FC criterion were subjected to KEGG pathway enrichment analysis to characterize their functional associations. Figure 5B summarizes the significantly enriched pathways (adjusted p < 0.05) for these high-Δ genes. Genes associated with multiple metabolic pathways exhibited greater transcriptional plasticity, notably amino acid metabolism (e.g., aromatic and branched-chain amino acids), central carbon metabolism, fatty acid degradation, and cofactor metabolism. Pathways related to cell envelope synthesis (e.g., peptidoglycan biosynthesis) and antibiotic resistance were also significantly enriched. Additionally, pathways involved in protein translation and nucleotide synthesis (including ribosomal proteins, tRNA biogenesis, and purine metabolism) were prominently represented. These pathways represent functional modules with markedly different expression patterns between in vivo and in vitro conditions, indicating that Mab-R undergoes extensive remodeling of its metabolic network, nutrient acquisition strategies, and cell envelope maintenance to adapt to the host environment.¹⁸

Single-cell characterization reveals dynamic remodeling of immune and structural cell populations during Mab-R lung infection

We performed single-cell RNA sequencing of lung tissues from control and Mab-R-infected mice (at weeks 1 and 6 post-infection), identifying all major immune and structural cell types (Figure 6A). Dimensionality reduction, clustering, and annotation based on canonical marker genes (Table S1) enabled clear delineation of resident and infiltrating lung cell populations, generating a comprehensive atlas of lung cellular composition (Figures 6A and 6C). Cellular composition changes were evident between control and infected lungs over time (Figure 6B). Following Mab-R infection, substantial changes occurred in lung cellular composition (Figure 6B). At one week post-infection, innate immune cells significantly expanded: neutrophils increased dramatically from ∼6.4% in controls to 41.5%, and monocyte/macrophages rose from 14.1% to 25.9%, indicative of a robust acute inflammatory response.⁵ T cell proportions also increased from 11.2% to 19.5%. Conversely, B cells and NK cells were markedly reduced, with dendritic cells showing a slight decrease. By 6 weeks, the immune cell profile shifted, with T cells becoming the dominant lymphocyte subset (28.7%) and neutrophils returning to baseline levels (9.4%). Monocyte/macrophage populations remained mildly elevated compared to controls, whereas B cells and NK cells persisted at low frequencies, and dendritic cell numbers recovered to control levels. Overall, Mab-R infection prompted rapid early recruitment of neutrophils and monocytes, followed by sustained T cell accumulation in chronic infection, while B cells and NK cells constituted a minor component throughout.

Figure 6.

Single-cell RNA-seq analysis of lung immune cells in control vs. Mab-infected mice

(A) UMAP visualization identifying 13 distinct cell clusters from integrated lung datasets (all groups): T cells, naive T cells, B cells, NK cells, neutrophils, and dendritic cells (DCs), monocyte/macrophages, epithelial cells, endothelial cells, lymphatic endothelial cells, smooth muscle cells, fibroblasts, and doublets.

(B) Stacked bar chart showing proportional composition of cell types across conditions (control vs. Mab-R infection at weeks 1 and 6).

(C) Heatmap of representative marker gene expression (rows) for identified cell clusters (columns), color indicating relative expression (high, red; low, blue). Key markers for each major cell type are shown.

(D) t-SNE plot illustrating subclustering of monocyte/macrophage populations into seven subsets: alveolar macrophages (AMs), proliferative AMs (AMs_Mki67), Marco_Ace resident macrophages, C1qa+ interstitial macrophages (C1qa⁺IMs), recruited monocyte-derived macrophages (Ccr2⁺recMacs, Ly6c2⁺recMacs), and others.

(E) Proportional changes of macrophage subsets under infection versus control conditions.

(F) Heatmap displaying selected marker genes distinguishing macrophage subsets defined in (D). Columns indicate subsets; rows denote specific marker genes.

Detailed analysis of lung mononuclear phagocytes revealed dynamic macrophage subset changes during Mab-R infection (Figure 6D). Guided by the established murine lung macrophage atlas,¹⁹ we identified multiple transcriptionally defined subsets (Table S2), including resident AMs, proliferative AMs (AMs_Mki67), macrophages with high ACE expression (MARCO_ACE), interstitial macrophages (C1qa⁺IMs), and two recruited inflammatory macrophage subsets (Ly6c2⁺recMacs and Ccr2⁺recMacs). Resident AMs subsets expressed canonical markers Siglecf, Car4, and Pparg, while AMs_Mki67 showed high expression of proliferation markers Mki67, indicating active proliferation. MARCO_ACE represents a macrophage subset characterized by high ACE expression. In contrast, recruited macrophages (recMacs) expressed monocyte-derived inflammatory markers such as Ly6c2 and Ccr2, indicative of recent bone marrow-derived recruitment. The C1qa⁺IMs subset expressed transitional tissue-resident markers including C1qa, Mafb, and Apoe. Collectively, these data demonstrate significant heterogeneity and plasticity within lung macrophages during Mab-R infection, reflecting complex adaptive processes involving loss and regeneration of resident AMs and differentiation of recruited monocytes into distinct functional subsets.

