Experimental evolution in the cystic fibrosis chemical environment reveals early TCA cycle flux as a central regulator of Mycobacterium abscessus biofilm formation

Abstract

Mycobacterium abscessus (MAB) is an emerging opportunistic pathogen that can cause severe, recalcitrant pulmonary infections in susceptible groups, including people with cystic fibrosis (CF). MAB forms biofilms during human infection and in environmental reservoirs such as household plumbing systems, and biofilm formation renders MAB more drug tolerant. However, our limited understanding of the regulatory systems governing mycobacterial biofilm formation undercuts our ability to disperse MAB biofilms and potentially increase treatment efficacy. Using experimental evolution, we demonstrate that selective pressure from synthetic cystic fibrosis sputum medium (SCFM1) drives the emergence of evolved MAB lineages that more readily aggregate in different environmental conditions. Whole-genome sequencing identified mutations in genes coding for two putative IclR family transcriptional regulators, which we named MraA and MraB, as responsible for the increase in aggregation. Using RNA-seq, we revealed that MraA and MraB share a regulon composed largely of genes involved in the early tricarboxylic acid (TCA) cycle and glutamate metabolism. Targeted metabolomic analysis confirmed that both mutants had increased levels of TCA cycle intermediates citrate/isocitrate and α-ketoglutarate (AKG), suggesting that in WT both MraA and MraB suppress flux through those metabolites. We found we could increase both citrate/isocitrate and AKG pools in WT MAB by supplementing SCFM1 with acetate, thereby increasing biofilm formation without increasing expression of the MraA/B regulon and demonstrating a specific causal relationship between those metabolites and biofilm formation. Finally, we show that acetate-induced, agar-suspended biofilms confer antibiotic tolerance. Altogether, we demonstrate how MAB carbon flux can be redirected by selective pressures in a CF sputum-like chemical environment to increase biofilm formation and drug tolerance. We propose a model in which MraA and MraB control flux of citrate/isocitrate/AKG and thereby feed into a metabolism-based biofilm regulatory system in MAB.

1. Introduction

Nontuberculous mycobacteria (NTM) are a group of environmental microbes that includes a subset of opportunistic pathogens [[1], [2], [3], [4]]. Over the past few decades, the global prevalence of NTM infections has seen a dramatic rise [5]. From natural and built environment reservoirs, pathogenic NTM, including Mycobacterium abscessus (MAB), can infect people with chronic pulmonary disorders such as cystic fibrosis (CF), chronic obstructive pulmonary disorder (COPD), and non-CF bronchiectasis, leading to severe, very difficult-to-treat, and sometimes lethal pulmonary infections [2,4]. This poor treatment outcome is attributed in part to MAB's strong inherent drug resistance [1,2,6]. In addition, there is often a disconnect between in vitro and in vivo MAB drug sensitivity, meaning even when MAB isolates appear susceptible to drugs in vitro, treatment of the patient with those same drugs often is not effective [3,7,8]. This phenomenon suggests that some difficulties in treating MAB infections are due to mechanisms of physiological tolerance; transient phenotypes such as biofilm formation that are not recapitulated in in vitro susceptibility tests [6,[9], [10], [11]].

The regulatory networks governing mycobacterial biofilm formation, or indeed biofilm formation in any Actinobacteria, are poorly understood compared to counterparts in Proteobacteria (Pseudomonas aeruginosa, Escherichia coli, etc.) or Firmicutes (Staphylococcus aureus, Bacillus subtilis, etc.). We know that MAB can form biofilms on both abiotic surfaces and in humans during pulmonary infection, and that MAB biofilm formation provides protection against stressors such as antibiotics and disinfectants in vitro [1,6,[12], [13], [14], [15]]. However, the environmental cues and regulatory networks controlling MAB biofilm formation, especially in the context of disease, are poorly understood. The tendency of mycobacteria to spontaneously clump in liquid medium has historically made it difficult to cultivate planktonic cells and elucidate the chemical cues and genetic elements underpinning the early stages of biofilm initiation, i.e. how free-living cells make the decision to aggregate together. Recent results, though, have shown that the planktonic to aggregate transition is dictated by the availability of carbon and nitrogen in both the model NTM Mycobacterium smegmatis and in MAB, enabling investigations into this heretofore elusive transition [16,17]. The chemical environments that trigger increased liquid aggregation also support more robust colony and pellicle biofilm formation, indicating shared regulatory pathways and suggesting that liquid aggregation can be used as a quantitative tool to study NTM biofilm regulation [16,17]. M. smegmatis uses intracellular glutamine as a nitrogen-availability sensor from which it controls aggregation [17]. However, the mechanisms by which carbon availability is sensed and translated to a biofilm regulatory network, how pathogenic NTM such as MAB regulate aggregation, and how the infection environment impacts these systems, remain unknown.

Insight into static biofilm regulatory networks of specific bacterial strains has the potential to reveal anti-biofilm targets, but it is also important to understand that pathogens adapt during chronic infections, and these regulatory networks can and do change. For example, Pseudomonas aeruginosa, a common CF pathogen, frequently acquires mutations leading to constitutive expression of the matrix component alginate and a mucoid phenotype, which can protect the cells from phagocytosis, opsonization, and antibiotics [[18], [19], [20]]. MAB, too, can show altered biofilm phenotypes during chronic infection. Most notably, the loss or decrease of glycopeptidolipid (GPL) surface molecules results in a constitutive aggregation phenotype (rough colony morphology) in roughly half of CF isolates [21,22]. Prior literature has largely focused on the immune system as a primary provider of selective pressure that drives adaptive changes to biofilm formation in MAB and other CF pathogens [11,20,21]. However, there is precedent that adaptation to the chemical environment of an infection site alone is sufficient to alter metabolism and biofilm formation [[23], [24], [25]]. For instance, evolution experiments with P. aeruginosa in medium that mimics the CF chemical environment resulted in changes to biofilm regulatory pathways, including the quorum sensing regulator lasR and c-di-GMP signaling regulators, leading to an increase in biofilm formation and drug tolerance in the evolved population [23,24]. Given that MAB infections can become chronic and last for many years, and given the recent results showing that carbon and nitrogen availability impact NTM aggregation, we hypothesized that MAB adapts to the chemistry of the CF environment and that such adaptation may alter biofilm phenotypes, potentially revealing insight into these systems [2,4,16].

Using experimental evolution of MAB in synthetic cystic fibrosis sputum medium (SCFM1), we demonstrate in this study that selective pressure from a CF-like chemical environment can drive the emergence of evolved populations with a higher propensity to aggregate [26]. Whole-genome sequencing analysis revealed that mutations in genes coding for two previously undescribed IclR-like transcriptional regulators (MraA and MraB) led to an increase in MAB aggregation. Genomic and phylogenetic analysis revealed that both regulators are partially conserved among MAB genomes and are frequently mutated in MAB isolates, including those from human respiratory infection. RNA-seq analysis of both mutants revealed a shared regulon composed largely of genes involved in the early TCA cycle and glutamate metabolism. Targeted metabolomic analysis substantiated this, as both mutants had higher levels of TCA cycle metabolites citrate/isocitrate and α-ketoglutarate (AKG). By simply altering the available carbon sources in SCFM1, we found we could mimic the citrate/isocitrate/AKG increase without altering expression from our MraA/B regulon and still increase biofilm formation in wild-type (WT) MAB, demonstrating a causative relationship between these metabolites and aggregation. Altogether, our results reveal a fundamental connection between central metabolism and biofilm formation in MAB. Specifically, flux through the early TCA cycle drives MAB aggregation, and when this flux changes in response to selective pressure from the chemical environment it can subsequently impact aggregation dynamics and, by extension, biofilm formation. We hope these results will both serve as a foundation to build a more complete model of mycobacterial biofilm regulation and will inform the development of novel anti-biofilm treatment strategies to be used in conjunction with MAB antimicrobial therapies.

2. Results

2.1. Aggregation dynamics of MAB in a CF sputum-mimicking environment

To understand the aggregation dynamics of MAB in CF sputum-mimicking conditions, we quantified the proportion of aggregates and planktonic cells over time in SCFM1 [16,26]. We tested two smooth colony (WT) M. abscessus subsp. abscessus CF isolates, 0253a and 0711a (Fig. 1A & C), and paired rough colony variants, 0253b and 0711b (Fig. 1B & D), that were isolated from the same person with CF as their respective smooth colony strain [16]. MAB rough colony variants are typically the result of mutations decreasing GPL production and causing a constitutive aggregation phenotype and rough colony morphology [27]. To verify strain relatedness between our paired isolates, we performed whole genome sequencing of all four strains (Table S1). Indeed, there were only five mutations in 0253b relative to 0253a, including a frameshift mutation in mps1. Similarly, there were only three mutations in 0711b compared to 0711a, including a 489 bp deletion in mps2. Mps1 and Mps2 are non-ribosomal peptide synthetases involved in the generation of GPLs and are commonly mutated in rough colony variants [28]. Altogether these results suggest 0253b resulted from the mps1 mutation in an 0253a ancestor strain, and 0711b resulted from the mps2 mutation in an 0711a ancestor strain. In rich medium, MAB 0253b and 0711b constitutively aggregate and never disperse [16]. Similarly, in SCFM1, the rough variants MAB 0253b and MAB 0711b grew constitutively as aggregated cells, never dispersing (Fig. 1B & D). The smooth colony variants 0253a and 0711a aggregate and then disperse in rich medium [16], but surprisingly these strains both aggregated very little in SCFM1 (Fig. 1A & C). MAB 0711a displayed slightly more aggregation than MAB 0253a, reaching a peak mean OD₆₀₀ value of ∼0.1 at 48 h post-inoculation (Fig. 1C). Adding glucose to rich medium increases aggregation in both M. smegmatis and MAB [16]. Interestingly, glucose supplementation to SCFM1 failed to substantially induce aggregation in any of the strains (Fig. 1A–D).

