Synthesis, photodynamic activity, and quantitative structure-activity relationship modelling of a series of BODIPYs

Article abstract of DOI:10.1016/j.jphotobiol.2017.01.012

Here we report the synthesis of eleven new BODIPYs (14–24) characterized by the presence of an aromatic ring on the 8 (meso) position and of iodine atoms on the pyrrolic 2,6 positions. These molecules, together with twelve BODIPYs already reported by us (1???12), represent a large panel of BODIPYs showing different atoms or groups as substituent of the aromatic moiety. Two physico-chemical features (¹O₂ generation rate and lipophilicity), which can play a fundamental role in the outcome as photosensitizers, have been studied. The in vitro photo-induced cell-killing efficacy of 23 PSs was studied on the SKOV3 cell line treating the cells for 24?h in the dark then irradiating for 2?h with a green LED device (fluence 25.2?J/cm²). The cell-killing efficacy was assessed with the MTT test and compared with that one of meso un-substituted compound (13). In order to understand the possible effect of the substituents, a predictive quantitative structure-activity relationship (QSAR) regression model, based on theoretical holistic molecular descriptors, was developed. The results clearly indicate that the presence of an aromatic ring is fundamental for an excellent photodynamic response, whereas the electronic effects and the position of the substituents on the aromatic ring do not influence the photodynamic efficacy.

Full text of DOI:10.1016/j.jphotobiol.2017.01.012

Accepted Manuscript
Synthesis, photodynamic activity, and quantitative structure-
activity relationship modelling of a series of BODIPYs
Enrico Caruso, Marzia Gariboldi, Alessandro Sangion, Paola
Gramatica, Stefano Banfi
PII:
DOI:
Reference:
S1011-1344(16)30985-X
doi: 10.1016/j.jphotobiol.2017.01.012
JPB 10713
To appear in:
Journal of Photochemistry & Photobiology, B: Biology
Received date:
Accepted date:
2 November 2016
9 January 2017
Please cite this article as: Enrico Caruso, Marzia Gariboldi, Alessandro Sangion, Paola
Gramatica, Stefano Banfi , Synthesis, photodynamic activity, and quantitative structure-
activity relationship modelling of a series of BODIPYs. The address for the corresponding
author was captured as affiliation for all authors. Please check if appropriate. Jpb(2017),
doi: 10.1016/j.jphotobiol.2017.01.012
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.

ACCEPTED MANUSCRIPT
Synthesis, Photodynamic activity, and Quantitative
Structure-Activity Relationship Modeling of a
series of BODIPYs
Enrico Caruso*1, Marzia Gariboldi1, Alessandro Sangion2, Paola Gramatica2, Stefano Banfi1
1 Department of Biotechnology and Life Sciences (DBSV). University of Insubria, Via J.H. Dunant 3, 21100
Varese (VA), Italy.
2 Department of Theoretical and Applied Sciences (DiSTA), University of Insubria, Via J.H. Dunant 3,
21100 Varese (VA), Italy.
*To whom correspondence should be addressed. Phone: +39-0332-421550. Fax: +39-0332-421554. E-mail:
enrico.caruso@uninsubria.it

ACCEPTED MANUSCRIPT
Abstract
Here we report the synthesis of eleven new BODIPYs (14 – 24) characterized by the presence of an
aromatic ring on the 8 (meso) position and of iodine atoms on the pyrrolic 2,6 positions. These
molecules, together with twelve BODIPYs already reported by us (1 – 12), represent a large panel
of BODIPYs showing different atoms or groups as substituent of the aromatic moiety. Two
physico-chemical features (1O2 generation rate and lipophilicity), which can play a fundamental role
in the outcome as photosensitizers, have been studied. The in vitro photo-induced cell-killing
efficacy of 23 PSs was studied on the SKOV3 cell line treating the cells for 24 h in the dark then
irradiating for 2 h with a green LED device (fluence 25.2 J/cm2). The cell-killing efficacy was
assessed with the MTT test and compared with that one of meso un-substituted compound (13). In
order to understand the possible effect of the substituents, a predictive quantitative structure-activity
relationship (QSAR) regression model, based on theoretical holistic molecular descriptors, was
developed. The results clearly indicate that the presence of an aromatic ring is fundamental for an
excellent photodynamic response, whereas the electronic effects and the position of the substituents
on the aromatic ring do not influence the photodynamic efficacy.
Keywords: Photodynamic Therapy, BODIPY, tumor cells, singlet oxygen, QSAR

ACCEPTED MANUSCRIPT
Introduction
Photodynamic therapy (PDT) is a light-based therapeutic approach used for the treatment of various
diseases, including precancerous skin pathologies and localized solid tumors. PDT involves the
administration of a photosensitizer (PS), topically applied directly on the diseased tissue or
intravenously administered. After the appropriate time of incubation, the effective step of PDT is
the area-selective irradiation with visible light to induce the electronic excitation of the PS.
In the presence of adequate concentrations of molecular oxygen in the tissue, the light activated PS
transfers its energy to molecular oxygen, leading to the generation of reactive oxygen species
(ROS) that are responsible for PDT-based cell-killing which could lead to tumor ablation [1,2].
Porfimer sodium (Photofrin®) a purified form of hematoporphyrin derivative consisting of a
mixture of porphyrins, was the first PS approved by the US Food and Drug Administration (FDA).
Despite the encouraging results in the treatment of a variety of cancers, including lung, esophageal,
stomach and bladder adenocarcinoma, porfimer sodium is not devoid of drawbacks and limits, such
as weak absorption in the red region of the visible spectrum, a poor selectivity between tumours and
normal tissues and a long lasting skin phototoxicity [3,4].
Most of the drug-development efforts aimed at overcoming these problems have resulted in novel
“second generation” PS with improved properties. Some second-generation PSs, such as verteporfin
(Visudine®), temoporfin (mTHPC, Foscan®), chlorophyll-a derivative (Radachlorin®), Pd-
bacteriopheophorbide-a derivative (Tookad®) and pyropheophorbide-a derivative (HPPH), belong
to the cyclic tetrapyrrolic class of molecules [5], exhibiting relatively high affinity for tumour
tissues [6], low intrinsic toxicity and intense absorption properties in the therapeutic window of the
visible electromagnetic spectrum (600 – 700 nm). On the other hand, the synthesis of the PSs from
chlorophyll is quite demanding because of the instability of the natural starting molecule.
An alternative class of non-porphyrin compounds, based on a difluoro-boradiazaindacenes skeleton,
also known as boron-dipyrromethenes (BODIPY), has emerged over the last decade. BODIPY
derivatives proved to be useful fluorophores in many applications, such as fluorescent labelling of

ACCEPTED MANUSCRIPT
biomolecules, material for energy transfer cassettes and fluorescent chemosensors for detection of
ionic species [7-11].
BODIPYs show many characteristics including: easy synthetic pathway, tuneable
hydrophilic/hydrophobic character [12-14], high extinction coefficients in the visible region
(relatively unaffected by environmental conditions) [15,16], partial resistance to photobleaching
[17] and high quantum efficiencies of fluorescence (fl) [18]. The last two features exert a negative
effect whenever BODIPYs are envisaged as PS; actually, while photodynamic activity may benefit
from a long half-life of the photoactivatable molecules, this property might be associated with
persistent skin photosensitization. Furthermore, although a high quantum yields of fluorescence is
fundamental when BODIPYs are exploited as biological probes, the light emission should be
limited when BODIPYs are engineered for PDT. To this purpose, fluorescence must be depressed,
thus avoiding the energy dispersion from the singlet excited PS, favouring the formation of a PS
triplet excited state from which occurs a process useful for the production of cell toxic reactive
oxygen species.
As far as it concerns the availability of these PSs, the straightforward “one pot” synthesis is a
practical advantage favouring the development of these molecules that can be isolated in fairly good
yields after a single chromatographic purification.
Here we report the synthesis of a panel of fully characterized BODIPY featuring an aromatic ring
on the meso (8) position, characterized by the presence of different substituents. These new
molecules, together with thirteen BODIPYs previously synthesized by us [19], were studied to
determine the singlet oxygen (1O2) production rates, the fluorescent quantum yield and, finally, the
relative degree of lipophilicity. The photodynamic activity of these compounds, all characterized by
intense absorbance in the 495 – 535 nm range, has been assessed on the ovarian carcinoma
(SKOV3) cell line, following irradiation with a low energy green LED light. Quantitative structure-
activity relationship (QSAR) analysis was also performed on the whole set of the BODIPYs here
reported.

