Quantitative structure-activity relationship studies of threo-methylphenidate analogs

Article abstract of DOI:10.1016/j.bmc.2010.08.034

Complementary two-dimensional (2D) and three-dimensional (3D) Quantitative Structure-Activity Relationship (QSAR) techniques were used to derive a preliminary model for the dopamine transporter (DAT) binding affinity of 80 racemic threo-methylphenidate (MP) analogs. A novel approach based on using the atom-level E-state indices of the 14 common scaffold atoms in a sphere exclusion protocol was used to identify a test set for 2D- and 3D-QSAR model validation. Comparative Molecular Field Analysis (CoMFA) contour maps based on the structure-activity data of the training set indicate that the 2′ position of the phenyl ring cannot tolerate much steric bulk and that addition of electron-withdrawing groups to the 3′ or 4′ positions of the phenyl ring leads to improved DAT binding affinity. In particular, the optimal substituents were found to be those whose bulk is mainly in the plane of the phenyl ring. Substituents with significant bulk above or below the plane of the ring led to decreased binding affinity. Suggested alterations to be explored in the design of new compounds are the placement at the 3′ and 4′ position of the phenyl ring of electron-withdrawing groups that lie chiefly in the plane of the ring, for example, halogen substituents on the 3′,4′-benzo analog, 79. A complementary 2D-QSAR approach - partial least squares analysis using a reduced set of Molconn-Z descriptors - supports the CoMFA structure-activity interpretation that phenyl ring substitution is a major determinant of DAT binding affinity. The potential usefulness of the CoMFA models was demonstrated by the prediction of the binding affinity of methyl 2-(naphthalen-1-yl)-2-(piperidin-2-yl)acetate, an analog not in the original data set, to be in good agreement with the experimental value.

Full text of DOI:10.1016/j.bmc.2010.08.034

Bioorganic & Medicinal Chemistry 18 (2010) 7221–7238
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Quantitative structure–activity relationship studies
of threo-methylphenidate analogs
Milind Misra a, , Qing Shi b, Xiaocong Ye b, Ewa Gruszecka-Kowalik b, Wei Bu b, Zhanzhu Liu b,
Margaret M. Schweri c, Howard M. Deutsch b, Carol A. Venanzi a,
⇑
a Department of Chemistry and Environmental Science, New Jersey Institute of Technology, Newark, NJ 07102, USA
b School of Chemistry and Biochemistry, Georgia Institute of Technology, Atlanta, GA 30332-0400, USA
c Retired from Mercer University School of Medicine, Macon, GA 31207-0001, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Complementary two-dimensional (2D) and three-dimensional (3D) Quantitative Structure–Activity Rela-
tionship (QSAR) techniques were used to derive a preliminary model for the dopamine transporter (DAT)
binding affinity of 80 racemic threo-methylphenidate (MP) analogs. A novel approach based on using the
atom-level E-state indices of the 14 common scaffold atoms in a sphere exclusion protocol was used to
identify a test set for 2D- and 3D-QSAR model validation. Comparative Molecular Field Analysis (CoMFA)
contour maps based on the structure–activity data of the training set indicate that the 20 position of the
phenyl ring cannot tolerate much steric bulk and that addition of electron-withdrawing groups to the 30
or 40 positions of the phenyl ring leads to improved DAT binding affinity. In particular, the optimal sub-
stituents were found to be those whose bulk is mainly in the plane of the phenyl ring. Substituents with
significant bulk above or below the plane of the ring led to decreased binding affinity. Suggested altera-
tions to be explored in the design of new compounds are the placement at the 30 and 40 position of the
phenyl ring of electron-withdrawing groups that lie chiefly in the plane of the ring, for example, halogen
substituents on the 30,40-benzo analog, 79. A complementary 2D-QSAR approach—partial least squares
analysis using a reduced set of Molconn-Z descriptors—supports the CoMFA structure–activity interpre-
tation that phenyl ring substitution is a major determinant of DAT binding affinity. The potential useful-
ness of the CoMFA models was demonstrated by the prediction of the binding affinity of methyl
2-(naphthalen-1-yl)-2-(piperidin-2-yl)acetate, an analog not in the original data set, to be in good agree-
ment with the experimental value.
Received 11 June 2010
Revised 8 August 2010
Accepted 14 August 2010
Available online 19 August 2010
Keywords:
Molecular descriptors
Dopamine transporter
Methylphenidate
Comparative molecular field analysis
Ó 2010 Elsevier Ltd. All rights reserved.
1. Introduction
transporter and less at the norepinephrine transporter.7,8 Unlike
cocaine,9 MP is not a sodium channel blocker and hence would
Cocaine binds to the dopamine transporter (DAT) in the brain
and is believed to produce its euphoric and addictive effects by
inhibiting the reuptake of synaptic dopamine into presynaptic neu-
rons.1,2 Structure–activity relationship (SAR) studies for several
classes of dopamine reuptake inhibitors have been reviewed.3–6
threo-Methylphenidate (MP) and its analogs comprise one such
class that has been studied as possible cocaine abuse therapeutics.
Although both MP and cocaine are potent inhibitors of dopamine
transport, they have markedly different effects at several other
neuronal sites. MP has significant inhibitory activity at the norepi-
nephrine transporter and virtually none at the serotonin trans-
porter, while cocaine has significant activity at the serotonin
not be expected to slow neuronal conduction. Perhaps because of
these important differences, MP has found use as an orally-admin-
istered medication for the long-term treatment of children with
attention deficit hyperactivity disorder.
The accumulated experience with MP in a large pediatric popu-
lation, coupled with a relatively low incidence of serious side ef-
fects and little evidence of abuse potential, have generated
considerable interest in MP and its analogs as potential pharmaco-
therapies for cocaine abuse. The wide range of MP SAR studies pro-
duced by ourselves and others7,8,10–23 provides a large amount of
data for pharmacophore modeling.24 However, although the amino
acid sequence of the DAT has been identified,25 its three-dimen-
sional structure is not known, necessitating a ligand-based rather
than structure-based modeling approach. As a first step in this
direction, we have recently carried out an exhaustive conforma-
tional analysis26 of neutral (nMP) and protonated (pMP) methyl-
phenidate, comparing their potential energy surfaces in various
⇑
Corresponding author. Tel./fax: +1 (973) 596 3596.
E-mail address: carol.a.venanzi@njit.edu (C.A. Venanzi).
Present address: Multiscale Dynamic Materials Modeling Department, Sandia
National Laboratories, Albuquerque, NM 87185-1322, USA.
0968-0896/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2010.08.034

7222
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
molecular mechanics, molecular orbital, and solvation models to
each other and to rotationally restricted MP analogs.8 The present
work takes this computer modeling a step further by using com-
plementary two-dimensional (2D) and three-dimensional (3D)
Quantitative Structure–Activity Relationship (QSAR) approaches
to generate a preliminary model for the DAT binding of 80 MP ana-
logs. The simple models produced by these studies identified phe-
nyl ring modification as a major determinant of DAT binding
affinity. This paves the way for our future comprehensive 3D-QSAR
studies of MP analogs with only phenyl ring substitutions. Previ-
ously, 2D-27–29 and 3D-QSAR30–42 studies have been carried out
on other dopamine reuptake inhibitors, but the present work is
the first application of these techniques to methylphenidate
analogs.
the 14 analogs in the test set represent all possible combinations
of substitutions. Table 1 shows that the substitution patterns on
these 14 test set analogs (18% of the data set) are wide-ranging
and representative, as are their binding affinity values which range
from 5.3 to 8.8 log units—a difference of greater than three log
units. The training set corresponding to this test set contains the
complementary 66 analogs (80 ꢀ 14), including MP.
2.3. 2D-QSAR study results
Tables 4–6 give detailed results for the nMP 2D-QSAR model.
2.3.1. Descriptors
Table 4 lists the 12 most chemically-relevant, nonredundant
Molconn-Z descriptors along with the six nonredundant atom-le-
vel E-state indices of the scaffold atoms identified earlier. The
molecular shape and electrotopological descriptors are discussed
in detail in the Molconn-Z manual55 and elsewhere.44,45,56–58
Briefly, they encode the structure of a molecule by reducing the
molecular formula to a hydrogen-suppressed skeleton or ‘graph’.
Molecular shape indices, such as diam, muldiam, rad, and ishape,
characterize the molecular graph. Diam is the graph diameter, mul-
diam is the multiplicity of the diameter, rad is the graph radius,
and ishape is (diam ꢀ rad)/rad. Descriptors beginning with ‘S’ are
sums of atom-level E-state indices for all atoms of a particular type
in a given molecule. For example, SsF, SsCl, and SsI in Table 4 refer
to the sum of the atom-level E-state indices for the F, Cl, and I atom
types, respectively. SssO and SssS refer to similar sums for atom
types –O– and –S–. SsNH2, SaaN, and SddsN refer to corresponding
sums for N in –NH2, N with lone pairs in an aromatic ring, and a
planar nitrogen with three bonds such as in NO2, respectively.
SHdsCH refers to the sum of the atom-level E-state indices for all
hydrogens bonded to a double-bonded carbon (e.g., the H of
@CHR).
2. Results
The structures of the 80 molecules of the ( )-threo-MP analog
data set are listed in Table 1 along with their DAT binding affinities.
Binding affinity is expressed as an IC50 (that concentration of com-
pound required to inhibit the specific binding of [3H]WIN 35,428 in
control samples by 50%). All the molecules share an important fea-
0
0
ture: a common scaffold consisting of 14 atoms (N1–C6, C1 –C6 , C7,
with the 14th atom, C8, being the first carbon of the R3 side chain;
see structure in Table 1). MP is compound 39 with R1 = H, R2 = H,
and R3 = CO2CH3. These three sites of substitutions on MP (the
phenyl ring (R1), the piperidinyl nitrogen (R2), and the side chain
(R3)) produce the other 79 analogs in the table.
2.1. Atom-level E-state indices
The atom-level electrotopological state (E-state) indices43–52
are a measure of the atom’s electron accessibility and topological
accessibility in a molecule. They are identified by the numbering
scheme on the structure in Table 1. For example, ESN1 is the
These 18 descriptors were used as input to the partial least
squares (PLS) analysis of the pIC50 of the 66 nMP training set ana-
logs, where pIC50 = ꢀlog IC50 with IC50 given in molar units. The ta-
ble also lists the normalized coefficient and fractional contribution
of each descriptor to the PLS model. The table shows that, of the
0
atom-level E-state index of N1, ESC2 is that of C2, ESC1 is that of
0
C1 , etc. For both the nMP and pMP data sets, the atom-level E-state
indices of eight scaffold atoms (ESC2, ESC3, ESC4, ESC5, ESC6, ESC7,
0
0
0
ESC1 , ESC6 ) were found by correlation analysis to be redundant.
Table 2 identifies which indices were selected for the training
and test set studies, and lists the values of the E-state indices of
the 14-atom scaffold of nMP and pMP. Except for ESN1, the values
of the E-state indices are very similar for nMP and pMP. The differ-
ence in ESN1 values is due to the difference in chemical environ-
ment of neutral and protonated N1.
atom-level E-state indices, ESC2 makes the largest contribution
0
0
(22%) of all the descriptors, with ESC3 and ESC4 contributing about
11% each, and ESC5 only 3%. Taken together, the E-state indices of
0
the 20, 30, 40, and 50 positions of the phenyl ring contribute 46% to
the model. Their large contribution is perhaps not surprising given
that 32 out of the 66 training set analogs have substituents on the
phenyl ring (i.e., Group R1 of Table 1). In contrast, ESN1, where N1
is the point of R2 substitution, and ESC8, where C8 is the first car-
bon of R3 substituent, together contribute only an additional 4%
to the model. This indicates that the activity of the analogs is par-
ticularly sensitive to substitution on the phenyl ring. This is sup-
ported by the fact that the descriptor SsCl contributes an
additional 15% to the model. Of the 66 training set analogs, nine
have chlorine substituents on the phenyl ring (as the R1 substitu-
ents) and five have substituents containing chlorine at the R2 posi-
tion (see structure of MP at the top of Table 1 for the location of R1
and R2.).
2.2. Identification of test and training sets
The sphere exclusion procedure53,54 (see Section 4.1.1.4) re-
sulted in the same test set for both the nMP and pMP data sets,
consisting of analogs 11, 26, 27, 30, 32, 33, 34, 41, 56, 57, 62, 67,
71, and 80. It is interesting to note how the test set analogs are dis-
tributed with respect to changes in R1, R2, and R3, compared to the
full data set of 80 analogs. The analogs can be grouped based on all
possible combinations of substitutions with respect to the parent
compound, MP (compound 39 in Table 1 with R1 = H, R2 = H, and
R3 = CO2CH3). Table 1 also lists this information for each analog
in the column titled ‘Groups’. It is also possible to consider the data
set such that the substitutions occur at least at R1 (this would
group analogs for R1, R1 + R2, R1 + R3, and R1 + R2 + R3), at least
at R2 (R2, R1 + R2, R2 + R3, and R1 + R2 + R3), and at least at R3
(R3, R1 + R3, R2 + R3, and R1 + R2 + R3). These additional groupings
Taken together, the atom-level E-state indices contribute about
50% to the model. They can be used to provide a qualitative inter-
pretation of the SAR data in Table 1. The table shows that analogs
that have a 20 position substituent on the phenyl ring have very
poor binding affinities (e.g., analogs 45, 55, and 56). This can be ex-
plained by examining Eq. (2) and the sign of the coefficients in col-
umn two of Table 4. In Eq. (2), the d value is in the denominator. By
definition, replacing the hydrogen substituent on the 20 position
*
*
*
are shown in Table 3 under R1 , R2 , and R3 , respectively.
Table 3 shows how the full data set (minus MP) is distributed
with respect to changes at R1, R2, and R3. The table shows that
carbon with a heavy atom increases that carbon’s d value, thereby
0
0
decreasing its E-state index (ESC2 ) relative to that of the 2 position

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7223
Table 1
Methylphenidate analog data seta
4'
R1
3'
5'
6'
2'
1'
5
2
8
R3
6
7
4
3
N
1
R2
R1b
R2
R3c
IC50 (nM)
pIC50
6.37
Groupse
R2
d
-CH2
N
C
S
1
–H
–CO2CH3
422 13
2
3
4
5
6
–H
–H
–H
–H
–H
–(CH2)3Ph
–CH2C„CH
–(CH2)4Ph
–(CH2)2Ph
–(CH2)3Ph
–CH2OH
–CO2CH3
–CH2OH
–CH2OH
–CO2CH3
194 14
821 100
623 64
1430 150
267 13
6.71
6.09
6.21
5.84
6.57
R2 + R3
R2
R2 + R3
R2 + R3
R2
Cl
7
–H
–CO2CH3
106 24
6.98
R2
-CH2
8
9
10
11
–H
–H
–H
–H
–(CH2)2Ph
–(CH2)4Ph
–(CH2)5Ph
–(CH2)6Ph
Cl
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
678 46
205 44
1570 80
656 17
6.17
6.69
5.80
6.18
R2
R2
R2
R2
12
–H
–CO2CH3
243 40
6.62
R2
-CH2
13
14
–H
–H
–CH2CH@CH2
–CO2CH3
–CO2CH3
597 4.0
31.2 5.7
6.22
7.51
R2
R2
-CH2
-CH2
-CH2
Cl
NO2
15
16
17
–H
–H
–H
–CO2CH3
–CO2CH3
–CO2CH3
113 3.0
79.1 1.4
392 15
6.95
7.10
6.41
R2
R2
R2
OCH3
S
-CH2
Cl
-CH2
18
–H
–CO2CH3
369 4.0
6.43
R2
N
-CH2
-CH2
19
20
–H
–H
–CO2CH3
–CO2CH3
173 15
128 13
6.76
6.89
R2
R2
N
N
O
-CH2
21
22
23
24
–H
–H
–H
–H
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
536 38
143 25
224 1.0
459 67
6.27
6.85
6.65
6.34
R2
R2
R2
R2
-CH2
S
S
-CH2
-CH2
O
25
26
27
–H
30,50-diCH3
30,50-diCl
–CH2CH3
–H
–H
–CH2OH
–CO2CH3
–CO2CH3
2340 780
4690 60
65.6 5.4
5.63
5.33
7.18
R2 + R3
R1
R1
Cl
28
–H
–CH2OH
25.8 0.20
7.59
R2 + R3
-CH2
(continued on next page)