Fig. 1.

Aggregation dynamics of MAB clinical isolates in SCFM1 with and without 30 mM glucose supplement.

MAB 0253a (A) and MAB 0711a (C) are smooth colony variants while MAB 0253b (B) and MAB 0711b (D) are rough colony variants. The chart shows individual OD₆₀₀ values of planktonic and aggregation fractions for each of three replicates taken at time points 0-, 44-, 48-, 64-, 68-, 88-, and 92-h post-inoculation. The line on the chart represents the mean OD₆₀₀ values at each time point.

2.2. Adaptive evolution in SCFM1 increases MAB biofilm formation

Next, we sought to assess how adaptation to the CF sputum-like chemical environment of SCFM1 impacts MAB biofilm formation. A single ancestor colony of MAB 0253a was passaged 15 times in SCFM1, SCFM1 with no glucose, or nutrient-rich TYEM medium (Fig. 2A). Bovine serum albumin was supplemented to all SCFM1 cultures at a concentration of 5 mg/mL to help promote mycobacterial growth (Fig. S1), consistent with an average total protein concentration in CF sputum of ∼4.78 mg/mL [29,30]. At the end of the evolution experiment, all passage 15 evolution lineages were spotted on TYEM agar plates and colony morphologies were characterized (Fig. 2B). Previously, we have shown that colony biofilm wrinkling in M. smegmatis correlates with more liquid aggregation [16,17]. Populations that evolved under SCFM1 and SCFM1 no glucose conditions developed novel colony phenotypes that were distinct from those of both ancestor and the TYEM-evolved populations and included a much more expansive colony with non-uniform, fuzzy edges and a distinct central ring structure. These structures were largely distinct from the hyper-wrinkled morphology characteristic of rough colony variants (Fig. 2B).

Fig. 2.

Experimental evolution of MAB 0253a in SCFM1 selects for novel biofilm phenotypes.

A) Graphical overview of the evolution experiment. SNG stands for SCFM1 no glucose B) Colony morphologies of all evolved populations from passage 15 of the evolution experiment on TYEM agar plates. The smooth variant MAB 0253a (ancestor) and the rough variant MAB 0253b are included for comparison. C) Aggregation dynamics of all evolved populations from passage 15 of the evolution experiment in TYEM media. The left plot depicts the peak aggregate OD₆₀₀ measurement of each lineage, taken within 40–52 h post-inoculation. The right plot depicts the aggregate OD₆₀₀ readout taken at 68 h post-inoculation. P-value obtained by unpaired Student's T-test against the Ancestor. ∗ denotes P-value <0.05. D) Graphical overview of MAB_0812 (mraA) and MAB_0813c (mraB) on the positive and negative strands, with nucleotide positions on the contiguous genome sequence labelled on the horizontal axis. Mutations observed in laboratory evolution experiments are denoted with a blue point. The helix-turn-helix (HTH) DNA-binding domain is green and the C-terminal effector-binding domain is red. Allelic variants for orthologs isolated from people with cystic fibrosis are denoted with red points and are labelled with their amino acid positions on each gene. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

We next assessed the aggregation dynamics of all evolved populations in TYEM. Nearly all SCFM1-evolved populations showed a significant increase in peak aggregation compared to the ancestor (Fig. 2C & Fig. S2). The one exception, SCFM1 lineage 3, displayed an interesting extended aggregation phenotype (Fig. 2C & Fig. S2C). TYEM lineage 1 displayed significantly higher peak aggregation than the ancestor (Fig. 2C & Fig. S2A), but neither TYEM lineage 2 nor 3 showed statistically significant differences in peak aggregation compared to ancestor. SCFM1 lineage 2 (S2) and SCFM1 no glucose lineage 3 (SNG3) showed the highest peak aggregation among their respective conditions, so we chose those strains for detailed follow-up studies. While S2 and SNG3 aggregated more than WT, they both displayed a temporal aggregation-dispersal dynamic that's typical of WT smooth colony MAB in rich medium (Fig. S2) [16,17].The constitutive aggregation phenotype seen in rough colony variants (Fig. 1) was absent. Rough colony isolates are also defective in sliding motility on soft agar plates [[31], [32]], but both S2 and SNG3 showed comparable sliding motility to ancestor 0253a (Fig. S3), further suggesting that these mutants are not traditional rough colony variants.

We conducted aggregation assays of SNG3 and S2 in SCFM1 with albumin to verify that the aggregation increase in TYEM was consistent under CF sputum-mimicking conditions. Indeed, both S2 and SNG3 showed considerably higher aggregation compared to ancestor in SCFM1 + albumin (Fig. 3A). Both strains also showed higher aggregation than ancestor in SCFM1 without supplemental albumin, indicating that albumin was not required for the effect (Fig. S4).

Fig. 3.

Mutations in genes coding for IclR-like putative transcriptional regulators MAB_0812 and MAB_0813c increase aggregation in MAB 0253a.

A) Aggregation dynamics of evolved populations from the SCFM1 2 (S2) and SCFM1 no glucose 3 (SNG3) lineages in SCFM1 with 5 mg/mL albumin supplement. The OD₆₀₀ measurements of planktonic and aggregated cells were taken at time points of 36-, 40-, 44-, 48-, 60-, 64-, and 68-h post-inoculation. The line on the chart represents the mean OD₆₀₀ values at each time point. B) Colony morphologies of the MAB_0812 (S2) and MAB_0813c (SNG3) evolution mutants alongside their respective complement strains on TYEM agar plates. C-D) Aggregation dynamics of the MAB_0812 (C) and MAB_0813c (D) evolution mutants and their respective complement strains in SCFM1 with 5 mg/mL albumin supplement. For MAB_0812 mutant lineage, OD₆₀₀ measurements of planktonic and aggregated cells were taken at time points of 36-, 40-, 44-, 48-, 60-, 64-, and 68-h post-inoculation. For MAB_0813c mutant lineage, the time points were 28-, 32-, 36-, 40-, 48-, 52-, and 56-h post-inoculation. The line on the chart represents the mean OD₆₀₀ values at each time point.

2.3. Identification of emergent mutations that arose during the evolution experiment

To understand the genetic basis for the new colony biofilm/aggregation phenotypes among SCFM1-evolved populations, whole-genome sequencing (WGS) and variant calling analysis were performed on all passage 15 evolved populations and the ancestor strain [33]. Six mutations reaching frequencies >50 % that were neither present in the ancestor nor in the TYEM-evolved strains were identified and are listed in Table S2. Mutations that were present in the TYEM-lineages are included in Table S3. Interestingly, S2 and SNG3, which had the highest aggregation in TYEM (Fig. 2C), had frameshift mutations in MAB_0812 (S2) and MAB_0813c (SNG3), each of which codes for a putative IclR family transcriptional regulator. IclR-like regulators have an N-terminal HTH DNA-binding domain and an effector-binding C-terminal domain, and they can be both activators and repressors of gene expression [34]. MAB_0812 and MAB_0813c are adjacent to each other on the genome but are convergently transcribed (Fig. 2D). Allelic variants of both genes are common among M. abscessus isolates from various sources, including in isolates from people with CF (Fig. 2D). The canonical IclR, encoded by E. coli, is a transcriptional repressor of the icl operon, which is comprised of genes involved in the glyoxylate cycle [[34], [35], [36]]. Other described IclR family members regulate various functions in a wide array of bacterial species, including biofilm formation control in the CF pathogen Burkholderia cenocepacia [[37], [38], [39]].

2.4. Complementation of MAB_0812 and MAB_0813c restores ancestor phenotypes

Both S2 and SNG3 had multiple mutations not present in ancestor or the TYEM-evolved strains (Table S2). Therefore, to test whether the changes in aggregation dynamics were due to the mutations in the IclR-like-regulators MAB_0812 and MAB0813c, S2 and SNG3 were complemented with a pSD5hsp60 expression vector carrying WT copies of those genes. Complementation of MAB_0812 and MAB_0813c restored the colony morphology to a similar appearance as ancestor (Fig. 3B). Additionally, the aggregation dynamics of the complemented strains were evaluated with aggregation assays in SCFM1 with albumin (Fig. 3C & D). Indeed, complementation of MAB_0812 and MAB_0813c mutants restored aggregation dynamics to that of the WT ancestor. These results suggest that mutations in MAB_0812 and MAB_0813c were responsible for the changes in aggregation dynamics of the evolution populations.

2.5. Genomic analysis of MAB_0812 and MAB_0813c

To further characterize MAB_0812 and MAB_0813c, genomic and phylogenetic analyses of both transcriptional regulators were performed. MAB_0812 is prevalent among Mycobacteria spp. and conserved across Actinomycetota whereas MAB_0813c is present in a smaller subset of Actinomycetota, with 49 % of MAB_0813c alleles restricted to M. abscessus (Figs. S5 and S6). Within MAB, there are 67 alleles of MAB_0812, with CF alleles not clustering in any single clade (Fig. S7). There are 75 alleles of MAB_0813c within MAB, and CF alleles are similarly distributed across clades (Fig. S8). Synteny of MAB_0812 and MAB_0813c is partially conserved across MAB genomes. The average amino acid identity (AAI) between MAB 0253a MAB_0812 and alleles from CF ranged from 24 % to 100 %. In contrast, the AAI of MAB_0813c and alleles from CF ranged from 98.50 % to 99.60 %, suggesting that MAB_0813c is more conserved than MAB_0812 within CF isolates. Since both of our S2 and SNG3 strains contained frameshift mutations, we assessed the lengths of the MAB alleles in comparison to the lab strain alleles to determine if truncations are common in nature. 26.87 % of MAB alleles of MAB_0812 were shorter than the MAB 0253a strain allele, while 20.00 % of MAB alleles of MAB_0813c were shorter than the MAB 0253a strain allele. Interestingly, there were no truncations observed in the CF alleles (Fig. S9).

