A Molecular Rotor that Measures Dynamic Changes of Lipid Bilayer Viscosity Caused by Oxidative Stress

Abstract Oxidation of cellular structures is typically an undesirable process that can be a hallmark of certain diseases. On the other hand, photooxidation is a necessary step of photodynamic therapy (PDT), a cancer treatment causing cell death upon light irradiation. Here, the effect of photooxidation on the microscopic viscosity of model lipid bilayers constructed of 1,2‐dioleoyl‐sn‐glycero‐3‐phosphocholine has been studied. A molecular rotor has been employed that displays a viscosity‐dependent fluorescence lifetime as a quantitative probe of the bilayer's viscosity. Thus, spatially‐resolved viscosity maps of lipid photooxidation in giant unilamellar vesicles (GUVs) were obtained, testing the effect of the positioning of the oxidant relative to the rotor in the bilayer. It was found that PDT has a strong impact on viscoelastic properties of lipid bilayers, which ‘travels’ through the bilayer to areas that have not been irradiated directly. A dramatic difference in viscoelastic properties of oxidized GUVs by Type I (electron transfer) and Type II (singlet oxygen‐based) photosensitisers was also detected.


Introduction
Unsaturated lipids are commonly found in av ariety of biological membranes and are vulnerable to reactive oxygen species (ROS) such as singlet oxygen and oxygen-based radicals. Oxidized lipid molecules have been shown to playar ole in the regulation of immune responses. [1] However,l ipid oxidation products are more commonly associated with disruptingn atural cellular processes, and contribute to aging and diseases such as Parkinson's, Alzheimer's,a therosclerosis, and cancer. [2,3] In the limiting case, extreme oxidation of cellular components can lead to cell death through apoptosis or necrosis, and this effect is successfully used in photodynamic therapy (PDT), at ype of cancer treatment. [4] PDT is al ight-activated process where a 'photosensitizer',am olecule that produces ROS upon excitation by an appropriate wavelength of light, is targeted to malignant cells and tissues. Locally produced ROS efficientlyo xidizec ellularc omponents, leading to the death of targeted cells. Given their abundance in cells, lipids serve as primary targets for ROS during PDT and membrane oxidation is ak ey step leadingt ocell apoptosis. [5,6] Consequently, significant effort has been made to understand physicochemical changes in membranes undero xidative stress.I nm odel membrane systemst he appearance of oxidized lipids was reported to increase the membrane surface area, causing spontaneous fluctuations of the membrane [7,8] and alterations in membranec urvature, [9] permeability, [10] and packing order. [11] Lipid oxidation has also been shownt oa ffect diffusion in model membranes. [9,12] However, the majority of the aforementioned effects have been observed in the bulk solution of model membranes (in large unilamellar vesicles, LUVs)l acking spatial resolutiona cross the bilayer of an individual vesicle.
Fluorescence microscopy is ap owerful tool for the visualization of lipid membranes. Consequently, aw ide range of fluorescent probes suitablef or probing multiple properties of lipid membranes was developed, [13] including probesf or sensing membrane potentiala nd fluidity, [14] for detecting lipid order in the outer lipid leaflet of the lipid bilayer [15] and for sensing changes in the membrane during apoptosis. [16] In this work, we utilized BODIPY-C 10 [17,18] (Figure 1), af luorophore that belongs [a] A. Vyšniauskas to ag roup of dyes termed 'molecular rotors' that have viscosity-dependent fluorescenceq uantum yields, lifetimes, [19,20] and depolarization. [21,22] When combined with fluorescence lifetime imaging microscopy (FLIM), molecular rotors can be used to obtain spatially resolved viscositym aps of microscopico bjects, [17,[23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40] as well as to observed ynamic change in viscosity during relevant processes of interest. [37,39,41,42] Thus,w ea imed to use BODIPY-C 10 ,w hich is known to completely embedi nto the fluid-phase lipid bilayers [40] to directly examine how photooxidation during PDT affects viscoelastic properties of model lipid membranes, with spatial-and time-resolution.A sam odel system,w eh ave employed giant unilamellar vesicles (GUVs) composed of an unsaturated lipid 1,2-dioleoyl-sn-glycero-3phosphocholine( DOPC), which is susceptible to oxidation by av ariety of ROS. By imaging GUVs we are able to monitor the effects of oxidation away from an initially irradiated bilayer site, providing information on the nature of ROS involved in viscosity change and the mechanism of its action. There are two distinct pathways for ap hotosensitizer to create ROS, Types Ia nd II. [43] The reactiond iagrams of the resulting speciesw ith lipids are shown in Figure S1 in the Supporting Information.