7224
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
Table 1 (continued)
d
R1b
R2
R3c
IC50 (nM)
pIC50
Groupse
29f
30
31
32
–H
–H
–CH2Ph
–H
–H
–CO2CH3
–CH2OH
–CO2CH3
–CO2CH3
52.9 2.3
448 8.0
98.0 10
620 40
7.28
6.35
7.01
6.21
R2
R3
R1
R1 + R2
g
40-OH
30-CH2OH,
40-OCH2OH
40-OH
40-NO2
30-NH2
40-NH2
40-OCH3
40-Cl
–CH3
33
–CH3
–H
–H
–H
–H
–H
–H
–H
–CH3
–H
–H
–CH3
–H
–H
–H
–H
–H
–H
–H
–H
–H
–H
–H
–H
–H
–H
–H
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–CH2OH
1220 140
494 33
265 5.0
34.5 4.0
83.0 11
20.6 3.4
83.0 7.9
13,500 450
160 18
5.92
6.31
6.58
7.46
7.07
7.69
7.08
4.87
6.79
7.85
8.38
6.85
5.73
4.00
6.54
4.64
6.49
8.29
7.46
8.28
6.09
8.16
5.71
5.85
7.39
7.48
7.67
6.55
8.38
8.77
7.39
5.76
6.42
7.63
7.75
7.01
7.12
8.57
8.38
7.79
5.99
6.97
6.16
5.76
6.30
6.13
7.96
6.21
R1 + R2
R1
R1
R1
R1
R1
—
R1
R1 + R2
R1
R1
R1 + R2
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1 + R3
R1 + R3
R1 + R3
R1 + R2
R2 + R3
R2 + R3
R2 + R3
R2 + R3
R3
R1 + R2
R1 + R2 + R3
R1 + R2 + R3
R1 + R3
R3
34g
35g
36g
37g
38g
39g
40g
41h
42g
43g
44f
45g
46g
47g
48g
49g
50g
51g
52g
53g
54g
55g
56g
57g
58g
59g
60
–H
40-t-Butyl
30-Cl
40-I
14.0 0.0
4.18 0.17
140 9.0
30-Br
40-CH3
20-Br
1870 135
101,000 10,000
288 52
23,100 50
321 1.0
5.10 1.6
35.0 3.0
5.30 0.70
810 10
6.90 0.10
1950 230
1420 120
40.5 4.5
33.0 1.2
21.4 1.1
281 32
4.20 0.52
1.70 0.24
41.2 3.4
1730 52
384 8.0
23.7 3.3
17.8 1.1
97.1 10.3
76.3 2.7
2.74 0.35
4.20 1.2
16.4 2.4
1020 130
108 16
20-OCH3
30-OCH3
20-OH
30-OH
30-Cl
40-F
30,40-diCl
30,40-diOCH3
40-Br
20-Cl
20-F
30-F
40-CH3
30-CH3
30-F
–H
–H
–H
61
62
63
64
30,40-diCl
30,40-diCl
30-Cl
–CH2OH
–CH2OCH3
–CO2CH3
–CON(CH3)2
–CONH2
–CH2Ph
–CH2Ph
–CH2Ph
–CH2Ph
–CH2Ph
–H
–CH2Ph
–CH2Ph
–CH2Ph
–H
–H
–CH3
–H
–H
–CH3
–H
–H
–H
–H
–H
65
66h
67h
68
–CH2OH
–CH2OCH3
–CH2OCH3
–CO2CH3
–CH2OH
–CH2OCH3
–CONH2
–CO2CH2Ph
–CO2CH3
–CH2O(CO)CH3
–CONH2
–CO2CH3
–CO2CH3
–CO2CH3
–CO2CH3
–H
69
70
71
72
73
74
75
30,40-diCl
30,40-diCl
30,40-diCl
30,40-diCl
–H
30-CH3
–H
–H
R1 + R2
R3
R3
R2
R1
690 270
1730 170
499 25
737 78
11.0 2.5
615 15
76
77h
78
–H
40-C2H5
30,40-Benzoj
40-CF3
79i
80i
–H
–H
R1
R1
a
Binding data were generated in the laboratory of MMS; values are expressed as the mean standard error of the mean (SEM) of 2–7 assays for each compound.
Methylphenidate is compound 39. Compounds 42, 43, 45, and 54 were provided by Dr. S. J. Gatley of Brookhaven National Laboratories.
b
Naming convention for R1 substituents is based on the six positions of the phenyl ring: for example, 48 (20-OH MP) has the –OH substituent at the 20 position of the
phenyl ring. For R2 substituents, Ph = phenyl.
c
The first carbon in the R3 substituent is the 14th atom of the scaffold, C8.
d
pIC50 = ꢀlog IC50, with IC50 in molar units.
e
‘Group’ refers to the substituents R1, R2, and/or R3, that have been modified compared to MP (39).
f
Synthesis and binding studies have been described elsewhere.16
Synthesis and binding studies have been described elsewhere.12
g
h
Synthesis and binding studies have been described elsewhere.19
i
Synthesis and binding studies have been described elsewhere.17
j
‘Benzo’ refers to a 1,2-disubstituted benzene ring (30,40-benzo = beta naphthyl analog).
0
0
carbon atom in MP. In the regression equation, ESC2 is multiplied
greater than (and its E-state index (ESC3 ) is less than) that of the
0
by the coefficient 1.017, so if it is assumed that the only change in
same carbon in MP. However, in the regression equation ESC3 is
0
0
the 2 position analog relative to MP is in the value of ESC2 , then
multiplied by a negative coefficient, ꢀ0.313. This results in the pre-
diction that (IC50 of a MP analog with a 30 position substituent) <
(IC50 of MP), in agreement with much of the experimental data
in Table 1. A similar analysis pertains to ESC4 and ESC5 . These
E-state indices are also multiplied by negative coefficients in the
(IC50 of a MP analog with a 20 position substituent) > (IC50 of MP).
In contrast, many analogs with 30 position phenyl ring substituents
have better binding affinity than MP (e.g., 57 and 63). As above, the
d value of a 30 position carbon with a heavy atom substituent is
0
0

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7225
Table 2
Table 4
Atom-level E-state indices of scaffold atoms for neu-
tral and protonated methylphenidate
Final set of 2D-QSAR descriptors for the nMP data set
Descriptor Coefficient in
regression equationa
Normalized
coefficientb
Fractional
contribution to
model
E-State
index
nMP
pMP
Redundanta
ESN1
ESC2
ESC3
ESC4
ESC5
ESC6
ESC7
ESC8
3.429
4.589
Diam
Muldiam
Ishape
SHdsCH
SsNH2
SaaN
SddsN
SssO
SsF
SssS
SsCl
SsI
ꢀ0.000
0.001
0.083
0.177
0.008
0.052
0.004
0.042
0.097
0.039
0.031
0.276
0.125
0.022
0.407
0.212
0.201
0.053
0.047
0.000
0.044
0.094
0.004
0.028
0.002
0.022
0.051
0.021
0.017
0.147
0.066
0.012
0.217
0.113
0.107
0.028
0.025
0.960
1.169
1.148
0.997
0.147
ꢀ0.264
ꢀ0.187
0.950
1.915
0.766
1.881
1.910
1.959
0.835
1.114
1.117
0.941
0.022
ꢀ0.319
ꢀ0.219
0.918
1.895
0.752
1.871
1.896
1.939
X
X
X
X
X
X
ꢀ0.139
ꢀ0.317
0.007
ꢀ0.031
0.004
ꢀ0.787
ꢀ0.029
ꢀ0.016
0.079
0.068
0.403
0.039
1.017
ꢀ0.313
ꢀ0.309
ꢀ0.370
0.220
0
ESC1
X
0
ESC2
0
ESC3
0
ESC4
0
ESC5
ESN1
0
0
ESC6
X
ESC2
0
ESC3
a
‘X’ indicates
deleted from the final set.
a redundant descriptor that was
0
ESC4
0
ESC5
ESC8
a
Intercept = 6.968.
regression equation. The model suggests that substitutions at these
positions will lead to lower IC50’s relative to MP. As Table 1 shows,
many of the analogs with the best binding affinities (such as ana-
logs 43, 50, 52, and 54) are those that are substituted at these posi-
tions. There are some exceptions to this analysis, indicating that it
is too simplistic to completely describe the data set. This supports
the use of the more detailed CoMFA study described below.
b
Normalized coefficient is the scaled actual coefficient such that the relative
contributions of descriptors become apparent.
the model would not be able to use the orthogonality of the elim-
inated descriptors to explain the associated variance. Indeed, when
the 29 original descriptors were combined with the set of six non-
redundant E-state indices, the resulting model explained as much
as 74% of the variance. However, this model was considered unsat-
isfactory because it lacked interpretability due to the presence of a
large number of descriptors. This was the reason for conducting
variable selection by retaining only the 12 chemically-relevant
Molconn-Z descriptors described above.
2.3.2. Model validation
In Table 5, the non cross-validated models were produced for
the optimal number of components obtained by the cross-vali-
dated models.The probability values, or p-values, for all non-
cross-validated runs are not shown because they were always very
small (0.000 to three decimal places). Thus, the probability that r2
equals zero is itself zero (or very small), signifying the existence of
a relationship between the descriptors used in a model and the
binding affinity. A stable or robust model is one with high Q2,
low cSDEP, and a slope, dq2=dry2y0 , near 1.0. Such a model will be af-
fected in proportion to the magnitude of the change in the under-
lying pIC50 activity values. The Q2 and cSDEP statistics tend to be
inversely related. The Q2 value is a conservative statistic and will
usually be low because it is based on randomized or noisy activity
values.
Table 5 also lists the results of progressive scrambling59 for the
2D nMP model. The high value of dq2=dry2y0 near 1.0 suggests that
this is a stable model, that is, a change in the underlying activity
values would produce a proportional change in this model. Valida-
tion of the 2D-QSAR model using the test set activity prediction is
shown in Table 6. Three (27, 41, and 57) out of the five best-mod-
eled compounds (with residuals less than 0.5) are esters with a hal-
ogen substituent at the 30 position of the phenyl ring. The largest
residual is for analog 26, which is predicted poorly by the CoMFA
models as well (see below). The root mean square deviation
(RMSD) of the residuals is 0.98.
Table 5 shows that the 2D-QSAR model (q2 = 0.358) explains
only 55% of the variance in the training set data. The predicted
pIC50’s and associated residuals (actual value ꢀ predicted value)
are given in the Supplementary data. For 23% of the training set,
the absolute value of the residual was greater than 0.5; for 9% of
the data set, it was between 1.0 and the maximum value, 1.8. How-
ever, it should be noted that prior to obtaining this model, 17
descriptors were eliminated from the original set of 29 nonredun-
dant descriptors obtained after correlation analysis. This would re-
duce the explanatory power of the 2D-QSAR model in Table 5 as
2.4. 3D-QSAR studies
The results for the Comparative Molecular Field Analysis
(CoMFA)60 PLS models for the nMP and pMP data sets are given
in Table 5. Both models explain 92% of the variance in the training
set data and both have q2 greater than 0.5. The predicted pIC50’s
and residuals for the training set analogs for both models are given
in the Supplementary data. For the nMP model, only 6% of the
Table 3
Distribution of data set and test set analogs by substituent groupsa
*
*
*
R1
R2
R3
R1 + R2
R1 + R3
R2 + R3
R1 + R2 + R3
Total
R1
R2
R3
Data setb
Test setc
29
6
23
1
5
1
7
3
4
1
9
1
2
1
79
14
42
11
41
6
20
4
a
Analogs that belong to the groups R1, R2, and R3 have a substitution (relative to MP, 39) only at these locations; analogs belonging to the groups (R1 + R2), (R1 + R3), and
(R2 + R3) have substitutions at both locations; analogs belonging to the group (R1 + R2 + R3) have substitutions at all three locations. Analogs belonging to groups R1*, R2*,
and R3* have at least one substitution at R1, R2, and R3, respectively. See Table 1 for identification of groups R1, R2, and R3.
b
The data set consists of the 79 analogs other than MP in Table 1.
The test set consists of analogs 11, 26, 27, 30, 32, 33, 34, 41, 56, 57, 62, 67, 71, and 80.
c

7226
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
Table 5
2D- and 3D-QSAR studies: resultsa and model validation
b
c
d
e
f
g
h
i
q2
SDEP
Compd
r2
SEE
F-Value
Q2
cSDEP
2
dq2=dryy0
j
Model
Data set
2D
3D
3D
nMP
nMP
pMP
0.358
0.556
0.568
0.734
0.631
0.622
2
6
6
0.550
0.924
0.918
0.614
0.261
0.271
38.548
119.396
109.644
0.251
0.407
0.537
0.791
0.727
0.644
0.845
0.617
0.759
a
b
c
d
e
f
See Section 4.1 for a complete description of parameters and associated mathematical formulae.
Cross-validated r2 statistic.
Standard error of prediction after leave-one-out cross-validation.
Optimal number of components.
Coefficient of determination.
Standard error of estimate.
Tests null hypothesis that none of the descriptors has any effect on the binding affinity.
Expected value of q2 at the user-specified critical point in Progressive Scrambling.
Estimated cross-validated standard error at the user-specified critical point.
g
h
i
j
Slope at the critical point, evaluated with respect to the correlation of the original biological activities versus the scrambled activities.
Table 6
Test set activitya predictions
Analog
Actual
2D-QSAR (nMP)
3D-QSAR (nMP)
3D-QSAR (pMP)
Predicted
Residualb
Predicted
Residualb
Predicted
Residualb
11
26
27
30
32
33
34
41
56
57
62
67
71
80
6.18
5.33
7.18
6.35
6.21
5.92
6.31
6.79
5.85
7.39
8.77
7.75
8.38
6.21
6.65
7.24
7.48
6.78
7.36
7.17
7.21
6.95
4.58
7.04
7.69
6.92
7.68
7.42
ꢀ0.47
ꢀ1.91
ꢀ0.30
ꢀ0.43
ꢀ1.15
ꢀ1.26
ꢀ0.91
ꢀ0.15
1.27
0.35
1.08
0.83
0.70
6.50
7.45
8.21
6.59
4.01
6.69
7.83
7.16
6.21
7.08
8.54
7.71
9.17
7.71
ꢀ0.32
ꢀ2.13
ꢀ1.02
ꢀ0.24
2.20
ꢀ0.78
ꢀ1.52
ꢀ0.36
ꢀ0.36
0.31
5.95
7.02
7.89
6.62
6.58
6.72
8.36
6.66
6.18
6.73
8.69
7.25
8.32
7.27
0.23
ꢀ1.69
ꢀ0.71
ꢀ0.28
ꢀ0.38
ꢀ0.80
ꢀ2.05
0.14
ꢀ0.33
0.66
0.23
0.04
0.08
0.50
0.06
ꢀ1.05
ꢀ0.79
ꢀ1.21
ꢀ1.50
a
Activities are expressed as pIC50 = ꢀlog IC50 with IC50 in molar units.
Residual = actual value ꢀ predicted value.
b
training set had a residual with absolute value between 0.5 and the
maximum value, 0.6. For the pMP model, only 4% of the training set
had a residual with absolute value between 0.5 and the maximum
value, 0.85. These results are in contrast to the 2D-QSAR nMP mod-
el for which 23% of the training set had an absolute value of the
residual greater than 0.5. Both 3D-QSAR models have a high F-va-
lue, implying that there is a good chance that a relationship exists
between the CoMFA predictors (steric and electrostatic field val-
ues) and bioactivity measured as pIC50. The (percent steric)/(per-
cent electrostatic) contribution to each model is 46/54 for nMP
and 81/19 for pMP. This is a somewhat counterintuitive result as
it can be imagined that the electrostatic contribution would be
greater for the protonated model than for the neutral model.
2.4.2. Prediction using models
The nMP and pMP CoMFA models are used to predict the IC50 of
a MP analog, methyl 2-(naphthalen-1-yl)-2-(piperidin-2-yl)ace-
tate, not in the original data set. The residuals are predicted to be
ꢀ0.20 and 0.33 by the nMP and pMP models, respectively. This cor-
responds to predicted IC50 values of 448 nM (nMP model) and
1515 nM (pMP model), which are in good agreement with the
experimental value of 716 55 nM.17 Since the residuals are small,
both models appear to be useful in predicting the IC50 value of a
compound closely related to those in the training set.
2.4.3. Interpretation of R1 (phenyl ring substitution) SAR using
CoMFA maps
Since most of the SAR changes that lead to significant altera-
tions in binding affinity involve substitution on the phenyl ring,
this section will focus on the R1 region. Discussion of the SAR of
the R2 and R3 regions is given in Section 3.3. Figure 1 shows the
steric and electrostatic CoMFA contour maps for the nMP CoMFA
model. The pMP maps are very similar and are not shown. In Fig-
ure 1, two orientations were chosen to display the CoMFA contours
of the R1 region to best advantage: a ‘full view’, with the phenyl
ring in the plane of the paper and a ‘side view’, with the molecule
rotated 90ꢀso that the phenyl ring is viewed from the side. The
maps are essentially a color-coded, three-dimensional summary
of the structure–activity relationships of the 66 training set ana-
logs in Table 1. The structures displayed in the maps are those of
the test set analogs (Table 6). In the steric maps, the green contours
indicate regions where the addition of bulky groups leads to better
2.4.1. Model validation
The results of progressive scrambling are listed in Table 5 for
the nMP and pMP 3D-QSAR models. The nMP CoMFA model is sta-
ble because its Q2 is large (0.407) and its slope is fairly good
(dq2=dry2y0 ¼ 0:617). The pMP CoMFA model has a higher Q2 value
(0.537) and a slope nearer to 1.0 (dq2=dry2y0 ¼ 0:759). The predicted
activities of the test set analogs are listed in Table 6. For the nMP
model, all but two analogs (26 and 32) were predicted with resid-
uals less than 2 units. For the pMP model, just one analog (34) had
a residual greater than 2 units. The qualitative trends in the test set
residuals are generally similar to those of the 2D nMP model. The
RMSD of the residuals is 1.09 and 0.86 for the nMP and pMP mod-
els, respectively, indicating that the 3D pMP model predicts more
accurately than both the 2D and 3D nMP models.

