Comparison between 5,10,15,20-tetraaryl- and 5,15-diarylporphyrins as photosensitizers: Synthesis, photodynamic activity, and quantitative structure-activity relationship modeling

Article abstract of DOI:10.1021/jm050997m

The synthesis of a panel of seven nonsymmetric 5,10,15,20- tetraarylporphyrins, 13 symmetric and nonsymmetric 5,15-diarylporphyrins, and one 5,15-diarylchlorin is described. In vitro photodynamic activities on HCT116 human colon adenocarcinoma cells were evaluated by standard cytotoxicity assays. A predictive quantitative structure-activity relationship (QSAR) regression model, based on theoretical holistic molecular descriptors, of a series of 34 tetrapyrrolic photosensitizers (PSs), including the 24 compounds synthesized in this work, was developed to describe the relationship between structural features and photodynamic activity. The present study demonstrates that structural features significantly influence the photodynamic activity of tetrapyrrolic derivatives: diaryl compounds were more active with respect to the tetraarylporphyrins, and among the diaryl derivatives, hydroxy-substituted compounds were more effective than the corresponding methoxy-substituted ones. Furthermore, three monoarylporphyrins, isolated as byproducts during diarylporphyrin synthesis, were considered for both photodynamic and QSAR studies; surprisingly they were found to be particularly active photosensitizers.

Full text of DOI:10.1021/jm050997m

J. Med. Chem. 2006, 49, 3293-3304
3293
Comparison between 5,10,15,20-Tetraaryl- and 5,15-Diarylporphyrins as Photosensitizers:
Synthesis, Photodynamic Activity, and Quantitative Structure-Activity Relationship Modeling
Stefano Banfi,* Enrico Caruso, Loredana Buccafurni, Roberto Murano, Elena Monti, Marzia Gariboldi, Ester Papa, and
Paola Gramatica
Department of Structural and Functional Biology, UniVersity of Insubria, Via H. Dunant 3, 21100 Varese (VA), Italy
ReceiVed October 6, 2005
The synthesis of a panel of seven nonsymmetric 5,10,15,20-tetraarylporphyrins, 13 symmetric and
nonsymmetric 5,15-diarylporphyrins, and one 5,15-diarylchlorin is described. In vitro photodynamic activities
on HCT116 human colon adenocarcinoma cells were evaluated by standard cytotoxicity assays. A predictive
quantitative structure-activity relationship (QSAR) regression model, based on theoretical holistic molecular
descriptors, of a series of 34 tetrapyrrolic photosensitizers (PSs), including the 24 compounds synthesized
in this work, was developed to describe the relationship between structural features and photodynamic activity.
The present study demonstrates that structural features significantly influence the photodynamic activity of
tetrapyrrolic derivatives: diaryl compounds were more active with respect to the tetraarylporphyrins, and
among the diaryl derivatives, hydroxy-substituted compounds were more effective than the corresponding
methoxy-substituted ones. Furthermore, three monoarylporphyrins, isolated as byproducts during diarylpor-
phyrin synthesis, were considered for both photodynamic and QSAR studies; surprisingly they were found
to be particularly active photosensitizers.
Introduction
ring are particularly attractive, as they combine some features
of the 5,10,15,20-tetraaryl derivatives (characterized by unsub-
stituted â-pyrrole positions) with those of the â-pyrrole octa-
alkylporphyrins [bearing a hydrogen atom at each 5,10,15,20
position (meso) of the tetrapyrrolic ring]. A large number of
structural modifications, including bromination, nitration, or
formylation, can be performed at the free 10,20-positions, which
makes 5,15-diarylporphyrins suitable for such diverse applica-
tions as light-harvesting antenna systems,3 liquid crystal por-
phyrins,4 and nonlinear optical devices.5 To date, only scattered
reports have been published on the potential use of 5,15-
diaryporphyrins or chlorins as PSs.6 Following up on the recent
development by our group of a series of tetrapyrrole derivatives
with potential use in medicine,7 the present study reports the
synthesis of a panel of novel tetraaryl- and diarylporphyrins
and their in vitro photodynamic activities against the cultured
human colon adenocarcinoma cell line HCT116. Quantitative
structure-activity relationship (QSAR) analysis was performed
on a larger set of PSs, including the newly synthesized
derivatives and the monoarylporphyrin byproducts, as well as
previously reported symmetric compounds.7
Photodynamic therapy (PDT) is a minimally invasive treat-
ment that uses the combination of a photosensitizing agent (PS)
and light to selectively target solid tumors, as well as several
nonneoplastic proliferating cell diseases. Systemic administration
of the PS is followed by localized irradiation with visible light;
in the presence of adequate concentrations of molecular oxygen,
PS photoactivation results in the formation of reactive oxygen
species (ROS) and related tissue damage.1,2
The first PS to be granted regulatory approval (in Canada,
1993) was porfimer sodium (Photofrin), a purified form of
hematoporphyrin derivative consisting of a mixture of porphy-
rins. Porfimer sodium is currently used in more than 40 countries
worldwide to treat a variety of cancers, including cancers of
the lung, stomach, cervix, and bladder and esophageal adeno-
carcinoma. However, despite its continuing effectiveness, por-
fimer sodium has a number of disadvantages, most notably low
tissue penetration of activating light (due to its weak absorption
at 630 nm), extended skin photosensitivity, and low selectivity
between tumor and healthy tissue in the early phases of
treatment, and this has spurred an active search for novel
(“second-generation”) PSs with improved properties over the
past 20 years. Most second-generation PSs (such as chlorins,
bacteriochlorins, and phthalocyanines) belong to the tetrapyrrole
class; while many agents have been identified with better
absorption at longer wavelengths than porfimer sodium, thereby
increasing the depth at which photodynamic cell kill can be
achieved, only one second-generation PS, the 5,10,15,20-
tetrakis(m-hydroxyphenyl)chlorin (m-THPC or temoporfin, cur-
rently marketed as Foscan) has been recently approved for the
palliative treatment of head and neck cancer, and the quest for
novel candidates is still ongoing.
Results and Discussion
A series of tetraaryl- and diarylporphyrins was synthesized,
based on the assumption that the presence of phenyls with
different substituents at the meso positions would affect the
hydrophobic/hydrophilic character of the resulting PSs. This is
a crucial feature for both cell penetration and subcellular
localization, and it is generally believed that amphiphilic
molecules, bearing both hydrophobic and hydrophilic moieties,
display improved tumor-selective uptake and retention, particu-
larly when polar and nonpolar groups are distributed nonsym-
metrically.8
Among porphyrins/chlorins, compounds bearing aromatic
substituents only at the 5 and 15 positions of the tetrapyrrolic
Chemistry. Nonsymmetric tetraarylporphyrins (1-7) (Figure
1) were synthesized via acid-catalyzed mixed condensation of
pyrrole with two different aromatic aldehydes, following the
general procedure described by Lindsey and co-workers.9 The
porphyrinogen intermediate was then oxidized to porphyrin with
* To whom correspondence should be addressed: phone +39-0332-
421550; fax +39-0332-421554; e-mail stefano.banfi@uninsubria.it.
10.1021/jm050997m CCC: $33.50 © 2006 American Chemical Society
Published on Web 04/29/2006

