Molecular switch controlling the binding of anionic bile acid conjugates to human apical sodium-dependent bile acid transporter

Article abstract of DOI:10.1021/jm1003683

The human apical sodium-dependent bile acid transporter (hASBT) may serve as a prodrug target for oral drug absorption. Synthetic, biological, NMR, and computational approaches identified the structure-activity relationships of mono- and dianionic bile acid conjugates for hASBT binding. Experimental data combined with a conformationally sampled pharmacophore/QSAR modeling approach (CSP-SAR) predicted that dianionic substituents with intramolecular hydrogen bonding between hydroxyls on the cholane skeleton and the acid group on the conjugates aromatic ring increased conjugate hydrophobicity and improved binding affinity. Notably, the model predicted the presence of a conformational molecular switch, where shifting the carboxylate substituent on an aromatic ring by a single position controlled binding affinity. Model validation was performed by effectively shifting the spatial location of the carboxylate by inserting a methylene adjacent to the aromatic ring, resulting in the predicted alteration in binding affinity. This work illustrates conformation as a determinant of ligand physiochemical properties and ligand binding affinity to a biological transporter.

Full text of DOI:10.1021/jm1003683

J. Med. Chem. 2010, 53, 4749–4760 4749
DOI: 10.1021/jm1003683
Molecular Switch Controlling the Binding of Anionic Bile Acid Conjugates to Human Apical
Sodium-Dependent Bile Acid Transporter
Rana Rais,† Chayan Acharya,† Gasirat Tririya, Alexander D. MacKerell, Jr.,* and James E. Polli*
Department of Pharmaceutical Sciences, School of Pharmacy, University of Maryland, 20 Penn Street, Baltimore, Maryland 21201.
†Contributed equally to the work.
Received March 23, 2010
The human apical sodium-dependent bile acid transporter (hASBT) may serve as a prodrug target for oral
drug absorption. Synthetic, biological, NMR, and computational approaches identified the structure-
activity relationships of mono- and dianionic bile acid conjugates for hASBT binding. Experimental data
combined with a conformationally sampled pharmacophore/QSAR modeling approach (CSP-SAR)
predicted that dianionic substituents with intramolecular hydrogen bonding between hydroxyls on the
cholane skeleton and the acid group on the conjugate’s aromatic ring increased conjugate hydrophobicity
and improved binding affinity. Notably, the model predicted the presence of a conformational molecular
switch, where shifting the carboxylate substituent on an aromatic ring by a single position controlled
binding affinity. Model validation was performed by effectively shifting the spatial location of the
carboxylate by inserting a methylene adjacent to the aromatic ring, resulting in the predicted alteration in
binding affinity. This work illustrates conformation as a determinant of ligand physiochemical properties
and ligand binding affinity to a biological transporter.
Introduction
hASBT crystal structure hampers understanding of its struc-
ture and function at the molecular level. In particular, the
contribution of the chemistry space beyond the bile acid’s C-24
position to hASBT binding has not been well characterized.6,7
We previously showed C-24 conjugation patterns impacted the
interaction of native bile acids with hASBT.1 Addition of steric
bulk to the C-24 position of CDCA enhanced binding. Inter-
estingly, although native bile acids are monoanions, a single
negative charge was not essential for binding to hASBT.6
Neutral and cationic conjugates exhibited potent inhibitory
activity, while dianions were weak inhibitors.
To further explore the chemistry space of bile acid con-
jugates extending beyond C-24region and its effect on hASBT
binding, a novel series of bile acid analogues were synthesized
and evaluated here for their inhibition potency. Glutamic acid
CDCA amide (glu-CDCA) served as the bile acid conjugate
moiety, to which various substituted anilines were coupled,
yielding anilinyl conjugates of glu-CDCA. The resulting
compounds were then subjected to hASBT inhibition and
subsequent QSAR model development.
The conformationally sampled pharmacophore (CSP)
method is a novel approach7-10 that can be combined with
3D QSAR methods (denoted CSP-SAR). CSP is well-suited
for compound classes with high structural flexibility. In this
method, all accessible conformations for each molecule are
considered during model development, thereby maximizing
the probability of including the bioactive conformer of each
ligand. This approach would appear particularly useful to
examine the important conformational and physical features
of bile acid conjugates, as these conjugates are relatively large
and flexible ligands.
Human apical sodium-dependent bile acid transporter
(hASBTa) is a member of the solute carrier genetic super-
family and is expressed at high levels in the terminal ileum of
the small intestine. It plays a key physiological role in the
active intestinal reabsorption of bile acids. With the bile acid
pool circulating 6-10 times each day, 20-30 g of bile acid are
taken up via hASBT daily. Because of its high transport
efficiency and capacity, hASBT is an attractive target for
prodrug delivery.1 Previously, acyclovir bioavailability was
increased several fold by conjugating acyclovir to cheno-
deoxycholic acid (CDCA) via a valine linker at the bile acid’s
C-24 position.2 Inhibition of hASBT is also an approach to
treat hypercholesterolemia.3-5
To date, a reliable rational design for prodrugs that targets
hASBT has yet to be developed. Moreover, the lack of an
*To whom correspondence should be addressed. For A.D.M.: phone,
410-706-7442; fax, 410-706-5017; E-mail, amackere@rx.umaryland.
edu; address, Department of Pharmaceutical Sciences, School of
Pharmacy, University of Maryland, 20 Penn Street, HSF2 Room 629,
Baltimore, MD 21201. For J.E.P.: phone, 410-706-8292; fax, 410-706-
5017; E-mail, jpolli@rx.umaryland.edu; Department of Pharmaceutical
Sciences, School of Pharmacy, University of Maryland, 20 Penn Street,
HSF2 Room 623, Baltimore, MD 21201.
aAbbreviations: hASBT, human apical sodium-dependent bile acid
transporter; CDCA, chenodeoxycholic acid; glu-CDCA, glutamic acid
CDCA amide; CSP, conformationally sampled pharmacophore; MD,
molecular dynamics; PSA, polar surface area; HD, hydrogen bond
donor; OC, overlap coefficient; SASA, solvent accessible surface area;
log Pc, calculated partition coefficient; NOE, nuclear Overhauser effect;
OSU, N-hydroxysuccinimide; OBT, benzotriazole; HBTU, benzotria-
zol-1-yloxytris-1,1,3,3 tetra methyl uranium hexaflourophosphate;
TEA, triethylamine; EtOH, ethanol; DMF, dimethylformamide; TFA,
triflouroacetic acid; TES, triethylsilane; DIPEA, N,N-diisopropylethyl-
amine; MDCK, Madin-Darby canine kidney.
From the combined experimental/CSP-SAR approach,
hydrogen bonding between the hydroxyl moieties on the
r
2010 American Chemical Society
Published on Web 05/26/2010
pubs.acs.org/jmc

