Identification of a potent and noncytotoxic inhibitor of melanin production

Article abstract of DOI:10.1016/j.bmc.2010.06.034

On the basis of a hit from random screening, biaryl amide derivatives were prepared in a combinatorial manner via parallel solution-phase synthesis, and their effects on melanocytes were investigated to discover new effective skin depigmenting agents. Amo

Full text of DOI:10.1016/j.bmc.2010.06.034

Bioorganic & Medicinal Chemistry 18 (2010) 5602–5609
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Identification of a potent and noncytotoxic inhibitor of melanin production
Soonho Hwang a, Sang Yoon Choi b, Jin Hee Lee c, Shinae Kim a, Jinkyung In a, Sang Keun Ha b, Eunjung Lee d,
Tae-Yoon Kim d, Sun Yeou Kim b, Sun Choi c, Sanghee Kim a,
*
a College of Pharmacy, Seoul National University, San 56-1, Shilim, Kwanak, Seoul 151-742, Republic of Korea
b Graduate School of East-West Medical Science, Kyunghee University, 1 Hoegi, Dongdaemun, Seoul 130-701, Republic of Korea
c College of Pharmacy, Division of Life and Pharmaceutical Sciences, and National Core Research Center for Cell Signaling and Drug Discovery,
Ewha Womans University, 11-1 Daehyun-Dong, Seodaemun, Seoul 120-750, Republic of Korea
d Laboratory of Dermatology-Immunology, College of Medicine, The Catholic University of Korea, 505 Banpo, Seocho, Seoul 137-040, Republic of Korea
a r t i c l e i n f o
a b s t r a c t
Article history:
On the basis of a hit from random screening, biaryl amide derivatives were prepared in a combinatorial
manner via parallel solution-phase synthesis, and their effects on melanocytes were investigated to dis-
cover new effective skin depigmenting agents. Among the 120 derivatives prepared, five members exhib-
Received 8 May 2010
Revised 10 June 2010
Accepted 11 June 2010
Available online 17 June 2010
ited a >30% reduction of melanin production at 30 lM with a cell viability of >90%. In particular,
compound A3/B5 exhibited effective inhibitory activity on melanin synthesis. Although the inhibition per-
centage of A3/B5 was slightly lower than that of the positive reference compound, phenylthiourea (PTU),
A3/B5 demonstrated a much better cell viability than PTU. In vivo evaluation of A3/B5 also showed a sig-
nificant decrease of melanin pigments. In addition, the in silico classification model was built based on
the experimental data of library members. Our results suggest that these biaryl amide derivatives may
act as potent skin depigmenting agents.
Keywords:
Biaryl amide library
Combinatorial chemistry
Depigmenting agents
Laplacian-modified naïve Bayesian
Structure–activity relationships
Ó 2010 Elsevier Ltd. All rights reserved.
1. Introduction
several mechanisms.11–13 Many chemical agents, including hydro-
quinone, arbutin, kojic acid, and phenylthiourea (Fig. 1), have been
The melanins are a group of complex phenolic biopolymers that
are distributed widely in the natural world.1 In animals, melanins
are generated by melanocytes throughout the epidermis.2,3 Mela-
nins play a pivotal role in protecting the host from various types
of ionizing sources such as free radicals and UV-radiation.4 How-
ever, the increased production and accumulation of melanins in-
duces a large number of hyperpigmentation diseases, including
skin cancer, freckles, brown spots, and senescence spots.5,6
The melanogenesis pathway in mammals is controlled at many
different levels.1,7 A variety of proteins and enzymes, such as tyros-
inase, tyrosinase-related protein-1 (TRP-1), and TRP-2, are known
to be involved in the biosynthetic pathway. Several factors have
been demonstrated to affect the activity of these enzymes and re-
lated proteins, and these factors include melanocyte stimulating
hormone (MSH), basic fibroblast growth factor (bFGF), endothe-
lin-1, and UV light.8,9
utilized for the treatment of hyperpigmentation disorders, but
none are completely satisfactory owing to adverse effects such as
toxicity.11 Therefore, the development of ideal depigmenting
agents is still in great demand. However, as far as we are aware,
The key regulatory enzyme in the melanogenesis pathway is
tyrosinase.10 Thus, many efforts have been devoted to screening
for inhibitors of this enzyme, and a majority of currently known
depigmenting agents act directly or indirectly on this enzyme via
* Corresponding author. Tel.: +82 2 880 2487; fax: +82 2 888 0649.
E-mail address: pennkim@snu.ac.kr (S. Kim).
Figure 1. Structures and biological activity of representative depigmenting agents.
0968-0896/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2010.06.034

S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
5603
amide template, a series of biaryl amides was designed in a combi-
natorial manner and synthesized (Fig. 3 and Scheme 1). One factor
to be considered in choosing the substituents on the aromatic rings
was skin permeability. Since hydrophobic compounds tend to pen-
etrate the skin much easier than hydrophilic compounds,16 methyl
and methoxy groups were primarily chosen as substituents in our
studies.
Figure 2. Chemical structure of compound 1.