2.6. MAB_0812 and MAB_0813c both repress an operon with genes involved in the early TCA cycle and glutamate metabolism

Since MAB_0812 and MAB_0813c are both putative transcriptional regulators, we expected that we could glean insight into their regulons with transcriptomic analysis. We therefore performed RNA-seq of our ancestor strain, S2, and SNG3 during aggregated growth in SCFM1 with albumin at 34 h post-inoculation. Fig. 4A shows genes with a ≥ 5-fold difference in expression in either S2 or SNG3 compared to ancestor. It was immediately apparent that these lists largely overlap and are concentrated in a gene cluster immediately upstream of MAB_0812 and MAB_0813c (MAB_0815-MAB_0824) (Fig. 4A & B). All members of this gene cluster were upregulated in both mutants compared to ancestor, suggesting that both MAB_0812 and MAB_0813c likely repress expression of those genes in WT MAB (Fig. 4A & B). Members of the MAB_0815-MAB_0824 gene cluster have various putative functions, but most revolve around fatty acid catabolism (MAB_0816, MAB_0822, MAB_0824) or glutamate, AKG, and gamma-aminobutyric acid (GABA) metabolism (MAB_0815, MAB_0817, MAB_0820c, and MAB_0821). For example, MAB_0817 is a Gamma-aminobutyraldehyde dehydrogenase, which converts the precursor GABA aldehyde into GABA [40,41]. MAB_0815, a putative glutamate/gamma-aminobutyrate antiporter similar to GadC permease, could export GABA in exchange for glutamate, which may be broken down into AKG and ammonium by MAB_0820c, an NAD-glutamate dehydrogenase [[42], [43], [44]]. Likewise, MAB_0822 is an acyl-CoA dehydrogenase (ACAD), which initiates the first step of β-oxidation, a metabolic process that breaks down fatty acids into acetate/acetyl-CoA [45]. MAB_0824, a CoA transferase, likely assists in the process of acetyl-CoA generation and fatty acid transport [46,47]. Finally, MAB_0821, a GABA transaminase, converts GABA and AKG into succinic semialdehyde, a succinate precursor, and glutamate [48].

Fig. 4.

MAB_0812 and MAB_0813c repress a gene cluster involved in the TCA cycle and glutamate metabolism. A) Heat map showing differentially expressed genes (DEGs) in S2 and SNG3 mutants comparied to the WT ancestor. DEGs from either S2 or SNG3 with ≥ 5-fold changes in gene expression are shown. MAB_0812 (mraA), MAB_0813c (mraB), and MAB_0811 (bracketed) are shown for reference and did not meet the ≥ 5-fold cutoff. MAB_0812 (mraA) and MAB_0813c (mraB) are highlighted in red text and denoted with ∗. The full RNA-seq results can be found in the Supporting Information. B) Gene map of MAB_0811-MAB_0824. The color corresponds to gene expression fold changes in A. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

A holistic examination of the MAB_0815-MAB_0824 gene cluster reveals that their functions mostly affect one specific region of metabolism, the pathway between acetyl-CoA and GABA. Fatty acid β-oxidation generates acetyl-CoA, which condenses with oxaloacetate to form citrate in the first step of the TCA cycle. Citrate is isomerized to isocitrate, which undergoes oxidative decarboxylation to produce AKG, which can then undergo reductive amination to form glutamate. Glutamate can subsequently be decarboxylated to yield GABA. While the RNA-seq data was successful in focusing us on this area of metabolism, gene annotation information cannot always reliably predict the direction of metabolic flow. Therefore, to understand how the MAB_0812 and MAB_0813c regulons affected the intracellular concentrations of metabolites within these pathways, we turned to targeted metabolomic analysis.

2.7. TCA cycle intermediates citrate/isocitrate and α-ketoglutarate increase in S2 and SNG3 and positively correlate with MAB aggregation

We performed targeted metabolomics of MAB 0253a ancestor, S2 (MAB_0812), and SNG3 (MAB_0813c) grown in SCFM1 with albumin at 32 (SNG3) and 34 (S2) hours post-inoculation, time points approaching peak aggregation but still in active growth. Metabolites were extracted and analyzed using liquid chromatography-high resolution mass spectrometry (LC-HRMS) (Fig. 5A and Fig. S10). Interestingly, the TCA cycle intermediates citrate/isocitrate (could not be differentiated in our methodology) and AKG were the only metabolites we analyzed that were significantly elevated in both S2 and SNG3 compared to the ancestor. Glutamate and glutamine levels did not significantly change between ancestor and either mutant; GABA level significantly rose in SNG3, albeit with high variation, but not in S2 (Fig. S10). Other TCA cycle intermediates, including fumarate and succinate, did not show a change in either S2 or SNG3 compared to WT, and malate showed a small decrease in only SNG3 compared to ancestor. We found the increase of citrate/isocitrate and AKG in conjunction with an increase in aggregation compelling for three reasons. For one, it agrees with our RNA-seq data in that both metabolites are part of the identified metabolic pathway likely regulated by MAB_0812 and MAB_0813c. Secondly, both of these metabolites are common flux-dependent sensors in bacteria; their intracellular concentrations increase with higher flux through the early part of the TCA cycle and these fluctuations can be sensed by various effectors, including transcriptional regulators, thereby altering cellular physiology [49]. Lastly, while it is not yet known how carbon availability is sensed and how it affects NTM aggregation, a prior transposon screen in M. smegmatis indicated that enzymes affecting both AKG and citrate metabolism may affect aggregation [17].

Fig. 5.

TCA cycle intermediates citrate/isocitrate and α-ketoglutarate positively correlate with MAB aggregation.

A) Graphical overview of the TCA cycle including the glyoxylate bypass. Quantification of intracellular citrate/isocitrate, AKG, succinate, fumarate, and malate pools in Ancestor (Anc), S2, and SNG3 evolution lineages are shown relative to total mg of cellular protein (assessed by a BCA assay). All cells were grown in SCFM1 with 0.5 mg/mL albumin supplement. Metabolite extraction was performed at 32 h (S2) or 34 h (SNG3) post-inoculation. P-values were obtained using an unpaired Student's T-Test. ∗ denotes P-value <0.05. ∗∗∗ denotes P-value <0.001. B) Metabolomic quantification of intracellular citrate/isocitrate and AKG pools in S2 and SNG3 mutants and complemented strains. Culture conditions and metabolite extraction times were unchanged. P-value obtained by unpaired Student's T-Test. ∗∗ denotes P-value <0.01. ∗∗∗ denotes P-value <0.001. ∗∗∗∗ denotes P-value <0.0001.

To confirm that the increase in citrate/isocitrate and AKG was connected to mutations in MAB_0812 and MAB_0813c, the same targeted metabolomic analysis was repeated with the SNG3 and S2 alongside their respective complement strains (Fig. 5B). When both strains were complemented, the citrate and AKG pools were restored to levels comparable to those of the ancestor. Altogether, these results indicate that flux through citrate/isocitrate and AKG is controlled by MAB_0812 and MAB_0813c, likely via the cluster of genes identified in our RNA-seq experiment, and that this flux can affect MAB aggregation. Given their roles in the regulation of aggregation through metabolic control, we therefore named MAB_0812 and MAB_0813c as Metabolic Regulator of Aggregation A and B (mraA and mraB) respectively.

2.8. Acetate increases AKG and citrate/isocitrate pools and induces MAB biofilm formation, independently of the MraA/B regulon

If the increase in AKG and citrate/isocitrate in S2 and SNG causes an increase in aggregation in MAB, we hypothesized that we should be able to mimic that effect in the ancestor strain by modifying available nutrients to increase flux through those two metabolite pools. To test this, we supplemented SCFM1, which contains 3 mM glucose, with either additional glucose or acetate, normalized to total carbon. Acetate is converted to acetyl-CoA, one of the initial reactants of the TCA cycle. Acetyl-CoA condenses with oxaloacetate to form citrate, which is isomerized to isocitrate and then forms AKG (Fig. 5A). By supplementing with 30 mM acetate (78 mM total carbon), we aimed to increase carbon flow through citrate/isocitrate and AKG. Indeed, targeted metabolomic analysis of MAB 0253a ancestor cultures in SCFM1 supplemented with acetate showed a significant increase in both citrate/isocitrate and AKG levels compared to MAB 0253a grown in SCFM1 supplemented with an additional 10 mM glucose (78 mM total carbon) (Fig. 6A & B).

Fig. 6.

Addition of acetate to SCFM1 induces MAB aggregation and biofilm formation.