4750 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12
Rais et al.
Scheme 1. Synthesis of Most Monoanionic and All Dianionic Anilinyl Conjugates of glu-CDCAa
a (a) TEA, HBTU, HOBT, DMF, 25 °C, 4 h; (b) TEA, HBTU, NHS, 25 °C, 4 h; (c) TEA, DMF, 40°C, 12 h; (d) TEA, HBTU, DMF, 40-60°C, 24 h;
(e) 10%Pd/C, EtOH, 1 h.
cholane skeleton and a carboxylate beyond the bile acid C-24
region improved activity. Notably, this interaction was pre-
dicted to be highly sensitive to location of the acid substituent
on the aromatic ring, effectively creating a molecular switch
that controls ligand binding to hASBT. This prediction was
then tested via the synthesis and evaluation of additional bile
acid conjugates. The resulting validated model provides a new
perspective to rationally exploit hASBT for enhanced absorp-
tion of drugs containing carboxylate moieties.
Results and Discussion
Results from synthetic efforts and hASBT inhibition ex-
periments are presented, followed by CSP-SAR analysis and
resulting inhibition models. CSP-SAR models are further
classified into quantitative CSP-SAR models, and a qualita-
tive CSP-SAR model.
Conjugate Synthesis. A total of 35 compounds, including
Figure 1. Inhibition profile for 5, the 3-flouro substituted anilinyl
conjugate of glu-CDCA.
33 initial conjugates and two validation compounds, were
synthesized in moderate to high yields, depending on the
nature and position of the substituent group on aniline. Meta
substituents with strong electron withdrawing groups were
not highly reactive, leading to low yields and in some cases
inability to react (e.g., trifluoromethyl and ethylaminoben-
zoate). All conjugates (excluding 2) were synthesized using
Scheme 1, where a large amount of intermediate glu-CDCA
was synthesized.
such as flouro, chloro, and triflouromethyl, provided stron-
ger inhibition than electron donating substituents, such as
methyl and methoxy at the para position. For monoanionic
conjugates, substituent position on the side chain aromatic
ring did not markedly affect binding affinity. Ki values for
monoanions were potent and ranged from 0.275 to 16.3 μM.
Interestingly, 30 was a potent inhibitor, even though the
charge on glutamic acid was protected by R-benzyl, such that
the negative charge was distal, as compared to other mono-
Inhibition of Taurocholate Uptake. All compounds were
potent inhibitors of hASBT, except for the dianions. Figure 1
illustrates inhibition of taurocholate uptake by the m-fluoro
substituted anilinyl conjugate of glu-CDCA in a concentra-
tion dependent manner (Ki = 1.25 μM). Table 1 shows conj-
ugate structures, along with their inhibition values. The table
is arranged with unsubstituted aniline first, followed by
monosubstituted anilines, disubstituted aniline, and then
dianions. In general, most compounds showed potent bind-
ing, with no drastic effect of substituent. Bulkier groups (e.g.,
t-butyl) were preferred. Electron withdrawing substituents,
anions 1-29. This observation is consistent with previous
results where a proximal single negative charge is not re-
quired for binding.6
Substituent position had a significant impact on the
affinity for dianions. Dianions 31, 32, and 35 were potent
(i.e., Ki < 50 μM), while 33 and 34 were weak inhibitors (Ki
values of 108 and 126 μM, respectively). Previous studies
have shown dianions to be weak inhibitors of hASBT.6 The
greater potency of the 2- and the 3-amino benzoic acid
congeners, compared to the 4-amino benzoic acid substitu-
ent, warranted further investigation and is discussed below.

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12 4751
Table 1. Structures and Ki Values of Anilinyl Conjugates of glu-CDCA
a Predicted Ki values were obtained using CSP-SAR quantitative model 1.

4752 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12
Rais et al.
CSP-SAR hASBT Inhibition Model Development: Quanti-
tative Model. In the absence of a crystal structure for hASBT
or that of an inhibitor-transporter complex, a ligand-based
drug design approach was applied.11 Conjugates 1-33 were
subjected to CSP-SAR model analysis to develop an inhibi-
tion model for Ki, in part because these conjugates exhibit
high molecular flexibility, for which CSP-SAR and replica-
exchange MD simulations are particularly advantageous.
For each conjugate, starting with the second-most potent
compound 14, Supporting Information Table S4 lists physico-
chemical descriptor values, as wells as overlap coefficients of
each structural descriptor. Figure 2 identifies conjugate
structural features used for pharmacophore development,
as described in Experimental Section. Overlap coefficients
were calculated relative to 15.
C18-GS-O7 and angle C19-GS-O3 and shows that the
conformational distribution pattern of 33 (in green) is almost
identical to 14 and 15 (in red), while the second and third
weakest inhibitors (31 and 32 in blue) have a differential
pattern for those two structural descriptors. All 33 com-
pounds in the initial data set were monoanionic, except for
31, 32, and 33, which were dianionic. Because dianionic
compounds have previously shown poor affinity toward
hASBT,6 31-33 being the three weakest inhibitors was not
unanticipated. However, 31 and 32 were estimated to be over
2-fold more potent than 33. These observations necessitated
a more detailed evaluation of the poor potency of 33, making
it difficult to include 33 in the general quantitative model. As
discussed below, compound 33’s low potency was qualita-
tively explainable by virtue of its combined dianionic nature
and inability for an intramolecular hydrogen bond. Hence,
33 was subsequently excluded from the quantitative model
but included in the qualitative analysis presented below.
Supporting Information Table S5 shows the best CSP-SAR
model for compounds 1-33; although r2 = 0.837, the model
failed validation due to Q2 = 0.347.
Initially, the efforts to obtain a comprehensive CSP-SAR
model for all 33 conjugates in Table 1 (1-33) were not
successful due to insufficient validation. Molecular descrip-
tors indicated that 33, in spite of being the weakest inhibitor
in the set, possessed high structural similarity with 14 and 15,
the two most potent inhibitors. Figure 3 considers the angle
Quantitative modeling for compounds 1-32 was initiated
using single variable regression analysis of all structural and
physicochemical descriptors with respect to Ki. Supporting
Information Table S6 represents the results from single-
variable linear regression analysis of the molecular descrip-
tors against Ki. ΔGw and log Pc showed the highest correla-
tions withKi. The negative coefficient of these two descriptors
suggested that conjugate hydrophobic character favors
hASBT inhibition (i.e., decrease in Ki). Among physicochem-
ical descriptors, HA ranked third in correlation with Ki. The
positive coefficient of HA was consistent with ΔGw and log Pc
in that HA indicated more hydrogen bond acceptors in-
creased polar character and hence disfavored binding.
Binding affinity was dominated by conjugate physico-
chemical properties. In Supporting Information Table S6,
all conformational descriptors, except for O7-AA-C19
(r2 = 0.02), exhibited negative linear coefficient for the over-
lap coefficient, as expected becaise similarity to 15 promoted
lower Ki value.
Figure 2. Structural features in anilinyl conjugates of glu-CDCA
used for pharmacophore development. Notation is as follows: O3,
position of 3-OH; O7, position of 7-OH; BC, centroid of B and C
rings of steroidal nucleus; C18, location of C-18; C19, location of
C-19; C20, location of C-20; AA, centroid of carboxylic acid
oxygens on R-acid; and GS, centroid of γ-substituent (R1).
Eighty-one candidate quantitative models were identified
via multivariable regression analysis. AIC analysis indicated
that only one model possessed more than 5% probability of
being the best model. Table 2 lists parameters for the three
best fitting quantitative models. Model 1 yielded r2 = 0.914
with 95.5% probability of being the best model. Model 2
yielded r2 = 0.890, with 2.14% probability of being the best
model, while model 3 yielded r2 = 0.886, with 1.31%
probability of being the best model. Figure 4 plots predicted
versus observed Ki from model 1. From leave-one-out vali-
dation, Q2 = 0.888, indicating the robustness of model 1.
Models 2 and 3 also yielded a high Q2 of 0.861 and 0.855,
respectively. Regression plots for models 2 and 3 are shown
in Supporting Information Figure S1.
Figure 3. Probability distributions of structural descriptors: (A)
angle C18-GS-O7 and (B) angle C19-GS-O3. The red line
represents the combined distribution of 14 and 15; the blue line
represents the same for 31 and 32. The green line represents 33.
Table 2. Results from Multivariable Regression Analysis and Modified Akaike Information Criterion (AICC) for the Selected CSP-SAR Modelsa,b
standardized
A (μM)
CSP-SAR models
r2
Q2
F
Wi
K (μM)
A (μM/kcal/mol)
B
standardized B
0.215 (μM)
(1) ΔGw, HD
(2) ΔGw, PSAside
(3) ΔGw, PSA
0.914 0.888 148 0.955
0.890 0.861 113 0.021
-45.9 ( 6.1
-31.9 ( 4.2
-0.365 ( 0.021
-0.342 ( 0.025
-0.348 ( 0.025
5.29 ( 1.4 (μM)
0.0505 ( 0.021 (μM/A )
0.0457 ( 0.022 (μM/A )
-0.950
-0.890
-0.904
2
2
˚
˚
0.154 (μM/A )
2
2
˚
˚
0.886 0.855 109 0.0131 -34.8 ( 5.8
0.138 (μM/A )
a Parameters include free energy of solvation for each compound in water (ΔGw), number of hydrogen bond donors (HD), polar surface area (PSA),
and PSA exclusively for the region beyond C-24 of the bile acid conjugates (PSAside). b A and B represent the coefficient of the independent variables
(molecular descriptors); K represents the coefficient of the intercept.

