Magnesium is an axis of nutritional competition between P. aeruginosa and C. albicans

We focused on the contributions of the PA4824, PA4825 (mgtA), and PA4826 genes during co-culture. We engineered gene deletion mutants and measured the fitness of these mutants relative to a fluorescently labeled wild-type (WT) strain in either monoculture or co-culture with C. albicans using competitive fitness assays, which resemble the pooled Tn-seq bulk selection conditions (S4A Fig). Our experiments showed that the relative fitness of the ΔPA4824 or ΔPA4825 (ΔmgtA) mutants, but not ΔPA4826, was approximately 10% lower than the WT in co-culture (S4B Fig). In contrast, neither of these mutants had any measurable fitness defect, relative to WT PAO1, in monoculture (S4B Fig). Co-culture colony formation assays also confirmed our findings; deletion of either ΔPA4824 or ΔmgtA led to a 10-fold reduction in colony-forming units (CFUs) of P. aeruginosa in co-culture relative to the CFUs of WT in co-culture (Fig 1D). CFUs of ΔPA4824 or ΔmgtA in co-culture were restored by expressing a WT copy of PA4824 or mgtA in trans (S4C and S4D Fig). Moreover, a double deletion ΔPA4824 ΔmgtA mutant was even more significantly impaired than single deletion mutants (Fig 1D). These results suggest that the protein products encoded by PA4824 and mgtA are independently required for optimal P. aeruginosa fitness under co-culture conditions, but PA4826 is not. We infer that the identification of PA4826 in our Tn-seq analyses may be due to the effects of insertional mutants in PA4826 on the expression of adjacent PA4824 and PA4825 (mgtA) genes.

Our finding that P. aeruginosa PA4825 (mgtA) is required for full bacterial fitness in co-culture led us to hypothesize that C. albicans depletes the vital Mg²⁺ cation in co-culture. Under Mg²⁺ depletion conditions, mgtA might become essential for fitness since its loss would result in lower intracellular Mg²⁺ levels. To test this hypothesis, we employed a Mg²⁺ genetic sensor [31] from Salmonella enterica serovar Typhimurium. Previous studies showed that S. Typhimurium mgtA expression levels are controlled by its 5′ untranslated region (5′UTR), which acts as a ribo-switch, adopting different stem-loop structures depending on the Mg²⁺ levels and regulating the transcription of mgtA [31]. We showed that this S. Typhimurium Mg²⁺ sensor can accurately detect changes in intracellular P. aeruginosa Mg²⁺ levels (S5 Fig). Using this sensor, we found that the intracellular Mg²⁺ levels in the WT P. aeruginosa are reduced by 2-fold in C. albicans-spent media (i.e., derived from the filtrate of C. albicans cultures to simulate Mg²⁺ depletion) compared to in fresh BHI media (Fig 1E). Loss of mgtA further reduced bacterial intracellular Mg²⁺ levels in the fungal spent media, but not in monoculture conditions (Fig 1E). Surprisingly, loss of PA4824 also reduced intracellular Mg²⁺ levels in fungal spent media compared to the WT strain, with a double deletion ΔPA4824 ΔmgtA mutant further reducing intracellular Mg²⁺ levels, compared to the single gene deletion mutants (Fig 1E). These experiments confirm fungal-imposed Mg²⁺ sequestration and demonstrate that mgtA and PA4824 are independently required for both maintaining intracellular P. aeruginosa Mg²⁺ levels as well as fitness of P. aeruginosa cells in co-culture with C. albicans.

Our findings suggest that PA4824 may play a crucial role in Mg²⁺ uptake. Although its function remains uncharacterized, Alphafold structural predictions [32] revealed that PA4824 likely encodes a transmembrane protein with a distinctive β-barrel structure commonly associated with proteins involved in ion transport or nutrient uptake (S6A Fig). The core of PA4824 is composed of hydrophilic and negative-charged residues (S6B and S6C Fig), suggesting it might transport positive-charged molecules. Based on our co-culture RNA-seq and Tn-seq experiments, results from the Mg²⁺ genetic sensor assay, and the Alphafold prediction of PA4824 protein structure, we speculate that PA4824 potentially acts as a novel Mg²⁺ transporter.

In addition to MgtA (and potentially PA4824), P. aeruginosa uses 2 other constitutively expressed Mg²⁺ transporters, CorA and MgtE, to obtain Mg²⁺ from the environments [33]. However, neither CorA nor MgtE was implicated in competition with C. albicans by our Tn-seq or RNA-seq analyses (S7A Fig). We also found that corA or mgtE loss-of-function mutants did not alter P. aeruginosa fitness in co-culture conditions (S7B Fig), nor did they significantly reduce intracellular Mg²⁺ levels in C. albicans-spent BHI compared to ΔmgtA mutant (S7C Fig). Our experiments reveal that MgtA is the primary Mg²⁺ transporter induced in co-culture, crucial for bacterial fitness and Mg²⁺ uptake in co-culture. Thus, we conclude that that MgtA is a key bacterial Mg²⁺ transporter required to overcome fungal sequestration of Mg²⁺.