3294 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11
Banfi et al.
phyrins serendipitously isolated during diarylporphyrin synthesis
(22-24) (Figure 3) were also included in the in vitro study of
photodynamic activity performed in HCT116 cells; to the best
of our knowledge, only one paper has appeared to date assessing
monosubstituted tetrapyrroles as potential PSs.13
To confirm our previous results about the relationship
between the photodynamic activity of porphyrins and that of
the corresponding chlorin derivatives,7 we also synthesized the
chlorin 21, obtained from the corresponding diarylporphyrin 8
by the diimide reduction method.14
Photobleaching. It is known that tetrapyrrolic compounds
undergo partial demolition in the presence of oxidizing species,
and it is generally believed that bleaching can be mediated by
singlet oxygen,15a although some experiments suggest that
radical-mediated photodegradation (type I) predominates.15b To
select an appropriate irradiation time for cytotoxicity studies,
we assessed the rate of photodegradation for the new compounds
(Figure 4), based on the decreasing intensity of the Soret band.
Our main concern here was that diaryl derivatives, characterized
by the presence of unsubstituted meso positions that can undergo
fast oxidation, might be largely photodegraded during irradiation
of the cells; furthermore, these assays allowed us to compare
the degradation rates of methoxy- and hydroxy-substituted
porphyrins, which, in principle, could exhibit different behaviors
upon irradiation. However, no significant differences were
observed among the three classes of PSs, namely, tetra-, di-,
and monoarylporphyrins, as well as between hydroxy- and
methoxyphenyl-substituted porphyrins; for most compounds
approximately 60% of the intensity of the initial Soret band
was maintained following 2 h exposure to a 500 W tungsten-
halogen lamp. The stability of temoporfin and porfimer sodium
was also assessed under these conditions; both compounds
showed a comparable extent of photobleaching. It may be worth
emphasizing that, to allow the UV-vis spectral determination
at 400-420 nm, the PS concentrations used in these experiments
(5 × 10-5 M) were significantly higher than those used for PDT
on cell cultures (nanomolar range); thus, it is reasonable to
hypothesize that, in the low PS concentration range used for
PDT experiments, the extent of photobleaching may be even
lower.
Cytotoxicity Assays. The IC50 values from dose/response
curves obtained in HCT116 cells following exposure to the
different PSs for 24 h and irradiation with visible light for 2 h
are reported in Table 1; porfimer sodium and temoporfin were
also included as reference compounds. The intrinsic cytotoxicity
of PSs was assessed by omitting the irradiation step from the
treatment protocol and was found to be negligible in all cases
up to PS concentrations 10-fold higher than those used for PDT
experiments.
All the diaryl compounds tested in the present study were
significantly more effective than the corresponding tetraaryl
derivatives; in fact, the range of IC50 values obtained for
diarylporphyrins was 1.06-52.46 nM (median 9.84 nM) versus
53.31-1886.79 nM (median 344 nM) for tetraaryl derivatives.
The most effective diaryl compound tested 19 was also
significantly more phototoxic than both porfimer sodium and
temoporfin (IC50 values: compound 19, 0.51 ( 0.07 ng/mL;
porfimer sodium, 73.67 ( 8.04 ng/mL; temoporfin, 5.17 ( 0.39
ng/mL; concentrations are expressed in nanograms per milliliter
because an exact molecular weight cannot be calculated for
porfimer sodium, which is a mixture of oligomers).
Figure 1. Structures of tetraarylporphyrins 1-7.
2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ). The yields
of the recovered products range between 10% and 20%.
A number of different procedures have been reported for the
synthesis of 5,15-diarylporphyrins,10 mostly based on a (2 +
2)-type condensation in which two dipyrrolic compounds
incorporating one type of carbon bridge are fused together,
thereby forming the other type of carbon bridge; the intermediate
is then oxidized in situ to the final porphyrin. While in principle
there are many (2 + 2)-type synthetic pathways to obtain 5,15-
diarylporphyrins,10a the most frequently reported procedures rely
on condensation of 5-aryldipyrromethane with a second moiety
of 1,9-functionalyzed dipyrromethane.11 For the present study,
after a few preliminary attempts, we found that better overall
yields could be obtained when 5-unsubstituted dipyrromethane
was reacted with aromatic aldehydes, following the procedure
described by Plater et al.12 Accordingly, pyrrole was reacted
with thiophosgene to give the di-2-pyrrolylthione, which was
then oxidized with 30% hydrogen peroxide to the corresponding
ketone; the final reduction with NaBH4 produces the desired
dipyrromethane. Attempts to carry out direct hydrodesulfuriza-
tion of thione group with LiAlH4 or NaBH4 did not afford the
desired product in appreciable yields.
Acid-catalyzed condensation of dipyrromethane with desired
aromatic aldehydes afforded a series of 5,15-diarylporphyrins
(8-20), isolated with a variable yields from 10% to 25% (Figure
2).
Interestingly, under the experimental conditions adopted, in
a few peculiar cases the reaction between dipyrromethane and
aromatic aldehyde generated a side product that was recognized
as the monoarylporphyrin. To date, we have not been able to
establish the actual mechanism leading to the formation of this
product; most probably, methylene units provided by dipyr-
romethane allow the closure of the tetrapyrrolic ring with only
one aldehyde unit. For the sake of comparison, the monoarylpor-
Cytotoxicity data reported in Table 1 indicate that hydroxy-
substituted porphyrins are significantly more effective that the
corresponding methoxy derivatives, thus confirming our previ-

Tetra- and Diarylporphyrins as Photosensitizers
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11 3295
Figure 2. Structures of diarylporphyrins 8-20 and of chlorin 21.
bear a hydroxyl group in the meta position; in addition, Patrice
and co-workers6b recently reported on a promising diaryl PS
again bearing four hydroxyl groups (two for each phenyl ring).
As mentioned above, the different activities of hydroxyl and
methoxy derivatives cannot be accounted for by different
intrinsic cytotoxicities or different photodegradation rates. A
tentative explanation for the better photodynamic performance
of hydroxyl-substituted porphyrins could depend on their greater
solubility in the aqueous medium. Unfortunately, we were
unable to obtain experimental data to support this hypothesis,
as the octanol/water partition coefficients could not be deter-
mined for most of the agents in this study, due to their highly
lipophilic nature, yielding concentrations in the aqueous phase
that were below the limits of detection (even when the aqueous
phase was in 10-fold excess to the octanol layer). On the basis
of log P values calculated according to different computational
approaches (Table 2 in Supporting Information), a small
difference can be detected between the hydroxyl-substituted and
the corresponding methoxy-substituted porphyrins, as expected
from the effect of a few small polar groups on a large
hydrocarbon skeleton. The slightly more hydrophilic nature of
Figure 3. Structures of monoarylporphyrins 22-24.
ous findings.7 It may be worth mentioning that mono- or
disubstitution of the phenyl rings with hydroxyl groups generally
yields good photodynamic agents: in temoporfin, the most
active compound currently in clinical use, all four phenyl rings

3296 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11
Banfi et al.
Figure 4. Photobleaching of tetraarylporphyrins (A) and diaryl- and monoarylporphyrins (B).
hydroxy-substituted derivatives might concur with other, as yet
unidentified factors, to the observed difference in photodynamic
activity.
reduced form of chlorins,16 thus providing a further rationale
for the use of porphyrins instead of chlorins for in vitro screens
of photodynamic activity. Once the best lead compounds have
been identified, conversion to the corresponding chlorins will
yield active PSs more suitable for in vivo applications.
Finally, an interesting finding regards the phototoxicity of
the serendipitously isolated monoaryl derivatives (22-24), with
IC50 values ranging between 1.85 and 4.8 nM (median 3.2 nM);
further studies will specifically address the photodynamic
properties of this promising series of compounds.
QSAR Analysis. To the best of our knowledge, very few
QSAR analyses of photosensitizers for photodynamic therapy17-19
have been published. Two QSAR models17,18 are based on the
octanol/water partition coefficient and simple structural ele-
ments, such as length and shape of alkyl chains, while the more
recent paper19 applies theoretical molecular descriptors in
multiple linear regression (MLR) and artificial neural network
(ANN) models of only 12 pyropheophorbides. A previous 3D-
QSAR model, based on CoMFA and PLS, was published by
Debnath et al.20 for the modeling of 21 porphyrin derivatives,
including meso-unsubstituted and tetraaryl-substituted porphy-
rins, with anti-HIV-1 activity. The model had a low predictivity
(Q2 ) 0.59), verified just by leave-one-out cross-validation,
The fact that we decided to perform the present study on
porphyrins, despite the higher in vivo activity of chlorins,
deserves some comments. The better in vivo performance of
chlorins can be attributed to their absorbance profile at longer
wavelengths as compared to the corresponding porphyrins,
allowing photoactivation of the PS at greater depths within the
irradiated tissues. Significant differences in cellular uptake or
subcellular distribution between chlorins and the corresponding
porphyrins, which could result in different activities, have never
been reported. Accordingly, results obtained for compounds 8
and 21 (diphenylporphyrin and the corresponding chlorin
derivative) indicate a nonsignificant difference between the
phototoxic activities of the two compounds (5.66 ( 1.16 nM
for the porphyrin vs 3.26 ( 2.06 nM for the chlorin). These
data, together with previously reported observations by our
group,7 indicate that a good correlation exists between the
photodynamic effects of porphyrins and those of the corre-
sponding chlorins. In addition, the synthesis of porphyrins is
straightforward and gives better yields, due to the greater
stability of the totally unsaturated skeleton, as compared to the

