Effect of cardiolipin on the antimicrobial activity of a new amphiphilic aminoglycoside derivative on Pseudomonas aeruginosa

Amphiphilic aminoglycoside derivatives are promising new antibacterials active against Gram-negative bacteria such as Pseudomonas aeruginosa, including colistin resistant strains. In this study, we demonstrated that addition of cardiolipin to the culture medium delayed growth of P. aeruginosa, favored asymmetrical growth and enhanced the efficiency of a new amphiphilic aminoglycoside derivative, the 3’,6-dinonylneamine. By using membrane models mimicking P. aeruginosa plasma membrane composition (POPE:POPG:CL), we demonstrated the ability of 3’6-dinonylneamine to induce changes in the biophysical properties of membrane model lipid systems in a cardiolipin dependent manner. These changes include an increased membrane permeability associated with a reduced hydration and a decreased ability of membrane to mix and fuse as shown by monitoring calcein release, Generalized Polarization of Laurdan and fluorescence dequenching of octadecyl rhodamine B, respectively. Altogether, results shed light on how cardiolipin may be critical for improving antibacterial action of new amphiphilic aminoglycoside derivatives.


Introduction
Bacterial resistance to all classes of antibiotics, including aminoglycosides, is becoming a global public health crisis [1] requiring urgent actions. Besides measures to encourage the appropriate use of antibiotics, discovery and development of new active molecules are urgently required. Driven by the unique architecture of bacterial envelopes and biophysical properties of bacterial cell membranes, the search for new membrane-active antibacterials is promising [2].
Membrane-active antibiotics target essential and preserved structures among various bacterial species. More and more they are described as chemosensitizer to increase the activity of other antibiotics [3]. They have also been developed for their potential alternatives to the resistance-prone conventional antibiotics and for their activity against slow-growing or dormant bacteria as well as on biofilms. PLOS

Membrane fusion experiment
Membrane fusion was monitored on LUVs. Octadecyl rhodamine B (R18), a lipid-soluble probe, was incorporated into a lipid membrane at a self-quenching concentration. Hence a decrease in its surface density [27;32] is associated with an increase in the fluorescence intensity of the preparation. The experiments were performed as described previously in details [27;33] by using an LS55 spectrofluorimeter (PerkinElmer) and excitation wavelength (λ exc ) a1nd emission wavelength (λ em ) fixed at 560 and 590 nm, respectively. The fusion percentage was calculated according to the Eq (1) where F t , F 0 and F 1 are the fluorescence intensities measured at time t, at time 0, and after the addition of detergent (0.5% triton X-100) to disrupt the membrane of the liposomes, respectively. Elimination of non-entrapped calcein and determination of calcein release from liposomes were performed as described in detail previously [27] by using an LS55 spectrofluorimeter (PerkinElmer) with λ exc and λ em fixed at 472 and 512 nm, respectively.

Calcein-release measurement for permeability study
The amount of calcein released after time t was calculated according to Eq (2): where RF-fraction of calcein released, F t , F 0 and F 1 are the fluorescence intensities measured at time t, at time o, and after the addition of Triton X-100, respectively.

Generalized Polarization of Laurdan
Amphiphilic Laurdan fluorescent probe resides in the interfacial region of the lipid bilayer membrane. Scanning electron microscopy P. aeruginosa was grown to mid log (OD: 0.3 at 600 nm). Bacteria were washed in phosphate buffered saline (0.1 M PBS, pH 7.4) and then treated with cardiolipin and/or 3',6-dinonylneamine for 1 hour at different concentrations. A suspension of bacterial cells was immobilized on poly-L-lysine coated coverslips for 10 min at room temperature. After washing in buffer-to remove the excess of free floating bacteria-coverslips were incubated in 1% glutaraldehyde in order to cross-link the fixed bacteria on poly-lysine coating. We further post-fixed the samples in 1% osmium tetroxide in cacodylate buffer for 2 h at 4˚C and washed in water to eliminate traces of remaining osmium tetroxide. Samples were then dehydrated in graded series of ethanol, critical point dried and coated with 10 nm of gold. Samples were observed in a CM12 Philips electron microscope at 80 kV with the secondary electron detector. The i-TEM imaging software was used for the analysis of bacterial length. The asymmetrical division was calculated by measuring the differences in daughter lengths from the position of cell division.