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12 4753
Figure 4. Regression plot for predicted vs observed Ki values for
anilinyl conjugates of glu-CDCA using CSP-SAR model 1.
Figure 5. Probability distribution of O3-GSmin and O7-GSmin
for conjugates 15, 31, 32, and 33. The maximum probability of
O7-GSmin for 32 was 0.213.
Model 1 comprised of the physicochemical descriptors
ΔGw and HD. As with the single variable regression analysis
results, the negative coefficient of ΔGw and the positive
coefficient of HD indicate hydrophobicity facilitated bind-
ing. Model 2 included ΔGw, PSAside; model 3 consisted of
ΔGw and PSA. Consistent with model 1, ΔGw yielded a
negative coefficient and both PSAside and PSA yielded
positive coefficients in model 2 and model 3, demonstrating
that hydrophobicity improved affinity. The standardized
coefficients of the descriptors suggested that ΔGw had the
highest impact on binding affinity.
It should be noted that a model based on log transformed
data (i.e., pKi rather than Ki) provided inferior models (Q2 <
0.6) than those above for Ki. Supporting Information Table
S7 lists parameters for the two best fitting quantitative
models for pKi. Supporting Information Figure S2 plots
predicted versus observed pKi for these models. Reasons
for these models to be inferior are that log transformed data
represents an order-of-magnitude focused analysis, but ob-
served Ki data ranges only 147-fold, which is less than three
orders-of-magnitude. Additionally, 31 and 32 are notably
less potent than other compounds (about 2-fold less).
CSP-SAR hASBT Inhibition Model Development: Quali-
tative Model. Qualitative analysis was undertaken to include
the nonpotent conjugate 33 in the analysis, as quantitative
models were developed without 33. Because 33 is a dianion,
analysis focused on the three dianionic compounds 31, 32, and
33 and their probability distributions for various pharmaco-
phore distances and angles. Figure 5 shows the probability
distributions for O7-GSmin and O3-GSmin. The distribution
for the potent conjugate 31 showed high probabilities for
Figure 6. Dependence of ΔGw and log Pc on distance between the
3-OH and carboxylate group on the aromatic ring (O3-GA) for 31
(A,B); between 7-OH and carboxylate group on the aromatic ring
(O7-GA) for 32 (C,D); and between carboxylate group on the
aromatic ring and 3-OH (E,F) and 7-OH (G,H) for 33.
˚
conformers with short distances of O3-GSmin (<3 A). Simi-
larly, the distribution of O7-GSmin for the other potent
dianionic conjugate, 32, displayed a high probability at dis-
˚
tances less than 3 A. Meanwhile, nonpotent compound 33 did
not sample shortdistances. Thesestructuraldescriptorsreflect
intramolecular hydrogen bonding between the γ-substituent
and the cholane hydroxyls for 31 and 32, leading to improved
binding as compared to 33.
These intramolecular hydrogen bonds promoted hydro-
phobicity, consistent with the quantitative CSP-SAR.
Figure 6 depicts the distances between carboxylate group
on the γ-substituent (GA) and the hydroxyls against ΔGw
and log Pc. In panels A-D of Figure 6, short γ acid-
hydroxyl distances correlated with ΔGw and log P for both
31 and 32. For these two potent inhibitors, conformers with
short distances exhibited higher ΔGw and log Pc values.
These findings illustrate that the conformers of dianionic
conjugates with intramolecular hydrogen bonding exhibit
greater hydrophobicity, which promotes hASBT binding, as
per the quantitative CSP-SAR models (Table 2).

4754 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12
Rais et al.
Figure 7. Representative conformers of 31, 32, and 33, illustrating the intramolecular hydrogen bond in 31 and 32 but not in 33.
Table 3. Tabulated Approximate Orientation of the Carboxylate at
para (33 and 35) and meta (32 and 34) Positions Associated with the
Insertion of an Additional Methylene Group between the Aromatic
Ring and the Remainder of the Bile Acid Conjugate
Meanwhile, in panels E-H of Figure 6, no conformers
with short distance interactions were observed for 33; con-
sequently, there was no significant elevation in the hydro-
phobicity of 33. Visual analysis of the compact conforma-
tions of 31 and 32, as well as various conformations of 33,
indicated that the para-position of the acid group in 33
disfavored sampling of the conformation for hASBT bind-
ing. Figure 7 depicts representative conformers of 31, 32, and
33. The intramolecular hydrogen bond between the γ acid
and a steroidal hydroxyl is evident in both 31 and 32 but
not in 33. What is most remarkable about the differences
in biological and conformational characteristics of 31 and
32 versus 33 is that the conjugates only differ in the loca-
tion of the carboxylate group on the aromatic ring of the
γ-substituent (Table 1); 31, 32, and 33 are ortho, meta, and
para, respectively. These results suggest that the location of
the acid group acts as a “molecular switch” to allow for
hydrogen bonding with the cholane hydroxyls, leading to
conformations where the acid is “concealed”, thereby afford-
ing hydrophobicity.
CSP-SAR hASBT Inhibition Model Development: Quali-
tative Model Validation. To challenge our molecular switch
hypothesis, two additional conjugates were synthesized.
Conjugate 35 was devised to have high affinity by effectively
shifting the para location of the carboxylate on low-activity
33 to an “effective-meta” position relative to the remainder
of the conjugate, thereby mimicking the location of the
carboxylate in 32. This switch was performed by inserting
a methylene between the aromatic ring and the remainder of
the conjugate in 33, yielding 35. Table 3 presents a schematic
representation of the shift of the carboxylate from the para
position in 33 to the “effective-meta” position in 35, although
the aromatic ring is still formally para substituted. The
structural change from 33 to 35 is predicted to allow for
the necessary intramolecular hydrogen bonding needed for
dianion high affinity binding.
Additionally, a second conjugate was designed as a nega-
tive control, where the molecular switch was applied to abo-
lish intramolecular hydrogen bonding. Conjugate 34 was
designed to have low affinity by shifting the location of the
meta carboxylate in high affinity 32 to an “effective-para”
position by again introducing a methylene bridge adjacent
to the aromatic ring to produce 34. Table 3 schematically
shows how the introduction of the methylene into 32 shifts
the meta substituent to an “effective-para” position in 34,
even though the ring is still formally meta substituted.
Conjugate 34 was predicted to not undergo intramolecular
hydrogen bonding, as the molecular switch was designed to
be turned off.
Conjugates 34 and 35 were initially subjected to in silico
analysis via MD simulations as part of the CSP-SAR meth-
odology. Figure 8 shows the dependence of ΔGw and log Pc
on the distance between 7-OH and carboxylate group on
the aromatic ring (O3-GA) for 34 and 35. Relative to 32,
the addition of the methylene bridge in 34 disrupted the
intramolecular hydrogen bonding that favors hASBT bind-
ing; in essence, the molecular switch was turned off. In
comparing Figure 8A,B for 34 to Figure 6C,D for 32, loss
of intramolecular hydrogen bonding yielded no conformer
with enhanced hydrophobicity. Meanwhile, in comparing
Figure 8C,D for 35 to Figure 6G,H for 33, short-distance
conformations with enhanced hydrophobicity were observed
when the switch was turned on. The CSP-SAR binding
affinity from quantitative model 1 for 35 was 39.6 μM. The