If C. albicans sequesters Mg²⁺ from P. aeruginosa, we hypothesized that supplementing cultures with Mg²⁺ might rescue the fitness defects of the ΔmgtA and ΔPA4824 mutants. Indeed, supplemental Mg²⁺ (10 mM) was sufficient to restore the fitness of ΔmgtA, ΔPA4824, and ΔPA4824 ΔmgtA double deletion mutants to WT levels in co-culture conditions (Fig 1F). To investigate whether Mg²⁺ could explain fitness effects of other candidate genes revealed by the Tn-seq analyses, we repeated our Tn-seq experiments in monoculture versus co-culture conditions except that we added 10 mM Mg²⁺ to BHI media. To our surprise, we found that Mg²⁺ supplementation ameliorated the fitness effects of most of the candidate genes identified in our initial Tn-seq analyses (S8A Fig), including all 8 fungal-defense genes and 17 of 19 dispensable genes crucial to fundamental cellular processes, including cell division, pyrimidine synthesis, and energy production—enzymes in essential processes that rely on Mg²⁺. Despite their importance to bacterial physiology, we hypothesize that these otherwise critical processes might become too costly to maintain in low Mg²⁺ co-culture conditions. Only 2 genes (PA4029 and PA5484) remained marginally essential in co-culture, even under supplemented Mg²⁺ conditions (S8A Fig). Notably, Mg²⁺ supplementation only affected the expression of PA4824-PA4826, but not other candidate genes (S8B Fig), suggesting that Mg²⁺ levels affect their functions posttranscriptionally. Thus, our finding underscores the influential role of Mg²⁺ competition in mediating the fitness effects of most candidate genes in co-culture conditions. We note, however, that Mg²⁺ supplementation did not fully restore the fitness of P. aeruginosa WT cells in co-culture to monoculture fitness levels (Fig 1F), consistent with prior observations of additional, Mg²⁺-independent, axes of antagonism between bacteria and fungi [26,34,35].

Nutritional competition for important metal ions, such as iron, zinc, and manganese, is known for competition among microbes and between microbes and vertebrate hosts [36–38]. However, competition for the vital Mg²⁺ cation is a novel finding. We reasoned that our finding might reflect different Mg²⁺ levels in various media conditions. Indeed, we found that MgtA is required for P. aeruginosa fitness in co-culture with C. albicans in BHI and tryptic soy broth (TSB) media (S9A Fig), but not in yeast extract-peptone-dextrose (YPD) or synthetic cystic fibrosis media (SCFM) [39] (S9B and S9C Fig). Mg²⁺ levels in BHI and TSB were lower (100 to 140 μm) than in YPD or SCFM (240 to 390 μm) (S9D Fig). As a result, fungal filtrates derived from BHI and TSB exhibited the lowest levels of Mg²⁺ (S9D Fig), leading to a significant reduction of intracellular Mg²⁺ in P. aeruginosa (S9E Fig). Additionally, by titrating Mg²⁺ levels in co-culture, we found that supplementation of BHI with greater than 300 μm Mg²⁺ was sufficient to relieve the fitness defect of the P. aeruginosa ΔmgtA mutant (S9F Fig). Thus, we conclude that BHI media represents a low Mg²⁺ condition that reveals a novel mode of nutritional competition between fungi and bacteria.

MgtA was previously described as a Mg²⁺-specific transporter [31]. To rule out the possibility that the fitness impairment of ΔmgtA mutant in co-culture could be caused by the reduction of other cations in BHI, we performed a series of supplementation experiments with alternate cations. We found that Mg²⁺ levels specifically determine P. aeruginosa viability in co-culture with C. albicans; supplementation of zinc (Zn²⁺) or copper (Cu²⁺) ions did not restore the viability of ΔmgtA mutant in co-culture to the monoculture levels (S9G Fig). The addition of molybdenum (Mo⁶⁺) or manganese (Mn²⁺) ions drastically decreased P. aeruginosa viability even in monoculture (S9G Fig), which rules out the possibility that either of these 2 ions was limited in BHI. Finally, we found that only a very high concentration of Ca²⁺ (10 mM) rescued the fitness defect of the ΔmgtA mutant in co-culture (S9H Fig), orders of magnitude higher than the concentration of Mg²⁺ required to rescue ΔmgtA fitness in co-culture (S9F Fig). These experiments further confirm that the fitness of ΔmgtA in co-culture conditions is primarily determined by Mg²⁺ sequestration.

Widespread Mg²⁺ competition between fungi and gram-negative bacteria

We next assessed whether Mg²⁺ competition is a general means of antagonism by fungi against bacteria. Escherichia coli and S. Typhimurium are gram-negative bacteria that coexist with C. albicans in the human gut [40]. E. coli encodes a single mgtA ortholog (like P. aeruginosa), whereas S. Typhimurium encodes 2 paralogs: mgtA and mgtB. We found that either a single deletion ΔmgtA in E. coli or a double deletion ΔmgtA ΔmgtB in S. Typhimurium severely reduced bacterial fitness in co-culture with C. albicans, and such fitness reduction was restored via Mg²⁺ supplementation (Fig 2A and 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mg²⁺ competition is widespread between multiple fungal and gram-negative bacterial species.