Results
We first investigated the effect of increasing contents in cardiolipin on changes in membrane fusion, membrane permeabilization and membrane hydration induced by 3',6-dinonylneamine by using membrane models mimicking the Pseudomonas aeruginosa membrane. Moving onto bacteria, we studied the potential effect of cardiolipin on morphology, growth rate and division of P. aeruginosa as well as the effect on the antimicrobial activity of a new promising antibiotic, the 3', 6-dinonylneamine.

Effect of cardiolipin on membrane fusion induced by 3',6-dinonylneamine
To investigate whether the ability of 3',6-dinonylneamine to induce lipid membrane fusion, a critical event in bacterial division, is dependent upon the presence of cardiolipin, we carried out experiments based on octadecyl rhodamine B (R18) fluorescence dequenching [32]. We used large unilamellar vesicles (LUVs) composed of POPE (60%) and POPG (21%) and varying amounts of cardiolipin (CL) (molar ratio 0, 11, and 20%). LUVs labeled with octadecyl rhodamine B (R18) were mixed with unlabeled LUVs and then 3',6-dinonylneamine was added at increasing concentrations (0-1.5 μM). In the absence of cardiolipin (POPE: POPG: CL [60:21:0] LUVs) (Fig 1A), the addition of 3',6-dinonylneamine causes a fast increase in fluorescence even at very low concentration (0.1 μM) of aminoglycoside derivative. The fusion process reached a plateau value after 8 minutes for all the concentration range of 3',6-dinonylneamine. The effect is inversely related to contents in cardiolipin since with cardiolipin present at 11% ( Fig 1B) the percentage of fusion was lower as compared to that observed with liposomes without cardiolipin (31.5% versus 60% at the highest 3',6-dinonylneamine concentration, 1.5 μM). At cardiolipin content of 20% (Fig 1C), no fusion was observed. Fig 1D shows at 10 min, the percentage of membrane fusion at the three selected molar ratio of cardiolipin, with increasing concentrations of 3',6-dinonylneamine. Membrane fusion decreases with increased ratios of cardiolipin content in despite of the concentration of the amphiphilic neamine derivative. At 20% of cardiolipin, no fusion was observed even at the highest concentration of 3',6-dinonylneamine. At 0 and 11% molar ratio of cardiolipin, the percentage of fusion reached a plateau value (45% and 30%, respectively) at 0.5 μM of 3',6-dinonylneamine.

Effect of cardiolipin on membrane permeability induced by 3',6-dinonylneamine
To know if cardiolipin is involved and critical for membrane permeabilization induced by 3',6-dinonylneamine, we monitored calcein release from liposomes in which calcein was entrapped at self-quenching concentrations. Fig 2A-2C shows time-dependent calcein release induced by 3',6-dinonylneamine, with variations of cardiolipin contents from 0-20% in POPE: POPG (60:21) LUVs. In a cardiolipin free membrane model system (Fig 2A), increased concentrations of 3',6-dinonylneamine induced a slow increase in fluorescence intensity. After 10 min, at the highest concentration of the amphiphilic neamine derivative used (2 μM), only 10% to 15% of the encapsulated calcein was released from liposomes. When cardiolipin was inserted within liposomes, a prompt enhancement of calcein release was observed (around 62.5% and 76.0% release of the encapsulated calcein) at 11 ( Fig 2B) and 20 ( Fig 2C) % cardiolipin respectively. The effect of cardiolipin on membrane permeability induced by increasing concentrations of 3',6-dinonylneamine is further illustrated in Fig 2D. A marked increase in membrane permeability correlated with increased molar ratio of cardiolipin in LUVs as observed for all concentrations of 3',6-dinonylneamine investigated.

Effect of cardiolipin on membrane hydration induced by 3',6-dinonylneamine
To further elucidate the mechanisms involved in the cardiolipin dependency changes of membrane fusion and membrane permeabilisation induced by 3',6-dinonylneamine, we used Laurdan for monitoring membrane order and hydration. We followed changes in fluorescence intensity and calculated Generalized Polarization (GP) to establish the ability of 3',6-dinonylneamine to modulate hydration of the lipid bilayers containing different percentages of cardiolipin. Together, these results obtained on lipid membrane models highlighted the critical role of cardiolipin on the effects induced by 3',6-dinonylneamine by decreasing membrane fusion and hydration, and increasing membrane permeability of lipid membranes. The further question was to demonstrate if cardiolipin also played a key role in bacterial growth and if it could be involved in the antibacterial activity of 3',6-dinonylneamine on P. aeruginosa.

