The Quorum-Sensing Inhibitor Furanone C-30 Rapidly Loses Its Tobramycin-Potentiating Activity against Pseudomonas aeruginosa Biofilms during Experimental Evolution

ABSTRACT

The use of quorum-sensing inhibitors (QSI) has been proposed as an alternative strategy to combat antibiotic resistance. QSI reduce the virulence of a pathogen without killing it and it is claimed that resistance to such compounds is less likely to develop, although there is a lack of experimental data supporting this hypothesis. Additionally, such studies are often carried out in conditions that do not mimic the in vivo situation. In the present study, we evaluated whether a combination of the QSI furanone C-30 and the aminoglycoside antibiotic tobramycin would be “evolution-proof” when used to eradicate Pseudomonas aeruginosa biofilms grown in a synthetic cystic fibrosis sputum medium. We found that the biofilm-eradicating activity of the tobramycin/furanone C-30 combination already decreased after 5 treatment cycles. The antimicrobial susceptibility of P. aeruginosa to tobramycin decreased 8-fold after 16 cycles of treatment with the tobramycin/furanone C-30 combination. Furthermore, microcalorimetry revealed changes in the metabolic activity of P. aeruginosa exposed to furanone C-30, tobramycin, and the combination. Whole-genome sequencing analysis of the evolved strains exposed to the combination identified mutations in mexT, fusA1, and parS, genes known to be involved in antibiotic resistance. In P. aeruginosa treated with furanone C-30 alone, a deletion in mexT was also observed. Our data indicate that furanone C-30 is not “evolution-proof” and quickly becomes ineffective as a tobramycin potentiator.

INTRODUCTION

Pseudomonas aeruginosa is an opportunistic pathogen that is listed as a “Priority Pathogen” by the World Health Organization (1). This organism is involved in various (nosocomial) infections, with respiratory tract, urinary tract, bloodstream, and wound infections being most frequent. One of the major concerns is its multidrug resistance (MDR); P. aeruginosa has been shown to possess several resistance mechanisms, including antibiotic-degrading enzyme production, target modification, and a reduced influx or increased efflux of antibiotics (2).

Colonization of the lungs of cystic fibrosis (CF) patients with bacterial pathogens, often forming biofilm aggregates, results in recurring inflammation and destruction of lung tissue, which can ultimately lead to respiratory failure (3). The conditions in the CF lung are very different from those in the lungs of non-CF individuals, and CF lung mucus is enriched in carbohydrates, lipids, DNA, and amino acids. This leads to a thick, stagnant mucus layer that provides the perfect environment for bacterial colonization (4, 5). P. aeruginosa is a key pathogen in respiratory tract infections in CF, and is associated with increased morbidity and mortality (6, 7). In the lungs of chronically infected CF patients, nonmucoid strains of P. aeruginosa typically acquire mutations in the mucA gene that lead to the continuous production of the exopolysaccharide alginate and a mucoid phenotype (8). The viscous mucus and alginate form a matrix in which P. aeruginosa grows as microcolonies, which protects against host immune defenses and antibiotics and, as a consequence, antibiotic therapy often fails to eradicate P. aeruginosa from the CF lung (9, 10).

Previous research has shown that quorum-sensing inhibitors (QSI) can increase the antimicrobial susceptibility of biofilm-grown pathogens (11–13). Quorum sensing (QS) allows microorganisms to synchronize their behavior based on cell-density and is based on the exchange of signaling molecules (14). Since the inhibition of QS does not lead to cell death, the general assumption is that QSI do not induce selective pressure. As a consequence, the development of resistance to QSI would be much slower or even absent (15). However, recent studies cast doubt on this hypothesis. For example, resistance against the QSI furanone C-30 (C-30) due to efflux and a reduced permeability was observed in clinical P. aeruginosa isolates recovered from CF patients (16, 17). In addition, a recent experimental evolution study showed that the tobramycin-potentiating activity of the QSI baicalin hydrate against Burkholderia cenocepacia is rapidly lost (18).

The goal of the present study is to evaluate the evolutionary robustness of the QSI C-30 for the treatment of P. aeruginosa biofilm infections. To this end, we repeatedly exposed P. aeruginosa biofilms to C-30 (with or without tobramycin) in a relevant in vitro model and characterized the genotype and phenotype of the evolved lineages.