Quantitative analysis of macrophage subset frequencies across conditions demonstrated a profound temporal reorganization (Figure 6E). In uninfected control lungs, the majority of lung macrophages were the Siglec-F+ AMs (∼21.6% at 1 week, increasing to 41.4% by 6 weeks as the mice matured). By contrast, in Mab-R-infected lungs, resident AMs were largely depleted during acute infection, comprising only ∼3.8% of macrophages at 1 week. Even at 6 weeks post-infection, AMs represented ∼10.3% of the macrophage pool in infected lungs, significantly below the ∼41% in age-matched controls, indicating an incomplete recovery of the resident macrophage niche. The acute macrophage pool in infected lungs was instead dominated by inflammatory monocyte-derived cells: Ly6Chi macrophages comprised 42.3% of all macrophages at 1 week (versus <1% in controls), and Ccr2hi recruited macrophages constituted 38.0% (versus 32.4% in controls). Together, these newly recruited subsets accounted for ∼80% of the lung macrophages in early infection. By 6 weeks, the proportion of Ly6Chi macrophages in infected lungs had dropped to 17.9%, consistent with their transient role, while Ccr2hi macrophages declined to 24.5%, near the level in controls (21.8%). During the transition to chronic infection, the C1qhi macrophage subset expanded dramatically. This subset made up only ∼9.9% of macrophages at 1 week infected, but surged to 33.95% by 6 weeks, becoming the largest macrophage population in chronic infection (compared to ∼17% in control lungs). The ACE+ regulatory macrophages were nearly absent in acute infection (1.07% at 1 week vs. 14.6% in controls), but recovered to 6.4% at 6 weeks (still about half of control levels, 11.2%). Proliferating Ki-67+ AMs remained rare in infected lungs at both time points (<0.5% of macrophages) compared to controls (where they were ∼1.6% at 1 week), underscoring the impaired self-renewal of the resident macrophage pool during infection. Collectively, these shifts show that Mab-R infection acutely replaces the resident AMs with incoming inflammatory monocytes, and over time these cells differentiate into C1q-rich tissue macrophages as the infection becomes chronic.

Innate immune and metabolic reprogramming of pulmonary macrophage subsets Following Mab-R infection

We compared the transcriptomic profiles of four major macrophage subsets during the acute phase of infection with those of controls, identifying significantly DEGs in each subset (|log₂FC| > 1, false discovery rate (FDR) <0.05). KEGG pathway enrichment analysis was then performed separately for upregulated and downregulated genes in each subset, with significantly enriched pathways (FDR <0.05) shown in Figures S3 and S4. To systematically assess functional similarities and differences across subsets, we integrated all enriched pathways and compared their distribution among subsets, as visualized in a heatmap (Figure S5).

Both AMs and IMs exhibited upregulation of pathways related to pathogen recognition (TLR, NLR, and CLRs), phagocytosis and degradation (phagosome, lysosome, and proteasome), cell death regulation (apoptosis, necroptosis, and ferroptosis), and pro-inflammatory immune responses (IL-17 and TNF signaling). Meanwhile, pathways such as cholesterol metabolism, Th17 cell differentiation, and IL-17 signaling were commonly downregulated.

Notably, in AMs and IMs, several pathways (e.g., NOD-like receptor, C-type lectin receptor, lysosome, and IL-17 signaling in AMs; IL-17 and TNF signaling in IMs) showed simultaneous up- and downregulation of different components, indicating dynamic regulation. This suggests a fine-tuned balance between inflammation and tissue homeostasis in response to infection or damage. Key genes dysregulated in both AMs and IMs during acute infection include Slc7a11 and Fth1 (oxidative stress and ferroptosis resistance), Pycard (inflammasome activation), Cd63, Ctsk, Atp6v0c (lysosomal function), and Acsl1, Lpcat2 (lipid metabolism and membrane remodeling) (Figure S3).

Among the recruited macrophages, Ly6c2⁺recMacs and Ccr2⁺recMacs displayed distinct enrichment profiles compared to controls, reflecting their different stages of functional differentiation and potential roles in immune regulation. During the acute phase of infection, we identified very few shared DEGs between Ly6c2⁺recMacs and Ccr2⁺recMacs (Figure S4), such as Slc7a11 and Fth1, which are involved in oxidative stress response and iron homeostasis. We still need detailed functional scoring to quantify the gene regulatory differences between different subsets and experimental groups.

Stage-specific remodeling of macrophage function during mab infection

During the acute phase of infection, we performed pseudotime analysis of pulmonary macrophages and identified five major lineage branches, all originating from the Ccr2⁺recMacs subset (Figure 7A). In lineage 1, Ccr2⁺recMacs differentiates into the inflammatory Ly6c2⁺recMacs subset, then transitions into AMs, and ultimately into AMs_Mki67. Lineage 2 differentiates into Marco_Ace macrophages with tissue-repair/regulatory characteristics, while lineage 3 forms Ms_C1qa macrophages associated with complement production, chronic inflammation, and tissue remodeling. Lineages 4 and 5 terminate in “Doublet” or “Other” states, likely representing doublet artifacts or immature precursors. These findings suggest that during acute Mab-R infection, peripherally recruited macrophages exhibit multipotent differentiation potential, leading to functionally diverse and spatiotemporally heterogeneous outcomes.

Figure 7.

Pseudotime analysis and functional characterization of macrophage subsets during Mab infection

(A) Single-cell pseudotime trajectories originating from recruited monocyte-derived macrophages (Ccr2⁺recMacs), differentiating into distinct macrophage subsets including Ly6c2⁺recMacs, alveolar macrophages (AMs), proliferative AMs (Mki67⁺AMs), interstitial macrophages (C1qa⁺IMs), Marco_Ace macrophages, and minor doublet/other populations.

