Antimicrobial activity of photosensitizers: arrangement in bacterial membrane matters

Porphyrins are well-known photosensitizers (PSs) for antibacterial photodynamic therapy (aPDT), which is still an underestimated antibiotic-free method to kill bacteria, viruses, and fungi. In the present work, we developed a comprehensive tool for predicting the structure and assessment of the photodynamic efficacy of PS molecules for their application in aPDT. We checked it on a series of water-soluble phosphorus(V) porphyrin molecules with OH or ethoxy axial ligands and phenyl/pyridyl peripheral substituents. First, we used biophysical approaches to show the effect of PSs on membrane structure and their photodynamic activity in the lipid environment. Second, we developed a force field for studying phosphorus(V) porphyrins and performed all-atom molecular dynamics simulations of their interactions with bacterial lipid membranes. Finally, we obtained the structure-activity relationship for the antimicrobial activity of PSs and tested our predictions on two models of Gram-negative bacteria, Escherichia coli and Acinetobacter baumannii. Our approach allowed us to propose a new PS molecule, whose MIC50 values after an extremely low light dose of 5 J/cm2 (5.0 ± 0.4 μg/mL for E. coli and 4.9 ± 0.8 μg/mL for A. baumannii) exceeded those for common antibiotics, making it a prospective antimicrobial agent.


TABLE OF CONTENTS
. 1 Table S1 provide information about LJ parameter assignment. Figure S7 shows atom numbers. For hydrogen atoms connected to C1, C2, C5, C6, C10, C11, C15, C16 we used atom type HGR51, and for hydrogen atoms connected to C9, C14, C19, C20 we used aromatic hydrogen atom type HGR61. For hydrogen atoms connected to the oxygen, we used HGP1 atom type. For ethoxy group we used CG321 and CG331 for carbon atoms and HGA2 and HGA3 for hydrogen atoms.  Partial charges Partial charges for N and P atoms determined using NPA are provided in Table S2, and partial charges of heave atoms determined with RESP method with restrained charges for N and P are presented in Table S3. Charges for aromatic and aliphatic hydrogen atoms were assigned their standard value of 0.15 and 0.09 from CHARMM force field. Bond and angle parameters We tested 3 sets of parameters based on the atom type assignment for nitrogen atoms: one atom type for all nitrogen atoms, three atom types based on obtained partial charges and four different atom types. Only parameter set with four different atom types for nitrogen atoms was able to correctly reproduce the optimized QM geometry (deviation less than 0.02 Å and 3° for bonds and angles, respectively).

PARAMETRIZATION OF THE FORCE FIELD Lennard-Jones (LJ) parameters assignment
Bonds and angles were fitted simultaneously to minimize the objective function described in ref. 1 with simplex optimization algorithm. We performed several runs of optimization until the objective function stopped decreasing. n(OH)2 model was used to obtain most of the bonded parameters for porphyrin ring. For n(OEt)2 model only bonded parameters connected to the atoms N1-N4 were reoptimized.
Dihedral parameters QM potential energy surface (PES) scans were calculated for dihedrals of porphyrin ring, dihedrals arising from the connection of axial and equatorial groups to the porphyrin ring (see Fig. S7). Dihedrals inside porphyrin ring were scanned by 50° in each direction with 10° increment, rotation of the ethoxy group around P-O bond were scanned by 180° in each direction with 15° increment, all other dihedrals were scanned by 90° in each direction with 15° increment. To reduce overfitting, we set to zero most of the dihedrals of porphyrin ring and dihedrals containing hydrogen atoms.
We  Table 4 provides the information about root mean square error (RMSE). Although RMSE for Ring and Phenyl dihedrals are bigger than recommended value for CGenFF force field (~0.5 kcal/mol), the obtained RMSE are comparable with what was derive previously for similar structure 3 and lowenergy configuration are reproduced within recommended RMSE.

Comparison of QM and MM optimized geometry
To elucidate the ability of our force field to reproduce the shape of porphyrin ring we compared QM and MM optimized geometry. The comparison between QM and MM optimized geometry is illustrated in Figure S11. MM-optimized geometry agrees well with geometry obtained from QM calculation. RMSD for both structures below 0.02 Å. Figure S11. Comparison of QM (red) and MM (blue) optimized geometry. Figure S12. Dependence of the optical density at 600 nm (OD600) for the bacterial suspension of E. coli incubated for 24 h with Ampicillin (A) and of A. Baumannii incubated for 24 h with Collicin (B). The study was carried out in 3 independent repetitions, the error was determined by ANOVA.