The uptake of NMB and ACS268 was studied by treating the cells with NaOH 0.1 M/1% sodium dodecyl sulphate (SDS) in order to extract the cell-bound PSs and analysing the cell lysates by fluorescence spectroscopy. Bulk concentrations of 10 µM and 7.5 µM were used, respectively. While the uptake of ACS268 reaches a plateau value after only 1 h, the uptake of NMB requires ca. 20 h for completion (see Supplementary Information). For ZnTMPyPz we adopted the uptake conditions described by Dupouy et al. [15], namely 30 min contact time at 37 °C and 10 µM bulk concentration.

It was critical in our experiments to eliminate any PS molecule from the external aqueous buffer in order to ensure that all the spectroscopic data was related to cell-bound molecules. Therefore, to unmistakably assign the spectroscopic data, the external aqueous solutions were routinely tested before and after the in vitro measurements, providing no significant signal in all cases.

The absorption spectrum of NMB bound to E. coli showed a red-shift of 8 nm relative to that in the buffer (Figure 2A), as well as a slight change in the relative amplitudes of the band and the shoulder. However, no significant change was observed in the fluorescence emission spectrum (Figure 2B). For ZnTMPyPz, the absorption spectra both in E. coli and buffer showed a similar structureless broad-band distinct from the spectrum obtained in DMF, where it is in a monomeric form (Figure 2C). A 5 nm red shift was observed in the fluorescence spectrum relative to that in DMF (Figure 2D). For ACS268, the absorption spectrum in E. coli showed a 7 nm red shift relative to that in buffer (Figure 2E). The fluorescence spectra in E. coli and buffer showed similar structures but with a 5 nm red shift (Figure 2F).

Time-resolved fluorescence measurements provided additional insight on the uptake of the PSs (see Electronic Supplementary Information). The fluorescence of NMB at 657 nm decayed with first-order kinetics with a lifetime of 0.5 ± 0.1 ns both in E. coli and in buffer. The fluorescence decay of ZnTMPyPZz in E. coli at 675 was also fitted with a single exponential function with a lifetime of 2.8 ± 0.5 ns. An additional exponential term was necessary to fit the data in buffer, namely with a lifetime of 0.5 ± 0.1 ns. Time-resolved fluorescence of ACS268 was inconclusive as the kinetics in solution was too complex to be fitted by any standard model.

Pulsed-laser irradiation of NMB-loaded E. coli cells (30 min contact time; 10 μM NMB bulk concentration) suspended in buffer at pH 7.4 produced clear time-resolved ¹O₂ phosphorescence signals at 1,270 nm (Figure 3). The contact time between the NMB solution and the bacteria had no effect on the ¹O₂ luminescence kinetics. In PBS, the signals grew with a time constant of τ₁^PBS = 2.5 ± 1 μs, and decayed monoexponentially with a lifetime τ₂^PBS = 3.7 ± 1 μs. In deuterated-PBS (D-PBS), the signal rise remained unchanged with a lifetime τ₁^D−PBS = 2.5 ± 1 μs, but it decayed more slowly, with a time constant of τ₂^D−PBS = 65 ± 2 μs. Addition of 0.75 mM bovine serum albumin (BSA), a ¹O₂ quencher that is not able to cross the cell wall due to its size [19], clearly modified the signal: the rise of the phosphorescence was now complete within the time resolution of our set-up, and the decay was also faster with lifetime τ₁^BSA = 7.0 ± 1 μs. Saturation of this suspension with oxygen led to a further decrease of the decay lifetime to τ₁^BSA,O2 = 3.5 ± 1 μs. Finally, measurements in E. coli spheroplasts, i.e., bacteria from which the cell wall had been almost completely removed, led to a 4-fold decrease in the ¹O₂ phosphorescence signal intensity, but with the same kinetics as in intact cells (see Supplementary Information).

Although ZnTMPyPz is a potent ¹O₂ photosensitizer in solution [20], irradiation of ZnTMPyPz-loaded E. coli cells suspended in D-PBS did not produce any clear time-resolved ¹O₂ phosphorescence signal at 1,270 nm (see Supplementary Information). Neither a higher contact time of ZnTMPyPz (3 h), nor a higher acquisition time of the ¹O₂ phosphorescence signal (12 million laser pulses) produced any detectable signal.