Tetra- and Diarylporphyrins as Photosensitizers
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11 3297
Table 1. IC50 Valuesa for the Tested Compounds
the response values, log (1/IC50)] the available experimental set
of 34 chemicals (24 reported in this work and 10 in previous
work7) into 22 training chemicals for model development and
12 validation chemicals for model evaluation. The model with
the highest external predictivity (verified by Q2ext) was selected
in the GA population of models, each based on various
combinations of different descriptors, and it is proposed here
as the reference QSAR for log (1/IC50) calculations.
molecule
IC50, ng/mL ( SE
IC50, nM ( SE
7.60 ( 0.57
porfimer sodium
temoporfin
73.67 ( 8.04
5.17 ( 0.39
Tetraarylporphyrins
893.76 ( 44.73
99.80 ( 1.21
1
2
3
4
5
6
7
1386.97 ( 69.42
154.88 ( 1.88
1006.14 ( 194.73
>1886.79
708.52 ( 137.13
>1500
The best predictive three-variable MLR model has the
following equation and statistical parameters:
145.03 ( 24.52
216.87 ( 17.02
35.30 ( 7.57
230.06 ( 38.90
344.03 ( 27
53.31 ( 11.43
log 1/IC50 ) -10.06((3.89) + 21.05((2.31)GATS6v -
40.55((8.27)PW3 +36.15((13.71)R4u+ (1)
Diarylporphyrins
2.63 ( 0.54
27.39 ( 1.51
12.92 ( 0.97
6.32 ( 0.35
6.06 ( 0.53
2.93 ( 0.42
4.28 ( 1.30
9.03 ( 2.49
9.21 ( 2.69
3.51 ( 1.26
3.35 ( 0.10
0.51 ( 0.07
4.97 ( 0.38
1.51 ( 0.96
8
9
5.66 ( 1.16
52.46 ( 2.89
24.74 ( 1.85
9.84 ( 0.54
12.31 ( 1.07
5.94 ( 0.85
7.75 ( 2.36
15.51 ( 4.27
15.81 ( 4.62
7.09 ( 2.54
7.01 ( 0.21
1.06 ( 0.15
9.75 ( 0.74
3.26 ( 2.06
10
11
12
13
14
15
16
17
18
19
20
21
where ntraining ) 22, nvalidation ) 12, R2 ) 0.86, Q2 ) 0.81,
Q2boot ) 0.80, Q2ext ) 0.77, s ) 0.388, F ) 36.49, Kxx ) 18.9,
Kxy ) 40.8, RMSE ) 0.41, and RMSEP ) 0.46. The reported
fitting and validation parameters have high values, indicating
that the model is stable and has very good descriptive (R2) and
predictive performance (Q2). The quality of Q2 (0.77) and
ext
the small RMSE and RMSEP values (similar for training and
validation sets) confirm the robustness and good predictivity
of this model.
The model is dominated by the autocorrelation descriptor
GATS6v (Geary autocorrelation, of lag 6, weighted by the
atomic van der Waals volumes)25,26 (standardized regression
coefficient 0.83), which accounts for the correlation among
atoms, weighted by the van der Waals volumes, with a distance
of six bonds (the lag) in the molecule; GATS6v is a distance-
type function and gives mainly information on the molecular
size. The second most informative descriptor is the topological
descriptor PW3 (path/walk 3-Randic shape index)27 (standard-
ized regression coefficient -0.45) mainly accounting for mo-
lecular shape information, whereas the least relevant descriptor
is the GETAWAY descriptor28 R4u+ (standardized regression
coefficient 0.25), the maximal value of autocorrelation descriptor
(atoms individually weighted, topological distance of four bonds)
based on the influence/distance matrix, which takes into account
local structural aspects of the molecules. Thus, the size and
shape features, described by specific molecular descriptors in
their multivariate combination, are selected as the dominant
structural aspects related to the photodynamic activity of the
set of PSs examined in the present study.
Monoarylporphyrins
1.23 ( 0.29
22
23
24
3.2 ( 0.76
1.85 ( 0.71
4.8 ( 0.29
0.77 ( 0.29
2.29 ( 0.14
a Porphyrin concentrations inhibiting tumor cell growth by 50%. IC50
value are reported as means of three independent replications ( SE (standard
error).
probably due to a high degree of heterogeneity of the structures
of the molecules in the panel. In the present paper, 34 chemicals,
including the newly synthesized PSs described above along with
previously reported ones,7 were tested for their photodynamic
activity, to develop a validated MLR QSAR model, based on
theoretical molecular descriptors for this specific chemical
domain. In addition, the size of the available data set allows
statistical external validation of the model as well as the more
common internal validation. Assessing the predictivity of the
chemical applicability domain of the proposed model, that is,
the ability of the model to yield reliable data, is crucial for
application of the model to the design of novel, more effective
derivatives belonging to the verified chemical domain.
It is noteworthy that models developed on the basis of
calculated log P (A log P, M log P, HYPER-log P) as a single
descriptor were unpredictive as regards the whole set of 34
compounds. In particular, the sulfonamido compounds 31 and
32 fall totally outside the domain; however, even without
considering the two sulfonamido compounds, log P-based
models developed on the remaining 32 chemicals had the
following unsatisfactory statistical parameters: for A log P,
QSAR model development was performed by the procedure
described in the Experimental Section. A wide set of theoretical
molecular descriptors21 was used as input set of variables for
modeling, to identify different structural features of the PSs
related to the relevant effect (i.e., photodynamic activity).
Different kinds of calculated log P (A log P,21 M log P,21 and
HYPER-log P22) were also used as input variables during model
development.
R2 ) 0.57 and Q2 ) 0.51; for M log P, R2 ) 0.06 and
loo
Q2 ) -0.27; and for HYPER-log P, R2 ) 0.45 and Q2
)
As we cannot know a priori which descriptors could be related
to photodynamic activity and therefore useful in models for
prediction, we applied an evolutionary variable selection
procedure (genetic algorithms,23 GA-VSS) to select only the
best combinations of descriptors most relevant to obtaining
models with the highest predictive power for photodynamic
cytotoxicity.
loo
loo
0.38. The QSAR model proposed here highlights the ability of
the selected molecular descriptors to identify the structural
aspects related to the photodynamic activity, whereas models
based on log P alone are unable to capture this information of
this set of porphyrins. In addition, it is important to remember
that, in apparent contradiction to its widespread use, log P is
not a universal descriptor and its value is strongly variable
depending on the experimental procedure or the calculation
method applied.29-31 The GA population of models developed
by adding various log P to the previously selected theoretical
molecular descriptors (Table 2) have lower robustness and worse
In addition to different internal validation procedures (leave-
one-out, bootstrap, and Y-scrambling), for a stronger and more
reliable evaluation of model applicability for the prediction on
new chemicals,24 statistical external validation of the models
was also performed by randomly splitting [in connection with

3298 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11
Banfi et al.
Table 2. Models Developed Including Various Calculated Log P
Values in Addition to Theoretical Descriptors
reported by the authors, who found that 35 was as effective as
temoporfin 34 in their experimental model.6b [Note: While this
paper was under revision, we have synthesized the unprec-
edented monoarylporphyrin 37, which was among the novel
photosensitizers hypothesized whose activity was predicted by
the QSAR model. Experimental evaluation of the photodynamic
activity of compound 37 on HCT 116 cells yielded an IC50 value
(1.11 nM) very similar to the value predicted by the model (1.04
nM). This new result strengthens our feeling that the QSAR
model, obtained from the set of molecules indicated, has very
good predictivity and therefore could be applied to the screening
of large sets of molecules belonging to the mono-, di-, and
tetraaryl-substituted porphyrin class.]
2
2
variables
R2
Q2 Qboot Qext Kx, % Kxy, %
PW3, GATS6v, R4u+
0.86 0.81 0.80 0.77 18.90 40.77
PW3, GATS6v, HYPER-log P 0.81 0.77 0.68 0.52 32.83 49.30
PW3, GATS6v, A log P
PW3, GATS6v, M log P
GATS6v, R4u+, A log P
0.80 0.74 0.72 0.49 35.32 50.29
0.81 0.65 0.43 0.49 21.01 41.44
0.74 0.64 0.54 0.82 25.40 43.77
GATS6v, R4u+, HYPER-log P 0.72 0.49 0.42 0.64 31.71 39.36
GATS6v, R4u+, M log P 0.67 0.22 0.28 0.61 25.10 36.75
Principal component analysis (PCA) of the three molecular
descriptors of the proposed model was performed and is reported
in Figure 7, where the chemicals (the points) are plotted in the
descriptors space (the arrows for the loadings) of the two PCs
most highly correlated with the response (respectively PC2,
85.5%, and PC3, 32.6%).
The explained variance of the molecular descriptors in these
two PCs is 29.1% and 20.1%, respectively. PC1, while explain-
ing 50.8% of the molecular descriptor variance, is the least
correlated with the response (-7.8%). The chemicals are clearly
separated along PC2 (related inversely to PW3 and directly to
the other two descriptors) according to their degree of aryl
substitution and activity (Figure 7A). As PC2 and PC3 correlate
positively with the response, obviously with the corresponding
weight (PC2 about three times more important than PC3), the
chemicals on the left, and particularly at the bottom, have the
lowest log 1/IC50 values. These compounds are the least active
and belong to the series of tetraarylporphyrins; however, it is
important to note that compound 34 (temoporfin, already used
in clinical practice) is predicted as the most active tetraaryl-
substituted porphyrin. Mono- and diaryl derivatives, generally
more active than tetraarylporphyrins, are on the upper right side
of the graph. Moreover, when the diarylporphyrin panel is
considered, phenyl- or hydroxyphenyl-substituted compounds
(dotted ellipse) show higher activity than the majority of
methoxypheny- diarylporphyrins (solid circle), compound 20
being the only exception (as evident in the expanded portion
of panel A plotted in panel B). Notably, among all the PSs
examined in the present study, 20 is the only diaryl derivative
featuring hydroxyl groups on 3,4,5 positions of one phenyl ring,
and this structural difference with respect to the other hydroxy
diaryl compounds is well highlighted by the descriptor space
of the proposed model.
Figure 5. Regression line of the proposed QSAR model. The training
and validation chemicals are differently labeled and the “unknown”
chemicals are also numbered.
performance, mainly in external predictivity, than the model
proposed here (eq 1).
The regression line of the QSAR model presented above is
reported in Figure 5, where it is possible to observe the good
and balanced distribution of the validation set into the training
set, highlighting the efficacy of the splitting.
The analysis of the chemical applicability domain of the
model with the leverage approach32,33 allows us to verify the
presence of outliers for the response or chemicals influential
for some peculiarities in their structure. In general, all the
chemicals are well predicted (residuals within 2.5 standard
deviations) and are also within the structural applicability
domain (into the leverage value of hat). It is also important to
note that the validation chemicals, which were not used for
model development, are predicted with similar accuracy as the
training chemicals. The QSAR model was also used to predict
the activities of five as yet unknown porphyrins (37-41), of
the parent porphyrin tetraphenyl derivative (36) and of the
diarylchlorin 35, recently reported by the Patrice group6b
(Figure 6).
We have also verified by the leverage that these unknown
compounds are also within the structural chemical domain of
the model; thus their predicted data should be considered not
as extrapolated from the model but as reliable predictions.
This part of the study was performed with two aims: (a) to
verify the model predictivity on the well-known compound 36
and on porphyrin 35, independently described by other authors
as a very active photosensitizer, and (b) to obtain indications
on the biological activity of newly designed chemicals, to
properly address the synthesis of more effective derivatives. It
is interesting to note that the model predicts good photodynamic
activity for compound 35, a hydroxy-substituted diarylchlorin,
structurally similar to our most active compounds 18 and 19.
This prediction is confirmed by the in vivo and in vitro activity
In Figure 7A, the predicted chemicals 35-41 are also plotted
in the same molecular descriptor space. Application of the
QSAR model to novel hypothetical derivatives (37-41),
belonging to the chemical domain of the model, predicts a higher
photodynamic activity for monoaryl- (37-39) than for diaryl-
porphyrins (40, 41), suggesting that synthesis of this class of
compounds might be worth pursuing.
Conclusions
The results obtained in the present study suggest that
diaryltetrapyrrole compounds are more effective that the cor-
responding tetraaryl derivatives in inducing photodynamic cell
kill of human colon adenocarcinoma cells and that hydroxy-
substituted compounds are more active than methoxy-substituted
derivatives. The unintended isolation of three monoarylporphy-
rins as side products of the 5,15-diarylporphyrin synthesis
allowed us to investigate their phototoxicity on HCT116 cells
and to include them in the QSAR analysis: these molecules