ACCEPTED MANUSCRIPT
Experimental section
General.
1H and 13C NMR spectra were recorded on a Bruker 400 MHz spectrometer in CDCl3 or [d6]
DMSO; chemical shifts are expressed in ppm relative to chloroform (7.28) and are reported as s
(singlet), bs (broad singlet), d (doublet), t (triplet) and m (multiplet). Mass spectrometric
measurements were performed on a Bruker Esquire 3000 Plus. Elemental analyses were performed
on a ThermoQuest NA 2100, C, H, N analyzer, equipped with an electronic mass flow control and
thermal conductivity detector. Merck 60 F254 silica gel precoated sheets (0.2 mm thick) were used
for analytical thin-layer chromatography (TLC). Silica gel 60 (70-230 mesh, Merck) was used for
column chromatography. Aromatic aldehydes and 2,4-dimethylpyrrole were commercial products
(Sigma-Aldrich) and used as received. Dichloromethane used for BODIPY synthesis was distilled
from CaCl2 directly into the reaction flask. HPLC analyses were conducted with an Agilent 1100
Series instrument coupled with a diode array detector. The instrument was fitted with a 250×4.6 mm
column (Supelco, Ascentis) packed with C-18 reversed-phase particles (5 m), operating with an
isocratic elution of MeCN at 0.5 mL/min. The green LED array is composed of 12×3 W diodes
distributed on an 11 cm diameter disk, and equipped with a heat sinker. The emitted light is
characterized by a lambda max (λmax) at 525 nm (fluence rate 3.5 x 10-3 W/cm2, light energy density,
or fluence, 25.2 J/cm2) and a width at half maximum of 70 nm. The electric supply was ensured by a
50 W current transformer. This array was placed above the plate at such a distance as to produce a
homogeneous area of irradiation with a fluence rate (irradiance) of 7.52×10-5 W/cm2 at 525 nm ( of
maximum emission), as determined with a LI-COR 1800 spectroradiometer.
Synthesis.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(3’-iodo-4’,5’-dimethoxyphenyl)-4-bora-3a,4adiaza-s-
indacene (1); 4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(4’-methoxyphenyl)-4-bora-3a,4adiaza-
s-indacene
(2);
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(3’-methoxyphenyl)-4-bora-
(3); 4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(4’-hydroxy-3’,5’-
3a,4adiaza-s-indacene

ACCEPTED MANUSCRIPT
diiodophenyl)-4-bora-3a,4adiaza-s-indacene (4); 4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(3’-
hydroxy-4’-iodophenyl)-4-bora-3a,4adiaza-s-indacene
(5);
4,4-difluoro-2,6-diiodo-1,3,5,7-
tetramethyl-8-(4’-methoxynaphthalene)-4-bora-3a,4adiaza-s-indacene (6); 4,4-difluoro-2,6-diiodo-
1,3,5,7-tetramethyl-8-(4’,7’-dimethoxynaphthalene)-4-bora-3a,4adiaza-s-indacene (7); 4,4-difluoro-
2,6-diiodo-1,3,5,7-tetramethyl-8-(2’,6’-dichlorophenyl)-4-bora-3a,4adiaza-s-indacene
difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’,3’,4’,5’,6’-pentafluorophenyl)-4-bora-3a,4adiaza-s-
indacene (9); 4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’,3’,5’,6’-tetrafluoro-3’-[(4’’-
methoxybenzyl)sulfanyl]phenyl)-4-bora-3a,4adiaza-s-indacene (10); 4,4-difluoro-2,6-diiodo-1,3,5,7-
(8);
4,4-
tetramethyl-8-(4’-pyridyl)-4-bora-3a,4adiaza-s-indacene
tetramethyl-8-(3’-pyridyl)-4-bora-3a,4adiaza-s-indacene
(11);
(12);
4,4-difluoro-2,6-diiodo-1,3,5,7-
4,4-difluoro-2,6-diiodo-1,3,5,7-
tetramethyl-4-bora-3a,4adiaza-s-indacene (13) were synthesized as previously described [19].
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’-chloro-6’-nitrophenyl)-4-bora-3a,4adiaza-s-
indacene (14). To a 30 mL solution of 2-chloro-6-nitrobenzaldehyde (213 mg, 1.15 mmol) and 2,4-
dimethylpyrrole (260 µL, 2.53 mmol) in dry CH2Cl2, kept under N2 atmosphere, were added ten
drops of TFA. The solution was then stirred at RT for the time necessary to obtain the complete
consumption of the aldehyde (from 6 to 12 h) as determined by TLC analysis. Thereafter the
oxidation of the dipyrrolylmethane to dipyrrolylmethene was carried out, in the same flask, by the
addition of DDQ (391.6 mg, 1.725 mmol) and the mixture was stirred for 30 min. The last step of
.
the BODIPY synthesis requires the addition of Et3N (3 mL) and BF3 OEt2 (3 mL) thus producing
the desired borinated final compound. This step requires about 12 h, then the organic layer was
washed two times with water, two time with a HCl 1 M solution and then other three times with
water. The organic solution was dried over Na2SO4, filtered and evaporated to dryness. The raw
material was purified by column chromatography (SiO2 50 g, petroleum ether-CH2Cl2, 3:7)
affording 197 mg (0.49 mmol, yield: 42.6%) of 14a in the form of orange needles, mp = 190 °C.
1
UV-vis(CH CN): 513 nm (=63600); Φfluo (526 nm): 0.68; H NMR (CDCl3)  1.46 (s, 6H, 2×CH3),
3

ACCEPTED MANUSCRIPT
2.59 (s, 6H, 2×CH3), 6.03 (s, 2H), 7.66 (t, 1H, J = 12 Hz), 7.85 (d, 1H, J = 12 Hz), 8.03 (d, 1H, J =
12 Hz).
The corresponding 2,6-diiodo compound (14) was synthesised treating with I2 (231.5 mg, 0.912
mmol) and HIO3 (160.4 mg, 0.912 mmol) an EtOH solution of compound 14a (184 mg, 0.456
mmol). This mixture was stirred for 15–18 h at RT, then washed with water and extracted three
times with CH2Cl2; the organic phase was dried over Na2SO4 and evaporated to dryness. The crude
product was purified by column chromatography (SiO2 40 g, petroleum ether-CH2Cl2, 1:1)
affording 79 mg (0.12 mmol; yield: 26.3%, overall yield: 10%) of 14 as red needles, mp > 300 °C;
EA: calcd for C19H15BClF2I2N3O2 C, 34.82; H, 2.31; N, 6.41; found: C, 35.01; H, 2.32; N, 6.45;
1
UV-vis(CH CN): 548 nm (=49700); Φfluo (555 nm): 0.02; H NMR (CDCl3)  1.47 (s, 6H, 2×CH3),
3
2.69 (s, 6H, 2×CH3), 7.73 (t, 1H, J = 12 Hz), 7.89 (dd, 1H, J1 = 12 Hz, J2 = 4 Hz), 8.11 (dd, 1H); 13C
NMR (CDCl3)  15.84, 16.32, 86.03, 124.39, 128.99, 130.41, 131.78, 132.99, 135.00, 139.52,
144.15, 148.15, 157.98; MS (ESI): M-1+ found: 654.32 calcd for C19H15BClF2I2N3O2 655.41;
HPLC: retention time = 8’28’’ (100%); Photo-degradation t60: 55%, t120: 71%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’-methoxyphenyl)-4-bora-3ª,4adiaza-s-
indacene (15). 2,4-Dimethylpyrrole (400 µL, 3.89 mmol) and 2-methoxybenzaldehyde (241 mg,
1.77 mmol) were reacted as described for compound 14a. The raw material was purified by column
chromatography (SiO2 45 g, petroleum CH2Cl2-MeOH, 95:5) affording 55 mg (0.16 mmol; yield:
1
9.0%) of orange needles. mp = 200 °C. UV-vis(CH CN): 500 nm (=95000); Φfluo (529 nm): 0.61; H
3
NMR (CDCl3)  1.45 (s, 6H, 2×CH3), 2.57 (s, 6H, 2×CH3), 3.79 (s, 3H, OCH3), 5.98 (s, 2H), 7.01
(d, 1H, J = 8 Hz), 7.11 (m, 2H), 7.47 (m, 1H).
Compound 15a (49.6 mg, 0.140 mmol) was iodinated in the presence of I2 (71 mg, 0.280 mmol)
and HIO3 (49.2 mg, 0.280 mmol) as described for compound 14. The crude product was purified by
column chromatography (SiO2 27 g, petroleum ether-CHCl3, 2:8) affording 56 mg (0.09 mmol;
yield: 64.3%) of 15 as red needles, mp > 300 °C; EA: calcd for C20H19BF2I2N2O C, 39.64; H, 3.16;
N, 4.62; found: C, 39.85; H, 3.18; N, 4.63; UV-vis(CH CN): 531 nm (=60400); Φfluo (559 nm): 0.03;
3