In aT ype II process, at riplet-state photosensitizer can transfer its energy to ag round-stateo xygen molecule, producing singlet oxygen, 1 O 2 . 1 O 2 is an oxidant, which is knownt or eact with unsaturated lipids, producing peroxidation products. [7] On the other hand, in aT ype Ir eaction, at riplet-state photosensitizer can act as an electron donor or can abstract hydrogen from surrounding molecules,c reating radicals. Thus, aT ype I oxidation process does not stopa taperoxidation stage,a nd the reactionp roceeds furtheru ntil the lipid molecule is cleaved along the double bond. [43] We have selected ar ange of photosensitizerst hat participate in either Type Io rT ype II chemistry. Their structures are shown in Figure S2 in the Supporting Information. Firstly,w e have used three differentp orphyrin-based photosensitizers that are known to resulti nT ype II oxidation by singlet oxygen, but occupy differentp ositions relative to the hydrophobic core of al ipid bilayer.N amely,1 )hydrophobic tetraphenylporphyrin (TPP) will reside in the tail region of the lipid membrane;2 )hydrophilic tetrakis(4-sulfonatophenyl)porphine (TPPS 4À )w ill reside in the aqueous solution on the outside of the bilayer; and 3) ap orphyrin dimer (PD), which was previously demonstrated to attach to the surface of the lipid bilayer. [39] Secondly, we used methylene blue (MB) as ap hotosensitizer,w hich is known to participate in Type Ireactions. [44] Here we demonstrate that FLIM of lipid bilayers containing molecular rotor BODIPY-C 10 is ap owerful toolf or studying change in viscoelastic properties of membranes during oxidation. We examine how the localization of photosensitizer affects the bilayer's viscosity.F inally,w es how ac lear difference in the evolution of the membrane's viscoelastic properties during Type Iand Type II photooxidation.

Results and Discussion
We have prepared as eries of DOPC GUVs that contained BODIPY-C 10 ,o ur molecular rotor,a nd variousp hotosensitizers. We made sure that the viscosity-sensitive fluorescences ignal of BODIPY-C 10 ,( recorded between5 10-600 nm) can be clearly separated from the fluorescenceo fp hotosensitizers. The absorptions pectra of all dyes used in this study are given in Figure S3 in the Supporting Information. PD alone absorbs at the excitation wavelength of BODIPY-C 10 (480 nm). However,e ven thoughP Da bsorbs at 480 nm, its fluorescence is centered at 630-750nm, [39] which does not overlap with fluorescence of BODIPY-C 10 (510-600nm). Thus, we made sure that the timeresolved fluorescenced ecays recorded belong to the molecular rotor and are not contaminated by the signal from the PDT photosensitizer used. We note that all the photosensitizers can be individually excited using internal microscope laser wavelengths as shown in Figure S3 in the Supporting Information.