The quantum yield of singlet oxygen production (Φ_Δ) of ACS268 was measured in two different solvents, DMF and D₂O, by comparison of the signal produced by optically-matched solutions of reference PSs (phenalenone and TMPyP) [21] with that produced by ACS268. The detailed procedure is described in the Supplementary Information. Thus, Φ_Δ values of 0.87 ± 0.01 and 0.04 ± 0.01 were determined in DMF and D₂O, respectively. Irradiation of ACS268-loaded E. coli cells produced time-resolved ¹O₂ phosphorescence signals clearly different from those observed for NMB (Figure 4). In PBS, the signals grew with a time constant of τ₁^PBS = 2.1 ± 1 μs, and decayed biexponentially with lifetimes τ₂^PBS = 8.6 ± 1 μs and τ₃^PBS = 34.1 ± 1 μs. In D-PBS, the rise of the signal was slightly slower, τ₁^D−PBS = 5.2 ± 1 μs, but the decay kinetics remained unchanged biexponential. Addition of 0.75 mM of BSA shortened only the slowest decay lifetime, i.e., τ₂^BSA = 8.6 ± 1 μs and τ₃^BSA = 34.0 ± 1 μs. Finally, measurements in E. coli spheroplasts led to a 3-fold decrease in the ¹O₂ phosphorescence signal intensity, but with the same kinetics as in intact cells (see Supplementary Information).

The main goal of this study was to gain insight into the kinetics of ¹O₂ production and decay by three different photosensitizers in E. coli cells, thereby shedding light on the mechanism of photodynamic inactivation of Gram-negative bacteria. Specifically, we were interested to learn whether representative members of three different families of photosensitizers would provide different photosensitization kinetics.

Towards this goal, it was first essential to obtain unambiguous time-resolved ¹O₂ phosphorescence signals from viable cells, i.e., (1) to reduce to a minimum the acquisition time as the ¹O₂ kinetics in cells changes with increasing light exposure [22,23], and (2) to identify the optimal concentration and contact time of each PS for efficient cell binding/uptake. A concentration ca. 10 μM, similar to those used in inactivation experiments for E. coli [15,24], and a 30 min contact time, typical in APDT [25,26], proved adequate in this respect notwithstanding the higher amount of NMB taken up at longer contact times (Figure 2). Under similar experimental conditions, all these PSs show no significant dark toxicity [15,17].

In homogeneous environments, the ¹O₂ signals can be appropriately described by Equation (1), where S₀ is a quantity proportional to the amount of ¹O₂ formed upon pulse excitation, τ_T is the lifetime of the PS’s triplet-excited state, and τ_Δ is the ¹O₂ lifetime. Thus, the signal is expected to show two monoexponential components, corresponding to the rise and decay of the signal: (1)

In microheterogeneous environments it is often found that different populations the PS’s triplet-excited state and ¹O₂ exist and therefore Equation (1) must be expanded to account for such different populations Equation (2). In such cases the rise and the decay of the signals follow multiexponential laws, the number of exponentials representing the number of independent populations. (2)

The kinetic models may be even more complex due to phenomena such as diffusion or cross-talk between the different populations.

