Photosensitizing properties and subcellular localisation of 3,4-dihydro-β-carbolines harmaline and harmalol

Harmaline (1) and harmalol (2) represent two 3,4-dihydro-β-carboline (DHβCs) most frequently reported in a vast number of living systems. Fundamental aspects including the photosensitizing properties, cellular uptake, as well as the cyto- and phototoxicity of 1 and 2 were investigated herein. The molecular basis underlying the investigated processes are elucidated. Data reveal that both alkaloids show a distinctive pattern of extracellular DNA photodamage. Compound 1 induces a DNA photodamage profile dominated by oxidised purines and sites of base loss (AP sites), whereas 2 mostly induces single-strand breaks (SSBs) in addition to a small extent of purine oxidative damage. In both cases, DNA oxidative damage would occur through type I mechanism. In addition, a concerted hydrolytic attack is suggested as an extra mechanism accounting for the SSBs formation photoinduced by 2. Subcellular internalisation, cyto- and phototoxicity of 1 and 2 and the corresponding full-aromatic derivatives harmine (3) and harmol (4) also showed quite distinctive patterns in a structure-dependent manner. These results are discussed in the framework of the potential biological, biomedical and/or pharmacological roles reported for these alkaloids. The subtle structural difference (i.e., the exchange of a methoxy group for a hydroxyl substituent at C(7)) between harmaline and harmalol, gives rise to distinctive photosensitizing and subcellular localisation patterns.


Introduction
3,4-dihydro-β-carbolines (DHβCs) comprise a set of natural products present in vast range of living organisms, including mammals. [1][2][3][4] Biosynthesis of DHβCs involves the Pictet-Spengler reaction catalysed by strictosidine synthases, [5] and followed by different enzymatic and non-enzymatic reactions to yield the extensive set of DHβCs and full-aromatic derivatives (βCs) found in nature. [6] Harmaline (1) and harmalol (2) (Scheme 1) are two of the most representative DHβCs showing a broad spectrum of biological and pharmacological activity: (1) DHβCs 1 and 2 showed antitumor and antimicrobial activity against a broad spectrum of cell lines and species. [7][8][9][10][11][12] (2) Due to their stimulatory effect on the cellular melanin biosynthesis as well as on the tyrosinase activity, these two DHβCs were suggested to be useful drugs for the treatment of hypopigmentation-related skin disorders such as vitiligo. [13] However, the lack of pigmentation might enhance the sunlight exposition of 1 and 2, triggering photosensitized processes and causing unwanted side effects. Thus, the photosensitizing properties of photoexcited DHβCs need to be investigated. In addition, both alkaloids would prevent dopamine-induced mitochondrial damage and PC12 cell death. Both alkaloids would exert their effects at the mitochondrial domain through a scavenging action on reactive oxygen species and inhibition of monoamine oxidase and thiol oxidation. [14] It is noteworthy that the latter in vitro study was performed with brain homogenates and/or with isolated mitochondria. In this context, identifying the actual subcellular localisation of 1 and 2 becomes relevant to shed light on the mechanism of action suggested.

UV-Vis spectrophotometric analysis
The interaction of 1 and 2 with calf thymus DNA (ctDNA) was studied by UV-Vis absorption spectroscopy according to the methodology described elsewhere. [28] Briefly, spectra were recorded on a Perkin Elmer lambda 25 spectrophotometer. Measurements were made in quartz cells of 1 cm optical-path length, at room temperature. The concentrations of βC phosphate buffer solutions were 25 μM at pH 7.4. ctDNA concentration varied from 0 to 610 μM. Assuming a 1:1 stoichiometry for complexes, the association constants (K G ) was estimated using the following equation where ε βC·B and ε βC are the molar absorption coefficients of βC·ctDNA complex (βC·B) and βC, respectively, at the titration wavelength. ΔA is the change of absorbance, at ctDNA concentration ([B]), relative to the completely free βC ([βC] = 0 M) at this wavelength. The corresponding K G values are obtained from the ratio of the slope and intercept of Eq. 1.