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12 4755
Table 4. Observed NOE Enhancements from NMR Studies and
Calculated Shortest Proton-Proton Distances from MD Simulationsa
compd
no.
proton
irradiatedb
observed NOE
enhancement
from NMR
studiesb
shortest
distance
from MD
percent of
MD conformers
satisfying NMR
data (%)
˚
simulations (A)
32
31
35
21
18
33
18
19
21
3.61
5.45
4.90
16.6
0.0500
5.68
18
19
21
3.34
3.28
4.99
33.0
1.35
0.775
31
18
19
4.90
1.82
5.16
5.68
100
13.9
31
21
19
3.61
1.82
2.11
16.6
100
100
Figure 8. Dependence of ΔGw and log Pc on distance between
7-OH and carboxylate group on the γ-substituent (O7-GA) for
(A,B) 34 and (C,D) 35.
34
18 (weak)
21 (weak)
2.10
4.54
5.97
100
27.2
model predicts 34 to be a weak inhibitor, as 34 did not exhibit
conformational distributions that were associated with po-
tent Ki.
0.0500
33
31
32
18
19
21
32
31
2.13
100
Conjugates 34 and 35 were then synthesized and subjected
to hASBT inhibition. In agreement with the predictions,
observed Ki values for 34 and 35 were 126 and 41.0 μM,
respectively (Table 1). The CSP-SAR method and the
experimental data demonstrate that intramolecular hydro-
gen bond plays a critical role in the binding potency of
dianionic conjugates of glu-CDCA. This hydrogen bond
acted as a molecular switch to control the binding of such
compounds with hASBT. Therefore, according to our model
and in contrast to the present knowledge, dianionic conju-
gates of bile acids can possess potent binding affinity if they
attain sufficient hydrophobicity via specific intramolecular
hydrogen bonding. Additionally, the previously reported
poor activity of glu-CDCA dianion6 is in agreement with
the evaluation here that glu-CDCA is hydrophilic and is
not able to benefit from a molecular switch due to the
location of the carboxylate group on the aromatic ring of
the γ-substituent.
2.13
100
none
none
none
2.40c
2.35c
2.02c
17.7c
1.73c
13.8c
a Interacting proton pairs between aromatic carbons and cholane
moiety were identified from 1D NOE spectra. NMR and MD simula-
tions indicate 32 exhibited compact conformations, while 33 was
extended. b Number represents the carbon at which the proton was
irradiated, or carbon with observed NOE enhancement. 18, 19, and 21
are on the cholane skeleton, while 31, 33, 34, and 35 are on the side chain
aromatic ring near the acid (see Supporting Information Table S8).
cAlthough no NOEs were observed when protons on carbons 18, 19, and
21 were irradiated, MD simulation results are reported for proton pairs
18-31, 19-31, and 21-31.
were only very weak. Likewise, irradiation on all three
methyl groups on 33 resulted only in very weak short-range
enhancements.
NMR results supported the qualitative CSP-SAR analy-
sis. NOE enhancements require the two interacting protons
Conformational Analysis via NMR. The 1D proton, 1D
carbon, and COSY spectra of 32 in D2O-phosphate were
almost identical to corresponding spectra in D2O, such that
chemical shift assignment of 32 in D2O-phosphate were
applied to spectra in D2O (Supporting Information Table
S8). Table 4 represents the pairs of protons on the carbon
atoms in 32 and 33 that showed peaks in 1D NOE. As shown
in Table 4, 1D steady-state NOE experiments of 32 indi-
cated that 4 s irradiation of protons on C31 (7.84 ppm) and
C35 (7.72 ppm) resulted in NOE enhancement of protons on
methyls C18 (0.5 ppm), C19 (0.8 ppm), and C21 (0.84 ppm)
of the cholane skeleton. Table 4 also lists the percentage of
MD conformers that satisfy NMR data. For example,
33.0% of 32’s conformers showed a short distance (less
˚
to be within approximately 6 A or less. Such short distance
conformers were observed from MD simulation of 32 but not
from 33. The percentage of conformers showing intramole-
cular hydrogen bonding (i.e., either O3-GA distance or
˚
O7-GA distance less than 3 A) for 32 and 33 were 71.8%
and 0%, respectively. Accordingly, both NMR and qualita-
tive CSP-SAR analysis found 32 to sample compact con-
formations, while 33 displayed extended conformations
only. From CSP analysis, 32 exhibited intramolecular hydro-
gen bond between the carboxylate group on the γ substituent
and 7-OH (Figure 7) resulting in enhanced hydrophobicity
and inhibition potency. NMR was consistent with this
observation, because NOEs between the aromatic ring and
the cholane skeleton in 32 indicate a folded conformation.
Meanwhile, no NOEs were observed between the aromatic
ring and the cholane skeleton in 33, suggesting lack of
hydrogen bonding.
˚
than 6 A) between protons on C-35 and C-18. Irradiation of
the aromatic protons on C33 and C34 did not yield any
notable NOE enhancements. The reciprocal proton irradia-
tions of C18 (0.50 ppm) and C21 (0.84 ppm) caused en-
hancement of the aromatic proton on C31 (7.84 ppm). In
contrast, for 33, irradiation of protons on C31, C32, C34,
and C35 produced no long-range enhancements; short-
range enhancements, such as those to immediate neighbors,
Comparison to Previous hASBT Inhibition Models. A 3D
QSAR model for inhibition of rabbit ASBT was developed
using neutral and monoanionic nonbile acids.12 In the
model, the five-membered D ring, the C-21 methyl group,