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7227
and 50 of the phenyl ring. Analogs 27, 38, 42, 43, 50, 51, 52, 54,
57, and 79 all exhibit the appropriate steric bulk and electron den-
sity in these regions. Analog 79 has electron density in that region
due to the aromatic ring fused between positions 30 and 40 of the
phenyl ring, while the other analogs have halogen substituents.
The fact that 27, 38, 42, 43, 50, 51, 52, 54, 57, and 79 exhibit the
combination of steric and electrostatic requirements shown in Fig-
ure 1 is consistent with their improved binding affinity compared
to MP. In contrast, 26 has the same steric bulk of 27 (at the 30 and 50
positions) but does not match the electrostatic requirement at
those positions. Analog 26 has poor binding affinity compared to
MP.
Beyond the green region off position 40 in the Figure 1a steric
map lies a yellow region, out of the plane of the aromatic ring,
which indicates that large bulky groups such as those of 40 and
78 are unfavorable to DAT binding. This out-of-plane orientation
is best noted in the side view of Figure 1b. Similarly, the broad yel-
low region off the 20 position (Fig. 1a) indicates that substitution in
this region is not favorable to binding, as seen in 45, 46, 48, 55, and
56. The 30 position tolerates some, but not all, substituents as seen
by the unfavorable yellow contours (due to bulkier groups like the
–OCH3 substituent of 47) lying beyond the green contours (due to
the –Cl (50) or –CH3 (59) substituents). Since the methoxy groups
are free to rotate, they can occupy the unfavorable yellow regions,
described above, which are out of the plane of the aromatic ring.
In summary, halogen substituents at R1 on the phenyl ring, par-
ticularly 30,40-diCl, improve the DAT binding affinity of this series of
MP analogs. The steric contour maps show the presence of yellow
regions (out of the plane of the aromatic ring) where bulky groups
are disfavored. The electrostatic maps show that the
p-electron
cloud above and below the plane of the phenyl ring lies in the
red region where higher electron density is correlated with better
binding affinity. This indicates that, in the design of new com-
pounds, the optimal substituents may be those that lie in the plane
of the phenyl ring and contain electron-withdrawing groups in the
general area of the 30 and 40 positions. The 30,40-benzo analog has
an aromatic group that extends into the favorable green region be-
tween the 30 and 40 positions. An area that might be fruitful for
exploration is that of halogen substituents on the 30,40-benzo
analog.
Figure 1. CoMFA maps for nMP (steric on left, electrostatic on right) based on the
training set of 66 analogs. Substituent regions (defined in the schematic figure
accompanying Table 1) are indicated: R1 (on phenyl ring), R2 (on nitrogen), and R3
(side chain). The 20, 30, 40, 50, and 60 positions on the phenyl ring are noted. The
superimposed test set analogs are displayed. Contribution levels: green/yel-
low = 70/40 and blue/red = 65/25. (a) Full view with phenyl ring in plane of paper.
(b) Side view with phenyl ring perpendicular to plane of paper.
binding affinity, whereas the yellow contours indicate regions
where steric bulk is not well tolerated. In the electrostatic maps,
the blue contours indicate regions where positively charged groups
enhance binding affinity, whereas the red contours are regions
where negatively-charged groups improve binding affinity. In
other words, electron-withdrawing groups are favored as substitu-
ents in the red, but not the blue, regions.
3. Discussion
3.1. Test set selection
Considering analogs with changes in R1 only (see Table 1), it is
clear that those analogs that have an electron-withdrawing halo-
gen substituent in the 30 and/or 40 position of the phenyl ring
(i.e., analogs 38, 42, 43, 50, 51, 52, 54, and 57) have better binding
affinity than MP, as does 27, which has a chlorine atom at both the
30 and 50 positions. Fusion of an aromatic ring onto the 30 and 40
positions, as in the 30,40-benzo analog, 79, also gives improved
affinity. The 40 position tolerates the addition of –OH (31), –NH2
(36), –OCH3 (37), and –CH3 (58) substituents, but not groups as
large as –C2H5 (78) or t-butyl (40). In contrast to their effect at
the 40 position, the substituents –OH (49), –NH2 (35) and –OCH3
(47) at the 30 position result in poorer binding affinity compared
to MP. This is seen also in the 30,40-diOCH3 analog (53). However,
a –CH3 substituent at the 30 position (59) gives improved binding
affinity. All substitutions at the 20 position result in poor binding
affinity (see 45, 46, 48, 55, and 56).
The test set identification was based upon the selection of the
most orthogonal E-state indices. The structural information en-
coded by the E-state indices allowed the identification of a truly
representative subset of analogs that span not only the structure
space but also the activity space. Since the particular test set iden-
tification technique used in this work is deterministic (i.e., for a gi-
ven set of input parameters, it identifies the same test set for a
given series of analogs), it may eliminate the need for generating
a whole collection of test and training sets and repeating model
development for each different combination. The input parameters
for calculating a test set (described in Section 4.1.1), such as the
pairwise correlation coefficient cutoff value, r2max, used in the unsu-
pervised forward selection (UFS) procedure,61 as well as the
threshold radius, t, and the particular dissimilarity-based com-
pound selection (DBCS) algorithm62,63 used, may be varied to suit
the modeler’s preferences. Two other test sets were selected using
the following combinations (r2max, t, DBCS algorithm) = (0.99, 0.10,
SE-MinMax) and (0.85, 0.15, SE-MinMax), where SE-MinMax63 is
a selection criterion for sphere exclusion (SE) in which the com-
pound with the smallest maximum dissimilarity is selected. The
These trends are summarized in the CoMFA maps. The steric
map on the left of Figure 1a shows the presence of favorable green
regions off positions 30, 40, and 50 of the phenyl ring, as well as
between these positions. The electrostatic map on the right of
Figure 1a shows the presence of red regions off positions 30, 40,

7228
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
distribution of resulting test set analogs among the various groups
in Table 3 was not as representative of the full data set, as was that
shown in the table, derived with (r2max, t, DBCS algorithm) = (0.85,
0.10, SE-MinMax).
substitutions at R2). It is clear from Table 1 that having more than
one methylene group in the substituent –(CH2)xPh leads to poor
binding affinity. Compare 29, a good binder with x = 1, to the poor
binders with increasing chain length x = 2–6 (i.e., 8, 6, 9, 10, and
11). The effect of chain length shows up in the steric maps as yel-
low contours in the R2 region of Figure 1. Similarly, the effect of
substitution on the phenyl ring of –CH2Ph is observed by compar-
ing the binding affinity of 29 to that of 12, 7, and 14 with –Cl in the
ortho, meta, and para positions, respectively. Table 1 shows that
only substitution at the para position is favorable, as indicated by
the green contour off that position in Figure 1.
Four pairs of compounds allow comparison of the effect of R2
substitution when the original phenyl ring of MP contains substitu-
tions, but the ester at R3 remains unchanged. In all four cases,
when the R1 substituent is held constant, introduction of a substi-
tuent at the nitrogen decreases the affinity of the resulting com-
pound by 4.2- to 31-fold (compare 58 vs 44, 31 vs 33, 52 vs 69,
and 50 vs 41).
Only five analogs have changes solely at R3 (30, 68, 73, 75, 76).
On average, these compounds were 10-fold less active than MP.
The order of increasing activity for those compounds with small
substituents at R3 [76 (amide at R3) < 30 (alcohol) < 68
(ether) ꢁ 39 (ester, MP)] mostly parallels the order of decreasing
polarity, indicating that electronic effects are important. This also
agrees with literature results for the even more polar carboxylic
acid (ritalinic acid, 2-phenyl-2-(piperidin-2-yl)acetic acid) which
is reported to have an IC50 >10,000 nM in another test system.10
However, the poor activity of ester analogs with more bulky groups
at R3 (73 and 75) indicates that not much steric bulk can be accom-
modated at the R3 position.
Of the nine analogs with substituents at both R2 and R3,
those with –CH2Ph at R2 and –CH2OH (66) or –COCH3 (67) at
R3 give improved binding affinity, as does 28 with a chlorine
at the meta position of the R2 phenyl ring and R3 = CH2OH. Of
the four analogs that have changes both at R1 and R3, those that
have 30,40-diCl at R1 (61, 62, and 72) have improved binding
affinity compared to MP, whereas 60 (with 30F at R1) does not.
In fact, with the notable exception of 69, all analogs that have
30,40-diCl at the R1 position (52, 61, 62, 70, 71, and 72) have
greatly improved affinity compared to MP, regardless of their
R2 and R3 substituents. Similarly, of the seven analogs that have
changes at R1 and R2 (32, 33, 41, 44, 63, 69, and 74), those with
–Cl substituents at the 30 position of the R1 phenyl ring and –
CH2Ph at R2 (63 and 69) have equivalent or better binding affin-
ity compared to MP. As in the discussion of R1 substituents, very
bulky groups at the 30 or 40 position are not well tolerated as
shown by the poor binding affinity of 32.
The in-house program D-SIM version 1.0 (see Section 4.1.1.4) al-
lows the user to vary the value of t and offers a choice of three
sphere-exclusion algorithms. The program output includes a log
file with a record of the order in which the analogs are selected
in the test set. This information can give some idea about the dis-
tribution of the analogs as points within the descriptor space. Thus,
analogs that are selected in succession would be closer to each
other in the descriptor space. Also, analogs that are selected earlier
would lie toward the center of the descriptor space. As an example,
in this work, the order of selection for the test set was 11, 80, 67,
41, 56, 32, 71, 27, 33, 57, 62, 30, 26, and 34. It could be imagined
that, as it nears completion, the sphere exclusion algorithm would
select the analogs located near the extremes of the descriptor
space. If the descriptor space is sparse (due to the series of analogs
being ‘incomplete’), it is likely that the analogs that are included
late in the test set might be outliers. Note that analog 34 in the test
set was the last to be selected and had the largest residual from the
pMP PLS model. Analog 26, which was selected just before analog
34, also had consistently large residuals.
3.2. Preliminary models
The work described herein was not intended to be a compre-
hensive comparison of 2D descriptor versus 3D CoMFA models,
nor an attempt to identify an optimal variable selection technique.
These studies are available elsewhere.54,64–71 Rather the goal of the
present work was to use complementary 2D- and 3D-QSAR model-
ing to identify the most significant molecular features that relate to
the DAT binding affinity of a set of 80 MP analogs. The simple, pre-
liminary models produced by these 2D- and 3D-QSAR studies iden-
tified phenyl ring modification as a major determinant of DAT
binding affinity and suggest that region as a focus area for future
modeling studies. However, it should be noted that 42 out of 80
compounds in the data set have phenyl ring substitutions, so that
the data set itself contains more information on the SAR of this re-
gion than others.
0
For the 2D-QSAR model, ESC2 contributes by far the largest
0
0
amount (22%) to the model, followed by ESC3 and ESC4 at 11%
apiece and ESC5 at 3%. This indicates that substitution at the 20,
0
30, 40, and 50 positions of the phenyl ring can dramatically affect
DAT binding affinity, accounting for almost 50% of the model. Ex-
cept for ESN1, the E-state index of the nitrogen, which is either
neutral or protonated depending on the model, the E-state indices
are very similar for nMP and pMP.
The nMP and pMP CoMFA models both explain 92% of the var-
iance in the training set data and have q2 greater than 0.5. Both
models are stable. The CoMFA maps identify regions that cannot
tolerate steric bulk to be those that lie off the 20 position of the phe-
nyl ring, as well as above and below the phenyl ring near the 30, 40,
and 50 positions. Suggested alterations to be explored in the design
of new compounds are the placement at the 30 and 40 position of
the phenyl ring of electron-withdrawing groups that lie chiefly in
the plane of the ring, for example, halogen substituents on the
30,40-benzo analog.
3.4. Future work
The results of CoMFA studies are known to be sensitive to the
template conformer used for alignment of all the analogs.72 Only
the global energy minimum (GEM) conformer was used as a tem-
plate in the present study. Yet, it is known that molecules do not
necessarily bind to proteins in their GEM conformations73–77 and
the importance of considering conformers other than the GEM in
pharmacophore modeling has been demonstrated.42,78–84 There-
fore, the next step in the development of a 3D-QSAR model for
the DAT binding of MP analogs will be to include alternative con-
formers as templates for MP CoMFA studies. Our recent conforma-
tional analysis of nMP and pMP26 identified such conformers by
comparison of MP conformers to the structure of a rotationally re-
stricted analog8 that has the same DAT binding affinity as MP.18
Therefore the preliminary studies described here pave the way
for our more extensive CoMFA studies of MP analogs with only
phenyl ring substitutions.
3.3. Additional SAR insights
Considering the 23 analogs with changes in R2 only, it appears
that substitution at the piperidinyl nitrogen usually decreases the
affinity of the resulting compounds. Twenty of the 23 compounds
exhibit losses of potency ranging from 1.3- to 19-fold compared to
MP; the three exceptions are 14, 16, and 29 (all contain benzyl (Bn)

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7229
atoms of a particular atom type.58 The E-state index for an atom
in a molecule is a measure of both the electron accessibility and
the topological accessibility of that atom.
4. Methods
4.1. Computer modeling
The atom-level E-state indices for the 14 scaffold atoms were
calculated by the Molconn-Z module in SYBYL. The scaffold atoms
were numbered identically in all analogs. For each molecule, Mol-
conn-Z was used to generate a ‘.S’ file containing the atom-level E-
state indices of all atoms in the molecule. The E-state indices for
the 14 scaffold atoms were extracted from each .S file and used
for all subsequent calculations.
SYBYL55 molecular spreadsheets were constructed for two types
of data sets: neutral analogs (nMP) and analogs protonated on N1
(pMP). The nMP data set was used for the 2D-QSAR studies; both
data sets were used for the CoMFA studies. The 2D-QSAR study in-
volves generation of hundreds of conformation-independent
descriptors, followed by variable selection through complete corre-
lation analysis to identify a set of least redundant descriptors. The
3D-QSAR CoMFA study uses steric and electrostatic conformation-
dependent descriptors. In these studies, PLS analysis relates the
DAT binding affinity of the 80 MP analogs to structural alterations
of a common 14-atom scaffold.
4.1.1.2. Variable selection: identification of redundant E-state
descriptors.
Complete correlation analysis was performed
on the unscaled atom-level E-state indices of the 14 scaffold
atoms to select an ‘ideal’ or ‘minimal’ set of descriptors to be used
for the calculation of the dissimilarity matrix. A program devel-
oped by Whitley et al.61 (available from http://www.vcclab.org/
lab/ufs/), which implements the unsupervised forward selection
(UFS) algorithm,61 was used to select the least redundant (or
most orthogonal) descriptors. This algorithm first selects the
two descriptors that have the smallest pairwise correlation coef-
ficient. Next, it rejects each of the remaining descriptors whose
pairwise correlation coefficient with the first two descriptors ex-
ceeds a user-specified value, r2max <1. In the present work, r2max
was set equal to 0.85. The algorithm then iterates until all
descriptors are either selected or rejected. A descriptor is selected
if it has the smallest squared multiple correlation coefficient (of
the present set of as-yet-unselected descriptors) with the previ-
ously-selected descriptors. All descriptors that have squared mul-
tiple correlation coefficients with currently-selected descriptors
greater than r2max are rejected. Since the selection of additional
descriptors is based upon their correlation with those already
chosen, the algorithm builds a subset of descriptors that is as
orthogonal as possible. Six atom-level E-state indices were se-
lected by the UFS procedure. These were then range-scaled for in-
put to the calculation of the dissimilarity matrix below. Range
scaling was performed as
4.1.1. Use of atom-level E-state indices in test set selection
The identification of a test set of analogs is an important part of a
model validation strategy.85,86 However, care must be taken to se-
lect the analogs for constructing the test set. The test set analogs
should span both the structure space as well as the activity space.
Maw and Hall27 demonstrated the utility of using the atom-level
E-state descriptors of a common 19-atom tropane scaffold in their
2D-QSAR study of a data set of 25 tropane dopamine reuptake inhib-
itors. Here we use the atom-level E-state indices of the common 14-
atom scaffold of MP analogs in a different manner: to identify a test
set that spans descriptor space for use in both 2D- and 3D-QSAR
studies of the MP data set (see Table 1), the assumption being that
the variations in the binding affinity may be correlated with the
variations in structure due to substitutions on the scaffold atoms.
First, variable selection through correlation analysis was used to
delete redundant E-state descriptors from the set of 14. Next, the
remaining, nonredundant E-state indices were range-scaled and
used to construct a dissimilarity matrix, D, the elements of which
are the complements of pairwise Tanimoto coefficients.87 This dis-
similarity matrix was then used along with a sphere exclusion algo-
rithm53,54 for identifying
a deterministic test set using a
dissimilarity-based compound selection (DBCS) algorithm.62,63 This
ensures that this procedure will always identify the same test set gi-
ven the same input parameters. The training set was defined as the
remaining analogs. The same test and training sets were used for the
2D- and 3D-QSAR studies. The steps are described in detail below.
xi ꢀ xmin
zi ¼
ð4Þ
xmax ꢀ xmin
where zi is the scaled descriptor, xi is the unscaled descriptor, and
xmin and xmax are the respective minimum and maximum values
for this descriptor in the data set.
4.1.1.1. Calculation of E-state indices of scaffold atoms.
The
4.1.1.3. Calculation of the dissimilarity matrix.
The Tanim-
atom-level E-state index of atom i, Si, is a combination of electronic
and topological information defined by45
oto coefficient is frequently used in the determination of intermo-
lecular similarities and database searching applications.63,86,88–90
The pairwise Tanimoto coefficient, T, for the similarity between
two molecules, Mi and Mj, is given by
X
Si ¼ Ii þ
D
Iij
ð1Þ
j
m
where j indicates all atoms other than i. The atom intrinsic state, Ii,
is defined as
X
EikEjk
E2ik þ E2jk ꢀ EikEjk
TðMi; MjÞ ¼
ð5Þ
k¼1
2
v
ð2=NiÞ di þ 1
Ii ¼
ð2Þ
where Mi and Mj are the ith and jth analogs; i, j = 1, 2, . . ., n, with
n = 80 for the MP data set; k = 1, 2, . . ., m, where m is the number
of nonredundant, atom-level scaffold atom E-state indices selected
by the UFS procedure (m = 6 in the present application). The E’s in
Eq. (5) are the atom-level E-state indices for the molecules being
used in the current calculation. Note that for i = j, Tij = 1.
The elements of the dissimilarity matrix, D, are then the com-
plement of the Tanimoto coefficients obtained above in the simi-
larity matrix, T, and are given by
di
where the simple delta value, d, is the number of electrons of atom i
in sigma orbitals minus the number of hydrogen atoms bonded to
v
atom i; the valence delta value, di , is the simple delta value plus
the number of electrons of atom i in
p orbitals plus the number
of electrons of atom i in lone pair electron orbitals; and Ni is the
principal quantum number of atom i. The perturbation due to all
the other atoms in the molecule, DIij, is given by
Iij ¼ ðIi ꢀ IjÞ=r2ij
ð3Þ
DðMi; MjÞ ¼ 1 ꢀ TðMi; MjÞ
ð6Þ
D
Thus, for i = j, dissimilarity D = 0. The complement of the Tanimoto
coefficient is also known as the Soergel distance.
where rij is the graph distance between i and j. The atom-type E-
state index is the sum of the atom-level E-state indices for all the