Fluorescence measurements
Steady-state fluorescence emission was measured using a Fluoromax4 (HORIBA Jobin Yvon). Corrected fluorescence emission spectra were recorded in a 1 cm × 1 cm path lengths quartz cell at room temperature. For determining the quenching of fluorescence of βCs by ctDNA, emission spectra of air-equilibrated βC solutions (25 μM, in phosphate buffer at pH 7.4) were recorded in the absence and in the presence of ctDNA (from 0 to 710 Mbp). To avoid interferences due to changes in the absorbance of the samples because of the βC·ctDNA complex formation, spectra were excited at the corresponding isosbestic points ( Table 1). The fluorescence intensity (I F ) was obtained by integration of the corrected fluorescence spectra over the entire emission profile. (1)

Quantitative analysis of endonuclease-sensitive modifications
The use of DNA (supercoiled extracellular DNA of bacteriophage PM2) relaxation assay, in combination with incision repair endonucleases, to characterise the DNA photosensitized damage has been previously described. [21] Briefly, this assay makes use of the fact that supercoiled PM2 DNA is converted by either a SSB or the incision of a repair endonuclease into a relaxed (nicked) form, which migrates separately from the supercoiled form in agarose gel electrophoresis. [29] This methods allows the evaluation of several types of putative photosensitizing mechanisms [22,29]. In this work, four different DNA repair enzymes were used: (1) formamidopyrimidine-DNA repair glycosylase or Fpg to detect oxidatively damaged purines and AP sites, (2) Endo III to recognize oxidised pyrimidines as well as AP sites, (3) T4 endo V to recognize cyclobutane pyrimidine dimers (CPDs) and several specific types of AP sites, and (4) Endo IV to sense all types of AP sites. The number of cyclobutane pyrimidine dimers (CPDs calc ) were calculated as the difference between the number of Table 1 Benesi-Hildebrand binding constants (K G ) and Stern-Volmer constants (K SS ) for static quenching of the fluorescence of compounds 1 and 2 and ctDNA a Data obtained from UV-Vis spectroscopy analysis (Benesi-Hildebrand correlation) b λ iso is the wavelength of the isosbestic points in UV-vis spectra (nm) c Stern-Volmer constants obtained from steady-state measurements. Samples were excited at the low-energy isosbestic point (λ exc = λ iso )  Fig. 1a and dark controls (mixtures of PM2 and βCs, at the highest concentrations, kept in the dark).

Fluorescence Imaging Microscopy
FLIM experiments were carried out on a MicroTime 200 instrument (PicoQuant), using an LDH-375 (PicoQuant) pulsed laser as an excitation source with a repetition rate of 20 MHz. The laser was directed into the sample using a dichroic mirror (F33-375RDC, AHF/Chroma) and an oil immersion objective (1.4 NA, 100 ×) of an inverted microscope (IX-71, Olympus). The emitted fluorescence was filtered by a cut-off filter (F76-405LP, AHF/Chroma) and focused onto a 75 μm pinhole. The detection filter was a FB450-40 bandpass filter (Thorlabs), and the fluorescence was detected using a single-photon avalanche diode (SPCM-AQR 14, PerkinElmer). Photon counting, imaging reconstruction and data acquisition were performed with a TimeHarp 200 TCSPC module (PicoQuant). Raster scanned images were recorded with a 512 × 512 pixels resolution. The fluorescence images were analysed using the image processing package Fiji [30]. For live-cell experiment samples were prepared as follow: HeLa and HEK293 cells grown on coverslips (in DMEM) were washed (2 times) with phosphate buffered saline medium (PBS), then incubated with the corresponding βC in PBS solution (extracellular concentration = 10-12 μM) at 37 °C. Colocalisation studies were performed in an Abberior scanning microscope (Abberior Instruments GmbH, Germany) equipped with 375, 560, and 640 nm source lasers. Excitations were sequentially switched between 375 and 560 nm (or 640 nm) to excite the βC and the biomarkers, respectively. Emission signals were recorded in two different emission ranges: 495-520 nm, and 580-630 nm (or 640-750 nm), to collect the βC (cyan channel) and the biomarkers (red channel) fluorescent emission, respectively. The objective used was an UPlanSApo 100 × /1.40 oil immersion. In all the measurements, the pinhole size was set to 1 AU (Airy unit). For samples prepared, HeLa cells in DMEM free of FBS were first incubated with Mitotracker®, Lysotracker® and ER-Tracker (16 nM) in three different sets of experiments. After 20 min incubation, βC (10 μM) was added and co-incubated during 2 h. Cells were then prepared for confocal microscope analysis.