(B) Functional scoring (phagocytosis, inflammation, oxidative stress, and tissue remodeling) of macrophage subsets at week 1 post-infection.

(C and D) Heatmaps displaying expression dynamics of key genes along lineage 1 (Ccr2⁺recMacs → Ly6c2⁺recMacs → AMs → Mki67⁺AMs) (C) and lineage 3 (Ccr2⁺recMacs → C1qa⁺IMs) (D).

(E) Comparison of functional and metabolic pathway scores among control subjects, acute infection (Week 1), and chronic infection (week 6) stages. Statistical analysis was performed using the Kruskal-Wallis test followed by Dunn’s multiple comparisons test. Statistical significance is indicated as follows: ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Using the functional framework proposed by Sanin et al.,¹⁰ which includes phagocytic regulation, inflammatory cytokine production, oxidative antimicrobial responses, and extracellular matrix remodeling, we integrated pseudotime analysis to assess the functional states of five pulmonary macrophage subsets (Table S3). The results showed that Ccr2⁺recMacs and Ly6c2⁺recMacs subsets exhibited the highest activity in phagocytosis and innate immune response pathways, surpassing AMs, indicating stronger phagocytic and immune-initiating capacity. Antigen presentation, oxidative stress, and complement pathways were enhanced during early differentiation but declined sharply upon transition to AMs (Figure 7B; Figures S6 and S7). Notably, Ly6c2⁺recMacs showed robust oxidative stress activity, highlighting its critical role in intracellular pathogen clearance.

In contrast, AMs and AMs_Mki67 scored highest in ECM and actin regulation pathways, while Ly6c2⁺recMacs scored lowest, indicating that AMs possess superior cytoskeletal remodeling and migratory capacity for rapid recruitment to infection sites. However, their limited antimicrobial activity may allow pathogen persistence and dissemination. The progressive increase in ECM scores along lineage 1 suggests enhanced tissue remodeling during differentiation.

C1qa⁺IMs appeared skewed toward homeostasis or chronic inflammation regulation, with high scores in complement, phagocytosis, and antigen presentation pathways, indicating key roles in pathogen processing and adaptive immunity. Despite moderate phagosome activity, low scores in innate immunity, oxidative stress, and ECM remodeling suggest limited pro-inflammatory and reparative functions (Figure 7B; Figures S6 and S7).

Analysis of key genes from the pseudotime results of different subsets reveals that peripheral monocyte-derived macrophages regulate key functions such as phagocytosis, inflammation, oxidative stress, cytoskeletal remodeling, cell cycle, antigen presentation, and immune regulation, which determines their differentiation into effector, regulatory, or resident phenotypes (Figures 6D and 7C).

We further compared the differences in macrophage functional and metabolic scores between the acute infection group, chronic infection group, and control group to reveal the underlying mechanisms of macrophage behavior during infection chronicization. In the macrophage functional scores, the antigen presentation pathway continuously upregulated across all macrophage subsets as the infection progressed into the chronic phase. From the acute to chronic phase, the innate immune response and oxidative stress pathways downregulated in all subsets, while the complement and phagocytosis pathway mainly upregulated (Figure 7E). From acute to chronic phase, the proportion of C1qa⁺IMs cells increased, but their phagocytic function decreased, whereas Ly6c2⁺recMacs showed an increase in phagocytic function, though its cell proportion significantly decreased. Notably, although Ccr2⁺recMacs had a higher proportion of cells in the acute phase, its phagosome pathway was downregulated compared to the control group, even though it remained the most potent in phagocytic function among the four subsets. During Mab infection, macrophage lipid metabolism and the TCA cycle exhibited clear temporal dynamics and subset differences (Figure 7E). Fatty acid synthesis pathways were broadly upregulated in week 1 of infection, indicating that acute-phase macrophages activate lipid synthesis to meet the demands of membrane reconstruction, inflammatory mediator production, and signal transduction, reflecting strong immune activation. As the infection progressed to the chronic phase (week 6), fatty acid synthesis generally decreased, possibly reflecting the host’s active downregulation of metabolic pathways to limit pathogen access to lipids, or Mab entering a dormant state with reduced lipid dependency. Notably, in terms of fatty acid degradation, AMs showed early suppression followed by recovery, suggesting a transition from inhibition to functional rebuilding; while C1qa⁺IMs and Ly6c2⁺recMacs gradually downregulated fatty acid degradation after w1, Ccr2⁺recMacs showed no significant changes. The temporal dynamics and subset differences in lipid metabolism may be exploited by Mab for its own cell wall synthesis and growth, particularly in AMs.

Regarding cholesterol metabolism, AMs maintained relatively stable metabolic levels, while Ccr2⁺recMacs and Ly6c2⁺recMacs significantly upregulated cholesterol metabolism during early activation, followed by a gradual decline. C1qa⁺IMs exhibited an initial suppression followed by recovery. The TCA cycle in C1qa⁺IMs showed a distinct pattern compared to the other three subsets, being upregulated during the acute phase and downregulated during the chronic phase. This suggests that Mab may adapt to the intracellular environment by utilizing the cholesterol metabolism of C1qa⁺IMs (Figure 7E). These results indicate that macrophages undergo functional switching through the remodeling of lipid metabolism and the TCA cycle, while Mab adapts to corresponding metabolic modes through transcriptional regulation, utilizing different carbon sources to prolong intracellular survival.

