Increasing selectivity of CC chemokine receptor 8 antagonists by engineering nondesolvation related interactions with the intended and off-target binding sites

7706 J. Med. Chem. 2009, 52, 7706–7723

DOI: 10.1021/jm900713y

Increasing Selectivity of CC Chemokine Receptor 8 Antagonists by Engineering Nondesolvation

Related Interactions with the Intended and Off-Target Binding Sites

†

§

Igor Shamovsky,*^,†Chris de Graaf,^#,OL_^isa Alderin, Malena Bengtsson, Hakan Bladh, Lena Borjesson, Stephen Connolly,

€

Hazel J. Dyke, Marco van den Heuvel, Henrik Johansson, Bo-Goran Josefsson, Anna Kristoffersson, Tero Linnanen,

^

†

†,[

€

Annea Lisius, Roope Mannikko, Bo Norden, Steve Price, Lena Ripa, Didier Rognan, Alexander Rosendahl,

†

,z

†

^

†

#

‡

€

ꢀ

Marco Skrinjar,^†and Klaus Urbahns³

^†Department of Medicinal Chemistry and ^‡Department of Biological Sciences, AstraZeneca R&D Lund, S-22187 Lund, Sweden, ^§Department

of Medicinal Chemistry, AstraZeneca R&D Charnwood, Loughborough, Leicestershire LE11 5RH, U.K., Department of Safety Pharmacology,

Safety Assessment U.K., AstraZeneca R&D Alderley Park, Macclesfield SK10 4TG, U.K., ^{^}Argenta Discovery Ltd, 8/9 Spire Green Centre, Flex

Meadow, Harlow, Essex CM19 5TR, U.K., ^#Structural Chemogenomics Group, Laboratory of Therapeutic Innovation, UMR 7200 CNRS-UdS

(Universiteꢀ de Strasbourg), 74 Route du Rhin, B.P. 24, F-67400 Illkirch, France, and ³Global Respiratory and Inflammation Research, Astra-

Zeneca R&D Charnwood, Loughborough, Leicestershire LE11 5RH, U.K. ^OCurrent address: Division of Medicinal Chemistry, Faculty of

Sciences, Leiden/Amsterdam Center for Drug Research, VU University Amsterdam, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands.

^[Current address: FeF Chemicals A/S, Københavnsvej 216, 4600 Køge, Denmark. ^zCurrent address: Department of Physiology, Anatomy and

Genetics, Sherrington Building, University of Oxford, Parks Road, Oxford OX1 3PT, UK.

Received May 26, 2009

The metabolic stability and selectivity of a series of CCR8 antagonists against binding to the hERG ion

channel and cytochrome Cyp2D6 are studied by principal component analysis. It is demonstrated that

an efficient way of increasing metabolic stability and selectivity of this series is to decrease compound

lipophilicity by engineering nondesolvation related attractive interactions with CCR8, as rationalized

by three-dimensional receptor models. Although such polar interactions led to increased compound

selectivity, such a strategy could also jeopardize the DMPK profile of compounds. However, once

increased potency is found, the lipophilicity can be readjusted by engineering hydrophobic substituents

that fit to CCR8 but do not fit to hERG. Several such lipophilic fragments are identified by two-

dimensional fragment-based QSAR analysis. Electrophysiological measurements and site-directed

mutagenesis studies indicated that the repulsive interactions of these fragments with hERG are caused

by steric hindrances with residue F656.

Introduction

A wealth of structure-activity relationships and mutagen-

esis studies on the class A GPCR antagonists suggests that

there is a generic binding pocket for small molecules inside the

heptahelical bundle common to this subfamily. With the release

of crystal structures of the bovine rhodopsin,²it is apparent

that this generic transmembrane (TM) binding pocket for small

molecules is reminiscent of the retinal-binding pocket in rho-

dopsin.^6,7The predicted high value of homology modeling of

GPCRs in the identification of the class A GPCR ligands^6,7has

been verified in several classes of therapeutics,⁸although

intrinsic conformational plasticity inherent in GPCR structure

and ligand binding sites has to be taken into consideration.⁹

The pro-inflammatory chemokine receptor 8 (CCR8) is

selectively expressed on a subset of T-helper-2 (Th2) cells,

eosinophils, and on monocyte derived dendritic cells, and is

upregulated upon activation.^10,11These effector cells migrate

to the site of inflammation by sensing a gradient of the

chemokine CCL1 by the cell surface expressed CCR8 recep-

tor. Once inside the inflamed tissue, the CCR8 positive cells

arrest and perform their effector function. Among other

functions, the activated Th2 cells produce the key cytokines

IL-4, IL-5, andIL-13, which are the predominantmediatorsin

inflammation and hyperresponsiveness in bronchial asthma

and atopic dermatitis. Clinical studies have demonstrated

significantly elevated numbers of CCR8 positive cells in

asthmatics and atopic dermatitis patients.^12-14The pivotal

role of CCR8 in driving allergic diseases like asthma was

Chemokines constitute a family of small secreted peptides,

which control the trafficking cascades of leukocytes to in-

flammatory sites. They act through binding to specific high-

affinity receptors, the chemokine receptors, which belong to a

superfamily of G-protein-coupled receptors (GPCRs^a). The

chemokine receptor family is the largest subfamily of known

peptide-binding GPCRs.¹They belong to the class A GPCRs,

which also include rhodopsin, β₁-adrenergic, β₂-adrenergic,

and A_2Aadenosine receptors, four GPCRs for which high-

resolution X-ray crystal structures have been solved.^2-5The

chemokine system regulates the development, activation, and

recruitment of leukocytes and plays important roles outside

the immune system.¹

*To whom correspondence should be addressed. Phone: þ4646-

338347. Fax: þ4646-338399. E-mail: igor.shamovsky@astrazeneca.

com.

^aAbbreviations: CCR8, CC chemokine receptor 8; DMPK, drug

ꢁ

metabolism and pharmacokinetics; hERG, human ether-a-go-go-re-

lated gene; GPCR, G-protein-coupled receptor; LLE, ligand lipophili-

city efficiency; PCA, principal component analysis; H-bond, hydrogen-

bond; TM, transmembrane; TM7, transmembrane R-helix number 7;

Cyp2D6, the 2D6 isoform of cytochrome P450; DSM, discrete structural

modifications; LMP2, local Møller-Plesset second-order perturbation

theory calculation; cc-pVTZ(-f), the Dunning correlation consistent

polarized valence triple-ζ basis set without f-functions; IFP, interaction

fingerprint bit-string; PLS-DA, partial least-squares discriminant ana-

lysis.

r

pubs.acs.org/jmc

Published on Web 07/17/2009

2009 American Chemical Society

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7707

these CCR8 antagonists 1-30 are detailed in Table 1 and Table

S1 of Supporting Information. The molecular basis of repulsive

interactions with the hERG ion channel is revealed by electro-

physiological experiments conducted with a focused set of 25 of

these compounds and eight hERG mutants. It is shown that

some lipophilic fragments built in to these CCR8 antagonists,

which are responsible for high intrinsic anti-hERG pro-

pensities, do not fit to the narrow part of the channel between

residues F656.