ACCEPTED MANUSCRIPT
1H NMR (CDCl3)  1.47 (s, 6H, 2×CH3), 2.67 (s, 6H, 2×CH3), 3.79 (s, 3H, OCH3), 7.03 (d, 1H, J =
8 Hz), 7.12 (m, 2H), 7.52 (m, 1H); 13C NMR (CDCl3)  16.02, 16.28, 55.69, 85.30, 111.37, 121.78,
123.55, 129.29, 131.25, 131.45, 138.91, 144.84, 156.21, 156.34; MS (ESI): M-1+ found: 604.88
calcd for C20H19BF2I2N2O 606.00; HPLC: retention time = 12’09’’ (98.1%); Photo-degradation t60:
50%, t120: 71%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(4’-nitrophenyl)-4-bora-3a,4adiaza-s-indacene
(16). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 4-nitrobenzaldehyde (226.7 mg, 1.50 mmol)
were reacted as described for compound 14a. The raw material was purified by column
chromatography (SiO2 50 g, petroleum CH2Cl2-petroleum ether, 6:4) affording 61 mg (0.16 mmol;
yield: 11%) of orange needles. mp = 189 °C. UV-vis(CH CN): 501 nm (=110500); Φfluo (521 nm):
3
0.78; 1H NMR (CDCl3)  1.38 (s, 6H, 2×CH3), 2.59 (s, 6H, 2×CH3), 6.04 (s, 2H), 7.56 (d, 2H, J = 9
Hz), 8.41 (d, 2H, J = 9 Hz).
The compound 16a (48 mg, 0.130 mmol) was treated with I2 (66 mg, 0.260 mmol) and HIO3 (45.7
mg, 0.260 mmol) as described for compound 14. The crude product was purified by column
chromatography (SiO2 20 g, petroleum ether-CH2Cl2, 1:1) affording 50 mg (0.080 mmol; yield:
61.5%) of 16 as red needles, mp > 300 °C; EA: calcd for C19H16BF2I2N3O2 C, 36.75; H, 2.60; N,
6.77; found: C, 36.82; H, 2.60; N, 6.82; UV-vis(CH CN): 535 nm (=65400); Φfluo (552 nm): 0.01; 1H
3
NMR (CDCl3)  1.40 (s, 6H, 2×CH3), 2.69 (s, 6H, 2×CH3), 7.55 (d, 2H, J = 9 Hz), 8.45 (d, 2H, J =
13
9 Hz); C NMR (CDCl3) 16.18, 17.35, 86.47, 124.67, 129.56, 130.54, 137.82, 141.59, 144.72,
148.64, 158.07; MS (ESI): M-1+ found: 620.01 calcd for C19H16BF2I2N3O2 620.97; HPLC: retention
time = 9’57’’ (100%); Photo-degradation t60: 57%, t120: 74%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(3’-cyanophenyl)-4-bora-3a,4adiaza-s-indacene
(17). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 3-cyanobenzaldehyde (197 mg, 1.50 mmol)
were reacted as described for compound 14a. The raw material was purified by column
chromatography (SiO2 40 g, petroleum CH2Cl2-petroleum ether, 6:4) affording 79 mg (0.22 mmol;

ACCEPTED MANUSCRIPT
yield: 14.7%) of 17a as an orange needles. mp = 195 °C. UV-vis(CH CN): 500 nm (=79000); Φfluo
3
(519 nm): 0.68; 1H NMR (CD3OD)  1.38 (s, 6H, 2×CH3), 2.58 (s, 6H, 2×CH3), 6.03 (s, 2H), 7.60
(d, 1H, J = 10 Hz), 7.68 (m, 2H), 7.83 (d, 1H, J = 10 Hz).
The 17a (67 mg, 0.20 mmol), I2 (101.5 mg, 0.40 mmol) and HIO3 (70.4 mg, 0.40 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 25 g, petroleum CH2Cl2-petroleum ether, 6:4) affording 55 mg (0.091 mmol; yield: 45.5%) of
17 as red needles, mp > 300 °C; EA: calcd for C20H16BF2I2N3 C, 39.97; H, 2.68; N, 6.99; found: C,
40.22; H, 2.67; N, 6.93; UV-vis(CH CN): 534 nm (=57000Φfluo (551 nm): 0.02; 1H NMR (CDCl3) 
3
1.40 (s, 6H, 2×CH3), 2.68 (s, 6H, 2×CH3), 7.58 (dt, 1H, J1 = 3 Hz, J2 = 15 Hz), 7.65 (t, 1H, J = 3
13
Hz), 7.72 (t, 1H, J = 9 Hz), 7.88 (dt, 1H, J1 = 3 Hz, J2 = 9 Hz); C NMR (CDCl3) 16.15, 17.35,
86.41, 113.99, 117.54, 130.47, 130.94, 131.65, 132.59, 133.21, 136.36, 137.49, 144.74, 158.00MS
(ESI): M+1+ found: 601.95 calcd for C20H16BF2I2N3 C 600.98; HPLC: retention time = 9’20’’
(99.80%) Photo-degradation t60: 54%, t120: 72%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’,5’-dibromophenyl)-bora-3a,4a-diaza-s-
indacene (18). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 2,5-dibromobenzaldehyde (396 mg,
1.50 mmol) were reacted as described for compound 14a. The raw material was purified by column
chromatography (SiO2 55 g, petroleum ether-CH2Cl2, 4:6) affording 95 mg (0.20 mmol; yield:
13.3%) of 18a as an orange needles. mp = 274 °C. UV-vis(CH CN): 505 nm (=151000); Φfluo (524
3
1
nm): 0.80; H NMR (CD3OD)  1.50 (s, 6H, 2×CH3), 2.58 (s, 6H, 2×CH3), 6.04 (s, 2H), 7.50 (d,
1H, J = 12 Hz), 7.51 (s, 1H), 7.59 (d, 1H, J = 12 Hz).
The 18a (91 mg, 0.19 mmol), I2 (96.4 mg, 0.38 mmol) and HIO3 (66.84 mg, 0.38 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 32 g, petroleum ether-CH2Cl2, 4:6) affording 78 mg (0.106 mmol; yield: 55.8%) of 18 as red
needles, mp > 300 °C; EA: calcd for C19H15BBr2F2I2N2 C, 31.10; H, 2.06; N, 3.82; found: C, 31.15;
H, 2.06; N, 3.84; UV-vis(CH CN): 540 nm (=53000); Φfluo (555 nm): 0.03; 1H NMR (CD3OD)  1.53
3

ACCEPTED MANUSCRIPT
(s, 6H, 2×CH3), 2.68 (s, 6H, 2×CH3), 7.47 (d, 1H, J = 6 Hz), 7.56 (dd, 1H, J1 = 12 Hz, J2 = 3 Hz),
7.62 (d, 1H, J = 12 Hz); 13C NMR (CDCl3) 16.16, 16.88, 85.79, 124.91, 125.32, 130.65, 131.13,
132.71, 133.12, 138.21, 139.52, 143.34, 157.01; MS (ESI): M-H+ found: 730.76 calcd for
C19H15BBr2F2I2N2 C 733.36; HPLC: retention time = 16’21’’ (96.0%); Photo-degradation t60: 49%,
t120: 66%.
4,4-difluoro2,6-diiodo-1,3,5,7-tetramethyl-8-[2’-(2’’-hydroxyethoxy)phenyl)]-4-bora-
3a,4adiaza-s-indacene (19). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 2-(2-
hydroxyethoxy)benzaldehyde (249 mg, 1.50 mmol) were reacted as described for compound 14a.
The raw material was purified by column chromatography (SiO2 42 g, petroleum CH2Cl2-MeOH,
98:2) affording 157 mg (0.409 mmol; yield: 27.3%) of 19a as an orange needles. mp > 300 °C. UV-
1
vis(CH CN): 500 nm (=125000); Φfluo (518 nm): 0.72; H NMR (CD3OD)  1.47 (s, 6H, 2×CH3),
3
2.57 (s, 6H, 2×CH3), 3.80 (t, 2H, J = 6 Hz), 4.09 (t, 2H, J = 6 Hz), 6.00 (s, 2H), 7.02 (d, 1H, J = 8
Hz), 7.13 (t, 1H, J = 3 Hz), 7.18 (d, 1H, J = 3 Hz), 7.46 (t, 1H, J = 8 Hz).
The 19a (146 mg, 0.38 mmol), I2 (193 mg, 0.76 mmol) and HIO3 (133.6 mg, 0.76 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 35 g, CH2Cl2-MeOH, 97:2) affording 110 mg (0.173 mmol; yield: 45.5%) of 19 as red
needles, mp > 300 °C; EA: calcd for C21H21BF2I2N2O2 C, 39.66; H, 3.33; N, 4.40; found: C, 39.88;
H, 3.32; N, 4.43; UV-vis(CH CN): 531 nm (=66000); Φfluo (559 nm): 0.01; 1H NMR (CD3OD)  1.49
3
(s, 6H, 2×CH3), 2.67 (s, 6H, 2×CH3), 3.82 (t, 2H, J = 6 Hz), 4.10 (t, 2H, J = 6 Hz), 7.06 (d, 1H, J =
12 Hz), 7.14 (m, 2H), 7.52 (td, 1H, J1 = 9 Hz, J2 = 3 Hz); 13C NMR (CDCl3)  16.04, 16.36, 61.03,
70.31, 85.39, 112.82, 122.34, 123.83, 129.60, 131.36, 138.52, 144.80, 155.52, 156.56; MS (ESI):
M+ found: 635.97 calcd for C21H21BF2I2N2O2 636.02; HPLC: retention time = 9’42’’ (98.5%);
Photo-degradation t60: 53%, t120: 66%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-[4’-(4’’-chlorobuthoxy)phenyl)]-4-bora-
3a,4adiaza-s-indacene (20). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 4-(4-
chlorobuthoxy)benzaldehyde (319 mg, 1.50 mmol) were reacted as described for compound 14a.