Cytotoxicity effect of 1 to 4 on A549 cells
The viability of host cells after treatment with the investigated βCs was evaluated using the colorimetric MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) assay as it was described elsewhere. [15] Briefly, A549 cells were trypsinized, counted and seeded into 96-well culture plates at a density of 15,000-20,000 cells/ Absorbance @365nm / a.u. well and grown in supplemented DMEM without phenol-red to confluence for 24 h. Cells were then exposed to several dilutions of the assayed compounds. Control cells (negative control) were treated with the same volume of dimethyl sulfoxide (DMSO) as vehicle alone (final DMSO concentration was 0.5%). To avoid effects of evaporation, outer wells were not used. After 24 h or 48 h of incubation with the βC, medium was removed, cells were washed with phosphate buffer saline (PBS) and treated with 0.5 mg/mL (100 μL/ well) of the tetrazolium dye MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, Sigma-Aldrich) in PBS, during 3 h, at 37 °C and 5% CO 2 . Supernatants were removed, and formazan crystals were solubilised in DMSO (100 μL/well). Absorbance at 540 nm (reference wavelength: 700 nm) was read in a plate reader (Synergy H1 Hybrid Reader, Biotek). Average absorbance from background wells (without cells, treated as the samples), was subtracted out, to obtain corrected values (A 540 corr ). The percentage of cell viability was calculated as: CC 50 values were obtained by non-linear regression analysis of Cell Viability (%) vs. Log [βC] with a variable Hill's slope (normalised response), using GraphPad Prism 5.03 software. Each concentration was assayed in triplicates, in five independent sets of experiments.

Phototoxicity of 1 and 2 on A549 cells.
A549 cells were counted and seeded into 96-well culture plates at a density of 20,000 cells/well and grown to confluence for 24 h following the procedure described above. Cells were incubated with DHβC in DMEM (without phenol-red) for 1.5 h. After incubation, cells were irradiated (Rayonet lamp @420 ± 20 nm, 1.5 J cm -2 ) for 10 min and then were incubated during 24 h prior to the determination of cell viability by MTT assay as it was described above.

DNA damage induced by photoexcited DHβCs under cell-free conditions
The photosensitizing properties of 1 and 2 (Scheme 1) were explored herein using supercoiled extracellular DNA of bacteriophage PM2, in combination with a set of four incision repair endonucleases. The dose dependence of the DNA damage was investigated first ( of single-strand breaks (SSBs) and modifications sensitive to Fpg were subsequently quantified. Interestingly, both compounds showed a linear dose-dependence trend with a distinctive behaviour. Compound 1 showed the typical dose dependence already reported for other related full-aromatic βCs, including 3. [19-22, 28, 31-35] Briefly, Fpg-sensitive sites (i.e., oxidatively modified purines including 8-oxo-7,8-dihydroguanine and formamidopyrimidines, as well as sites of base loss or AP sites) were found to be produced with higher efficiency than SSBs. On the contrary, 2 showed the opposite trend. In all the cases, DNA damage observed is mediated by the photoexcited βCs, since no damage was observed in both light and dark controls.