Figure 1. Core structure of CCR8 antagonists. R1 is normally a

lipophilic group; R2 is normally a polar group.

demonstrated in CCR8 deficient mice, in which it was shown

that CCR8 knock-down significantly ameliorated lung in-

flammation and airway function in several allergic airway

models.^15,16These clinical and experimental results strongly

suggest that CCR8 antagonists could be of therapeutic value

in allergic pulmonary diseases.^17,18

Methods

Synthesis. The compounds were prepared according to

Route A (Scheme 1) by reductive amination of tert-butyl 3,

9-diazaspiro[5.5]undecane-3-carboxylate hydrochloride 31

followed by amide coupling with a suitable acid after depro-

tection. Alternatively, compounds were prepared according

to route B (Scheme 2) by amide coupling of compound 31,

followed by deprotection and subsequent reductive amina-

tion to afford the desired compound. Compound 30 was

prepared by alkylation of 31 with 5-bromo-7-(bromo-

methyl)-3,3-difluoro-2,2-dimethyl-2,3-dihydro-1-benzofur-

an 38 according to Scheme 3. The intermediate 38 was

synthesized via alkylation of ortho-cresol 32, followed by

intramolecular Friedel-Crafts acylation to afford the ben-

zofuranone 35. The ketone was transformed to the corre-

sponding 1,3-dithiolane and the fluorine atoms were

introduced through geminal difluorination. Under these

reaction conditions, a bromine atom was also introduced

in the aryl ring to form 37. Following a radical bromination

of the benzylic position and coupling with compound 31 to

afford 39, the bromine atom was removed by hydrogenation

to give compound 40. Deprotection and coupling with 5-

aminopyridine-2-carboxylic acid gave the desired compound

30. Synthetic details and characterization of the compounds

can be found in ref 22 and in the Supporting Information.

Selectivity of CCR8 Antagonists through hERG Baseline

Lipophilicity Relationships. Baseline lipophilicity relationships

describe the link between compound lipophilicity (log D) and

potency of binding (pIC₅₀) (eq 1).^37-40,45,46Upon binding to the

target or an off-target binding site, e.g., hERG, a hydrophobic

drug molecule and the hydrophobic binding sites go through

energetically favorable desolvation processes. Accordingly, the

lipophilicity-driven term similarly adds to the overall binding

potencies, both desired and undesired, as is expressed in eqs 1 and 2.

Because binding sites in proteins normally represent hydropho-

bic clefts with few polar residues,^47,48the hydrophobicity factors

of binding sites rarely significantly deviate from 1,⁴⁵although

such deviations have been noted.³⁷Correspondingly, the selecti-

vity is virtually insensitive to compound lipophilicity per se but is

determined by an interplay of nondesolvation related specific

interactions with target and antitarget, such as H-bonding, π-π

interactions and geometric fit.^41,49This point is clearly seen in a

difference of the eqs 1 and 2.

The vast majority of nonpeptide chemokine receptor antago-

nists share a common pharmacophore, which includes lipophi-

lic groups and a centrally located protonated amine^18,19that

forms an ionic bond with the conserved glutamate E^7.39of

chemokine receptors.^20,21Different classes of known small-

molecule CCR8 antagonists also reveal this general pharmaco-

phore.^17,18,21-24This particular pharmacophore is also inherent

in hERG ion channel blocking agents^25-27and in inhibitors of

the 2D6 isoform of cytochrome P450 (Cyp2D6),²⁸one of the

most important enzymes involved in drug metabolism. Indeed,

much of the preclinical attrition of chemokine receptor anta-

gonists has been due to the challenges of increasing selectivity

against hERG and Cyp2D6, and metabolic liability.^29-34

In this communication, we consider a series of 464

CCR8 antagonists, which are based on the 3,9-diazaspiro-

[5.5]undecanescaffold(Figure1).²²Apart from general binding

of these compounds to the hERG ion channel, several com-

pounds of this series also exhibit binding to Cyp2D6 and have

poor metabolic stability. The overall promiscuity of bioactive

compounds is thought to be related to their lipophilicity,

typically quantified by Log P or Log D.³⁵A main focus of this

study is to investigate the connection of the intended and off-

target binding of the CCR8 antagonists with their lipophilicity.

We demonstrate that engineering attractive polar interactions

in this series of CCR8 antagonists with the primary binding site

results in an increase in potency and a simultaneous decrease in

lipophilicity, which in turn diminishes lipophilicity-linked pro-

miscuous interactions and also increases metabolic stability.

Both location and conformational flexibility of key polar

residues lining the consensus TM binding pocket³⁶within the

CCR8 homology model are shown to be critical to explain

the structure-activity relationships and to guide the inten-

ded decrease of compound lipophilicity. However, although

the reduction of the desolvation related component of po-

tency^37-41is a reliable general approach to increase compound

selectivity,^35,42,43this strategy can have certain limitations

caused by poor permeability and bioavailability of polar

compounds. In addition, contrary to the general strategy of

reducing lipophilicity, it has also been demonstrated that in

specially identified locations polar functionality may form

detrimental intrinsic attractive interactions within the hERG

ion channel cavity^26,42-44and therefore decrease selectivity

against hERG. In a previous communication,⁴⁴we demon-

strated that molecular fragments, which are responsible for

both beneficial and detrimental nondesolvation related inter-

actions⁴⁰with an antitarget, can be identified by two-dimen-

sional fragment-based QSAR analysis. We identified structural

features of chemokine receptor antagonists that caused intrinsic

attractive interactions with hERG leading to the loss of hERG

selectivity, whereas here we explore possibilities for anti-hERG

propensity of spirocyclic CCR8 antagonists. The structures of

CCR8

pIC₅₀

¼ a^CCR8log D -kΔG_intr

¼ a^hERGlog D -kΔG_intr

þconst

ð1Þ

ð2Þ

hERG

ꢀ

ꢁ

hERG

pIC₅₀^CCR8-pIC₅₀

¼

pIC₅₀^CCR8-a^hERGlog D

hERG

þkΔG_intr

ꢀ

þconst

ð3Þ

ꢁ

kΔG_intr

¼ - pIC₅₀^hERG-a^hERGlog D þconst ð4Þ

hERG

7708 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Table 1. Structures of the Focused Set of Compounds^a

Shamovsky et al.

^aThe common scaffold of the compounds is given in Figure 1. LogD denotes measured compound lipophilicity, where D is the apparent n-octanol/

water partition coefficient, CCR8 is potency at CCR8 (pIC₅₀^CCR8).

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7709

Scheme 1. Route A^a

^aReagents and conditions: (a) R1-CHO, NaBH(OAc)₃, AcOH, DCM; (b) HCl, MeOH; (c) R2-COOH, HATU, NEt₃, DCM.

Scheme 2. Route B^a

^aReagents and conditions: (a) R2-COOH, HATU, NEt₃, DCM; (b) HCl, MeOH; (c) R1-CHO, NaBH(OAc)₃, AcOH, DCM.

Scheme 3^a

^aReagents and conditions: (a) (i) ethyl-2-bromoisobutyrate, potassium carbonate, DMF, (ii) NaOH, H₂O; (b) SOCl₂; (c) AlCl₃, toluene;

(d) ethanedithiol, BF₃OEt₂, CHCl₃; (e) 1,3-dibromo-5,5-dimethylhydantoin, 70% HF/pyridine complex, DCM; (f) N-bromosuccinimide, benzoylper-

3

oxide, CCl₄; (g) 31, DIPEA, THF; (h) 10% Pd/C, H₂, MeOH; (i) TFA, DCM, H₂O; (j) 5-aminopyridine-2-carboxylic acid, PS-carbodiimide, HOBt,

DCM.

where pIC₅₀is potency of a compound, D is the apparent

n-octanol/water partition coefficient, a^CCR8and a^hERGare

hydrophobicity factors of the corresponding binding sites, k is

coefficient, equal to (2.303RT)^-1, and ΔG_intris the intrinsic

binding energy, which is not related to desolvation and describes

direct interactions of the ligand with the binding site, including

polar interactions and geometric fit.⁴⁰

Expression 3 gives the compound selectivity against

hERG. The two terms in the right side of eq 3 signify two

approaches to increasing compound selectivity in a structur-

al class. The first term increases when chemical alterations

result in forming new direct attractive interactions with

CCR8,^41,49whereas the increase of the second term is an

indication that new molecular fragments do not fit to hERG.