7230
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
4.1.1.4. Selection of the test set.
A sphere-exclusion53 (SE)
the critical point, evaluated with respect to the correlation of the
dissimilarity-based compound selection (DBCS) algorithm was
used to select a structurally-diverse test set that spans the biolog-
ical activity space.91 Sphere-exclusion algorithms use a predefined
value of a ‘threshold dissimilarity’, t, and in each iteration reject all
compounds in the remaining data set that have a dissimilarity less
than t with respect to a compound in the current test set. Thus, t
defines the radius of a sphere in the descriptor space and at each
stage compounds lying within this sphere are excluded from fur-
ther consideration as candidates for the test set. The size of the
resulting test set is determined by the choice of t: a smaller t would
generate a larger test set, and vice versa. The algorithm executes as
long as there remain compounds in the data set that lie outside all
current spheres. A value of 0.15 has been suggested for t;91,92
t = 0.10 was used in the present work. The SE-MinMax63 selection
criterion for sphere exclusion, in which the compound with the
smallest maximum dissimilarity is selected, was used in the pres-
ent work. An in-house C++ program (D-SIM Version 1.0) was devel-
oped for fast, automated test set identification based on various
combinations of t and sphere exclusion selection criteria. Once
the analogs in the test set were identified, the remaining com-
pounds formed the training set.
original activities vs the scrambled activities). Given a redundant
data set, the value of Q2 is considered conservative such that a
fairly low value of Q2 might indicate a robust original model. A va-
lue greater than 1.2 for the slope is considered to indicate an unsta-
ble model, that is, a model that changes greatly with small changes
in the underlying activity values.97
Progressive scrambling was performed in SYBYL with 50 scram-
blings using a maximum of 10 bins to a minimum of 2 bins. The
critical point was specified at the default value of 0.85 and the ran-
dom seed for every scrambling analysis was set to 123456.
4.1.3.2. Test set validation.
External validation of selected
models was performed using the test set identified above. The
activities of the test set analogs were predicted using these models
and the residuals were noted.
4.1.4. 2D-QSAR study
The six atom-level E-state indices of the scaffold atoms identified
in the test set section above were combined with selected Molconn-
Z descriptors to form an input set of descriptors for PLS analysis. The
Molconn-Z descriptors were selected as follows: 524 descriptors
(other than specific atom-level E-state indices) were calculated for
each of the 80 nMP analogs using the Molconn-Z module of SYBYL.
The same type of correlation analysis using the UFS algorithm was
applied to these descriptors as to the atom-level E-state indices of
the scaffold atoms in the test set selection section above, resulting
in 29 nonredundant Molconn-Z descriptors. Of these, the 12 most
chemically-relevant descriptors were selected by inspection. The
deleted descriptors were related to vertex properties, simple cluster
counts, count of bromine atoms, maximum atom-level E-state and
maximum hydrogen atom-level E-state indices, sums of internal
hydrogen bonding E-state indices, and sums of various carbon atom
E-state indices. The final set of 12 nonredundant Molconn-Z descrip-
tors contained descriptors related to molecular shape and diameter,
and sums of nitrogen, oxygen, fluorine, chlorine, iodine, and sulfur
E-state indices, discussed in detail in Section 2. This set of 12 Mol-
conn-Z descriptors was combined with the six nonredundant
atom-level E-state indices of the 14-atom scaffold to form a set of
18 descriptors for PLS analysis of the binding affinity of the 80 MP
analogs. The PLS analysis and model validation were carried out as
described above.
4.1.2. Partial least squares analysis of training set analogs
Partial least squares (PLS) analysis was carried out using SYBYL to
model the pIC50 values of the 66 training set analogs in terms of
various descriptors. The types of descriptors used in each study
are described in detail in the 2D-QSAR Study and 3D-QSAR Study
sections below. The nMP data set was used in the former study;
CoMFA studies were carried out for both the nMP and pMP data
sets. The PLS method has been shown to work effectively for matri-
ces such as those encountered in QSAR studies.93–96 The leave-one-
out cross-validation method was used to calculate the cross-vali-
dated r2 (or q2), the standard error of prediction (SDEP), and the
optimum number of components. The maximum number of com-
ponents allowed in a model was six. Non-cross-validated PLS mod-
els used the optimum number of components identified in the
cross-validated run.
In the leave-one-out cross-validation procedure, one analog (or
row) in the data set was deleted and a QSAR equation was derived
from the remaining analogs (or rows). The binding affinity of the
deleted analog was predicted using the derived equation and the
deviation from its experimental value (residual) was calculated.
The procedure continued until every analog was deleted once
and its residual was calculated. The sum of all the squared residu-
als was calculated as the PRESS (predictive residual sum of squares)
statistic. The PRESS statistic was used to calculate the cross-vali-
dated correlation coefficient (q2) and the cross-validated standard
error of prediction (SDEP). The q2 is given by
4.1.5. 3D-QSAR study
Separate CoMFA studies were carried out for the nMP and pMP
data sets. The CoMFA steric and electrostatic descriptors were cal-
culated for each data set using cutoffs of 30 kcal/mol each and
standard CoMFA scaling.
ꢀ
ꢁ
PRESS
SS
q2 ¼ 1 ꢀ
ð7Þ
where SS is the sum of squared deviations from the mean.
4.1.3. Model validation
The QSAR models were validated using both activity shuffling
(progressive scrambling) and external test set validation.
4.1.3.1. Progressive scrambling.
Progressive scrambling59 is
recommended for large, redundant data sets that have been used
to generate models employing cross-validation techniques such
as the leave-one-out cross-validation procedure that was used in
this work. Three main statistics are generated: (1) cSDEP (which
is the estimated cross-validated standard error at a user-specified
‘critical point’), (2) Q2 (which is the expected value of q2 at the
specified critical point), and (3) dq2=dry2y0 (which is the slope, at
Figure 2. Torsional angles of the neutral (nMP) and protonated (pMP) methylphe-
nidate global energy minimum (GEM) structures. Angles in degrees.

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7231
4.1.5.1. Data preparation.
The respective nMP and pMP GEM
umn filtering value for cross-validated CoMFA models was set to
2.0 kcal/mol. Three-dimensional steric and electrostatic CoMFA
maps were generated for the CoMFA models and used to interpret
the SAR data in Table 1.
conformations (see Fig. 2) from our previous extensive conforma-
tional analysis study of the potential energy surface of MP26 were
selected as templates for the CoMFA studies. The active, (+)(R,R)-
threo-methylphenidate stereoisomer [(R)-methyl 2-phenyl-2-((R)-
piperidin-2-yl)acetate]12,98 was used to model MP and all the ana-
logs. All 80 analogs were constructed in conformations with tor-
sional angles T1–T4 (see Fig. 2) similar to those of the respective
template GEM conformers to give the three-dimensional nMP
and pMP data sets. As each analog was built, systematic search fol-
lowed by energy minimization was carried out to locate the GEM
conformer for the analog. In a few cases (described below) the ana-
log’s GEM conformer differed significantly from the template’s
GEM due to significant differences in T1. In those cases, a low-en-
ergy structure of the analog with T1 similar to that of the template
was selected instead so that all analogs in the series would have
the same relative alignment for the CoMFA study.
4.1.5.1.1. Neutral MP data set. For each analog, the piperidine
ring was maintained in the chair configuration and its substituents
were constructed in the equatorial positions. The energy of the
analog was minimized using the MMFF9499 force field and charges.
Systematic search was performed in 30° increments on all non-ring
torsional angles in the analog and the resulting conformations
were ranked by energy. As implemented in SYBYL, systematic
search increments each specified torsional angle by the specified
increment, examines the resulting conformations for van der
Waals contacts, and calculates the energy of each conformation
without geometry optimization. A 20 kcal/mol cutoff was used
during the systematic search to eliminate high energy conforma-
tions and to speed up the search. The lowest energy conformation
from the systematic search was selected and its geometry was
optimized by minimization of its energy. If this new, energy-min-
imized conformation had a lower energy than that of the analog
prior to systematic search, then the new conformation was se-
lected for the data set, otherwise the original was kept. The result-
4.2. Pharmacology
The binding protocol has been described in detail else-
where.12,19 Briefly, a P2 fraction was prepared from the striatal tis-
sue of the brains of male Sprague–Dawley rats which had been
anesthetized with CO2 gas. This crude synaptosomal preparation
was homogenized in 25 mM sodium phosphate buffer, pH 7.7, at
0 °C using a Tekmar tissuemizer. Following a 2 h incubation of
the tissue preparation with the analog to be assayed or water (con-
trols) in the presence of 2 nM [3H]WIN 35,428, the samples were
rapidly filtered through glass microfiber filters mounted in Milli-
pore filtration manifolds. The bound radioactivity was extracted
from the filters using liquid scintillation fluid and counted on a li-
quid scintillation counter. Amfonelic acid (10 lM final concentra-
tion) was used to define nonspecific binding. The KD for [3H]WIN
35,428 binding was 18.3 0.9 nM.
4.3. Experimental
4.3.1. Chemistry
The syntheses of 47 new compounds (structures shown in Ta-
ble 1) are presented in this work. All compounds in Table 1 are
racemic and are the threo isomer. Compounds 26, 27, 72, 76, and
78 were made using our modified phenylacetonitrile method.12,19
All of the others were synthesized by straightforward functional
group manipulation of known compounds we had previously
made, some substituted at R1 and some not. Table 7 outlines the
synthetic methods used for each type of compound. Section 4.3.2
gives the details of the synthetic methods.
0
ing analogs were aligned on the five central atoms (C6, C7, C1 , C8
4.3.1.1. Synthetic modifications on nitrogen (R2) and of the
and the hydrogen on C7; see schematic structure of MP in Table 1).
There was one exception to the above procedure. For analog 48
(20-OH nMP), the above procedure led to hydrogen bonding be-
tween the oxygen of the 20 position phenyl substituent and the
hydrogen on the piperidinyl nitrogen. This distorted the analog sig-
nificantly from the template nMP GEM conformation. An alterna-
tive conformation was chosen in which the phenyl ring was
rotated by 180°, causing the 20-OH substituent to move out of
the way of the piperidine ring while still maintaining the basic
nMP GEM conformation.
ester group (R3).
method 1.
Compounds 2, 4, 5, and 25 were made by
4.3.1.2. Synthetic modifications on nitrogen (R2).
Com-
pounds 3 and 13 were made by method 2. Compounds 6,–11 and
21–24 were made by method 3. Compounds 7, 63, and 69 were
made by method 4. Compounds 12, 14, 15, 16, 17, 18, 19, 20, and
28 were made by method 5. Compounds 32 and 33 were made
by method 7. Compound 74 was made by method 13. Compound
1 was made by method 15.
4.1.5.1.2. Protonated MP data set. The procedure outlined above
for the nMP data set (analog construction, minimization, systematic
search, and selection) was also followed for the pMP data set. How-
ever, the presence of the proton on the nitrogen atom in pMP affects
the rotational barrier around T1 compared to nMP.26 This effect was
alsonotedinsomepMPanalogs(2, 4, 5, 25, 28, 66, 70, 71, and75)that
have a somewhat different conformational potential energy surface
than pMP. In these cases, the protonated analog’s GEM conformer
has a very different T1 than that of the pMP GEM conformer. The
hydrogen bonding observed between the ester carbonyl oxygen
and the hydrogen on the nitrogen of the pMP GEM conformer26 does
not occur in these analogs’ GEM conformers. So, for these nine ana-
logs, the systematic search was repeated, but torsional angle T1
was frozen at the pMP GEM value. The GEM conformer from this
search was selected and the side chain was oriented such that it as-
sumed the same hydrogen bonding orientation as that of the pMP
GEM conformation, producing a pMP-like conformer.
4.3.1.3. Synthetic modifications of the ester group (R3).
Com-
pounds 30, 60, 61, and 70 were made by method 6. Compound 62
was made by method 8. Compounds 64 and 65 were made by
method 9. Compound 68 was made by method 10. Compound 71
was made by method 11. Compounds 73 and 75 were made by
methods 12 and 14.
4.3.2. General methods
Starting materials were purchased from Aldrich Chemical Co.
and used without further purification. Tetrahydrofuran (THF) was
dried over sodium benzophenon ketyl prior to distillation under
Argon. Flash chromatography was run using 230–400 mesh silica
gel. Melting points were determined on a Mel-Temp apparatus
and are uncorrected. 1H (300 MHz) spectra were obtained on a Var-
ian Gemini-300 Spectrometer. High-resolution mass spectra [EI
(electron ionization), CI (chemical ionization) or FAB (fast atom
bombardment)] were recorded on a VG Analytical 70-SE mass
spectrometer equipped with a 11-250J data system. Elemental
analyses were obtained from Atlantic Microlabs, Atlanta, GA. Free
4.1.5.2. PLS analysis and model validation.
The PLS analysis
and model validation were carried out as described above. The col-