4756 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12
Rais et al.
the 7- or 12-OH in R-orientation, and a negative charge on
the C-24 side chain promoted affinity. The 3-OH and the cis-
configuration of A and B rings of the cholane nucleus did not
impact binding. This model did not include dianionic bile
acid analogues or explain the contribution of conforma-
tional distributions to activity.
We recently studied the inhibition properties of amino-
piperidine conjugates of glu-CDCA.7 The quantitative CSP-
SAR models for aminopiperidine analogues included the
electrostatic contribution of solvation free energy (GBener)
and log Pc, where hydrophobic character favored binding,
consistent with the present study’s CSP-SAR models for
aniline conjugates. The quantitative model for aminopiper-
idine conjugates included two angle parameters (O7-AS-
BC and O7-AS-C20, where AS represents the R-sub-
stituents), which were evaluated for aniline conjugates here.
However, these two angles did not impact activity for
anilinyl conjugates here.
Conformation, Hydrophobicity, and Binding Activity. Results
here indicate that ligand conformation can impact ligand
physical properties, in this case hydrophobicity, and subse-
quent binding. For example, Gomar et al.13 demonstrated
that 3D QSAR analysis for the IC50 of anti-HIV compounds
was improved when the lipophilicity of the relevant con-
former(s) was applied. Gaillard et al.14 found that 3D QSAR
analysis for R1-adrenoceptor ligands was improved by con-
sidering conformationally dependent lipophilicity. Findings
here about the conformation of dianions and hASBT inhibi-
tion are similar to these previous works. Hydrophobicity is
well-known to be a crucial factor for ligand-protein binding
interaction. For example, Parker et al.15 observed that activity
of 5-HT2R agonists was strongly dependent on compound
lipophilicity.
While the associations of conformation, lipophilicity, and
activity have been described previously, the present study
elucidates the role of conformation in ligand hydrophobicity
and subsequent binding activity for a transporter protein.
Perhaps an analogous finding is the unusually high passive
permeability of cyclosporine A across lipid bilayers. In spite
of cyclosporine A being a relatively large cyclic peptide, this
drug shows high permeability across membrane and is well
absorbed from the gut after oral administration. Drug con-
formation, including intramolecular hydrogen bonding, has
been considered as a determinant in cyclosporine A’s high
permeability.16
This hypothesis was tested by designing conjugates 34 and
35 that should have altered binding properties based on
intramolecular hydrogen bonding ability. MD simulations
showed that the insertion of a methylene bridge in 32 to yield
34 resulted in loss of intramolecular hydrogen bond, while the
insertion of a methylene bridge in 33 to yield 35 resulted in
conformers with an intramolecular hydrogen bond between
carboxylate group on the aromatic ring and 7-OH. Accord-
ingly, 35 attained conformations with enhanced hydrophobi-
city, which led to improved binding affinity. Synthesis and
assay of these compounds validated the predictions.
Overall, the present work indicates that dianionic conju-
gates can be designed with high binding affinity to hASBT by
exploiting an underpinning molecular switch mechanism.
Activity of potent dianions depended on the location of the
acid group and the acid group’s function as a “molecular
switch” to allow for intramolecular hydrogen bonding, lead-
ing to conformations where the acid is “concealed”; thereby
enhancing conjugate hydrophobicity. Results here focus on
the transporter hASBT and analogues of its native ligand, bile
acids. These findings have broader applicability in illustrating
ligand conformations that are responsible for ligand binding
to biological targets. Enhanced ligand hydrophobicity via a
molecular switch or other conformation-dependent mechan-
isms may contribute to the activity in other ligand-receptor
scenarios.
Experimental Section
Materials. [3H]-Taurocholic acid (10 μCi/mmol) was pur-
chased from American Radiolabeled Chemicals, Inc. (St. Louis,
MO). Taurocholate was obtained from Sigma Aldrich (St. Louis,
MO). Chenodeoxycholate (CDCA) was obtained from TCI
America (Portland, OR). Protected glutamic acid analogues
were purchased from Novabiochem (Gibbstown, NJ). Aniline
and substituted anilines were obtained from Sigma Aldrich (St.
Louis, MO). Geneticin, fetal bovine serum (FBS), trypsin, and
DMEM were purchased from Invitrogen (Rockville, MD). All
other reagents and chemicals were of the highest purity commer-
cially available.
Synthetic Procedures: Overview. A series of substituted anili-
nyl conjugates glu-CDCA were synthesized using two synthetic
approaches. In the first approach (used for synthesis of all
targets except 2), R-benzyl glutamic acid was first coupled to
CDCA via either N-hydroxysuccinimide (OSU) ester or benzo-
triazole (OBT) ester. Various substituted aniline probes were
then coupled to glutamic acid by stirring either at RT or 60 °C
using O-benztriazol-1-yloxytris-1,1,3,3 tetra methyl uranium
hexaflourophosphate (HBTU) as the activating agent and
triethylamine (TEA) as the base. The resulting neutral com-
pounds were then subjected to hydrogenation in a Parr shaker
for 1-2 h in ethanol (EtOH) and 10% palladium to remove the
R-benzyl group, yielding the mono and dianionic targets. To
synthesize 2 (Scheme 2), a second approach was needed.
All neutral compounds intermediates were purified by col-
umn chromatography using a gradient of hexane and ethyl
acetate. All final target compounds were obtained as solids after
deprotection. Identity and purity were confirmed by TLC, MS,
NMR, and elemental analysis. All final target compounds
possessed g95% purity.
Conclusion
Thirty-three anilinyl conjugates of glu-CDCA were
synthesized and provided a range of hASBT inhibition
potencies, ranging from 0.275 to 108 μM. Three quantitative
CSP-SAR models that excluded the weak dianion 33 were
developed and identified physicochemical features that favor
binding. In addition, a qualitative CSP-SAR model was
derived to explain affinity values of the dianionic molecules,
including 33. The model predicted the presence of a molec-
ular switch associated with the spatial location of the car-
boxylate group on the aromatic ring, thereby controlling
intramolecular hydrogen bonding between the carboxylate
group on the aromatic ring and either 3-OH or 7-OH of the
cholane skeleton. This interaction afforded ligand confor-
mations that promoted ligand hydrophobicity, thereby im-
proving binding, consistent with the quantitative CSP-SAR
models.
Synthetic Procedure: Method A. Synthesis of HOBT Ester of
CDCA (ii).7 To a solution of CDCA (i) (5 g, 12.74 mmol) in
DMF (15 mL), HBTU (5.07 g, 13.37 mmol), HOBt (0.86 g, 6.37
mmol), and TEA (2.33 mL, 13.37 mmol) were added and stirred
at RT for 4 h. After the reaction was complete, it was quenched
with H2O, yielding a crude product, which was then extracted
into EtOAc (3ꢀ). The combined organic layers were washed
with brine (1ꢀ), dried over Na2SO4, and filtered. EtOAc was

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12 4757
Scheme 2. Synthesis of p-Chloro Substituted Anilinyl Conjugates of glu-CDCAa
a (a) EDC, HOBt, DIPEA, DMF 25 °C, 2 h; (b) TFA/TES 25 °C, 48 h; (c) CDCA-OBt (ii), DIPEA, DMF 25 °C, 12 h.
removed under reduced pressure, resulting in CDCA-OBt (ii), a
white fluffy solid which was used without any further purifica-
tion. MS showed appropriate peaks: [M þ 1] 510.8, [M þ Na]
532.8.
Synthesis of N-Hydroxysuccinimide Ester of CDCA (iii). To a
solution of CDCA (i) (5 g, 12.74 mmol) in DMF (15 mL), HBTU
(5.07 g, 13.37 mmol), anhydrous N-hydroxysuccinimide (1.48 g,
12.8 mmol), and TEA (1.7 mL, 12.8 mmol) were added and
stirred at RT for 4 h. Water was added to quench the reaction,
followed by extraction as described above for CDCA-OBT
ester. MS showed appropriate peaks: [M þ 1] 512.8, [M þ Na]
534.8.
Synthesis of r-Benzyl Ester Glutamic Acid CDCA Amide (glu-
CDCA).7 The activated ester of CDCA (either ii or iii) (5.2 g,
10.2 mmol) was reacted with R-benzyl-glutamic acid (2.4 g, 10.2
mmol), along with 1.42 mL of triethylamine (TEA), in anhy-
drous DMF (20 mL). After heating the reaction at 40 °C over-
night, the reaction was quenched with 50 mL of cold water. The
resulting sticky white precipitate was extracted into 40 mL
volumes of EtOAc (3ꢀ) and washed with 1N HCl (3ꢀ), 15 mL
of water (ꢀ3), and 15 mL of brine (1ꢀ). The organic layer was
dried over Na2SO4 and filtered. Removal of EtOAc under
reduced pressure yielded an off-white fluffy solid, which was
used without any further purification. MS showed appropriate
peaks: [M þ 1] 612.8, [M þ Na] 634.8.
vacuum. The solid was extracted with EtOAc (3ꢀ) and dried
over Na2SO4, followed by filtration.
Synthetic Procedure: Method B. Synthesis of p-Chloroaniline
Conjugates of r-Benzyl Glutamic Acid CDCA Amide. To a
solution of N-R-t-boc-L-glutamic acid R-t-butyl ester (vi) (3 g,
9.89 mmol) in DMF (30 mL) was added p-chloroaniline (10.88
mmol), 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC,
2.08 g, 10.88 mmol), HOBt (1.47 g, 10.88 mmol), and DIPEA
(3.8 mL, 21.75 mmol). The reaction mixture was stirred at RT for
2 h. After reaction was complete, 30 mL of 1 M HCl was added,
and the mixture was extracted with EtOAc (3ꢀ). The combined
organic layers were washed with brine (1ꢀ), dried over MgSO4,
and filtered. EtOAc was removed under vacuum, yielding vii,
which was further purified by column chromatography (SiO2/3:7
ethyl acetate/hexane) to give a white solid.
To vii was added trifluoroacetic acid/triethyl silane (TFA/
TES, 95:5 ratio, 10 mL/1 g of vii). The resulting solution was
stirred at RT for 48 h, followed by the removal of the solvent
under reduced pressure to yield an oil of viii, which was solidi-
fied upon trituration with ether. The pure solid viii was collec-
ted by filtration and washed with ether. To a solution of ii
(1.2 equiv) in DMF, viii (1 equiv) was added, followed by the
addition of DIPEA (1.2 equiv). The reaction mixture was stirred
at RT for 12 h. The reaction was quenched by addition of 0.5 M
HCl solution (30 mL). The mixture was stirred until product 2
solidified, which was then filtered and washed with water. NMR
and MS results (Supporting Information Table S1) and ele-
mental analysis (Supporting Information Table S2) verified the
identity and purity of the reported compounds.
Binding Affinity Studies. Stably transfected hASBT-MDCK
cells were cultured as previously described.17 Briefly, cells were
grown at 37 °C, 90% relative humidity, 5% CO2 atmo-
sphere, and fed every two days. Culture media consisted on
DMEM supplemented with 10% FBS, 50 units/mL penicillin,
and 50 μg/mL streptomycin. Geneticin was added at 1 mg/mL to
maintain selection pressure. Cells were passaged after four days
or after reaching 90% confluency.
The binding affinities of 1-35 were evaluated via taurocholate
inhibition studies, where taurocholate is a potent substrate for
hASBT. cis-Inhibition of taurocholic acid (3H-TCA) by conju-
gate of interest was performed usinghASBT-MDCK monolayers
in 12-well plates (3.8 cm2, Corning, Corning, NY). Briefly, cells
were seeded at a density of 1.5 million/well. Prior to study, on
day 4, cells were induced with 10 mM sodium butyrate for about
12-17 h. To conduct the inhibition assay, cells were washed three
times with Hank’s balanced salt solution (HBSS) and then
exposed to donor solution containing taurocholic acid (2.5 μM
with 0.5 μCi/mL 3H-TCA) and conjugate (0-100 μM for mono-
anions and 0-1000 μM for dianions). Inhibition studies were
conducted at 37°Cfor 10minon anorbital shaker (50rpm). After
10 min, donor solution was removed and cells were washed three
Synthesis of Aniline Conjugates of r-Benzyl Glutamic Acid
CDCA Amide (v). To a solution of glu-CDCA iv (1.25 g 2 mmol)
in DMF (15 mL) was added TEA (0.265 mL) and HBTU (0.76 g
2 mmol) for 15 min, followed by the addition of appropriate
aniline analogue. The reaction mixture was stirred at 40-60 °C
overnight. The DMF solution was then extracted with 40 mL
volumes of EtOAc (3ꢀ). The organic extract was washed with
1N HCl (3ꢀ), 1N NaOH (3ꢀ), 15 mL of water (3ꢀ), and 15 mL
of brine (1ꢀ). The organic extract was dried over Na2SO4 and
filtered. The EtOAc extract was evaporated under vacuum to
yield crude product. The neutral compounds were further
purified by column chromatography (SiO2/8:2 EtOAc/hexane)
to give a white solid. MS showed appropriate peaks for neutral
compounds.
Monoanionic target conjugates 1 and 3-30 were obtained
from v via catalytic hydrogenation to remove the R-benzyl ester
(10% Pd/charcoal in EtOH at 40 psi for 1 h). The product was
filtered through celite, and the solvent was removed under
reduced pressure to give targets 1 and 3-30. Dianionic con-
jugates 31-35 were obtained from v as above via catalytic
hydrogenation to remove the R-benzyl ester (10% Pd/charcoal
in EtOH at 40 psi for 1 h). The resulting solution was filtered,
and solvent was removed under reduced pressure. The benzoate
esters were then hydrolyzed, using 5% NaOH in methanol for
2 h. The resulting solution was acidified using 12N HCl drop-
wise to yield a precipitate. Methanol was evaporated under