Analysis of eq 3 is useful in lead identification and lead

optimization projects only when the target and off-target

proteins are not related, in which case it is possible to

separate changes in the two terms. When the hydrophobicity

factor of the binding site of an antitarget is close to 1, the first

term of eq 3 is roughly equal to a previously described value

of ligand lipophilicity efficiency, LLE.³⁵The first term is not

associated with any particular antitarget, correspondingly

7710 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

the approach of increasing this term deals with general

potential of compounds for multiple promiscuous interac-

tions, in agreement with previous observations.³⁵

CCR8.²¹The H-bond acceptor and donor interaction fin-

gerprint bits of Q^2.60, Q^45.49, N^5.39, and N^5.43were adapted to

the side-chain amide rotamer of the receptor model, depend-

ing on whether the amide oxygen (H-bond acceptor) or

nitrogen (H-bond donor) atom was facing the binding

pocket, yielding 16 different IFPs corresponding to the 16

different receptor models. The reason for this was to account

for protein flexibility and the existence of both H-bond

acceptor and H-bond donor functional groups in the ligand

(Table 1 and Table S1 of Supporting Information). For each

of the diaza-spiro-alkanes, docking poses with the highest

IFP similarity score and forming an H-bond-assisted salt

bridge with E^7.39were selected as representative binding

modes.

Site Directed Mutagenesis Experiments of the hERG Ion

Channel. Generation of hERG Mutant/Wild-Type hERG

Expressing Cell Lines. The hERG gene was cloned into the

pTight expression vector (Clonetech Laboratories, Inc.,

Mountain View, CA) to keep expression of the channel

under tetracycline control. The mutations T623S, S624A,

S624T, Y652A, Y652F, F656M, F656T, and F656W were

introduced into the gene by standard polymerase chain

reaction (PCR)-based site-directed mutagenesis techniques

(QuickChange, Stratagene, La Jolla CA) and confirmed by

sequencing both strands of the entire gene. Linearized hERG

DNA was transfected into Tet-On CHO K1 cells together

with a linear hygromycin selection marker using the lipofec-

tamine method (Invitrogen). Cells were then incubated in

presence of 0.6 mg/mL hygromycin until cells in untrans-

fected control flask had died (2-3 weeks). Cells were then

incubated in the presence of 0.8 μg/mL doxycyclin for 24-

48 h to induce channel expression, which was confirmed

using the IonWorks HT high-throughput electrophysiology

measurement platform (Molecular Devices, Sunnyvale

CA).⁵⁷Clonal cell lines were then created by dilution cloning

and selected by measuring functional expression, using Ion-

Works HT.

Pharmacological Comparison. An automated, plate-based

electrophysiology device (IonWorks HT)⁵⁷was used to study

the pharmacological profile of wild-type hERG (wt-hERG)

channel and each hERG mutant by directly assessing the

channel function.⁵⁸For each experimental run of IonWorks

HT, the device made perforated whole-cell recordings at

21 ꢀC, usually from more than 200 of the 384 wells in a 384-

well substrate (Molecular Devices). The extracellular solu-

tion was Dulbecco’s phosphate-buffered saline containing

calcium (0.9 mM) and magnesium (0.5 mM) (PBS; In-

vitrogen) and the “pipette” solution was (in mM): KCl

140, EGTA 1, MgCl₂1 and HEPES 20 (pH 7.25-7.30 using

10 M KOH) plus 100 μg/mL amphotericin B (Sigma-

Aldrich). After attainment of the whole-cell configuration,

a precompound hERG current was evoked in each cell in the

presence of PBS by the following voltage protocol: a 20 s

period holding at -70 mV, a 160 ms step to -60 mV (to

obtain an estimate of leak), a 100 ms step back to -70 mV, a

1 s step to þ40 mV, a 2 s step to -30 mV, and finally a 500 ms

step to -70 mV. The amplitude of the hERG current was

measured by subtracting the baseline current at -70 mV

from peak current measured during the step to -30 mV. Test

compounds, vehicle, or 31.6 μM terfenadine controls were

then added to each well, and after 3 min, the voltage pulse

was reapplied to generate a postcompound hERG current.

Terfenadine at 31.6 μM fully blocked currents in wt-hERG

and all hERG mutants in our setup. In between the pre- and

It has been shown that the hydrophobicity factor of the

hERG ion channel cavity, a^hERG, is indeed close to 1.⁴⁴

Correspondingly, according to eq 3, a normally expected

average gain in hERG selectivity of compounds in chemical

series is expected to be equal to the increase in the value

of LLE. When the hydrophobicity factors of both target

and antitarget binding sites are roughly equal, as in the

present study, eq 3 is virtually free of desolvation effects

and gives an opportunity to analyze the selectivity as

an interplay of intrinsic interactions with the alternative

binding sites. As opposed to the general character of the first

term of eq 3, the second term, which is detailed in eq 4, is

linked to a particular antitarget and represents an additional

opportunity to improve selectivity in a chemical series by

increasing the specific anti-antitarget propensity of com-

pounds by engineering direct repulsive interactions with the

binding site of this antitarget. Both terms of eq 3 are analyzed

in the present study.

Molecular Modeling of CCR8 Receptor-Ligand Interac-

tions. Construction of Three-Dimensional CCR8 Receptor

Models. A ground-state homology model of the CCR8

receptor was constructed following a previously defined

protocol.⁵⁰A preliminary high-throughput receptor model,

including only the seven TM helices, was derived from a

validated CCR5 model⁵¹using the GPCRgen program.⁵²

The amino acid sequence alignments used for constructing

the receptor models are shown in Figure S1 of Supporting

Information. Notice that the Ballesteros-Weinstein residue

numbering scheme⁵³and a recently proposed numbering

scheme of the extracellular loop 2⁵⁰are used throughout

the manuscript. Compound 13 (see Table 1) was docked into

the receptor model using the “3 times speed-up” settings

of the Gold v3.3 program, guided by experimentally driven

H-bond constraints between the protonable tertiary amine of

the diaza-spiro scaffold (Figure 1) and both E^7.39carboxylate

oxygen atoms. The active site center determined by the PASS

program⁵⁴was taken as the starting position of the GOLD

flood fill algorithm. This preliminary model of CCR8 with

bound 13 was used to construct the extracellular and intra-

cellular loops and for further refinement of the receptor

model by in vacuo energy minimization and molecular

dynamics simulation in a solvated membrane layer as re-

cently described.⁵⁵The χ₂torsion angles of N^5.39and N^5.43

and χ₃torsion angles of Q^2.60and Q^45.49of the final opti-

mized CCR8 model were independently rotated by 180ꢀ,

yielding 16 (= 2⁴) unique receptor models.

Automated Docking of Diaza-spiro-alkane Antagonists.

The 3,9-diazaspiro[5.5]undecane antagonists presented in

Table 1 were automatically docked in the 16 CCR8 homol-

ogy models as described in the preceding section. The bind-

ing pose of compound 13 was used to generate an interaction

fingerprint (IFP) bit-string as previously described.⁵⁶Seven

different interaction fingerprints, namely (1) hydrophobic

contact, (2) aromatic face-to-face, (3) aromatic edge-to-face,

(4) donor-acceptor and (5) acceptor-donor H-bond inter-

actions, (6) positive-negative, and (7) negative-positive

ionic interactions, were used to define the IFP bit-string.