Minor chemical modifications of the molecular structure of the βC-ring can dramatically modify its photodamage pattern and, as such, the DNA damage profile [16,19,22]. Therefore, the number of four types of repair enzymes-sensitive modifications (i.e., DNA damage profile) photoinduced by 1 and 2 were additionally determined (Fig. 1b). Data evince a strong qualitative and quantitative structure-activity relationships: (1) Both 1 and 2 induce specific DNA base damage. In the case of 1, the damage profile observed is characterised by a relatively high number of oxidatively damaged purines and AP sites (i.e., Fpg-sensitive base modifications). This is consistent with the fact that DHβCs bind preferentially with dG·dC-rich DNA regions. [36,37] AP sites, which are specifically recognized by endonuclease IV, and SSBs are generated in lower amounts. On the contrary, SSBs represent the main types of DNA damage when the genetic material is irradiated in the presence of 2.
(2) It is generally accepted that Fpg-sensitive sites are produced either by type I (via electron transfer) or type II (via reactive oxygen species such as singlet oxygen, 1 O 2 ) photosensitization mechanisms [29]. Taking into account the extremely low or null efficiency of singlet oxygen production (Φ Δ ~ 0.02) showed by both 1 and 2, [23] type I would be the operative mechanism in the Fpg-sensitive sites photosensitized formation. The extended oxidative damage depicted by non-negligible amounts of Endo IV sensitive sites photoinduced by both DHβCs further support this hypothesis. (4) SSBs can be formed from a cumulative damage on the DNA moiety. However, this does not seem to be the case for compound 2, where a relatively large amount of SSBs is observed since the very early stage of the reaction (Fig. 1b). In a recent work, a concerted photoinduced hydrolytic attack of the DNA was suggested as the operative mechanism responsible for SSBs formation photoinduced by a related βC derivative (6-chloroharmine). [22] This could also be the case for 2. Note that the acidity of both -N(9)H and -OH group placed at C(7) of 2 are enhanced when subject to photoexcitation [23]. This is more evident when analysing the electronic density over the molecule depicted in the molecular orbital diagrams reported by Villarruel et al. [23] and from the emission spectra depicted in Figure Table 1). Insets: normalised emission spectra are unknown, the large changes in the relative energy of their electronic states (i.e., bathochromic shift ~ 30 nm) [23] with respect to the full-aromatic derivatives 3 and 4, respectively, support this conclusion.

Interaction between 2 and 4 with calf thymus DNA
The interaction between 1 and 2 with genetic material, including calf thymus DNA (ctDNA), has been previously described. [16,36,37,[39][40][41][42][43] Due to the strong solvent dependence of the photophysical and photochemical properties of βCs, a comparative UV-Vis absorption and fluorescence spectroscopic analysis of the interaction between 1 and 2 with ctDNA in phosphate buffered solutions at pH 7.4 (i.e., under the same experimental conditions used in the PM2 experiments) was performed herein. UV-Vis spectra of 1 and 2 were recorded in the presence of increasing amounts of ctDNA (Fig. 2). The spectroscopic pattern observed was the same to that previously reported by Duportail et al. [39] and Sarkar et al. [43] for 1 and 2, respectively. Briefly, in the presence of ctDNA, the absorption spectra of both compounds show a large hypochromism and bathochromic shift, accompanied by the appearance of isosbestic points (see λ iso in Table 1). The corresponding association constants (K G ) between βCs and ctDNA were obtained from Eq. 1 (Table 1). Representative plots depicted as insets in Fig. 2 show a linear dependence with the DNA concentration. The comparison of K G values confirms results previously reported suggesting that, in the electronic ground state, the interaction of 1 with ctDNA is slightly stronger than that observed for 2 (K G 1 > K G 2 ). [39] When comparing the binding constants reported herein, it appears very clearly that DHβCs show lower affinity to ctDNA than the corresponding full-aromatic derivatives 3 and 4. [19,37] Thus, the hydrogenation of a double bond at C(3)-C(4) of the pyridine ring in the DHβCs greatly reduces the extent of interactions.