We have previously reported a solution-phase strategy for the
synthesis of a biaryl amide library that utilized Girard’s reagent T
as an acid chloride scavenger.17 This clean, high-yielding protocol
was employed in the present study. The selected 15 aryl amines
A (A1–A15) (Fig. 3) were coupled with an excess (1.5 equiv) of eight
different aromatic acid chlorides B (B1–B8) in the presence of Et3N
(2.0 equiv) and a catalytic amount of DMAP (0.1 equiv) in CH2Cl2
(Scheme 1). When total consumption of the amines was observed
by thin layer chromatography, the reaction mixtures were purified
by employing the work-up procedure previously described by us.17
Briefly, the reaction mixture was treated with a solution of Girard’s
reagent T (2) in AcOH to convert the excess acid chlorides B to
water soluble hydrazine derivatives 3, and then the mixture was
diluted with ethyl acetate. Subsequent washing with 1 N HCl,
brine, saturated aqueous NaHCO3, and brine, followed by concen-
tration of the organic solution after drying over Na2SO4, yielded
the desired combinatorial biaryl amide library consisting of 120
members A1–15/B1–8. In all cases, the members obtained after
extractive work-up were virtually free from byproducts and impu-
rities with an average HPLC purity of 98%.18
there has been little systematic medicinal chemistry research on
the development of inhibitors of melanin production.
As a part of our efforts to discover new depigmenting agents, we
have screened our compound collections. This resulted in the iden-
tification of compounds 1a and 1b (Fig. 2), which had modest po-
tency with low cytotoxicity.14,15 Unlike other most depigmenting
agents, these compounds 1 had no effect on the tyrosinase enzyme
function.14,15 The biaryl amide template of 1 was attractive for a
new class of depigmenting agents due to its simple structure, ease
of synthesis, and chemical stability. As part of a continued effort to
develop a potent melanin synthesis inhibitor, we report herein the
generation of this class of compound library through an efficient
solution-phase synthesis protocol, biological evaluations, and an
in silico classification model based on the experimental data.
2. Results and discussion
2.1. Chemistry
The initial objective of our study was to identify analogues of
compound 1 with improved potency without increased toxicity.
To define the optimal substituents and their positions in the biaryl
Figure 4 shows the distribution of the library members by the
molecular properties: molecular weight, number of hydrogen-
bond acceptors, ALog P, and polar surface area. The ALog P values
Figure 3. Building blocks, A and B, used for synthesis of the biaryl amide library.

5604
S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
Scheme 1. Biaryl amide library synthetic route using Girard’s reagent T.
of members were not extremely high. Although Lipinski’s rule of
five might not be accurately applied to the topical compounds,
all of the members were drug-like according to the rule of five.
ing agents phenylthiourea (PTU)22,23 and compound 1b were also
tested for purposes of comparison. The level of reduced melanin
was expressed as the difference between the percentage of cell via-
bility and the percentage of melanin content to demonstrate the rel-
ative melanin content changes in only the viable cells. Among the
members, A3/B4 exhibited the highest reduction of melanin produc-
tion (41.5%) with a cell viability of ca. 86%. The depigmenting activity
and cell viability of compound A3/B4 was higher than that of refer-
ence compound PTU at the tested concentration. A number of library
members exerted a potent inhibitory effect on melanin production
with a low cell toxicity; five members produced a >30% reduction
in melanin production with a cell viability of >90%. For instance,
A3/B5 exerted no discernible cytotoxicity for cultured melan-a cells
and showed considerably improved inhibitory effect on melanin
production when compared with compound 1b.
There appears to be a correlation between the cell viability and
the substituent type at the biaryl amide template. For example, a
majority of the library members of A10/Bx and A14/Bx, having the
electron withdrawing group at para position of aryl amines,
showed a high level of cell viability. On the other hand, chemset
A15/Bx, which contains the trifluoromethyl group, exhibited a pro-
found reduction in cell viability. In addition, these data revealed
that the locations of substituents on the template are also crucial
in terms of affecting cell viability. For instance, chemsets Ax/B1,
Ax/B2, and Ax/B3, which differed only in the positions of their
methyl groups, showed differing levels of cell viability. The effects
of library members on melanin production were also sensitive to
the substitution pattern on the biaryl amide template. For exam-
ple, chemsets A1/Bx, A2/Bx, A5/Bx, A6/Bx, A7/Bx, A8/Bx, A10/Bx, and
A11/Bx exerted no significant inhibitory effects on melanin produc-
tion. However, chemsets A3/Bx, A12/Bx, A15/Bx, Ax/B3, Ax/B4, Ax/B5,
Ax/B6, and Ax/B7 contained many members with notable depig-
mentation activity.
2.2. In vitro assays
To date, the in vitro screening of depigmenting agents depends
predominantly on tyrosinase inhibitory activity.19 However, many
of the compounds having tyrosinase inhibitory activity failed to ex-
ert efficient inhibitory effects on melanin synthesis in melanocyte
cells. For example, kojic acid, a well-known tyrosinase inhibitor,
did not show significant depigmentation activity in either the mel-
anocytes or in vivo.20 Moreover, oxyresveratrol, which was identi-
fied as a potent tyrosinase inhibitor,21 showed high cell toxicity in
melanocyte cells.14 Thus, we believed that the melanocyte assay
was a more applicable and reliable method than the in vitro tyros-
inase assay in terms of discovering compounds with both low tox-
icity and potent depigmentation ability.14,15 In this study, we
employed the melan-a melanocytes assay to identify effective
depigmenting agents.
When screening combinatorial libraries, only one or two dilu-
tions are generally tested, because multi-point assay for all sam-
ples is not common practice. Thus, we evaluated the biaryl
amide library at a single concentration (30 lM). Although the
tested concentration was somewhat high, we felt it was appropri-
ate to identify the depigmenting compounds that have low cyto-
toxicity. The library members were dissolved in propylene glycol/
EtOH/H2O (5:3:2) solvent and added to the cell culture medium.