Type II photooxidation
The first set of GUVs studied contained PD, ak nown singlet oxygen photosensitizer [45,46] that is bound to the lipid bilayer surface [39] and BODIPY-C 10 ,w hich probesv iscosity of the inner part of the lipid bilayer. [40] Previously,w eh ave utilizedP Da s both the photosensitizer and as am olecular rotor and recorded al arge increasei nt he DOPC monolayer viscosity upon PDT. [39] Here, we aimed to separateo ut the dual function of PD as ap hotosensitizer and am olecular rotor.B yd oing so we aimed to test whether:1 )the viscosity increasec an be observedi ndependently of the rotor used and its positioning in the bilayer and 2) whether ROS can penetrate the bilayer from the surface of the membrane when produced by an externally boundp hotosensitizer and can cause av iscosity increase within the bilayer,a sp robedb yB ODIPY-C 10 .T he results are presented in Figure 2. We have selected as ingleG UV (shown by the red arrow) by zooming in and irradiated it at 453 nm, where PD absorbs. Throughouti rradiation we acquireds everal FLIM images of BODIPY-C 10 .I ti sc leart os ee that progressiveirradiation caused ac ontinuous increase in fluorescencel ifetime from 1509 AE 26 to 2254 AE 53 ps, corresponding to av iscosity increasefrom 170 AE 5t o332 AE 14 cP ( Figure 2B).
Furthermore,w ep erformed three control experiments( see Figure S4 in the Supporting Information). First, we prepared DOPC GUVs containing BODIPY-C 10 only,without PD, and irradiated it at 453 nm. Secondly, we prepared GUVs using DPhPC, as aturated lipid that does not contain double bonds, making the bilayer unreactivet oR OS. Finally,w et ested the irradiation effects in the presence of 0.11 m NaN 3 ,a ne fficient singlet oxygen quencher.I na ll three controle xperimentsn oc hange in fluorescencel ifetimeo fB ODIPY-C 10 was observed during irradiation ( Figure S4 in the Supporting Information). This data allowed us to concludet hat, for the viscosity change seen in Figure2 to take place, both singlet oxygen and unsaturated bonds are required;t herefore, this change is likely caused by the oxidation of unsaturated bonds in lipid molecules by singlet oxygen.
It was suggested previously that singlet oxygen reacts with double bonds in an 'ene' type reaction, which leads to the insertion of ah ydroperoxideg roup next to ad ouble bond in lipid molecules. [7,43] Such ac hange is likely to force the reacted lipid molecule to curve in order to insertthe formedhydrophilic hydroperoxide group into the aqueous phase. [7] We hypothesize that this changeleads to an increaseofm icroviscosity in the hydrophobic core of the membrane where BODIPY-C 10 resides. We stress the fact that the lifetimec hanges gradually in the whole vesicle, which meanst hat microviscosity in the GUV increases gradually with the increasing amount of the oxidizedl ipid molecules without any visible phase separation. The change is accompanied by the loss of lipid material from the vesicle, which was observed during the imaging. Examples of such behavior can be seen in Figure2Aa fter 48, 127, and 520 so fi rradiation, and in Figure 3C after 125 so fi rradiation, where the irradiated vesicles show thin structurese xtending away from the lipid shell. The loss of lipid materialisconsistent with data reported previously. [47][48][49] We note that the fluorescence lifetimes of other DOPC vesicles in the large fieldo f view ( Figure 2) increases lightly as well, even thought hey were not directly irradiated. This could be due either to oxidation during FLIM imaging of BODIPY-C 10 or by longer-lived ROS, whichdiffused away from the initially irradiated region.
We next set out to investigate if the dynamics of microviscosity increasec an be affected by the location of the photosensitizer relative to the rotor within the bilayer,F igure 3. We used ah ydrophobic porphyrin, TPP,w hich,d ue to its hydrophobic structurea nd neutral charge, is expected to be fully embedded in the hydrophobic core of the lipid bilayer.N ext, aw ater soluble porphyrin,T PPS 4À ,w as used, whichr eadily dissolves in an aqueous solution and does not strongly interact with the lipid bilayer,a sc onfirmed by fluorescencei maging ( Figure S5 in the SupportingInformation).