Tetra- and Diarylporphyrins as Photosensitizers
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11 3299
Figure 6. Structures of the seven molecules included in the QSAR model as “unknown” and predicted.
recorded on a Bruker 400 MHz spectrometer in CDCl3 or deuterated
proved to be particularly promising photosensitizers, with IC50
values comparable to, or even better than, the most active
diarylporphyrins.
dimethyl sulfoxide (DMSO-d6); chemical shifts are expressed in
parts per million (ppm) relative to chloroform (7.28) and are
reported as s (singlet), d (doublet), t (triplet), m (multiplet), or br
s (broad singlet). Mass spectrometric measurements were performed
on a Finnigan LCQ-MS instrument. Elemental analyses were
performed on a ThermoQuest NA 2100 C, H, N analyzer, equipped
with an electronic mass flow control and thermal conductivity
detector. Analytical thin-layer chromatography (TLC) was per-
formed on Merck 60 F254 silica gel (precoated sheets, 0.2 mm
thick). Silica gel 60 (70-230 mesh, Merck) was used for column
chromatography. Aromatic aldehydes were commercial products
(Sigma-Aldrich) and were used as received. Pyrrole and BF3‚Et2O
were freshly distilled prior to use. Dichloromethane used for
porphyrin synthesis was distilled from CaCl2 directly into the
reaction flask.
Synthesis of Free Base Tetraarylporphyrins. 5,10,15-Triphen-
yl-20-(4-methoxyphenyl)-21H,23H-porphyrin 1, 5,10,15-triphenyl-
20-(3-methoxyphenyl)-21H,23H-porphyrin 2, 5,10,15-triphenyl-20-
(3,4,5-trimethoxyphenyl)-21H,23H-porphyrin 3, and 5,10,15-tri-(4-
methoxyphenyl)-20-(3,4,5-trimethoxyphenyl)-21H,23H-porphyrin 4
were synthesized via condensation of the corresponding aromatic
aldehydes and pyrrole under mixed-acid catalysis, as recently
reported by Lindsey and co-workers.9 The general procedure for
porphyrin synthesis is fully described for the first compound, the
others being prepared under similar conditions.
QSAR studies performed on a representative set of PSs
support this conclusion: a good predictive QSAR model,
statistically externally validated and based on three theoretical
molecular descriptors, was developed and allows for reliable
predictions about the activity of seven PSs that were not
experimentally tested in the present study, directing the synthesis
of future potential PSs toward diaryl- or monoaryltetrapyrrolic
compounds.
Interestingly, some of the new molecules (19, 21, 22, and
23) were found to be more active than the clinically approved
PSs porfimer sodium and temoporfin, at least in the in vitro
model adopted in this study. However, in vivo studies are
required to establish the actual therapeutic potential of these
PSs and to significantly compare their pharmacodynamic/
pharmacokinetic profiles with those of the agents already in
clinical use for PDT.
Experimental Section
Chemistry. UV-vis absorption spectra were measured on a
Perkin-Elmer Lambda 10 instrument. 1H NMR spectra were

3300 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11
Banfi et al.
Figure 7. (A) Loadings and score plots from principal component analysis (PCA: PC2-PC3) of the three molecular descriptors of the proposed
QSAR model for log (1/IC50). (B) Enlarged view of PCA from panel A only for diarylporphyrins: methoxy-, hydroxyl-, and nonsubstituted derivatives
are differently labeled.
5,10,15-Triphenyl-20-(4-methoxyphenyl)-21H,23H-porphy-
rin (1). BF3‚Et2O (12 µL, 1.2 × 10-2 mmol) and 0.35 mL (4.5
mmol) of trifluoroacetic acid (TFA) were added to a solution of
0.379 mL (3.75 mmol) of benzaldehyde, 0.152 mL (1.25 mmol)
of p-anisaldehyde, and 0.35 mL (5 mmol) of freshly distilled pyrrole
in 500 mL of CH2Cl2; the mixture was kept at room temperature
for 2 h. The reaction was controlled over time by TLC (SiO2;
hexane/CH2Cl2 ) 1/1). Then 570 mg (3.73 mmol) of DDQ was
added when the aldehydes were completely reacted, and the mixture
was kept at room temperature for 2 h. The solvent was evaporated
and the crude product was purified by column chromatography
(SiO2; CH2Cl2/hexane 6/4). Three different compounds were
separated: 5,10,15,20-tetraphenyl-21H,23H-porphyrin, 110 mg
(14.3%); 5,10-diphenyl-15,20-(4-methoxyphenyl)-21H,23H-porphy-
rin, 55 mg (6.4%); and the desired product 5,10,15-triphenyl-20-
(4-methoxyphenyl)-21H,23H-porphyrin, 140 mg (17.4%); Rf 0.63
(SiO2; CH2Cl2/hexane ) 6/4); MS-ESI+ m/z 645.4 (M + 1) (100%).
Anal. Calcd for (C45H32N4O): C, H, N.
5,10,15-Triphenyl-20-(3-methoxyphenyl)-21H,23H-porphy-
rin (2). Compound 2 was synthesized as described above. Three
different compounds were separated: 5,10,15,20-tetraphenyl-21H,-
23H-porphyrin, 110 mg (14.3%); 5,10-diphenyl-15,20-(3-methoxy-
phenyl)-21H,23H-porphyrin, 150 mg (17.5%); and the desired
product 5,10,15-triphenyl-20-(4-methoxyphenyl)-21H,23H-porphy-
rin, 125 mg (19%); Rf 0.35 (SiO2; CH2Cl2/hexane ) 1/1); MS-
ESI+ m/z 645.4 (M + 1) (100%). Anal. Calcd for (C45H32N4O):
C, H, N.
5,10,15-Triphenyl-20-(3,4,5-trimethoxyphenyl)-21H,23H-por-
phyrin (3). 3,4,5-Trimethoxybenzaldehyde (245.5 mg, 1.25 mmol),
0.380 mL (3.75 mmol) of benzaldehyde, and 0.35 mL (5 mmol) of
pyrrole were reacted as described for compound 1. The crude
material was purified by means of two subsequent column chro-