(A) The fitness of the E. coli ΔmgtA mutant was impaired relative to the WT strain in co-culture with C. albicans (orange) but not in monoculture (blue) in BHI media. The fitness of the ΔmgtA strain was restored in BHI media supplemented with 10 mM Mg²⁺. (B) The fitness of the double deletion ΔmgtA ΔmgtB mutant S. Typhimurium strain was impaired relative to WT in co-culture with C. albicans (orange) but not in monoculture in BHI media (blue). (C) The fitness of the ΔmgtA mutant P. aeruginosa strain (green) was impaired relative to WT (magenta) in BHI in co-culture with fungal species: C. albicans, C. tropicalis, C. parapsilosis, C. glabrata, and S. cerevisiae. (D) Mg²⁺ supplementation restored the fitness of the ΔmgtA mutant P. aeruginosa strain (green) in co-culture with fungal species in Fig 2C. (E) The fitness of the P. aeruginosa ΔmgtA mutant (green) does not significantly differ from WT (magenta) in co-culture with gram-positive S. aureus. (F) The fitness of the P. aeruginosa ΔmgtA mutant (green) was similar to WT (magenta) in co-culture with gram-negative bacteria, such as B. multivorans, E. coli, or S. Typhimurium. (G) The functions and locations of proposed Mg²⁺ transporters in S. cerevisiae are illustrated with Biorender.com (top panel). Alr2 localizes to the plasma membrane, Mrs2 and Mme1 to the mitochondria, and Mnr2 to the vacuole. The fitness of the P. aeruginosa ΔmgtA mutant (black) is not significantly different from the WT strain (orange) in monoculture. However, the relative fitness of P. aeruginosa ΔmgtA was impaired in co-culture with either WT S. cerevisiae or either of 3 S. cerevisiae mutant strains: Δalr2, Δmrs2, or Δmme1. In contrast, the relative fitness of P. aeruginosa ΔmgtA was restored in co-culture with S. cerevisiae Δmnr2 strain. Mean ± std of 4 biological replicates is shown in panels A–G. (** p < 0.01 unpaired two-tailed Student’s t test used; n.s. indicates not significant). The data underlying this Fig 2A–2G can be found in S11 Data. BHI, brain heart infusion; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3002694.g002

Next, we expanded our analyses to 3 additional fungal species that coexist with P. aeruginosa in polymicrobial infections [41]: Candida tropicalis, Candida parapsilosis, Candida glabrata, as well as a non-pathogenic species, Saccharomyces cerevisiae. In each case, we found that the P. aeruginosa ΔmgtA mutant had reduced fitness relative to WT cells in fungal co-culture (Fig 2C), which could be restored by Mg²⁺ supplementation (Fig 2D).

This mode of competition might be highly specific between fungi and diverse gram-negative γ-proteobacteria we have tested; it does not manifest in inter-bacterial competition. For example, lack of mgtA did not lower P. aeruginosa fitness in co-culture with bacteria that commonly co-exist in CF airways, such as gram-positive Staphylococcus aureus (Fig 2E) or gram-negative Burkholderia multivorans (Fig 2F). Similarly, P. aeruginosa ΔmgtA mutant had similar fitness as WT cells in co-culture with other gut-colonizing gram-negative like E. coli or S. Typhimurium (Fig 2F). Despite BHI representing a low-Mg²⁺ condition, our results showed that different bacterial species do not necessarily compete for Mg²⁺ during co-culture, suggesting fungi may have a higher demand or propensity for Mg²⁺ uptake and sequester Mg²⁺ more effectively than gram-negative bacteria. Overall, our findings confirm that competition for Mg²⁺ is a widespread mode of fungal antagonism against gram-negative bacteria and that MgtA plays an important role in counteracting fungal-mediated Mg²⁺ competition. Whether fungi can suppress gram-positive bacteria through the same mechanism of Mg²⁺ competition remains an open question.

How do fungi impose competition for Mg²⁺? We investigated whether the loss of S. cerevisiae genes in Mg²⁺ uptake or sequestration restores the fitness of P. aeruginosa ΔmgtA mutant. S. cerevisiae cells store 80% of cellular Mg²⁺ in their vacuole [42]. To investigate whether Mg²⁺ sequestration in the fungal vacuole is involved, we generated a deletion mutant in Mnr2 [43], previously proposed to be a vacuolar Mg²⁺ transporter in S. cerevisiae. We found that the fitness of P. aeruginosa ΔmgtA mutant was restored to WT levels in co-culture with the mnr2Δ mutant (Fig 2G). Yet, S. cerevisiae mnr2Δ mutant had similar fitness as WT S. cerevisiae both in media only as well as in co-culture (S10 Fig), suggesting that lack of Mnr2 does not cause a strong growth defect. In contrast, P. aeruginosa ΔmgtA fitness was not restored by competition with S. cerevisiae mutants lacking Alr2, a cell-wall-associated Mg²⁺ transporter used to acquire environmental Mg²⁺ into the cytosol [44], or those lacking Mrs2/Mme1, mitochondria-specific Mg²⁺ transporters (Fig 2G) [45]. Thus, fungal vacuolar Mg²⁺ sequestration might broadly mediate Mg²⁺ antagonism against gram-negative bacteria.

Fungal Mg²⁺ sequestration protects bacteria from polymyxin antibiotics

Mg²⁺ is required for many fundamental cellular processes by neutralizing negatively charged biomolecules and functioning as an enzyme cofactor [33]. Mg²⁺ is also important for stabilizing the outer membrane structure of gram-negative bacteria by binding to negatively charged lipopolysaccharides (LPS) on the bacterial cell membrane [46]. These Mg²⁺-stabilized LPS moieties can be targeted by polymyxins [47], last-resort antibiotics for treating multidrug-resistant bacterial infections [48,49] (Fig 3A). In Mg²⁺-limited conditions, bacteria activate dual two-component signaling pathways, PmrAB and PhoPQ, modifying lipid A of LPS with L-Ara4N (4-amino-4-deoxy-L-arabinose) and PEtN (phosphoethanolamine), leading to resistance to polymyxins [50–53] (Fig 3A). Thus, fungal-mediated Mg²⁺ depletion might confer polymyxin resistance in bacteria.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. C. albicans-mediated Mg²⁺ sequestration protects gram-negative bacteria from colistin.

