Synthesis and characterisation of flavins
Flavins F1-4 (2A) were prepared according to Schemes S1 and S2 (see ESI). Inspired by the classical amphiflavins developed by Trissel, Schmidt and Hemmerich,30–32 we chose to include an alkyl chain (C8) within the structure to improve phospholipid membrane incorporation. Additionally, the flavin chromophore itself was substituted with bromo groups in the case of F3-4 in order to harness the heavy-atom effect that increases the rate of ISC from singlet to triplet excited states, thereby potentially enhancing photosensitised generation of singlet oxygen.33 Heavy atom substitution has been shown to improve both the rate of ISC and the singlet oxygen quantum yield (ФΔ) in flavin derivatives used for synthetic photooxidation reactions.34 Although the heavy-atom effect has been used to boost photodynamic efficacy for other PS dyes in PDI applications, it has not yet been explored for flavin derivatives.35
To afford both methylated (F1-2) and brominated derivatives (F3-4), a Boc-protected ethylene amino component was first installed to the methylated or brominated arene core prior to cyclisation of the isoalloxazine ring system. Following N3-alkylation with an octyl chain on the isoalloxazine, N-Boc protecting groups were removed to yield the amino-functionalised F1 or F3 that were subsequently converted to guanidino moieties to achieve F2 or F4 respectively (Figure 2a). Guanidino groups were chosen as they are known to increase membrane coordination and penetration through strong guanidinium-phosphate H-bonding and we hypothesised would enhance PDI efficacy.
The UV-Vis absorption spectra of the flavins in DMSO revealed very similar absorption properties at the λ1 (S0→S1) band but a blue shift of 10-13 nm for the higher energy λ2 (S0→S2) bands of the brominated compounds (F3 and F4) presumably due to the electron withdrawing effect of Br atoms (Figure 2b, Table 1). In terms of emission properties, the heavy-atom effect of bromination can be clearly observed with F3 and F4 exhibiting severely reduced emission intensity (ФF <2% in DMSO, Table 1) when compared to the methylated F1 and F2 (ФF = 13% and 14% respectively in DMSO, Table 1). This effect can also be observed when comparing the efficiency of 1O2 production upon excitation, where brominated F3 and F4 demonstrate up to a 30% increase in activity over the methylated F1 and F2 in MeCN (Table 1), indicating their potential to be potent photodynamic agents. Despite this, F1 and F2 are still efficient 1O2 photosensitisers comparable with riboflavin (ФΔ = 0.54 ± 0.07). However, a smaller ФΔ value for guanidinylated F4 (62%) is observed compared to aminated F3 (85%) which could be explained by fast reverse ISC and/or solvent-dependent aggregation.
Table 1. Photophysical properties of flavins F1-4. λn = S0→Sn absorption band, λem = emission wavelength, φF = fluorescence quantum yield, φΔ = singlet oxygen quantum yield.
Flavin
|
λ1 (nm)a
|
λ2 (nm)a
|
λem (nm)a
|
φFb
|
φΔc
|
F1
|
444
|
349
|
507
|
0.129 ± 0.012
|
0.49 ± 0.12
|
F2
|
446
|
350
|
512
|
0.141 ± 0.010
|
0.55 ± 0.10
|
F3
|
443
|
336
|
508
|
0.009 ± 0.013
|
0.85 ± 0.09
|
F4
|
444
|
340
|
513
|
0.011 ± 0.010
|
0.62 ± 0.10
|
a measured in DMSO b calculated using riboflavin as the reference (ϕF = 0.226 ± 0.001 in DMSO) c caluclated using Ru(bpy)32+ as the reference (ϕΔ= 0.57 ± 0.06 in MeCN)
|
After successful synthesis and initial characterisation, we monitored the photostability of the compounds in PBS (1x, pH 7.4) which was used for PDI assays as distilled or ultrapure water destabilises bacterial cells and coronaviruses through osmotic pressures,36,37 thereby augmenting inactivation results. Flavins F1-4 and riboflavin (100 µM in PBS) were therefore irradiated with a 6200K white LED light source (18W, 400-700 nm, see ESI Figure S1 for emission spectrum) at an illuminance of 1x105 lx (35 mW/cm2 irradiance) to resemble typical daylight,38 and the changes in their UV-Vis absorption were monitored over time (see ESI, Figure S4). For riboflavin, rapid photodecomposition (80%) was observed over 30 min irradiation which is known to be due to intramolecular dealkylation of the ribityl chain (Figure S4a, S4b). This yields lumichrome as the major degradation product which can only act as a PS under UV-irradiation.39 Amino-containing F1 and F3 exhibited around 40% and 60% degradation respectively after the same irradiation time (Figure S4a, S4c and S4e), whereas F2 and F4 degraded by approximately 10% and 30% respectively (Figure S4a, S4d and S4f). These findings corroborate previous work showing amino-containing flavins photodegradation in the presence of phosphate ions which resulted in diminished bacterial PDI efficiency,40 however it appears that guanidino substitution improves photostability in this case. Interestingly, bromo-substituted flavins show higher rates of photodegradation which could be explained by their higher ФΔ values resulting in greater 1O2-induced degradation.