There was no visible precipitation of library members from the
complete assay solution.
The tests were conducted using cultured melan-a cells, and cell
toxicity and melanin production were assessed. The results of the
screening are summarized in Table 1. The known potent depigment-

S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
5605
Figure 4. Distribution of the library members: (a) molecular weight (MW), (b) number of hydrogen-bond acceptors (HBA), (c) ALog P, (d) polar surface area (PSA).
2.3. Computational study
model was evaluated, and the accuracy, sensitivity, and specificity
for the training set were 95.9%, 100%, and 95.5%, respectively.
Those for the test set were slightly lower but satisfactory (details
are as shown in Table 2.) The area under the receiver operating
characteristic (ROC) score is a measure of the accuracy of a model,
and this score between 0.9 and 1.0 is considered excellent.28 In our
model, the ROC scores were 0.96 and 0.90 in the training and test
sets, respectively. An ROC plot is the true positive rate vs. the false
positive rate, and the ROC plot of the test set was as shown in
Figure 5a. An enrichment plot explains how fast all of the actives
could be identified if the compounds were resorted by the model
score.29 Using the scores generated from our Laplacian-modified
naïve Bayesian model, the enrichment plot revealed that 70% of
the actives were found when ca. 10% of the compounds were tested
(Fig. 5b). The overall results from our model showed that it was
successful at classifying the data set compounds, and could be uti-
lized to discover new biaryl amide templated depigmenting
agents.
Since our preliminary results of quantitative structure–activity
relationship (QSAR) and quantitative structure–property relation-
ship (QSPR) analyses were not satisfactory, we decided to classify
the depigmenting effect of biaryl amides. An in silico classification
model was built by Laplacian-modified naïve Bayesian.24,25 This is
an effective machine learning method to classify actives from inac-
tives, and has become more frequently used in recent drug discov-
ery. One hundred and twenty compounds of the biaryl amide series
with depigmenting activity values were used to build the classifi-
cation model. If a library member had depigmenting activity great-
er than or equal to 20% with a cell viability of P80%, it was defined
as active; other members were defined as inactive. All the data
were split randomly into the training and test sets using the ran-
dom percent filter of Pipeline Pilot,26 resulting in 73 compounds
in the training set and 47 compounds in the test set (as marked
in Table 1). To obtain a reliable model, five descriptors were finally
selected: ALog P, molecular weight, number of hydrogen-bond
acceptors, polar surface area, and the extended-connectivity fin-
gerprint (FCFP_8).27 All of the compounds had one hydrogen-bond
donor, so it was excluded from the descriptors.
2.4. In vivo assays
Although a member A3/B4 exhibited the highest inhibitory
activity of melanin production, we were more interested in A3/B5
because it manifested a profile superior to that of A3/B4 in terms
In the model building, we used leave-on-out (LOO) cross-valida-
tion to avoid overfitting of the training set. Performance of the

5606
S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
Table 1
Biological data for compounds A1–15/B1–8
A/B
Cell viabilitya (%)
Depigmenting
effecta,b (%)
A/B
Cell viability (%)
Depigmenting
effect (%)
A/B
Cell viability (%)
Depigmenting
effect (%)
d
d
A1/B1
A1/B2
A1/B3
A1/B4
A1/B5
A1/B6
A1/B7
A1/B8
A2/B1
A2/B2
A2/B3
A2/B4
A2/B5
A2/B6
A2/B7
A2/B8
A3/B1
A3/B2
A3/B3
A3/B4
A3/B5
A3/B6
A3/B7
A3/B8
A4/B1
A4/B2
A4/B3
A4/B4
A4/B5
A4/B6
A4/B7
A4/B8
A5/B1
A5/B2
A5/B3
A5/B4
A5/B5
A5/B6
A5/B7
A5/B8
87.5 3.5
79.9 3.9
96.1 2.3
100.0 2.7
103.1 6.0
100.8 4.2
97.4 5.2
93.7 1.9
85.4 5.5
67.4 4.6
86.0 3.3
75.1 5.8
84.6 6.6
99.8 3.1
92.2 5.3
100.2 1.9
101.7 6.4
81.0 1.4
94.8 4.8
85.8 10.6
100.4 4.1
90.5 3.3
98.4 1.8
99.2 1.9
87.0 2.9
71.1 6.7
91.0 3.5
83.5 3.1
95.7 2.4
94.0 2.6
94.5 1.8
85.7 6.3
93.4 2.5
77.0 6.9
100.0 3.4
66.3 7.4