Regarding the phenothiazine group, NMB was chosen as the model due to its highest singlet oxygen quantum yield and higher hydrophobicity [27], as well as its ability to properly inactivate many bacterial species [27]. Two major changes can be observed in the absorption spectrum of NMB upon binding to E. coli: on one hand, the relative contribution of the shoulder at 600 nm increases, which is consistent with the report by Usacheva et al. [28] that phenothiazine dyes tend to dimerize in the presence of bacteria. On the other, the band maximum shows a slight red shift that reveals a change in the NMB’s microenvironment [29]. Somewhat surprisingly, such microenvironmental changes have no effect on the fluorescence spectrum or in the fluorescence lifetime. The fact that there is no change in the fluorescence lifetime indicates that, despite the different microenvironment, NMB is still surrounded by water molecules (see Supplementary Information). This conclusion is supported by the ¹O₂ phosphorescence formation and decay kinetics. Both in PBS and D-PBS, the lifetimes obtained for the ¹O₂ phosphorescence signal matched those obtained in neat H₂O or D₂O, a 2.5 μs rise component assigned to the triplet state of NMB, and a 3.7 or 65 μs component due to ¹O₂, respectively. This indicates that NMB is localized in an aqueous-like environment and that ¹O₂ is deactivated mainly by interactions with H₂O or D₂O [30]. Of course, it cannot be excluded that another fraction of ¹O₂ molecules reacts with or is quenched by cell components so rapidly that it does not contribute to the observed signal. Care was taken to ensure that the signals did not originate on NMB molecules that have leaked into the bulk solution during the experiments. The supernatant solution was checked before and after the experiments and in all cases we could find no signal. Addition of BSA to the D₂O suspensions significantly changed the signal kinetics, causing both a faster rise (<1 μs) and a faster decay (lifetime of 7 μs) in air-saturated suspensions. The latter component was decreased further to 3.5 μs upon bubbling with oxygen, which increased five-fold the concentration of oxygen in the system. These results suggest a dual effect of BSA: on one hand it quenches ¹O₂ as one would expect [19] decreasing its lifetime below the time resolution of our setup; on the other, it interacts with NMB shielding it from oxygen [31], which effectively increases the triplet NMB lifetime from 2.5 μs to 7 μs (in air) or 3.5 μs (under oxygen) as the bimolecular energy transfer process is slowed down. Therefore, NMB must be located somewhere in the external structure of the cell wall, accessible to BSA, or must be able to relocate and bind to BSA upon its addition to the solution. This conclusion is consistent with Usacheva’s et al. report that phenothiazinium dyes interact with the outer-wall bacterial lipopolysaccharides (LPS) [32,33]. Overall, our results are in line with the report that an initial population of 10⁴ CFU·mL⁻¹ of E. coli could be totally eliminated by a minimum lethal concentration of 7.8 μM NMB at a light fluence of 3.15 J cm⁻² [17]. Thus the suggestion that ¹O₂ plays an important role in the mechanism of action of NMB [17] is supported by our spectroscopic data.

Concerning the phthalocyanine family, ZnTMPyPz was chosen as a model since Dupouy et al. [15] had shown that, using the same uptake conditions as in this work, irradiation with visible light reduced the cell viability by 99.5%. These results can now be rationalized in the light of our findings: First, as observed in the absorption spectrum (Figure 2C), ZnTMPyPz is largely aggregated in E. coli as deduced from the broadening of the bands relative to DMF. This aggregation would explain why no signal is observed in the ¹O₂ luminescence experiments under conditions similar to those used for the other PSs, even increasing the contact time and the acquisition time. Spesia et al. showed that alterations in the cell membrane appear to be the major cause of E. coli inactivation upon APDT with visible light and a tetracationic zinc(II) phthalocyanine derivative (ZnPc⁴⁺) [34]. Assuming that (1) ZnTMPyPz and ZnPc⁴⁺ localize in a similar microenvironment because of the structural similarity, and that (2) ZnTMPyPz is basically aggregated like in an aqueous media, as observed by the absorption spectrum and the absence of ¹O₂ luminescence signal, a localization to the outer structure of the cell wall, comparable to the one obtained with NMB, is tentatively attributed. This conclusion is not contradictory to the 99.5% reduction in cell viability, as there is still a small quantity of monomer, detectable by steady-state (Figure 2D) and time-resolved (Figure S3; see the Supplementary Information) fluorescence measurements, that could be responsible for the production of ROS and the resulting photodynamic effect shown by ZnTMPyPz.

We previously demonstrated that a tetracationic porphyrin such as TMPyP, with the cationic charges peripherally-distributed, localizes in E. coli at two different sites: (1) externally bound to the cell wall, probably in a similar localization as NMB and ZnTMPyPz, and (2) inside the cell, bound to nucleic acids [14]. We hypothesize that ACS268, which has a unique cationic charge in the alkyl chain, would localize in a completely different environment than TMPyP, as the cationic charge is far away from the photoactive core and the lipophilicity/hydrophilicity of the molecule is completely different. The totally different Φ_Δ values in DMF and D₂O clearly demonstrate that ACS268 is able to produce ¹O₂ only when it is in monomeric form.

Both the spectroscopic measurements and the ¹O₂ phosphorescence in E. coli are in agreement with the aforementioned hypothesis. The red shift observed in the absorption and fluorescence spectra in E. coli indicates a different environment surrounding the PS relative to that in solution [29,35,36].