A-B) Metabolomic quantification of intracellular citrate/isocitrate (A) and AKG pools (B), normalized to mg of total cellular protein, in MAB 0253a grown in plain SCFM1 and SCFM1 supplemented with either 10 mM glucose or 30 mM acetate. Metabolite extraction was performed at 44 h post-inoculation. P-values were obtained with an unpaired Student's T-Test. ∗∗∗ denotes P-value <0.001. C) Aggregation dynamics of MAB 0253a in plain SCFM1 and SCFM1 supplemented with either 10 mM glucose or 30 mM acetate. OD₆₀₀ measurements of planktonic and aggregated cells were taken at time points of 36-, 40-, 44-, 48-, 60-, 64-, and 68-h post-inoculation. The line on the chart represents the mean OD₆₀₀ values at each time point. D) Aggregation dynamics of MAB 0711a under the same conditions. Time points were taken at 36-, 40-, 44-, 48-, 60-, 64-, and 68-h post-inoculation. E) Pellicle formation of MAB 0253a and MAB 0711a grown in SCFM1 supplemented with 10 mM glucose or 30 mM acetate. Pellicle photos were taken after 4 (MAB 0711a) or 5 (MAB 0253a) days of incubation at 37 °C.

Consistent with our overall hypothesis that higher citrate/isocitrate/AKG increases aggregation, MAB 0253a aggregated in SCFM1 supplemented with acetate but did not aggregate in SCFM1 supplemented with glucose (Fig. 6C). Acetate increased MAB 0711a aggregation in SCFM1, and MAB 0711a maintained the aggregated state longer than MAB 0253a, never dispersing even at 68 h (Fig. 6D). Similar acetate-induced aggregation was also observed in MAB GD91, a smooth colony variant M. abscessus subsp bolletii clinical isolate (Fig. S11). Altogether, our results suggest that acetate addition to SCFM1 is a strong inducer of aggregation, and that this increase in aggregation is achieved by increasing intracellular citrate/isocitrate and AKG levels.

We next wanted to test whether the increase in citrate/isocitrate and AKG levels, and the subsequent aggregation, in the SCFM1 plus acetate condition was dependent on the induction of the MraA/B regulon. Therefore, we performed RNA-seq on MAB 0253a grown in SCFM1 plus acetate and SCFM1 plus glucose during peak aggregated growth at 44 h post-inoculation (Fig. S12). The resulting heat map suggested a contrasting transcriptomic landscape compared to MAB S2 and SNG3 (Fig. 4A). DEGs with ≥ 5-fold change were largely uncharacterized hypothetical proteins, and we observed no induction of the MraA/B gene cluster (MAB_0815-MAB_0824) in SCFM1 plus acetate, nor were mraA and mraB induced. These observation suggest that the increases in citrate/isocitrate and AKG in SCFM1 plus acetate were independent of the MraA/B regulon. Moreover, they suggest that the increase in citrate/isocitrate and AKG, not other effects of MraA/B induction, is responsible for driving MAB aggregation.

The cell-cell aggregation intrinsic to our liquid aggregation assay is an essential step in early biofilm development, so we expect that changes we observe with our aggregation assay will also manifest in other biofilm models that incorporate later stages of biofilm development [16,17,50]. We therefore tested MAB pellicle biofilm development in SCFM1 supplemented with 30 mM acetate or 10 mM glucose (Fig. 6E). MAB 0253a developed pellicles when grown in SCFM1 plus acetate but failed to initiate pellicle development in SCFM1 plus glucose. MAB 0711a developed pellicles under both growth conditions, with the acetate-induced pellicles being considerably more robust and structured. These results mirror our liquid aggregation assays with these strains, providing further evidence that the cell-cell adhesion we measure in liquid aggregation assays is an important step in biofilm development.

2.9. Nitrogen counters acetate-induced aggregation

Increasing nitrogen availability in rich medium favors planktonic growth in both M. smegmatis and MAB [16,17]. To understand whether nitrogen similarly affects MAB under CF sputum-mimicking conditions, aggregation was assessed for MAB 0253a cultures grown in SCFM1, SCFM1 with 30 mM acetate, and SCFM1 with 30 mM acetate plus 15 mM ammonium chloride (Fig. 7A). Ammonium ablated aggregation in MAB 0253a and eventually facilitated the early dispersal of MAB 0711a and MAB GD91 aggregates even in the presence of acetate (Fig. 7B & S11). Interestingly, the evolution mutants MAB S2 and SNG3, which aggregate more and accumulate intracellular citrate/isocitrate and AKG in SCFM1 (Fig. 3, Fig. 5), can be forced into early dispersal through ammonium supplementation (Fig. 7C). In contrast, the rough variants MAB 0253b and 0711b remained aggregated even with addition of ammonium. (Fig. S13). Overall, it appears that nitrogen availability is capable of dispersing smooth colony MAB strains in SCFM1 as it does in rich medium [16]. The slight difference in how MAB 0253a, MAB 0711a, and MAB GD91 react to both acetate and ammonium indicates strain-specific variation in MAB aggregation dynamics. MAB S2 and SNG3 dispersing in response to ammonium indicates that, unlike rough colony variants, the metabolic shifts responsible for an increase in aggregation in those strains can be chemically countered.

Fig. 7.

Addition of nitrogen to SCFM1 induces dispersal.

Aggregation dynamics of MAB 0253a (A) and 0711a (B) in plain SCFM1, SCFM1 plus 30 mM acetate, and SCFM1 plus 30 mM acetate and 15 mM ammonium chloride. OD₆₀₀ measurements of planktonic and aggregated cells were taken at time points of 36-, 40-, 44-, 48-, 60-, 64-, and 68-h post-inoculation. The line on the chart represents the mean OD₆₀₀ values at each time point. C) Aggregation dynamics of S2 and SNG3 in SCFM1 plus albumin with or without 15 mM ammonium chloride. Time points were taken at 28-, 32-, 36-, 40-, 48-, 52-, and 56-h post-inoculation. MAB 0253a ancestor was included as a control.

2.10. Acetate-induced aggregation confers antibiotic tolerance

In the context of CF and other respiratory infections, bacterial communities are often embedded in a semi-solid matrix composed, at least in part, of mucins, eDNA, and host cells [50,51]. To understand how acetate-induced aggregation impacts antibiotic sensitivity in a similar physical environment, we utilized the agar block biofilm assay (ABBA), in which soft-agar-suspended MAB 0253a cells grow as embedded aggregates [52,53] (Fig. 8A). MAB communities in both glucose- and acetate-supplemented SCFM1 displayed similar distribution and sizes (Fig. 8B). Closer observation with a 40× objective revealed a distinction in aggregate morphologies; the acetate-supplemented aggregates showed a slightly more-spindled morphology compared to the smoother glucose-supplemented aggregates (Fig. 8B). Despite their similar size, antibiotic sensitivity tests revealed that acetate-supplemented aggregates displayed consistently higher survival after treatment with increasing concentration of amikacin and clarithromycin (Fig. 8C). As a comparison, traditional minimal inhibitory concentration (MIC) testing with MAB 0253a demonstrated similar susceptibilities against both amikacin and clarithromycin after three days of incubation (Table S11), suggesting aggregation in the ABBA is a key factor in conferring drug tolerance. Altogether, our results demonstrate that selective pressures from the CF environment can impact carbon flux, and that these simple metabolic redirections can have dramatic effects on MAB biofilm formation as well as drug tolerance.

Fig. 8.

Acetate-induced aggregation confers antibiotic tolerance to agar-suspended MAB.

A) Schematic of the agar block biofilm assay (ABBA). B) Confocal microscopy images of MAB aggregates in the ABBA with either SCFM1 plus 10 mM glucose or 30 mM acetate, viewed under an 10X or 40X objective. Aggregates are labelled green with the SYTO™ 9 Green Fluorescent Nucleic Acid Stain. C) Antibiotic sensitivity of MAB aggregates in the ABBA with either SCFM1 plus 10 mM glucose or 30 mM acetate. ABBA samples were treated with amikacin (AMK) at concentrations of 0, 10, 15, and 25 μg/mL or clarithromycin (CAM) at concentrations of 0, 25, 50, and 100 μg/mL at 48 h post-inoculation. Antibiotic sensitivity was represented as percentage of cells that survived antibiotic treatment. P-value obtained by unpaired Student's T-Test. ∗ denotes-value <0.05. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3. Discussion

The disconnect between in vitro and in vivo drug sensitivity, coupled with the in situ observation of MAB biofilm formation, suggests that MAB biofilms may form during infection and increase drug tolerance [7,12]. Dispersal of MAB biofilms in vitro renders them more drug tolerant [15], and disruption of biofilms in combination with drug delivery can be an effective treatment strategy to counter chronic infections in animal models [54], so insight into biofilm regulation has the potential to lead to more effective MAB treatments. Biofilm development involves individual cells aggregating to each other and/or a surface, followed by late-stage 3-D biofilm maturation [50,55]. Because the spontaneous aggregation of mycobacteria in most media has masked the planktonic to aggregate transition, most NTM biofilm studies have focused on the maturation process [[55], [56], [57], [58]]. For example, the chaperone protein GroEL1 modulates synthesis of M. smegmatis mycolic acids, a major component of NTM biofilms, and is involved in biofilm development [55,56]. In addition, iron is essential for late-stage maturation of M. smegmatis biofilms [55,57]. Here, we build off recent studies showing that NTM aggregation is not constitutive and that planktonic cultures of M. smegmatis and MAB can be achieved in detergent-free medium by tuning carbon and nitrogen availability [16,17]. We performed our MAB evolution experiment in SCFM1 with the expectation that it would reveal insight into the ambiguous regulatory networks controlling MAB aggregation and, by extension, biofilm formation, in addition to helping us understand how these networks can change with selective pressure from the infection environment chemistry.