90.8 3.0
92.6 2.1
97.8 0.6
97.9 3.1
99.8 2.4
5.8
À0.1
14.3
À2.2
12.0
À1.7
11.1
À1.3
À0.2
À2.4
13.9
1.8
7.5
0.7
10.9
7.9
25.6
6.0
A6/B1
A6/B2
A6/B3
A6/B4
A6/B5
A6/B6
A6/B7
A6/B8
A7/B1
A7/B2
A7/B3
A7/B4
A7/B5
A7/B6
A7/B7
A7/B8
A8/B1
A8/B2
A8/B3
A8/B4
A8/B5
A8/B6
A8/B7
A8/B8
A9/B1
A9/B2
A9/B3
A9/B4
A9/B5
A9/B6
A9/B7
A9/B8
97.8 3.2
94.0 4.1
100.6 1.8
76.5 3.3
99.0 4.1
93.7 2.8
98.5 2.6
95.5 2.1
95.5 4.4
103.2 4.8
99.4 1.6
78.5 5.5
98.9 0.7
94.5 3.2
101.8 3.3
71.9 5.9
95.6 1.5
89.9 5.3
84.1 7.7
72.7 2.7
101.6 5.0
94.4 5.2
96.0 3.6
91.6 4.6
89.1 3.0
89.3 2.9
92.1 3.1
94.8 5.4
95.2 0.9
93.6 3.7
97.9 4.4
93.7 1.9
97.9 1.8
95.4 3.8
98.0 0.5
99.7 2.1
96.3 5.0
96.2 2.8
96.9 4.8
93.2 3.3
99.8 3.9
12.8
À0.4
4.9
A11/B1
A11/B2
A11/B3
A11/B4
A11/B5
A11/B6
A11/B7
A11/B8
A12/B1
A12/B2
A12/B3
A12/B4
A12/B5
A12/B6
A12/B7
A12/B8
A13/B1
A13/B2
A13/B3
A13/B4
A13/B5
A13/B6
A13/B7
A13/B8
A14/B1
A14/B2
A14/B3
A14/B4
A14/B5
A14/B6
A14/B7
A14/B8
A15/B1
A15/B2
A15/B3
A15/B4
A15/B5
A15/B6
A15/B7
A15/B8
PTU
81.6 4.0
96.5 1.4
100.7 1.5
90.4 2.6
90.4 5.0
97.1 1.1
89.0 4.2
87.8 0.9
90.4 6.5
85.3 3.9
90.1 6.5
91.9 2.6
92.7 5.3
98.2 2.4
99.6 2.0
97.4 4.3
99.2 8.6
89.6 2.2
94.2 0.9
79.7 6.5
92.4 8.9
99.2 4.7
86.4 3.6
92.5 2.9
98.6 3.1
98.7 3.5
97.1 4.2
96.7 6.1
102.9 2.0
96.2 0.6
97.4 2.9
85.6 3.7
35.5 6.0
90.4 4.2
69.7 2.2
30.5 3.9
45.5 3.2
37.4 6.8
67.5 7.0
72.1 14.6
81.1 6.6
9.9
4.8
À2.4
À10.8
19.1
1.0
d
d
d
d
d
d
À9.9
0.5
d
d
À10.6
8.8
13.6
À3.5
3.2
14.4
À6.5
À3.9
À9.4
7.2
À4.4
3.7
d
d
9.7
d
d
d
d
29.6
20.1
18.9
11.1
31.1c
8.0
d
d
d
d
d
d
À24.1
À2.3
8.6
4.6
0.3
10.5
16.4
37.0
20.1
18.8
0.8
30.9c
41.5
32.0c
31.9c
31.6c
14.2
7.0
11.1
18.5
13.5
10.1
À2.6
À3.1
À5.4
13.4
À0.3
2.7
À3.2
À19.8
17.8
À8.3
À1.7
À17.0
6.4
À12.6
19.2
10.1
9.6
d
d
d
d
d
d
d
d
0.4
0.8
12.9
11.2
2.0
13.7
À5.6
14.2
À8.4
15.1
À2.3
0.3
16.5
9.0
21.0
33.3
31.9
38.0
d
d
d
d
13.6
9.6
d
d
d
d
À1.3
d
d
d
d
d
d
d
A10/B1
A10/B2
A10/B3
A10/B4
A10/B5
A10/B6
A10/B7
A10/B8
1b
0.1
À5.9
11.2
3.4
6.3
d
d
À8.4
À5.0
0.6
0.7
À9.4
0.4
0.0
6.9
À2.6
Vehicle
10.2
a
b
c
Cell viability and depigmenting effect was tested at 30 lM.
Depigmenting effect was expressed as the difference between the percentage of cell viability and the melanin content {cell viability (%) À melanin content (%)}.
Bold-face indicates biaryl amide members with >30% depigmenting activity and >90% cell viability.
d
In the Laplacian-modified naïve Bayesian classification model, these compounds were used as the test set.
effectively inhibited UVB-induced pigmentation and suggests that
biaryl amide derivatives could act as efficient therapeutic agents
for hyperpigmentation in skin.
Table 2
Results of the Laplacian-modified naïve Bayesian model
Set
Actives
TP FN
Inactives
Acc (%)
Sen (%)
Spe (%)
ROC
FP
TN
3. Conclusion
Training
Test
6
3
0
1
3
2
64
41
95.9
93.6
100
75.0
95.5
95.4
0.96
0.90
In conclusion, our study was conducted using melanocytes to
discover a compound that fulfills the requisite criteria of reducing
melanin production and cell toxicity. Based on the structure of the
hit from the screening, we designed and synthesized a series of
biaryl amides in a combinatorial manner. The use of parallel solu-
tion-phase synthesis techniques enabled the rapid preparation and
identification of optimal substituents for each aryl moiety exam-
ined. The selected compound exhibited an effective inhibitory
activity on melanin synthesis, with low cell toxicity in melan-a
cells and a significant decrease of melanin pigment in the UVB-
stimulated hyper-pigmented brown guinea pig skin. Our findings
indicate that these biaryl amide derivatives may act as potent skin
depigmenting agents. Additionally, our in silico model, generated
by Laplacian-modified naïve Bayesian, showed an effective classifi-
cation of the actives from the inactives and could be utilized for the
prediction of new depigmenting agents. In addition, we believe
TP: true positive, FN: false negative, FP: false positive, TN: true negative, Acc:
accuracy, Sen: sensitivity, Spe: specificity, ROC: the area under the receiver oper-
ating characteristic curve scores, Definitions of the Acc, Sen, and Spe are described
in Section 4.