(A) Mg²⁺ binds to negatively charged lipid A of LPS on the outer membrane of gram-negative bacteria. In high Mg²⁺ conditions, colistin (and other polymyxin antibiotics) targets Mg²⁺-bound LPS to disrupt cell membranes of gram-negative bacteria, leading to bacterial cell death. Low Mg²⁺ conditions activate two-component signaling pathways, PmrAB and PhoPQ, that modify LPS (PEtN and L-Ara4N), making bacteria resistant to colistin. (B) Schematic of the colistin survival assay in C. albicans-spent media. P. aeruginosa strains were grown to stationary phase in BHI, C. albicans-spent BHI, or C. albicans-spent BHI supplemented with 10 mM Mg²⁺, and then passaged until exponential phase, when they were treated with 25 μg/ml colistin. Bacterial survival is shown as the percentage of viable cells after the colistin treatment. (C) WT P. aeruginosa strains were highly susceptible to colistin treatment in BHI media. However, they could survive colistin treatment in C. albicans-spent BHI media but not in spent BHI media supplemented with 10 mM Mg²⁺. In contrast to WT strains, pmrB or phoP Tn-mutants of P. aeruginosa (which cannot modify LPS) remained highly susceptible to colistin even in C. albicans-spent BHI media. (D) WT S. Typhimurium was highly susceptible to 3 μg/ml colistin in either BHI media or C. albicans-spent media supplemented with 10 mM Mg²⁺ but became more resistant to colistin in fungal spent media alone. (E) Schematic of the colistin survival assay in co-culture. The P. aeruginosa WT strain was cultured only in BHI or co-culture with C. albicans for 18 h. Monocultures or co-cultures were then treated with colistin. Bacterial survival is calculated as the percentage of viable cells after the colistin treatment, just as in (B). (F) P. aeruginosa co-cultures with C. albicans (orange) were more resistant than monocultures (blue) over a series of colistin concentrations, from 1.5 μg/ml colistin (clinical MIC) to 12.5 μg/ml colistin (8× clinical MIC). (G) Like in (F), co-cultures of S. Typhimurium with C. albicans (orange) were significantly more resistant than monocultures of S. Typhimurium alone over a range of colistin concentrations. Mean ± std of 3 biological replicates is shown in panels C–D and F–G. (**p < 0.01, Dunnett’s one-way ANOVA test and unpaired two-tailed Student’s t test were used in panel C–D and F–G separately). The data underlying this Fig 3C–3D and 3F–3G can be found in S13 Data. Fig 3A, 3B, and 3E are created with Biorender.com. BHI, brain heart infusion; LPS, lipopolysaccharide; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3002694.g003

To test this possibility, we grew P. aeruginosa in fresh BHI media or C. albicans-spent BHI media, which mimics Mg²⁺ depletion by fungi, and examined bacterial survival when treated with colistin (also referred to as polymyxin E) (Fig 3B). Consistent with our hypothesis, we found that cells grown in fresh media showed no survival after colistin treatment, whereas 20% of cells grown in fungal spent media survived colistin treatment (Figs 3C and S11A). Supplementation with extra Mg²⁺ in C. albicans-spent media restored the sensitivity of P. aeruginosa cells to colistin, to levels resembling sensitivity in BHI fresh media (Figs 3C and S11A). Since low Mg²⁺ conditions activate the PmrAB and PhoPQ signaling pathways required to modify LPS and confer colistin resistance, we tested whether pmrB or phoQ loss-of-function transposon mutants [54] would abolish the observed colistin resistance in fungal spent media; we found that this was the case (Fig 3C).

We found that fungi-dependent increased colistin survival is not restricted to P. aeruginosa or colistin. For example, we found that Mg²⁺ depletion by C. albicans also increased the survival of S. Typhimurium to colistin, a phenotype that is also suppressed by extra Mg²⁺ (Figs 3D and S11B). Furthermore, Mg²⁺ depletion increased P. aeruginosa survival to another polymyxin antibiotic, polymyxin B (S12 Fig). Thus, fungal sequestration of Mg²⁺ provides a general means of conferring polymyxin resistance on gram-negative bacteria.

Since fungal spent media cannot fully recapitulate fungal presence in co-culture conditions, we tested whether fungal co-culture also conferred increased colistin survival (Fig 3E). We found that co-culture with C. albicans was sufficient to protect P. aeruginosa and S. Typhimurium from colistin across a range of concentrations (1.5 μg/ml to 12.5 μg/ml; the clinical MIC of P. aeruginosa to colistin is 2 μg/ml), in contrast to monoculture conditions at both 30 °C (Figs 3F, 3G, S13A and S13B) and 37 °C (S14A and S14B Fig). Additionally, the protection from antibiotics by fungi in co-culture applies to different P. aeruginosa strains. We found that co-culture with C. albicans also protects several colistin-sensitive P. aeruginosa isolates from cystic fibrosis patients from colistin (S15 Fig). Thus, C. albicans can protect gram-negative bacteria from colistin across a range of conditions.

Fungal co-culture fundamentally alters the evolution of bacterial colistin resistance

Understanding how microbial interactions drive the evolution of antibiotic resistance is important for biological and biomedical reasons; increased antibiotic resistance is often observed in polymicrobial communities [55]. We hypothesized that long-term coexistence with C. albicans may fundamentally alter the onset and evolution of enhanced colistin survival in P. aeruginosa. To test this possibility, we passaged 8 replicate WT populations of P. aeruginosa PAO1 for 90 days, either in monoculture or in co-culture with C. albicans, with each population exposed to gradually increasing concentrations of colistin (from 1.5 μg/ml to 192 μg/ml) (Fig 4A). As controls, we also passaged mono- and co-culture populations in the absence of colistin (Fig S16). After evolution, both monoculture- and co-culture-evolved populations gained increased survival to 192 μg/ml colistin (Fig 4B). Surprisingly, we found that enhanced colistin survival in P. aeruginosa co-culture populations was dependent on C. albicans; colistin survival was almost completely lost upon removal of C. albicans using antifungal treatment or the addition of Mg²⁺ (Figs 4C and S17). Our findings indicate that co-culture-evolved populations acquire increased colistin survival due to fungal protection induced by Mg²⁺ sequestration but continue to depend on the fungus for increased protection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Long-term coexistence with C. albicans alters the evolution of P. aeruginosa colistin resistance.