Research on biofilm formation in pathogens associated with chronic infections often focuses on isolates with dramatically altered matrix or adhesion production. For example, in P. aeruginosa, the shift from WT to mucoid strains is a result of the constitutive expression of the matrix component alginate, leading to an excessively goopy and very notable phenotype on agar plates [20]. Similarly, in MAB, the smooth-to-rough colony variant transition involves a dramatic colony morphotype and complete or near-complete loss of surface GPLs [21,59]. Other surface molecules, specifically arabinogalactan and lipoarabinomannan (LAM), are also commonly mutated in isolates from people with CF, and these mutations can have profound impacts on biofilm formation and host interactions [60,61]. However, just as mucoidy itself does not confer in vitro drug tolerance to P. aeruginosa [18], the transition to a rough morphotype does not by itself confer in vitro drug tolerance to MAB [62]. Instead, both smooth and rough MAB strains become more drug tolerant when they form biofilms in vitro, as shown in prior studies that used Tween 80 to grow planktonic cultures as comparators [11,63]. Therefore, if biofilm formation confers drug tolerance specifically during infection, there must be MAB regulatory systems in place, in WT, smooth colony strains at least, that respond to features of the infection environment to drive in vivo biofilm development. Through artificial evolution of MAB in SCFM1, we provide insight into what those environmental signals are and reveal a central piece of the regulatory system controlling MAB aggregation and by extension biofilm initiation.

Mycobacteria co-catabolize carbon sources with no obvious import preference for any given nutrient [64,65]. As a broad hypothesis, we predict that two IclR-like regulators, MraA and MraB, redirect carbon generated from catabolic pathways away from the early TCA cycle intermediates citrate/isocitrate and AKG. The limited flux through those intermediates then prevents MAB aggregation in SCFM1. While many IclR-like regulators have been described, the role of IclR-like regulators in mycobacteria, and specifically in NTM pathogenesis, is poorly understood [34,66]. The select few that have been characterized are mostly in the context of M. tuberculosis [[67], [68], [69]]. An example of which includes Rv2989, which regulates the leuCD operon involved in leucine biosynthesis [67,68]. MAB_0813c has been previously identified as one of many transcriptional regulators actively expressed during late-stage MAB biofilm formation, although its exact role was not clear [70]. In this study, we utilized transcriptomic and metabolomic analysis to identify MraA/B's function as regulators of multiple metabolic pathways that affect the metabolism of fatty acids, glutamate, GABA, semi-aldehyde/aldehyde, carboxylic acids, and TCA cycle intermediates citrate/isocitrate and AKG. RNA-seq analysis of the MraA/B regulon further suggested that both regulators control to some degree the same regulon, as evidenced by increased expression of MAB_0815-MAB_0824 in correlation with the loss of MraA/B function. This raises questions about the purpose and necessity of having two regulators with similar functions. Also interesting and unclear is the suggestion from our results that neither regulator seems to be sufficient on its own to suppress the shared regulon. ChIP-seq analysis will be important in identifying the regulatory binding sites of MraA/B, allowing for a deeper understanding of MraA and MraB's respective regulatory dynamics. Given the importance of the MraA/B regulon in affecting MAB aggregation, it could also potentially reveal new therapeutic targets. Future work will therefore entail a detailed investigation into the metabolic networks controlled by the MraA/B regulon, specifically in identifying the directionality of metabolic flow and its end products.

Genomic analysis of mraA (MAB_0812) and mraB (MAB_0813c) revealed that both regulators were present in a wide range of mycobacterial species, with mraB having a significant predominance in MAB, including multiple MAB clinical isolates from human respiratory samples. Allelic variation of mraA and mraB were found among MAB respiratory isolates and isolates from environmental reservoirs, suggesting that alteration of these regulators could promote persistence of MAB in diverse environments. When compared to the mutations found in other MAB respiratory isolates, which were mostly missense mutations, the mutants identified in our evolution experiment stood out as both were frameshift mutations. The implication of either type of mutation on regulator functions is yet unclear, although because complementation with the intact gene restored function, the frameshift mutations we acquired are likely deleterious. This suggests a fitness advantage of overexpression of the MraA/B regulon in vitro in SCFM1. Future work will therefore explore the evolutionary history of mraA and mraB, including exploring how mutations common in respiratory isolates impact function.

SCFM1 represents an average concentration of nutrients – each one can vary between sputum samples and more broadly between people with CF [26,71]. In this study, nutrient supplements, including glucose, ammonium, acetate, and albumin, were maintained at levels previously observed directly in CF sputum or, in the case of acetate, other sputum-like media [[71], [72], [73], [74]]. Future studies will aim to identify the specific factors in SCFM1 that provided selective pressure for our mraA/B mutations. Furthermore, understanding how varying other SCFM1 components, including cations such as magnesium, within the bounds of direct sputum measurements impact MraA/B activity, MAB aggregation, AKG and citrate/isocitrate levels, and antibiotic tolerance will provide a more complete picture of biofilm regulation in the infection environment [75]. The activity of the glyoxylate shunt in MAB grown in SCFM1 is also in our purview. In many bacteria, including E. coli and M. tuberculosis, acetate supplementation triggers the glyoxylate shunt, which bypasses AKG and leads to an increase in succinate [35,65]. Our finding that both citrate/isocitrate and AKG increased in WT after acetate supplementation suggests that MAB may regulate the glyoxylate shunt differently than most bacteria, at least in the context of SCFM1. Finally, repeating evolution and phenotypic experiments with the more complex SCFM2 formula and with a more diverse MAB strain repertoire could provide a more generalizable conclusion on MAB adaptation to CF-specific/mimicking environments.

AKG and citrate are both well-documented flux-dependent sensors [49,76,77]. AKG can serve as an indicator of carbon availability and, alongside Gln, is a known regulatory node for nitrogen availability in bacteria [[77], [78], [79]]. In S. aureus, citrate binds and activates the transcriptional regulator CcpE, modulating the expression of genes involved in central carbon metabolism, iron uptake, and virulence [76]. Therefore, follow-up studies will aim to identify the regulators that modify aggregation through AKG/citrate/isocitrate availability and will clarify if any of these metabolites are dominant compared to others. One regulator of interest is the PII protein (GlnK in mycobacteria) [77,80]. PII proteins can directly bind to ATP, ADP, and AKG [77,81]. These allosteric interactions change the PII protein structure, leading to downstream regulatory response pathways. Since AKG can fluctuate with both carbon and nitrogen availability, and since glutamine levels can also affect PII activity, PII proteins sense and can impact both carbon and nitrogen metabolism [77,81]. Indeed, our finding that ammonium addition can counter acetate-induced aggregation would fit well into a model wherein both signals converged on PII.

Biofilm formation confers drug tolerance by both erecting a physical barrier against antibiotic penetration and by allowing development of cellular subpopulations with altered metabolic states that limit drug effectiveness [1,9,10,15,82]. In our ABBA experiments, the survival of some percentage of the acetate-supplemented MAB population contrasted sharply with the rapid collapse of the glucose-supplemented population, demonstrating a drug-tolerant state induced by acetate. In our planktonic MIC tests, the lack of difference in susceptibility between the acetate- and glucose-supplemented MAB provides support for the hypothesis that aggregation is driving tolerance in the ABBA. For a non-motile pathogen growing in an embedded environment, the demarcation between biofilm formation and physically constrained planktonic growth becomes hazy and interesting. The morphological changes observed between acetate and glucose-grown aggregates suggest that matrix production differs between the two conditions. This may include altered surface structures such as GPL, LAM, trehalose dimycolate (TDM), mycolic acids, or other characterized or uncharacterized components [1,27,54,60,83]. Determining which surface molecules drive liquid aggregation in response to acetate in WT MAB, and drive drug tolerance in the ABBA, will be the goal of future studies. The loss of GPLs in rough colony variants presumably makes those strains physically incapable of dispersal, even if the regulatory systems driving that dispersal are intact in the rough colony variants. Elucidating those regulatory systems is therefore best accomplished with investigation into WT (smooth colony variant) MAB strains. The differences between our evolved S2 and SNG3 strains and the rough colony variant 0253b in terms of aggregation dynamics, colony morphology, sliding motility, and responsiveness to ammonium suggests that GPLs can still be produced by our evolved strains. Whether regulated changes to GPL abundance or GPL modifications are responsible for controlling aggregation and dispersal in WT MAB, or whether fluctuations of other surface molecules drive those dynamics, will likewise be the focus of future work.

Overall, several key conclusions can be drawn from this study. First, through artificial evolution, we demonstrated that passage in SCFM1 is sufficient to drive the emergence of a pro-aggregation MAB genotype/phenotype. Second, we found that the pro-aggregation phenotype was the result of mutations in two novel IclR-like regulators, MraA and MraB, which control the size of intracellular citrate/isocitrate and AKG pools through repression of a complex genetic/metabolic network. Third, we identified that acetate induces the formation of antibiotic-tolerant biofilms in WT MAB by increasing flux through citrate/isocitrate and AKG. Future studies will probe various aspects of this system. These include investigating which specific MraA/B regulon components influence changes in citrate/isocitrate and AKG flux, understanding the intracellular cues that control activity of each regulator, investigating how a shift in citrate/isocitrate and AKG flux facilitates the downstream shift in MAB aggregation, understanding which adhesins change in abundance in response to TCA cycle flux, and directly assessing these activities in in situ biofilms. We hope that future work into the specific regulatory nodes that control the MAB planktonic:aggregate transition will aid in the development of novel anti-biofilm compounds that could be used in synergy with traditional antibiotic treatments.