With ACS268, two components with lifetimes 8.6 μs and 34 μs are observed in the ¹O₂ phosphorescence experiments in PBS, D-PBS and upon addition of BSA. The presence of two lifetimes suggests two different localizations for ACS268 in E. coli., likely in the cell wall as deduced from the 3-fold decrease in the ¹O₂ phosphorescence signal intensity observed in spheroplasts. Both ACS268 populations are clearly shielded from oxygen as evidenced by the long triplet lifetimes as compared to the 2.5 μs found for NMB. This in turn indicates that only a minor fraction of triplets is quenched by oxygen and therefore that the amount of ¹O₂ produced by ACS269 in the cells must be very low. Moreover, the τ_Δ values for ¹O₂ in E. coli are significantly shorter (2.1 μs in PBS; 5.2 μs in D-PBS suspensions) than the ones observed in buffer (3.5 μs in PBS; 67 μs in D-PBS) [30] indicating a fast deactivation of ¹O₂ by quenchers surrounding the PS molecules that, however, does not lead to cell death. The inefficient production of singlet oxygen and its fast deactivation are consistent with the very low photodynamic activity found for this compound against E. coli (this work, data not shown).

The rate constant of ¹O₂ decay in microheterogeneous systems under exchange equilibrium conditions, k_d, is described by Equation (3) [37]: (3) where K_eq = [¹O₂]_cell/[¹O₂]_water is the partition equilibrium of ¹O₂ between the two phases, f_m and (1-f_m) are the volume fractions of the cell and aqueous phase, respectively, and k_d,cell and k_d,water are the decay rate constants inside and outside the cell, respectively. Assuming K_eq ≈ 1 and f_m ≈ 0.0012 for our suspensions in 5 × 10⁸ CFU/mL, the lifetime of ¹O₂ within the bacterial cell can be estimated as τ_Δ,cell ≈ 7 ns, a value that would perfectly correlate with the ¹O₂ phosphorescence kinetics observed upon addition of BSA, which traps any ¹O₂ molecule escaping from the cell wall into the external aqueous media.

It has previously been demonstrated that the presence of external proteins clearly modifies the effectiveness of APDT to inactivate microbial cells [38,39,40]. Such effect can also be observed in the kinetics of ¹O₂ in the presence of BSA. The addition of BSA clearly modifies the ¹O₂ phosphorescence signal produced by NMB, decreasing τ_Δ and increasing τ_T, demonstrating an interaction between NMB and BSA (Figure 3). However, the effect is barely observed when ACS268 is used as the PS (Figure 4), suggesting a deeper localization of ACS268 into the cell wall, thereby making it less accessible to BSA.

New Methylene Blue N (NMB) was supplied by Sigma-Aldrich Co. (St. Louis, MO, USA). Zinc-tetramethyltetrapyridino[3,4-b:39,49-g:30,40-l:3-,4-q]porphyrazinium salt (ZnTMPyPz) was synthesized as described by Marti et al. [20]. 5,10,15-tris(2,6-Dichlorophenyl)-20-[N-(12-trimethylammonium chloride)-dodecyl-p-sulphonamidophenyl]porphyrin (ACS268) was synthesized as described in the Supplementary Information. Deuterium oxide (D₂O, >99.9%) and dimethylformamide were purchased from Solvents Documentation Synthesis (Peypin, France). Bovine serum albumin (98%), lysozyme, sodium dodecyl sulphate (SDS), trizma, and Dubelcco’s phosphate buffered saline (PBS) were purchased from Sigma-Aldrich Co. Saccarose and ethylendiamine tetraacetic acid (EDTA) were supplied by Panreac S.A. (Barcelona, Spain). Deuterated PBS (D-PBS) was prepared by dissolving PBS powder in D₂O instead of H₂O.

Escherichia coli CECT101 was grown overnight in Luria-Bertani (LB) media, then transferred into new LB media to obtain an initial optical density value of 0.1 per cm at 660 nm, and allowed to grow aerobically at 37 °C. Cells in the middle logarithmic phase (optical density 0.35 per cm at 660 nm) were harvested by centrifugation (10 min, 3,000 rpm) and then resuspended in sterile PBS at pH 7.4. The procedure was repeated three times, yielding a bacterial concentration of 10⁸ colony forming units (CFU) mL⁻¹, as assessed by serial dilutions of the cell suspension followed by a colony formation assay.