The cavity used for the IFP analysis consisted of the follow-

ing 13 residues: Y^1.39, Q^2.60, S^3.29, Y^3.32, Y^3.33, Q^45.49, Y^45.51

^5.39, M^5.42, N^5.43, F^6.51, L^6.55, and E^7.39. Eight residues of

this set have been shown to be involved in ligand binding to

,

N

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7711

postcompound voltage pulses, there was no clamping of the

membrane potential.

Compounds to be tested were first dissolved in DMSO at a

concentration 300-fold the top test concentration and then

diluted to the final test concentrations. IonWorks HT is set

up such that each well of a 96-well plate containing the test

compounds maps to four wells of the 384-well substrate in

which recordings are made. The 96-well compound plate was

made up of two identical halves. Each half was made up of

PBS containing eight half log-spaced concentrations of a

single test compound at 3-fold its final concentration, and

this was repeated five times. For the remaining eight wells of

the half plate, four contained 1% DMSO and four contained

95 μM terfenadine. Each well of a 384-well substrate initially

contained 3 μL of PBS and 3 μL of PBS containing a cell

suspension of a single cell line at a concentration of 250000

cells/mL. When 3 μL of test compound from the compound

plate was added to each well of a 384-well substrate, this

meant that the cells were exposed to the final compound test

concentration, 0.33% DMSO or 31.6 μM terfenadine.

For each IonWorks HT run, the hERG tail current

amplitude in each cell in the presence of PBS was compared

with that in the presence of test compound for the same cell.

All the data were then scaled by defining the effect of 0.33%

DMSO as 0% inhibition and the effect of the supramaximal

blocking concentration of terfenadine as 100% inhibi-

tion. Because the contents of each well of the test plate

were added to four wells of the 384-well substrate, there

could be percentage inhibition data from between 0 and 4

cells for each test concentration. Because only one com-

pound was tested in a 384-well substrate containing wt-

hERG and a hERG mutant cell line, each point of the

noncumulative curve came from up to 20 cells. The pharma-

cological data were fitted using a four parameter nonlinear

regression (curve top, curve slope, and IC₅₀value, with the

curve bottom being normalized to 0) with in-house Origin

package.

Figure 2. The main loadings plot of the PCA of properties of 464

in-house CCR8 antagonists of the common scaffold indicated in

Figure 1. Each compound was described by 17 properties. PC-1 and

PC-2 represent the first and the second principal components of the

covariance matrix. Projections of coordinate axes of intended

properties are shown in green, those of unwanted properties in

red. Labels CCR8, hERG(b), hERG(el), and 2D6 designate poten-

cies in logarithmic units at CCR8, at hERG in binding experiments,

at hERG in electrophysiological experiments, and at Cyp2D6 in

binding assay, respectively. Labels LogD and AZLogD are mea-

sured and predicted compound lipophilicities at pH = 7.4, respec-

tively. Labels Mics(h), Mics(r), and Heps(r) are rates of metabolic

clearance in logarithmic units in human miscosomes, rat micro-

somes and rat hepatocytes, respectively. Labels CCR8-hERG(b),

CCR8-hERG(el), CCR8-2D6, CCR8-Mics(h), and CCR8-Mics(r)

are the corresponding selectivities. Labels LLE, hERG(b)-LogD,

and hERG(el)-LogD are nondesolvation related components of

potencies at CCR8, hERG in binding experiments, and hERG in

electrophysiological experiments, respectively. Locations of major-

ity of promiscuous, inactive, polar, and selective compounds are

indicated. The arrows designate structural alterations within the

compound class that usually result in lipophilicity-linked promis-

cuous binding (quadrant IV), specific binding to the primary target

(quadrant III), or specific binding to the hERG ion channel

(quadrant II). The first two components in the PCA plot explain

70% of the variance.

(pIC₅₀^CCR8) and selectivities against undesirable binding,

i.e., differences between the primary potency and unwanted

properties such as off-target potencies or rates of metabolic

clearance expressed in logarithmic units. Statistically, com-

pounds that fall in this parameter space exhibit both high

primary potency and high selectivities. The opposite para-

meter space, which is located in the first quadrant, refers to

relatively inactive compounds.

The subset of descriptors of unwanted properties (shown

in red) includes potency of hERG binding and potency of

inhibition of hERG in the electrophysiological assay, po-

tency of inhibition of Cyp2D6, and rates of microsomal and

hepatic metabolic clearance. In addition, measured and

calculated compound lipophilicity at pH=7.4 (LogD and

AZLogD, respectively), ligand lipophilicity efficiency

(LLE), and intrinsic hERG-binding propensity (potency at

hERG minus Log D)⁴⁴have been added to the set of proper-

ties for PCA. As is seen in Figure 2, the majority of unwanted

descriptors, including binding to and inhibition of the hERG

ion channel, inhibition of Cyp2D6, and metabolic instability

in microsomes and hepatocytes, are grouped together in

the fourth quadrant, which indicates that this particular

space represents the area of promiscuity. In addition to

the unwanted properties, this area also includes descrip-

tors of compound lipophilicity, LogD and AZLogD, which

suggests that the fourth quadrant of the plot is the common

place of the most lipophilic compounds in the set.

Results and Discussion

Multivariate Analysis of Properties of CCR8 Antagonists.

An analysis of CCR8 antagonists of the general structure

depicted in Figure 1 indicated that the following unwanted

properties are inherent in many compounds of this class:

inhibition of the hERG potassium ion channel, inhibition of

Cyp2D6, and high microsomal and hepatic metabolic clear-

ance. Therefore, the lead optimization strategy in this class

was predominantly focused on increasing compound selec-

tivity toward CCR8 and against these properties. Principal

component analysis (PCA) of relevant properties of a series

of 464 compounds of the structural class under study was

performed by the program SIMCA-Pþ (version 11; Ume-

trics, Kinnelon, NJ) to define the most efficient way to

separate the intended potency from the unwanted properties.

Figure 2 illustrates the plot of loadings of PCA, which

represents the projections of the initial coordinate system

of scaled compound properties into the plane of two first

eigenvectors (PC-1 and PC-2) of the covariance matrix. The

center of the coordinate system corresponds to the average

values of properties. This plot illustrates major correlations

that exist between properties in the data set. As is seen, the

subset of descriptors of desired properties (shown in green) is

grouped together and roughly occupies the third quadrant of

the plot. This subset includes the primary potency at CCR8

7712 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

This implies that lipophilicity of the CCR8 antagonists

correlates with their binding to multiple antitargets such

as hERG and metabolic enzymes because of high contribu-

tion of nonspecific hydrophobic interactions to the off-

target binding affinities, which is in line with previous

observations.^35,37,59,60

The major correlations between descriptors in the series of

CCR8 antagonists revealed by the PCA work in a statistical

sense, which leaves many outliers, both good and bad, out of

the picture. At the same time, it is in the outliers, where

potential value might be found. The next section deals with

special structural features of CCR8 antagonists that make

them particularly effective in increasing hERG selectivity.