RESULTS AND DISCUSSION

Growth rate and microcolony formation of P. aeruginosa PAO1 in SCFM2.

In order to study the behavior of P. aeruginosa under conditions that mimic those encountered in the CF lung, synthetic cystic fibrosis medium 2 (SCFM2) was used. The concentration of amino acids, ions, and sugars in SCFM2 is based on the analysis of CF sputum samples and the medium is supplemented with molecules that are also present in CF sputum and play an important role in viscosity and microcolony formation (including DNA, dioleoylphosphatidylcholine, N-acetylglucosamine, and mucin) (19, 20). We first confirmed growth and biofilm/microcolony formation of P. aeruginosa strain PAO1 in SCFM2. The doubling time during exponential phase was 1.25 h, which is in line with the 1.3 h previously reported (4), and the stationary phase was reached after approximately 12 h (Fig. S1). The maximal density reached was approximately 10⁹ CFU/ml.

After 6 h of growth in SCFM2, P. aeruginosa PAO1 already had formed aggregates ranging in size from approximately 50 to 100 μm in diameter, increasing in size to up to 2 mm after 48 h (Fig. 1). In contrast to that, no aggregates were observed in cultures in LB medium.

FIG 1.

Microcolonies observed in the supernatant of a culture of P. aeruginosa PAO1 in SCFM2 after 6 h (left) and 48 h (middle). After 48 h growth in LB, no microcolonies could be observed (right).

A green fluorescent protein (GFP)-expressing derivative of P. aeruginosa PAO1 was used to further quantify microcolony formation using flow cytometry. By measuring the GFP signal, living cells can be distinguished from dead cells, and microcolonies can be distinguished from macromolecules present in the medium. The scatterplot from the culture grown in SCFM2 showed two populations: a planktonic population mainly consisting of single cells (or small aggregates) and a second population characterized by a higher side and forward scatter, corresponding to larger biofilm aggregates (4.29 ± 3.06% of all particles detected). After growth in LB medium, only a single population of cells was observed on the scatterplot, with only 0.05 ± 0.01% of all particles identified as aggregates (P < 0.05). A scatterplot of a representative experiment is shown in Fig. S2. Our data clearly show that the growth of P. aeruginosa in SCFM2 is distinct from the growth in LB and that SCFM2 promotes the formation of biofilm aggregates in the supernatant.

Experimental evolution shows the potentiating activity of the QSI C-30 is quickly lost.

In preliminary experiments, we determined the optimal concentrations of the test compounds. The concentration of the compound QSI C-30 was chosen so that it did not affect the number of CFU when used alone. Second, a concentration of tobramycin was chosen that, when used alone, markedly reduced the number of CFU in the biofilm without eradicating them completely. These preliminary experiments indicated that C-30 had no effect on the number of CFU in the biofilm in a concentration of up to 200 μg/ml. Tobramycin alone decreased the number of cells recovered from the biofilm by approximately 92% and 98% when 10 μg/ml (P = 0.001) and 20 μg/ml (P < 0.001) were used, respectively (Fig. 2). When tobramycin and C-30 were combined, the number of CFU in the biofilm was significantly lower than in the biofilms treated with tobramycin alone (Fig. 2). This confirms that the QSI C-30 increases the susceptibility of P. aeruginosa biofilms to tobramycin, without having a noticeable antibiofilm effect on its own, which is in line with previous findings (13, 21, 22).

FIG 2.

Number of culturable cells (CFU/ml) recovered from a 24-h P. aeruginosa PAO1 biofilm grown in SCFM2 after exposure to different concentrations of tobramycin, C-30, or a combination of both. Errors bars represent the standard deviation (n = 3). *, indicates a significant difference compared to the control (P < 0.001); **, indicates a significant difference compared to tobramycin alone (P < 0.05).