Steady-state fluorescence spectroscopy was used to further analyse the interaction between βCs (in both the electronic ground (S 0 ) and excited states (S 1 )) and ctDNA. Emission spectra of 1 and 2 were recorded in the presence of increasing amounts of ctDNA (Fig. 2c, d). The increase in the ctDNA concentration causes a decrease in the fluorescence intensity of both alkaloids. Interestingly, in the case of 2, the shape of the emission spectra changes with the increase of the ctDNA concentration. This is more evident when comparing the normalised emission spectra shown as inset in Fig. 2d. Briefly, changes in the relative intensity of the bands ascribed to the cationic (C* ~ 480 nm) and zwitterionic (Z(II)* ~ 540 nm) species ( Figure SI.3) [23] are evident. Thus, the presence of proton acceptor/donor groups (e.g., DNA nucleobases) affect the equilibria among the different acid-base and/or prototropic species of 2 changing their steady-state relative concentrations. Such a proton exchange might induce shifts in the local pH that can lead to the hydrolysis of DNA phosphodiester-bounds with the concomitant SSBs formation. The latter hypothesis deserves to be further investigated with additional spectroscopic and analytical techniques.
In both cases, a linear trend was observed in the Stern-Volmer plots (not shown) and the corresponding slope values (K SS = τ F k q F ) obtained are listed in Table 1. Bearing in mind the βCs fluorescence lifetime (~ 5-6 ns) [23] values of ~ 10 11 M -1 s -1 can be calculated for the bimolecular quenching constant (k q F ). This fact suggests a large contribution of a static quenching between βCs and ctDNA. When comparing the magnitude of K SS values, it is evident that 1 shows a stronger interaction than 2 (K SS ). This trend agrees with that one observed from the UV-Visible spectroscopic titration. It is worth mentioning that despite both techniques show values that are in the same order of magnitude for each compound, values are slightly different. This could be accounted for the fact that in fluorescence experiments other processes can be detected in addition to the formation of a ground state complex. Another key aspect that might also explain the difference is the fact that both compounds are producing fluorescent βC·ctDNA complexes. [39] This is clear when analysing the evolution of the absorbance and emission intensity as a function of the ctDNA concentration that a plateau is reached at the highest ctDNA concentration (Figure SI.4).
Regarding the mode of action, circular dichroism studies suggest that intercalation of 1 into the stacked base pairs of double-stranded DNA is unlikely. The lack of full planarity in the DHβC moiety would be the reason for this behaviour. [39] On the contrary, the molecule would probably adopt a tilted position in one of the DNA grooves. [40] Another key factor that might also be responsible of a non-intercalative mode of action is the fact that, under these experimental conditions (pH 7.4), the cationic (protonated) species of 1 and 2 are present in the solution. Thus, as it was previously reported, cationic βCs are typically bound parallelly to the DNA base pairs, by partial insertion through the double helix groove, [41] placing the N(9) atom in a polar-protic environment towards the negatively charged DNA backbone. [16]

Cellular uptake of βCs
The cellular internalisation of a particular group of fullaromatic βCs was recently investigated by FLIM microscopy, [16] showing that mild changes in the βC ring (i.e., N-methylation) modulate both the dynamics of the drug uptake (and release) by HeLa and HEK293 cells as well as the intracellular localisation. Having this in mind, the cellular uptake of the two DHβCs 1 and 2 as well as their respective full-aromatic derivatives 3 and 4 was investigated herein (Figs. 3, SI.5 and SI.6).
When co-incubated with βCs in PBS, cells were able to take up all four investigated compounds within 25-40 min depending on the chemical structure ( Figure SI.5). In the case of both 3,4-dihydro-derivatives (1 and 2), the maximum emission intensity (plateau) is reached 10-15 min later than their respective full-aromatic derivatives (3 and 4). This fact would be compatible with a passive uptake mechanisms where a reduced cellular membrane permeability is expected for polar cationic species [16]. Note that 1 and 2 show pK a values up to 2-times larger than 3 and 4, respectively. [18,23] Thus, under physiological pH conditions protonated (cationic) species of 1 and 2 are present in more than 99.9%, whereas ~ 50-60% of 3 and 4 are present in the neutral form. Additional experiments are needed to further address the actual uptake mechanism.