Detection of Nondesolvation Related Interactions with the

CCR8 and the hERG Binding Sites. A subset of 30 spirocyclic

CCR8 antagonists with diverse R1 and R2 fragments were

chosen to form a focused set for detailed investigation. All

types of fragments that maintained the primary potency of

compounds below 2 μM were included in the set. Molecular

structures and experimental data obtained for this set are

presented in Tables 1, 2, and Table S1 of the Supporting

Information. The two-dimensional fragment-based QSAR

analysis utilizing eq 3 was carried out to detect intrinsic

interactions with CCR8 and hERG in this data set. As

suggested in our previous communication,⁴⁴average con-

tributions of different molecular fragments R1 and R2 to the

LLE and to the hERG selectivity were calculated by solving

the Free-Wilson system of equations⁶⁵using an in-house

AstraZeneca plug-in of the program Spotfire DecisionSite

Descriptors of two unwanted properties, namely intrinsic

potency for binding to and inhibition of the hERG channel,

are located in the PCA plot far beyond the area of promis-

cuity, in the second quadrant, which is the common place for

relatively polar compounds. The results suggest that an

increase of polarity of compounds moves them away from

the area of promiscuity toward the second quadrant in the

plot. Although this allows one to get rid of desolvation-

linked promiscuous off-target binding, it does not increase

the required selectivities because the desolvation effects

similarly decrease binding affinity at the primary target

(see eqs 1 and 2). On the other hand, the intended selectivities

are colocated in the PCA plot with LLE, which suggests that

increasing polarity while maintaining the required level of

the primary potency is the most efficient way of increasing

selectivity against promiscuous interactions, which is in

agreement with current opinion.^35,41,49The key to this path

is to find new direct attractive interactions with the primary

target in the form of H-bonding, π-πinteractions, or improved

geometric fit.^41,44,49This path, shown in Figure 2 by the green

arrow, results most of the time in reducing compound lipo-

philicity, which has clear limitations. It is well recognized

that compound lipophilicity has to be within a certain interval

to maintain the required DMPK profile of drug candidates,

including bioavailability and permeability. In addition, an

increase of the number of polar functions in compounds,

designed to lead to more potency at the intended target, may

also lead to polar interactions with binding sites of antitargets

and thereby not result in increased selectivity.^26,39,42-44Speci-

fically, it has been noted that general pharmacophores of

chemokine receptor antagonists are similar to that of the hERG

ion channel blockers such that accumulation of solubili-

zing groups on one side of the molecule often leads to the loss

of hERG selectivity because of polar interactions within the

hERG channel vestibule cavity.^42,43A distinctive sign of simi-

larity of pharmacophores for direct attractive interactions in

CCR8 and hERG within the considered compound class is the

fact that the descriptors of such interactions (denoted as LLE

and hERG-LogD, respectively) are colocated in the second

quadrant of the PCA plot in Figure 2.

Furthermore, introduction of direct attractive interactions

of polar functional groups of compounds to increase

selectivity may also bring unexpected problems with intrinsic

metabolic clearance or inhibition of metabolic pathways

by interacting positively with the active sites of metabolic

enzymes.^37-39,61-63In this case, it is valuable to be able to

increase compound selectivity against a specific antitarget

without lowering lipophilicity. This can be done by disrupt-

ing existing polar interactions or by introducing steric in-

compatibility with the antitarget, which increases the anti-

antitarget propensity (eq 4). This strategy, which addresses

the issues linked to one particular antitarget, is widely used in

lead optimization. The method of discrete structural mod-

ifications (DSM) of reducing hERG liability⁴³and a com-

mon practice of removing metabolically vulnerable spots in

compounds to increase metabolic stability⁶⁴are just two

examples of nondesolvation related increasing of intrinsic

anti-antitarget propensities.

€

(version 9.1; TIBCO Spotfire, Goteborg, Sweden). The

central assumption of the Free-Wilson approach, indepen-

dent contributions of R1 and R2 to compound properties,

is unequivocally accepted in this study. The objective was

to find direct attractive interactions with the target binding

site and direct repulsive interactions with hERG as required

by eq 3.

Figure 3 gives the results of the two-dimensional QSAR

analysis. Relative contributions of each molecular fragment

of the focused set to attractive nondesolvation related inter-

actions of compounds with the binding sites of CCR8 and

hERG

hERG are quantified by LLE and (pIC₅₀

- log D),

respectively. Structures of the fragments R1 and R2 are

presented in Figure 4. Consistent with eq 3 and 4, this plot

visualizes an interplay of direct interactions with the binding

sites of the target and the antitarget, with desolvation-driven

components of the corresponding potencies being taken out

of the picture. The diagonal line, called the average hERG

lipophilicity baseline, signifies the average fit of fragments in

the data set to the hERG channel cavity. Consistent with eq

3, the line illustrates how a fragment, either R1 or R2, with an

average intrinsic hERG-binding propensity, contributes to

the hERG selectivity of compounds depending on its intrin-

sic fit to the primary target. Fragments that exhibit unusually

high hERG-binding propensities are located below the line,

whereas those fragments that make unusual repulsive inter-

actions with the hERG cavity are above the line. The vertical

line is the average CCR8 lipophilicity baseline or the average

LLE line, which splits all fragments into two groups such

that the fragments located to the left and to the right from

this line form less and more efficient direct interactions with

the CCR8 binding site than an average fragment, respec-

tively. The resulting nonorthogonal coordinate system di-

vides the plot into four quadrants, with the fragments located

in the first two quadrants being the main focus of the present

investigation.

The fragments are classified in Figure 4. Those that appear

in the first quadrant have the highest potential for selective

interactions with the primary target not only because they

fit to the primary target and do not fit to the hERG channel

but also because they are less prone to desolvation-linked

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7713

Table 2. Potencies at hERG and hERG Mutants of the Focused Set of Compounds^a

No.

hERG el-phys

hERG binding

T623S

S624A

S624T

Y652A

Y652F

F656M

F656T

F656W

1

6.22

7.08

7.06

5.31

5.72

5.63

5.29

7.34

6.66

7.15

6.35

5.71

5.72

4.59

6.64

4.85

4.95

4.71

4.68

7.09

8.19

8.6

0.60

0.91

0.81

0.22

0.08

0.06

0.77

0.93

0.42

0.51

0.47

0.27

0.11

0.10

0.48

0.42

0.53

0.71

0.70

1.04

0.74

1.00

-0.73

-0.14

-0.20

-0.42

-0.64

-1.07

-0.32

0.42

0.78

0.11

0.21

0.39

0.23

0.48

0.46

1.32

0.81

1.65

1.82

2.26

1.42

0.39

2

-0.08

-0.30

-0.42

-0.87

-0.60

-0.54

3

0.08

0.12

4

4.72

4.82

5.96

4.77

7.52

6.68

6.92

6.44

5.85

5

-0.10

0.20

6

7

-0.08

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

0.94

0.58

1.20

1.79

-0.16

0.18

1.49

-0.03

0.44

0.39

1.16

0.51

0.39

0.24

0.56

0.67

0.29

0.72

0.54

0.74

1.09

0.61

-0.25

-0.30

-0.28

0.53

0.67

0.98

1.61

0.79

1.19

0.82

0.83

1.63

-0.01

-0.29

-0.35

-0.48

0.35

-0.21

-0.20

0.21

0.22

0.58

0.87

1.23

1.92

0.46

-0.09

1.41

-0.57

-0.15

-0.39

-0.16

-0.18

-0.14

-0.35

-0.58

5.07

5.7

0.24

0.22

-0.39

-0.28

-0.12

0.11

0.47

-0.30

-0.66

1.04

4.49

5.09

5.11

-0.21

4.42

4.36

4.78

5.38

4.54

4.78

5.15

4.95

5.46

4.51

0.20

0.22

-0.06

0.41

0.01

0.49

0.48

0.82

0.59

1.32

0.55

0.79

-0.62

-0.05

-0.33

0.43

0.93

0.83

1.22

1.81

0.89

1.23

0.04

0.35

0.37

-0.31

0.62

1.32

0.51

1.37

0.77

1.32

0.20

-0.75

-0.99

-0.39

-0.32

-0.61

-0.66

4.43

6.12

4.49

4.77

4.89

4.52

4.51

5.4

-0.19

0.64

-0.26

0.14

-0.58

0.31

-0.37

0.68

-0.22

0.11

0.42

0.01

0.70

0.90

0.36

0.00

0.17

0.39

0.70

0.87

-0.05

0.12

0.02

-0.73

-0.54

-0.49

0.05

0.10

-0.58

-0.46

-1.07

4.32

-0.62

-0.17

^aAbbreviations hERG/el-phys and hERG/binding denote potencies (pIC₅₀) of a compound at hERG in electrophysiological and binding

experiments, respectively. Data presented in columns T623S, S624A, S624T, Y652A, Y652F, F656M, F656T, and F656W characterize differences

between potencies of the compounds at wild-type hERG and at the corresponding hERG mutant, (pIC₅₀^wt-hERG- pIC₅₀^mutant), in electrophysiological

experiments.

interactions with other off-target promiscuous binding

sites, including metabolic enzymes and plasma proteins.