of cell viability at the tested concentration (30 lM), and also exhib-
ited a comparable inhibitory effect on melanin production. To as-
sess the effect of A3/B5 on the inhibition of pigmentation in vivo,
it was resynthesized and purified by chromatography. Compound
A3/B5 (10 lg, 0.1% in propylene glycol/EtOH/H2O) were applied
topically to the UVB-stimulated hyper-pigmented30 dorsal skin of
brown guinea pigs. As shown Figure 6, Fontana masson staining
showed a significant decrease in the amount of melanin pigment
in the A3/B5-treated brown guinea pig skin (Fig. 6b) compared to
the control (Fig. 6c). This result indicated that A3/B5 treatment

S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
5607
Figure 5. ROC and enrichment plots of the Laplacian-modified naïve Bayesian model for the test set. (a) ROC plot. (b) Enrichment plot of our model (red), compared with
those of the perfect model (blue) and random model (green).
Figure 6. Chemical structure and effect of compound A3/B5 on UV-induced pigmentation in brown guinea pig skin. (a) Chemical structure of A3/B5. (b) 10 lg of 0.1% A3/B5
was applied topically twice a day for 4 weeks after the last UV tanning. (c) Control experiments were performed with use of the vehicle only.
that our study gives useful direction for further studies into the
development of other suitable depigmenting agents. Detailed
mechanistic studies directed towards identification of the molecu-
lar target of biaryl amides and biological evaluations of this newly
disclosed class of depigmenting agents are ongoing and will be re-
ported in due course.
violet absorption detector at 225 nm; gradient, 30–100% MeOH/
H2O, 40 min). Amines, aromatic acid chlorides, triethylamine, Gir-
ard’s reagent T, and phorbol 12-myristate 13-acetate (TPA) were
obtained from Sigma–Aldrich Inc. (USA). PTU was purchased from
the TCI Co. (Japan). FBS, AA, and RPMI were purchased from Gibco
BRL (USA). The melanin content and cell viability were measured
using a Tecan Sunrise Remote microplate reader.
4. Experimental
4.1. General
4.2. General procedure for parallel synthesis of biaryl amides
To a solution of aryl amine A (0.2 mmol) in CH2Cl2 (1 mL) was
added acid chloride B (0.3 mmol), Et3N (0.4 mmol), and a catalytic
amount of DMAP (0.02 mmol). The mixture was stirred at room
temperature for 5 h, after which a solution of Girard’s reagent T
in AcOH (0.5 M, 0.5 mL) was added. It was stirred overnight at
room temperature and then diluted with ethyl acetate (3 mL).
The reaction mixture was purified by subsequent washing with
1 N HCl (2 mL Â 2), brine (2 mL), saturated aqueous NaHCO3
(2 mL Â 2), and brine (2 mL). The organic layer was passed through
a short pad of drying agent (Na2SO4) and concentrated to give
biaryl amide A/B (120 compounds). Purities of all products were
All chemicals were reagent grade and used as purchased. All
reactions were performed under an inert atmosphere of dry argon
or nitrogen using distilled dry solvents. Reactions were monitored
by TLC analysis using Silica Gel 60 F-254 thin layer plates. 1H NMR
and 13C NMR spectra were recorded in d units relative to deuter-
ated solvent as the internal reference at 300 or 400 MHz. Mass
spectra (MS) were recorded using fast atom bombardment (FAB).
High resolution mass spectra (HRMS) were recorded using FAB.
Purities of products were determined by HPLC peak area analysis
(ZORBAX Eclipse plus C18 column, 4.6 mm  150 mm, 5
lm; ultra-

5608
S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
determined by HPLC peak area analysis. All operations in this syn-
thesis were conducted on a Quest 210 parallel organic synthesizer,
and solvent evaporation was performed on Genevac EZ-2 vacuum
centrifuge.
2.6, 7.7 Hz, 1H), 6.87 (d, J = 9.0 Hz, 2H), 3.83 (s, 3H), 3.80 (s, 3H);
MS (FAB): m/z (%): 258 (11) [M+H]+, 176 (20), 154 (100), 136
(72); HRMS-FAB: m/z [M+H]+ calcd for C15H16NO3: 258.1130, found
258.1125; purity by HPLC: 97.3%.
4.3. Analytical data of representative biaryl amide compounds
4.3.9. N-(3,5-Dimethoxyphenyl)-4-methylbenzamide (A9/B3)
Yield 92%; 1H NMR (300 MHz, CD3OD) d 7.80 (td, J = 2.0, 8.3 Hz,
2H), 7.30 (d, J = 7.9 Hz, 2H), 6.96 (d, J = 2.2 Hz, 2H), 6.27 (dd, J = 2.2,
2.4 Hz, 1H), 3.77 (s, 6H), 2.40 (s, 3H); MS (FAB): m/z (%): 272 (46)
[M+H]+, 154 (100), 136 (68), 119 (35); HRMS-FAB: m/z [M+H]+
calcd for C16H18NO3: 272.1287, found 272.1292; purity by HPLC:
99.5%.
Compounds A5/B8, A6/B3, and A6/B5 were previously reported in
2002 by us.17 However, HPLC purity, mass, and HRMS data were
reanalyzed.