4. Materials and methods

4.1. Bacterial strains and growth conditions

All bacterial strains and plasmids used in this study are summarized in Table S4 in the supplementary material. All primers are listed in Table S5. For general culturing, MAB was grown in TYEM plus 0.05 % Tween 80 for 48–72 h at 37 °C, with shaking at 250 RPM [16]. When performing aggregation assays, ABBA, and pellicle formation assay, Tween 80 was excluded from TYEM. For indicated aggregation assays and the evolution experiment, MAB was grown in Synthetic Cystic Fibrosis Medium (SCFM1). The composition of SCFM1 was the same as described in Palmer et al., 2007 with some modifications [26]. First, the final concentration of Morpholinepropanesulfonic acid (MOPS) was increased from 10 mM to 20 mM in order to increase the buffer capacity. Second, NH₄Cl was removed from SCFM1 in order to separately assess the effects of nitrogen on MAB aggregation. Additional nutrients, including NH₄Cl, glucose, acetate, and others, were supplemented to SCFM1 in the indicated amount. For the evolution experiment and indicated aggregation assays, bovin serum albumin was added to a final concentration of 5 mg/mL in SCFM1 [29,72]. The amount of albumin added was based on the typical amount used in BD BBL™ Middlebrook OADC Growth Supplement (Fisher Scientific) when supplemented at a 1:10 ratio to the Middlebrook 7H9 broth, a common media for mycobacteria cultivation. The amount of albumin added was reflective of the amount used in previous synthetic sputum medium iterations from the literature [72].

4.2. in vitro aggregation assay

Aggregation assays were performed as described previously, with some modifications [16]. During the preparatory stage, SCFM1 or TYEM was prepared in a 250 mL flask. Filter-sterilized nutrient supplements were added to the medium at the indicated concentrations. MAB was diluted to an OD₆₀₀ of 1 and added to the culture flask at a 1:100 dilution factor. The resulting culture was thoroughly mixed and 5 mL aliquots were pipetted into sterile borosilicate disposable culture tubes. The cultures were incubated at 37 °C with shaking at 250 rpm. Aggregates were harvested at the indicated time points by pouring the cultures through a 10 μm cell strainer, thereby separating the aggregates from planktonic cells. The OD₆₀₀ of the planktonic fraction was immediately recorded. The original culture tube was then washed with 5 mL of 1x PBS and poured through the filter a second time to wash off any remaining planktonic cells. Aggregates collected by the strainer were washed off with 5 mL of 1x PBS plus 6 % Tween 20 back into the original culture tube. 500 μL of Tween 20 was added to the aggregate fraction and resuspended through vertexing. Aggregate fractions were then water bath sonicated until no visible clumps remained, and the OD₆₀₀ of the aggregate fraction was recorded. At least three technical replicates were harvested at each time point.

4.3. Cloning and plasmid construction

Escherichia coli was grown in Lysogeny broth (LB) medium and the cells were grown for 16–24 h at 37 °C, shaking at 250 RPM. 1.7 % agar (Fisher Bioreagents) was used for both LB and TYEM agar plates. The antibiotic kanamycin (Kan) was supplemented to TYEM as indicated, at 125 μg/mL to select for or maintain pSD5hsp60 plasmids in MAB. For E. coli, the amount of Kan was decreased to 50 μg/mL. No antibiotic was supplemented to the media during aggregation assays or the evolution experiment.

Genes of interest were first amplified from MAB 0253a genomic DNA using PCR. All PCR reactions were performed with either OneTaq 2x Master Mix with Standard Buffer (New England Biolabs) or Q5 High-Fidelity 2x Master Mix (New England Biolabs). Correct fragment size of the PCR product was confirmed using an Invitrogen™ E-Gel™ Power Snap Electrophoresis System with E-Gel EX 1–2 % agarose gels (Thermo Scientific™). Purification of PCR products was done using the Monarch PCR & DNA Cleanup Kit (New England Biolabs), and the DNA concentration was determined with a NanoDrop™ One/OneC Microvolume UV–Vis Spectrophotometer (Thermo Scientific™) in ng/μL. pSD5hsp60 vectors were extracted from E. coli DH10B cells using the Monarch Plasmid Miniprep Kit (New England Biolabs). Restriction enzyme digestion was performed on pSDhsp60 vector using NdeI restriction enzyme (New England Biolabs), which linearized the vector. The digested product was purified again using the Monarch PCR & DNA Cleanup Kit. The PCR product was then ligated into the linearized vector by Gibson assembly using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs). The resulting plasmid was then either transformed into chemically competent E. coli DH10B cells or stored at −20 °C for later use. All molecular genetic analysis, cloning visualization, primer design, and sequence alignments performed using the SnapGene software (Dotmatics).

For transforming plasmids into MAB cells, a stationary MAB culture was inoculated into 100 mL of TYEM plus Tween 80 media, with an initial OD₆₀₀ of 0.02–0.04. The culture was then incubated at 37 °C with shaking until the OD₆₀₀ reached 0.5–0.9. The culture was then chilled on ice or in a 4 °C cold room for 4–24 h. Cells were then spun down in a centrifuge at 6000×g for 10 min at 4 °C. The resulting pellets were progressively washed with 50, 25, 10, and 5 mL of ice-cold glycerol wash buffer (250 mM d-glucose, 500 μM potassium acetate, and 10 % glycerol). During resuspension, the pellet was broken up using a 10 mL syringe with an 18g blunt-head needle. The final pellet was resuspended in 1 mL of glycerol wash buffer and 100 μL of electrocompetent cells were aliquoted to pre-chilled Eppendorf tubes. 1 μg or more plasmid DNA was added to the tube and thoroughly mixed by pipetting. The plasmid cell mix was left on ice for 30 min before being transferred to a pre-chilled electroporation cuvette, 0.2 cm gap. The cells were electroporated at 2.5 kV, 25 μF, 1000 Ω with a Bio-Rad Gene Pulser Xcell Electroporation System. Immediately, the cells were added to 1 mL of TYEM plus Tween 80 media and incubated overnight at 37 °C with shaking. The next day, transformant cells were concentrated in the microcentrifuge at 16,000×g for 1 min, and the pellet was resuspended in 100 μL of TYEM plus Tween 80 media. The cells were plated onto a TYEM plate with Kan selection (125 μg/mL) and incubated at 37 °C for 5 days or until colonies start to emerge. Once visible colonies were identified, individual colonies were selected and plated onto new a TYEM plate with Kan selection. For colony PCR, a single MAB colony was inoculated into 50 μL of sterile water in an Eppendorf tube. The cells were heated to 95 °C on a heat block for 10 min. Then, 1 μL of the heated cell mix were taken and used as template for colony PCR in similar manner as described above. The same primers were used for the PCR reaction. PCR products with the correct fragment size were purified and sequenced using the Sanger Sequencing Services at Azenta Life Sciences. The sequences were verified using SnapGene, and successful transformant cells were frozen down at −80 °C.

4.4. Targeted metabolomics

Metabolite extraction was performed as described with modifications [17,65]. MAB cultures were grown as described in the Results section. For each experimental condition, five replicate cultures were set up per time point. Each replicate contained a total of 15 mL of cell culture. At each indicated time point, all 15 mL of culture was poured through a Durapore 0.22 μm PVDF Membrane Filter (MilliporeSigma). The membrane was then washed with 5 mL of 1xPBS, pre-heated to 37 °C. The filter membrane was immediately moved to 2 mL of dry ice-chilled 2:2:1 acetonitrile:methanol:water solution for cell lysis. The cells were scraped off the filter membrane with a cell scraper and collected into a 2 mL bead beating tube from the Precellys Tough Micro-organism Lysing Kit VK05 (Bertin Instruments). The cells were bead-beaten with an Omni International Bead Ruptor 12 machine (4 speed, 3 cycles, 2 min per cycle, 10 s break in between cycles; all done at 4 °C). The resulting cell lysate was pelleted in a microcentrifuge at 20,000 RCF for 5 min at 4 °C. From the supernatant, a 600 μL aliquot was transferred to a Costar Spin-X 0.22 μm Centrifuge Tube Filter (Corning), and the rest was decanted. The supernatant centrifuged using the same setting, and the resulting 600 μL metabolite extract was immediately capped and frozen at −80 °C until metabolomic analysis by LC-HRMS.

The remaining cell lysate pellet was set on a heat block in a chemical fume hood at 55 °C for 2 h to evaporate any remaining acetonitrile:methanol:water solution. Once dried, the pellet was resuspended in 500 μL of 3 % SDS in 10 mM Tris-HCl solution, pH 7.5. The resuspended pellet was bead-beaten again in the same manner as described, and the resulting liquid was left in a 4 °C cold room to settle for 20 min. From the settled liquid, a 1:10 dilution in water was made and the protein concentration within the cell lysate was assessed using the Pierce™ Dilution-Free™ Rapid Gold BCA Protein Assay (Thermo ScientificTM).