To determine whether the potentiating effect of C-30 is stable over time, we performed an experimental evolution study using three independent lineages for each condition. To analyze the data, a simple effects model was designed with “number of CFU” as the dependent variable and “treatment cycle” and “treatment” as independent variables. This model allowed us to make a pairwise comparison, at each cycle, between the number of CFU after treatment with the tobramycin/C-30 combination and after treatment with tobramycin alone. Overall, no meaningful differences could be observed between the three lineages. At cycle 5, the potentiating effect of C-30 was lost, as the number of CFU after treatment with the tobramycin/C-30 combination no longer differed from the number of CFU after treatment with tobramycin alone (Table S1). Moreover, the overall eradicating effect of the combination treatment was also lost starting from cycle 5, as the number of CFU after exposure to the tobramycin/C-30 combination at that treatment cycle became significantly higher than the number of CFU after the first treatment (Table S2), and did not differ from the untreated control (Fig. 3A and D). The susceptibility to tobramycin alone also changed during evolution, but varied slightly between lineages. Starting from cycle 14, treatment with tobramycin led to significantly less reduction in the number of CFU in all lineages, likely due to an increased tolerance or resistance to tobramycin (Table S3). No change in the number of CFU was observed in P. aeruginosa PAO1 biofilms treated with C-30 alone or in the untreated control group (Fig. 3), as was expected.

FIG 3.

The number of CFU/ml recovered after treatment of P. aeruginosa PAO1 biofilms with the tobramycin/C-30 combination (A), tobramycin (B), or C-30 alone (C) compared to an untreated control (D). The different curves represent three independent lineages of the evolution experiment (L1, L2, and L3).

Resistance to tobramycin develops faster in the presence of C-30 during evolution.

To assess the effect of repeated biofilm treatment on the susceptibility of P. aeruginosa PAO1 to tobramycin, the MIC_SCFM2 was determined. As shown in Table 1, the MIC_SCFM2 of tobramycin for all independent lineages that were exposed to the combination therapy increased 8-fold during the evolution experiment. Exposure to tobramycin alone resulted in a 4-fold increase of the MIC_SCFM2 for all lineages. The reduced susceptibility to tobramycin occurred earlier (cycle 5) in the biofilms treated with the combination than in the biofilms treated with tobramycin alone (cycle 10 or 16). This is not surprising if we take into account that the reduction in the number of CFU was larger after treatment with the combination therapy, which likely indicates stronger selection toward development of resistance to tobramycin. The concentration of tobramycin in sputum of CF patients is variable, but concentrations after inhalation therapy are typically 5- to 100-fold higher than the 20 μg/ml used in the present study (23, 24). Whether resistance would develop at the same rate in vivo as observed in our in vitro study remains to be determined.

TABLE 1.

Median (n = 3) MIC_SCFM2 of tobramycin (μg/ml) for the different P. aeruginosa PAO1 lineages

Lineage + treatmenta	Cycle 0	Cycle 5	Cycle 10	Cycle 16
L1 TOB	10	10	20	40
L2 TOB	10	10	10	40
L3 TOB	10	10	40	40
L1 TOB + C-30	10	40	40	80
L2 TOB + C-30	10	80	80	80
L3 TOB + C-30	10	80	40	80
L1 C-30	10	10	10	10
L2 C-30	10	10	10	10
L3 C-30	10	10	10	10

^aL1, lineage 1; L2, lineage 2; L3, lineage 3; TOB, treatment with tobramycin alone; C-30, treatment with the QSI C-30 alone; TOB + C-30, treatment with both together.

Metabolic activity is altered after exposure to C-30, tobramycin, and the combination of both.

The calScreener device was used to continuously (1 measurement/sec) measure the heat flow of the P. aeruginosa PAO1 biofilms grown in SCFM2. Previous studies have demonstrated that the calScreener can detect small changes in metabolism that are not necessarily reflected in differences in biomass of the samples or number of CFU (25, 26). A summary of the data obtained is shown in Fig. 4. Thermograms, the time to peak, the maximum metabolic rate, and the maximum metabolic decay velocity for each lineage separately can be found in Fig. S3. We observed that in evolved populations that had been treated with C-30, the maximum metabolic rate was reached earlier than in the treated wild-type (WT) P. aeruginosa PAO1 (P ≤ 0.001); this became apparent from cycle 5. For evolved populations repeatedly exposed to the tobramycin/C-30 combination (from cycle 5 onward) or to tobramycin alone (from cycle 10 onward), the maximum metabolic rate was higher than in the treated WT (P < 0.05); this maximum was also reached earlier (P < 0.001). These microcalorimetry data clearly show that the metabolic activity of the evolved lineages in response to the different treatments differs from that of the WT. Metabolic adaptations in biofilms have been shown to contribute to a reduced antimicrobial susceptibility (27), and the microcalorimetric changes detected in the treated biofilms of the evolved P. aeruginosa populations in the present study likely contribute to the reduced susceptibility toward tobramycin and the tobramycin/furanone C-30 combination.