Tetra- and Diarylporphyrins as Photosensitizers
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11 3301
matographies (SiO2; CH2Cl2/hexane ) 1/1). The total amount of
3, isolated as pure product, was 150 mg (0.21 mmol, 16.8%); Rf
0.32 (SiO2; CH2Cl2/hexane ) 1/1); MS-ESI+ m/z 705.2 (M + 1)
(100%). Anal. Calcd for (C47H36N4O3): C, H, N. In addition to 3,
380 mg (0.62 mmol, 49%) of 5,10,15,20-tetraphenyl-21H,23H-
porphyrin was also recovered from the reaction.
ered: the first was 5-mono-(3,4,5-trimethoxyphenyl)-21H,23H-
porphyrin 24 (184 mg, 20%), and the second was the desired
product (176 mg, 11%).
5,15-Di-(3,4,5-trimethoxyphenyl)-21H,23H-porphyrin 11: Rf
0.54 (SiO2; CH2Cl2/hexane/Et2O ) 20/3/2); MS-ESI+ m/z 643.2
(M + 1) (100%). Anal. Calcd for (C38H34N4O6): C, H, N.
5-Mono-(3,4,5-trimethoxyphenyl)-21H,23H-porphyrin 24: Rf
0.75 (SiO2; CH2Cl2/hexane/Et2O ) 20/3/2); MS-ESI+ m/z 477.2
(M + 1) (100%). Anal. Calcd for (C29H24N4O3): C, H, N.
5-Phenyl-15-(4-methoxyphenyl)-21H,23H-porphyrin (12). TFA
(0.215 mL, 2.18 mmol) and 5.4 µL of BF3‚Et2O were added to a
500 mL of CH2Cl2 solution containing 0.173 mL (1.71 mmol) of
benzaldehyde, 0.208 mL (1.71 mmol) of p-anisaldehyde, and 500
mg (3.42 mmol) of 2,2′-dipyrromethane. The oxidation was obtained
with 780 mg (3.59 mmol) of DDQ. Compound 12 (80 mg, 16%)
was obtained after two subsequent chromatographies (SiO2; CH2-
Cl2 and CH2Cl2/hexane ) 6:4): Rf 0.52 (SiO2; CH2Cl2/hexane )
6/4); MS-ESI+ m/z 493.3 (M + 1) (100%). Anal. Calcd for (C33-
H24N4O): C, H, N. In addition to the desired product, 55 mg (8.3%)
of 5,15-diphenyl-21H,23H-porphyrin 8 was also isolated as second-
ary product.
5-Phenyl-15-(3-methoxyphenyl)-21H,23H-porphyrin (13). Com-
pound 13 was synthesized from m-anisaldehyde and benzaldehyde
as described in the case of compound 12. The isolated pure product
was 100 mg (20% yield): Rf 0.55 (SiO2; CH2Cl2/hexane ) 6/4);
MS-ESI+ m/z 493.3 (M + 1) (100%). Anal. Calcd for (C33H24N4-
O): C, H, N. In this reaction, 80 mg (12.1%) of 5,15-diphenyl-
21H,23H-porphyrin was also isolated.
5-Phenyl-15-(3,4,5-trimethoxyphenyl)-21H,23H-porphyrin (14).
The title compound was synthesized from 0.140 mL (1.36 mmol)
of benzaldehyde and 266 mg (1.36 mmol) of 3,4,5-trimethoxy-
benzaldehyde as described above and was isolated in 4% yield (38
mg) after two column chromatographies: Rf 0.35 (SiO2; CH2Cl2/
hexane ) 6/4); MS-ESI+ m/z 553.2 (M + 1) (100%). Anal. Calcd
for (C35H28N4O3): C, H, N.
5-(3,4,5-trimethoxyphenyl)-15-(4-methoxyphenyl)-21H,23H-
porphyrin (15). 4-Methoxybenzaldehyde (0.270 mL, 2.22 mmol),
3,4,5-trimethoxybenzaldehyde (436.1 mg, 2.22 mmol), 650 mg (4.44
mmol) of 2,2′-dipyrromethane, 0.279 mL of TFA, and 7 µL (0.056
mmol) of BF3‚Et2O were reacted as described for compound 8.
The crude material was purified by column chromatography (SiO2;
CH2Cl2) affording 200 mg (14%) of pure product 15: Rf 0.44 (SiO2;
CH2Cl2); MS-ESI+ m/z 583.3 (M + 1) (100%). Anal. Calcd for
(C35H30N4O4): C, H, N. In addition to compound 15, 90 mg (9.7%)
of porphyrin 9 was also isolated from the mixture.
5-(3,4,5-Trimethoxyphenyl)-15-(3-methoxyphenyl)-21H,23H-
porphyrin (16). Compound 16 was synthesized as described for
compound 15, and the isolated pure product yield was 65 mg
(11%): Rf 0.46 (SiO2; CH2Cl2); MS-ESI+ m/z 583.3 (M + 1)
(100%). Anal. Calcd for (C35H30N4O4): C, H, N. From the reaction
mixture, 15 mg (1.6%) of compound 10 was also recovered.
Synthesis of Hydroxy-Substituted Porphyrins from Meth-
oxyporphyrins. Porphyrins bearing hydroxyls as substituents on
meso-phenyls were obtained by demethylation of corresponding
methoxy groups, following the BBr3 general method.34
5,10,15-Triphenyl-20-(4-hydroxyphenyl)-21H,23H-porphy-
rin (5). A solution of 80 mg (0.124 mmol) of porphyrin 1 in 15
mL of CH2Cl2 was stirred at 0 °C for 15 min, and then 2.48 mL
(2.48 mmol) of 1 M BBr3 dichloromethane solution was added.
The mixture was kept at 0 °C for 1 h and at room temperature for
18 h; after this period, 30 mL of CH2Cl2 was added to the reaction
flask together with the desired amount of Na2CO3 saturated solution
to neutralize the mixture. The layers were separated and the organic
phase was thoroughly washed with water, dried (Na2SO4), and
concentrated to dryness, yielding 5 as solid pure product after
crystallization (65 mg, 83%): Rf 0.38 (SiO2; CH2Cl2); MS-ESI+
m/z 631.4 (M + 1) (100%). Anal. Calcd for (C44H30N4O): C, H,
N.
5,10,15-Tri-(4-methoxyphenyl)-20-(3,4,5-trimethoxyphenyl)-
21H,23H-porphyrin (4). 3,4,5-Trimethoxybenzaldehyde (245.5 mg,
1.25 mmol), 0.426 mL (3.75 mmol) of p-anisaldehyde, and 0.35
mL (5 mmol) of pyrrole were reacted as described for compound
1. The crude product was purified by means of two subsequent
column chromatographies (SiO2; CH2Cl2); 100 mg (14%) of 3 was
isolated as pure product; Rf 0.3 (SiO2; CH2Cl2; MS-ESI+ m/z 795.2
(M + 1) (100%). Anal. Calcd for (C50H42N4O6): C, H, N. From
this reaction was also isolated 95 mg (10.2%) of 5,10,15,20-tetra-
(4-methoxyphenyl)-21H,23H-porphyrin as secondary product.
Synthesis of Free Base Diarylporphyrins. The general proce-
dure for porphyrin syntheses is fully described for the first
compound, the others being prepared under similar conditions.
5,15-Diphenyl-21H,23H-porphyrin (8). TFA (0.168 mL, 2.18
mmol) was added to a solution of 0.275 mL (2.72 mmol) of
benzaldehyde and 398 mg (2.72 mmol) of 2,2′-dipyrromethane in
400 mL of CH2Cl2; the mixture was kept at room temperature for
2 h. The disappearance of the aldehyde from the reaction mixture
was determined by TLC (SiO2; CH2Cl2); then 650 mg (2.86 mmol)
of DDQ was added and the mixture was kept at room temperature
for 2 h. The solvent was evaporated and the crude product was
purified by column chromatography (SiO2; hexane/CH2Cl2 ) 1/1).
Two fractions were collected: the first was 5-monophenyl-21H,-
23H-porphyrin 22 (14 mg, 2.6%), and the second was the desired
product.
5,15-Diphenyl-21H,23H-porphyrin 8: 158 mg (25%); Rf 0.56
(SiO2; hexane/CH2Cl2 ) 1/1); MS-ESI+ m/z 463.1 (M + 1) (100%).
Anal. Calcd for (C32H22N4): C, H, N.
5-Monophenyl-21H,23H-porphyrin 22: Rf 0.76 (SiO2; hexane/
CH2Cl2 1/1); MS-ESI+ m/z 387.2 (M + 1) (100%). Anal. Calcd
for (C26H18N4): C, H, N.
5,15-Di-(4-methoxyphenyl)-21H,23H-porphyrin (9). To a solu-
tion of 0.242 mL (2.12 mmol) of p-anisaldehyde and 310 mg (2.12
mmol) of 2,2′-dipyrromethane in 500 mL of CH2Cl2 was added
0.038 mL (0.30 mmol) of BF3‚Et2O; the mixture was then treated
with 502 mg (2.22 mmol) of DDQ and worked up as described
above. The crude product was purified by column chromatography
(SiO2; CH2Cl2/hexane/AcOEt ) 20/3/2). The product was further
purified with a second column chromatography (SiO2; CH2Cl2/
hexane ) 6/4). Two fractions were recovered: the first was 5-mono-
(4-methoxyphenyl)-21H,23H-porphyrin 23 (16 mg, 3.6%), and the
second was the desired product (45 mg, 8%).
5,15-Di-(4-methoxyphenyl)-21H,23H-porphyrin 9: Rf 0.48
(SiO2; hexane/CH2Cl2 ) 6/4); MS-ESI+ m/z 523.1 (M + 1) (100%).
Anal. Calcd for (C34H26N4O2): C, H, N.
5-Mono-(4-methoxyphenyl)-21H,23H-porphyrin 23: Rf 0.69
(SiO2; hexane/CH2Cl2 ) 6/4); MS-ESI+ m/z 417 (M + 1) (100%).
Anal. Calcd for (C27H20N4O): C, H, N.
5,15-Di-(3-methoxyphenyl)-21H,23H-porphyrin (10). 3-Meth-
oxybenzaldehyde (0.330 mL, 2.72 mmol), 398.5 mg (2.72 mmol)
of 2,2′-dipyrromethane, and 0.168 mL of TFA were reacted as
described for compound 8. The crude product was purified by
column chromatography (SiO2; CH2Cl2) and the recovered product
was crystallized by dissolving the solid in the minimum amount of
dichloromethane and then settling as a precipitate by careful addition
of hexane. Compound 10 (145 mg, 20%) was isolated as pure
product: Rf 0.72 (SiO2; CH2Cl2); MS-ESI+ m/z 523.1 (M + 1)
(100%). Anal. Calcd for (C34H26N4O2): C, H, N.
5,15-Di-(3,4,5-trimethoxyphenyl)-21H,23H-porphyrin (11). A
solution of 980 mg (5 mmol) of 3,4,5-trimethoxybenzaldehyde and
750 mg (5 mmol) of 2,2′-dipyrromethane in 500 mL of CH2Cl2
were treated with 0.094 mL (0.75 mmol) of BF3‚Et2O; the addition
of DDQ (1186 mg 5.25 mmol) allows the oxidation of porphy-
rinogen. The crude product was purified by column chromatography
(SiO2; CH2Cl2/hexane/Et2O ) 20/3/2). Two fractions were recov-
5,10,15-Triphenyl-20-(3-hydroxyphenyl)-21H,23H-porphy-
rin (6). A solution of 70 mg (0.108 mmol) of porphyrin 2 in 20
mL of CH2Cl2 was treated with 2.16 mL (2.16 mmol) of 1 M BBr3