(A) Schematic of the evolution experiment in co-culture, created with Biorender.com. Eight independent WT P. aeruginosa populations were passaged daily in BHI with colistin or in BHI with both C. albicans and colistin. Colistin levels were gradually increased from 1.5 μg/ml to 192 μg/ml (128× clinical MIC) over 90 days (see Methods). (B) After evolution, 8 monoculture replicate populations (blue) were not significantly different from 8 co-culture populations (orange) for resistance to 192 μg/ml colistin (co-culture populations were treated with colistin in the presence of co-evolving fungal cells). For each population, the average survival rate of 4 replicates is shown (n.s. indicates not significant, Mann–Whitney U test). (C) P. aeruginosa cells from co-cultured-evolved populations (orange circle) lost colistin resistance upon removal of C. albicans with antifungal nystatin treatment (gray circle) or upon Mg²⁺ supplementation (yellow triangle). Mean ± std of 4 biological replicates is shown (* p < 0.05, ** p < 0.01, Mann–Whitney U test). (D) Genome sequencing revealed that P. aeruginosa strains that evolved enhanced colistin survival either in monoculture or in co- had different numbers of fixed mutations per population (** p < 0.01, Mann–Whitney U test) due to recurrent hypermutator mutations in monoculture populations. (E) Monoculture- and co-culture-evolved populations acquired enhanced colistin survival via a distinct spectrum of putative adaptive mutations. Columns indicate each individual replicate monoculture (blue) or co-culture (orange)-evolved population. Rows indicate genes that acquire mutations, grouped by cellular processes. We list all genes with fixed mutations (gray) in 5 or more replicate populations in either condition (* indicates a mutation in the promoter region). The only genes common between monoculture and co-culture-adapted populations are colS, previously known for colistin resistance, and PA3969 (indicated with a #), which is not unique to colistin-treated populations. A more comprehensive list of all fixed mutations is presented in S3 and S4 Tables. The data underlying this Fig 4B–4D can be found in S19 Data. BHI, brain heart infusion; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3002694.g004

Based on our findings, we suspected that P. aeruginosa populations evolved in co-culture with fungi likely enhance colistin survival via fundamentally different mutational mechanisms than monoculture-evolved populations. To test this hypothesis, we performed genome sequencing of all experimentally evolved populations. We found that all colistin-treated populations grown in monoculture became hypermutators (Fig 4D) by acquiring at least 1 mutation in genes involved in DNA replication or repair (uvrB, recQ, and mutS) (Fig 4E and S3 Table). Hypermutation is a hallmark of antibiotic-resistance acquisition that has been observed in several previous experimental evolution studies [56]. As a result of hypermutation, each monoculture population acquired hundreds of fixed mutations (Fig 4D), including mutations previously shown to confer colistin resistance, in genes such as pmrB [57], ptsP [58,59], pqsR [58], colS [60,61], porins [58], and proteins related to bacterial motility [62] (Fig 4E and S3 Table). In contrast, none of the co-culture-evolved populations became hypermutators; each only harbored 6 to 8 fixed mutations (Fig 4D), primarily in other targets for colistin resistance, such as genes in LPS biosynthesis [63] and an outer-membrane protein oprH [52] (Figs 4E and S18 and S3 Table). Intriguingly, we found that PA4824, another co-culture-specific defense gene identified in our Tn-seq experiments, was also mutated in all 8 replicates of co-culture-evolved populations (Figs 4E and S14 and S3 Table). We did not find fixed mutations in any of these genes in co-culture populations that were not exposed to colistin (S4 Table). Thus, the dual stressors of colistin and fungal co-culture specifically select for distinct classes of adaptive mutations, fundamentally altering the mode of bacterial colistin survival.

Bacterial and fungal strains and culture conditions

All bacterial and fungal strains used in this study are listed in S5 Table in the supplemental material. Bacterial strains in this study were derived from P. aeruginosa PAO1, E. coli BW25113, S. Typhimurium 14028S, S. aureus SA113, and B. multivorans AMT17616. Yeast strains were derived from C. albicans SC5314, S. cerevisiae W303. C. tropicalis CBS94, C. parapsilosis CBS604, and C. glabrata CBS562 were purchased from ATCC. Individual P. aeruginosa Tn mutant strains were obtained from Dr. Colin Manoil’s lab [28,54]; correct transposon insertions were confirmed by PCR and Sanger sequencing. Clinical P. aeruginosa isolates from CF patients were obtained from Dr. Pradeep Singh’s lab. Unless otherwise specified, all experiments were performed in Brain Heart Infusion Broth Media (BHI, Sigma-Aldrich), buffered with 10% MOPS to pH 7.0 followed by filter sterilization. All strains were grown in BHI and incubated at 30 °C or 37 °C. Other media used in this study were YPD (1% yeast extract, 2% peptone, and 2% D-glucose), SCFM [39], and TSB (Sigma-Aldrich). For antibiotics used in this study, 100 μg/ml gentamicin was used to select against gram-negative bacteria, and 50 μg/ml nystatin was used to select against Candida species and S. cerevisiae. Colistin and polymyxin B (Sigma-Aldrich) were prepared as 10 mg/ml and 5 mg/ml stock solutions, respectively.