Extracted metabolite samples were sent to the University of Pittsburgh Health Sciences Mass Spectrometry Core (MSC), where the intracellular pools of citrate/isocitrate, α-ketoglutarate, succinate, fumarate, malate, glutamine, and glutamate were quantified using LC-HRMS. Briefly, a deuterated internal standard mix that included creatinine-d₃, alanine-d₃, taurine-d₄ and lactate-d₃ (Sigma-Aldrich) was added to the sample lysates at a final concentration of 10 μM. After 3 min of vortexing, the supernatant was cleared of protein by centrifugation at 16,000×g. Cleared supernatant (2 μL) was injected via a Thermo Vanquish UHPLC and separated over a reversed phase Thermo HyperCarb porous graphite column (2.1 × 100 mm, 3 μm particle size) maintained at 55 °C. For the 20-min LC gradient, the mobile phase consisted of the following: solvent A (water/0.1 % FA) and solvent B (ACN/0.1 % FA). The gradient was the following: 1 %B for the first minute increasing to 15 %B over 5 min, followed by an increase to 98 %B over 5 min that was held for 5 min before equilibration at initial conditions for 5 min. The Thermo ID-X tribrid mass spectrometer was operated in both positive and negative ion mode, scanning in ddMS² mode (2 μscans) from 70 to 800 m/z at 120,000 resolution with an AGC target of 2e5 for full scan, and 2e4 for MS² scans using HCD fragmentation at stepped 15,35,50 collision energies. Source ionization was 3.0 and 2.4 kV spray voltage respectively, for positive and negative mode. Source gas parameters were 35 sheath gas, 12 auxiliary gas at 320 °C, and 8 sweep gas. Calibration was performed prior to analysis using the Pierce™ FlexMix Ion Calibration Solution (Thermo Fisher Scientific). Integrated peak areas were extracted manually using Quan Browser (Thermo Fisher Xcalibur ver. 2.7). Targeted metabolite values are reported as the ratio of the analyte to the internal standard before conversion to absolute concentration via calibration curves ranging from 15 fmol/μL to 100 pmol/μL and utilizing total cellular protein values obtained with the BCA assay.

4.5. Evolution experiments and phenotype characterization

To initiate the evolution experiment, a single colony of MAB 0253a was inoculated into 5 mL of TYEM plus Tween 80 broth and grown to stationary phase. The MAB culture was diluted to an OD₆₀₀ of 1 and inoculated at a ratio of 1:100 into test tubes containing 5 mL of SCFM1 (with albumin), SCFM1 with no glucose (with albumin), or TYEM (all without Tween 80). For each condition, three replicates were set up. Each replicate represented a single evolution lineage. The cultures were grown at 37 °C, with shaking at 250 RPM, for exactly 72 h. At 72 h, each culture's OD₆₀₀ was determined and adjusted to 1 in a 1 mL aliquot. From the 1 mL aliquot, the MAB culture was passed at a ratio of 1:100 again into fresh media. This serial passage was done for a total of 15 passages (45 days), with each passage roughly corresponding to 6.64 generations (2.213 generations per day). At each passage, an aliquot of culture from each lineage was frozen down at −80 °C for later phenotypic characterization, whole-genome sequencing, and lab record.

At the end of the evolution experiment (passage 15), the resulting evolution populations were assessed for biofilm-relevant phenotypes. For checking colony morphology, passage 15 populations were cultured in their respective media. The OD₆₀₀ of each culture was adjusted to 1, and a 4 μL aliquot was spotted onto a TYEM agar plate. The agar plate was left to incubate at 37 °C for 4–5 days. The colonies were imaged using the Zeiss Axiocam 305 color camera, viewed under a Zeiss SteREO Discovery.V8 microscope with a Plan S 1.0× Objective. To characterize changes in aggregation dynamics, aggregation assays in both TYEM, SCFM1, and SCFM1 with albumin were performed, as previously described, for all passage 15 evolution populations [16].

4.6. Whole genome sequencing and variant calling of experimentally evolved MAB

To prepare genomic DNA for whole genome sequencing (WGS), all passage 15 MAB strains and the MAB 0253a ancestor were cultured in 50 mL of TYEM plus Tween 80. Cells were concentrated by centrifugation at 6000×g for 10 min. The resulting pellet was resuspended in 2 mL of sterile water. An aliquot of 500 μL was transferred to an Eppendorf tube and placed in −80 °C with no glycerol to induce cell lysis through the freeze-thaw process. The next day, the cells were heated on an 80 °C heat block for 30 min and then cooled to room temperature and 500 μL of 2:1 chloroform/methanol solution was added to the tube. The cells and solvent were mixed periodically over the course of 1 h. The cells were then centrifuged at 2500×g for 20 min to separate the organic and aqueous phases. A white band of cells was located in between the two liquid phases. Both the organic and aqueous phases were carefully pipetted out, and the Eppendorf tube was placed on a 55 °C heat block, uncapped, for 10–15 min to fully evaporate any remaining organic solvent. The leftover cell pellet was once again resuspended in 500 μL of sterile water and mixed vigorously by vortexing. Then, 50 μL of 1 M Tris-HCl (pH 9.5) solution and 25 μL of lysozyme (10 mg/mL) in 10 mM Tris-HCl (pH 7.5) was added. The cell-enzyme mix was incubated at 37 °C overnight with no shaking. After overnight lysozyme treatment, 70 μL 10 % SDS and 50 μL proteinase K (10 mg/mL) in 10 mM Tris-HCl (pH 7.5) was added to the mixture and incubated at 60 °C for 1 h with periodic mixing. Then, 100 μL 5 M NaCl solution and 100 μL 10 % Cetyltrimethylammonium bromide (CTAB) solution was added to the mixture. The resulting mixture was incubated at 60 °C with periodic mixing for 15 min. The resulting cell lysate was left to cool down to room temperature. 700 μL of chloroform/isoamyl alcohol (24:1) solution was added to the cell lysate and inverted 20–25 times until there was a homogenous white-opaque solution. The cell lysate was centrifuged at 13,000×g for 10 min at room temperature. The resulting aqueous layer was transferred to a sterile Eppendorf tube and 10 μL of RNaseA (10 mg/mL) was added. The enzyme mix was incubated at 37 °C for 30 min. Genomic DNA was precipitated by adding 0.1 volume of 3 M sodium acetate (pH 5.2) and 1 volume of isopropanol to the RNase-treated extract and invert 25–30 times to mix. DNA was pelleted in a microcentrifuge at 12,000×g at 4 °C for 15 min. Supernatant was removed and the resulting DNA pellet was further washed with ice-cold 70 % ethanol. The genomic DNA was then pelleted again in a microcentrifuge at the same speed and setting. The ethanol was completely removed by pipetting and evaporation. The genomic DNA was finally dissolved in 50–100 μL of sterile water.

Whole genome sequencing of genomic DNA was performed on an Illumina NextSeq500 at a depth of 400 Mbp (2.67 M Reads) and with a MinIon (Oxford Nanopore Technologies). The resulting short and long reads were hybrid assembled into a complete genome with unicycler version 0.4.8, using a MAB 0253a reference genome previously sequenced. The reference genome was annotated with prokka version 1.14.5 [84]. The genes within the reference genome were matched to genes in MAB ATCC 19977 genome (NCBI accession: CU458896.1) by BLASTn. The breseq variant calling pipeline was used to identify mutations in experimentally evolved populations [85]. The genomes of all passage 15 evolution strains and the ancestor were compared to the MAB 0253a reference genome, and mutations that arose during the evolution experiment were identified. Mutations with a population frequency of less than 50 % were excluded. Mutations that were present in either the ancestor or the TYEM lineages were also excluded.

4.7. Whole genome sequencing and variant calling between MAB 0253a/0711a (smooth) and MAB 0253b/0711b (rough) variants

Genomic DNA extracted from MAB 0253a and 0253b was sequenced using the Illumina NextSeq 2000 platform to generate at least 400 Mbp of 2 x 151 bp paired-end reads. Reads were trimmed with Trimmomatic version 0.36 and quality checked with FastQC version 0.11.5. Variant calling between the two isolates was performed using breseq v0.31.0 in consensus mode with default parameters, with MAB 0253a as the reference genome [85]. The same method was applied to the comparison between MAB 0711a and 0711b, with MAB 0711a used as the reference genome.

4.8. RNA extraction and RNA-sequencing

Stationary phase cultures of MAB 0253a in TYEM plus Tween 80 were normalized to OD₆₀₀ of 1 using 1 mL of PBS. From the 1 OD₆₀₀ culture, the cells were diluted into 50 mL of SCFM1, with chemical supplements as described in the text, at a ratio of 1:100 and aliquoted into 10 test tubes (5 mL per tube). The cells were left to grow at 37 °C, shaking at 250 RPM, until 34 h post-inoculation. Cells were then combined into a 50 mL conical and centrifuged at 7000×g for 10 min at 4 °C. An aliquot of 1 mL residual medium was taken out of the supernatant and the rest was decanted. The pellet was resuspended in the 1 mL residual medium and transferred to a new Eppendorf tube, where it was flash-frozen in an ice-cold ethanol bath. The frozen sample was stored at −80 °C until needed.

RNA extracts were prepared using the Invitrogen PureLink™ RNA Mini Kit (Protocol for purification from bacterial cells, with on-column DNase treatment) (Thermo Fisher Scientific) and were sent to the Health Sciences Sequencing Core at the UPMC Children’s Hospital of Pittsburgh for library preparation and sequencing. Fasta files were concatenated for each sample. For RNA-seq, RNA was first assessed for quality using an Agilent TapeStation 4150 and RNA concentration was quantified on a Qubit FLEX fluorometer. Libraries were generated with the Illumina Stranded Total Library Prep kit with Ribo-Zero Plus Microbiome (Illumina: 20072063) according to the manufacturer's instructions. Briefly, 250 ng of input RNA was used for each sample. Following adapter ligation, 15 cycles of indexing PCR were completed, using Illumina RNA UD Indexes. Library quantification and assessment was done using a Qubit FLEX fluorometer and an Agilent TapeStation 4150. Libraries were normalized and pooled to 2 nM by calculating the concentration based off the fragment size and the concentration of the libraries. Sequencing was performed on an Illumina NextSeq 2000 with read lengths of 2x101 bp, and a target of 10 million reads per sample.