FIG 4.

Average time to peak activity (left) and maximum metabolic rate (right) of the biofilms of all lineages after treatment with the tobramycin/C-30 combination (TOB + C-30), tobramycin alone (TOB), C-30, or the untreated control. Bars show data at the start of the experiment (no fill), after 5 (dots), 10 (diagonal stripes), and 16 (black) cycles. Error bars indicate the standard deviation (n = 6); *, indicates a significant difference compared to the treated WT P. aeruginosa PAO1 (P < 0.05).

Whole-genome sequencing.

The genomes of the P. aeruginosa PAO1 populations were sequenced at the start of the experiment and after 5, 10, and 16 treatment cycles. A complete overview of all genes that acquired mutations (with a mutation frequency cutoff of 10%) in the treated evolved populations can be found in Table S4, and their genomic location is shown in Fig. S4. A summary of the most abundant mutations (mutation frequency > 50%) and their corresponding amino acid changes are shown in Table 2. No mutations in coding regions were found in the untreated control lineages. Genes with the most mutations were mexT and fusA1. Mutations in mexT were found in multiple lineages evolved in the presence of C-30 and the tobramycin/C-30 combination, while mutations in fusA1 were found when bacteria were exposed to tobramycin and the tobramycin/C-30 combination. Disruptions in the parS and ptsP genes were detected in P. aeruginosa exposed to tobramycin alone (parS and ptsP) and exposed to the tobramycin/C30 combination (parS). A few genes that could not be linked to antimicrobial resistance showed mutations in only one or a few lineages; these included rne (encoding RNase E) (28), rpsE (encoding protein S5 of the 30S ribosomal subunit) (29), htpG (encoding a heath shock/chaperone protein) (30), and dipA (encoding a phosphodiesterase) (31).

TABLE 2.

Nonsynonymous mutations detected in the evolved P. aeruginosa PAO1 lineages (with mutation frequency > 50%) after 5, 10, and 16 treatment cycles

Gene	Lineage; treatment; cycle(s)a	Locationb	Mutation type	Amino acid change
mexT (PA2492)	L1; C-30; 5, 10, 16	2807694	Deletion of 8 bp	NA
	L2; C-30; 5, 10, 16	2807694	Deletion of 8 bp	NA
	L3; C-30; 10, 16	2807694	Deletion of 8 bp	NA
	L1; TOB + C-30; 16	2807694	Deletion of 8 bp	NA
	L2; TOB + C-30; 10, 16	2807694	Deletion of 8 bp	NA
	L3; TOB + C-30; 16	2807694	Deletion of 8 bp	NA
fusA1 (PA4266)	L1; TOB + C-30; 16	4769700	Point mutation G to A	Pro → Ser 486
	L2; TOB + C-30; 10, 16	4769267	Point mutation T to C	Tyr → Cys 630
	L3; TOB + C-30; 10, 16	4769118	Point mutation G to A	Arg → Cys 680
	L1; TOB; 10, 16	4769118	Point mutation G to A	Arg → Cys 680
	L2; TOB; 16	4769370	Point mutation A to G	Phe → Lys 596
	L3; TOB; 16	4769118	Point mutation G to A	Arg → Cys 680
parS (PA1798)	L1; TOB; 10, 16	1951521	Point mutation A to G	Trp → Arg 69
	L3; TOB; 16	1951521	Point mutation A to G	Trp → Arg 69
	L1; TOB + C-30; 516	1951521	Point mutation A to G	Trp → Arg 69
	L3; TOB + C-30; 5, 10, 16	1951521	Point mutation A to G	Trp → Arg 69
htpG (PA1596)	L2; TOB + C-30; 16	1737860	Deletion of 134 bp	NA
rne (PA2976)	L1; TOB + C-30; 16	3332890	Point mutation A to G	Met → Val 4
	L3; TOB + C-30; 16	3335073	Insertion of 2 bp	NA
ptsP (PA0337)	L3; TOB; 16	380551	Point mutation C to T	Pro → Ser 625
rpsE (PA4246)	L1; TOB + C-30; 16	4759497	Point mutation G to C	Thr → Ser 25
dipA (PA5017)	L1; C-30; 16	5643432	Point mutation T to A	Leu → Gln 808