Upon irradiation of individual vesicles ( Figure 3), in both cases the lifetime of BODIPY-C 10 increased gradually,inasimilar manner that shown in Figure 2w ith PD, from 1550-1600 to 2500-2750 ps, corresponding to viscosity change from 180 to 400-500cP. Throughout the oxidation, vesiclesleaked lipid material and went through as tage of rapid fluctuations ( Figure 3), similar to what was observed using PD as as ensitizer.W hen the irradiation was paused during as hape fluctuation, the majority of vesiclesr etained their deformed shape (e.g.,F igure 3b at 190 s). This deformed shape did not produce inhomogeneous viscosity distribution across the vesicle. Surprisingly,t he GUV returned to its spherical shape following further irradiation. We hypothesize that these temporary changes in shape are due to changes in the curvature of the bilayer induced by the presence of oxidized lipids. However,t hesea re then released by excess lipid shedding during furtherirradiation.
Ta ken together,t he results recorded in the presence of three singlet-oxygen photosensitizers seem to indicate that the membrane is efficientlyo xidized by singlet oxygen, irrespective of whether it was produced inside, outside, or on the surface of the bilayer.W ew ould like to point out the factt hat, due to as hort lifetimeo f 1 O 2 in water (3.5 ms), [50] it has limitedt ime to diffusei nto the membrane unless sensitized in close proximity to the bilayer,byt he water-soluble photosensitizer TPPS 4À .
Next, we set out to examine the mobility of oxidized lipids within asingle GUV.Weirradiated part of asingle DOPC vesicle and followed ac hange in lifetimeo fB ODIPY-C 10 in the whole vesicle using FLIM (Figure4). TPP was chosen as ap hotosensitizer because it embeds in the hydrophobic part of the lipid bilayer and is not present in the aqueous solution outside the bilayer.
It is clear to see that the irradiated section of GUV (showed by the red rectangleinFigure 4) has an increased lifetimecompared to the rest of the vesicle and the rest of the image. The fluorescencel ifetimeo fB ODIPY-C 10 in the irradiated section of the vesicle (Figure 4, Region 1) increased faster during irradiation, compared to the value in the non-irradiated part in the same GUV (Region 2). Both these values werec onsiderably higher than the lifetime in an on-irradiated GUV (Region 3) adjacent to the irradiated vesicle. Though the lifetimei ncrease was the highest in the directly irradiated area (ca. 1500 to ca. 2500 ps, equivalent to 170 to 400 cP), we observed that the lifetimeofB ODIPY-C 10 ,i ndicative of microviscosity,g raduallyi ncreasedi nt he whole vesicle. For example, the non-irradiated part of the same vesicle (Region 2) showed an increase from % 1500 to % 2250 ps, equivalent to 170 to 330 cP.A tt he same time, an adjacent vesicle (Region3)s howeda lmost no lifetime change at all. This lacko fc hangei nR egion 3m usti ndicate that, as expected, the ROS responsible form icroviscosity increase in the DOPC bilayer can efficiently travel through the bilayer (to Region 2), but cannot travel longd istances in an aqueous solution.
Although single oxygen is the main ROS produced by the Type II photosensitizer TPP,w en ote that, given the known lifetime of 1 O 2 in the DOPC lipid bilayers of % 35 ms, [51] it is only expected to travel approximately 3 mmd istance within three times its lifetime. [5] An alternative explanation is that the ROS initially produced cause lipid peroxidationa tt he point of irra-diation and then the lipid oxidation productst ravel within the same vesicle (but not in an aqueous solution), leading to ag radual viscosity increaseint he whole vesicle.
To distinguish between these two scenarios, we partially irradiated aG UV (see Figure 5) for 1min, to achieve ac ontrast in viscosity between the irradiated and the non-irradiated part (equivalent to 125 ps lifetimed ifference). We consequently recorded aF LIM image of the same vesicle 15 min later,t os ee if the viscosity variations within the vesicle remained after the irradiation was complete.T he premise here was that most of the short-lived ROS, and in particulars inglet oxygen, willd ecay and/or react during the irradiation period only,a nd will not be able to causef urthero xidation after the irradiation was completed.T he oxidized lipids, on the other hand, should be able to diffuse across the whole vesicle even after the irradiation has been completed.