ACCEPTED MANUSCRIPT
The raw material was purified by column chromatography (SiO2 55 g, CH2Cl2) affording 149 mg
(0.345 mmol; yield: 23.0%) of 20a as an orange needles. mp > 300 °C. UV-vis(CH CN): 502 nm
3
(=118000); Φfluo (520 nm): 0.79; 1H NMR (CD3OD)  1.45 (s, 6H, 2×CH3), 2.03 (m, 4H, 2×CH2),
2.57 (s, 6H, 2×CH3), 3.67 (t, 2H, J = 7 Hz), 4.08 (t, 2H, J = 7 Hz), 6.00 (s, 2H), 7.02 (d, 2H, J = 10
Hz), 7.18 (t, 2H, J = 10 Hz).
The 20a (138 mg, 0.32 mmol), I2 (162.4 mg, 0.64 mmol) and HIO3 (112.6 mg, 0.64 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 35 g, CH2Cl2) affording 98 mg (0.143 mmol; yield: 44.7%) of 20 as red needles, mp > 300
°C; EA: calcd for C23H24BClF2I2N2O C, 40.47; H, 3.54; N, 4.10; found: C, 40.21; H, 3.52; N, 4.11;
1
UV-vis(CH CN): 528 nm (=52000); Φfluo (549 nm): 0.04; H NMR (CD3OD)  1.51 (s, 6H, 2×CH3),
3
2.05 (m, 4H, 2xCH2), 2.66 (s, 6H, 2×CH3), 3.68 (t, 2H, J = 8 Hz), 4.10 (t, 2H, J = 8 Hz), 7.04 (d,
13
2H, J = 11 Hz), 7.15 (d, 2H, J = 11 Hz); C NMR (CDCl3)  16.00, 17.18, 26.63, 29.31, 44.61,
67.24, 85.53, 115.36, 126.79, 129.15, 131.74, 141.76, 145.35, 156.61, 159.86;MS (ESI): M+ found:
682.30 calcd for C23H24BClF2I2N2O 682.52; HPLC: retention time = 17’19’’ (99.1%); Photo-
degradation t60: 59%, t120: 75%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’,6’-dichloro-4’-nitrophenyl)-4-bora-
3a,4adiaza-s-indacene (21). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 2,6-dichloro-4-
nitrobenzaldehyde (220 mg, 1.50 mmol) were reacted as described for compound 14a. The raw
material was purified by column chromatography (SiO2 55 g, petroleum ether-CH2Cl2, 1:1)
affording 100 mg (0.228 mmol; yield: 15.2%) of 21a as an orange needles. mp > 300 °C. UV-
vis(CH CN): 504 nm (=112000); Φfluo (522 nm): 0.77; 1H NMR (CDCl3)  1.52 (s, 6H, 2×CH3), 2.65
3
(s, 6H, 2×CH3), 6.02 (s, 2H), 8.38 (s, 2H).
The 21a (87.6 mg, 0.20 mmol), I2 (101.5 mg, 0.40 mmol) and HIO3 (70.4 mg, 0.40 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 30 g, petroleum ether-CH2Cl2, 6:4) affording 82 mg (0.119 mmol; yield: 59.5%) of 21 as red
needles, mp > 300 °C; EA: calcd for C19H14BCl2F2I2N3O2 C, 33.08; H, 2.05; N, 6.09; found: C,

ACCEPTED MANUSCRIPT
33.11; H, 2.05; N, 6.07; UV-vis(CH CN): 530 nm (=56000); Φfluo (547 nm): 0.02; 1H NMR (CDCl3) 
3
13
1.52 (s, 6H, 2×CH3), 2.70 (s, 6H, 2×CH3), 8.38 (s, 2H); C NMR (CDCl3)  15.87, 16.43, 85.18,
123.38, 130.58, 135.37, 138.45, 140.32, 144.63, 149.09, 158.13; MS (ESI): M+ found: 689.68 calcd
for C19H14BCl2F2I2N3O2 689.86; HPLC: retention time = 10’57’’ (99.8%); Photo-degradation t60:
57%, t120: 71%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(2’,4’,6’-trimethoxyphenyl)-4-bora-3a,4adiaza-s-
indacene (22). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 2,4,6-trimethoxybenzaldehyde (294.3
mg, 1.50 mmol) were reacted as described for compound 14a. The raw material was purified by
column chromatography (SiO2 55 g, petroleum CH2Cl2-MeOH, 99:1) affording 148 mg (0.357
mmol; yield: 23.8%) of 22a as an orange needles. mp > 300 °C. UV-vis(CH CN): 502 nm (=113000);
3
1
Φfluo (521 nm): 0.65; H NMR (CDCl3)  1.57 (s, 6H, 2×CH3), 2.62 (s, 6H, 2×CH3), 3.72 (s, 6H,
2×OCH3), 3.91 (s, 3H, OCH3), 6.00 (s, 2H), 6.26 (s, 2H).
The 22a (141 mg, 0.34 mmol), I2 (170 mg, 0.67 mmol) and HIO3 (118 mg, 0.67 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 30 g, CH2Cl2-MeOH, 99:1) affording 121 mg (0.182 mmol; yield: 53.5%) of 22 as red
needles, mp > 300 °C; EA: calcd for C22H23BF2I2N2O3 C, 39.67; H, 3.48; N, 4.21; found: C, 39.70;
1
H, 3.47; N, 4.23; UV-vis(CH CN): 530 nm (=45000); Φfluo (554 nm): 0.01; H NMR (CDCl3)  1.57
3
13
(s, 6H, 2×CH3), 2.65 (s, 6H, 2×CH3), 3.72 (s, 6H, 2×OCH3), 3.91 (s, 3H, OCH3), 6.24 (s, 2H); C
NMR (CDCl3)  16.10, 17.01, 55.03, 57.65, 84.99, 108.58, 124.34, 131.97, 137.45, 145.09, 156.13,
156.45, 156.89; MS (ESI): M+ found: 665.96 calcd for C22H23BF2I2N2O3 666.05; HPLC: retention
time = 10’52’’ (95.6%); Photo-degradation t60: 55%, t120: 74%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(4’-cyanophenyl)-4-bora-3a,4adiaza-s-indacene
(23). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 4-cyanobenzaldehyde (197 mg, 1.50 mmol)
were reacted as described for compound 14a. The raw material was purified by column
chromatography (SiO2 50 g, CH2Cl2) affording 169 mg (0.484 mmol; yield: 32.3%) of 23a as an