Analysis of T cell subsets revealed a progressive remodeling of T cell composition in Mab-infected lungs compared with the control group. During early infection, naive/memory T cells remained the dominant population, whereas by week 6, the proportions of effector-associated subsets, including Gamma-delta T cells, Th1-like CD4 T cells, NKT-like T cells, and cytotoxic CD8 T cells, were markedly increased (Figure S8), suggesting that the T cell immune microenvironment gradually shifts from a relatively quiescent steady-state condition to a more activated and functionally specialized state during chronic Mab infection. This trend was consistent with the dynamic changes observed in macrophage subsets: early infection was characterized by the recruitment of inflammatory Ly6C2⁺CCR2⁺ macrophages, indicating a predominance of innate immune responses, whereas in the chronic stage, the expansion of repair-associated macrophages such as MARCO_ACE was accompanied by enrichment of effector T cell subsets, suggesting a gradual strengthening of adaptive immunity. Collectively, these findings indicate that macrophage remodeling may be closely linked to T cell activation and functional differentiation, together contributing to the dynamic balance between inflammatory control and tissue repair during chronic Mab infection.

Spatial distribution of Mab infection and alveolar macrophages

To investigate the spatial relationship between Mab and macrophages, particularly AMs, during infection, we performed multicolor immunofluorescence and acid-fast staining on lung sections from control, Mab-R-W1, and Mab-R-W6 groups. Markers included DAPI (nuclei), ab905 (Mab), F4/80 (macrophages), and Siglec-F (AMs).

Acid-fast staining revealed abundant mycobacteria at both week 1 and week 6, indicating sustained infection (Figure 8A). Immunofluorescence analysis showed that the proportion of Mab co-localized with F4/80⁺ macrophages increased from 20.0% at week 1–28.3% at week 6 (Figure 8C). Notably, only 1% of Mab co-localized with Siglec-F⁺ AMs at week 1, whereas this proportion rose to 19.9% at week 6, suggesting increased association with AMs during chronic infection (Figure 8D).

Figure 8.

Spatial co-localization analysis of Mab and macrophages

(A) Acid-fast staining.

(B) Immunofluorescence staining.

(C) and (D) Co-localization analysis of Mab with macrophages and alveolar macrophages.

Furthermore, the spatial distribution of Mab and macrophages shifted over time. At week 1, bacteria and macrophages were scattered throughout the tissue, whereas by week 6, they formed clustered, granuloma-like structures, reflecting organized immune containment during chronic infection (Figure 8B).

Discussion

In this study, we demonstrate that Mab undergoes rapid adaptive evolution within the host by transitioning to the rough morphotype, thereby enhancing its virulence. We further elucidate the critical role of GPL loss in this morphotypic shift and its global impact on Mab transcriptional reprogramming. In addition, we systematically characterized the dynamic host immune responses during chronic infection using the Mab-R infection model, as this strain induces a more pronounced immune response in vivo. Using a high-dose murine infection model, we show that R-type strains establish a higher bacterial burden in the lungs and significantly reduce host survival, indicating increased pathogenicity in vivo. These findings are consistent with previous studies in mouse⁷ and zebrafish models,⁵ reinforcing the high-virulence phenotype of R-type Mab.

Loss of GPL in Mab triggers global transcriptional reprogramming, markedly enhancing its metabolic adaptability, antibiotic resistance, virulence, and immune evasion. At the metabolic level, the rough morphotype showed upregulation of pathways involved in fatty acid β-oxidation, the glyoxylate shunt, and lipid uptake (e.g., MCE operons). Combined with metabolic signatures inferred from host single-cell transcriptomic data, we propose that Mab-R may shift toward utilizing host-derived lipids as a carbon source, thereby adapting to the hypoxic and glucose-limited microenvironment characteristic of inflamed lung tissue. Concurrently, they exhibit increased iron acquisition (via ESX-3 and ESX-4 systems) and strengthened oxidative stress defenses, improving resistance to host immune attacks. ESX-3 facilitates iron/zinc scavenging through EsxG-EsxH secretion, supporting bacterial metabolism and virulence, and may suppress host NF-κB/TLR signaling to reduce pro-inflammatory cytokine production (e.g., TNF-α, IL-1β), aiding immune evasion.²⁰ ESX-4 disrupts phagosomal membranes via EsxU/T secretion, a function that can be compensated by ESX-3.²⁰ Moreover, R-type strains show broad upregulation of antibiotic resistance genes, contributing to increased drug tolerance within the host. We found that 91.7% of clinical strains harbored nonsynonymous mutations within the GPL locus, exhibiting clear clustering patterns associated with specific subspecies or DCC groups. The massiliense subspecies showed a higher frequency of mutations, particularly in MAB_4098c and MAB_4099c, while the abscessus subspecies had mutations predominantly in genes involved in methylation and glycosylation. This genetic variation pattern likely contributes to subspecies-specific differences in pathogenicity and immune evasion, and offers valuable insights for the development of subspecies- or DCC-specific typing markers and targeted therapeutic strategies.