Generation of P. aeruginosa mutants

For generating gene deletions in P. aeruginosa, a pEXG2-mediated allelic exchange method was used [83]. In short, pEXG2 deletion constructs were transformed into E. coli S17. S17 donor was subsequently mixed with P. aeruginosa recipients on an LB agar plate at a 5:1 ratio of donor to recipient cells, and the cell mixture was incubated at 30 °C overnight. The cell mixture was then scraped, resuspended into 200 μl LB, and plated on LB agar plates with 100 μg/ml gentamicin to select for cells containing the deletion plasmid integrated into the P. aeruginosa genome. P. aeruginosa merodiploid colonies were streaked on LB-no salt-agar plates with sucrose. Gentamicin-sensitive and sucrose-resistant colonies were screened for allelic replacement by colony PCR, and gene deletions were confirmed by Sanger sequencing of PCR products. Primers and plasmids used for strain constructions are listed in S6 Table.

Generation of S. cerevisiae mutants

S. cerevisiae gene deletion mutants were generated by homologous recombination at endogenous gene loci. Gene deletion fragments were amplified from pFA6a-natMX4 and fused with 500 bp upstream and downstream sequences of the targeted coding sequence by PCR. The PCR fragment with homology arms was transformed into S. cerevisiae cells with the standard LiOAc-based protocol [84]. Gene deletions were confirmed by colony PCR and Sanger sequencing to check the absence of endogenous genes. Primers and plasmids used for strain constructions are listed in S6 Table.

Fungal–bacterial co-culture CFU assay

All experiments began with bacterial or fungal cultures that were grown overnight. The starter cultures were diluted either 1:100 (for bacterial cultures) or 1:50 (for fungal cultures) in BHI and cultured for approximately 4 to 5 h to reach the log phase. Refreshed bacterial and fungal strains were added to BHI to reach a final bacterial cell density of 2.5 × 10⁴ cells/ml and a fungal cell density of 5 × 10⁵ cells/ml, for the co-culture experiment. The same amounts of bacterial or fungal cells were also added separately in fresh BHI, as monoculture controls. These cultures were incubated for 18 h at 30 °C or 37 °C with shaking. We chose 30 °C for the initial co-culture assays for 2 reasons. First, C. albicans cells reached higher CFU at 30 °C than 37 °C, which would impose a stronger competition with bacteria. Second, C. albicans cells form hyphae at 37 °C, which can have multiple cells in one filament and thus confound CFU measurements. We further confirmed that our findings of Mg²⁺ competition are independent of temperature by setting up co-culture assays at both 30 °C and 37 °C. Cultures were serially diluted and spotted on LB plates with 50 μg/ml nystatin for measuring bacterial growth and on YPD plates with 100 μg/ml gentamicin for measuring fungal growth.

PA4824 or mgtA complementation assay

To complement the deletion of mgtA, ΔmgtA was transformed with a genome-integrated copy of mgtA under its endogenous promoter (miniTn7-mgtA) [85] or a plasmid copy of mgtA under an arabinose-inducible promoter (pBAD) [86]. As PA4824 is in the same operon as mgtA, we used pBAD to express PA4824 to ensure its full expression. These deletion-complemented strains were prepared in the same way as the co-culture CFU assay, except that 0.2% arabinose was used to induce PA4824 or mgtA expression.

Bacterial–bacterial co-culture CFU assay

Overnight bacterial cultures were diluted 1:100 in BHI and cultured for approximately 4 to 5 h to reach log phase. Log-phase cultures of bacterial competitors (S. aureus, S. multivorans, E. coli, or S. Typhimurium) were added in fresh BHI at 5 × 10⁵ cells/ml as final cell density while 2.5 × 10⁴ cells/ml of log-phase P. aeruginosa culture was added. The same number of bacterial cells were also added separately in fresh BHI, as monoculture controls. These cultures were incubated for 18 h at 30 °C with shaking. Cultures were serially diluted and spotted on Pseudomonas Isolation Agar (Millipore-17208) to measure P. aeruginosa CFUs.

Measurements of bacterial intracellular Mg²⁺

Cytosolic Mg²⁺ concentrations in P. aeruginosa were measured using a modification of the previously reported Mg²⁺ genetic sensor assay [31]. First, we cloned the Mg²⁺ genetic sensor containing S. Typhimurium mgtA leader sequence (P_lac-mgtA_leader-lacZ) or the promoter-only fragment (P_lac-lacZ) into a P. aeruginosa replicating plasmid, pBBR [87]. Then, these 2 plasmids were transformed into the indicated P. aeruginosa strains separately, with the former acting as a reporter for Mg²⁺ and the latter acting as a control. Both strains were cultured in BHI or fungal spent media (80% v/v final concentration) containing 100 μg/ml gentamicin for 16 h and then diluted 1:100 in the corresponding media with gentamicin for 4 h to reach log phase. The cell density of these cultures was adjusted to OD600 = 0.4, and we followed the same protocol to measure β-galactosidase activity using Galacto-Light Plus System (Thermo Fisher). Finally, cytosolic Mg²⁺ concentrations were derived from the following equation, where the control strain was used to normalize transcription or translation efficiency between strains:

Colistin survival assay in C. albicans-spent media

To prepare C. albicans-free supernatant, a single C. albicans colony was inoculated in 10 ml BHI and cultured at 30 °C for 24 h. The fungal culture was centrifuged, and the supernatant was filtered with a Steriflip unit with a 0.22 μm Millipore Express PLUS membrane (MilliporeSigma), and 8 ml fungal filtrate was added to 2 ml fresh BHI (80% v/v final concentration) to prepare C. albicans-spent media, mimicking nutrient exhaustion by C. albicans. To perform colistin treatment, we first diluted overnight bacterial culture in 3 ml BHI or C. albicans-spent media at 1:100 for 4 h. Then, bacterial cultures were adjusted to OD600 = 0.3 in BHI or C. albicans-spent media. Each bacterial culture was treated with colistin at the indicated concentration, and the samples were incubated at 30 °C for 1.5 h. To assay bacterial viability, bacterial cultures after colistin treatment were serially diluted on LB plates. Bacterial cultures in the same growth condition without colistin treatment were used as a control. Bacterial survival was calculated by the ratio of bacterial CFUs, relative to those observed without treatment, at 1.5 h.