Quadrants III and IV contain fragments that exhibit direct

attractive interactions with the hERG ion channel and there-

by are responsible for the unusually low hERG selectivity

of the corresponding compounds. Fragments of the quad-

rants II and III reveal poor fit to the primary binding site and,

thereby, the entire molecules have to have higher desolvation

contribution to the intended potency to compensate the

insufficient intrinsic complementarity with the primary binding

site, which predisposes them to unwanted promiscuous inter-

actions with target-unrelated binding sites.

The results of the QSAR analysis presented in Figures 3

and 4 reveal fragments that form selective interactions with

CCR8. As is seen, fragments R2 at the right-hand side of the

molecule (Figure 1) tend to move from the left to the right

the LLE and the hERG selectivity, according to eq 3. Structures of the

Figure 3. Results of fragment-based QSAR analysis of intrinsic

interactions in the focused set of CCR8 antagonists with CCR8 and

hERG obtained by Free-Wilson analysis performed in two directions,

side of the plot in Figure 3 as soon as the number of polar

heteroatoms increases, which indicates that the subsite of the

CCR8 binding site they bind to is polar. They have to possess

two polar functions that are able to form H-bonds with the

CCR8 binding site to demonstrate a better than average

contribution to LLE. Thus, 4-pyridine 7 is 14 times more

efficient than p-Cl-phenyl 5 in direct interactions to the

CCR8 binding site, which suggests that the 4-pyridine nitro-

gen makes an H-bond to CCR8. 2,4-Pyrimidine 2 is 4 times

more efficient than 4-pyridine, therefore both nitrogens of

fragment 2 are capable of H-bonding to CCR8. On the other

hand, N-oxide-4-pyridine 3 is as efficient as 2,4-pyrimidine 2,

which is likely to be caused by a stronger H-bonding capacity

fragments are given in Figure 4. Fragments R1 are shown in blue, R2 in

red. The nonorthogonal system of coordinates is made of the average

lipophilicity baselines (the zero level of relative intrinsic potencies in the

compound set) of hERG (the 45ꢀ diagonal line) and CCR8 (the

vertical line) in binding assays. Fragments located below and above

the diagonal axis are intrinsic hERG binding and intrinsic hERG

nonbinding fragments, respectively. Fragments that appear to the left

and to the right from the vertical axis are characteristic of promiscuous

and direct attractive interactions with the primary target, respectively.

Fragments that appear in the first quadrant of the coordinate system

exhibit selective interactions with the primary target and, therefore,

demonstrate the greatest potential for further optimization. The R1

fragment 17 was placed in the first quadrant according to its position

obtained in the analogous plot that used hERG potency in the

electrophysiological assay instead of hERG binding.

7714 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

Figure 4. Structures of fragments of CCR8 antagonists of the focused set classified according to their potential for selective binding to the

primary target as derived from the two-dimensional fragment-based QSAR analysis (Figure 3).

of the N-oxide-pyridine oxygen than the pyridine nitrogen.

Surprisingly, replacement of any of the two H-bond accep-

tors in 2,4-pyrimidine 2 by a H-bond donor in fragments 1, 4,

and 8 does not decrease stability of the H-bonding network

with CCR8. This result is suggestive of flexibility of the polar

subsite of the CCR8 binding site and possible involvement of

residues Gln or Asn. Out of the five R2 fragments, 1-4 and 8,

that demonstrate productive polar interactions with CCR8,

fragments 4 and 8 stand out as most efficient in direct

attractive interactions with CCR8 and direct repulsive inter-

actions with hERG, respectively.

The structure of fragments R1 also dramatically affects

LLE, but their efficiency in direct interactions with CCR8 is

determined more by their geometric complementarity rather

than their H-bonding capacity. Among the 10 R1 fragments

that exhibit a better than average intrinsic binding to the

CCR8 binding site, 3-9 and 15-17, there are fragments with

one and two oxygens. However, the phenoxyphenyl oxygens

in fragments 1, 2, 6, and 7 have been previously established to

have no H-bonding capacity because of delocalization of

their lone pairs in two benzene rings.⁶⁶Results suggest that

one of the oxygens capable of intermolecular H-bonding in

R1 fragments is critical for efficient intrinsic interactions

with the CCR8 binding site, but the second is not. This is the

added first polar oxygen in the ortho-methoxy groups in frag-

ments 6 and 7, which considerably improves otherwise poor

values of LLE of 1 and 2. The second polar oxygen that exists

in fragments 9, 12, and 13 does not further improves the LLE

values, which are already achieved with the single oxygen in

8, 16, and 17. The position of the first polar oxygen in R1 is

important for the binding efficiency. Thus, the ortho oxygen

that exists in bicyclic fragments 8, 9, 16, and 17 is more

efficient than the meta oxygen inherent in 15, whereas the

only fragment with the para oxygen in the data set, 11, does

not fit at all. These observations strongly suggest a single

H-bonding capacity of the corresponding CCR8 subsite,

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7715

which implies that hydrophobic interactions and geometric

complementarity are likely to be critical in this subsite.

The effects of methyl and short alkyl groups in bicyclic

R1 fragments 8, 9, and 12-17 on the nondesolvation rela-

ted interactions with the lipophilic subsite of CCR8 are

remarkable. As is seen in Figure 3, the sole methyl group in

fragment 8 increases the LLE of compounds by 0.9 loga-

rithmic units (compare fragments 3 and 8), which indi-

cates that this particular methyl group significantly imp-

roves geometric complementarity of compounds to the

CCR8 binding site. It should be emphasized that this effect

of the methyl group in 8 on LLE is something that comes in

addition to the normally expected desolvation-driven con-

tribution of 0.5 logarithmic units of a typical methyl group

to compound potency.^47,67It is well documented that in

special cases small alkyl groups in bioactive compounds

may cause unexpected “magic” effects on geometric com-

plementarity or biological outcomes, effects that are far

beyond their desolvation related interactions with bio-

polymers.^{43,44,47,51,68-70}The dramatic effect of the sole

methyl group in 8 on the nondesolvation related interactions

with CCR8 allows us to regard this methyl as “magic”.

Surprisingly, the second methyl group in the bicyclic R1

fragments 9 and 15-17 does not add anything more to the

LLE than is already achieved by the first methyl. The second

methyl increases the CCR8 potency of compounds but adds

just about as much to compound lipophilicity, thereby

having a virtually zero effect on LLE, which is an indication

of its purely desolvation-driven effects.

The dramatic effect of the “magic methyl” in R1 fragment

8 is especially surprising as this fragment is a mixture of two

enantiomers. Our attempts to separate the enantiomers of

the corresponding compound 19 failed. This may be because

a chiral center located at the periphery of the molecule

imparts little steric difference to the stereoisomers of 19,

which makes the separation virtually impossible. Normally

only one of two enantiomers of bioactive compounds ex-

hibits a special fit to the binding site, so the racemate is

supposed to be less efficient than that particular enantiomer,

which contradicts with the exceptional LLE-boosting per-

formance of this “magic methyl”. The results suggest that the

single methyl in 8 is likely equally superior in both config-

urations. To shed light on this apparent discrepancy,

we calculated the 3D structure of this fragment by

fully optimized correlated ab initio calculations at the

LMP2/cc-pVTZ(-f) level^71-73using the program Jaguar 7.5

Figure 5. Superimposed structures of stereoisomers of the LLE-

boosting R1 fragment 8 (Figure 4) with the single methyl being in

the equatorial conformation obtained by ab initio calculations

at the LMP2/cc-pVTZ(-f) level of theory. Carbons of the S- and

R-enantiomers are shown in cyan and dark green, respectively.