In addition to characterizing pathogen-specific traits, we performed high-resolution single-cell RNA sequencing of lung immune cells from mice at 1 and 6 weeks post-infection, compared with uninfected controls, to delineate the temporal evolution of host immune responses during Mab chronic infection. Early infection was marked by a robust innate inflammatory response. Mab-R caused a significant reduction in the numbers of AMs and C1qa⁺IMs, accompanied by strong signatures of programmed cell death, including apoptosis, necroptosis, and ferroptosis. Despite reduced numbers, residual AMs and IMs upregulated genes associated with extracellular matrix remodeling and actin cytoskeleton regulation, suggesting enhanced migratory and phagocytic capacity and initiation of tissue repair to mitigate inflammation-induced damage. These findings indicate that AMs, as the first line of defense encountering Mab,⁸ may undergo programmed cell death due to the pathogen’s immune evasion mechanisms, thereby limiting intracellular replication at the cost of host cell survival. The rapid recruitment and differentiation of Ly6C^hi inflammatory monocytes into macrophages is driven by pathogen-induced inflammatory signaling and the loss or dysfunction of resident AMs.²¹ We identified a macrophage subset characterized by high expression of Ly6C2 and CCR2, exhibiting potent phagocytic activity, pro-inflammatory cytokine production, and antigen presentation—key effector functions for early infection control. Consistent with CFU measurements, bacterial burden significantly declined in the lungs during weeks 1–2 post-infection but was not completely cleared. We speculate that this residual population escapes initial phagocytic clearance and disseminates via AM-mediated migration between macrophage subsets, facilitating chronic infection establishment. This stage reflects a host immune balance between inflammation control and tissue preservation, and reveals a key strategy by which Mab disrupts AM function and induces their programmed death to evade immunity and establish long-term persistence.

By six weeks post-infection, the immune landscape shifted toward a more regulated chronic profile, with pro-inflammatory macrophages waning and repair-oriented subsets MARCO_ACE macrophages becoming more prominent.²² All macrophage subsets in the chronic stage maintained high phagocytic and antigen-presenting activity, but showed reduced acute inflammatory and oxidative stress signatures compared to week 1. By week 6, we noted an increased presence of T lymphocytes, indicating engagement of adaptive immunity. Taken together, our single-cell analysis depicts a temporal “dance” of immunity: an early burst of inflammatory innate defense that gives way to a more controlled, chronic immune state with greater involvement of regulatory myeloid cells and adaptive lymphocytes. Integrating the transcriptional adaptation profile of the Mab-R strain with functional and metabolic scoring of pulmonary macrophages during acute and chronic infection stages, we propose that Mab-R possesses an enhanced capacity to exploit host-derived lipids, primarily driven by significant upregulation of MCE operons. This facilitates active uptake of exogenous lipids, including fatty acids and cholesterol. In the hypoxic, nutrient-limited lung environment, Mab-R adapts to metabolic stress by flexibly switching between carbon utilization pathways—such as the TCA cycle, lipid metabolism, and amino acid metabolism—to maintain survival advantages.

Single-cell metabolic profiling further revealed differential lipid utilization preferences across macrophage subsets. In AMs, upregulation of fatty acid synthesis coupled with downregulation of fatty acid degradation suggests a reliance on fatty acids as the primary carbon source. In contrast, IMs exhibited downregulation of cholesterol metabolism, indicating a shift toward cholesterol as a preferred nutrient source in this subset. These findings highlight that Mab-R adopts distinct metabolic strategies tailored to specific host cell environments, enabling efficient exploitation of diverse lipid resources. This multifaceted metabolic adaptability likely underpins its ability to persist at high bacterial loads and evade immune clearance during chronic infection.

Previous studies have typically focused on the role of individual molecular mechanisms or specific signaling pathways in Mab pathogenicity, whereas relatively little attention has been paid to the systemic interplay between pathogen-wide metabolic adaptation and the host immune environment. In this study, by integrating a time-resolved infection model with single-cell transcriptomic analysis, we systematically characterized the global metabolic reprogramming that occurs during Mab adaptation within the host. Furthermore, we revealed potential links between these metabolic changes and the dynamic remodeling of host immune cell populations, thereby providing a more comprehensive perspective for understanding the mechanisms underlying chronic Mab infection.

Limitations of the study

This study has several limitations. First, although we identified a high-frequency mutation in the GPL locus, particularly in MAB_4105c, we did not perform direct genetic validation to establish its causal link to enhanced virulence. It therefore remains unclear which phenotypic and transcriptional changes are directly caused by this mutation and which arise secondarily from altered growth behavior, increased aggregation, or changed cell-cell interactions in the rough variant. Although we used low-speed centrifugation and collected the supernatant to obtain more dispersed bacterial suspensions, aggregation of rough variants could not be fully controlled during in vitro culture or in vivo infection. Future studies combining genetic complementation with stricter control of bacterial aggregation are needed to distinguish direct from secondary effects.

Second, because single-cell transcriptomic data were not available for the critical transition period between weeks 2 and 4 post-infection, we could not fully define the dynamic shift of host immune responses from the acute to the chronic stage. In addition, in vivo Mab transcriptomic analysis was performed only at 1 week post-infection, because sufficient bacterial RNA could be obtained only at this early stage. As bacterial burden declined at later time points, reliable bacterial transcriptomic profiling became technically difficult. Thus, the bacterial gene expression patterns reported here mainly reflect the early stage of infection.

Finally, although single-cell analysis identified macrophage subsets associated with infection progression, their functional roles were not validated in vivo. Further studies using genetic or depletion-based approaches will be needed to define their contributions to immune regulation and infection persistence.