Photodynamic inactivation of E. coli
Having observed clear trends in the photophysical properties of methylated and brominated derivatives, we were interested to see how this would affect the PDI efficacy of F1-4 against pathogens. First, we investigated the inactivation of the Gram-negative bacterium, E. coli BL21(DE3). The cell envelope of Gram-negative bacteria presents a formidable barrier to antimicrobial compounds that consequently inhibits PDI efficacy compared to the analogous structure in Gram-positive bacteria.41 Following an initial 20 min incubation of the flavin compounds with E. coli in PBS at various concentrations in the dark, the mixture was irradiated (1x105 lx) and the number of surviving colony forming units (CFUs) were determined. After 15 min (31.5 J/cm2 light dose) of irradiation, no inactivation was observed at 1 µM for either riboflavin or F1-4. At 10 µM approximately 1 log10 reduction of E. coli CFUs was observed for F2 and a 2.8 log10 reduction in the presence of F4 (Figure 3a, Table S1). At higher concentrations, this effect was greatly enhanced with both F2 and F4 exhibiting >6.0 log10 reduction of CFU/mL at 100 µM. From these data, it can be derived that 11 µM of F4 and 42 µM of F2 can achieve >3 log reduction (>99.9%) of bacterial load which is considered a minimum level of decontamination according to EMA guidance, whereas >6 logs is the highest.42
Interestingly, riboflavin and F1 demonstrated no activity at 100 µM, whereas F3 exhibited a 1.8 log10 reduction in bacterial load. To ensure effective PDI was occurring, the flavin compounds were incubated in the dark under the same experimental irradiation conditions (100 µM, 15 min) to reveal no bacterial toxicity (see ESI, Figure S5). The rate of E. coli inactivation over time was then monitored at a 100 µM flavin concentration revealing an extremely rapid reduction of CFUs in the presence of F4, facilitating >6.0 log10 reduction after just 5 min of irradiation (10.5 J/cm2 light dose). Similarly, F2 shows effective bactericidal activity with a 3.4 log10 reduction (>99.9%) after 5 min irradiation. A closer investigation into the speed of F4’s activity at 100 µM showed that after only 1 min of irradiation (2.10 J/cm2 light dose) a 4.1 log10 reduction (>99.99%) of bacteria was achieved (Figure 3b).
These data clearly show that the introduction of a guanidino moiety, as in the case of F2 and F4, greatly increases the flavin’s PDI efficacy against the bacterium when compared to amino-containing F1 and F3. This may be attributed to better photostability in PBS, as well as the guanidino group facilitating better coordination to the cytoplasmic phospholipid membrane resulting in improved permeability and lipid peroxidation under irradiation.43 In addition, the replacement of methyl substituents with bromines improves efficacy when the same cationic group is compared. This can be explained by more efficient generation of singlet oxygen, as predicted by their ФΔ values shown in Table 1. As a result, greater lipid and biomolecule oxidation can be achieved to inactivate the pathogen. It should also be noted that the inclusion of bromine atoms increases the lipophilicity of the molecule which can further improve cell membrane permeability and incorporation. However, despite having the highest predicted ФΔ value (85%), F3 did not outperform F2 (ФΔ= 55%) which demonstrates the overarching impact of guanidino substitution.