4.3.1. 4-Cyano-N-m-tolylbenzamide (A1/B7)
Yield 92%; 1H NMR (400 MHz, CD3OD) d 8.05 (d, J = 8.4 Hz, 2H),
7.87 (d, J = 8.4 Hz, 2H), 7.51 (s, 1H), 7.47 (d, J = 8.2 Hz, 1H), 7.24 (t,
J = 7.8 Hz, 1H), 6.99 (d, J = 7.0 Hz, 1H), 2.35 (s, 3H); MS (FAB): m/z
(%): 237 (7) [M+H]+, 176 (25), 154 (100), 136 (87); HRMS-FAB:
m/z [M+H]+ calcd for C15H13N2O: 237.1028, found: 237.1033; pur-
ity by HPLC: 99.1%.
4.3.10. Methyl 2-(4-methoxybenzamido)benzoate (A12/B6)
Yield 91%; 1H NMR (400 MHz, CD3OD) d 8.75 (d, J = 8.5 Hz, 1H),
8.05 (dd, J = 1.2, 7.9 Hz, 1H), 7.94 (d, J = 8.8 Hz, 2H), 7.56 (dt, J = 1.4,
8.6 Hz, 1H), 7.11 (t, J = 7.4 Hz, 1H), 7.00 (d, J = 8.8 Hz, 2H), 3.94 (s,
3H), 3.85 (s, 3H); MS (FAB): m/z (%): 286 (9) [M+H]+, 176 (18),
154 (100), 136 (75); HRMS-FAB: m/z [M+H]+ calcd for
4.3.2. N-(3,5-Dimethylphenyl)-4-methylbenzamide (A3/B3)
Yield 91%; 1H NMR (300 MHz, CDCl3) d 7.71 (d, J = 8.1 Hz, 3H),
7.24–7.22 (m, 3H), 6.75 (s, 1H), 2.38 (s, 3H), 2.28 (s, 6H); MS
(FAB): m/z (%): 240 (72) [M+H]+, 239 (51), 154 (100), 136 (72),
119 (47); HRMS-FAB: m/z [M+H]+ calcd for C16H18NO: 240.1388,
found 240.1379; purity by HPLC: 99.2%.
C16H16NO4: 286.1079, found 286.1081; purity by HPLC: 98.8%.
4.4. Biological assay
4.4.1. Cell line and culture procedure
A cell line of pigmented melanocytes, ‘melan-a’, was derived
from normal epidermal melanoblasts from the embryos of inbred
C57BL mice.31 The melan-a cells were generously donated by Dr.
B. G. Lee at the Skin Research Institute, Amore-Pacific Co, Korea.
The cells were cultured at 37 °C under an atmosphere of 5% CO2
in RPMI1640 medium with 10% FBS, 1% AA, and 200 nM phorbol
12-myristate 13-acetate (TPA). Melan-a cells were seeded at
105 cells/well in 24-well plates and incubated for 24 h. The media
4.3.3. N-(3,5-Dimethylphenyl)-4-ethylbenzamide (A3/B4)
Yield 96%; 1H NMR (300 MHz, CDCl3) d 7.74 (d, J = 8.3 Hz, 3H),
7.26–7.22 (m, 3H), 6.74 (s, 1H), 2.46 (q, J = 7.7 Hz, 2H), 2.27 (s,
6H), 1.22 (t, J = 7.7 Hz, 3H); MS (FAB): m/z (%): 254 (100) [M+H]+,
133 (67); HRMS-FAB: m/z [M+H]+ calcd for C17H20NO: 254.1545,
found 254.1551; purity by HPLC: 99.6%.
in each well were exchanged with 990
lL of medium each day and
4.3.4. N-(3,5-Dimethylphenyl)-3-methoxybenzamide (A3/B5)
Yield 94%; 1H NMR (400 MHz, CD3OD) d 7.48–7.45 (m, 2H), 7.39
(t, J = 8.0 Hz, 1H), 7.29 (s, 2H), 7.11 (dd, J = 2.1, 8.2 Hz, 1H), 6.80 (s,
1H), 3.85 (s, 3H), 2.23 (s, 6H); MS (FAB): m/z (%): 256 (100) [M+H]+,
255 (50), 154 (60), 135 (69); HRMS-FAB: m/z [M+H]+ calcd for
treated for three days with 10 L of test samples. The test samples
were dissolved in propylene glycol/EtOH/H2O = 5:3:2 solvent.
l
4.4.2. Measurement of the cell viability
The percentage of viable cells was determined by staining the
cells with crystal violet. After removing the medium from each
C16H18NO2: 256.1338, found 256.1332; purity by HPLC: 99.4%.
well, the cells were washed with PBS. 200 lL of crystal violet (CV
4.3.5. 4-Cyano-N-(3,5-dimethylphenyl)benzamide (A3/B7)
Yield 95%; 1H NMR (400 MHz, CD3OD) d 8.04 (d, J = 8.3 Hz, 2H),
7.87 (d, J = 8.4 Hz, 2H), 7.30 (s, 2H), 6.83 (s, 1H), 2.31 (s, 6H); MS
(FAB): m/z (%): 251 (4) [M+H]+, 176 (23), 154 (100), 136 (88);
HRMS-FAB: m/z [M+H]+ calcd for C16H15N2O: 251.1184, found
251.1181; purity by HPLC: 99.4%.
0.1%, 10% EtOH, the rest was PBS) was added. The mixture was
incubated for 5 min at room temperature and washed twice in
water. After the addition of 1 mL of EtOH, the mixture was shaken
for 10 min at room temperature. The UV absorption was measured
at 590 nm.