4.9. RNA-seq data analysis

Filtering, trimming, alignment, count, and normalization of RNA-seq reads were performed using the ProkSeq (version 2.0:v2) analysis pipeline with 4 processors and single-end reads [86]. RNA-seq reads were mapped to the MAB 0253a reference genome. Normalized read counts were used to calculate gene expression values in transcripts per million (TPM). Differential gene expression analysis was performed by calculating the log₂ fold change of TPM values between the treatment and control groups. The full RNA-seq analysis results and TPM values for all genes in MAB 0253a can be found in the Supporting Information (Tables S6–8). The heat map of the differentially expressed genes (DEGs) was generated using the GraphPad Prism software ver. 10.4.1.

4.10. Agar block biofilm assay (ABBA) and confocal microscopy

The ABBA was performed as described with modifications [52]. A stationary phase MAB 0253a culture in TYEM medium with no Tween 80 was first adjusted to an OD₆₀₀ of 1 in 1 mL of PBS. From the 1 OD₆₀₀ culture, the cells were diluted into 1 mL of SCFM1 (supplemented with 10 mM glucose or 30 mM acetate) with 0.4 % warm, melted noble agar in an Eppendorf tube at a ratio of 1:100. The cells were mixed by vortexing and 200 μL of the agar-cell mix was aliquoted into each well of an 8-well Lab-Tek II Chambered Coverglass slides (Thermo Scientific). The ABBA cassettes were set up in triplicate. For comparing MAB 0253a grown with glucose versus acetate, four wells on each cassette were allocated for SCFM1 plus glucose conditions, while the rest were allocated for SCFM1 plus acetate conditions. The agar block was left to solidify, then transferred to a humidity chamber and left to incubate at 37 °C for 48 h. At 48 h, 175 μL of clarithromycin or amikacin was added to the top of the ABBA blocks. Clarithromycin concentrations were at 25, 50, and 100 μg/mL. Amikacin concentrations were at 10, 15, and 25 μg/mL. As a negative control, one well received water in place of antibiotic. The ABBA cassette was then placed back in the humidity chamber and left to incubate at 37 °C overnight. After overnight incubation, the antibiotic was removed by pipetting. The remaining antibiotic was rinsed off with 200 μL of sterile water.

For assessing antibiotic killing, the ABBA blocks were transferred from the cassette using a sterile spatula to an Eppendorf tube with 500 μL of sterile PBS. Using a Fisherbrand 150 Handheld Homogenizer, the ABBA blocks were broken up and serial dilution was performed in a 90-well plate to a dilution of 10⁻⁷ in PBS. The diluted cells were plated onto TYEM agar plates using the drip plate method. The plates were incubated at 37 °C until colonies could be counted. CFUs/mL for each antibiotic concentration were calculated. Survival percentage was calculated by dividing the CFU/mL of the antibiotic-treated well with the CFU/mL of the untreated control well.

For preparing the ABBA samples for confocal imaging, a master mix of 1245 μL sterile water plus 5 μL of the SYTO™ 9 Green Fluorescent Nucleic Acid Stain (Invitrogen) was prepared. For each well in the ABBA cassette, 125 μL of the diluted SYTO 9 mix was applied. The ABBA cassette was wrapped in aluminum foil and left to incubate in the humidity chamber at 37 °C for 2–4 h. Confocal microscopy was performed at the Cell Imaging Core Laboratory at the UPMC Children's Hospital of Pittsburgh using a Leica Confocal Microscope Platform STELLARIS 5 with both a HC PL APO 10x/0,40 CS2 and a 40x/1,1 W CORR CS2 objective (Leica). Carl Zeiss Immersol Immersion Oil W 2010 was used as the immersion fluid. All images and Z-stacks were collected in 8-bit mode at 600 scan speeds with a 1024-by-1024 scan format and line averaging of 1. 10X Z-stacks were collected at a length of 200 μm. Image processing, reconstruction, and analysis were all done with Imaris imaging software ver. 10.1.0 (Bitplane).

4.11. Antibiotic susceptibility assays

Antibiotic susceptibility assay was performed based on the guideline issued by the Clinical and Laboratory Standards Institute (CLSI) for the purpose of identifying the planktonic MIC of MAB 0253a against multiple antibiotics in SCFM1 with and without acetate supplement [[87], [88], [89]]. MAB cultures were first grown up to stationary phase in TYEM medium with Tween 80. Using the broth dilution method, 100 μL of SCFM1 plus either 10 mM glucose or 30 mM acetate and 0.1 % Tween 20 was added to the first 11 columns of a 96 well flat bottom plate (Fisherbrand, Cat # FB012931). Then, 200 μL of SCFM1 plus glucose or acetate with 256 μg/mL clarithromycin or amikacin was pipetted into the final column. Serial dilutions of the antibiotic were performed, with 100 μL being transferred, starting from the final column, between wells with pipetting to mix to the second to last well. One column was intentionally kept clean of any antibiotic to serve as a positive control. The extra 100 μL of media was ejected. Culture of MAB 0253a was then diluted to an inoculum of 0.0005 OD₆₀₀ in SCFM1 with glucose or acetate. 100 μL of the inoculum was added to each well of the plate in triplicate per strain. The addition of inoculum brought the final antibiotic concentration to a range of 0.5 μg/mL to 128 μg/mL indexed to base 2. The 96-well plates were sealed with parafilm and incubated at 37 °C in a humidity chamber. MICs were read visually on Day 3, Day 5, Day 7, and Day 14 as recommended by the CLSI for testing rapid growing mycobacteria susceptibility.

4.12. Pellicle biofilm assays

MAB cultured to stationary phase in TYEM medium with no Tween 80 was first adjusted to an OD₆₀₀ of 1 in 1 mL of PBS. From the 1 OD₆₀₀ culture, cells were diluted into SCFM1 (supplemented with 10 mM glucose or 30 mM acetate) at a ratio of 1:100. 2 mL of culture was aliquoted to each well of a 24-well plate. Three replicates were set up for each experimental condition. The plate was incubated at 37 °C for four or five days. Pictures of the representative pellicle biofilms were taken using the Zeiss Axiocam 305 color camera, viewed under a Zeiss SteREO Discovery.V8 microscope with a Plan S 1.0× Objective.

4.13. Sliding motility assays

MAB 0253a (ancestor), MAB 0253b, S2, and SNG3 were cultured in TYEM with Tween 80 and adjusted to an OD₆₀₀ of 1 in 1 mL of PBS. The cells were centrifuged at 16,000×g and the resulting pellet was washed with 1 mL of PBS to rinse off any leftover Tween 80. From the 1 OD₆₀₀ culture, 2 μL was spotted onto a semi-solid Middlebrook 7H9 agar plate with 0.3 % agar [31]. Triplicate plates were prepared for each MAB strain. After spotting, the plates were sealed with parafilm and incubated at 37 °C for 12 days. The motility morphology of the MAB strains was visually inspected at the end of the 12-day incubation period, and representative images of the plates were taken using the Zeiss Axiocam 305 color camera, viewed under a Zeiss SteREO Discovery.V8 microscope with a Plan S 0.3× Objective.

4.14. Genomic analysis of MAB_0812 (mraA) and MAB_0813c (mraB)

Alleles of MAB_0812 and MAB_0813c were identified using BLASTp, restricting the results to Mycobacteria [90]. The isolation sources of the genomes that encode the Mycobacteria spp. alleles were identified using the NCBI Entrez database. A multiple sequence alignment (MSA) of the Mycobacteria alleles, as well as an MSA of a smaller subset of MAB alleles, was created using Clustal Omega [91]. Further, an MSA of MAB genomes sourced from people with CF was created. The resulting MSAs were used to construct phylogenetic trees of MAB alleles with RAxML [92]. Isolation source and synteny of MAB_0812 and MAB_0813c for the genomes that encode the alleles present in the trees was annotated using iTOL [92,93]. In addition, a phylogenetic tree of Mycobacteria spp. that had at least three MAB_0812 alleles present on NCBI was created, annotated with isolation source and synteny information. The NCBI accession numbers of the representative ortholog used for each species in the MAB_0812 Mycobacteria spp. tree is included in the Supplemental Materials (Table S9). Synteny was assessed using Pynteny [94]. Lollipop diagrams of MAB_0812 and MAB_0813c were constructed to display mutations found in MAB CF alleles using lollipops [95]. Amino acid identity (AAI) of the MAB CF alleles against the MAB 0253a strain allele was calculated using EMBOSS [96]. Lengths of Mycobacteria alleles were analyzed and summarized in histograms showing the distribution of lengths of alleles in comparison to the lab strain alleles and the mutated evolved alleles.

4.15. Statistical analysis

All statistical analysis in this study was done using the two-tailed, unpaired Student's T-test. Statistical analysis was performed using the data analysis package in GraphPad Prism software ver. 10.4.1. Statistical significance is achieved when the P-value is less than 0.05.

CRediT authorship contribution statement

Yu-Hao Wang: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Isabelle D'Amico: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation. Jocelyn Whalen: Writing – review & editing, Visualization, Methodology, Formal analysis, Data curation. Steven J. Mullett: Writing – original draft, Resources, Methodology, Formal analysis, Data curation. Stacy L. Gelhaus: Writing – review & editing, Writing – original draft, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Data curation. Vaughn S. Cooper: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Conceptualization. Catherine R. Armbruster: Writing – review & editing, Visualization, Supervision, Project administration, Methodology, Funding acquisition, Data curation, Conceptualization. William H. DePas: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: William DePas reports financial support was provided by National Institutes of Health. William DePas reports financial support was provided by Cystic Fibrosis Foundation. Catherine Armbruster reports financial support was provided by National Institutes of Health. Catherine Armbruster reports financial support was provided by Cystic Fibrosis Foundation. Stacy Gelhaus reports financial support was provided by National Institutes of Health. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article.

Data availability

Data will be made available on request.