Interestingly, two distinctive image patterns were observed (left column in Figs. 3 and SI.6). The four compounds investigated stain the same intracellular structures, including additional stain of intranuclear structures (see below). This is more evident for both compounds bearing a hydroxyl group at C(7) (i. e., 2 and 4). The second column in Figs. 3 and SI.6 shows the lifetime images from the entire cells. The corresponding histograms showing the changes in intensity and lifetime along the arrow traced in the first and second columns are depicted in the third column of Figs. 3 and SI.6. At a first glance, the four compounds show quite a similar histogram pattern. I.e., the regions with the highest fluorescence intensity show the lowest fluorescence lifetimes (τ F ). This agrees with the fact that fluorescence properties of βCs 1-4 strongly depend on the environment [6,18,23,44]. Briefly, in polar-protic solvents, photoexcited cationic species (C*) of βCs is the dominant emitting species, showing τ F values of ~ 5 ns (e. g., in water τ F = 5.9 ns and 4.7 ns for compounds 1 and 2, respectively). On the contrary, in nonprotic solvents, photoexcited neutral species (N*) of βCs is the main emitting species showing slightly larger τ F values (e. g., in ACN or DMSO τ F of ~ 7 ns [44] and 6.0 ns for compounds 1 and 2, respectively). In this context, data depicted in the histograms (Fig. 3) would be compatible with a model where βCs alkaloids would be sensing a polar-protic environment (aqueous media in the interior of the organelles) and, to a lower extent, a non-protic environment (cellular membranes). The fact that the fluorescence quantum yield (Φ F ) of DHβC in aqueous environment is lower than in organic solvents [44] support the hypothesis that the highest fluorescent intensity regions depicted in the histograms represent the subcellular domain where the alkaloids accumulate in greater proportion (see below). Intracellular localisation of 1 to 4 was further characterised by using three different red fluorescent biomarkers: Lysotracker®, Mitotracker® and ER-Tracker, that selectively stain mitochondria, lysosome, and endoplasmic reticulum, respectively (Figs. 4, SI.7 and SI.8). The investigated compounds showed two distinctive localisation patterns depending on the chemical nature of the Fig. 3 Fluorescence images of HeLa cells incubated with a compound 1 and b compound 2. Column 1: fluorescence intensity images of βCs (λ exc = laser at 375 nm, λ em 495-520 nm). Column 2: fluo-rescence lifetime images of dye in the whole cell. Column 3: plot of intensities and average lifetimes along the selected arrows (histograms). Scale bars represent 10 µm substituent placed at C(7) (i.e., -OCH 3 or -OH). This is better represented by the white pixels in the merged images, and it is also accounted for by the Pearson's coefficients depicted in Fig. 4c. Briefly, 1 and 3 showed quite a large localisation in the lysosomes (R ~ 0.5-0.7) and endoplasmic reticulum (R ~ 0.4) and, to some extent, a minor colocalisation in the mitochondria (R ~ 0.15). On the contrary, 2 and 4 showed quite a homogenous distribution among the three investigated subcellular organelles (i.e., R ~ 0.3, ~ 0.2 and ~ 0.2 for lysosomes, endoplasmic reticulum, and mitochondria, respectively). Interestingly, R values for Lysotracker® and ER-Tracker observed for 2 and 4 were slightly lower than the corresponding values observed for 1 and 3. This fact can be rationalised in terms of the qualitative information depicted in Fig. 3 showing that, although the four investigated βCs are able to stain other subcellular domains (such as intranuclear structures), 2 and 4 showed the highest signal.