ACCEPTED MANUSCRIPT
1
orange needles. mp > 300 °C. UV-vis(CH CN): 498 nm (=111000); Φfluo (516 nm): 0.81; H NMR
3
(CDCl3)  1.37 (s, 6H, 2×CH3), 2.67 (s, 6H, 2×CH3), 6.02 (s, 2H), 7.47 (d, 2H, J = 10 Hz), 7.88 (d,
2H, J = 10 Hz).
The 23a (148 mg, 0.44 mmol), I2 (223.3 mg, 0.88 mmol) and HIO3 (155 mg, 0.88 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 40 g, CH2Cl2) affording 189 mg (0.314 mmol; yield: 71.4%) of 23 as red needles, mp > 300
°C; EA: calcd for C20H16BF2I2N3 C, 39.97; H, 2.68; N, 6.99; found: C, 40.01; H, 2.70; N, 7.02; UV-
1
vis(CH CN): 530 nm (=76000); Φfluo (553 nm): 0.03; H NMR (CDCl3)  1.39 (s, 6H, 2×CH3), 2.68
3
(s, 6H, 2×CH3), 7.47 (d, 2H, J = 10 Hz), 7.87 (d, 2H, J = 10 Hz); 13C NMR (CDCl3)  16.14, 17.25,
86.35, 113.91, 117.84, 129.23, 130.61, 133.16, 138.15, 139.73, 144.78, 157.98; MS (ESI): M+
found: 601.00 calcd for C20H16BF2I2N3 600.98; HPLC: retention time = 9’53’’ (96.9%); Photo-
degradation t60: 53%, t120: 70%.
4,4-difluoro-2,6-diiodo-1,3,5,7-tetramethyl-8-(3’,5’-dimethoxyphenyl)-4-bora-3a,4adiaza-s-
indacene (24). 2,4-Dimethylpyrrole (340 µL, 3.30 mmol) and 3,5-dimethoxybenzaldehyde (249
mg, 1.50 mmol) were reacted as described for compound 14a. The raw material was purified by
column chromatography (SiO2 50 g, CH2Cl2-MeOH, 99:1) affording 189 mg (0.492 mmol; yield:
32.8%) of 24a as an orange needles. mp > 300 °C. UV-vis(CH CN): 498 nm (=111000); Φfluo (517
3
1
nm): 0.74; H NMR (CD3OD)  1.55 (s, 6H, 2×CH3), 2.63 (s, 6H, 2×CH3), 3.80 (s, 6H, 2×OCH3),
6.01 (s, 2H), 6.42 (d, 2H, J = 2.5 Hz), 6.60 (d, 1H, J = 2.5 Hz).
The 24a (181 mg, 0.47 mmol), I2 (238.6 mg, 0.94 mmol) and HIO3 (165.3 mg, 0.94 mmol) were
reacted as described for compound 14. The crude product was purified by column chromatography
(SiO2 40 g, CH2Cl2) affording 223 mg (0.351 mmol; yield: 74.7%) of 24 as red needles, mp > 300
°C; EA: calcd for C21H21BF2I2N2O2 C, 39.66; H, 3.33; N, 4.40; found: C, 39.71; H, 3.33; N, 4.39;
1
UV-vis(CH CN): 533 nm (=46000); Φfluo (556 nm): 0.02; H NMR (CDCl3)  1.58 (s, 6H, 2×CH3),
3
2.66 (s, 6H, 2×CH3), 3.82 (s, 6H, 2×OCH3), 6.43 (d, 2H, J = 2.5 Hz), 6.60 (d, 1H, J = 2.5 Hz); 13C

ACCEPTED MANUSCRIPT
NMR (CDCl3)  16.26, 17.03, 57.02, 85.39, 109.08, 112.87, 122.99, 131.89, 139.22, 144.76,
156.03, 157.20; MS (ESI): M+ found: 636.09 calcd for C21H21BF2I2N2O2 636.02; HPLC: retention
time = 12’37’’ (99.0%); Photo-degradation t60: 56%, t120: 68%.
Fluorescence properties and Quantum Yield of Fluorescence.
Steady-state fluorescence spectroscopic studies were performed with a Jasco FP-750
spectrofluorimeter. The slit width was 2.5 nm for both excitation and emission. Values of
absorbance, necessary for the equation described below, were obtained on a Varian Cary 50 Scan
instrument by using the same solutions of the fluorescence measurements. The relative quantum
efficiencies of fluorescence of each BODIPY derivative was obtained comparing the emission
spectra of the test sample, prepared as dichloromethane solution (dieletric constants 9.14), with that
of a 0.1 M NaOH solution of fluorescein excited at 490 nm, whose quantum efficiency literature
value is 0.85 [20]. The quantum efficiencies of fluorescence (Φfluo) of the BODIPYs were
calculated according to the following equation (Ifluo is the fluorescence intensity at the specific
excitation wavelength, Abs denotes the absorbance at the excitation wavelength):
Φfluosample=Φfluostandard×(Ifluosample/Ifluostandard)×(Absstandard/Abssample
)
Comparative Singlet-Oxygen generation measurements.
A solution of aerated isopropanol containing 50 μM of 1,3-diphenylisobenzofuran (DPBF) and 2.5
μM of photosensitizers was prepared in a glass flask and kept in the dark; 2 mL of this solution
were transferred into a cuvette, placed in the UV-vis instrument and irradiated from the open top
side with the green LED source, at room temperature and for the time required to observe the
chemical event (maximum 20 min.). The DPBF absorbance decrease was measured at 410 nm at
fixed time intervals, as defined for each experiment. Two controls were also set up, the first to study
the effect of the light irradiating a DPBF-isopropanol solution in the absence of photosensitizer
(negative control) and the second irradiating a DPBF-isopropanol and Rose Bengal (2.5 μM)
(comparative normalizing data). The rate of singlet oxygen production for each PS was obtained
plotting the variation of DPBF absorbance over the irradiation time and the relative singlet oxygen

ACCEPTED MANUSCRIPT
generation rates were normalized with respect to the rate determined by using Rose Bengal as
standard [21].
Photo-degradation of the BODIPYs.
Phosphate Buffer Solution (PBS) of each PS (10 μM) was prepared in a flask. These solutions were
irradiated, for 2 h at room temperature, with the green LED; within this period 1 mL sample was
collected after 60 min and at 120 min and analysed with the spectrophotometer. The percentage of
photo-degradation was determined at 60 (t60) and 120 (t120) min as the ratio of the absorption
intensity with respect to absorption measured at t0.
Cytotoxicity studies.
The human ovarian carcinoma cell line SKOV3 was obtained from the American Type Culture
Collection (Rockville, MD, USA) and maintained in Dulbecco’s modified Eagle medium-F12
(DMEM-F12, Mascia-Brunelli, Milano, Italy) supplemented with 10% fetal bovine serum, 1%
glutamine and 0.5% antibiotic mixture (penicillin, streptomycin and neomycin), in standard culture
conditions at 37 °C in a humidified 5% CO2 atmosphere.
The antiproliferative effect of each BODIPY was assessed using the MTT assay [22]. Briefly,
1.5×103 cells/well were seeded onto 96-wells plate and allowed to grow for 48 h prior to the
treatment with 100 L of BODIPY solutions at different concentration. These solutions were
prepared by dilution with PBS of a stock BODIPY solution, prepared in acetone or DMSO at 1 mM
concentration, according to the PS solubility.
The dilution of the stock BODIPY solutions were made in order to get a cell non toxic
concentration of the organic solvent in the solution used in the cells treatment (below 1%).
After 24 h, the BODIPY-containing medium was replaced by fresh PBS, and cells were irradiated
under visible light for 2 h using a green LED (fluence 25.2 J/cm2). At the end of irradiation, cells
were incubated in the dark for 24 h at 37 °C in drug-free medium then MTT was added to each well
(final concentration 0.4 mg/mL) and left reacting for 3 h at 37 °C. The formazan crystals, formed

ACCEPTED MANUSCRIPT
through MTT metabolism by viable cells, were dissolved in DMSO and the optical densities were
measured at 570 nm using a Universal Microplate Reader EL800 (Bio-Tek Instruments).
Possible intrinsic cytotoxic effects (i.e. not photo induced cell death) of the BODIPYs were
assessed on control cultures kept in the dark and treated, as described above, with concentrations of
BODIPY tenfold higher than those used in the PDT assays.
IC50 values (i.e. the concentration affecting 50% of cell survival fraction) were obtained by
nonlinear regression analysis, using the GraphPad PRISM 3.03 software (GraphPad Software Inc.,
San Diego CA).
Intra-Cellular localization.
Non-iodinated BODIPYs, namely those featuring a hydrogen atom on the 2,6 positions (H2-
BODIPY), were used in these experiments because of these molecules are characterized by high
fluorescence, a useful property for the detection of the PS inside the cells. Thus, SKOV3 cells were
seeded onto coverslips (10000 cells), allowed to grow for 48 h and subsequently treated with the
desired H2-BODIPY at a concentration ten time higher than the IC50 values, as determined in the
MTT assays with the iodinated derivatives. After 24 h, cell attached to coverslips were washed
three time in 1X PBS and fixed in 3% paraformaldehyde for at least 10 min and then washed in 1X
PBS three times. Coverslips were mounted on microscope slides. Images were acquired using a
LEICA TCS SP8 X confocal laser scanning microscope.
Flow cytometric studies.
The effects of compound 2 were assessed at different concentrations (0.5, 1 and 2 nM) on SKOV3
cell line. Following the described treatment (24 h exposure to the drugs, 2 h irradiation in drug-free
PBS and 24 incubation in drug-free medium) the percentage of apoptotic cells and cell cycle
distribution were assessed using Propidium Iodide (PI) staining. Cells were fixed in 0.5 mL of ice-
cold 70% ethanol and stored at −20 °C. Cells were then centrifuged, washed with PBS, resuspended
in the dying solution (50 μg/mL propidium iodide, 20 μg/mL RNase in 1× PBS) and analyzed. The