4.4.3. Determination of the melanin level
4.3.6. N-(3-Methoxyphenyl)furan-2-carboxamide (A5/B8)
After removing the media from each well, the samples were
washed with PBS. This was followed by the addition of 1 mL of
1 N NaOH and shaking to dissolve the melanin. UV absorption
was measured at 400 nm, and the percentage of melanin content
per well was calculated by comparison with the controls.
Yield 92%; 1H NMR (300 MHz, CDCl3) d 8.17 (br s, 1H), 7.48 (dd,
J = 0.9, 1.8 Hz, 1H), 7.45 (t, J = 2.1 Hz, 1H), 7.22 (m, 2H), 7.11 (ddd,
J = 0.9, 1.8, 8.1 Hz, 1H), 6.69 (ddd, J = 0.9, 2.4, 8.4 Hz, 1H), 6.53 (dd,
J = 1.8, 3.3 Hz, 1H), 3.80 (s, 3H); MS (FAB): m/z (%): 218 (47)
[M+H]+, 154 (100), 136 (74); HRMS-FAB: m/z [M+H]+ calcd for
C
12H12NO3: 218.0817, found 218.0814; purity by HPLC: 99.3%.
4.4.4. UVB-induced hyperpigmentation in brown guinea pigs
UVB-induced hyperpigmentation was induced on the backs of
the brownish guinea pigs weighing approximately 500 g using a
slight modification of the methods reported by Ando et al.32 and
Imokawa et al.33 The guinea pigs were anesthetized with pentobar-
bital (30 mg/kg), and separate areas (1 cm diametrical circle) of the
back of each animal were exposed to the UVB radiation (Wald-
mann UV 800, Herbert Waldmann GmbH, Philips TL/12 lamp emit-
ting 280–305 nm). The total UVB dose was 500 mJ/cm2. Groups of
four animals were used in the experiments. Animals were exposed
to the UVB radiation once a week for three consecutive weeks. Ten
micrograms of compound A3/B5 (0.1% in propylene glycol/EtOH/
H2O = 5:3:2) was given topically to the hyper-pigmented areas
4.3.7. N-(4-Methoxyphenyl)-4-methylbenzamide (A6/B3)
Yield 93%; 1H NMR (300 MHz, CDCl3) d 9.35 (br s, 1H), 7.88 (d,
J = 8.4 Hz, 2H), 7.73 (d, J = 9.3 Hz, 2H), 7.28 (d, J = 8.4 Hz, 2H),
6.90 (d, J = 9.3 Hz, 2H), 3.77 (s, 3H), 2.38 (s, 3H); MS (FAB): m/z
(%): 242 (19) [M+H]+, 176 (25), 154 (100), 136 (77); HRMS-FAB:
m/z [M+H]+ calcd for C15H16NO2: 242.1181, found 242.1174; purity
by HPLC: 97.4%.
4.3.8. 3-Methoxy-N-(4-methoxyphenyl)benzamide (A6/B5)
Yield 92%; 1H NMR (300 MHz, CDCl3) d 7.95 (br s, 1H), 7.53 (d,
J = 9.0 Hz, 2H), 7.41 (m, 1H), 7.42–7.30 (m, 2H), 7.04 (ddd, J = 1.7,

S. Hwang et al. / Bioorg. Med. Chem. 18 (2010) 5602–5609
5609
twice a day for 4 weeks from the next day of the last tanning.
Acknowledgments
4 weeks later, skin biopsies were carried out and processed for
Fontana masson staining. Fontana masson staining was performed
on 5 lm sections embedded in paraffin. The animals were housed
and maintained in a barrier facility at the Institute for Animal Stud-
ies, School of Medicine, Catholic University of Korea. All of the ani-
mal protocols used in this study were approved by the Catholic
Research Institute of the Medical Science Committee for Institu-
tional Animal Care and Use.
This work was supported by the WCU program (R32-2008-000-
10098-0) through KOSEF funded by MEST and the NCRC program
(R15-2006-020) of the MEST and NRF through the Center for Cell
Signaling & Drug Discovery Research at Ewha Womans University.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2010.06.034.
4.5. Computational study
All computation calculations were undertaken on a Linux (Cent
OS release 4.6) workstation.
References and notes
1. Prota, G. Pigment Cell Res. 2000, 13, 283.
2. Sulaimon, S. S.; Kitchell, B. E. Vet. Dermatol. 2003, 14, 57.
3. Lin, J. Y.; Fisher, D. E. Nature 2007, 445, 843.
4. Sugumaran, M. Pigment Cell Res. 2002, 15, 2.
4.5.1. Data set and descriptors
All of the compounds were sketched and cleaned up by Concord
in Tripos SYBYL v.8.1.1. To generate the classification model, the
data set was split into two groups, actives and inactives, and
randomly divided into the training and test sets in a ratio of 1:1.
The descriptors used in this study were: ALog P, molecular weight,
number of hydrogen-bond acceptors, polar surface area, and the
functional-class fingerprint (FCFP_8). FCFPs are 2D and extended-
connectivity fingerprints based on the functional roles of an atom
and its neighbors (i.e., hydrogen-bond donor/acceptor, positively/
negatively ionizable, aromatic, and halogen). Splitting of data sets,
calculation of descriptors, and building and evaluation of the classi-
fication model were carried out using SciTegic Pipeline Pilot v.7.5.2.
5. Urabe, K.; Nakayama, J.; Hori, Y. In The Pigmentary System: Physiology and
Pathophysiology; Nordlund, J. J., Boissy, R. E., Hearing, V. J., King, R. A., Ortonne,
J.-P., Eds.; Oxford University Press: New York, 1998; pp 909–911.