Our data ( Figure 5) show that, immediatelyf ollowing irradiation, the fluorescencel ifetime of BODIPY-C 10 in the irradiated and the non-irradiatedp arts were 1950 and 2075 ps, respectively,c orresponding to viscosities of 260 and 290 cP,w hereas in the image recorded following a15min delay,wedid not ob-serve any variations in the viscosity across the vesicle, giving the overall lifetime of % 2100 ps. This data appears to indicate that diffusion of the lipid peroxidation product contributes to the observedp rocesses, because singlet oxygen is not ablet o survivef or 15 min and causet he observed changes.O nt he other hand, we note that the viscosity across the vesicle did not simply stagnate between 260 and 290 cP;i nstead, the non-irradiated region displayed av iscosity increase, producing the high final viscosity of 290 cP in the whole vesicle. This result suggests that ar elativelyl ong-lived oxidizing species is presenti nt he system and is able to diffuse within the vesicle, causingf urther oxidation even after the irradiation was complete. One possibility is that lipid peroxides that diffusea cross the vesicle can slowly decompose, leadingt of urther ROS formationt hat oxidize other lipids in the vicinity.W en ote that the fluctuations of the vesicle shape observed in the courseo f our experiments must assist in lipid mixinga nd diffusion within the continuous bilayer length.

Type Iphotooxidation
As previously mentioned, Type Ip hotosensitization involving electront ransfer or hydrogen abstraction, can occur from the photosensitizer triplet state. [43] In the previous section we utilized porphyrin-based photosensitizers characterized by high  singlet-oxygen yields that were likely to participatei nT ype II photosensitization. Here, we comparet his with microviscosity evolution in the presence of methylene blue (MB), ap hotooxidant with Type Iproperties. [44] The results of 633 nm irradiation of as ingle vesicle in the presenceo fM Bi nt he incubation medium are showni n Figure 6. The observed evolution of the BODIPY-C 10 fluorescence lifetimew as surprising and strongly contrasted the results obtained with porphyrin-based photosensitizers. During irradiation of asingle vesicle with MB presentint he incubation solution, the mean fluorescencel ifetime of BODIPY-C 10 decreasedr ather than increased, asw eo bserved with Type II photosensitizers ( Figure 6). Furthermore, the BODIPY-C 10 fluorescenced ecaysi nt he irradiated vesicles were best-fitted with abiexponential function (see Figure S6 in the Supporting Information).
BODIPY-C 10 is characterizedb ym onoexponential fluorescence decays, if recorded in homogeneousm ediuma nd in the absence of the aggregates. [17] Hence, the biexponential decays observedd uring Type Ip eroxidation can be either the result of two kinds of lipid environments createdb yp eroxidation or the presenceo fa ggregates. [17] We first tested if biexponential decays of BODIPY-C 10 are the result of the aggregation of BODIPY-C 10 .T he aggregates are known to cause quenching of the BODIPY-C 10 monomeric species, with the long-lived fluorescence of the aggregates appearing in the red region of the spectrum( > 570 nm), overall leading to ab iexponential decay. [17] Hence, we recorded the fluorescence decay traces over two detection windows, 500-550a nd 600-650 nm (Figure S6 in the Supporting Information). We observed no difference between the two traces recorded over these different detection ranges, which ruled out the presence of aggregates in the oxidized GUVs.
We have also tested if the biexponential decay was caused by oxidation of BODIPY-C 10 itself by MB. For this, we repeated the oxidation experiment using GUVs made out of saturated DPhPC lipid, which is resistantt oo xidation by ROS.T he DPhPC GUVs showed no change in lifetimeu poni rradiation, confirming that the BODIPY-C 10 itself was not affectedb yt he ROS produced ( Figure S7 in the Supporting Information). It should be noted that, in all experiments involving MB as ap hotooxidant, bleaching of BODIPY-C 10 was observed. However,t he constant lifetimeo bserved in the DPhPC controle xperiment confirms that the bleaching does not affect the validity of FLIM data.