Effect of cardiolipin on P. aeruginosa growth and division
Based on results obtained from membrane models, we wanted to investigate dependence of cardiolipin on the antimicrobial activity of 3',6-dinonylneamine in P. aeruginosa (ATCC 27853). We first monitored the growth rate of P. aeruginosa in the presence of increasing concentrations of cardiolipin in the growth medium (Fig 4). At 1 μg/ml of cardiolipin, no effect on P. aeruginosa growth was observed. Above this concentration, cardiolipin significantly affected the growth curve of P. aeruginosa by delaying the time required to observe 50% of the maximal effect and by decreasing the maximal value of log (OD t /OD 0 ).
Interestingly, by using scanning electron microscopy (Fig 5), we showed that the presence of cardiolipin in the growth medium (15 μg/ml) induced asymmetric growth of bacteria. This effect was not dependent upon the presence of the amphiphilic neamine derivative.

Effect of cardiolipin on growth of P. aeruginosa in the presence or absence of 3',6-dinonylneamine
The minimum inhibitory concentration of 3',6-dinonylneamine against P. aeruginosa is 1-4 μg/ml (23;25-27). When cardiolipin was added to the medium (1 and 15 μg/ml) (Fig 6B  and 6C), the growth curves of P. aeruginosa in presence of increasing amounts of 3',6-dinonylneamine were monitored in comparison with data obtained in absence of added cardiolipin (Fig 6A). The bacterial growth was delayed and the growth rate decreased with increase in concentrations of 3',6-dinonylneamine with or without cardiolipin. When cardiolipin was replaced by POPG (15 μg/ml), the bacterial growth in presence of 3',6-dinonylneamine was only slightly decreased and delayed (S2 Fig). To highlight the effect of cardiolipin on 3',6-dinonylneamine activity, the P. aeruginosa growth rate upon addition of cardiolipin (15 μg/ml) and 3',6-dinonylneamine (0-2 and 5 times MIC) is illustrated in Fig 6D.

Imaging of the effect of cardiolipin and/or 3',6-dinonylneamine on P. aeruginosa growth
In order to visualize potential effect of cardiolipin, 3',6-dinonylneamine and both compounds on growth of P. aeruginosa, we used scanning electron microscopy (Fig 7A-7D). We normalized the number length found (which is represents Counts) in each range (range divided in 400 nm interval from 600 nm to 4000 nm) with maximum number of length in a range (Fig 7A'-7D'). Total number of length (or counts) is approximately 200. As compared to control (Fig 7A'), the three main features are (i) a decrease in normalized counts in the range 1800-2200 nm, (ii) a shift in maximum counts from 1800-2200 nm to 1400-1800 range from control to conditions where P. aeruginosa was incubated with cardiolipin ( Fig 7B') and 3',6-dinonyl neamine (Fig 7C') or both compounds (Fig 7D'), and (iii) a decrease of the size heterogeneity. These effects were observed in presence of cardiolipin (Fig 7B'), 3',6-dinonyl neamine (Fig 7C'), and the mixture of both (Fig 7D') with an increased in their extent (S3 Fig).