Cytotoxicity of 1-4 on eukaryotic cells
The toxicity of these four alkaloids have been largely studied. [15,36,[45][46][47][48] However, these studies were performed under different experimental conditions (incubation times, % of foetal bovine serum (FBS) and/or % of the vehicle DMSO, etc.), making a direct comparison difficult. All these facts might account for the large variability on the CC 50 values reported. In this work, the effect of 1 to 4 on the A549 cells' viability was evaluated. Data obtained from dose-response curves (Fig. 5a, b) revealed a strong qualitative and quantitative structure-activity relationship ( Table 2).
(1) Both DHβCs (1 and 2) showed lower toxicity on eukaryotic cell culture than their corresponding full-aromatic derivatives 3 and 4 (entry 1). Interestingly, a similar trend has been described for 1 and 3 (entries 5-10) [10,45] and for 2 and 4 (entries [11][12][13] [47] on different cell lines including A549. (2) Among the DHβCs, the presence of the methoxy group placed at C(7) in the βC ring of 1 might enhance the cytotoxicity of the alkaloid (entries 1 and 3). This could be related to the fact that 1 showed a stronger interaction with biomolecules than 2 and/or to the fact that both alkaloids show mild differences in the subcellular internalisation patterns.  Fig. 4 Representative fluorescence images of HeLa cells co-incubated with: a compound 1 and b compound 2 with three different biomarkers (Lysotracker®, Mitotracker® and ER-Tracker). White pixels in the merge from the biomarker (first column in red) and compound (second column in cyan) images depict the colocalisation extent (third column). c Pearson's coefficients calculated for the four alkaloids with the three biomarkers investigated. Scale bars represent 10 µm the apparent cytotoxicity of the alkaloids investigated. The higher the %FBS, the lower the cytotoxicity (i.e., the higher the CC50 values observed) (entries 1-3). A similar trend was reported by Ghosh et al. [49] The latter fact suggests that βCs might strongly interact with the proteins present in FBS and therefore a lower βC' effective concentration is observed. (5) With the exception of HepG2 (entries 19 and 22) and HBL-100 (entry 7), cytotoxicity of βCs against different cell lines, including A549 and HeLa, is quite similar. This is reflected in the similarity of the CC 50 values reported for each experimental condition.

Phototoxicity of 1-4 on eukaryotic cells.
Under in vitro conditions, 1 and 2 showed quite an important photosensitizing activity (Fig. 1) as well as intracellular localisation on eukaryotic cells (Fig. 3) in a dose-and structure-dependent manner. To further evaluate the photodynamic activity on cell culture, A549 cells treated pre-incubated with the DHβCs were irradiated (420 ± 20 nm) during 10 min. Under the experimental conditions explored herein, only 1 showed a photodynamic effect (Fig. 5c). This fact can be explained, on one hand, in the framework of the differential mechanisms of photodamage described above for these two DHβCs. A larger oxidative damage is photoinduced by 1 comparing with 2 that, upon photoexcitation, shows an additional and dominant non-oxidative deactivation pathway (i.e., proton exchanges). Considering genetic material as a potential intracellular target, the purine bases oxidation induced by 1 can lead to mutations and hence to cell death, whereas SSBs induced by photoexcited 2 represents a type of damage that can be promptly and efficiently reverted by living cells. In addition, photoexcited 1 and 2 can also induce distinctive lysosomes, ER membranes and/or mitochondrial photodamage profiles and/or organelle dysfunction that might trigger differential photodynamic responses. [50] In this context, images of cells acquired 3 h after irradiation in the presence of 1 show a cytoplasmic vacuolisation associated with a paraptotic-like response ( Figure SI.9). [51] Despite this, further studies are needed to clearly assess the mechanism of photodynamic action of these alkaloids. The potential use of 1 and 2 to treat hypopigmentation disorders (e. g., vitiligo) was suggested by Park et al. [13] The results obtained herein, showing a very low cytotoxicity (CC 50 > 600 μM, Fig. 5b) and low or null intracellular    (Fig. 5c), indicate that 2 would be the best candidate for this particular pharmacological application to avoid potential photoinduced side effects.