3302 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11
Banfi et al.
solution. The isolated pure product yield was 47 mg (69%): Rf
0.53 (SiO2; CH2Cl2); MS-ESI+ m/z 631.4 (M + 1) (100%). Anal.
Calcd for (C44H30N4O): C, H, N.
5,10,15-Triphenyl-20-(3,4,5-hydroxyphenyl)-21H,23H-por-
phyrin (7). A solution of 40 mg (0.056 mmol) of porphyrin 4 in
25 mL of CH2Cl2 was treated with 2.0 mL (2.0 mmol) of BBr3
solution. The isolated pure product yield was 30 mg (75%): Rf
0.21 (SiO2; CH2Cl2); MS-ESI+ m/z 663.2 (M + 1) (100%). Anal.
Calcd for (C44H30N4O3): C, H, N.
5,15-Di-(3-hydroxyphenyl)-21H,23H-porphyrin (17). BBr3 1
M dichloromethane solution (1.57 mL, 1.57 mmol) was added to
a solution of 41 mg (0.078 mmol) of porphyrin 10 in 10 mL of
CH2Cl2. After crystallization, 18 mg (46%) of 17 was isolated: Rf
0.32 (SiO2; CH2Cl2); MS-ESI+ m/z 495.2 (100%). Anal. Calcd for
(C32H22N4O2): C, H, N.
5-Phenyl-15-(4-hydroxyphenyl)-21H,23H-porphyrin (18). A
solution of 50 mg (0.101 mmol) of porphyrin 12 in 5 mL of CH2-
Cl2 was treated with 3.03 mL (3.03 mmol) of BBr3 solution, and
30 mg (62.5%) of pure compound 18 was isolated: Rf 0.37 (SiO2;
CH2Cl2); MS-ESI+ m/z 479.1 (M + 1) (100%). Anal. Calcd for
(C32H22N4O): C, H, N.
5-Phenyl-15-(3-hydroxyphenyl)-21H,23H-porphyrin (19). A
solution of 60 mg (0.12 mmol) of porphyrin 13 in 15 mL of CH2-
Cl2 was treated with 2.5 mL (2.5 mmol) of BBr3 solution, and 36
mg (62%) of the desired product was recovered: Rf 0.38 (SiO2;
CH2Cl2); MS-ESI+ m/z 479.1 (M + 1) (100%). Anal. Calcd for
(C32H22N4O): C, H, N.
5-Phenyl-15-(3,4,5-trihydroxyphenyl)-21H,23H-porphyrin (20).
A solution of 20 mg (0.036 mmol) of porphyrin 14 in 25 mL of
CH2Cl2 was treated with 2 mL (2 mmol) of BBr3 solution. The
isolated pure product was 15 mg, corresponding to 81% yield: Rf
0.26 (SiO2; CH2Cl2); MS-ESI+ m/z 511.4 (M + 1) (100%). Anal.
Calcd for (C32H22N4O3): C, H, N.
Synthesis of Chlorin from Porphyrin: 5,10-Diphenyl-21H,-
23H-chlorin (21). Toluene-4-sulfonylhydrazide (128 mg, 0.688
mmol) and 475.4 mg (3.44 mmol) of K2CO3 were added to a
solution of 40 mg (0.086 mmol) of 5,15-diphenyl-21H,23H-
porphyrin 8 in 6 mL of pyridine, and then the mixture was refluxed
for 2 h; the reaction progress was monitored by means of UV-vis
spectroscopy, based on the occurrence of the chlorin absorption
band at 650 nm. Equivalent amounts of hydrazide and K2CO3 were
added every hour until the reaction was completed (determined by
evaluating the ratio of the intensity between the band at 400 nm
and that at 650 nm; this ratio must be equal or lower than 6). At
the end of the reaction, 10 mL of AcOEt and 20 mL of H2O were
added and the mixture was kept at 70 °C for 1 h. Afterward the
organic layer was separated, washed a few times with 10% HCl
and then with H2O, and finally dried (Na2SO4). Solvent evaporation
under vacuum afforded chlorin 21 as a solid product (24 mg, 59%).
Once more the absorption spectrum of the isolated compound was
controlled; the presence of the peak at 730 nm indicates the
formation of some overreduction product (bacteriochlorin); the
crude solid was then dissolved in CH2Cl2 and treated with a diluted
toluene solution of chloranyl. This procedure must be carried out
with particular attention to avoid oxidation of the chlorin back to
the initial porphyrin. As the final step of the reaction, the solution
was concentrated and purified by column chromatography (SiO2;
hexane/CH2Cl2 ) 1/1) to give 15 mg (37%) of the pure product:
Rf 0.54 (SiO2; hexane/CH2Cl2 ) 1/1); MS-ESI+ m/z 465.1 (M +
1) (100%). Anal. Calcd for (C32H22N4): C, H, N.
fetal bovine serum (Mascia-Brunelli) at 37 °C in a humidified 5%
CO2 atmosphere. The antiproliferative effect of the different PS
was assessed by the model transformation tools (MTT) assay.35
Briefly, 5 × 104 cells/mL were seeded onto 96-well plates and
allowed to grow for 48 h prior to treatment with different PS
concentrations. After 24 h, the PS-containing medium was replaced
by PBS, and cells were irradiated under visible light (tungsten-
halogen lamp 500 W) for 2 h (average value between 380 and 780
nm determined with a Licor-1800 spectroradiometer; light irradiance
22 mW/cm2 and a light energy of 158.4 J/cm2). At the end of this
time, cells were incubated for 24 h at 37 °C in drug-free medium;
MTT was then added to each well (final concentration 0.4 mg/
mL) for 3 h at 37 °C and formazan crystals formed through MTT
metabolism by viable cells were dissolved in dimethyl sulfoxide
(DMSO). Optical densities were measured at 570 nm by use of a
Universal Microplate Reader EL800 (Bio-Tek Instruments).
Possible intrinsic (i.e., nonphotodynamic) cytotoxic effects of
the PS were assessed on control cells treated as described above,
with PS concentrations up to 10-fold higher than those used for
PDT experiments, omitting cell irradiation.
IC50 (i.e., the concentration affecting 50% of cells) values were
obtained by nonlinear regression analysis, using the GraphPad
PRISM 3.03 software (GraphPad Software Inc., San Diego, CA)
QSAR: Experimental Data Set. The IC50 values, expressed in
nanomolar concentrations, used to develop the QSAR model were
obtained from this work (24 data) and from previously reported
results (10 data).7 The final set consists of IC50 values of 34 PSs
expressed as log 1/IC50.
In Supporting Information are reported experimental and pre-
dicted data (Table 3) and the structures of the 10 photosensitizerss
previously synthesized (Figure 1).
Molecular Descriptors. The geometries of the 41 studied
porphyrins were fully optimized without constraints by use of the
semiempirical quantum-mechanical PM3 Hamiltonian as imple-
mented in HYPERCHEM22 as molecular modeling software. The
three-dimensional structures of minimum energy conformation were
used as input for the DRAGON software,21 which was employed
to numerically encode the topology and geometry of these molecules
by theoretical molecular descriptors. A total of about 1200 molecular
descriptors of different kinds (mono-, bi-, and three-dimensional)
were calculated to describe compound chemical diversity.
To compare the modeling power of variables, traditionally used
to model biological end points, with the above-listed theoretical
molecular descriptors, different kinds of log P (calculated values
for A log P,21 M log P,21 and HYPER-log P22) were used. Quantum-
chemical descriptors such as HOMO (highest occupied molecular
orbital), LUMO (lowest unoccupied molecular orbital), HOMO-
LUMO gap (DHL), the ionization potential (Pion) and the heat of
formation (H), calculated by the semiempirical PM3 Hamiltonian
for the geometry optimization method available in the HYPER-
CHEM package,22 were always added and used for descriptor
selection during model development.
Constant values and descriptors found to be correlated pairwise
were excluded in a prereduction step (when there was more than
98% pairwise correlation, one variable was deleted), and the genetic
algorithm was applied for variable selection to a final set of about
270 descriptors.
The values of the selected molecular descriptors are reported in
Table 3 of Supporting Information.
Chemometric Methods. Multiple linear regression (MLR) and
genetic algorithm--variable subset selection (GA-VSS)23 were
performed by the software MOBY DIGS of Todeschini et al.36 using
the ordinary least-squares regression (OLS) method, optimizing the
Photobleaching Measurements. About 0.2 mL of DMSO
porphyrin mother solution was diluted to 15 mL with 0.1 M
phosphate-buffered saline (PBS) in order to obtain a 5 × 10-5
M
porphyrin concentration. This solution was exposed at 37 °C to a
500 W tungsten-halogen lamp fitted with aqueous filter for 2 h..
Every 15 min, 0.4 mL samples were withdrawn and diluted with
1.6 mL of PBS, and their UV-vis absorption was measured.
Cytotoxicity Studies. Human adenocarcinoma HCT116 cells
were obtained from the American Type Culture Collection (Rock-
ville, MD) and maintained in Dulbecco’s modified Eagle medium
(DMEM; Mascia-Brunelli, Milano, Italy) supplemented with 10%
prediction power Q2 (leave-one-out procedure). This algorithm
loo
provides the subsets of the most predictive molecular descriptors
for the selected property, automatically chosen among all the
available descriptors.
To avoid multicollinearity with or without “apparent” prediction
power (chance correlation), the regressions were calculated only
for variable subsets with an acceptable multivariate correlation with
response, by applying the QUIK rule (Q under influence of K),37