Thus, the presence of the two lifetimec omponents in the fluorescenced ecays recorded during Type Io xidation of DOPC GUVs indicate the presence of two environmentsi nt he lipid bilayer,a ss ensed by the molecular rotor.A gain, we did not detecta ny large-scale phase separation in the oxidized vesicle, which meanst hat the domains that form were small and below our resolution limit. Alternatively,i ti sp ossible that, in the bilayer produced by Type Io xidation, BODIPY-C 10 is able to adopt two positions,s imilart ow hat was previously detected in gel-phase bilayers constructed from 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC)o rs phingomyelin. [40] The fluorescence lifetimes extracted from the biexponential fitting of oxidized GUVs were both decreasing during the irradiation (Figure S8 in the Supporting Information), indicating the decrease in the effective viscosity sensedb yB ODIPY-C 10 .
This intriguing decrease in viscosity observed during Type I lipid photooxidation can be rationalized based on the literature data. It is known that Type Ioxidation results in acleavage through ad ouble bond in lipid molecules, [43,47,48] which in turn could lead to av ery loose packing of lipids and increased volumef or BODIPY's intramolecularr otation. Pore formation [43,47,48] and ah igher fluidity of the membrane [12] were previously reported as ar esult of Type Io xidation, which agree with our findings using directv iscositym easurements with the molecular rotor.
Finally,w eperformed oxidation experiment of DOPC GUVs, with MB as ap hotosensitizer in the presence of NaN 3 as as inglet-oxygen quencher ( Figure S9 in the Supporting Information). Here, we also observedb iexponential fluorescence decays of BODIPY-C 10 following irradiation, and saw ad ecrease in mean fluorescence lifetime. However,t he lifetimec hange observed in the presence of NaN 3 was significantly slower, which was consistent with MB being mixed Type Ia nd Type II oxidant.N aN 3 stops Type II oxidation path, which must otherwise assist Type Ioxidation through peroxidef ormation.Nevertheless, ap ure Type Io xidation (through peroxide formation followed by lipid cleavage) [22] is able to proceed, which led to ad ecrease of BODIPY-C 10 lifetime, as in the absence of NaN 3 .

Conclusion
In conclusion, we have utilized ah ydrophobic molecular rotor that was fully embedded in al ipid bilayer to probe the effect of photooxidationo nt he bilayer viscosity.W ee stablished that Type II photooxidation of unsaturated lipid bilayersc auses al arge increase of bilayer microviscosity,i rrespectiveo ft he relative positiono ft he photosensitizer used to produce singlet oxygen:o utside, within, or on the surfaceo ft he bilayer.B yi nvestigating the evolution of viscosity in partially irradiated vesicles, we concluded that the viscosityc hange is likely caused by the diffusion of lipid peroxides within the bilayer.I nc ontrastto Type II oxidation, we have detected al arge decrease in viscosity during Type Iphotooxidation using methylene blue as apho- Figure 6. FLIM of DOPC GUVs containingBODIPY-C 10 as av iscosity probe and awater-solubleMBa sp hotosensitizer (10 mm). The GUV irradiated at 633 nm is shownbyt he red arrow; irradiation times are shown above the images. The distributions of intensity-weighted mean lifetimes in an irradiated GUV are shownint he panelsb elow eachimage.
tosensitizer.T hus, molecular rotor BODIPY-C 10 can clearly discriminate between the microviscosity changes during Type I and Type II chemistry.O verall, our resultsd emonstrate that viscosity sensorB ODIPY-C 10 is av ery useful tool for investigating photooxidationo fmodel lipid membranes, which provides spatialand temporal resolution unavailable previously.