Conclusions
This work reports a comparative study of the photosensitizing properties of two 3,4-dihydro-β-carboline alkaloids (DHβCs), harmaline (1) and harmalol (2), under extracellular and intracellular domain. In addition, the cyto-and phototoxic effects of these alkaloids as well as their subcellular uptake and localisation was also investigated. For comparative reasons, results obtained for the two full-aromatic related βC harmine (3) and harmol (4) were included. Data reported herein show that a subtle structural change (i.e.,-OCH 3 or-OH substituent at C(7) in the βC ring gives rise to quite distinctive action patterns: (1) Regarding the extracellular DNA photodamage, 1 mainly induces oxidation at the purines through type I mechanism. The presence of a methoxy group at C(7) enhances the photosensitizing properties of 1 either by increasing its binding affinity with ctDNA and/or by increasing its oxidation potential. This agrees with the fact that 1 showed a selective damage profile where oxidised purines represent the main type of DNA photoproducts formed. In the case of 2, single-strand breaks (SSBs) in addition to DNA oxidative damage are mostly induced. In this case, a concerted hydrolytic attack (due to a strong pH change in the surroundings of the DNA chain triggered by photoexcited 2) is suggested as an additional mechanism accounting for the SSBs formation. (2) Both DHβCs as well as the full-aromatic derivatives 3 and 4 show quite a strong intracellular accumulation into different compartments of HeLa cells including lysosomes, endoplasmic reticulum and to a lower extent mitochondria. Interestingly, nuclear accumulation of 1-4 was also observed, being the later fact more evident for those βCs bearing a hydroxyl group (i.e., 2 and 4). Interestingly, the four βCs investigated herein showed a distinctive internalisation pattern with respect to the quaternary (cationic) 2,9-dimethyl-βC reported by Denofrio et al. [16] that mainly accumulates into mitochondria. Targeting key cell's compartments (including the lysosomes) with photoactive molecules represent an emerging strategy for Photodynamic Therapy. For all the compounds, lifetime changes depending on the subcellular domain. This can have particular ramifications in the development of biotechnological tools based on small molecules derived from βC alkaloids. [52] (3) The non-negligible mitochondrial colocalisation of 1 and 2 reported herein would support the hypothesis formulated by Lee et al. [14] suggesting that the neuroprotective role observed for these two alkaloids would occur at the submitochondrial domain. In addition, since monoamine oxidase enzymes (MAO) are located in the external membrane of mitochondria, DHβCs would still exert the reported enzyme inhibitory effect [16]. (4) Both DHβCs (1 and 2) showed lower cytotoxicity on eukaryotic cell culture than the corresponding fullaromatic derivatives 3 and 4. However, among the two DHβCs, 2 was less cytotoxic than 1. (5) Upon photoexcitation, 1 showed a strong photodynamic effect, whereas 2 did not show any effect under the experimental conditions explored herein. This fact could be a consequence of the slightly different internalisation pattern shown by both DHβCs. In addition to this, the distinctive photosensitized DNA damage profiles and photooxidative extent shown by these two alkaloids (i.e., purine oxidation and AP sites vs. singlestrand breaks in the case of 1 and 2, respectively) would also lead to a distinctive photoinduced oxidative damage (either on intracellular genetic material or other biologically relevant targets), triggering distinct biochemical processes. [50] To elucidate this, intracellular (nuclear and/or mitochondrial) DNA damage needs to be further investigated. (6) The understanding of the photosensitizing properties of DHβCs in a biological environment lies not only on their biological relevance, but also in the fact that these alkaloids have been suggested as potential therapeutic drugs. In particular, the potential use of 1 and 2 in hypopigmentation skin disorders was suggested. [13] In this context, the rather low or null efficiency of reactive oxygen production (i.e., singlet oxygen, Φ Δ ~ 0.02) reported for compound 2, [23] in addition to its very low cytotoxicity (CC 50 > 600 μM) as well as the lack of intracellular photodynamic activity reported herein, indicate that 2 (harmalol) would be the best candidate for this particular pharmacological application.