The bacterial uptake of PS was determined by fluorescence spectroscopy following the procedure described by Hamblin et al. [26,41]. Bacterial suspensions (10⁸ CFU·mL⁻¹) were incubated in the dark at room temperature (NMB, ACS268) or at 37 °C (ZnTMPyPz) with varying concentrations of PS and contact times under gentle stirring. Afterwards, the cells were washed by centrifugation (10 min, 3,000 rpm, 3 times) to remove any excess of PS. After the last centrifugation, a solution of 1% SDS in 0.1 M NaOH was added to the pellets and shaken for a minimum of 24 h. Each experiment was repeated three times. Stock PS solutions were sterilized by filtration through a 0.22 µm sterile filter and stored in the dark at 4 °C for a maximum of 1 week.

Spheroplasts were obtained by resuspending treated or untreated E. coli cells in Tris-HCl 0.05 M buffer at pH = 6.8 containing 0.01 M EDTA, 0.3 M sucrose and lysozyme (1 mg mL⁻¹) under stirring. After 1 h incubation at 37 °C, tubes were centrifuged twice at 2,000 rpm for 10 min [42]. The pellet was then resuspended in PBS solution at pH = 7.4. PS-loaded spheroplasts were obtained by first incubating whole E. coli cells with the PS and then subjecting them to the procedure above.

Absorption spectra were recorded on a Cary 4E spectrophotometer (Varian, Palo Alto, CA, USA), equipped with a 110 mm diameter integrating sphere and high performance photomultiplier tube for transmittance measurements. Fluorescence emission spectra were recorded in a Spex Fluoromax-2 spectrofluorometer (Horiba Jobin-Yvon, Edison, NJ, USA). Fluorescence decays were recorded with a time-correlated single photon counting system (Fluotime 200, PicoQuant GmbH, Berlin, Germany) equipped with a red sensitive photomultiplier. Excitation was achieved by means of a 375 nm picosecond diode laser, or a 596 or 657 nm LED working at 10 MHz repetition rate. The counting frequency was always below 1%. Fluorescence decays were analysed using the PicoQuant FluoFit 4.0 data analysis software.

¹O₂ phosphorescence detected by means of a customized PicoQuant Fluotime 200 system described in detail elsewhere [43]. Briefly, a diode-pumped pulsed Nd:Yag laser (FTSS355-Q, Crystal Laser, Berlin, Germany) working at 10 kHz repetition rate at 532 nm (12 mW, 1.2 μJ per pulse) was used for excitation. A 1,064 nm rugate notch filter (Edmund Optics, Barrington, NJ, USA) was placed at the exit port of the laser to remove any residual component of its fundamental emission in the near-IR region. The luminescence exiting from the side of the sample was filtered by a cold mirror (CVI Melles Griot, Albuquerque, NM, USA) to remove any scattered laser radiation, and focused on the entrance slit of a Science Tech 9055 dual grating monochromator. A near-IR sensitive photomultiplier tube assembly (H9170-45, Hamamatsu Photonics, Hamamatsu City, Japan) was used as detector at the exit port of the monochromator. Photon counting was achieved with a multichannel scaler (Becker&Hickl MSA 300 or PicoQuant’s Nanoharp 250). The system is similar to that used in other laboratories [44]. Time-resolved ¹O₂ signals were analysed by fitting either Equation (1) or (2) to the data using the FluoFit software. The lifetime values quoted throughout the manuscript refer to mean and standard deviation values derived from at least three independent measurements. All spectroscopic measurements were carried out in 1-cm quartz cuvettes (Hellma, Müllheim, Germany) at room temperature.

In a typical experiment, bacterial suspensions were incubated in the dark with the desired amount of PS for a given period of time. The cells were then washed three times with PBS and resuspended in PBS to a final concentration of ~5 × 10⁸ CFU·mL⁻¹. For time-resolved phosphorescence measurements, the bacterial suspensions (3 mL) were irradiated with 3 million laser pulses at 532 nm. The suspensions were gently stirred during the measurements. Appropriate controls were performed to ensure that the signals originated from PS molecules bound to the bacterial cells (see Supplementary Information).