Tetra- and Diarylporphyrins as Photosensitizers
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11 3303
which does not consider models with a K multivariate correlation
index of the [X] variable block greater than the correlation within
the [X + y] block variables, where X is the molecular descriptors
and y is the response variable.
In the MLR equation of the model, reported in this paper, the
variables are listed in order of relative importance by their
standardized regression coefficients. In fact, since molecular
descriptors do not have equal variance (i.e., they are not autoscaled),
their relative importance in the model is measured better by
standardized regression coefficients (i.e., the coefficients multiplied
by the standard deviation of the corresponding predictor). The errors
of the regression coefficients have also been reported for each
equation.
1-24; structures of the 10 photosensitizers previously synthesized;7
list of the calculated log P values (A log P, M log P, and HYPER-
log P), molecular descriptors, and experimental and predicted log
(1/IC50) (IC50 in nanomolar). This material is available free of charge
via the Internet at http://pubs.acs.org.
References
(1) Johnson, S.; Johnson F. N. ReViews in Contemporary Pharmaco-
therapy-Photodynamic Therapy; Marius Press: Carnforth, U.K.,
1999.
(2) Macdonald, I. J.; Dougherty, T. J. Basic principles of photodynamic
therapy. J. Porphyrins Phthalocyanines 2001, 5 (2), 105-129.
(3) Lin, V. S.; Therien, M. J. The Role of Steric and Electronic Effects
in the Extensive Modulation of the Absorptive and Emissive
Properties of a Series of Ethynyl- and Butadiynyl-Bridged Bis- and
Tris-Porphinato(zinc) Chromophores. Chem. Eur. J. 1995, 1, 645-
651.
(4) Wang, Q. M.; Bruce, D. W. Low-Melting, Liquid-Crystalline
Metalloporphyrins. Angew. Chem., Int. Ed. Engl. 1997, 36, 150-
152.
(5) Osuka, A.; Shimidzu, H. meso,meso-Linked Porphyrin Arrays. Angew.
Chem., Int Ed. Engl. 1997, 36, 135-137.
(6) (a) Macalpine, J. K.; Boch, R.; Dolphin, D. Evaluation of tetraphenyl-
2,3-dihydroxychlorins as potential photosensitizers. J. Porphyrins
Phthalocyanines 2002, 6, 146-155. (b) Bourre´, L.; Simonneaux, G.;
Ferrand, Y.; Thibaut, S.; Lajat, Y.; Patrice, T. Synthesis, and in vitro
and in vivo evaluation of a diphenylchlorin sensitizer for photody-
namic therapy. J. Photochem. Photobiol. B: Biol. 2003, 69, 179-
192.
(7) Banfi, S.; Caruso, E.; Caprioli, S.; Mazzagatti, L.; Canti, G.; Ravizza,
R.; Gariboldi, M.; Monti, E. Photodynamic effects of porphyrin and
chlorin photosensitizers in human colon adenocarcinoma cells.
Bioorg. Med. Chem. 2004, 12, 4853-4860.
(8) Osterloh, J.; Vicente, M. G. H. Mechanisms of porphyrinoid
localization in tumors. J. Porphyrins Phthalocyanine 2002, 6 (5),
305-324.
(9) Chevalier, F.; Geier, G. R., III; Lindsey, J. S. Acidolysis of
intermediates used in the preparation of core-modified porphyrinic
macrocycles. J. Porphyrins Phthalocyanines 2002, 6 (3), 186-197.
(10) (a) Bruckner, C.; Posakony, J. J.; Johnson, C. K.; Boyle, R. W.; James,
B. R.; Dolphin, D. Novel and improved synthesis of 5,15-diphenyl-
porphyrin and its dipyrrolic precursors. J. Porphyrins Phthalocyanines
1998, 2 (6), 455-465. (b) Clarke, O. J.; Boyle, R. W. Selective
Synthesis of Asymmetrically Substituted 5,15-Diphenylporphyrins.
Tetrahedron Lett. 1998, 39, 7167-7168.
(11) (a) Tamaru, S.; Yu, L.; Youngblood, J.; Muthukumaran, K.; Tan-
iguchi, M.; Lindsey, J. S. A tin-complexation strategy for use with
diverse acylation methods in the preparation of 1,9-diacyldipyr-
romethanes. J. Org. Chem. 2004, 69, 765-777. (b) Arumagam, N.;
Won, D.; Lee, C. Synthesis of regioselectively functionalized,
asymmetric porphyrins via “2+2” condensation. J. Porphyrins
Phthalocyanines 2002, 6 (7-8), 479-483.
(12) Plater, M. J.; Aiken, S.; Bourhill, G. A new synthetic route to donor-
acceptor porphyrins. Tetrahedron 2002, 58, 2405-2413.
(13) Pandey, S. K.; Gryshuk, A. L.; Graham, A.; Ohkubo, K.; Fukuzumi,
S.; Dobhal, M. P.; Zheng, G.; Ou, Z.; Zhan, R.; Kadish, K. M.;
Oseroff, A.; Ramaprasad, S.; Pandey, R. K. Fluorinated photosen-
sitizers: synthesis, photophysical, electrochemical, intracellular
localization, in vitro photosensitizing efficacy and determination of
tumor-uptake by 19F in vivo NMR spectroscopy. Tetrahedron 2003,
59, 10059-10073.
(14) Bonnett, R.; White, R. D.; Winfield, U. J.; Berenbaum, M. C.
Hydroporphyrins of the meso-tetra(hydroxyphenyl)porphyrin series
as tumor photosensitizers. Biochem. J. 1989, 261, 277-280.
(15) (a) Bonnett, R.; Martinez, G. Photobleaching of sensitisers used in
photodynamic therapy. Tetrahedron 2001, 57, 9513-9547. (b)
Georgakoudi, I.; Foster, T. H. Singlet Oxygen- Versus Nonsinglet
Oxygen-Mediated Mechanisms of Sensitizer Photobleaching and
Their Effects on Photodynamic Dosimetry. Photochem. Photobiol.
1998, 67, 612-625.
Validation. The robustness of the models and their predictivity
were evaluated by both Q2 and bootstrap. In this last procedure
loo
K n-dimensional groups are generated by a randomly repeated
selection of n objects from the original data set. The model obtained
on the first selected objects is used to predict the values for the
excluded sample and then Q2 is calculated for each model. The
bootstrapping was repeated 5000 times for each validated model.
The proposed models are also checked for reliability and
robustness by permutation testing;32 new models are recalculated
for randomly reordered response (Y scrambling) in order to exclude
by chance models.
The actual predictive capability of each model developed on the
training set is verified on an external validation set24 and is
calculated from Q2ext ) 1 - PRESS/SD, where PRESS is the sum
of squared differences between the measured response and the
predicted value for each molecule in the validation set, and SD is
the sum of squared deviations between the measured response for
each molecule in the validation set and the mean measured value
of the training set. A measure to define the accuracy of the proposed
QSARs is also the RMSE (root mean square of errors) 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 (RMSE and RMSEP, calculated separately
for the training and the test/validation sets):
(yi - yˆi)2
∑
i
RMSE )
x
n
Splitting Training/Test for External Validation. To have
compounds for external validation, the original data set of 34
compounds was split into a training set and an external validation
set. The splitting of the data set was realized considering the
distribution of the response value: 2/3 of the studied compounds
(22 objects) were selected as a training set and used for the model
developing, and the remaining 12 molecules were used as a
validation set, to allow the external statistical validation of the
model.
Chemical Domain. QSAR models must always be verified for
their applicability with regard to chemical domain. The presence
of outliers (i.e., compounds with cross-validated standardized
residuals greater than 2.5 standard deviation units) and chemicals
very structurally influential in determining model parameters [i.e.,
compounds with high leVerage value (h)33 greater than 3 p′/n (h*),
where p′ is the number of model variables plus one, and n is the
number of the objects used to calculate the model] was also verified.
The reliability of the predicted data with regard to chemical domain
was verified by the leVerage approach: the predictions for chemicals
of the validation set with a leverage value smaller than h* must be
considered reliable, being into the structural chemical domain of
the training set.
(16) Bonnett, R.; Martinez, G. Photobleaching of compounds of the
5,10,15,20-tetrakis(m-hydroxyphenyl)-porphyrin series (m-THPP, m-
THPC, and m-THPBC). Org. Lett. 2002, 4 (12), 2013-2016.
(17) Henderson, B. W.; Bellnier, D. A.; Greco, W. R.; Sharma, A.; Pandey,
R. K.; Vaughan, L. A.; Weishaupt, K. R.; Dougherty, T. J. An in
vivo quantitative structure-activity relationship for a congeneric
series of pyropheophorbide derivatives as photosensitizers for pho-
todynamic therapy. Cancer Res. 1997, 57, 4000-4007.
(18) Potter, W. R.; Henderson, B. W.; Bellnier, D. A.; Pandey, R. K.;
Vaughan, L. A.; Weishaupt, K. R.; Dougherty, T. J. Parabolic
quantitative structure-activity relationships and photodynamic
therapy: Application of a three-compartment model with clearance
Principal component analysis (PCA) for data exploration was
performed on autoscaled data by the SCAN38 and STATISTICA39
packages.
Supporting Information Available: Spectroscopic data (1H
NMR and UV-vis) and elemental analysis data for compounds