4758 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12
Rais et al.
times with chilled sodium-free buffer (where NaCl was replaced
by137 mM tetraethylammonium chloride). Cells werelysed using
250 μL of 1N NaOH and allowed to stand for at least 2 h. After
this time, cell lysate was neutralized with 250 μL of 1N HCl and
counted for associated radioactivity using an LS6500 liquid
scintillation counter (Beckmann Instruments, Inc., Fullerton,
CA).
GSmax, C18-GSmin, C19-AA, C19-GS, C19-GSmax, C19-
GSmin, O3-AA, O3-GS, O3-GSmax, O3-GSmin, O7-AA, O7-
GS, O7-GSmax, O7-GSmin) and 26 pharmacophore angles (AA-
GS-O3, AA-GS-O7, C18-GS-O3, C18-GS-O7, C19-GS-
O3, C19-GS-O7, GS-O3-AA, GS-O3-C18, GS-O3-C19,
GS-O7-AA, GS-O7-C18, GS-O7-C19, O3-AA-BC, O3-
AA-GS, O3-C18-GS, O3-C19-GS, O7-AA-BC, O7-AA-
GS, O7-C18-GS, O7-C19-GS, O3-AA-GS, O3-C18-GS,
O3-C19-GS, O7-AA-GS, O7-C18-GS, O7-C19-GS, O3-
AA-BC, O3-AA-C18, O3-AA-C19, O7-AA-BC, O7-AA-
C18, O7-AA-C19) as shown in Figure 2. Probability distributions
of conformational sampling involving the structural descriptors
were obtained from the MD trajectories. GSmax and GSmin repre-
sent the maximum and minimum distances of the γ-substituent
from the corresponding point. For example, C18-GSmax denotes
the maximum distance of the γ-substituent from C-18. One-dimen-
sional probability distributions of all structural descriptors were
Inhibition data was regressed using a modified Michaelis-
Menten equation to determine the inhibition constant Ki, as
previously described,7 including parallel studies on each occa-
sion to estimate Kt, Jmax, and Pp for taurocholic acid.
Computational Methods. In silico models of conjugates
were built using the program CHARMM18 with the all-atom
CHARMM general force field.19 Each molecule was subjected
to 1000 steps of steepest descent (SD) and 500 steps of Newton-
Raphson (NRAP) energy minimization in the gas phase to a
gradient of 10-4 kcal/mol/A. CHARMM-minimized structures
˚
˚
were subjected to replica exchange molecular dynamics (MD)
simulations to obtain the conformational distribution of each
molecule.20-22 Replica exchange MD simulations involved
20 ns simulations with four replicas of each molecule bet-
ween 300 and 400 K using an exponential scale (300, 330, 363,
and 400 K). Exchange of replicas was attempted after every
250 MD steps. MD simulations were performed using Langevin
dynamics23 with an integration time step of 0.002 ps, and the
aqueous solvation was modeled implicitly via the generalized
Born continuum solvent model (GBMV).24,25 SHAKE was
applied to all covalent bonds involving hydrogens.26 Conforma-
tions saved every 20 ps and obtained from all four replicas were
used for the analysis. The protonation states of the ionizable
groups present in the molecules were determined based on the
experimental pH of 6.8. All free acids at R1 and R2 (Table 1)
were assumed to be deprotonated.
Several physicochemical descriptors were calculated for
each conjugate by averaging over the full trajectories. The free
energy of solvation for each compound in water (ΔGw) and
octanol (ΔGo) was calculated using the semiempirical quantum
chemical program AMSOL6.8 by invoking the AM1-SM5.42R
continuum solvent model.27 The free energy of solvation in
a solvent medium was calculated by taking the difference
between the heat of formation plus solvation energy of the
solvated system in its relaxed electronic state and that for the
gas-phase system. Partition coefficients (log Pc) were calculated
using eq 1.28
derived with a bin size of 0.1 A and 1° for distances and valence
angles, respectively.
Overlap coefficient (OC) of each structural descriptor were
used to quantify the extent of similarity of the regions of
conformational space as defined by the various distances,
angles, and dihedrals sampled by each conjugate. OC for
continuous probability density functions is given by Reiser
and Faraggi,37
N
X
OC ¼
minf f1ðxÞ, f2ðxÞg dx
ð2Þ
i ¼ 1
where f1(x) and f2(x) are probability density functions of the two
distributions. For discrete probability functions, eq 2 can be
written as,
N
X
OC ¼
minðPAi, PB Þ
ð3Þ
i
i ¼ 1
where PA and PB are the probability in bin i for compounds
i
i
A and B, and N is the total number of bins. OC values for each
compound were calculated with respect to the most potent
inhibitor in the data set (Compound 15, Ki = 0.275 μM).
The overlap coefficients of all molecules were regressed with
respect to their inhibition constants. Similarly, physicochemical
descriptors were also subjected to linear regression against the
respective Ki values of the molecules. Statistical program “R”
was used for the regression analysis.38 The structural and
physicochemical descriptors were then combined to make a
complete set of molecular descriptors for model develop-
ment. All descriptors were selected for multivariable regression
analysis.
Molecular descriptors were subjected to multivariable regres-
sion in all possible combinations of two. However, any combi-
nation having correlation (r) greater than 0.8 were not
considered. The correlation matrix used to determine the inter-
nal correlation between descriptors is presented in Supporting
Information Table S3. All pairs of descriptors with p-values of
the independent variables, as well as the intercept, less than 0.05
were selected for further analysis. Additional descriptors were
sequentially added to each selected pair of descriptors. This
approach was applied to all selected groups until all possible
combinations of descriptors yielded p-values greater than 0.05.
All combinations of descriptors with p-values less than 0.05
formed the set of candidate models. The standardized coeffi-
cients of the independent variables39 of the candidate models
were also calculated using eq 4.40
1
ð1Þ
log Pc ¼ -
ðΔGw - ΔGoÞ
2:303RT
Average solvent accessible surface area (SASA) and polar sur-
face area (PSA) were calculated using the Lee and Richards
method as implemented in CHARMM.29 The radius of the
˚
solvent molecule was taken as 1.4 A, which approximates the
radius of a water molecule.30 PSA for each compound was
calculated by adding the contribution of polar atoms (N, O,
F, Cl, Br, and hydrogen atoms covalently attached to any of
these atoms). For calculation of PSA, halogens were included as
they could potentially take part in polar interaction with other
polar atoms.31-34 SASA and PSA were also calculated exclu-
sively for the region beyond C-24 of the bile acid conjugates
(SASAside and PSAside) to access the contribution of the sub-
stituents alone toward the interaction with hASBT. Surface area
terms were obtained as averages over the conformations saved
from the MD simulations. Molecular weight (MW), molar
refractivity (MR), number of rotatable bonds (brotN), and
flexibility parameter (KierFlex) were calculated using the pro-
gram MOE.35 Additionally, number of hydrogen bond donors
(HD) and hydrogen bond acceptors (HA) for each molecule was
calculated using Discovery Studio (Accelrys Inc.).36
Sx
k
b0k ¼ bk
ð4Þ
Sy
Conformational properties of molecules required for CSP model
development included 20 pharmacophore distances (BC-AA,
BC-GS, BC-GSmax, BC-GSmin, C18-AA, C18-GS, C18-
where bk and b0k are the metric and standardized coefficient of
independent variable xk, respectively, Sx and Sy are the stan-
dard deviation of xk and dependent variable y. Standardized
k