^aL1, lineage 1; L2, lineage 2; L3, lineage 3; TOB, treatment with tobramycin alone; C-30, treatment with the QSI C-30 alone; TOB + C-30, treatment with both together.

^bLocation of the mutation in the P. aeruginosa PAO1 reference genome.

Deletion in mexT results in a reduced susceptibility to C-30.

Cells recovered from P. aeruginosa PAO1 biofilms treated with both the tobramycin/C-30 combination and with C-30 alone had a deletion of 8 bp in the mexT gene, suggesting this mutation is linked to the exposure to C-30. The mutation could be detected in all lineages with a high frequency (>75% at cycle 16, Table S4). MexT is a LysR-type transcriptional activator of the mexEF-oprN operon (32). There is a wide variety of P. aeruginosa PAO1 strains of which some have an insertion of GGCCAGCC in mexT. This 8-bp insertion originates from a duplication and results in a premature stop codon (Fig. S5) (32–34). The sequence of the WT P. aeruginosa PAO1 that was used in this experiment does, indeed, contain the duplicated GGCCAGCC sequence. The 8-bp deletion of CGGCCAGC that is present in the mexT gene of all lineages treated with C-30 (either alone or in combination with tobramycin) shifts the reading frame and allows translation of the full mexT sequence, which can then act as a transcriptional activator of MexEF (32).

MexEF is a multidrug efflux pump that exports fluoroquinolones, chloramphenicol, and trimethoprim and has not yet been linked to reduced susceptibility to aminoglycosides (35–37). In line with the latter observation, we did not observe differences in the MIC_SCFM2 of tobramycin between WT and evolved strains with a mexT deletion (MIC values of 10 to 20 μg/ml, Table 1). In addition, the tobramycin susceptibility of a mexEF deletion mutant (PAO1 background) did not differ from that of the WT PAO1 strain (10 μg/ml) and no meaningful difference in biofilm-eradicating activity of tobramycin was observed between both strains (Fig. S6). In conclusion, as WT PAO1, a PAO1 mexEF deletion strain, and a PAO1 mutant with a deletion in mexT (leading to expression of mexEF) all have the same tobramycin susceptibility under various conditions, we can rule out a role for MexEF in tobramycin efflux under the conditions of our study. As an indirect measure of increased MexEF-OprN activity, and to confirm that this is due to a deletion in mexT, we compared the MIC_SCFM2 of the fluoroquinolone antibiotic ciprofloxacin (a substrate of MexEF-OprN) between WT P. aeruginosa and the evolved populations treated with C-30 alone (i.e., lineages that had acquired the mexT mutation). The MIC_SCFM2 of ciprofloxacin was 8-fold higher for the evolved lineages 8 μg/ml compared to 1 μg/ml for the WT), strongly suggesting that the mexT deletion indeed leads to an increased expression of MexEF.

Mutations in mexT were not observed in biofilms treated with tobramycin alone, while biofilms exposed to C-30 alone (after cycle 5) and the tobramycin/C-30 combination (after cycle 10 or 16) had acquired this mutation, suggesting this mutation is linked to the exposure to C-30. Possibly, the mutation in mexT allows cells to export C-30 by means of activating the transcription of mexEF. In previous work, efflux of C-30 in P. aeruginosa PA14 has been attributed to activity of the MexAB-OprM efflux pump (17).