Colistin survival assay in co-culture

Log-phase bacterial cells were cultured in BHI alone or in co-culture with fungal cells in BHI for 18 h at 30 °C or 37 °C, as the co-culture CFU assay indicated. To equilibrate bacterial cell numbers in these 2 culturing conditions prior to the colistin treatment, the cell density of monoculture samples was adjusted to OD600 = 0.3 to match the bacterial cell density after 18-h growth in co-culture. Both equilibrated monoculture and co-culture samples were split into 1 ml in 2 different tubes for colistin-treated versus untreated conditions. Colistin was added to the first tube at the indicated concentration. Bacterial cells in both tubes were incubated at 30 °C for 1.5 h and the bacterial viability was determined by counting CFUs after serial dilution. Bacterial survival was determined by the ratio of bacterial CFUs upon colistin treatment, relative to without treatment (average of 4 replicates). Bacterial survival was set to 100% if it was over 100% due to stochastic variation in colony counts.

We used the same fungal protection assay to measure colistin resistance in co-culture-evolved populations. These evolved populations were split into 2 culturing conditions: BHI only, as co-culture conditions, or BHI + 50 μg/ml nystatin, to remove fungal cells (as monoculture conditions). Cells were grown for 18 h at 37 °C. After overnight growth, populations in both conditions were treated with 192 μg/ml colistin following the above protocol. To measure bacterial viability after colistin treatments, cells in co-culture and monoculture conditions were spotted on LB + 50 μg/ml nystatin or LB plates, respectively.

Competitive fitness assay

A WT PAO1 strain with chromosomally integrated miniTn7-mCherry was used as the reference strain in fitness competition with deletion mutants, either in BHI only or in co-culture with C. albicans. All bacterial and yeast strains were incubated in BHI at 30 °C to log phase, and cell density was quantified by OD600 using a spectrophotometer (biowave CO8000 Cell Density Meter). First, a non-fluorescent strain, either the PAO1 wild-type or deletion mutants, was mixed with the fluorescence-labeled reference strain at a ratio of 1:1 to make a bacterial mixed culture with 2 × 10⁶ cells/ml. Part of the cell mixture was used to determine the initial ratio of sample and reference strains using flow cytometry (BD FACSymphony A5 Cell Analyzer). Then, 10 μl of cell mixture was inoculated separately in 3 ml BHI alone (monoculture) or 3 ml BHI with 2 × 10⁵ C. albicans cells (co-culture) and grown in BHI at 30 °C for 18 h. After 18 h, monoculture samples were diluted 100-fold to measure the ratio of sample and reference strain. For co-culture samples, a low-speed spin (1,000xg for 3 min) was applied to separate bacterial and fungal populations. Supernatants enriched with bacterial cells were measured using flow cytometry. For each sample, at least 30,000 cells were collected and the FACS data were analyzed by the FlowJo 10.4.1 software. The reference strain was cultured separately to estimate the number of generations during the experiment. Each experiment was conducted in at least 2 biological and 2 technical replicates. To calculate the relative fitness, w, of each sample strain to the reference strain, we followed the formula: w = 1+s, where selection coefficient s = [ln(sample/reference)_t- in(sample/reference)_t=0]/t, where t = number of generations and (sample/reference) is the ratio between a sample strain and the reference strain [88].

RNA sequencing and analysis

A log-phase P. aeruginosa WT culture (OD600 = 0.1~0.2) was diluted at 5 × 10⁶ cells in 100 ml media alone or 100 ml media with 5 × 10⁷ C. albicans cells and grown at 30 °C. Cells were collected for extracting RNA at 8 h when there was no significant growth difference of P. aeruginosa between monoculture and co-culture. Prior to RNA extraction, a slow-speed spin was applied to co-culture samples to enrich bacteria in the supernatant. For each sample, total RNA was harvested from 10⁹ cells, preserved in RNA-protected QIAzol, and purified by RNAeasy mini kit (Qiagen) following the commercial protocol. RNA was quantified by Qubit fluorometric quantitation system (Thermo Fisher), and RNA integrity was examined by 4200 TapeStation System (Agilent). We generated an rRNA depleted library and obtained >20 million 150 bp pair-end Illumina NovaSeq reads for each sample commercially (Azenta). Reads were examined using FastQC (Barbaham Bioinformatics, Babraham Institute) and trimmed by Trim Galore! (v0.6.10; https://github.com/FelixKrueger/TrimGalore). Reads were mapped to the P. aeruginosa PAO1 genome (GCF_00006765.1) using Subread/FeatureCounts (https://github.com/ShiLab-Bioinformatics/subread). Differential gene expression was analyzed with DESeq2 [89] (https://bioconductor.org/packages/release/bioc/htmL/DESeq2.htmL), following the default pipeline of data normalization. A log₂ fold change cutoff of 2 and a false discovery rate (FDR) of 0.05 were used to filter significantly differentially expressed genes between samples from 3 biological replicates. Gene Ontology analysis was done using the P. aeruginosa STRING database on ShinyGO 0.77(http://bioinformatics.sdstate.edu/go/).