7232
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
Table 7
Outline of methods.
Method Brief outline of reagents
1
2
3
4
5
6
7
8
Free base in CH2Cl2, add acid chloride, isolate amide, reduce with LiAlH4. See compound 2a
Free base in acetone, add alkyl bromide. See compound 13a
HCl salt in MeOH, add Et3N and aldehyde then NaBH3CN. See compound 21a
Free base in DMF, add K2CO3 and benzyl bromide. See compound 7a
Free base in acetone, add alkyl chloride and KI. See compound 14a
Free base in ether, add LiAlH4. See compound 61a
Free base in HCO2H add formaldehyde. See compound 32a
Free base in chloroform add (BOC)2O, isolate N-BOC alcohol, treat with KOH/DMSO followed by CH3I, isolate ether, treat with
TFA. See compound 62a
9
10
11
12
13
14
15
Ester treated with concentrated HCl, isolate acid, treat with DMF + SOCl2, add appropriate amine. See compound 65a
Free base in MeOH, add Pd/H2. See compound 68a
Free base in DMSO, treat with KOH/DMSO followed by CH3I. See compound 71a
Ester treated with HCl, isolate acid, treat with benzyl alcohol. See compound 73a
Free base in MeOH, add formaldehyde and acetic acid, Pd/H2. See compound 74a
Alcohol treated with acetyl chloride. See compound 75a
Nitro compound in EtOAc, 5% Pt/C, H2, isolate amine, treat with CSCl2. See compound 1a
a
Details of the method used to synthesize each compound can be found in Section 4.3.2.
bases were dissolved in EtOAc or MeOH and converted to HCl salts
by the addition of 1 M HCl (1.5 equiv) in diethylether. The excess
HCl was removed under reduced pressure and the solid was recrys-
tallized from various mixtures of MeOH and EtOAc. Nomenclature
of compounds: methylphenidate is methyl 2-phenyl-2-(piperidin-
2-yl)acetate, ritalinic acid is 2-phenyl-2-(piperidin-2-yl)acetic acid
and ritalinol is 2-phenyl-2-(piperidin-2-yl)ethanol. All structures
are shown in Table 1.
and evaporation of solvents to dryness in vacuo to give a very vis-
cous colorless liquid (75% based on methylphenidate). 1H NMR
(CDCl3): d 1.6–0.99 (m, 6H), 1.98–1.81 (m, 2H), 2.75–2.60 (t, 3H),
2.81–2.75 (t, 2H, N–CH2), 3.2–3.1 (d, 1H), 3.38–3.25 (t, 1H), 3.41–
3.40 (t, 1H), 3.80–3.75 (dd, 1H, from CH2OH), 4.0 (t, 1H, from
CH2OH), 7.38–7.10 (m, 10H, aromatic-H) ppm; CIMS m/z 324.20,
calcd for C22H30NO (MH)+, 324.23. Anal. (C22H29NO) C, H, N.
4.3.2.3. ( )-threo-N-PropargylmethylphenidateꢃHCl (3). Using
the same method as in the synthesis of 13, propargylbromide
(48 mg, 0.40 mmol) and ( )-threo-methylphenidate (89 mg,
0.38 mmol) gave the free amine (100 mg, 81%) was obtained. The
hydrochloride salt showed: mp 129–130 °C; 1H NMR (CDCl3): d
1.60–1.10 (m, 6H), 2.28–2.22 (t, 1H), 2.80–2.70 (m, 1H), 3.10–
3.00 (m, 1H), 3.45–3.44 (m, 1H), 3.50–3.49 (d, 1H, J = 2.5 Hz),
3.61–3.60 (d, 1H, J = 2.5 Hz), 3.70 (s, 3H), 4.00–3.90 (d, 1H,
J = 10.2 Hz), 7.40–7.20 (m, 5H) ppm; CIMS m/z 272.160, calcd for
4.3.2.1. ( )-threo-N-(4-Isothiocyanantobenzyl)methylphenidate
(1).
A mixture of 400 mg of compound 15, 90 mg of 5% Pt/C
and 15 mL EtOAc was exposed to H2 at 48 psi for 2 h. Filtration
and evaporation of solvents gave a solid amine (360 mg) that
was mixed with 25 mL of HCCl3, 15 mL water, 350 mg NaHCO3
and 90 lL of CSCl2. After 15 min of vigorous stirring, the organic
layer was dried and evaporated. The solid was recrystallized from
EtOAc to give 292 mg of the pure final product which showed: mp
112–113 °C; 1H NMR (CDCl3): d 1.60–1.10 (m, 6H), 2.58–2.53 (m,
1H), 3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 3.69 (s, 3H), 3.76 (d,
1H, J = 13.3 Hz), 3.94 (d, 1H, J = 13.3 Hz), 4.16 (d, 1H, J = 10.2 Hz),
7.40–7.20 (m, 9H) ppm; CIMS m/z 381.18, calcd for C22H25N2O2S
(MH)+, 381.16. Anal. (C22H24N2O2S) C, H, N.
C
17H22NO2 (MꢀCl)+, 272.165. Anal. (C17H22ClNO2) C, H, N, Cl.
4.3.2.4. ( )-threo-N-(4-Phenylbutyl)ritalinol (4). Using the same
method as in the synthesis of 2, 4-phenylbutyrylchloride gave a
very viscous colorless liquid (40 mg, 69% based on methylpheni-
date). 1H NMR (CDCl3): d 1.6–0.99 (m, 6H), 1.79–1.60 (m, 4H),
2.75–2.60 (t, 3H), 2.81–2.75 (t, 2H, N-CH2), 3.2–3.1 (d, 1H), 3.38–
3.25 (t, 1H), 3.41–3.40 (t, 1H), 3.80–3.75 (dd, 1H, from CH2OH),
4.0 (t, 1H, from CH2OH), 7.38–7.10 (m, 10H, aromatic-H) ppm;
4.3.2.2. ( )-threo-N-(3-Phenylpropyl)ritalinol (2).
A solution
of ( )-threo-methylphenidate (0.03 g, 0.13 mmol) in dry CH2Cl2
(15 mL) was mixed with anhydrous MgSO4 (2 g) and anhydrous
Na2CO3 (2 g). The suspension was vigorously stirred for 15 min be-
fore a solution of 3-phenylpropionylchloride (0.24 moles) in dry
CH2Cl2 (10 mL) was added dropwise over 5 min at rt. The suspen-
sion was stirred for 30 min and filtered. The organic layer was
washed successively with 1.2 M HCl (100 mL), 25% brine
(100 mL), 5% aqueous NaOH (2 ꢂ 100 mL), and 25% brine
(100 mL) and dried over anhydrous MgSO4. The residue obtained
after removal of solvents was recrystallized in a minimum amount
of hot methanol. In all cases the 1H NMR (CDCl3) was complicated
due to amide rotomers, but was consistent with the expected prod-
uct. The amide (1 equiv) was dissolved in anhydrous Et2O and
cooled in an ice-bath for 15 min. A solution of LiAlH4 in ether
(1 M, 5 equiv) was added and the mixture was stirred in an ice-
bath for an additional 5 min. The resulting mixture was heated un-
der reflux for 2 h, and then cooled in an ice-bath. Unreacted LiAlH4
was quenched by careful addition of excess water (80 mL). The
ether layer was separated and washed with water (2 ꢂ 50 mL),
dried (MgSO4), and evaporated in vacuo to give a residue. The
hygroscopic hydrochloride salt was prepared and converted to
the free base with 10% Na2CO3 solution, extraction with EtOAc
CIMS m/z 338.24, calcd for
C
23H32NO (MH)+, 338.24. Anal.
(C23H31NO) C, H, N.
4.3.2.5. ( )-threo-N-(2-Phenylethyl)ritalinol (5). Using the same
method as in the synthesis of 2, 2-phenylacetylchloride gave a very
viscous colorless liquid (45 mg, 71% based on methylphenidate). 1H
NMR (CDCl3): d 1.6–0.99 (m, 6H), 2.98–2.80 (t, 3H), 3.15–3.09
(t, 2H, N–CH2), 3.2–3.1 (d, 1H), 3.38–3.25 (t, 1H), 3.41–3.40
(t, 1H), 3.80–3.75 (dd, 1H, from CH2OH), 4.0 (t, 1H, from CH2OH),
7.38–7.10 (m, 10H, aromatic-H) ppm; CIMS m/z 310.20, calcd for
C
21H28NO (MH)+, 310.22. Anal. (C21H27NO) C, H, N.
4.3.2.6.
( )-threo-N-(3-Phenylpropyl)methylphenidateꢃHCl
(6). Using the same method as in the synthesis of 21, a mixture
of methylphenidate hydrochloride (0.16 g) and 3-phenylpropion-
aldehyde (0.2 mL) gave the hydrochloride salt (white crystals,
49 mg) which showed: mp 160–161 °C; 1H NMR (CDCl3):
d
1.60–0.90 (m, 6H), 1.80–1.60 (t, 2H), 2.40 (m, 1H), 2.80–2.60
(m, 4H), 3.10–3.20 (m, 1H), 3.4 (m, 1H), 3.61 (s, 3H), 4.20–4.10
(d, 1H, J = 10.4 Hz), 7.40–7.20 (m, 10H) ppm; CIMS m/z 352.22,

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7233
calcd for C23H30NO2 (MꢀCl)+, 352.23. Anal. (C23H30ClNO2) C, H,
3.69 (s, 3H), 3.76 (d, 1H, J = 13.3 Hz), 3.94 (d, 1H, J = 13.3 Hz),
4.16 (d, 1H, J = 10.2 Hz), 7.40–7.20 (m, 9H) ppm; CIMS m/z
N, Cl.
358.158, calcd for
C
21H25ClNO2 (MꢀCl)+, 358.157. Anal.
4.3.2.7.
(7).
( )-threo-N-(3-Chlorobenzyl)methylphenidateꢃHCl
(C21H25Cl2NO2) C, H, N, Cl.
Methylphenidate (267 mg, 1.26 mmol) was dissolved in
DMF (10 mL), 1.0 g K2CO3 added and the mixture stirred for
20 min. Then 3-chlorobenzylbromide (250 L) was added and the
4.3.2.13. ( )-threo-N-AllylmethylphenidateꢃHCl (13).
A mix-
l
ture of ( )-threo-methylphenidate (89 mg, 0.38 mmol) and allyl
bromide (48 mg, 0.40 mmol) in acetone (20 mL) was stirred over-
night at rt. The mixture was filtered and the filtrate and the wash-
ings were evaporated. The residue was chromatographed on 5 g of
silica gel using 5% MeOH in CHCl3 to give 100 mg (80%) pure (by 1H
NMR analysis) titled compound. The hydrochloride salt showed:
mp 159–160 °C; 1H NMR (CDCl3): d 1.60-1.10 (m, 6H), 2.70-2.58
(m, 1H), 3.18–3.00 (m, 1H), 3.48–3.39 (m, 1H), 3.40–3.30 (d, 2H),
3.70 (s, 3H), 4.15–4.10 (d, 1H, J = 10.2 Hz), 5.20–5.00 (m, 1H),
5.90–5.70 (m, 1H), 7.40–7.20 (m, SH) ppm; CIMS m/z 274.180,
calcd for C17H24NO2 (MꢀCl)+, 274.181. Anal. (C17H24ClNO2) C, H,
N, Cl.
resulting mixture was stirred at rt overnight. Ether (50 mL) was
added, washed with water (5 ꢂ 15 mL) and dried (MgSO4). After
evaporation of the solvent, the off-white solid (300 mg) was puri-
fied by silica gel column chromatography (2:98 EtOAc/hexane) to
give pure product as a white solid (217 mg, 63%). The HCl salt
showed: mp 179–180 °C; 1H NMR (CDCl3): d 7.42–7.21 (m, 9H),
4.15 (d, 1H, J = 11.6 Hz), 3.94 (d, 1H, J = 13.7 Hz), 3.79 (d, 1H,
J = 13.2 Hz), 3.67 (s, 3H), 3.50–3.44 (m, 1H), 2.96–2.91 (m, 1H),
2.58–2.52 (m, 1H), 1.58–1.05 (m, 6H) ppm. CIMS m/z 358.17, calcd
for C21H25ClNO2 (MꢀCl)+, 358.16. Anal. (C21H25ClNO2) C, H, N, Cl.
4.3.2.8.
(8).
( )-threo-N-(2-Phenylethyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 21 (with
4.3.2.14.
(14).
( )-threo-N-(4-Chlorobenzyl)methylphenidateꢃHCl
Using a method similar to that in making 13 (except that
the exception that an aqueous workup procedure was employed
in the isolation process), a mixture of methylphenidate hydrochlo-
ride (0.16 g) and 2-phenylacetaldehyde (0.2 mL) gave the hydro-
chloride salt (87 mg, 50%) which showed: mp 154–155 °C; 1H
NMR (CDCl3): d 1.60–0.90 (m, 6H), 2.40 (m, 1H), 2.80–2.60 (t,
2H), 2.95–3.00 (t, 2H), 3.40 (m, 1H), 3.65 (s, 3H), 4.20–4.10 (d,
1H, J = 10.4 Hz), 7.40–7.20 (m, 10H) ppm; CIMS m/z 338.22, calcd
for C22H28NO2 (MꢀCl)+, 338.21. Anal. (C22H28ClNO2) C, H, N, Cl.
0.25 g KI was added to the reaction mixture, the reaction time was
increased to 48 h and the crude product was filtered through a sil-
ica gel column), methylphenidate (89 mg, 0.38 mmol) and 4-chlo-
robenzylchloride (84 mg, 0.40 mmol) gave 70 mg (71%) titled
compound. The hydrochloride salt showed: mp 169–170 °C; 1H
NMR (CDCl3): d 7.42–7.21 (m, 9H), 4.15 (d, 1H, J = 11.6 Hz), 3.94
(d, 1H, J = 13.7 Hz), 3.79 (d, 1H, J = 13.2 Hz), 3.67 (s, 3H), 3.50–
3.44 (m, 1H), 2.96–2.91 (m, 1H), 2.58–2.52 (m, 1H), 1.58–1.05
(m, 6H) ppm. CIMS m/z 357.17, calcd for C21H25ClNO2 (MꢀCl)+,
357.15. Anal. (C21H25Cl2NO2) C, H, N, Cl.
4.3.2.9.
(9).
( )-threo-N-(4-Phenylbutyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 21, a mix-
ture of methylphenidate hydrochloride (0.172 g), 4-phenylbutyral-
dehyde (0.19 g, 1.28 mmol) gave the hydrochloride salt (white
crystals, 91 mg) which showed: mp 140–141 °C; 1H NMR (CDCl3):
d 1.60–0.90 (m, 6H), 1.80–1.60 (t, 4H), 2.40 (m, 1H), 2.60–2.58 (m,
4H), 3.10–3.20 (m, 1H), 3.4 (m, 1H), 3.61 (s, 3H), 4.20–4.10 (d, 1H,
J = 10.4 Hz), 7.40–7.20 (m, 10H) ppm; CIMS m/z 366.27, calcd for
4.3.2.15.
(15).
( )-threo-N-(4-Nitrobenzyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 14, a mix-
ture of methylphenidate (89 mg, 0.38 mmol) and 4-nitrobenzyl-
chloride (86 mg, 0.40 mmol) gave 70 mg (65%) titled compound.
The hydrochloride salt was recrystallized from EtOAc/hexane and
showed: mp 170–172 °C; 1H NMR (CDCl3): d 1.60–1.10 (m, 6H),
2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 3.69 (s,
3H), 3.78–3.74 (d, 1H, J = 13.3 Hz), 3.96–3.91 (d, 1H, J = 13.3 Hz),
4.18–4.14 (d, 1H, J = 10.2 Hz), 7.40–7.20 (m, 5H), 8.15 (d, 2H),
8.75 (d, 2H) ppm; CIMS m/z 369.180, calcd for C21H25N2O4 (MꢀCl)+,
369.189. Anal. (C21H25ClN2O4) C, H, N, Cl.
C
24H32NO2 (MꢀCl)+, 366.24. Anal. (C24H32ClNO2) C, H, N, Cl.
4.3.2.10.
(10).
( )-threo-N-(5-Phenylpentyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 21, a mix-
ture of methylphenidate (0.16 g) and phenylpentyraldehyde (0.2 g,
1.3 mmol) gave the hydrochloride salt (white crystals, 81 mg): mp
136–137 °C; 1H NMR (CDCl3): d 1.60–0.90 (m, 6H), 1.80–1.70 (t,
4H), 2.40 (m, 1H), 2.50–2.20 (m, 4H), 2.6 (t, 2H), 3.10–3.20 (m,
1H), 3.40 (m, 1H), 3.61 (s, 3H), 4.20–4.10 (d, 1H, J = 10.4 Hz),
7.40–7.20 (m, 10H) ppm; CIMS m/z 380.25, calcd for C25H34NO2
(MꢀCl)+, 380.26. Anal. (C25H34ClNO2) C, H, N, Cl.
4.3.2.16.
(16).
( )-threo-N-(4-Methoxybenzyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 14, a mix-
ture of methylphenidate (89 mg, 0.38 mmol) and 4-methoxyben-
zylchloride (63 mg, 0.40 mmol) gave 80 mg (61%) titled
compound. The hydrochloride salt was recrystallized from EtOAc/
hexane and showed: mp 164–165 °C; 1H NMR (CDCl3): d 1.60–
1.10 (m, 6H), 2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H), 3.66–3.49
(m, 1H), 3.70 (s, 3H), 3.78–3.74 (d, 1H, J = 13.3 Hz), 3.85 (s, 3H),
3.96–3.91 (d, 1H, J = 13.3 Hz), 4.18–4.14 (d, 1H, J = 10.2 Hz), 6.88
(d, 2H, J = 8.8 Hz), 7.40–7.20 (m, 7H) ppm; CIMS m/z 354.20, calcd
for C22H28NO3 (MꢀCl)+, 354.21. Anal. (C22H28ClNO3) C, H, N, Cl.
4.3.2.11.
(11).
( )-threo-N-(6-Phenylhexyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 21, a mix-
ture of methylphenidate (0.17 g) and phenylhexyraldehyde (0.23 g,
1.3 mmol) gave the hydrochloride salt (white crystals, 81 mg)
which showed: mp 136–137 °C; 1H NMR (CDCl3): d 1.60–0.90 (m,
6H), 1.80–1.70 (t, 4H), 2.40 (m, 1H), 2.50–2.20 (m, 4H), 2.6 (t,
2H), 3.10–3.20 (m, 1H), 3.40 (m, 1H), 3.61 (s, 3H), 4.20–4.10 (d,
1H, J = 10.4 Hz), 7.40–7.20 (m, 10H) ppm; CIMS m/z 394.25, calcd
for C26H36NO2 (MꢀCl)+, 394.27. Anal. (C26H36ClNO2) C, H, N, Cl.
4.3.2.17. ( )-threo-N-((5-Chlorothiophene-2-yl)methyl)methyl-
phenidateꢃHCl (17).
Using the same method as in the synthe-
sis of 14, a mixture of ( )-threo-methylphenidate hydrochloride
(0.040 g, 0.038 mmol) and 2-chloro-5-(chloromethyl)thiophene
(0.05 mL, excess) gave the titled compound. The hydrochloride salt
was recrystallized from EtOAc/hexane and showed: mp 174–
176 °C; 1H NMR (CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m, 1H),
3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 3.71 (s, 3H, OCH3), 3.97–
3.93 (d, 1H, J = 13.4 Hz), 3.84–3.97 (d, 1H, J = 13.2 Hz), 4.2–4.1 (d,
1H, J = 10.4 Hz), 7.3 (dd, 2H, thiopene-H), 7.4–7.2 (m, 5H, aro-
4.3.2.12.
(12).
( )-threo-N-(2-Chlorobenzyl)methylphenidateꢃHCl
Using the same method as in the synthesis of 14, a mix-
ture of methylphenidate (89 mg, 0.38 mmol) and 2-chlorobenzyl-
chloride (103 mg, 0.50 mmol) gave 100 mg (78%) titled
compound. After chromatography on silica gel, the hydrochloride
salt showed: mp 151–152 °C; 1H NMR (CDCl3): d 1.60–1.10 (m,
6H), 2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H),