Hydrogens are not shown for clarity. Distance between carbon

atoms of the methyl groups in the superimposed enantiomers is

0.78 A.

propensities being well above the average hERG lipophili-

city baseline. All these fragments contain a bicyclic ring

system plus two small alkyl chains attached to the same

carbon. As opposed to the less efficient R1 fragments 6 and 7,

they have only one aromatic ring, which limits their potential

for hydrophobic interactions with the channel and partially

explains their lower contribution to hERG potency but does

not explain their superior anti-hERG propensity given by eq

4. Note that the desolvation-driven components of potencies

are virtually removed from the plot in Figure 3 so the

fact that fragments 12-17 are separated from 6 and 7 by a

hERG lipophilicity baseline is an indication that they lose

much more in hERG potency than one would expect from

their loss of lipophilicity. This extra drop in hERG potency

contributions is due to their geometric incompatibility

with the hERG channel cavity. These fragments also have

significantly higher anti-hERG propensities than monocyc-

lic fragments 4 and 5, which is linked to their conformational

rigidity. The rigidity plus two alkyl chains at the periphery

of the bicyclic system make fragments 12-17 very special

in terms of anti-hERG propensity and set them apart from

more flexible fragments 6 and 7 and also from ordinary

bicyclic fragments 3 and 8. The presence of only one hetero-

atom capable of intermolecular H-bonding inherent in

12-17 is also an important feature, as the bicyclic fragment

9, which has two methyls but also two oxygens, is similar to

8 and nothing special in terms of the anti-hERG propensity.

Thus, rigidity, bulkiness, and one polar heteroatom are the

features that likely make these lipophilic left-hand side

fragments of CCR8 antagonists significantly less predis-

posed for hERG binding than could be expected from their

contribution to compound lipophilicity.

The roles the two peripheral alkyl groups in the bicyclic

lipophilic fragments 12-17 play in hERG selectivity are

different. As opposed to its effect on the intrinsic interactions

with the primary target, the first “magic methyl” group does

not contribute to the anti-hERG propensity because the line

connecting the R1 fragments 3 (no methyls) and 8 (one

methyl) in Figure 3 is about parallel to the hERG lipophi-

licity baseline. On the other hand, it is the second methyl,

which is crucial for the anti-hERG propensity of the bicyclic

R1 fragments, as is seen by comparing locations of R1

fragments 8 and 16 in the plot. Thus, both peripheral alkyl

groups attached to the same sp³-carbon in the bicyclic R1

fragments are “magic” contributors to the hERG selectivity

of the CCR8 antagonists, with the first alkyl improving the

geometric complementarity with the CCR8 binding site and

€

(Schrodinger, LLC). Calculations demonstrate that the di-

hydrofuran ring of 8 is not coplanar with the benzene ring

and adopts an envelope conformation, with each of the

stereoisomers having two equilibrium conformations of the

same energy in which the “magic methyl” is in either

equatorial or axial conformation. The results indicate that

the 3D structures of the stereoisomers with the equatorial

conformations of the “magic methyl” virtually coincide,

whereas the structures with the axial conformations signifi-

cantly differ. Figure 5 presents the overlaid 3D structures of

the stereoisomers with the equatorial “magic methyls” by

superposition of the benzene rings. Thus, we propose that the

“magic methyl” of the R1 fragment 8 is equally efficient in

boosting the stereochemical fit of compounds to the CCR8

binding site in both enantiomers being in the equatorial

conformation.

The data presented in Figure 3 indicate that the R1

fragments 12-17 exhibit remarkable intrinsic anti-hERG

7716 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

Figure 6. Refined docking poses selected by IFP of compounds 13 ((A) green carbon atoms), 16 ((B) yellow carbon atoms), 22 ((C) orange

carbon atoms), 23 ((D) magenta carbon atoms), 7 ((E) dark-cyan carbon atoms), and the R-stereoisomer of 19 ((F) light-brown carbon atoms)

in the CCR8 receptor. Parts of the backbone of TM helices 2, 3, 5, 6, and 7 are represented by light-yellow ribbons in (A). Important binding

residues are depicted in a ball-and-stick fashion with gray carbon atoms. Oxygen, nitrogen, fluorine, and hydrogen atoms are colored red, blue,

brown, and cyan, respectively. H-bonds described in the text are depicted by black dots.

the second alkyl causing steric clashes within the hERG

pocket.

affect binding of nonpeptide ligands to CCR8.²¹Residues at

corresponding positions in the closely related chemokine

receptors CCR1 and CCR5 have been shown to be involved

in antagonist binding.^74-79Figure 6 shows the typical bind-

ing modes of 3,9-diazaspiro[5.5]undecanes automatically

docked in the CCR8 homology model, in line with the

reported site-directed mutagenesis studies and the struc-

ture-activity relationships presented in the present study.

The protonated amine of the 3,9-diazaspiro[5.5]undecane

scaffold forms a salt bridge with the negatively charged

carboxylate group of E^7.39. Modification of this centrally

located basic amine leads to a dramatic loss of CCR8

potency. Thus, the neutral N-oxide of 10 is 1.4 logarithmic

units less potent than its positively charged parent com-

pound. The amide oxygen of the scaffold forms an H-bond

with the Y^3.33hydroxyl group and in some compounds also

with the hydroxyl group of Y^45.39in the second extracellular

loop. CCR8 antagonists occupy both subpockets i (TMs 1, 2,

3 and 7) and ii (TMs 3, 4, 5, 6), making aromatic interactions

Only three lipophilic R1 fragments, 15, 16, and 17, and one

polar R2 fragment, 8, are located in the first quadrant in

Figure 3, which emphasizes their potential for further opti-

mization. Although CCR8 represents the primary target, the

overall success in the lead optimization campaign in increas-

ing the LLE within this particular class is less impressive than

the success in increasing their intrinsic anti-hERG propen-

sity. As is seen in Figure 3, contributions of different frag-

ments to the LLE in the focused set span 3.0 and 2.4

logarithmic units for R1 and R2 fragments, respectively,

whereas their contributions to the anti-hERG propensity

span correspondingly 4.3 and 2.1 logarithmic units. As a

result, hERG selectivity in binding assays within the com-

pound class span 6.3 logarithmic units, with about 2 loga-

rithmic units being attributed to the discovery of the “magic”

compact lipophilic bicyclic R1 fragments with the duo of two

peripheral alkyl chains. This is the lipophilic periphery of the

molecule, which seems to be the most efficient tool in

removing hERG liability in the chemical class under inves-

tigation. The following site-directed mutagenesis studies of

the hERG ion channel provide the structural explanation for

high anti-hERG propensity of the distinct bicyclic lipophilic

left-hand side fragments.

(Y^1.39, Y^3.32, andY^3.33) and hydrophobic contacts (M^3.24

^45.51, M^5.43, F^6.51, and L^6.55) with the receptor. In our

,

Y

receptor model, residue F^2.57does not directly interact with

the antagonist but is a part of the aromatic cluster at the

bottom of subpocket i (Y^1.39, F^2.53, F^2.57, Y^3.32, and F^7.43, not

displayed in Figure 6 for reasons of clarity). Mutation of S^3.29

to a bulkier residue diminishes CCR8 agonist activity, while

the smaller S^3.29A mutant shows comparable agonist activity

as the wild-type receptor.²¹Increasing the size of this residue

would interrupt the course of the narrow gorge between TM3

and TM7 in the receptor model and obstruct binding of the

3,9-diazaspiro[5.5]undecane scaffold (Figure 6).