During chronic Mab infection, metabolic regulation serves as a key driver of host-pathogen interactions, influencing host infection tolerance, modulating immunopathology, and maintaining organ homeostasis, while also facilitating Mab’s persistence and the establishment of chronic infection. This metabolic interplay is highly bidirectional, cell specific, and temporally dynamic, representing a core mechanism that tightly couples immunity and metabolism throughout infection. It provides a novel conceptual framework for understanding refractory infections and developing metabolism-targeted therapeutic strategies.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Bijie Hu (hu.bijie@zs-hospital.sh.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The raw data supporting the findings of this study have been deposited in the NCBI database under BioProject accession number PRJNA1475709. The data are publicly accessible at https://dataview.ncbi.nlm.nih.gov/object/PRJNA1475709. Additional data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgments

We appreciate the technical support provided by the Department of Laboratory Medicine, Zhongshan Hospital.

This study was funded by Zhongshan Hospital, Fudan University (2023ZSLC24), and Shanghai Hospital Development Center Foundation (SHDC22024315). The funders had no role in the design of the study, collection and analysis of data, decision to publish, or preparation of the manuscript.

Author contributions

Y.C. and Q.M. contributed equally to this work; Y.C., T.F., and Q.M. designed the study; L.H., J.P., X.G., L.Q., and Q.M. performed the experiments; Y.C., Q.M., and T.F. analyzed the data; Y.C. and Q.M. wrote the paper; B.H. and T.F. revised the manuscript; all authors gave final approval of the version to be published.

Declaration of interests

The authors declare no conflict of interest.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Experimental models: Organisms/strains
Mouse: C57BL/6J, male, 6–8 weeks old	Shanghai Ji Cui Yao Kang Biotechnology Co., Ltd.	N/A
Chemicals, peptides, and recombinant proteins
Middlebrook 7H10 Agar	BD Difco	Cat# 262710
Middlebrook 7H9 Broth	BD Difco	Cat# 271310
Middlebrook OADC Enrichment, 10 × 20 mL	BD BBL	Cat# 211886/Fisher Cat# B11886
Middlebrook OADC Enrichment, 6 × 100 mL	BD BBL	Cat# 212240/Fisher Cat# B12351
Collagenase, Type IV, powder	Gibco	Cat# 17104019
Papain, lyophilized powder	Sigma-Aldrich	Cat# P5306
DNase I from bovine pancreas	Sigma-Aldrich	Cat# DN25
MACS Red Blood Cell Lysis Solution (10×)	Miltenyi Biotec	Cat# 130-094-183
Dead Cell Removal Kit/Dead Cell Removal MicroBeads	Miltenyi Biotec	Cat# 130-090-101
Trypan Blue Solution, 0.4%	Gibco	Cat# 15250061
PBS	Gibco/Thermo Fisher Scientific	Cat# 10010023
BSA	Sigma-Aldrich	Cat# A9418
Tween 20	Sigma-Aldrich	Cat# P1379
Paraformaldehyde (PFA)	Sigma-Aldrich	Cat# P6148
Mineral oil	Sigma-Aldrich	Cat# M8410
Tryptic Soy Agar (TSA)	BD Difco	Cat# 236950
Critical commercial assays
RNA 6000 Nano Kit for Agilent 2100 Bioanalyzer	Agilent Technologies	Cat# 5067-1511
MobiCube® Single-Cell 3′ RNA-seq Kit	MobiDrop	N/A
Other
MicroSprayer® Aerosolizer for mouse	Penn-Century	Model IA-1C-M
MicroSprayer® Aerosolizer (general IA-1B series)	Penn-Century	Model IA-1B
MobiNova®-100 Single-Cell System	MobiDrop	N/A
BACTEC MGIT 960 automated culture system	BD	N/A
Countess II Automated Cell Counter	Thermo Fisher Scientific	N/A
Illumina HiSeq X Ten platform	Illumina	N/A
Illumina NovaSeq 6000 platform	Illumina	N/A

Experimental model and study participant details

Study design

This study employed multiple high-throughput techniques, including bulk RNA sequencing (RNA-seq) and single-cell RNA sequencing (scRNA-seq), to systematically analyze transcriptional differences between Mab-S and Mab-R strains under nutrient-rich laboratory culture conditions and the complex host microenvironment in mice. Additionally, we focused on the dynamic functional changes of host macrophages during acute (1 week) and chronic (6 weeks) lung infections induced by Mab-R. All animal experiments were conducted with randomized grouping; detailed biological and technical replicates are described in the figure legends. All in vivo experiments comparing Mab-R and Mab-S in this study, including single-cell RNA sequencing and bacterial burden measurements, were performed using the agar bead–based infection model (agar-bead model). All experimental protocols were performed in accordance with relevant guidelines and regulations and were approved by the Institutional Animal Care and Use Committee of Ji Cui Yao kang Laboratory Animal Company (Ethical Approval ID: SHAP20241010-1).

Bacterial strains

Rough-type Mab strains were isolated from the lung tissues of infected mice, and single-clone colonies were obtained by picking individual colonies. All 109 clinical isolates were obtained from bronchoalveolar lavage fluid (BALF) samples collected by the Department of Infectious Diseases, Zhongshan Hospital, Fudan University, and were isolated using the BACTEC MGIT 960 automated culture system. These strains were cultured on Middlebrook 7H10 solid medium supplemented with oleic acid–albumin–dextrose–catalase (OADC) to expand bacterial populations for downstream analyses. DNA extracted from collected colonies was quality-assessed, and sequencing libraries with ∼400 bp insert sizes were constructed. Paired-end sequencing (PE150) was performed on the Illumina HiSeq X Ten platform with a minimum coverage depth of 100× per sample to ensure data quality and analytical accuracy.