Discussion
Binding of antibiotics at the septum of dividing Gram-positive [37;38] and Gram-negative [19] bacteria where cardiolipin is enriched [39] could be a unique target in order to design new antibacterial molecules. Interactions with bacterial cell membrane at discrete domains [26; 40], could result in the dispersion of these domains and often in the disruption of functions governed by those domains [4;40-42]. This is exactly what we demonstrated on the life-threatening Gram-negative bacteria, P. aeruginosa with the amphiphilic neamine derivative, 3',6-dinonylneamine [26] even the mechanisms are still unclear, requiring further study.
Increasing contents in cardiolipin in growth medium delayed P. aeruginosa growth and favored the asymmetric growth. Heterogeneity in bacterial length arises because of the unusual, unipolar nature in growth. In presence of cardiolipin, the division of the asymmetrically growing single mother cell gives rise to a number of daughter cells that differ in size [43]. This probably constitutes a physiological response of bacteria to stress since asymmetric division generates distinct cell types that may help bacteria to exploit patchy and variable, environments more effectively [44]. Interestingly, in the same time, the antimicrobial activity of 3',6-dinonylneamine, an amphiphilic aminoglycoside which binds to and requires cardiolipin for its proper activity, increased with increasing contents in cardiolipin. This effect is in agreement with that reported by Molohon and coll [37] who have shown that exogeneous cardiolipin increases sensitivity to plantazolicin, a linear azole-containing peptide active against the Gram-positive Bacillus anthracis. Cardiolipin could increase bacterial respiration resulting in potentialisation of the killing effect of bactericidal antibiotics [45]. However, the mechanism is probably more complex and the metabolic state of bacteria could impact antibiotic efficacy. In this line, daptomycin exhibits an antagonistic relationship with cardiolipin in Gram-positive Enterococci [46]. The role of cardiolipin on the activity of membranes targeting antibiotics is probably more complex than it seems at first sight. Several non-exclusive processes could be involved including (i) potential effect on cardiolipine synthases, (ii) changes in location of cardiolipin from the inner to outer membranes and (iii) equilibrium between the antibiotic and cardiolipin. In addition, the cross-talk between PG and cardiolipin metabolism [47] could be critical.
Based on our previous studies on P. aeruginosa [26], we suggest that amphiphilic aminoglycoside derivatives recruit cardiolipin in regions of high negative curvature leading to changes in its location which can in turn result in inhibition of membrane scission through changes of the topology of proteins involved in in cell division (FtsZ, FtsA) and cell shape regulation (MreB). Indeed, 3',6-dinonylneamine unsettled rod shape regulation protein mCherry-MreB and inhibited L-spherosplasts to restore their initial rod shape [26]. In agreement with results on lipid models, interaction between 3',6-dinonylneamine and cardiolipin could result in decreasing line tension and/or the discontinuity in elastic properties between the cardiolipinrich and cardiolipin-poor phases which are known to spontaneously induce fission. Alternatively, binding of amphiphilic neamine derivatives to cardiolipin could trigger a conformational change in FtsZB, ultimately decreasing protein activity. Thus, FtsZ and MreB functions would be fine-tuned through modulation of cardiolipin levels.
Additional insight on the effect of cardiolipin on the antibacterial activity of amphiphilic aminoglycoside derivatives results from studies performed on membrane models mimicking bacterial membranes of P. aeruginosa. We demonstrated that cardiolipin increased membrane Effect of cardiolipin on the antimicrobial activity of an amphiphilic aminoglycoside derivative permeability and decreased hydration (as assessed by increase of calcein release and increase of Generalized Polarization (GP) of Laurdan). In the absence of cardiolipin, the Generalized Polarization (GP) value slightly decreased with an increase in 3',6-dinonylneamine concentrations suggesting an increase in inter backbone distance between POPE and POPG and penetration of water molecules into the bilayer. At the opposite, in presence of cardiolipin, 3',6-dinonylneamine induced a huge increase in the Generalized Polarization GP, reflecting a decrease of hydration likely due to interaction of 3',6-dinonylneamine with cardiolipin. This could explain enhanced pore formation. Decrease of membrane hydration upon increasing cardiolipin contents could also explain the decreased hemi-fusion process induced by 3',6-dinonylneamine as evidenced by measuring fluorescence dequenching of octadecyl rhodamine B (R18) upon increasing contents in cardiolipin. One potential molecular mechanism could result from the complementary inverted cone shaped of 3',6-dinonylneamine and coneshaped of cardiolipin preventing hexagonal phase formation and decreasing fusion process.
Expanding on these studies will provide insights into how increase in cardiolipin contents, related to emergence of resistance, could influence antibiotic efficiency and modulate proteinprotein interactions [48].

Conclusion
We evidenced cardiolipin-dependency on antimicrobial activity of 3',6-dinonylneamine. Cardiolipin is responsible for decrease in growth rate, asymmetric growth defects and enhanced antimicrobial activity of 3',6-dinonylneamine. Upon increase in cardiolipin contents in membrane models mimicking membranes of P. aeruginosa, 3',6-dinonylneamine induced increase in membrane permeabilization and decreased in membrane hydration probably related with inhibition of membrane fusion. We anticipate that these findings will be influential for understanding the role of lipids in modulating functions of proteins or drug activity.