MexAB is an efflux pump similar to MexEF, and mutations in its transcriptional repressors mexR and nalC lead to an increased expression of mexAB (17). Interestingly, in the present study a point mutation was detected in nalD, another transcriptional repressor of MexAB, in lineage 1 treated with the tobramycin/C-30 combination (Table S4) (mutation frequency of 13.4% after 10 treatment cycles). Because this mutation was not detected after cycle 16, we cannot conclude that it contributes to the reduced potentiating activity of C-30 in the present study, but its presence indicates that a second efflux pump (besides MexEF) could potentially be activated after repeated exposure to tobramycin and C-30. Additional studies will be required to determine the exact contribution of MexEF-OprN and MexAB-OprM to C-30 efflux in P. aeruginosa PAO1.

Genes involved in tobramycin resistance.

Different point mutations in fusA1 were detected in P. aeruginosa PAO1 treated with tobramycin and the tobramycin/C-30 combination (Table 2). Mutations were not detected in populations treated with C-30 alone, indicating that this mutation is linked to tobramycin. In P. aeruginosa, fusA1 together with fusA2 encodes elongation factor G (EF-G), which is essential for translation (38). EF-G catalyzes the translocation of the transcript during elongation. In addition, EF-G facilitates the disassembly of the ribosome upon binding of the ribosome recycling factor, which leads to the release of the ribosomal subunits (38, 39). Amino acids targeted by aminoglycoside antibiotics are located on the small ribosomal subunit and are involved in this recycling process. When the aminoglycoside antibiotic binds to the ribosome, ribosome recycling is blocked, inhibiting protein synthesis (40). In multiple recent studies, the presence of point mutations in the fusA1 gene of aminoglycoside-resistant P. aeruginosa strains was described (38, 40, 41). The point mutations in the fusA1 gene found in this study result in changes of single amino acids (Fig. S7), which could disable the interaction of the recycling complex with tobramycin.

A point mutation could be detected in the parS gene of P. aeruginosa treated with tobramycin and the tobramycin/C-30 combination (in lineages 1 and 3). ParS is the response regulator of the ParS/ParR system, and mutations in this system have been shown to cause multidrug resistance (42). A recent evolutionary study also showed that P. aeruginosa isolates repetitively exposed to tobramycin had mutations in both parR and parS (43).

P. aeruginosa PAO1 lineage 3 exposed to tobramycin alone had a point mutation in the ptsP gene that occurred in cycle 16. This gene encodes the phosphoenolpyruvate-protein-phosphotransferase and has recently been connected to tobramycin resistance in clinical isolates, as well as in in vitro evolved P. aeruginosa populations, although the exact mechanism remains to be determined (44, 45).

As mentioned, the MIC_SCFM2 for tobramycin of P. aeruginosa biofilms treated with tobramycin alone had increased 4-fold by cycle 16 and we speculate this can be attributed to the point mutations in the fusA1 gene. The faster decline in susceptibility (MIC_SCFM2 increased 4- to 8-fold by cycle 5) of the P. aeruginosa biofilms treated with the tobramycin/C-30 combination seems to be linked to an early mutation in the parS gene. Surprisingly, no mutations could be detected in P. aeruginosa PAO1 lineage 2, exposed to the tobramycin/C-30 combination for 5 cycles (Table S4), while the MIC_SCFM2 had already increased 8-fold by then (Table 1). Additionally, the metabolic profile of this population had already changed at cycle 5 compared to the starting culture (Fig. S3). The fact that we could not find any mutations but did observe an altered metabolism and reduced susceptibility suggests that changes in gene expression are at the basis of our observations.

In conclusion, our data demonstrate that repeated exposure of P. aeruginosa biofilms to a combination of the QSI C-30 and tobramycin in an environment that mimics the physicochemical environment of the lungs of CF patients rapidly results in a loss of the tobramycin-potentiating activity of C-30. Exposure to C-30 alone did not affect the number of CFU, but still induced mutations and led to an altered metabolic profile, suggesting that exposure to C-30 does cause selective pressure on P. aeruginosa. These findings highlight the need to carefully evaluate combination therapy with QSI (or any other potentiator) for the development of resistance.

MATERIALS AND METHODS

Strains, culture conditions, and chemicals.

