Photodynamic therapy (PDT) combines a nontoxic photoactivatable dye, or photosensitizer (PS), with harmless visible light of the correct wavelength to excite the dye to its reactive triplet state, which will then generate reactive oxygen species, such as singlet oxygen and superoxide, that are toxic to cells (
4). Although discovered more than 100 years ago by its killing effect on microorganisms (
25), PDT has found most success as a treatment for cancer (
9) and age-related macular degeneration (
3).
A relatively novel application of PDT is to employ its ability to kill pathogenic microbes in the treatment of localized infections (
12). Most of the PSs that are under investigation for the treatment of cancer and other tissue diseases are based on the tetrapyrrole nucleus, such as porphyrins, chlorins, bacteriochlorins, and phthalocyanines (
4). However, other dyes that are frequently proposed as antimicrobial PSs have different molecular frameworks. These include halogenated xanthenes, such as Rose Bengal (RB) (
33); perylenequinones, such as hypericin (
17); and phenothiazinium salts, such as toluidine blue O (TBO) (
1), methylene blue (MB) (
11), and azure dyes (
37). It is known that gram-positive bacterial species are much more sensitive to photodynamic inactivation (PDI) than gram-negative species (
22) and that the ideal PS for killing bacteria should possess an overall cationic charge and preferably multiple cationic charges (
12). Phenothiazinium salts possess one intrinsic quaternary nitrogen atom and have been used as PSs to kill tumor cells in vitro (
39) as well as to treat tumors in animal models (
10). There is an increasing body of evidence for their phototoxic efficiency against a broad range of microorganisms (
29,
38), such as
Escherichia coli,
Staphylococcus aureus (
29), streptococci (
28),
Listeria monocytogenes (
32), and
Vibrio vulnificus (
41). At present, the only PSs used clinically for antimicrobial treatments are phenothiazinium salts. For instance, MB or TBO and red light are used to disinfect blood products and sterilize dental cavities and are proposed to treat periodontitis (
36).
Efflux mechanisms have become broadly recognized as major components of microbial resistance to many classes of antibiotics (
30). Some efflux pumps selectively extrude specific antibiotics, while others, referred to as multidrug resistance pumps (MDRs) expel a variety of structurally diverse compounds with differing modes of action. It has been suggested that amphipathic cations represent the existing natural substrates of MDRs (
19), and these molecules have frequently been used to study bacterial MDR-mediated efflux. A classical example of an amphipathic cation is ethidium bromide, and its uptake by microbial cells can easily be quantified by measuring intracellular fluorescence generated when the dye binds to nucleic acids (
2). The existence of MDRs makes the discovery of antimicrobial compounds that are recognized by them problematic in standard screens that employ cells carrying MDRs. The development of MDR mutants gave an answer to this apparent paradox (
13). Disabling of MDRs in gram-negative species led to a striking increase in antimicrobial activity for numerous plant substances (
35).
In the present study, we asked whether phenothiazinium salts, which are structurally characterized as amphipathic cations, could potentially be substrates of microbial MDRs. We used MDR-deficient and MDR-overexpressing mutants of the human pathogens S. aureus, E. coli, and Pseudomonas aeruginosa, together with a range of PSs, including both phenothiazinium salts and non-phenothiazinium-based PSs.
DISCUSSION
This study has demonstrated for the first time that phenothiazinium-based PSs are substrates of MDRs in bacteria. The fact that this effect was observed with three separate but related molecular structures (MB, TBO, and DMMB), which have all been frequently used in the literature as antimicrobial PSs, suggests that it is a general phenomenon applicable to all photoactive phenothiazinium dyes. By contrast, the complete absence of any differences in susceptibility by use of the non-phenothiazinium-based antimicrobial PSs (RB and pL-c
e6) provides evidence that the MDR phenotype of the bacterial cells does not affect other physiological parameters that could influence susceptibility to PDI. These parameters could have included such variables as membrane structure, DNA and other cellular repair systems, and levels of antioxidant enzymes. RB and light probably kill bacteria by generating extracellular singlet oxygen that destroys the membrane from the outside in (
7,
8,
33), while polycationic polymeric pL-c
e6 is probably taken up into bacterial cells by a self-promoted uptake pathway, as described for other polycationic peptides (
24). The finding that uptake levels of both MB and TBO by the NorA knockout mutant, wild-type, and overexpressing strains of
S. aureus were proportional to levels of NorA expression suggests that the role of the MDRs is to pump out the PS from the cells and thereby lessen the phototoxicity observed upon illumination. The similarity of the levels of uptake of the nonphenothiazinium PSs between the various NorA phenotypes suggests that these compounds are not recognized by MDRs. Alternatively, a less likely explanation is that the nonphenothiazinium PSs were not taken up into the cells but merely bound to the outside layers of the cell coat. This is unlikely because pL-c
e6 was the most potent PS active at a 1-μM concentration, and this high activity is best explained by the polycationic structure mediating intracellular uptake. The correlation of phenothiazinium dye (MB and TBO) uptake with TolC levels in