ACCEPTED MANUSCRIPT
percentage of apoptotic cells in each sample was determined based on the sub-G1 peaks detected in
monoparametric histograms.
Intracellular ROS levels at the end of compound 2 treatment was assessed using 2,7-
Dichlorodihydrofluorescein diacetate (DCHF-DA) as a fluorogenic probe. DCHF-DA is cleaved by
intracellular esterases to release the corresponding dichlorodihydrofluorescein derivative (non-
fluorescent), which is readily oxidized to dichlorofluorescein (DCF) by reactive oxygen species
(ROS).
Cells were treated with compound 2 at 0.5, 1 and 2 nM for 24 h, followed by 2 h irradiation. Cells
were then detached, washed and resuspended in PBS containing 10 μM DCFH-DA and analyzed.
after 45 min dark incubation at 37 °C.
All samples were analyzed with FACScalibur flow cytometer (Becton–Dickinson, Mountain View,
CA, USA) equipped with a 15 mW, 488 nm air cooled Argon ion laser. Fluorescent emissions were
collected through a 530 nm band-pass filter for fluorescein, a 575 nm bandpass filter for PI. At least
10,000 events were analyzed for each sample. Data were processed using CellQuest software
(Becton-Dickinson).
QSAR Analysis: experimental procedures.
In order to perform Quantitative Structure Activity Relationship (QSAR) analysis, the IC50
concentrations for the 24 molecules were expressed in logarithmical form, log (IC50). The three-
dimensional structures of the studied BODIPYs were designed in the computational software
HYPERCHEM [23] and their conformation was optimized at the minimum energy by the semi-
empirical Austin Model 1 (AM1) method. The topological, geometrical and chemical diversity of
compounds was numerically encoded by theoretical molecular descriptors calculated by DRAGON
(ver. 5) software [24] using the atomic coordinates (x, y, z) of the minimum energy conformation of
the three dimensional structures as input. More than 1200 mono-, bi- and three-dimensional
molecular descriptors were calculated. Moreover, some quantum chemicals descriptors such as
HOMO, LUMO, HOMO-LUMO Gap (DHL), Ionization Potential, Binding Energy (calculated by

ACCEPTED MANUSCRIPT
semi-empirical AM1 method in HYPERCHEM), and experimental properties such as HPLC
1
retention time (Rt) and O2 production rates (1O2) were added as potential modeling variables for
the descriptors selection during the model development. Then variables were analyzed in a pre-
reduction phase where constant (for more than the 80%) and pairwise correlated descriptors (one
variables was deleted when there was more than 98% pairwise correlation,) were excluded to delete
null and redundant information. The final reduced set of 427 molecular descriptors was used as
input for explorative chemometric analyses and QSAR models development. The Genetic
Algorithm variable subset selection (GA-VSS) and Multiple Linear Regression (MLR) by the
Ordinary Least Square (OLS) were performed by the QSARINS (QSAR-INSubria) software
[25,26]. All the possible combination of two descriptors were investigated by the all-subset
procedure. Then, the GA was applied (3000 iterations) to find the best three variable combinations
starting from a population of 100 models; Q2 (leave one out) was used as fitness function to be
loo
optimized during the variable selection procedure. The robustness and stability of the models were
evaluated by Q2loo, Q2 (leave more out) and CCC (concordance correlation coefficient) [27] in
lmo
cross validation. To guarantee good internally predictive models, the values of the Q2 functions
should be around 0.7 and the CCC values should be around 0.85 [28]. Cross Validation is an
iterative process that involves a sequential splitting of the original training set in different test sets
on which the fitting is re-tested. In particular, Q2 and Q2
measure the internal predictive
lmo
loo
capacity when, respectively, one or more compounds, in a series of iterations, are excluded from the
computation and then predicted by the model recalculated on the remaining objects. For the leave
more out procedure we run 2000 iterations with the 30% of compounds excluded from computation
for each iteration. Moreover, also permutation testing, by y-scrambling procedure, was performed to
exclude by chance models. The test is performed calculating, in a series of iterations, the quality of
the model (usually R2), randomly modifying the sequence of the response vector y, by assigning to
each object a response randomly selected from the true responses. Low values of the averaged R2
scrambled (R2 ) are indicative of a well founded (not by chance) original model [29,30]. A further
ys

ACCEPTED MANUSCRIPT
measure to define the accuracy of the proposed QSAR model is also the RMSE (Root Mean Square
Error) that summarizes the overall error of the model. It is calculated as the root square of the sum
of the squared errors in predictions divided by their total number (RMSEtr and RMSEcv, calculated
separately for the training and the cross validation). Finally, QSAR models must always be verified
for their applicability with regard to chemical domain. For this reason we investigated, by Williams
plot [29], the presence of outliers (molecules with cross-validated standardized residuals greater
than 2.5 deviation unit) and molecules structurally influent in the selection of modelling descriptors
and parameters (compounds with leverage value (h) higher than h* (3 p’/n).
Results and Discussion
A series of BODIPYs was synthesized, based on the assumption that the presence of different
substituents on an aromatic ring on the meso (8) position, would affect both the
hydrophobic/hydrophilic balance and the rate of singlet oxygen production of these PSs. The first
chemical property is crucial as it can influence the cell penetration and the target of the subcellular
localization, both depending on the lipophilicity degree of the photosensitizer [19]. As far as it
concerns another fundamental aspect influencing the PDT outcome, i.e. the singlet oxygen
generation rate, we have already reported that an aromatic moiety on the BODIPY meso position
induces different rates depending on the nature of the substituent tethered on the phenyl ring [19].
Chemistry.
The BODIPY derivatives were synthesized via acid catalysed condensation of the desired aromatic
aldehyde (11 different aldehydes were used) with 2,4-dimethylpyrrole in the presence of catalytic
amount of trifluoroacetic acid (TFA), following the general methods described by Akkaya and Liu
[31,32]. Either dichloromethane (DCM) or tetrahydrofuran (THF) were used as solvent in the
BODIPY syntheses, depending on the solubility of the aldehyde; in particular, THF was
conveniently used when hydroxy substituted aldehydes were the reactant. The disappearance of the
aldehyde spot in the TLC analysis, fulfilled for the reaction control, indicates the formation of
dipyrrolylmethane that was subsequently dehydrogenated to dipyrrolylmethene with 2,3-dichloro-

ACCEPTED MANUSCRIPT
5,6-dicyano-1,4-benzoquinone (DDQ) added in the same flask. In the end, the mixture was treated
.
with BF3 Et2O and Et3N thus achieving the complete synthesis of the desired BODIPYs. Each
compound has been isolated as pure solid after a single chromatographic step of purification
(Scheme 1).
The 2,6-positions of each BODIPY were modified with the insertion of iodine atoms (Scheme 1) to
favour the passage of the PS from the singlet to the triplet excited state following irradiation, thus
increasing the singlet oxygen production rate. Regarding the iodination process, we found that the
conditions reported in the literature (I2, HIO3, 60 °C, 20 min) [14] cause a partial degradation of the
dipyrrolylmethene skeleton, leading to extremely low yields in the desired iodinated boron-
compounds. The successful recovery of the iodinated derivatives in good yields (> 45%) was
obtained using a lower temperature (20–25 °C) and increasing the reaction time to 15–18 h.
Scheme 1
The BODIPYs synthesized in this work and those already reported [19] are listed in Figure 1. Each
compound was completely characterized in terms of UV-vis absorption spectra, quantum yield of
fluorescence, NMR and MS spectra. In addition, for a better understanding of the photodynamic
process, HPLC retention time (Rt) and singlet oxygen generation were also measured.
Figure 1
The lipophilicity of organic compounds is generally measured by considering the partition
coefficients (P) in a 1-octanol /water mixture. The P values are commonly obtained following Uv-vis
analysis of both the organic and the aqueous phase according to the equation Log P = Log
[PS]octanol/[PS]water This method cannot be applied for those compounds that hardly distribute in both
the phases (being or highly lipophilic or, vice versa, extremely hydrophilic). When the method based
on absorbance (or the more sensible fluorescence) cannot be applied, as for our molecules, a
comparative degree of lipophilicity of chemicals can be determined evaluating the HPLC retention
time (Rt) on a RP-C18 column, in the event that the molecules belong to a homogeneous series and
performing the analyses under standardized conditions.