3304 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 11
Banfi et al.
to the in vivo quantitative structure-activity relationships of a
congeneric series of pyropheophorbide derivatives used as photo-
sensitizers for photodynamic therapy. Photochem. Photobiol. 1999,
70, 781-788.
(29) Renner, R. The K-ow controversy. EnViron. Sci. Technol. 2002, 36,
410.
(30) Benfenati, E.; Gini, G.; Piclin, N.; Roncaglioni, A.; Vari, M. R.
Predicting logP of pesticides using different software. Chemosphere
2003, 53, 1155-1164.
(31) Papa, E.; Villa, F.; Gramatica, P. Statistically Validated QSARs,
Based on Theoretical Descriptors, for Modeling Aquatic Toxicity of
Organic Chemicals in Pimephales promelas (Fathead Minnow). J.
Chem. Inf. Comput. Sci. 2005, 45, 1256-1266.
(32) Eriksson, L.; Jaworska, J.; Worth, A.; Cronin, M.; McDowell, R.
M.; Gramatica, P. Methods for Reliability, Uncertainty Assessment,
and Applicability Evaluations of Regression Based and Classification
QSARs. EnViron. Health Perspect. 2003, 111, 1361-1375.
(33) Atkinson, A. C. Plots, Transformation and Regression; Clarendon
Press: Oxford, U.K., 1985; p 282.
(19) Vanyur, R.; Heberger, K.; Kovesdi, I.; Jakus, J. Prediction of
tumoricidal activity and accumulation of photosensitizers in photo-
dynamic therapy using multiple linear regression and artificial neural
networks. Photochem. Photobiol. 2002, 75, 471-478.
(20) Debnath, A. K.; Jiang, S.; Strick, N.; Lin, K.; Haberfield, P.; Neurath,
R. 3-Dimensional Structure-Activity Analysis of a series of Por-
phyrin Derivatives with Anti-HIV-1 Activity Targeted to the V3 Loop
of the gp120 Envelope Glycoprotein of the Human-Immunodefi-
ciency-Virus Type-1. J. Med. Chem. 1994, 3, 1099-1108.
(21) Todeschini, R.; Consonni, V.; Mauri, A.; Pavan, M. DRAGON
Software for the Calculation of Molecular Descriptors, version 5.2
for Windows; Talete S.r.l.: Milan, Italy, 2005.
(34) Williard, P. G.; Fryhle, C. B. Boron Trihalide-Methyl Sulfide
Complexes as for Dealkylation of Aryl Ethers. Tetrahedron Lett.
1980, 21, 3731-3734.
(22) HYPERCHEM/CHEMPLUS version 7.03 for Windows; Autodesk
Inc.: Sausalito, CA, 2002.
(23) Leardi, R.; Boggia, R.; Terrile, M. Genetic Algorithms as a Strategy
for Feature Selection. J. Chemom. 1992, 6, 267-281.
(24) Tropsha, A.; Gramatica, P.; Gombar, V. K. The Importance of Being
Earnest: Validation is the Absolute Essential for Successful Ap-
plication and Interpretation of QSPR Models. QSAR Comb. Sci. 2003,
22, 69-76.
(35) Alley, M. C.; Scudiero, D. A.; Monks, A.; Hursey, M. L.; Czerwinski,
M. J.; Fine, D. L.; Abbott, B. J.; Mayo, J. G.; Shoemaker, R. H.;
Boyd, M. R. Feasibility of drug screening with panels of human tumor
cell lines. Cancer Res. 1988, 48, 589-601.
(36) Todeschini R.; Ballabio D.; Consonni V.; Mauri A.; Pavan M. MOBY
DIGS-Models By Descriptors In Genetic Selection, version 1 beta
for Windows; Talete S.r.l.: Milan, Italy, 2004.
(25) Geary, R. C. The contiguity ratio and statistical mapping, Incorp.
Statist. 1954, 5, 115-145.
(37) Todeschini, R.; Maiocchi, A.; Consonni, V., The K correlation
index: theory development and its application in chemometrics.
Chemom. Intell. Lab. Syst. 1999, 46, 13-29.
(38) SCAN-sSoftware for Chemometric Analysis, version 1.1 for
Windows; Minitab: State College, PA, 1995.
(39) STATISTICA, release 5.1 for Windows; StatSoft, Inc.: Tulsa, OK,
1987.
(26) Todeschini, R.; Consonni, V. Handbook of Molecular Descriptors;
Wiley-VCH: Weinheim, Germany, 2000; p 667.
(27) Randic, M. Novel shape descriptors for molecular graphs. J. Chem.
Inf. Comput. Sci. 2001, 41, 607-613.
(28) Consonni, V.; Todeschini, R.; Pavan, M. Structure/Response Cor-
relation and Similarity/Diversity Analysis by GETAWAY descriptors.
Part 1. Theory of the Novel 3D Molecular Descriptors. J. Chem. Inf.
Comput. Sci. 2002, 42, 682-692.
JM050997M