Experimental Section
Materials PD was synthesized as described [52] previously and kindly provided by the group of Prof H.L. Anderson. BODIPY-C 10 was synthesized by the method previously reported. [53] TPP was obtained from Aldrich (97 %p urity), TPPS 4À was obtained from Sigma-Aldrich, and MB was obtained from Hopkin and Williams. Stock solutions of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) and 1,2-diphytanoyl-snglycero-3-phosphocholine (DPhPC) in chloroform were obtained from Avanti Polar Lipids. All solvents used were of spectroscopic grade.

Preparation of giant unilamellar vesicles (GUVs)
GUVs were prepared by electroformation. A5mLo fmixture of lipid (2 mg mL À1 )w as spread on ITO (indium-tin oxide) slide on 1cm 2 area. Then the solution was evaporated under 2M Pa pressure for 30 min. The electroformation chamber was then assembled out of two ITOs lides with polydimethylsiloxane (PDMS) spacer in between. The chamber was filled with 200 mm sucrose solution in water,c onnected to aT Ti TG550 function generator and 1.2 Vv oltage at 10 Hz frequency was applied for 1.5 h. The chamber was kept incubated at 60 8Cd uring electroformation. The newly formed GUVs were transferred into aqueous 200 mm glucose solution in Lab-Tek chamber slides for imaging. For appropriate experiments, water-soluble photosensitizers (TPPS 4À ,M B) were dissolved in the mixture used for imaging at 10 mm concentration. Other photosensitizers (PD, TPP) were incorporated by mixing them with lipids in chloroform before electroformation at 1000:1 lipid-to-dye ratio. BODIPY-C 10 was either included in the mixture with lipids at 300:1 lipid-to-dye ratio or 10 mLs olution of BODIPY-C 10 in methanol (50 mm)w as added into 400 mLo ft he mixture used for imaging. The fluorescence lifetime of BODIPY-C 10 matched that previously reported for DOPC bilayers, independent of the preparation method. [40] Acquisition of fluorescence lifetime images FLIM images were recorded using the Leica SPII confocal laserscanning microscope together with aC oherent Chameleon Vision II mode-locked femtosecond Ti :sapphire laser and aB ecker &H ickl SPC-830 (time-correlated single-photon counting) TCSPC card. The pulse length and pulsing frequency was 140 fs and 80 MHz. The output wavelength was tuneable between 680 and 1080 nm. The required laser wavelength was obtained by frequency doubling the output of the Ti :sapphire laser with second harmonic generation crystal (SHG, Harmonic, Coherent). BODIPY-C 10 was excited at 480 nm. Photosensitizer excitation wavelengths were 453 nm for PD, 420 nm for TPP and TPPS 4À ,a nd 633 nm for MB. An internal HeNe laser was used for 633 nm excitation. The BODIPY-C 10 lifetime for FLIM was detected between 510 and 600 nm using aphotomultiplier tube (PMC-100-1, Hamamatsu) with ax 63 water immersion objective, the confocal pinhole was half-open at 300 mm, which is equivalent to 2.7 Airy units. The optimal pinhole size of 1Airy unit for maximizing axial resolution was not used because it results in smaller fluorescence collection efficiencya nd the high axial resolution was not required for imaging vesicles larger than 10 mmi nd iameter.T he irradiation of photosensitizers in part of the image to achieve PDT was performed by zooming in 4-6 times and by continuous scanning across the zoomed area for al ength of time indicated. FLIM images were acquired at 256 256 pixel resolution using 256 time bins. The scanning frequency was 400 Hz. The instrument response function (IRF) was obtained by recording the scattering curve from aglass coverslip.

Data analysis
FLIM images were fitted and analyzed using the FLIMfit software tool developed at Imperial College London (v4.6.1). [54] 55p ixel binning was needed in order to get > 100 counts at the peak of each decay.D ata processing and analysis was done in MATLAB R2012a and OriginPro 8.6. The lifetime values were converted to viscosities using the calibration curve recorded earlier. [17]