E. coli and with MexAB levels in
P. aeruginosa further confirms the role of MDRs to pump out these PSs.
We found that the MDR recognition of phenothiazinium PSs applied equally to two different molecular efflux systems present in three different bacterial species. The NorA MDR of
S. aureus is a member of the MF family and protects the cells from norfloxacin and a number of amphipathic cations, such as the common disinfectants benzalkonium chloride and cetrimide (
13). RND pumps are frequently found in gram-negative bacteria. The TolC protein of
E. coli, through its interaction with AcrA and AcrB, is thought to form a tripartite continuous protein channel that expels substrates from the cell. AcrAB-TolC assembles into an alpha-helical transperiplasmic tunnel, which is embedded in the outer membrane by a contiguous beta-barrel channel (
16). AcrAB-TolC and its homologues thus provide large exit ducts for a wide range of substrates, including organic solvents, fluorescent lipids, bile acids, erythromycin, and cloxacillin (
40,
42), recognized by the AcrAB proteins.
P. aeruginosa carries genes for at least 11 distinct (but related) RND family pumps (
30). Mex-Opr substrates include biocides, dyes, detergents, metabolic inhibitors, organic solvents, and molecules involved in bacterial cell-cell communication. The total number of MexAB-OprM units per wild-type cell was calculated to be about 400 assemblies, and the turnover rate of a single pump unit was predicted to be about 500 molecules per second (
27).
There have been some reports of compounds with structures somewhat similar to those of phenothiazinium dyes being substrates or inhibitors of bacterial MDRs. The cationic xanthene dye pyronin Y was reported to be a substrate of RND pumps in
P. aeruginosa (
23) and also of NorA in
S. aureus (
15). Similarly, the cationic acridinium antimicrobial acriflavine was reported to be a substrate of NorA (
15). Noncationic phenothiazine compounds such as chlorpromazine derivatives have been reported to be inhibitors of MDRs in
E. coli (
26), and prochlorpromazine inhibited NorA in
S. aureus (
15).
It has been said that the present times represent the “end of the antibiotic era” (
31), due to increasing development of bacterial resistance to multiple classes of antibiotics. PDT uses otherwise harmless dyes and light, and it has been proposed that microbes would be unlikely to develop resistance to the destructive effects of photochemically generated reactive oxygen species that can cause irreversible oxidative damage to essential cellular constituents such as proteins, lipids, and nucleic acids. There is one report of the failure to produce resistance by repeated cycles of PDI using a nonphenothiazinium polycationic-PS conjugate (
18). Our data however raise the possibility of bacteria developing resistance to phenothiazinium-based PDI due to selective survival of strains with increased MDR expression levels.
We generally found bigger differences in susceptibility between MDR knockout and wild-type strains than between wild-type and overexpressing strains. This suggests that the wild-type species we tested had high levels of functioning MDRs and provides a possible reason why the concentrations of phenothiazinium PS necessary to efficiently kill both gram-positive and gram-negative bacteria upon illumination is significantly higher (10 to 300 μM [
37,
41]) than the concentrations of alternative cationic PS that have been used; for instance, Maisch et al. (
21) reported a cationic porphyrin derivative in combination with blue light that mediated killing
S. aureus at a concentration of only 5 nM.
Phenothiazinium dyes have been long established as nontoxic and clinically useful compounds both for staining living tissues (
6) (for instance, in the detection of dyplasias [
20]) and for some pharmacological indications (
5,
34). This consideration together with their ready availability has probably been important in their selection as antimicrobial PSs for the few clinical indications in which antimicrobial PDT is carried out. The present discovery that these compounds are substrates for bacterial MDRs raises the possibility of combining the phenothiazinium dye with an MDR inhibitor. For instance, the dye solution applied to sterilize a dental cavity when illuminated could contain an MDR inhibitor applicable to the MDRs expressed by the oral pathogens causing dental caries.
Our future work will ask whether phenothiazinium dyes are also substrates of MDRs expressed by fungi (e.g., Candida albicans) and to what degree selected MDR inhibitors can potentiate antimicrobial PDI of pathogenic microorganisms.
Acknowledgments
This work was supported by the National Institute of Allergy and Infectious Diseases (grant AI050875 to M.R.H.).
We are grateful to Kim Lewis (Northeastern University, Boston, MA), Olga Lomovskaya (Mpex Pharmaceuticals, Inc., San Diego, CA), and David C. Hooper (Massachusetts General Hospital, Boston, MA) for gifts of bacterial strains. We thank Tatiana N. Demidova for helpful suggestions and a critical reading of the manuscript.