7234
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
matic-H) ppm; FABMS m/z 364.10, calcd for C19H23ClNO2S (MꢀCl)+,
(m, 1H), 3.66–3.49 (m, 1H), 3.71 (s, 3H, OCH3), 3.97–3.93 (d, 1H,
J = 13.4 Hz), 3.84–3.97 (d, 1H, J = 13.2 Hz), 4.2–4.1 (d, 1H,
J = 10.4 Hz), 7.1 (s, 1H, thiopene-H), 7.4–7.2 (m, 5H, aromatic-H),
7.50 (d, 1H, thiopene-H), 7.60 (d, 1H, thiopene-H) ppm; FABMS
364.11. Anal. (C19H23Cl2NO2S) C, H, N, Cl.
4.3.2.18. ( )-threo-N-(2-Methylpyridine)methylphenidateꢃ2HCl
(18).
Using the same method as in the synthesis of 14, a mix-
m/z 330.17, calcd for
C
19H24NO2S (MꢀCl)+, 330.15. Anal.
ture of ( )-threo-methylphenidate (89 mg, 0.38 mmol) and 2-chlo-
romethylpyridine (66 mg, 0.4 mmol) gave the titled compound.
The dihydrochloride salt showed: mp 153–154 °C; 1H NMR
(CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H),
3.66–3.49 (m, 1H), 3.69 (s, 3H, OCH3), 3.96–3.91 (d, 1H,
J = 13.4 Hz), 3.79–3.75 (d, 1H, J = 13.2 Hz), 4.18–4.14 (d, 1H,
J = 10.4 Hz), 7.1 (t, 1H, pyridyl-H), 7.3–7.2 (m, 9H, aromatic-H),
7.4 (d, 1H, pyridyl-H), 7.8 (t, 1H, pyridyl-H), 8.45 (d, 1H, pyridyl-
H) ppm; FABMS m/z 325.20, calcd for C20H25N2O2 (MꢀHClꢀCl)+,
325.19. Anal. (C20H26Cl2N2O2) C, H, N, Cl.
(C19H24ClNO2S) C, H, N, Cl.
4.3.2.23. ( )-threo-N-(2-Methylthiopene)methylphenidateꢃHCl
(23). Using the same method as in the synthesis of 21, a mix-
ture of ( )-threo-methylphenidate hydrochloride (0.260 g,
0.96 mmol) and thiopene-2-aldehyde (0.146 g, 0.11 mL, 1.3 mmol)
gave the hydrochloride salt which showed: mp 180–181 °C; 1H
NMR (CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m, 1H), 3.03–2.95
(m, 1H), 3.66–3.49 (m, 1H), 3.71 (s, 3H, OCH3), 3.97–3.93 (d, 1H,
J = 13.4 Hz), 3.84–3.97 (d, 1H, J = 13.2 Hz), 4.2–4.1 (d, 1H,
J = 10.4 Hz), 6.9 (d, 1H, thiopene-H), 7.3 (dd, 2H, thiopene-H),
7.4–7.2 (m, 5H, aromatic-H), ppm; FABMS m/z 330.16, calcd for
4.3.2.19. ( )-threo-N-(3-Methylpyridine)methylphenidateꢃ2HCl
(19).
Using the same method as in the synthesis of 14, a mix-
C
19H24NO2S (MꢀCl)+, 330.15. Anal. (C19H24ClNO2S) C, H, N, Cl.
ture of ( )-threo-methylphenidate (89 mg, 0.38 mmol) and 3-chlo-
romethylpyridine (66 mg, 0.4 mmol) gave the titled compound.
The dihydrochloride salt showed: mp 159–160 °C; 1H NMR
(CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H),
3.66–3.49 (m, 1H), 3.69 (s, 3H, OCH3), 3.96–3.91 (d, 1H,
J = 13.4 Hz), 3.79–3.75 (d, 1H, J = 13.2 Hz), 4.18–4.14 (d, 1H,
J = 10.4 Hz), 7.3–7.2 (m, 9H, aromatic-H), 7.4 (s, 1H, pyridyl-H),
7.6 (d, 1H, pyridyl-H), 8.5 (d, 2H, pyridyl-H) ppm; FABMS m/z
4.3.2.24.
(24).
( )-threo-N-(3-Methylfuran)methylphenidateꢃHCl
Using the same method as in the synthesis of 21, a mix-
ture of methylphenidate (260 mg, 0.96 mmol) and 3-furaldehyde
(125 mg, 1.3 mmol) gave the hydrochloride salt which showed:
mp 181–182 °C; 1H NMR (CDCl3): d 1.60–1.10 (m, 6H), 2.58–2.53
(m, 1H), 3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 3.71 (s, 3H), 3.84–
3.97 (d, 1H, J = 13.4 Hz), 3.97–3.93 (d, 1H, J = 13.4 Hz), 4.20–4.10
(d, 1H, J = 10.4 Hz), 6.40 (s, 1H), 6.60 (d, 1H), 7.40–7.20 (m, 5H),
7.50 (dd, 2H) ppm; FABMS m/z 314.20, calcd for C19H24NO3
(MꢀCl)+, 314.18. Anal. (C19H24ClNO3) C, H, N, Cl.
325.20, calcd for
C
20H25N2O2 (MꢀHClꢀCl)+, 325.19. Anal.
(C20H26Cl2NO2) C, H, N, Cl.
4.3.2.20. ( )-threo-N-(4-Methylpyridine)methylphenidateꢃ2HCl
(20).
Using the same method as in the synthesis of 14, a mix-
4.3.2.25. ( )-threo-N-Ethylritalinol (25).
Using the same
ture of methylphenidate (89 mg, 0.38 mmol) and 4-pyridylmethyl-
chloride (66 mg, 0.4 mmol) gave the titled compound. The
dihydrochloride salt showed: mp 153–154 °C; 1H NMR (CDCl3): d
1.60–1.10 (m, 6H), 2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H), 3.66–
3.49 (m, 1H), 3.69 (s, 3H), 3.79–3.75 (d, 1H, J = 13.4 Hz), 3.96–
3.91 (d, 1H, J = 13.4 Hz), 4.18–4.14 (d, 1H, J = 10.4 Hz), 7.10 (t,
1H), 7.30–7.20 (m, 9H) ppm; FABMS m/z 325.20, calcd for
method as in the synthesis of 2, acetyl chloride gave a very viscous
colorless liquid (140 mg, 69% based on methylphenidate). 1H NMR
(CDCl3): d 0.99 (m, 1H), 1.20 (t, 2H), 1.30–1.60 (m, 5H), 2.75–2.60
(t, 3H), 2.81–2.75 (t, 2H, N–CH2), 3.2–3.1 (d, 1H), 3.35–3.20 (t,
1H), 3.40 (t, 1H), 3.79 (dd, 1H, from CH2OH), 4.0 (t, 1H, from
CH2OH), 7.38–7.10 (m, 5H, aromatic-H) ppm; CIMS m/z 234.200,
calcd for C15H24NO (MH)+, 234.186. Anal. (C15H23NO) C, H, N.
C
20H25N2O2 (MꢀHClꢀCl)+, 325.19. Anal. (C20H26Cl2NO2) C, H, N, Cl.
4.3.2.26.
(26).
( )-threo-30,50-DimethylmethylphenidateꢃHCl
4.3.2.21.
(21).
( )-threo-N-(2-Methylfuran)methylphenidateꢃHCl
A mixture of ( )-threo-methylphenidate hydrochloride
Using the same method as previously described,12
10.0 g of 3,5-dimethylphenylacetonitrile and 10.9 g of 2-bromo-
pyridine gave 0.425 g of the HCl salt which showed: mp 229–
230 °C. 1H NMR (D2O): d 7.1 (s, 1H), 6.8 (s, 2H), 3.84 (d, 1H),
3.75–3.72 (m, 1H), 3.59 (s, 3H), 3.31–3.13 (m, 1H), 2.83–2.79 (m,
1H), 2.2 (s, 6H), 1.94–1.40 (m, 6H) ppm; CIMS m/z 262.18, calcd
for C16H24NO2 (MꢀCl)+, 262.18. Anal. (C16H24ClNO2ꢃ0.50H2O): C,
H, N, Cl.
(0.260 g, 0.96 mmol) and 2-furaldehyde (0.125 g, 0.11 mL,
1.3 mmol) was dissolved in methanol (25 mL) followed by the
addition of Et3N (0.4 g). Into the solution was added 4 Å molecular
sieves (4 g), and the reaction mixture stirred at room temperature
for 1.5 h. NaBH3CN (0.15 g, 2.38 mmol) was added into the solu-
tion, and the reaction was continued for an additional 18 h. The
mixture was filtered and the filtrate and the washings were evap-
orated in vacuo. The residue was chromatographed on 10 g of silica
gel using CHCl3/MeOH (25:1) to give 110 mg (88%, pure by TLC
analysis) titled compound. The hydrochloride salt showed: mp
170–171 °C; 1H NMR (CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m,
1H), 3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 3.71 (s, 3H, OCH3),
3.97–3.93 (d, 1H, J = 13.4 Hz), 3.84–3.97 (d, 1H, J = 13.2 Hz), 4.2–
4.1 (d, 1H, J = 10.4 Hz), 6.4 (d, 1H, furan-H), 6.6 (d, 1H, furan-H),
7.4–7.2 (m, 5H, aromatic-H), 7.58 (d, 1H, furan-H) ppm; FABMS
4.3.2.27.
(27).
( )-threo-30,50-DichloromethylphenidateꢃHCl
Using the same method as previously described,12 4.0 g
of 3,5-dichlorophenylacetonitrile and 3.4 g of 2-bromopyridine
gave 0.105 g of the HCl salt which showed: mp 198–200 °C. 1H
NMR (D2O): d 7.4 (s, 1H), 7.2 (s, 2H), 3.84 (d, 1H), 3.75–3.72 (m,
1H), 3.59 (s, 3H), 3.31–3.13 (m, 1H), 2.83–2.79 (m, 1H), 1.94–
1.40 (m, 6H) ppm; CIMS m/z 302.06, calcd for C14H18Cl2NO2
(MꢀCl)+, 302.07. Anal. (C14H18Cl3NO2ꢃ0.75H2O): C, H, N, Cl.
m/z 314.20, calcd for
(C19H24ClNO3) C, H, N, Cl.
C
19H24NO3 (MꢀCl)+, 314.18. Anal.
4.3.2.28.
(28).
( )-threo-N-3-ChlorobenzylmethylphenidateꢃHCl
Using the same method as in the synthesis of 14, a mix-
4.3.2.22. ( )-threo-N-(3-Methylthiopene)methylphenidateꢃHCl
(22). Using the same method as in the synthesis of 21, a mix-
ture of ( )-threo-methylphenidate hydrochloride (0.260 g,
0.96 mmol) and thiopene-3-aldehyde (0.146 g, 0.11 mL, 1.3 mmol)
gave the hydrochloride salt which showed: mp 175–176 °C; 1H
NMR (CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m, 1H), 3.03–2.95
ture of methylphenidate (89 mg, 0.38 mmol) and 3-chlorobenzyl-
chloride (82 mg, 0.40 mmol) gave the titled compound. The
hydrochloride salt was recrystallized from EtOAc/hexane and
showed: mp 179–180 °C; 1H NMR (CDCl3): d 1.60–1.10 (m, 6H),
2.58–2.53 (m, 1H), 3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 3.69 (s,
3H), 3.78–3.74 (d, 1H, J = 13.3 Hz), 3.96–3.91 (d, 1H, J = 13.3 Hz),