6. Cullen, M. K. In The Pigmentary System: Physiology and Pathophysiology;
Nordlund, J. J., Boissy, R. E., Hearing, V. J., King, R. A., Ortonne, J.-P., Eds.;
Oxford University Press: New York, 1998; pp 760–766.
7. Ito, S.; Wakamatsu, K.; Ozeki, H. Pigment Cell Res. 2000, 13, 103.
8. Martínez-Liarte, J. H.; Solano, F.; García-Borrón, J. C.; Jara, J. R.; Lozano, J. A. J.
Invest. Dermatol. 1992, 99, 435.
9. Miyamura, Y.; Coelho, S. G.; Wolber, R.; Miller, S. A.; Wakamatsu, K.; Zmudzka,
B. Z.; Ito, S.; Smuda, C.; Passeron, T.; Choi, W.; Batzer, J.; Yamaguchi, Y.; Beer, J.
Z.; Hearing, V. J. Pigment Cell Res. 2007, 20, 2.
10. Wang, N.; Hebert, D. N. Pigment Cell Res. 2006, 19, 3.
11. Briganti, S.; Camera, E.; Picardo, M. Pigment Cell Res. 2003, 16, 101.
12. Okombi, S.; Rival, D.; Bonnet, S.; Mariotte, A.-M.; Perrier, E.; Boumendjel, A. J.
Med. Chem. 2006, 49, 329.
13. Parvez, S.; Kang, M.; Chung, H.-S.; Cho, C.; Hong, M.-C.; Shin, M.-K.; Bae, H.
Phytother. Res. 2006, 20, 921.
14. Choi, S. Y.; Kim, S.; Hwang, J. S.; Lee, B. G.; Kim, H.; Kim, S. Y. Biochem.
Pharmacol. 2004, 67, 707.
15. Choi, S. Y.; Hwang, J. S.; Kim, S.; Kim, S. Y. Biochem. Biophys. Res. Commun. 2006,
349, 39.
16. Bos, J. D.; Meinardi, M. M. H. M. Exp. Dermatol. 2000, 9, 165.
17. Kim, S.; Ko, H.; Kim, S.; Lee, T. J. Comb. Chem. 2002, 4, 549.
18. See the Supplementary data for details.
4.5.2. Laplacian-modified naïve Bayesian
The Bayesian classifier is a statistical modeling method based
on Bayes’s rule of conditional probability. The actual implementa-
tion of naïve Bayesian in Pipeline Pilot employs the Laplacian cor-
rection which assumes that the descriptors in the training set are
independent and of equal importance.24 This method produces a
high-dimensional representation of the molecule by using ex-
tended-connectivity fingerprints. In this study, the model building
and analysis were performed by Pipeline Pilot v.7.5.2 with the de-
fault parameters.
19. Ando, H.; Kondoh, H.; Ichihashi, M.; Hearing, V. J. J. Invest. Dermatol. 2007, 127,
751.
20. Kong, Y. H.; Jo, Y. O.; Cho, C.-W.; Son, D.; Park, S.; Rho, J.; Choi, S. Y. Biol. Pharm.
Bull. 2008, 31, 946.
21. Kim, Y. M.; Yun, J.; Lee, C.-K.; Lee, H.; Min, K. R.; Kim, Y. J. Biol. Chem. 2002, 277,
16340.
22. Poma, A.; Bianchini, S.; Miranda, M. Mutat. Res. 1999, 446, 143.
23. Kahn, V. Pigment Cell Res. 1995, 8, 234.
24. Xia, X.; Maliski, E. G.; Gallant, P.; Rogers, D. J. Med. Chem. 2004, 47,
4463.
25. The preliminary results of our quantitative structure–activity relationship
(QSAR) and quantitative structure–property relationship (QSPR) analyses were
not satisfactory.
4.5.3. Model validation
The performance of classification model was evaluated by cal-
culating the accuracy, sensitivity, and specificity, receiver operat-
ing characteristic (ROC) curve, and enrichment plot. Accuracy of
a model is defined to correctly classify the actives from the inac-
tives. Sensitivity is the ability of the model to avoid false negatives
and specificity is its ability to avoid false positives. The accuracy,
sensitivity, and specificity were calculated in the following way:
26. Pipeline Pilot, http://accelrys.com.
27. Accelrys Software Inc., Chemistry Collection: Basic Chemistry User Guide,
Pipeline Pilot Release 7.0, Accelrys Software Inc., San Diego, 2008.
28. Accelrys Software Inc., Data Modeling Collection: Basic Data Modeling User
Guide, SciTegic Pipeline Pilot Release 7.0, Accelrys Software Inc., San Diego,
2008.
Accuracy = (TP + TN)/(TP + FN + TN + FP)
Sensitivity = TP/(TP + FN)
Specificity = TN/(TN + FP)
29. Sun, H. ChemMedChem 2006, 1, 315.
30. Bolognia, J.; Murray, M.; Pawelek, J. J. Invest. Dermatol. 1989, 92, 651.
31. Bennett, D. C.; Cooper, P. J.; Hart, I. R. Int. J. Cancer 1987, 39, 414.
32. Ando, H.; Ryu, A.; Hashimoto, A.; Oka, M.; Ichihashi, M. Arch. Dermatol. Res.
1998, 290, 375.
33. Imokawa, G.; Kawai, M.; Mishima, Y.; Motegi, I. Arch. Dermatol. Res. 1986, 278,
352.
where TP is the number of true positives, TN is the number of true
negatives, FP is the number of false positives, and FN is the number
of false negatives.