ACCEPTED MANUSCRIPT
In the case of the described BODIPYs, we can assert that the aromatic group in meso position
modifies the hydrophilicity of these molecules (table 1). Considering the retention time values, these
BODIPYs could be arbitrarily divided in three groups: the first group, characterized by Rt values
below 10 min, includes the hydrophilic BODIPYs (8, 14, 16, 17, 19, 21 and 23) among which the
most polar is BODIPY 8, characterized by the presence of two chlorine atom in the ortho, ortho’
position of the phenyl ring. With the only exception of BODIPY 19, characterized by the presence of
a ethoxy-alcohol appendix, the molecules belonging to the hydrophilic set are characterized by the
presence of electron withdrawing substituent such as chloride atoms, cyano and nitro groups. A
second group of compounds can be envisaged considering the Rt interval between 10 and 15 min.
This is the most numerous set as eleven molecules (2, 3, 4, 5, 9, 11, 12, 13, 15, 22 and 24) show an
intermediate level of the hydrophilic/lipophilic ratio. The presence of heterogeneous substituents on
the aromatic ring characterizes the structure of these BODIPYs. In fact, the aromatic ring of the
second group show the presence of either electron-withdrawing or -donating group. Many phenyls
show the presence of one or more methoxy groups (one in compounds 2, 3 and 15; two in compound
24 and three in 22). Two molecules show one phenolic group that should ensure a high
hydrophilicity. They are expected with shorter Rt then to be part of the first set of polar molecules
rather than the second set. However, the presence of large iodine atom(s) nearby the OH masks the
hydrophilic effect of the OH (compounds 4 and 5). Two BODIPYs, belonging to this second set of
polarity, are characterized by the presence of pyridyl moiety (compound 11 and 12) that could be
protonated in acidic aqueous medium. At pH 7.2 (the buffered aqueous phase used in these HPLC
analyses), the pyridyl group is present as free base thus showing a weak interaction with the
lipophilic stationary phase of the C18 columns. The pentafluorophenyl substituted compound 9 fall
in this group of molecules because of the two opposite effects of the fluorine atoms. Their
electronegativity increases the polarity of the BODIPY but, it is also known, the high hydrophobicity
of polyfluorinated compounds. The observed Rt is the result of a balance of these two effect. Quite
surprisingly, compound 13, the only one characterized by the presence of a hydrogen atom on the

ACCEPTED MANUSCRIPT
meso position, belongs to the medium polarity set of BODIPYs. The last group of BODIPYs (Rt >
15 min,) includes the most lipophilic compounds (1, 6, 7, 10, 18 and 20) i.e. those characterized by
the highest affinity for the stationary phase. More precisely, the molecules featuring the nonpolar
naphthyl group (6, 7) show a greater affinity for the C18 reversed phase HPLC column, as witnessed
by the highest Rt. A similar strong interaction between the BODIPYs and the stationary phase was
also observed for compounds 10 and 20, both characterized by the presence of a long, nonpolar
appendix. As far it concerns the presence of large halogens (Br and I), as in compounds 1 and 18, it
can be assumed that the presence of these atoms increases the lipophilicity of the molecules. It is
worth noting the striking difference of polarity between the compounds bearing Br and I with respect
to the chlorine substituted BODIPY (8).
The fluorescence data, reported in Table 1, have been normalized with respect to the value obtained
using fluorescein as standard [20]. As expected, the data show a very low value of fluorescence
quantum yield as the iodine atoms exert the well known ISC effect, thus shifting the singlet excited
state into a lower energetic triplet state. This event prompts a nearly complete inhibition of
fluorescence to the advantage of a very high rate of singlet oxygen generation.
1
The production of O2 for each photosensitizer was indirectly determined measuring the
disappearance of the absorption band (at 410 nm) of the 1,3-diphenylisobenzofuran (DPBF), a well-
1
known O2 scavenger. The experiments were carried out following a literature method, using
isopropanol as solvent with a 50 M initial concentration of DPBF and in the presence of 2.5 M of
BODIPY. Irradiation was carried out with a green LED device. The choice of green light ensures a
high level of electronic excitation of these PSs because of the good overlap between the max of
absorbance of BODIPYs and the emission wavelength of the green LED, as well as with the
absorbance profile of the standard dye.
The relative rates of singlet oxygen (1O2) production were obtained comparing the kinetic of DPBF
degradation obtained with the BODIPY and with a reference molecule. In this work Rose Bengal,
widely recognized as a high singlet oxygen producer, was used as standard [21].

ACCEPTED MANUSCRIPT
The relative rates reported in Tables 1 show that all the tested BODIPYs are characterized by
intense singlet oxygen production with rates comparable or higher than Rose Bengal. Eleven
compounds showed a very high rate of 1O2 production and, among them, compounds 11, 15, 16 and
17 produce 1O2 twice as much with respect to the reference compound.
Table 1
In in vivo PDT applications, the photo-degradation of the photosensitizer plays an important role.
Indeed, any PS with a high stability can turn out to be highly phototoxic for a prolonged time after
the treatment. For this reason, we have evaluated the stability of the BODIPYs irradiating an
aqueous 10 M solution with the green LED then recording the absorbance at 60 and 120 min. The
results show that the photo-degradation of BODIPYs has an average value of 55% after 60 min
whereas it is approximately 70% after 120 min. It is evident that the degradation process rapidly
occurs at high concentrations, whereas it slows down when the majority of the PS has already
decomposed, to confirm the result previously reported by us [19]. According to the literature, it
seems that the extension of the degradation of BODIPYs under green light irradiation strongly
depends on the molecular structure [17].
Cytotoxicity Assays.
The IC50 values from dose/response curves obtained in SKOV3 cells following exposure to the
different PSs for 24 h and irradiation with Green LED device for 2 h (fluence 25.2 J/cm2) are
reported in Table 1. The intrinsic cytotoxicity of PSs was assessed omitting the irradiation step in the
treatment protocol observing a negligible effect in all cases, although PS concentrations 10-fold
higher than those used for PDT experiments were considered.
All the meso aryl substituted compounds showed a very high in vitro phototoxicity with IC50 value in
the range 0.65 – 8.03 nM. The comparison of the activity of compound 13 (IC50 = 28.75 nM),
characterized by the presence of a hydrogen atom in position 8, with those of all the other BODIPYs
clearly indicates that the presence of an aromatic ring (independently from the presence of whatever
substituents on it) on the meso position of the BODIPY frame is fundamental to ensure a high cell-

ACCEPTED MANUSCRIPT
killing efficacy. Even the less active “meso” aryl substituted BODIPYs, compounds 10 and 14, show
a remarkable higher efficacy as their IC50 values are 8.03 and 7.77 nM, respectively.
Besides the striking evidence of the negative effect exerted by the absence of substituent on the meso
BODIPY position, the tumor cell-killing efficacy of the aryl substituted BODIPYs does not give a
clear indication about which kind of substituent can ensure an activating effect or which has a clear
inhibiting effect. The two less active compounds (10 and 14) are characterized by the presence of
electron withdrawing and/or bulky substituents on the phenyl and are among the most polar. In this
context, the presence of a nitro group seems to decrease the efficacy of the photodynamic action as
evidenced by the low activity of compound 21 (2,6-dichloro-4-nitrophenyl; IC50 = 3.93 nM) and of
compound 14 (2-chloro-6-nitrophenyl; IC50 = 7.77 nM) as compared with the second best BODIPY
8 (2,6-dichlorophenyl; IC50 = 0.77 nM). However, the effect of the presence of a nitro group is not so
clear as compound 16 (4-nitrophenyl) has shown a good activity (IC50 = 2.01 nM).
It seems that the electronic properties, position(s) or steric hindrance of the substituents, present on
the meso aromatic ring, do not allow to point out a specific rule to predict the cytotoxicity of these
photosensitizers. Actually, considering the two more active compounds, BODIPY 2 and 8 (IC50 =
0.65 and 0.77 nM respectively) are characterized by groups located in different positions and with
different electronic effects (p-OMe and o,o’-chlorine atoms as appendix, respectively).
Thus we made the effort to correlate the cell-killing efficacy (IC50 Values) with two chemico-
physical properties (HPLC Rt and 1O2 production rates) reported for these compounds with the only
exclusion of compound 13 because of its striking different efficacy in the photo-induced cell-killing.
In figure 2 the IC50 values were correlated with the singlet oxygen generation (A) and with the
lipophilicity degree (B). For the correlation, we normalize the data to respect the maximum value of
each series.
1
Unfortunately we could not observe any correlation, in fact the R2 = 0.0064 for IC50 versus O2 and
R2 = 0.0111 for IC50 versus Rt were found, respectively, then it is impossible understand the activity
of these molecules based on lipophilicity or singlet oxygen production.