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7235
4.18–4.14 (d, 1H, J = 10.2 Hz), 7.40–7.20 (m, 9H) ppm; CIMS m/z
very slowly and the ether was washed with water and dried
(MgSO4). The solvent was removed to give the N-BOC alcohol as
a colorless oil (233 mg, 96% in two steps). Dry DMSO (23 mL)
was added to ground KOH (490 mg) and the suspension was stirred
at rt under a nitrogen atmosphere for 20 min. The N-BOC alcohol
(610 mg) was dissolved in DMSO (15 mL) and added into the
KOH/DMSO mixture in one portion, followed by immediate addi-
tion of CH3I (0.57 mL). The mixture was stirred at rt overnight
and then diluted by methylene chloride (150 mL), washed with
water (5 ꢂ 30 mL), and dried (MgSO4) to give a pale yellow oil
(565 mg) after evaporation of the solvent. The crude product was
then purified using silica gel column chromatography (5% EtOAc/
hexane followed by 10% EtOAc/hexane) to give the pure BOC-pro-
tected methyl ether product as a colorless oil (251 mg, 46% in two
steps). 1H NMR (CDCl3): d 7.41–7.11 (m, 3H), 4.47–4.31 (m, 1H),
4.19–3.98 (m, 1H), 3.59–3.45 (m, 2H), 3.38–3.31 (m, 1H), 3.22 (s,
3H), 2.89–2.81 (m, 1H), 1.66–1.25 (m, 15H) ppm. The BOC-pro-
tected methyl ether (250 mg, 0.68 mmol) was mixed with trifluo-
roacetic acid (3.5 mL) and stirred at rt for 30 min. The TFA was
then evaporated to give a pale yellow oil (360 mg). The oil was
then dissolved in EtOAc and washed with 5% sodium carbonate
and dried (MgSO4). A colorless oil (153 mg) was obtained as the
pure product (4d) (83%) after evaporation of the solvent. 1H NMR
(CDCl3): d 7.37 (d, 1H, J = 8.2 Hz), 7.31 (s, 1H), 7.07–7.04 (dd, 1H,
J = 1.7, 8.2 Hz), 3.67 (d, 2H, J = 8.2 Hz), 3.32 (s, 3H), 3.09 (d, 1H,
J = 12.1 Hz), 2.79–2.75 (m, 2H), 2.63 (t, 1H, J = 9.3 Hz), 2.07 (br s,
1H), 1.73–0.99 (m, 6H) ppm. The free base was converted to its
HCl salt which showed: mp 202–205 °C, 1H NMR (D2O): d 7.41
(d, 1H, J = 8.2 Hz), 7.35 (s, 1H), 7.06 (d, 1H, J = 8.2 Hz), 3.73 (t,
1H), 3.60–3.5 (m, 1H), 3.38 (t, 1H), 3.29–3.15 (m, 4H), 3.07–3.04
(q, 1H), 2.86 (t, 1H), 1.67–1.18 (m, 6H) ppm. Anal.
(C14H20Cl3NOꢃ0.50 H2O) C, H, N, Cl.
358.160, calcd for
C
21H25ClNO2 (MꢀCl)+, 358.157. Anal.
(C21H25Cl2NO2) C, H, N, Cl.
4.3.2.29. ( )-threo-Ritalinol (30).
Using the same procedure
as in the synthesis of 61, methylphenidate (1.61 g, 6.91 mmol)
gave the pure product as a white solid (mp 95.4–97 °C, 0.95 g,
99%). 1H NMR (CDCl3): d 7.35–7.20 (m, 5H), 3.99–3.93 (dd, 1H),
3.88–3.83 (dd, 1H), 3.11–3.06 (m, 1H), 2.97–2.90 (q, 1H), 2.67–
2.55 (m, 2H), 1.80–1.76 (m, 1H), 1.63–1.59 (m, 1H), 1.45–1.21
(m, 3H) ppm. CIMS m/z 206.159, calcd for C13H20NO (MH)+,
206.155. Anal. (C13H19NO) C, H, N.
4.3.2.30. ( )-threo-4-Hydroxy-N-methylmethylphenidate (33)
and by-product (32).
A 200 mg sample of 39 was converted
to the free base, and mixed with 1 mL of 37% formaldehyde and
1 mL of HCO2H. After heating under reflux for 2 h, the mixture
was evaporated, taken up in EtOAc, washed with NaHCO3 solution
and the organic layer was dried and evaporated. This gave 225 mg
of a clear oil which was carefully chromatographed on silica gel
using EtOAc/MeOH (90:10). This gave 57 mg of 33 and 28 mg of
32. Compound 33; 1H NMR (CDCl3): d 7.3 (s, 1H), 7.2 (d, 2H), 6.7
(d, 2H), 3.83 (d, J = 9.9 Hz, 1H), 3.68 (s, 3H), 3.22–3.1 (m, 1H),
3.05–2.9 (m,1H), 2.6–2.5 (m, 1H), 2.4 (s, 3H), 1.8–1.0 (m, 6H)
ppm. Anal. (C15H21NO3): C, H, N. Compound 32; 1H NMR (CDCl3):
d 7.1 (d, 1H), 6.95 (s, 1H), 6.8 (s, 1H), 5.25 (s, 2H), 4.9 (s, 1H),
3.77 (d, J = 9.9 Hz, 1H), 3.68 (s, 3H), 3.1 (m, 1H), 2.95 (m, 1H),
2.55 (m, 1H), 2.4 (s, 3H), 1.6–1.0 (m, 6H) ppm. Anal.
(C17H25NO5): C, H, N.
4.3.2.31. ( )-threo-3-Fluororitalinol (60).
Using the same
procedure as in the synthesis of 61, 57 (103 mg, 0.36 mmol) affor-
ded 60 as a white solid after recrystallization from ether (48 mg,
60%) which showed: mp 98–99 °C. 1H NMR (CDCl3): d 7.28 (dd,
1H, J = 14.0, 8.2 Hz), 7.1–6.9 (m, 3H), 3.95 (dd, 1H), 3.85 (dd, 1H),
3.09–3.00 (m, 1H), 2.93 (q, 1H), 2.65–2.52 (m, 2H), 1.76 (m, 1H),
1.60–1.57 (m, 1H), 1.40–1.20 (m, 3H) ppm. Anal. (C13H18FNO): C,
H, N.
4.3.2.34.
(63).
( )-threo-N-Benzyl-30-chloromethylphenidateꢃHCl
Using the same procedure as in the synthesis of 7, 30-
chloromethylphenidate (49 mg, 0.16 mmol) and bnezylbromide
afforded 63 as a off-white solid (mp 86.4–87.9 °C, 39.7 mg, 69%);
1H NMR (CDCl3): d 7.42–7.21 (m 9H), 4.15 (d, 1H, J = 11.6 Hz),
3.94 (d, 1H, J = 13.7 Hz), 3.79 (d, 1H, J = 13.2 Hz), 3.67 (s, 3H),
3.50–3.44 (m, 1H), 2.96–2.91 (m, 1H), 2.58–2.52 (m, 1H), 1.58–
1.05 (m, 6H) ppm. The HCl salt showed mp 116–162 °C (dec). Anal.
(C21H24ClNO2) C, H, N, Cl.
4.3.2.32. ( )-threo-30,40-Dichlororitalinol (61).
To the solu-
tion of 30,40-dichloromethylphenidate (162 mg, 0.54 mmol) in
ether (15 mL) was added 1 M lithium aluminum hydride (4.5 mL,
4.5 mmol) via syringe at 0 °C under a nitrogen atmosphere. The
solution was stirred at 0 °C for 2 h and water carefully added drop
by drop until no gas was evolved. The ether was washed with
water three times and then dried (MgSO4). Evaporation of the sol-
vent gave a white solid that proved to be the pure product 61 (mp
4.3.2.35. ( )-threo-N-Benzylmethylphenidate dimethyl ami-
deꢃHCl (64).
Using the same procedures as in the synthesis
of 65 (using dimethyl amine and 60 mg of N-benzylritalinic acid),
the pure amide was obtained as a pale yellow solid (mp 110–
111 °C, 45 mg, 74%). 1H NMR (CDCl3): d 7.41–7.20 (m, 10H), 4.18
(d, 1H, J = 9.9 Hz), 3.75–3.68 (m, 3H), 2.95 (d, 6H, J = 15 Hz),
2.75–2.73 (m, 1H), 2.48–2.43 (m, 1H), 1.69–1.07 (m, 6H) ppm;
EIMS m/z 337.221, calcd for C22H29N2O (MH)+, 337.228. The HCl
salt showed mp 201 °C (dec). Anal. (C22H29ClN2Oꢃ0.25H2O): C, H,
N, Cl.
134–136 °C, 146 mg, 99%). 1H NMR (CDCl3):
d 7.38 (d, 1H,
J = 8.2 Hz), 7.33 (d, 1H, J = 2.2 Hz), 7.08 (dd, 1H, J = 2.2, 8.2 Hz),
3.97 (dd, 1H, J = 7.7 Hz, 11.0 Hz), 3.86 (dd, 1H, J = 3.7, 11.0 Hz),
3.11–3.06 (m, 1H), 2.97–2.90 (m, 1H), 2.67–2.55 (m, 2H), 1.80–
1.76 (m, 1H), 1.63–1.59 (m, 1H), 1.45–1.21 (m, 3H) ppm; CIMS
m/z 274.07, calcd for C13H18Cl2NO (MH)+, 274.08; HCl salt, mp
164–167 °C (dec). Anal. (C13H18Cl3NOꢃ0.40 H2O) C, H, N, Cl.
4.3.2.36. ( )-threo-N-Benzylmethylphenidate amideꢃHCl (65).
N-Benzylmethylphenidate (29) (400 mg, 1.24 mmol) was mixed
with 6 N HCl (2.5 mL) and the solution was heated under reflux
for 1 h. The solvent was removed to afford N-benzylritalinic acid
as a white solid (426 mg, 99.5%) which showed mp 121–160 °C.
1H NMR d 7.40–7.09 (m, 10H), 4.61 (d, 1H, J = 13.2 Hz), 4.39–4.30
(m, 2H), 4.03–3.80 (m, 1H), 3.50–3.33 (m, 1H), 3.21–3.06 (m, 1H),
1.92–1.01 (m, 6H) ppm. The acid (84 mg, 0.24 mmol) was sus-
pended in chloroform (3 mL) at room temperature. A few drops
of DMF were added and followed with the addition of thionyl chlo-
ride (0.1 mL). The suspension changed to a clear solution after a
few minutes and then was cooled via ice-bath. This cooled solution
4.3.2.33. ( )-threo-30,40-Dichlororitalinol methyl etherꢃHCl
(62).
A mixture of 52 (185 mg, 0.55 mmol), chloroform
(10 mL), water (5 mL), NaHCO3 (105 mg), NaCl (120 mg) and
(BOC)2O (128 mg, 0.59 mmol) was heated under a gentle reflux
for 90 min. The chloroform layer was washed with ice-cold HCl
(1 N, 10 mL) and then water (3 ꢂ 10 mL), and dried (MgSO4) to give
a colorless oil (270 mg, 111%). To the solution of the N-BOC ester in
anhydrous diethyl ether (8 mL) was added lithium aluminum hy-
dride (0.6 mL, 0.6 mmol, 1.0 M in ether solution) via syringe at
0 °C under a nitrogen atmosphere. The solution was then stirred
at 0 °C for 1 h and then at rt for 25 min. Water was added dropwise

7236
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
was then added to a concentrated ammonium hydroxide solution
(4 mL) dropwise at 0 °C and the resulting mixture was stirred at
0 °C for 10 min and then rt for 2 h. The organic layer was separated
and the aqueous layer was extracted with chloroform (2 ꢂ 10 mL).
The combined organic layers were washed with water (2 ꢂ 10 mL)
and dried (MgSO4). After removal of the solvent, a yellow solid was
obtained (79.3 mg). The crude product was purified by silica gel
column chromatography (1:1 EtOAc/hexane) to give the pure prod-
uct as a yellow solid (mp 134–135 °C, 24 mg, 32%). 1H NMR
(CDCl3): d 7.88 (br s, 1H), 7.36–7.22 (m, 10H), 5.52 (br s, 1H),
3.92–3.79 (m, 3H), 3.44–3.37 (m, 1H), 2.98–2.90 (m, 1H), 2.60–
2.54 (m, 1H), 1.75–1.41 (m, 5H), 1.26–1.16 (m, 1H) ppm. EIMS
m/z 309.198, calcd for C20H25N2O (MH)+, 309.197. HCl salt mp
157–180 °C (dec). Anal. (C20H25ClN2Oꢃ0.60 H2O), C, H, N, Cl.
4.3.2.41. ( )-threo-30,40-Dichloromethylphenidate amideꢃHCl
(72).
Using the same method as previously described,12
30.4 g of 3,4-dichlorophenylacetonitrile and 21.5 g of 2-bromopyr-
idine gave 15.6 g of the crude amide (third intermediate) as an er-
ythro/threo mixture (2:1).
A 200 mg portion was carefully
chromatographed on silica gel using EtOAc/MeOH/Et2NH
(10:90:0.1). This gave 54 mg of the threo isomer which was con-
verted to the HCl salt and crystallized from MeOH/Ether (39 mg).
The HCl salt showed: mp 251–253 °C; 1H NMR (CDCl3): d 8.1 (br
s, 1H), 7.40 (d, 1H, J = 8.2 Hz), 7.45 (s, 1H), 7.04 (d, 1H, J = 8.2 Hz)
, 5.62 (br s, 1H), 3.92–3.79 (m, 3H), 3.44–3.37 (m, 1H), 2.98–2.90
(m, 1H), 2.60–2.54 (m, 1H), 1.75–1.41 (m, 5H), 1.26–1.16 (m, 1H)
ppm. CIMS m/z 287.074, calcd for
287.072. Anal. (C13H17Cl3N2Oꢃ0.75 H2O): C, H, N, Cl.
C
13H17Cl2N2O (MꢀCl)+,
4.3.2.37. ( )-threo-Ritalinol methyl etherꢃHCl (68).
The HCl
4.3.2.42. ( )-threo-BenzylphenidateꢃHCl (73).
Methylpheni-
salt of 67, (68 mg, 0.20 mmol) was dissolved in methanol (5 mL)
and palladium hydroxide (20 mg, 20% Pd, wet) was then added.
Hydrogenation was carried out at 45 psi for 2 h. The catalyst was
filtered off and the solvent was removed to afford pure 68 as a
white solid (mp 154–155 °C, 50 mg, 99%). 1H NMR (D2O): d 7.32–
7.14 (m, 5H), 3.76 (dd, 1H, J = 8.2, 9.9 Hz), 3.73 (dd, 1H, J = 6.0,
9.9 Hz), 3.43–3.37 (m, 1H), 3.23–3.21 (m, 4H), 3.10–3.03 (m, 1H),
2.91–2.82 (m, 1H), 1.65–1.17 (m, 6H) ppm. Anal. (C14H22ClNOꢃ0.50
H2O): C, H, N, Cl.
date (39) (190 mg, 0.70 mmol) was mixed with 6 N HCl (2.5 mL)
and the solution was heated under reflux for 1 h. The solvent
was removed to afford a white solid (ritalinic acid) which was
mixed with 5 mL of benzyl alcohol at 90 °C for 3 h. After cooling
to rt, 40 mL of ether was added, the solid material removed and
the filtrate concentrated to give white crystals (65 mg), mp 194–
196 °C. 1H NMR (CDCl3): d 1.6–1.1 (m, 6H), 2.58–2.53 (m, 1H),
3.03–2.95 (m, 1H), 3.66–3.49 (m, 1H), 4.2–4.1 (d, 1H, J = 10.4 Hz),
4.95 (d, 1H, J = 13.4 Hz), 4.90 (d, 1H, J = 13.4 Hz), 7.4–7.2 (m,
10H) ppm. Anal. (C20H24ClNO2): C, H, N, Cl.
4.3.2.38. ( )-threo-N-Benzyl-30,40-dichloromethylphenidateꢃHCl
(69).
Using the same procedure as in the synthesis of 7,
4.3.2.43. ( )-threo-N-Methyl-30-methylmethylphenidate (74).
A 100 mg sample of 59 was converted to the free base, dissolved in
2 mL of MeOH and 40 lL of 37% formaldehyde, 10 lL of HOAc and
20 mg of 5% Pd/C added. After 2 h of H2 treatment at 40 psi, the
mixture was filtered and evaporated, taken up in EtOAc, dried
and evaporated again. This gave 74 mg of a clear oil. 1H NMR
(CDCl3): d 7.2–7.1 (m, 4H), 3.83 (d, J = 9.9 Hz, 1H), 3.68 (s, 3H),
3.22–3.1 (m, 1H), 3.05–2.9 (m, 1H), 2.6–2.5 (m, 1H), 2.4 (s, 3H),
1.8–1.0 (m, 6H) ppm. Anal. (C16H23NO2): C, H, N.
30,40-dichloromethylphenidate (360 mg, 1.1 mmol) and excess ben-
zyl bromide afforded 69 as a pale yellow solid (mp 94.7–95.8 °C,
348 mg, 84%). 1H NMR (CDCl3): d 7.52 (d, 1H, J = 2.2 Hz), 7.39 (d,
1H, J = 8.3 Hz), 7.34–7.23 (m, 6H), 4.13 (d, 1H, J = 11.5 Hz), 3.93
(d, 1H, J = 13.7 Hz), 3.78 (d, 1H, J = 13.7 Hz), 3.67 (s, 3H), 3.44 (d,
1H, J = 11.0 Hz), 2.98–2.90 (dt, 1H), 2.57–2.51 (m, 1H), 1.57–1.52
(m, 4H), 1.49–1.32 (m, 1H), 1.06–1.04 (m, 1H) ppm; EIMS m/z
392.114, calcd for C21H24Cl2NO2 (MH)+, 392.118. HCl salt, mp
116–162 °C (dec). Anal. (C21H24Cl3NO2) C, H, N, Cl.
4.3.2.44. ( )-threo-Ritalinol acetate (75).
A 355 mg portion
4.3.2.39. ( )-threo-N-Benzyl-30,40-dichlororitalinolꢃHCl (70).
Using the same procedure as in the synthesis of 61, 69 (618 mg,
1.58 mmol) afforded a colorless oil as pure 70 (601 mg, 95%). 1H
NMR (CDCl3): d 7.34–7.23 (m, 6H), 7.07 (d, 1H, J = 2.2 Hz), 6.81
(dd, 1H, J = 2.2, 8.3 Hz), 3.90 (s, 3H), 3.68 (t, 1H, J = 11.0 Hz), 3.58
(dd, 1H, J = 3.9, 11.0 Hz), 3.35 (dt, 1H, J = 3.3, 11.5 Hz), 3.19–3.02
(m, 2H), 2.69–2.64 (m, 1H), 1.75–1.44 (m, 4H), 1.30–1.27 (m,
1H), 0.88–0.83 (m, 1H) ppm. The HCl salt showed: mp 215 °C
(dec); 1H NMR (D2O): d 7.47–7.36 (m, 7H), 7.14–7.11 (d, 1H,
J = 8.2 Hz), 4.77–4.62 (m, 1H), 4.41–4.26 (m, 1H), 4.03–3.62 (m,
4H), 3.45–3.38 (m, 1H), 3.08–2.96 (m, 1H), 1.95–1.32 (m, 6H)
ppm; Anal. (C20H24Cl3NO): C, H, N, Cl.
of 30–HCl was mixed with excess acetyl chloride. After 2 h the
mixture was evaporated, added to acetone and activated C, stirred
for 3 h, filtered and evaporated. The crude product (250 mg) was
converted to the free base and it was crystallized first from acetone
and then EtOAc/hexane to give 52 mg of white solid (mp 116–
117 °C). 1H NMR (CDCl3):
d 7.4–7.2 (m, 5H), 5.08 (d, 1H,
J = 12.1 Hz), 3.78–3.64 (m, 3H), 3.55 (d, 1H, J = 11.4 Hz), 3.06 (d,
1H, J = 12.1 Hz), 2.24 (s, 3H), 1.8–1.3 (m, 6H) ppm. Anal.
(C15H21NO2) C, H, N.
4.3.2.45. ( )-threo-Methylphenidate amide (76).
Using ex-
actly the same procedure as described in the synthesis of 72,
29.4 g of phenylacetonitrile and 39.5 g of 2-bromopyridine gave
19.4 g of the crude amide (fourth intermediate, after epimeriza-
tion) as an threo/erythro mixture (9:1). A 230 mg portion of the free
base was carefully crystallized from EtOAc to yield 194 mg of the
pure threo isomer which showed mp 174–176 °C; 1H NMR (CDCl3):
d 7.8 (br s, 1H), 7.4–7.2 (m, 5H), 5.71 (br s, 1H), 3.9–3.8 (m, 3H),
3.49–3.44 (m, 1H), 2.95–2.90 (m, 1H), 2.59–2.54 (m, 1H), 1.71–
1.45 (m, 5H), 1.21–1.16 (m, 1H) ppm; EIMS m/z 218.149, calcd
for C13H18N2O (M)+, 218.142. Anal. (C13H18N2O): C, H, N.
4.3.2.40.
( )-threo-N-Benzyl-30,40-dichlororitalinol
To 3 mL of DMSO was added KOH (68 mg,
methyl
etherꢃHCl (71).
powder, 1.21 mmol). After stirring for 20 min at rt, 70 (120 mg,
0.4 mmol) in DMSO (3 mL) was added in one portion, followed
immediately by methyl iodide (45 lL, 0.73 mmol). The mixture
was stirred at rt for 24 h and methylene chloride (60 mL) was
added; after washing with water, the organic layer was dried, fil-
tered and evaporated to give 104 mg of a white solid (mp 74–
76 °C, 83%). 1H NMR (CDCl3): d 7.36–7.23 (m, 7H), 7.03 (dd, 1H,
J = 2.2, 8.2 Hz), 3.95 (d, 1H, J = 13.7 Hz), 3.88 (dd, 1H, J = 3.4,
9.3 Hz), 3.75 (d, 1H, J = 13.7 Hz), 3.60 (t, 1H, J = 9.3 Hz), 3.38 (dt,
1H, J = 3.4, 9.5 Hz), 3.28 (s, 3H), 2.98–2.89 (m, 1H), 2.83–2.76 (m,
1H), 2.52–2.44 (m, 1H), 1.67–1.26 (m, 5H), 1.07–1.02 (m, 1H)
ppm; EIMS m/z 378.133, calcd for C21H26Cl2NO (MH)+, 378.139.
HCl salt mp 202–204 °C. Anal. (C21H26Cl3NO): C, H, N, Cl.
4.3.2.46. ( )-threo-40-EthylmethylphenidateꢃHCl (78).
Using
the same method as previously described,12 5.0 g of 4-ethylphenyl-
acetonitrile and 5.4 g of 2-bromopyridine gave 0.39 g of the HCl
salt which showed: mp 204–205 °C; 1H NMR (D2O): d 7.28 (d,
2H), 7.08 (d, 2H), 3.80 (d, J = 9.4 Hz, 1H), 3.64 (m, 1H), 3.58 (s,
3H), 3.30 (m, 1H), 2.93 (m, 1H), 2.92 (q, 2H), 1.8–1.2 (m, 6H),