Overnight cultures of P. aeruginosa PAO1, of a mexEF operon deletion mutant (46), and of a derivative expressing green fluorescent protein (GFP) (47) were statically grown for 18 h in Luria Bertani broth (LB; Lab M, Moss Hall, UK) in aerobic conditions at 37°C. Pure cultures and serial dilutions of P. aeruginosa were grown on tryptic soy agar (TSA; Lab M, Moss Hall, UK). Tobramycin was obtained from TCI Europe (Zwijndrecht, Belgium) and stock solutions of 2 mg/ml were prepared in milliQ water. C-30 was obtained from Sigma-Aldrich (Bornem, Belgium) and stock solutions of 40 mg/ml were prepared in dimethyl sulfoxide (DMSO) (Alfa Aesar, Germany). All stock solutions were aliquoted and stored at −20°C. Synthetic cystic fibrosis medium 2 (SCFM2) was prepared as described previously (19) with the modification that the mucin was autoclaved rather than sterilized using UV.

Biofilm formation in SCFM2.

An overnight culture of P. aeruginosa was centrifuged (5,000 rpm, 5 min) and resuspended in SCFM2 to a density of approximately 5 × 10⁷ CFU/ml. This bacterial suspension (100 μl) was added to a flat-bottom 96-well plate and incubated at 37°C for 24 h.

Determination of growth curve in SCFM2.

An overnight culture of P. aeruginosa was centrifuged (5,000 rpm, 5 min) and resuspended in SCFM2 to a density of approximately 5 × 10⁷ CFU/ml. Of this culture, 200 μl was added to a flat-bottom 96-well plate and statically incubated at 37°C. Three wells were included per time point. At each time point (at the start of the experiment and after 1, 2, 4, 6, 12, 16, 24, 48, and 72 h) the number of CFU was determined by serial 10-fold dilution and plating. To disintegrate the microcolonies that were formed, the samples were vortexed and sonicated, both for 5 min prior to dilution.

Microscopy.

To characterize the growth phenotype of P. aeruginosa, cultures in SCFM2 and LB were set up as described above for the growth curve. After incubation, the sample was diluted 1/100 in 0.9% (wt/vol) NaCl (physiologic saline, PS) and subsequently visualized using an EVOS FL Auto cell imaging system (Thermo Fisher Scientific, Waltham, USA).

Flow cytometry.

An overnight culture of a GFP-producing P. aeruginosa PAO1 strain was diluted and resuspended as described above for the growth curve. After incubation for 24 h, the culture was diluted 1/100 in filtered PS (PES 0.2 μm, VWR, Haasrode, Belgium), and analyzed on a flow cytometer (Attune 113NxT, Thermo Fisher Scientific). The GFP signal was analyzed by excitation at 488 nm and detection through a 530/30 bandpass filter (48).

Evolution experiment.

P. aeruginosa biofilms were grown in SCFM2 as described above and treated after 24 h. The stock solutions of tobramycin and C-30 were diluted in fresh SCFM2 medium and 100 μl was added to the biofilm to obtain a final concentration of 20 μg/ml and 100 μg/ml, respectively. For the negative control, 100 μl of fresh SCFM2 was added, containing the same amount of DMSO (0.25%) as for the biofilms treated with C-30. After 24 h of static incubation at 37°C, biofilms were first sonicated (5 min) and vortexed (5 min, 900 rpm) in order to disintegrate the biofilm aggregates. The samples were subsequently serially diluted and plated on TSA. A subsample from the treated biofilm was used to prepare a new overnight culture, in order to start a new cycle. After each cycle, an aliquot of the biofilm was stored at −80°C in Microbank vials (Prolab Diagnostics, Richmond Hill, Canada) to allow further tests on the evolved populations. For each treatment, three independent lineages were established, which were each exposed for 16 cycles. The process is illustrated in Fig. S8.

Evaluation of the antimicrobial susceptibility to tobramycin.