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12 4759
coefficients were calculated to determine the relative signifi-
cance of the independent variables to influence the biological
activity. Model evaluation was performed using AICC (modified
Akaike information criterion)41,42 analysis and leave-one-out
cross validation method, as described previously.7
reevaluation of hydrophobic moieties essential for biological ac-
tivity. J. Med. Chem. 2007, 50, 1799–1809.
(10) Bernard, D.; Coop, A.; MacKerell, A. D., Jr. Conformationally
sampled pharmacophore for peptidic delta opioid ligands. J. Med.
Chem. 2005, 48, 7773–7780.
(11) Ekins, S.; Mirny, L.; Schuetz, E. G. A Ligand-Based Approach to
Understanding Selectivity of Nuclear Hormone Receptors PXR,
CAR, FXR, LXRR, and LXRβ. Pharm. Res. 2002, 19, 1788–1800.
(12) Baringhaus, K. H.; Matter, H.; Stengelin,þS.; Kramer, W. Substrate
specificity of the ileal and the hepatic Na /bile acid cotransporters
of the rabbit. II. A reliable 3D QSAR pharmacophore model for
the ileal Naþ/bile acid cotransporter. J. Lipid Res. 1999, 40, 2158–
2168.
(13) Gomar, J.; Giraud, E.; Turner, D.; Lahana, R.; Carrupt, P. A. 3D-
LogP: An Alignment-Free 3D Description of Local Lipophilicity
for QSAR Studies. Methods Mol. Biol. 2008, 275, 215–260.
(14) Gaillard, P.; Carrupt, P. A.; Testa, B.; Boudon, A. Molecular
lipophilicity potential, a tool in 3D QSAR: method and applica-
tions. J. Comput.-Aided Mol. Des. 1994, 8, 83–96.
(15) Parker, M. A.; Kurrasch, D. M.; Nichols, D. E. The role of lipophi-
licity in determining binding affinity and functional activity for
5-HT2A receptor ligands. Bioorg. Med. Chem. 2008, 16, 4661–4669.
(16) Augustijns, P. F.; Brown, S. C.; Willard, D. H.; Consler, T. G.;
Annaert, P. P.; Hendren, R. W.; Bradshaw, T. P. Hydration
Changes Implicated in the Remarkable Temperature-Dependent
Membrane Permeation of Cyclosporin A. Biochemistry 2000, 39,
7621–7630.
NMR Methods. NMR experiments were performed on a
Varian Inova instrument operating at a proton Larmor fre-
quency of 500 MHz. Saturated solutions of each 32 and 33
were prepared in D2O. Analysis of the tumbling times indicated
that the compounds are not aggregating in solution. One-
dimensional proton steady-state NOE difference experiments
were conducted using irradiation times of 1.5, 2, and 4 s.
Aromatic protons on the aniline substituent were irradiated to
evaluate NOE enhancements in the steroidal region in order to
elucidate differences between 32 and 33 conformational proper-
ties. NMR data were processed with Varian software, NMR
Pipe (Delaglio and BAX, NIH), and SpinWorks (University of
Manitoba, Kurt Marat). Sparky (UCSF, Kneller and Goddard)
was used for analysis of the 2D spectra. NMR assignments of 32
were obtained from 1D proton and carbon experiments in D2O-
phosphate (pH 7.4) to achieve a high concentration, 2D studies
(i.e., COSY, HMQC, and HMBC) and published assignments of
CDCA.43
(17) Balakrishnan, A.; Sussman, D. J.; Polli, J. E. Development of
stably transfected monolayer overexpressing the human apical
sodium-dependent bile acid transporter (hASBT). Pharm. Res.
2005, 22, 1269–1280.
(18) Brooks, B. R.; Brooks, C. L., III; MacKerell, A. D., Jr.; Nilsson, L.;
Petrella, R. J.; Roux, B.; Won, Y.; Archontis, G.; Bartels, C.;
Boresch, S.; Caflisch, A.; Caves, L.; Cui, Q.; Dinner, A. R.; Feig,
M.; Fischer, S.; Gao, J.; Hodoscek, M.; Im, W.; Kuczera, K.;
Lazaridis, T.; Ma, J.; Ovchinnikov, V.; Paci, E.; Pastor, R. W.;
Post, C. B.; Pu, J. Z.; Schaefer, M.; Tidor, B.; Venable, R. V.;
Woodcock, H. L.; Wu, X.; Yang, W.; York, D. M.; Karplus, M.
CHARMM: The biomolecular simulation program. J. Comput.
Chem. 2009, 30, 1545–1614.
(19) Vanommeslaeghe, K.; Hatcher, E.; Acharya, C.; Kundu, S.;
Zhong, S.; Shim, J.; Darian, E.; Guvench, O.; Lopes, P.; Vorobyov,
I.; MacKerell, A. D., Jr. CHARMM General Force Field
(CGenFF): A Force Field for Drug-Like Molecules Compatible
with the CHARMM All-Atom Additive Biological Force Fields.
J. Comput. Chem. 2010, 31, 671–690.
Acknowledgment. This work was support in part by
National Institutes of Health grants DK67530 and the Com-
puter-Aided Drug Design Center, University of Maryland,
Baltimore, for computational resources. We thank Professor
Swaan (University of Maryland) for his helpful suggestions
and Kellie Hom (University of Maryland) for NMR assis-
tance.
Supporting Information Available: Complete list of 13C NMR
data, combustion analysis, and computational tables for all
synthesized analogues. This material is available free of charge
via the Internet at http://pubs.acs.org.
References
(20) Feig, M.; Karanicolas, J.; Brooks, C. L., III. MMTSB Tool Set:
enhanced sampling and multiscale modeling methods for applica-
tions in structural biology. J. Mol. Graphics Modell. 2004, 22,
377–395.
(21) Garcia, A. E.; Sanbonmatsu, K. Y. Exploring the energy landscape
of a beta hairpin in explicit solvent. Proteins 2001, 42, 345–354.
(22) Sugita, Y.; Okamoto, Y. Replica-Exchange Molecular Dynamics
Method for Protein Folding. Chem. Phys. Lett. 1999, 314, 141–151.
(23) Allen, M. P.; Tildesley, D. J. Computer Simulation of Liquids;
Clarendon Press: Oxford, 1987; p 385.
(1) Balakrishnan, A.; Polli, J. E. Apical sodium dependent bile acid
transporter (ASBT, SLC10A2): a potential prodrug target. Mol.
Pharmaceutics 2006, 3, 223–230.
(2) Tolle-Sander, S.; Lentz, K. A.; Maeda, D. Y.; Coop, A.; Polli, J. E.
Increased acyclovir oral bioavailability via a bile acid conjugate.
Mol. Pharmaceutics 2004, 1, 40–48.
(3) Alrefai, W. A.; Sarwar, Z.; Tyagi, S.; Saksena, S.; Dudeja, P. K.;
Gill, R. K. Cholesterol modulates human intestinal sodium-depen-
dent bile acid transporter. Am. J. Physiol. Gastrointest. Liver
Physiol. 2005, 288, G978–G985.
(24) Lee, M. S.; Feig, M.; Salsbury, F. R.; Brooks, C. L. New analytic
approximation to the standard molecular volume definition and its
application to generalized born calculations. J. Comput. Chem.
2003, 24, 1348–1356.
(4) Sakamoto, S.; Kusuhara, H.; Miyata, K.; Shimaoka, H.; Kanazu,
T.; Matsuo, Y.; Nomura, K.; Okamura, N.; Hara, S.; Horie, K.;
Baba, T.; Sugiyama, Y. Glucuronidation converting methyl 1-(3,4-
dimethoxyphenyl)-3-(3-ethylvaleryl)-4-hydroxy-6,7,8-trimethoxy-
2-na phthoate (S-8921) to a potent apical sodium-dependent bile
acid transporter inhibitor, resulting in a hypocholesterolemic ac-
tion. J. Pharmacol. Exp. Ther. 2007, 322, 610–618.
(5) Kitayama, K.; Nakai, D.; Kono, K.; van der Hoop, A. G.; Kurata,
H.; de Wit, E. C.; Cohen, L. H.; Inaba, T.; Kohama, T. Novel
nonsystemic inhibitor of ileal apical Naþ-dependent bile acid
transporter reduces serum cholesterol levels in hamsters and mon-
keys. Eur. J. Pharmacol. 2006, 539, 89–98.
(25) Lee, M. S.; Salsbury, F. R.; Brooks, C. L. Novel generalized Born
methods. J. Chem. Phys. 2002, 116, 10606–10614.
(26) Ryckaert, J. P.; Ciccotti, G.; Berendsen, J. C. Numerical integra-
tion of the cartesian equations of motion of a system with con-
straints: molecular dynamics of n-alkanes. J. Comput. Phys. 1977,
23, 327–341.
(27) Hawkins, G. D.; Giesen, D. J.; Lynch, G. C.; Chambers, C. C.;
Rossi, I.; Storer, J. W.; Li, J.; Zhu, T.; Thompson, J. D.; Winget,
P. e. a. AMSOL version 6.8; University of Minnesota: Minneapolis,
MN.
(6) Balakrishnan, A.; Wring, S. A.; Coop, A.; Polli, J. E. Influence of
charge and steric bulk in the C-24 region on the interaction of bile
acids with human apical sodium-dependent bile acid transporter.
Mol. Pharmaceutics 2006, 3, 282–292.
(28) Orozco, M.; Colominas, C.; Luque, F. J. Theoretical determination
of the solvatioon free energy in water and chloroform of the nucleic
acid bases. J. Chem. Phys. 1996, 209, 19–29.
ꢀ
(7) Gonzalez, P. M.; Acharya, C.; MacKerell, A. D., Jr.; Polli, J. E.
(29) Lee, B.; Richards, F. M. The interpretation of protein structures:
estimation of static accessibility. J. Mol. Biol. 1971, 55, 379–400.
(30) Chothia, C. Structural invariants in protein folding. Nature 1975,
254, 304–308.
(31) Auffinger, P.; Hays, F. A.; Westhof, E.; Ho, P. S. Halogen bonds in
biological molecules. Proc. Natl. Acad. Sci. U.S.A. 2004, 101,
16789–16794.
Inhibition Requirements of the Human Apical Sodium-Dependent
Bile Acid Transporter (hASBT) Using Aminopiperidine Conju-
gates of Glutamyl-Bile Acids. Pharm. Res. 2009, 26, 1665–1678.
(8) Bernard, D.; Coop, A.; MacKerell, A. D., Jr. 2D conformationally
sampled pharmacophore: a ligand-based pharmacophore to differ-
entiate delta opioid agonists from antagonists. J. Am. Chem. Soc.
2003, 125, 3101–3107.
(9) Bernard, D.; Coop, A.; MacKerell, A. D., Jr. Quantitative con-
(32) Panigrahi, S. K.; Desiraju, G. R. Strong and weak hydrogen bonds
formationally sampled pharmacophore for delta opioid ligands:
in the protein-ligand interface. Proteins 2007, 67, 128–141.