M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
7237
1.02 (t, 3H) ppm; CIMS m/z 262.180, calcd for C16H24NO2 (MꢀCl)+,
262.181. Anal. (C16H24ClNO2ꢃ0.25 H2O): C, H, N, Cl.
35. Davies, H. M. L.; Gilliatt, V.; Kuhn, L. A.; Saikali, E.; Ren, P.; Hammond, P. S.;
Sexton, G. J.; Childers, S. R. J. Med. Chem. 2001, 44, 1509.
36. Yuan, H.; Kozikowski, A. P.; Petukhov, P. A. J. Med. Chem. 2004, 47, 6137.
37. Yuan, H.; Petukhov, P. A. Bioorg. Med. Chem. Lett. 2006, 16, 6267.
38. Newman, A. H.; Izenwasser, S.; Robarge, M. J.; Kline, R. H. J. Med. Chem. 1999,
42, 3502.
Acknowledgments
39. Robarge, M. J.; Agoston, G. E.; Izenwasser, S.; Kopajtic, T.; George, C.; Katz, J. L.;
Newman, A. H. J. Med. Chem. 2000, 43, 1085.
40. Kulkarni, S. S.; Grundt, P.; Kopajtic, T.; Katz, J. L.; Newman, A. H. J. Med. Chem.
2004, 47, 3388.
41. Kulkarni, S. S.; Newman, A. H.; Houlihan, W. J. J. Med. Chem. 2002, 45, 4119.
42. Gilbert, K. M.; Boos, T. L.; Dersch, C. M.; Greiner, E.; Jacobson, A. E.; Lewis, D.;
Matecka, D.; Prisinzano, T. E.; Zhang, Y.; Rothman, R. B.; Rice, K. C.; Venanzi, C.
A. Bioorg. Med. Chem. 2007, 15, 1146.
This work was supported in part by NIH grants DA06305
(M.M.S. and H.M.D.), DA11541 (H.M.D., M.M.S., and C.A.V.) and
DA018153 (C.A.V.). M.M.S. gratefully acknowledges the technical
assistance of Mr. Adam Eason, Ms. Monica Stafford, Ms. LaKeisha
Nelson, and Mr. Eric Shields.
43. de Gregorio, C.; Kier, L. B.; Hall, L. H. J. Comput. Aided Mol. Des. 1998, 12, 557.
44. Kier, L. B.; Hall, L. H. Molecular Connectivity in Structure–Activity Analysis;
Research Studies Press, John Wiley and Sons: Chichester, UK, 1986.
45. Kier, L. B.; Hall, L. H. Molecular Structure Description: The Electrotopological State;
Academic Press: San Diego, 1999.
46. Kier, L. B.; Hall, L. H. Adv. Drug Res. 1992, 22, 1.
47. Kier, L. B.; Hall, L. H. Pharm. Res. 1990, 7, 801.
48. Hall, L. H.; Mohney, B. K.; Kier, L. B. J. Chem. Inf. Comput. Sci. 1991, 31, 76.
49. Hall, L. H.; Kier, L. B. J. Math. Chem. 1991, 7, 229.
50. Kim, H.-J.; Cho, Y. S.; Koh, H. Y.; Kong, J. Y.; No, K. T.; Pae, A. N. Bioorg. Med.
Chem. 2006, 14, 1454.
Supplementary data
Supplementary data (predicted pIC50’s and associated residuals
for the training set for the 2D- and 3D-QSAR models and the ele-
mental analysis data for the analogs) associated with this article
can be found, in the online version, at doi:10.1016/
j.bmc.2010.08.034.
51. Votano, J. R.; Parham, M.; Hall, L. M.; Hall, L. H.; Kier, L. B.; Oloff, S.; Tropsha, A.
J. Med. Chem. 2006, 49, 7169.
52. Varnek, A.; Kireeva, N.; Tetko, I.; Baskin, I.; Solov’ev, V. P. J. Chem. Inf. Model.
2007, 47, 1111.
53. Hudson, B. D.; Hyde, R. M.; Rahr, E.; Wood, J. Quant. Struct.-Act. Relat. 1996, 15,
285.
54. Kovatcheva, A.; Golbraikh, A.; Oloff, S.; Xiao, Y.-D.; Zheng, W.; Wolschann, P.;
Buchbauer, G.; Tropsha, A. J. Chem. Inf. Comput. Sci. 2004, 44, 582.
55. Available from Tripos Inc., St. Louis, MO.
56. Kier, L. B.; Hall, L. H. Molecular Connectivity in Chemistry and Drug Research;
Academic Press: New York, 1976.
57. Kier, L. B. Quant. Struct.-Act. Relat. 1985, 4, 109.
58. Hall, L. H.; Kier, L. B. J. Chem. Inf. Comput. Sci. 1995, 35, 1039.
[59]. Clark, R. D.; Sprous, D. G.; Leonard, J. M. In Rational Approaches to Drug Design:
Proceedings of the 13th European Symposium on QSAR; Höltje, H.-D., Sippl, W.,
Eds.; Prous Science, 2001, p 475.
60. Cramer, R. D., III; Patterson, D. E.; Bunce, J. D. J. Am. Chem. Soc. 1988, 110, 5959.
61. Whitley, D. C.; Ford, M. G.; Livingstone, D. J. J. Chem. Inf. Comput. Sci. 2000, 40,
1160.
62. Lajiness, M. S. Perspect. Drug Discovery Des. 1997, 718, 65.
63. Snarey, M.; Terrett, N. K.; Willett, P.; Wilton, D. J. J. Mol. Graphics Modell. 1997,
15, 372.
64. Hoffman, B.; Cho, S. J.; Zheng, W.; Wyrick, S.; Nichols, D. E.; Mailman, R. B.;
Tropsha, A. J. Med. Chem. 1999, 42, 3217.
65. Tropsha, A.; Zheng, W. Curr. Pharm. Des. 2001, 7, 599.
66. Xiao, Z.; Xiao, Y.-D.; Feng, J.; Golbraikh, A.; Tropsha, A.; Lee, K.-H. J. Med. Chem.
2002, 45, 2294.
67. Oloff, S.; Mailman, R.; Tropsha, A. J. Med. Chem. 2005, 48, 7322.
68. Itskowitz, P.; Tropsha, A. J. Chem. Inf. Model. 2005, 45, 777.
References and notes
1. Kuhar, M. J.; Ritz, M. C.; Boja, J. W. Trends Neurosci. 1991, 14, 299.
2. Giros, B.; Jaber, M.; Jones, S. R.; Wightman, R. M.; Caron, M. G. Nature 1996, 379,
606.
3. Chen, N. H.; Reith, M. E. In Contemporary Neuroscience: Neurotransmitter
Transporters: Structure, Function, and Regulation; Reith, M. E. A., Ed.; Humana
Press: Totowa, NJ, 1997; p 53.
4. Singh, S. Chem. Rev. 2000, 100, 925.
5. Carroll, F. I. J. Med. Chem. 2003, 46, 1775.
6. Prisinzano, T.; Rice, K. C.; Baumann, M. H.; Rothman, R. B. Curr. Med. Chem.
(CNSA) 2004, 4, 47.
7. Gatley, S. J.; Pan, D.; Chen, R.; Chaturvedi, G.; Ding, Y. S. Life Sci. 1996, 58, PL 231.
8. Kim, D.-I.; Deutsch, H. M.; Ye, X.; Schweri, M. M. J. Med. Chem. 2007, 50, 2718.
9. Reith, M. E. A. NIDA Res. Monogr. 1988, 88, 23.
10. Schweri, M. M.; Skolnick, P.; Rafferty, M. F.; Rice, K. C.; Janowsky, A. J.; Paul, S.
M. J. Neurochem. 1985, 45, 1062.
11. Schweri, M. M. Neuropharmacology 1990, 29, 901.
12. Deutsch, H. M.; Shi, Q.; Gruszecka-Kowalik, E.; Schweri, M. M. J. Med. Chem.
1996, 39, 1201.
13. Froimowitz, M.; Deutsch, H. M.; Shi, Q.; Wu, K.-M.; Glaser, R.; Adin, I.; George,
C.; Schweri, M. M. Bioorg. Med. Chem. Lett. 1997, 7, 1213.
14. Glaser, R.; Adin, I.; Shiftan, D.; Shi, Q.; Deutsch, H. M.; George, C.; Wu, K.-M.;
Froimowitz, M. J. Org. Chem. 1998, 63, 1785.
15. Deutsch, H. M. Med. Chem. Res. 1998, 8, 91.
16. Wayment, H. K.; Deutsch, H. M.; Schweri, M. M.; Schenk, J. O. J. Neurochem.
1999, 72, 1266.
17. Deutsch, H. M.; Ye, X.; Shi, Q.; Liu, Z.; Schweri, M. M. Eur. J. Med. Chem. 2001, 36,
303.
18. Deutsch, H. M.; Kim, D. I.; Holtzman, S. G.; Schweri, M. M.; Spealman, R. D. In
College on the Problems of Drug Dependence, 64th Annual Meeting: Quebec City,
Canada, 2002.
19. Schweri, M. M.; Deutsch, H. M.; Massey, A. T.; Holtzman, S. G. J. Pharmacol. Exp.
Ther. 2002, 301, 527.
20. Davies, H. M. L.; Hopper, D. W.; Hansen, T.; Liu, Q.; Childers, S. R. Bioorg. Med.
Chem. Lett. 2004, 14, 1799.
21. Froimowitz, M.; Gu, Y.; Dakin, L. A.; Nagafuji, P. M.; Kelley, C. J.; Parrish, D.;
Deschamps, J. R.; Janowsky, A. J. J. Med. Chem. 2007, 50, 219.
22. Patrick, K. S.; Williard, R. L.; VanWert, A. L.; Dowd, J. J.; Oatis, J. E., Jr.; Middaugh,
L. D. J. Med. Chem. 2005, 48, 2876.
23. Froimowitz, M.; Gu, Y.; Dakin, L. A.; Kelley, C. J.; Parrish, D.; Deschamps, J. R.
Bioorg. Med. Chem. Lett. 2005, 15, 3044.
ˇ
69. Marini, F.; Roncaglioni, A.; Novic, M. J. Chem. Inf. Model. 2005, 45, 1507.
70. Dutta, D.; Guha, R.; Wild, D.; Chen, T. J. Chem. Inf. Model. 2007, 47, 989.
71. Scior, T.; Medina-Franco, J. L.; Do, Q.-T.; Martínez-Mayorga, K.; Yunes Rojas, J.
A.; Bernard, P. Curr. Med. Chem. 2009, 16, 4297.
72. Kim, K. H.; Greco, G.; Novellino, E.. In 3D QSAR in Drug Design: Recent Advances;
Kubinyi, H., Folkers, G., Martin, Y. C., Eds.; Kluwer: Academic, Dordrecht, 1998;
Vol. 3, p 257.
73. Nicklaus, M. C.; Wang, S.; Driscoll, J. S.; Milne, G. W. Bioorg. Med. Chem. 1995, 3,
411.
74. Veith, M.; Hirst, J. D.; Brooks, C. L., III J. Comput. Aided Mol. Des. 1998, 12,
563.
75. Boström, J.; Norrby, P.-O.; Liljefors, T. J. Comput. Aided Mol. Des. 1998, 12,
383.
76. Debnath, A. K. J. Med. Chem. 1999, 42, 249.
77. Perola, E.; Charifson, P. S. J. Med. Chem. 2004, 47, 2499.
78. Guarnieri, F.; Weinstein, H. J. Am. Chem. Soc. 1996, 118, 5580.
79. Hopfinger, A. J.; Tokarski, J. S. In Practical Application of Computer-Aided Drug
Design; Charifson, P. S., Ed.; Marcel Dekker: New York, 1997; p 105.
80. Barnett-Norris, J.; Guarnieri, F.; Hurst, D. P.; Reggio, P. H. J. Med. Chem. 1998, 41,
4861.
24. Froimowitz, M.; Patrick, K. S.; Cody, V. Pharm. Res. 1995, 12, 1430.
25. Giros, B.; El Mestikawy, S.; Bertrand, L.; Caron, M. G. FEBS Lett. 1991, 295, 149.
26. Gilbert, K. M.; Skawinski, W. J.; Misra, M.; Paris, K. A.; Naik, N. H.; Buono, R. A.;
Deutsch, H. M.; Venanzi, C. A. J. Comput. Aided Mol. Des. 2004, 18, 719.
27. Maw, H. H.; Hall, L. H. J. Chem. Inf. Comput. Sci. 2000, 40, 1270.
28. Hoffman, B. T.; Kopajtic, T.; Katz, J. L.; Newman, A. H. J. Med. Chem. 2000, 43,
4151.
29. Christensen, H. S.; Boye, S. V.; Thinggaard, J.; Sinning, S.; Wiborg, O.; Schiott, B.;
Bols, M. Bioorg. Med. Chem. 2007, 15, 5262.
30. Carroll, F. I.; Gao, Y.; Rahman, M. A.; Abrams, P.; Parham, K.; Lewin, A. H.; Boja,
J. W.; Kuhar, M. J. J. Med. Chem. 1991, 34, 2719.
31. Carroll, F. I.; Mascarella, S. W.; Kuzemko, M. A.; Gao, Y.; Abraham, P.; Lewin, A.
H.; Boja, J. W.; Kuhar, M. J. J. Med. Chem. 1994, 37, 2865.
32. Lieske, S. F.; Yang, B.; Eldefrawi, M. E.; MacKerell, A. D., Jr.; Wright, J. J. Med.
Chem. 1998, 41, 864.
33. Muszynski, I. C.; Scapozza, L.; Kovar, K.-A.; Folkers, G. Quant. Struct.-Act. Relat.
1999, 18, 342.
81. Barnett-Norris, J.; Hurst, D. P.; Lynch, D. L.; Guarnieri, F.; Makriyanis, A.; Reggio,
P. H. J. Med. Chem. 2002, 45, 3649.
82. Greenidge, P. A.; Merette, S. A.; Beck, R.; Dodson, G.; Goodwin, C. A.; Scully, M.
F.; Spencer, J.; Weiser, J.; Deadman, J. J. J. Med. Chem. 2003, 46, 1293.
83. Bernard, D.; Coop, A.; MacKerell, A. D., Jr. J. Am. Chem. Soc. 2003, 125, 3101.
84. Bernard, D.; Coop, A.; MacKerell, A. D., Jr. J. Med. Chem. 2005, 48, 7773.
85. Golbraikh, A.; Tropsha, A. J. Comput. Aided Mol. Des. 2002, 16, 357.
86. Golbraikh, A.; Shen, M.; Xiao, Y.-D.; Lee, K.-H.; Tropsha, A. J. Comput. Aided Mol.
Des. 2003, 17, 241.
87. Willett, P.; Winterman, V. Quant. Struct.-Act. Relat. 1986, 5, 18.
88. Johnson, M. A. J. Math. Chem. 1989, 3, 117.
89. Willett, P.; Winterman, V.; Bawden, D. J. Chem. Inf. Comput. Sci. 1986, 26, 36.
34. Zhu, N.; Harrison, A.; Trudell, M. L.; Klein-Stevens, C. L. Struct. Chem. 1999, 10, 91.

7238
M. Misra et al. / Bioorg. Med. Chem. 18 (2010) 7221–7238
90. Molecular Similarity in Drug Design; Dean, P. M., Ed.; Blackie Academic &
Professional: Glasgow, 1995.
Mathematics and Statistics in Chemistry; Kowalski, B. R., Ed.; D. Reidel
Publishing Company: Dordrecht, Holland, 1984.
91. Ferguson, A. M.; Patterson, D. E.; Garr, C. D.; Underiner, T. L. J. Biomol. Screening
1996, 1, 65.
92. Matter, H. J. Med. Chem. 1997, 40, 1219.
93. Wold, H. In Multivariate Analysis; Krishnaiaah, P. R., Ed.; Academic Press: New
York, 1966; p 391.
94. Wold, S.; Albano, C.; Dunn, W. J., III; Edlund, U.; Esbensen, K.; Geladi, P.;
Hellberg, S.; Johansson, E.; Lindberg, W.; Sjöström, M. In Chemometrics:
95. Höskuldson, A. J. Chemom. 1988, 2, 211.
96. Frank, I. E.; Friedman, J. H. Technometrics 1993, 35, 109.
97. Clark, R. D. J. Comput.-Aided Mol. Des 2004, 18, 563.
98. Ding, Y.-S.; Fowler, J. S.; Volkow, N. D.; Dewey, S. L.; Wang, G.-J.; Logan, J.;
Gatley, S. J.; Pappas, N. Psychopharmacology 1997, 131, 71.
99. Halgren, T. A. J. Comput. Chem. 1996, 17, 490.