To gain initial mechanistic insight, we investigated the cellular localisation of fluorescent riboflavin, F1 and F2 by structured illumination microscopy (SIM) (Figure 3c, see ESI Figures S6-8). After incubation with the flavin compounds (100 µM), large field of view images revealed marked differences in the number and degree of fluorescent bacterial populations (see ESI Figures S6-8). Although the exact quantification of flavin uptake was not possible due to uncharacterised optical properties in a complex biological environment, the qualitative comparison of F2 and F1 (which have similar ФF in DMSO), revealed a higher number of fluorescent bacteria with bright fluorescent intensity after incubation with F2, especially within membranes (Figure 3c, Figure S6 and S7). This indicates that guanidino substitution does improve uptake, however the distribution of fluorescent intensity was not homogeneous across different bacterial cells which could be related to the amphiphilicity of F2, resulting in variable uptake due to aggregation (Figure S6). For hydrophilic riboflavin, very weak fluorescent populations were observed, most likely explained by the controlled transport of the compound through outer membrane porins of the bacterium and therefore unable to bind effectively to the outer or cytoplasmic membranes (Figure 3c, Figure S8). This therefore helps to explain the lack of PDI efficacy observed when using riboflavin.
Furthermore, we monitored the pDNA (pUC18) cleavage in the presence of the flavins (10 μM) under irradiation which indicated guanidino derivatives possess superior photocleavage ability (Figure 3d). The photocleavage of supercoiled (SC) to nicked coiled (NC) pDNA structures was far enhanced for F2 and F4 (39% and 21% respectively) when compared to F1 and F3 (≤3%) after irradiation for 15 min (Figure 3d). This difference in activity could be explained by favourable guanidinium-phosphate interactions that increase the likelihood of electron transfer events between flavin and guanosine which are known to primarily contribute to DNA cleavage alongside 1O2-mediated oxidation.44,45 Therefore, even if the amino-flavin compounds diffuse into the cytoplasm, it is unlikely that damage to constituent nucleic acids would contribute to PDI of the pathogen. Our control compound, riboflavin (Rbf) also exhibits photocleavage of the plasmid (11%) which has been reported previously,44,45 however it is clear from the SIM data that, due to its hydrophilic nature and lack of cationic substituent, little cell uptake of the compound is achieved for this to contribute to PDI.
Collectively, this initial mechanistic study demonstrates that even in the case of increased singlet oxygen production, the key component to achieve effective Gram-negative bactericidal activity under irradiation is the presence of the guanidino group, which facilitates enhanced cell uptake and nucleic acid degradation.
Photodynamic inactivation of murine hepatitis virus (MHV-A59)
Encouraged by the identification of such highly effective guanidino-flavins for bacterial inactivation, we were interested to see if the trend would be similar for coronaviruses, specifically, murine hepatitis virus A59 strain (MHV-A59) which is often used as a surrogate for other coronaviruses.4 It has already been demonstrated that riboflavin can effectively inactivate both enveloped and non-enveloped viruses in blood products using UV light,20–23 but much lower efficacy was observed using visible light (0.4x105 lx, 0.5-2 h) against hepatitis B virus (HBV).46,47 In order to evaluate possible applicability towards virus-inactivating surface coatings or textiles, we used an in vitro TCID50 assay to evaluate the viral titre of MHV-A59 through inoculation into murine fibroblast 17Cl-1 cells after irradiation at varying concentrations of flavin in PBS. The cytotoxicity of the flavins towards this cell line was first investigated by 24 h incubation MTS assay allowing us to obtain a working concentration range of ≤10 μM for the in vitro TCID50 assay to evaluate viral PDI efficacy (see ESI Figures S9 and Table S3 for 17Cl-1 cytotoxicity data).
We started our investigation by varying the concentration of flavin (1-10 μM in PBS) with 10 min of white LED exposure (21.0 J/cm2 light dose, Figure 4a). Even at 1 μM, the brominated guanidino flavin F4 demonstrated a 3.8 log10 reduction in viral titre, while F2 and F3 showed 2.1 log10 reduction (Figure 4a, Table S3). It should be noted that reductions of the order of 4 logs or more (>99.99%) are considered highly effective by EMA guidance and that a >1 log10 reduction in necessary to be considered reliable.48 At the same concentration, riboflavin achieved a 1.2 log10 reduction in titre, however F1 showed no effect under these conditions. Nevertheless, the activity of all flavins improved by the increase of their concentrations resulting in viral load reductions of >99.9% in the presence of riboflavin or F2, and >99.99% with F3 or F4 at 5 μM. Further increase in concentration (to 10 µM) only substantially improved virucidal activity for F4 (>5 log10). The irradiation time was then explored for 10 μM flavin to reveal high degrees of inactivation (≥5 log10) for F2, F3 and F4 when irradiated for 15 min (31.5 J/cm2 light dose) whereas shorter irradiation times (5 min, 10.5 J/cm2 light dose) still provided efficient inactivation of MHV-A59 ≥3 log10 steps using those same flavins (Figure 4b). It should be noted that variability within all experiments was noticeably high which is typical for TCID50 assays.49 However, no reliable log reduction of MHV-A59 was observed when irradiated without flavin in PBS containing 0.1% DMSO (Figure 4b). To confirm that the mechanism of viral inactivation by the flavins was dependent on light, dark control experiments (10 μM, 10 min incubation) resulted in no effective virucidal activity being observed (>1 log10 reduction, Figure 4c).