4760 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 12
Rais et al.
(33) Berski, S.; Ciunik, Z.; Drabent, K.; Latajka, Z.; Panek, J. Domi-
nant role of C-Br---N halogen bond in molecular self-organization.
Crystallographic and quantum-chemical study of Schiff-base-con-
taining triazoles. J. Phys. Chem. B 2004, 108, 12327–12332.
(34) Acharya, C.; Seo, P. R.; Polli, J. E.; MacKerell, A. D., Jr.
Computational model for predicting chemical substituent effects
on passive drug permeability across parallel artificial membranes.
Mol. Pharmaceutics 2008, 5, 818–828.
(35) Hanke, P. R.; Timm, P.; Falk, G.; Kramer, W. Behavior of
different suture materials in the urinary bladder of the rabbit with
special reference to wound healing, epithelization and crystalliza-
tion. Urol. Int. 1994, 52, 26–33.
(38) Ihaka, R.; Gentleman, R. R: A Language for Data Analysis and
Graphics. J. Comput. Graph. Stat. 1996, 5, 299–314.
(39) Schroeder, L. D.; Sjoquist, D. L.; Stephan, P. E. Understanding
Regression Analysis: An Introductory Guide; Sage Publications:
Newbury Park, CA, 1986.
(40) Olobatuyi, M. E. A User’s Guide to Path Analysis; University Press
of America: New York, 2006.
(41) Akaike, H. Likelihood of a Model and Information Criteria.
J. Econometrics 1981, 16, 3–14.
(42) McQuarrie, A. D. R.; Tsai, C. L. Regression and Time Series Model
Selection; World Scientific: Singapore, 1998.
(43) Waterhous, D. V.; Barnes, S.; Muccio, D. D. Nuclear magnetic
resonance spectroscopy of bile acids. Development of two-dimen-
sional NMR methods for the elucidation of proton resonance assign-
ments for five common hydroxylated bile acids, and their parent bile
acid, 5 beta-cholanoic acid. J. Lipid Res. 1985, 26, 1068–1078.
(36) Discovery Studio; Accelrys Inc, http://www.accelrys.com/dstudio.
(37) Reiser, B.; Faraggi, D. Confidence Intervals for the Overlapping
Coefficient: The Normal Equal Variance Case. Statistician 1999,
48, 413–418.