The MIC of tobramycin in SCFM2 (MIC_SCMF2) for the evolved lineages was determined at different time points and compared to the MIC_SCFM2 for the WT P. aeruginosa PAO1 strain. The protocol was based on the EUCAST broth microdilution assay (http://www.eucast.org) but Mueller-Hinton (MH) medium was replaced by SCFM2. A 2-fold serial dilution of tobramycin was prepared in a flat-bottom 96-well plate with a range of 40 to 0.625 μg/ml. The plates were inoculated with 1 × 10⁵ CFU/ml P. aeruginosa and incubated under aerobic conditions at 37°C during 24 h. Due to the high turbidity of SCFM2, accurate readings based on optical density were not possible, and the number of cells was determined by plate counts. The MIC_SCFM2 was defined as the lowest tobramycin concentration that kept the number of CFU/ml at or below 10⁵ (i.e., the inoculum size).

Characterization of metabolic activity with isothermal microcalorimetry.

To measure heat production as a result of metabolic activity in biofilms, the calScreener microcalorimeter (Symcel, Stockholm, Sweden) was used. Biofilms were grown and treated as described above, with the exception that they were grown in plastic inserts that fit in the titanium cups of the calScreener. After starting the treatment, the plastic inserts were transferred into the titanium cups, which were placed in a 48-well plate and inserted in the instrument. The heat flow was measured at 37°C for 48 h and the resulting data were analyzed using the CalView software (Symcel).

DNA extraction.

Overnight cultures were prepared in LB and 3 ml of the cultures with OD₅₉₀ = 2 was centrifuged and the pellet was resuspended in 200 μl Tris-EDTA buffer (1 mM EDTA, 10 mM Tris-HCl, pH 8.5; Gibco). This suspension (100 μl) was added to 500 μl of lysis buffer (50 mM Tris-HCl [pH 8], 70 mM EDTA [pH 8], 1% sodium dodecyl sulfate; Sigma), containing 0.5 μg/ml pronase (Roche, Mannheim, Germany) and acid-washed glass beads (Sigma). After vigorous vortexing for 10 s, the samples were incubated at 37°C for 4 min. Next, 400 μl of saturated ammonium acetate was added and the samples were vortexed for 10 s prior to centrifugation (2 min, 13,000 rpm). Subsequently, 600 μl of chloroform was added and the samples were vortexed for another 10 s. The phases were separated by centrifugation (5 min, 13,000 rpm) and 400 μl of the top aqueous phase was transferred to a DNase-free Eppendorf tube (Lobind tube, Eppendorf, Aarschot, Belgium) containing 1 ml of 100% ethanol. After centrifugation (5 min, 13,000 rpm) the pellet was washed with 500 μl of 70% ethanol and dried in a SpeedVac (Thermo Fisher Scientific, USA) for 10 min. Finally, the extracted DNA was dissolved in 100 μl of low EDTA-Tris buffer (0.1 mM EDTA, 10 mM Tris-HCl [pH 8.5]) containing 0.5 μg/ml RNase (Qiagen, Venlo, The Netherlands) and was incubated at 37°C for 1 h. The DNA was subsequently quantified using a BioDrop μLITE (BioDrop, Cambridge, UK).

Whole-genome sequencing and data analysis.

Libraries were prepared using an in-house adapted protocol of the NEB prep kit and sequenced with the Illumina Novaseq6000, PE150. The resulting 150 bp paired reads were analyzed with the CLC Genomics Workbench (Qiagen, Aarhus, Denmark) and mapped to the P. aeruginosa PAO1 reference genome with accession number NC_002516 (49). For the mapping, a length fraction cutoff of 0.5 and a similarity fraction cutoff of 0.8 were maintained with a priority order of (i) mismatch score, (ii) mismatch cost, and (iii) insertion and deletion cost. Basic variant detection was performed with a minimum frequency of 10%. Mutations that were also present in the P. aeruginosa PAO1 starting culture were manually excluded and each mutation was manually checked for false positives originating from incorrect mapping.

Statistical analysis.

Statistical analysis was performed using the SPSS software (version 27, IBM, New York, US). The normal distribution of the data was verified with a Shapiro-Wilk test. Independent sample t tests were used to evaluate differences between the two treatment groups. If normality could not be assumed, nonparametric Mann-Whitney U tests were performed. Comparisons between multiple groups were performed using one-way ANOVA with Bonferroni correction. A simple effects model was utilized to compare the effect (CFU/ml) of the combination therapy with tobramycin alone, pairwise for every time point during the evolution experiment.

Data availability.

The mapped and raw reads generated in this study are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-10124.