Sequencing files are stored in the BioProject PRJNA1021673 of the NCBI repository.

Transposon-sequencing and analysis

We used a pool of ~10⁵ transposon-insertion mutants, generated by Tn5-based transposon T8 (ISlacZhah-tc) insertion in the P. aeruginosa strain PAO1 [28], and 20 μl of frozen Tn mutants (approximately containing 10⁸ cells) was thawed and inoculated in 50 ml media or 50 ml media with 10⁸ C. albicans cells in a 250-ml flask, to establish monoculture and co-culture conditions, each with 3 biological replicates. To screen the fitness of Tn mutants effectively, cells were grown at 30 °C with shaking at 210 rpm for 10 h (approximately 10 generations) and collected for genomic DNA extraction using the DNAeasy Blood & Tissue kit (Qiagen). DNA samples were prepared for transposon sequencing following the previously published TdT/two-step PCR method [90]. In short, 2 μg DNA was end-repaired by NEB end-repair reaction and subsequently C-tailed in the TdT tailing reaction. Two PCR reactions were used to enrich DNA fragments spanning the junction of transposon insertion. PCR-1 was performed using a transposon-specific primer and a polyG primer (olj376) on 0.5 to 1 μg DNA; at least 2 reactions which were pooled in the following purification steps. For PCR-2, a mixture of Tn8 primers combined with 3N, 5N, 7N, and 9N random bases were used to increase the diversity of Tn-seq libraries. The size of library DNA was examined by 4200 High Sensitivity DNA TapeStation System (Agilent) and quantified using qPCR with standard Illumina primers (Kapa Biosystems). Libraries were pooled and sequenced commercially using an Illumina NovaSeq or NextSeq sequencer (Novogene). For each sample, at least 25 million pair-end 150 bp sequencing reads were generated after which, we proceeded for downstream analysis. Sequencing files are stored in the BioProject PRJNA1021673 of the NCBI repository.

The analysis scripts used in this study are adapted from previously published protocols and are available in the GitHub repository (https://github.com/PhoebeHsieh-yuying/P.-aeruginosa_Tnseq_paper) and the Zenodo repository (DOI: 10.5281/zenodo.11404260). Briefly, sequencing reads were groomed by FastQC (Barbaham Bioinformatics, Babraham Institute), filtered to keep those with the correct Tn8 sequence (TATAAGAGTCAG) near the 5′ end using Cutadapt [91] and mapped to the P. aeruginosa PAO1 genome (GCF_00006765.1) with Bowtie2 [92] using the Tnseq2.sh script. The output table with transposon insertion site assignments of each sample was concatenated and annotated with known gene features of P. aeruginosa using a custom R script, Tnseq_reformat.R. Then, DESeq2 [89] statistical package in R was used to identify genes with significant changes in Tn insertion between co-culture and monoculture. A log₂ fold change cutoff of 1 and an FDR of 0.05 were used as the criteria for significance.

Experimental evolution of colistin survival

Eight independent colonies of a WT P. aeruginosa strain PAO1 were used as ancestral strains for the experimental evolution experiment. Each ancestral clone was inoculated and grown in BHI at 37 °C to 10⁹ cells/ml. At the beginning of the experiment, approximately 10⁷ cells of each ancestral population were transferred either into 1 ml BHI with colistin or into 1 ml BHI with colistin and 10⁶ C. albicans cells to select colistin resistance. Each day, populations were passaged at a ratio of 1:200 in the former condition, while co-cultured populations were passaged at 1:50 in colistin-containing BHI to maintain comparable population size between both conditions. The concentration of colistin was started from 1.5 μg/ml, gradually increased 2-fold every week or every 2 weeks until it reached 192 μg/ml at the end of the 90-day evolution experiment. As controls for adaptation to BHI only or co-culture with C. albicans, the same ancestral populations were passaged at a ratio of 1:200 into 1 ml BHI or 1 ml BHI with the same amount of C. albicans cells, but without colistin. After every 7 days, 1 ml evolving cultures were mixed with 500 μl 80% glycerol and frozen at −80 °C. On day 90, we expected that populations evolved in colistin-containing BHI with C. albicans had been passaged for approximately 600 generations, while the rest of the evolved populations had been passaged for approximately 700 generations.

Whole-genome sequencing and analysis

We sequenced 2 ancestral populations and all 8 evolved populations per treatment. We revived each population from a freezer stock in the growth condition under which they evolved and grew for 24 h. For co-culture evolved populations, we treated them with nystatin to remove fungal cells, and 3 ml bacterial culture was collected for DNA extraction using the DNAeasy Blood & Tissue kit (Qiagen). Sequencing libraries were made and sequenced commercially by Illumina sequencers in SeqCenter (https://www.seqcenter.com/). Variant calling was done using the breseq [93] software v0.37.1 and the P. aeruginosa PAO1 genome (GCF_00006765.1) was used as the reference genome. The average depth of sequencing for populations was 222.5 fold (+/− 32.7) and the average genome coverage was 98.8 fold (+/− 0.04). Variants at >95% in each population were filtered for mutation calling. Mutations were manually curated by identifying unique variants in each evolved population compared to the ancestor.

Sequencing files are stored in the BioProject PRJNA1021673 of the NCBI repository.

Mg²⁺ measurements in media

Mg²⁺ concentrations were measured using the Magnesium Assay kit (Sigma-Aldrich) according to the protocol provided. Absorbances at OD₄₅₀ after the enzymatic reaction were compared to a standard curve of Mg²⁺ concentration to determine the absolute concentration in media.