Mouse lung infection model

All animal experiments were performed using specific pathogen-free (SPF) male C57BL/6J mice (6–8 weeks old, n = 90) purchased from GemPharmatech Co., Ltd (China). Mice were housed in individually ventilated cages under SPF conditions with free access to food and water. All in vivo procedures were conducted in the provider's SPF-grade animal facility under the direct supervision of professional veterinarians, and all researchers held valid national certifications for laboratory animal care. After one week of acclimation, mice were randomly assigned (stratified by body weight) to three groups (n = 30 each): Mab-R infection, Mab-S infection, and PBS control. On day 0, mice in the infection groups received intratracheal inoculation of 1 × 10⁷ CFU of Mab-R or Mab-S in a 100 μL suspension using a MicroSprayer® Aerosolizer under light isoflurane anesthesia. This relatively large volume ensures rapid aerosolization for deep lung penetration without compromising animal welfare or causing airway occlusion. Control mice received 100 μL of sterile PBS via the same aerosolization technique. Mycobacterium abscessus was cultured on 7H10 agar plates at 37 °C for 5 days. Three to four colonies were inoculated into 7H9 broth supplemented with 10% OADC and 0.2% glycerol and grown at 37 °C with shaking (190 rpm) for 24 h to reach approximately OD600 ≈ 1. Bacterial density was determined by serial dilution and CFU plating on 7H10 agar. For agar bead preparation, bacterial cultures were pelleted (2700 g, 4 °C), washed twice with PBS, and resuspended in PBS. The suspension was mixed with 1.5% TSA agar (48 °C) and emulsified in pre-warmed mineral oil with stirring to generate agar beads. The emulsion was rapidly cooled to solidify the beads, which were subsequently collected and washed multiple times with sterile PBS to remove mineral oil. The final suspension was homogenized, serially diluted, and plated on 7H10 agar to determine CFU. The agar bead suspension was adjusted with PBS to the desired concentration prior to mouse infection, and inoculum CFUs were confirmed by plating.

The condition of the mice was monitored daily after infection, and their body weight was measured every three days. At weeks 1, 2, and 6 post-infection, five mice from each group were sacrificed to collect lung tissues. Fresh lung tissues were immediately processed for single-cell sequencing, while the remaining samples were stored in PBS containing 0.1% Tween-20 for subsequent analyses. All deaths and humane endpoints were recorded for Kaplan–Meier survival analysis.

Method details

Bacterial culture and CFU assay

The whole lung and the left lobe were weighed separately. The left lobe was homogenized in 2 mL of 0.1% PBST, followed by 10¹–10⁷ serial dilutions. Aliquots of 100 μL from each dilution were plated onto Middlebrook 7H10 agar in triplicate and incubated at 37°C for 5–7 days. Colony-forming units (CFUs) were counted and expressed as CFU/mL to assess bacterial burden. Whole-lung bacterial loads were calculated from these counts and are presented in the Table S4.

Whole genome analysis

Raw sequencing reads were preprocessed using Fastp to remove low-quality reads, adapter sequences, and trim read ends. Core genome single nucleotide polymorphisms (SNPs) were identified by mapping reads to the Mab ATCC 19977 reference genome (GenBank accession GCA_000069185) using Snippy. SNP calling was conducted with SAMtools (v1.3.1), retaining only high-quality SNPs with mapping quality >30. VarScan (v2.3.9) was further employed to detect fixed mutations and indels supported by at least five reads. Functional annotation of variants, including coding, non-coding regions and protein impact, was performed with SnpEff v5.1. Variants in low coverage regions were recalibrated using joint variant calling, and highly polymorphic, repetitive, prophage, and PE/PPE regions were excluded. Variants clustered within 10 bp intervals with ≥3 SNPs were also filtered out.

Cell suspension preparation

Collected lung tissues were transferred into precooled culture dishes containing 1× PBS (RNase-free, Ca²⁺ and Mg²⁺ free) on wet ice, minced into ∼0.5 mm² pieces, and washed repeatedly with 1× PBS to remove blood and lipids. Tissue fragments were enzymatically digested in a solution containing 0.35% collagenase IV, 2 mg/mL papain, and 120 U/mL DNase I at 37°C on a shaking water bath (100 rpm) for 20 minutes. Digestion was terminated by adding 1× PBS with 10% fetal bovine serum (FBS), followed by gentle pipetting to dissociate cells. Cell suspensions were filtered sequentially through 70 μm and 30 μm strainers, centrifuged at 300 ×g for 5 minutes at 4°C, and resuspended in 100 μL 1× PBS with 0.04% BSA. Red blood cells were lysed with 1 mL of MACS RBC lysis buffer (MACS 130-094-183, 10×) at room temperature or on wet ice for 2–10 minutes, followed by centrifugation at 300 ×g for 5 minutes. Dead cells were removed using 100 μL Dead Cell Removal MicroBeads (MACS 130-090-101) per the manufacturer’s protocol. After reagent removal, cells were washed twice with 1× PBS (0.04% BSA) and centrifuged at 300 ×g for 3 minutes at 4°C. The final viable cell suspension was adjusted to 700–1200 cells/μL in 50 μL 1× PBS (0.04% BSA), with viability >85% confirmed by trypan blue staining and cell counting using a hemocytometer or automated counter (Countess II).