ACCEPTED MANUSCRIPT
Figure 2
Someway better results should be obtained with a more sophisticated QSAR analyses comparing the
results of photoinduced cell-killing activity with many different structure parameters.
QSAR analysis.
Quantitative Structure Activity Relationship (QSAR) modelling, based on theoretical molecular
descriptors, was performed in order to verify which structural features are correlated to
phototoxicity of BODIPYs against a tumor cell line.
The log (IC50) values for the 24 BODIPYs were used as responses (y), while mono-, bi- and three-
dimensional theoretical molecular descriptors were used as predictive variables (x) in the
development of descriptive MLR QSAR models in a specified structural Applicability Domain. As
described in the Experimental section, we applied an evolutionary variable selection procedure,
called Genetic Algorithm (GA-VSS), to find the best combination of descriptors able to justify the
photodynamic cytotoxicity among the whole molecular descriptors dataset. This is done because we
cannot know, a priori, which descriptors could be related with the activity of the BODIPY
describing significant structural features. Statistical internal validation was performed to assess the
stability of the models when subjected to perturbation and to ensure a certain degree of generality,
independently from the training set used for their development. Validation parameters (Q2loo, Q2
,
lmo
CCCcv) were used as criteria for the model selection in the GA population. In table 2 the best two
models, with similar statistical performances and based on combination of three descriptors with
similar structural meaning, are proposed.
Table 2
Both models have high fitting values and good validation parameters indicating that the models
have very good descriptive performances and are stable, robust and internally predictive. The
RMSE values, expressed in log unit, are low, indicating that the models have an accuracy of at least
the 90%, both in fitting and in cross validation.

ACCEPTED MANUSCRIPT
The plot of predicted and experimental data, the equation of the presented QSAR models and the
Williams plot are reported in figure 3.
Figure 3
The most important descriptors able to model the phototoxicity in both models are the GETAWAY
descriptors (GEometry, Topology, and Atom-Weights AssemblY) R2e+, R3p+ and R3v+.
GETAWAY descriptors derive from the molecular influence matrix (MIM) or from the
influence/distance matrix and encode information for structural fragments and structure-properties
correlations. In other words, these descriptors are based on the spatial autocorrelation combining 3-
D features of the molecules and their physiochemical properties such as atomic mass, polarizability,
van der Waals volume, and electronegativity. In particular, R2e+ encodes for the maximal
autocorrelation (at topological distance of two bonds) weighted by the atomic Sanderson
electronegativity. This descriptor is present in both models, with a positive coefficient meaning that
the presence of electronegative substituents is influent in decreasing the phototoxicity.
R3p+ encodes for the maximal autocorrelation (at topological distance of three bonds) weighted by
the atomic polarizability and R3v+ calculates the maximum autocorrelation value (at topological
distance of three bonds) weighted by atomic van der Waals volume. These two descriptors are
selected alternatively in the two proposed models; they have negative coefficients, therefore
chemicals with higher values of these descriptors are among the most active.
Overall, these descriptors give information regarding to the influence of electronegativity,
polarizability and molecular volume of different molecules within our data set. The other two
molecular descriptors, directly correlated with activity, are JIG8 and ARR. The JIG8 encodes for
the mean topological charge index and gives information about charge transfer between pairs of
atoms and therefore global charge transfer in the molecule. The ARR encodes for the aromatic ratio
(number of aromatic bonds over the total number of bonds excluded those with Hydrogen)
assuming higher values for compounds with more aromatic rings (such as 6, 7, 10).

ACCEPTED MANUSCRIPT
The analysis of the chemical applicability domain with the leverage approach (figure 3 b and d)
allows us to verify the presence of molecules influential for some peculiarities in their structure or
of outliers for the calculated response. In both models BODIPY 13 has h values greater than the
cutoff h*, therefore it is a high leverage compound, which has a great influence in selecting model
descriptors. It is important to note that 13 is the only one BODIPY of our set without an aryl
substituent and characterized by the lowest activity, thus it can be considered as reference
compound for the subsequent evaluation of the effect of the meso aryl group on the photoinduced
toxicity. For this reason 13 has been included in the dataset. Concerning outliers for the response,
only compound 15 was predicted with a standardized residual greater than 2.5 in model 2, while it
was well predicted by model 1.
In order to verify the contemporaneous influence of all the modelling descriptors in the distribution
of the studied chemicals, we performed a Principal Component Analysis (PCA) of the selected
molecular descriptors. In figure 4 the biplot of this PCA is reported where BODIPYs (the points)
are plotted in the descriptors space (the arrows for the loadings) of the first two principal
components (total variance explained 81.15%).
Figure 4
In particular, PC2 (EV% = 39.76) is the principal component more correlated with the response and
thus has been plotted in the abscissa axis. Compounds with lowest PC2 values are the less active
BODIPYs (left side) while high values of PC2 correspond to more effective molecules (right side).
Therefore in the PCA a trend (from left to right) of the photodynamic activity is evident. The least
active BODIPY 13 is correctly isolated in left side on the graph.
ARR has in general a positive influence in phototoxicity, highlighting that the photodynamic
activity is enhanced by aryl substituents; in fact all the compounds with aromatic rings are located
far from 13. R3v+ and R3p+ are strictly correlated and have a positive influence on the activity
trend. The compound 10 is isolated in the upper part of PC1 since it is the biggest with the lowest
values of R3v+ and R3p+. On the contrary, R2e+ has a negative correlation with the activity trend.

ACCEPTED MANUSCRIPT
Thus, less active compounds seem to be mainly characterized by the presence of highly electron
withdrawing groups, such as nitro groups (i.e. 14, 21). In these compounds, the presence of one or
two chlorine atoms, the other substituent present, seem not to contribute to the decrease of the
BODIPY efficacy, in fact, the BODIPY 8 is one of the most active compound of the series. It is
important to remember that in a multivariate analysis and modelling such as PCA and QSAR the
single descriptors are not able to justify independently the end-point; it is only a combination of the
selected descriptor set that allows modelling the studied response.
BODIPY cell penetration and percentage of apoptotic cell.
Another possibility to understand the BODIPY’s efficacy, is to evaluate the localization of the
photosensitizers following cell penetration. To perform this analysis, the non-iodinated derivatives
have to be used (H2-BODIPY) in order to exploit their high quantum yield of fluorescent, assuming
that the difference in the structure and chemical weight, between the iodinated and non–iodinated
BODIPYs, does not influence the cellular uptake. For all compounds the cellular cultures were
treated, over a period of 24 h of incubation, with a concentration of PS ten time higher than those
corresponding to the IC50 values obtained with the iodinated derivatives. An illustrative example of
the result is shown in figure 5. From this analysis become obvious that all BODIPYs were quickly
and efficiently internalized by cells and, as already reported for other PSs (porphyrins, chlorins, etc.)
[33], they are localized in the cytosol district whereas none fluorescence could be observed in the
nucleus.
Figure 5
We also investigated some possible mechanism of action of the most potent BODIPY found in this
study (compound 2). Figure 6 shows that 2 induce a significant increase in the percentage of
apoptotic cells, over control samples, at all the tested concentrations. The observed percentage of
apoptotic cells never exceed 12%, indicating that cell death induced by compound 2 occurs only in
part through apoptosis. A further contribution to the photodynamic effect of compound 2 is
evidenced in Figure 7, showing the levels of intracellular oxidizing species in compound 2 treated

ACCEPTED MANUSCRIPT
SKOV3 cells. Treatment with the BODIPY induce a dose-dependent increase in ROS generation
over control levels, even though statistical significance is only achieved when cells were treated with
a BODIPY concentration 1.0 nM, which is slightly greater than the IC50 value and with a
concentration three time higher (2.0 nM).
Cell cycle distribution analysis did not show any difference following treatment with the BODIPY
(data not shown).
Figure 6
Figure 7
Conclusions
The panel of iodinated BODIPY here reported has shown a very interesting singlet oxygen
production rate, in most cases higher than the one obtained with the reference compound Rose
Bengal. In agreement with these data a very promising cell-killing effect following irradiation with a
green light emitting LED was observed in in vitro assays. The efficacy is undoubtedly correlated to
the presence of an aromatic ring on the BODIPY 8 (meso) position although none evident
indications can be inferred from the presence of electron-withdrawing or electron-donating
substituents on the aromatic moiety or according to their position and then their steric hindrance.
Actually, the two most active BODIPYs are 2 and 8, the former characterized by the presence of a
methoxy group on the para position whereas the latter features two chlorine atoms on the ortho,
ortho’ positions. The complete difference in regio-isomerism and in the electronic effects confirming
the absence of correlation between the substituents and the efficacy of the photo-induced action. The
lipophilic degree also does not influence the PSs efficacies as the compound 2 is among the most
lipophilic BODIPYs while compound 8 is part of the hydrophilic set. In conclusion, we have
demonstrated the excellent potentiality of BODIPYs as green light activated photosensitizers.
Actually the meso aryl substituted BODIPYs, also featuring iodine atoms on the pyrrole 2,6
positions, are characterized by high rate of singlet oxygen production and an extensive tumor cells
penetration. Consequently, the photo-induced cell-killing efficacy observed in the in vitro studies is