The activity of the flavin derivatives towards viral inactivation show a different trend than previously observed with our model bacterium. For example, in the presence of riboflavin under the same conditions, effective PDI of MHV-A59 was measured whereas there was no activity towards E. coli. Therefore, it can be assumed that the requirements for a PS to interact and/or diffuse through the coronavirus membrane are less dependent on lipophilicity or the presence of cationic charge. Despite this, our flavin derivatives again show that guanidino group incorporation improves antiviral PDI activity when compared to amino groups. However, bromination now seems to play a more important role in viral deactivation as brominated amino F3 outperformed methylated guanidino F2.
To try and explain our findings, we investigated the interaction of the flavins with RNA to elucidate how its light-induced cleavage could lead to coronavirus inactivation, a model ssRNA (~1k nt) was irradiated (1x105 lx) in the presence of 10 µM flavin in PBS for 15 min and the extent of cleavage was observed via agarose gel electrophoresis (Figure 4d). It was observed that F1 exhibits the least RNA photocleavage (12%) whereas riboflavin shows greater activity (28%) which corroborates the result of viral inactivation and confirms previous findings.50-52 Similar to our results with pDNA, guanidino-containing F2 and F4 show the best photocleavage ability (46% and 42% respectively). Interestingly, F3 also exhibits effective cleavage of RNA (35%) which was not observed in the case of pDNA (Figure 3d). This can be rationalised by the mechanism of flavin-mediated RNA photocleavage which has been previously shown to depend more upon 1O2 oxidation than electron transfer events between flavin and nucleobase.50-52 Taken together, it can be concluded that efficient 1O2 production is key to viral PDI using flavin derivatives.
In vitro toxicity towards human cells
Finally, before considering the applications of our flavin derivatives for PDI of topical pathogen infections, surface coatings or textiles, we investigated their impact on human cells. In general, an ideal photosensitiser for these applications should show no cytotoxic effects on human cells in the dark and limited effects given the same irradiation conditions used for inactivation of the target pathogen, also referred to as a therapeutic window.12 Accordingly, the inherent cytotoxicity of the flavins was evaluated through incubation with human lung fibroblast cells (WI-38). A 24 h MTS assay was used to calculate inhibitory concentration values (IC50) of >100 µM for riboflavin and guanidino-functionalised F2 and F4 which can therefore be considered as non-toxic (Figure 5a, Table S4). On the other hand, amino-functionalised F1 and F3 had IC50 values of 96.1 µM and 30.9 µM respectively (Table S4). It has been shown previously that the cytotoxicity of amino-containing compounds may be derived from an increase in intracellular amine oxidase activity that induces excess oxidative stress leading to cell apoptosis.53
We then studied the light-induced toxicity of the flavins towards the WI-38 fibroblast cells by observing decreases in cell viability over irradiation time using an MTS assay (Figure 5b). Interestingly, the cytotoxicity trends observed in the dark after 24h are reversed when exposed to light (1x105 lx) over shorter time periods. For example, after 15 min of irradiation, 10 μM of methylated guanidino F2 induces ~75% reduction in cell viability. Under the same conditions, brominated guanidino F4 exhibits less of an effect with ~20% reduction in cell viability. The degree of variability in these experiments could be attributed to the amphiphilic nature of the flavins leading to aggregation of the compounds in aqueous conditions. On the other hand, riboflavin, amino-functionalised F1 and F3 display no decrease in cell viability over 15 min irradiation. These data show that despite being non-toxic to the cells in the dark, the guanidino-functionalised flavins can induce cytotoxicity upon irradiation most likely through similar mechanisms discussed for pathogens. It is our ongoing work to understand the differences in light induced cytotoxicity that were observed between methylated and brominated guanidino flavin derivatives. Nevertheless, F4 could be suitable for further PDI applications thanks to effective bacterial and viral inactivation (>3 log10 reduction after 15 min irradiation at 11 μM and 1 μM respectively) coupled with low toxicity to human fibroblast cells under the same conditions or in the dark.