Single-cell RNA sequencing library construction and sequencing

High-quality single-cell suspensions (without prior sorting of specific cell populations) were loaded into the MobiNova®-100 Single-Cell System and processed according to the MobiCube® Single-Cell 3′ RNA-seq Kit protocol for cell capture (targeting 5,000 cells), cDNA synthesis, amplification, and library preparation. Libraries were sequenced on the Illumina NovaSeq 6000 platform using 150 bp paired-end multiplexing runs, targeting a minimum sequencing depth of 20,000 reads per cell to ensure sufficient coverage for downstream analyses.

Single-cell RNA-seq data analysis

Raw sequencing data were processed using the MobiVision® software and MobiVision quantification tool to align reads to the mouse reference genome and transcriptome (GRCm39), and low-quality reads were filtered to generate a feature-barcode matrix. Quality control was performed in Seurat (v5.2.0) to exclude cells with fewer than 200 detected genes or greater than 35% mitochondrial gene expression. Mutual nearest neighbors (MNN) integration was applied to correct batch effects and integrate data across samples. Highly variable genes were identified using the FindVariableFeatures function, followed by principal component analysis (PCA) for dimensionality reduction. Cells were clustered using the FindClusters function with resolution set to 2, and visualization was performed via UMAP or t-SNE. Cluster-specific marker genes were identified using the FindAllMarkers function with a bimodal likelihood ratio test. Clusters were manually annotated based on canonical marker genes and recent single-cell literature. Functional pathway scoring of macrophages was conducted using AddModuleScore, with gene sets listed in supplementary tables.

Developmental trajectory inference

Slingshot was used to infer developmental trajectories of macrophage subpopulations without predefined start or end points, allowing data-driven identification of differentiation paths. For biologically relevant lineages, tradeSeq was applied to analyze dynamic gene expression along trajectories. Genes with adjusted p-values <0.05 and among the top 25 ranked by Wald statistics were visualized by heatmaps, revealing key regulatory genes involved in functional transitions and differentiation processes.

Hematoxylin and eosin (H&E) staining

Mouse lung tissues were collected, washed with PBS, and fixed in 4% paraformaldehyde (PFA). Samples were dehydrated through graded ethanol series and embedded in paraffin. Four-micron sections were prepared, deparaffinized, and stained sequentially with hematoxylin and eosin. Histopathological changes were assessed under light microscopy to evaluate pulmonary inflammation and tissue morphology.

Imaging method for co-localization of pulmonary infection in vivo

To delineate the spatial relationship between Mab and lung macrophage populations with high precision, we performed high-resolution confocal laser scanning microscopy (CLSM) on 4 μm lung sections. To quantify bacterial internalization versus extracellular adherence, z-stack images (step size: 0.3 μm) were acquired and subjected to 3D volumetric reconstruction using Imaris v9.5 software. Co-localization was defined based on a stringent threshold where ≥90% of the bacterial fluorescence signal was encapsulated within the 3D-rendered cellular boundaries of F4/80+ or Siglec-F+ cells. This approach ensured a precise topographic analysis of host-pathogen interactions and effectively excluded pseudo-co-localization artifacts during the transition from acute to chronic infection phases.

Bulk RNA sequencing and analysis

Infected lung tissues were homogenized in TRIzol, followed by low-speed centrifugation (150 × g, 5 min) to remove host debris. The supernatant was then subjected to high-speed centrifugation (12,000 × g, 10 min) to pellet mycobacteria. The bacterial pellet was mechanically lysed by bead-beating (FastPrep, 6.0 m/s, 40 s), after which RNA was extracted using the TRIzol method followed by DNase I treatment. Total RNA was extracted from bacterial samples using standard protocols, and RNA integrity was assessed with the Agilent 2100 Bioanalyzer. Specialized probes designed for bacteria (such as Ribo-Zero) were used to hybridize with high-abundance rRNA, which was subsequently removed using magnetic beads, thereby retaining target RNAs such as mRNA. Ribosomal RNA was depleted to enrich mRNA, which was then fragmented in NEB Fragmentation Buffer using divalent cations. Libraries were constructed following NEB standard or strand-specific protocols. Qualified libraries were pooled based on effective concentration and desired sequencing depth, then sequenced on an Illumina platform using sequencing-by-synthesis technology. Raw reads were quality-controlled and aligned to the Mab ATCC19977 reference genome using Bowtie2. Unique alignments were sorted with SAMtools, and gene counts were quantified using HTSeq (v0.11.3). Gene expression levels were calculated as TPM values, and differential expression analysis was performed using the DESeq2 package (v1.26.0). Raw read counts were normalized using the median-of-ratios method implemented in DESeq2 to correct for library size differences. Differentially expressed genes were identified using a threshold of |log₂FoldChange| ≥ 1, with adjusted p-values calculated using the Benjamini–Hochberg method to control the false discovery rate (FDR).

Quantification and statistical analysis

Statistical analysis

Data are presented as mean ± standard error of the mean (SEM) for groups of five mice. Differences between groups were assessed using Student’s t-test or one-way analysis of variance (ANOVA), followed by post hoc multiple comparison correction when appropriate (e.g., Tukey’s HSD test). The specific statistical methods used for each dataset are indicated in the figure legends. Survival analyses were performed using the Kaplan–Meier method with comparisons by log-rank test. All statistical analyses were conducted in R software (version 4.1.0), with p-values < 0.05 considered statistically significant.