Structure-Based Rationalization of CCR8-Antagonist In-

teractions. CCR8 Binding Pocket and 3,9-Diazaspiro-

[5.5]undecane Binding Mode. The proposed binding

mode of the 3,9-diazaspiro[5.5]undecane scaffold to the

CCR8 receptor model (Figure 6) is consistent with available

experimental data. Like other chemokine receptors, CCR8

binds its ligands through the conserved glutamate residue

Structure-Based Rationalization of Structure-Activity Re-

lationships. The CCR8 homology model can satisfactorily

rationalize the observed structure-activity relationships and

E

S

.

^{7.39 21,74-79}Site-directed mutagenesis of Y^1.39, F^2.57, Q^2.60

^3.29, Y^3.32, Y^3.33, F^6.51, and L^6.55to Ala indicated that they

,

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7717

the results of the fragment-based QSAR analysis of CCR8

antagonists, which reveal polar interactions with the recep-

tor binding site. The common spirocyclic core of the com-

pound class (Figure 1) exhibits similar binding modes in

different compounds docked to the receptor binding site and

displays two direct attractive interactions, namely one ionic

interaction of the basic amine with E^7.39and one H-bond of

the amide oxygen with Y^3.33. Consistent with the results of

the QSAR analysis (Figure 3), the docking simulations show

that the amide nitrogen atoms of Q^45.49and Q^2.60, as well as

the hydroxyl group of Y^1.39, can form hydrogen bonds with

important polar oxygens in the lipophilic left-hand side (R1)

of the antagonists in subpocket i. As is seen, the Asn residues

that are able to be both H-bond donor and H-bond acceptor

are indeed present in the polar subsite of the CCR8 TM

pocket. The amide groups of N^5.39and N^5.43can form

hydrogen bonds with both H-bond acceptors and donors

of the polar right-hand side (R2) of the antagonists in

subpocket ii. Figure 6 presents the binding modes of six

docked antagonists and demonstrates how modifications in

the R1 and R2 groups are reflected by the number and

character of H-bond interactions between the antagonist

and the CCR8 receptor. Table S2 of Supporting Information

presents the H-bond interaction patterns with R1 and R2

groups of all 30 docked antagonists along with their LLE

values.

4 versus 7). According to our docking simulations and in line

with the results of the fragment-based QSAR analysis, R2

fragment 1 (Figure 6D) forms only one H-bond with the

receptor binding site (compare compounds 25 and 26), while

the amine group of fragment 6 is ineffective because it only

forms an intramolecular H-bond with the diazaspiro oxygen

(compound 23 versus 24). It should be noted, however, that

water mediated contacts, as in the recently published crystal

structure of the adenosine A_2Areceptor,⁵are not taken into

account in the present study.

Analysis of the binding mode of compound 7 suggests that

the size and shape of the R1 fragment 10 might affect the

binding orientation of the R2 group as the H-bond between

the pyridine nitrogen in the ligand and the amide nitrogen of

N

^5.43is disrupted for this compound (Figure 6E). Also in the

complexes of compounds 12 and 15 with CCR8, the lipo-

philic R1 fragments, which possess a weak H-bonding

capacity, disrupt H-bonding between their polar R2 frag-

ments with N^5.43(Table S2 of Supporting Information). The

R1 fragments 3 and 11-14, however, are not predicted to

affect the number of H-bond interactions between fragment

R2 and the receptor (see compounds 17-18 and 28-30 in

Table S2 of Supporting Information) while they have a

negative relative contribution to LLE (Figure 3). Although

more extensive computational efforts would be needed to

accurately determine the geometric complementarity be-

tween R1 fragments and the receptor, the binding mode of

compound 7 in Figure 6E suggests that highly bulky R1

fragments like 10-14 bump into the bottom of subpocket i

between Y^1.39, Q^2.60, Y^3.32, and E^7.39. The R1 fragment 10,

however, has by far the largest negative contribution

to intrinsic CCR8 potency (Figure 3), and its binding

is mainly driven by lipophilicity. The R1 fragment 3 in

compound 20, on the other hand, is probably too compact

and therefore loosely bound (Figure 3), as it lacks

the essential methyl group “anchor” of compound 19

(Figure 6F) to make hydrophobic and packing interactions

in lipophilic subpocket i.

Figure 6A,B shows how the methoxy oxygens of R1

fragments 7 (in compound 13) and 6 (in compound 16)

accept one H-bond from residues Y^1.39and Q^45.49, respec-

tively. The phenoxyphenyl oxygen of fragments 1, 2, 6, and 7

are generally also within H-bond distance from Q^45.49

(Figure 6A,B, Table S2 of Supporting Information), but

aromatic ether oxygen atoms are considered to have a poor

H-bonding capacity,⁶⁶which explains significantly lower

contributions of R1 fragments 1 and 2 to LLE than those

of fragments 3, 4, and 5 (see Figure 3). Alkoxy oxygen atoms

in most other R1 fragments form H-bond interactions with

Q

^45.49and/or Q^2.60in the CCR8 receptor models as demon-

strated for compounds 22 and 23 (Figure 6C,D) and reported

in Table S2 of Supporting Information. Site-directed muta-

genesis of Q^2.60into a tryptophan residue (which is present in

most other chemokine receptors³⁶caused a significant de-

crease in the potency of CCR8 agonists,²¹supporting the

involvement of this residue in ligand binding.

Site-Directed Mutagenesis of the hERG Channel. It is well

documented that potency of hERG blockers within each

compound class is linked to their lipophilicity.^26,80-83Cor-

respondingly, one normally has to expect an increase of

hERG potency when adding lipophilic fragments to a hERG

ion channel blocker (see eq 2). On the other hand, some

distinct exceptions from this rule, when hERG potency

remains unchanged or even decreased with the increase of

compound lipophilicity within a structural class, have also

been described.^43,44It is important to identify the physical

origin of the difference between these alternative effects of

added lipophilicity on hERG potency. Both types of the

effects of lipophilicity are present in the compound set

(Tables 1, 2 and Table S1 of Supporting Information). The

lipophilicity-driven increase of hERG potency is seen when

comparing compounds 5 and 11, 9 and 8, 12 and 11, 20 and

11, 27 and 11, or 26 and 29, respectively. The alternative

effect is illustrated by compound pairs 5 and 7, 20 and 27, 23

and 21, 22 and 30, or 29 and 28, respectively. As is seen in

Figure 3, lipophilic left-hand side fragments, 10-17, reveal

significant intrinsic anti-hERG propensities. Lipophilic

fragments 12 and 14 are most efficient to separate hERG

binding from lipophilicity because they contribute to hERG

binding more than 100 times less than it would be expected

from the “average” left-hand side fragments in this com-

pound series. To shed light on the alternative effects of

The H-bond acceptor and H-bond donor groups in the R2

fragments of antagonists can form one or more H-bonds in

subpocket ii between TMs 3, 5, and 6. The H-bond acceptors

positioned at the 2-position of the R2 ring generally form H-

bonds with the amide nitrogen of N^5.39(Figure 6A-C, Table

S2 of Supporting Information), while H-bond acceptors or

donors at the 3- and 4-positions can form H-bonds with the

amide nitrogen or oxygen atoms of N^5.43(Figure 6A-D,F

and Table S2 of Supporting Information). This is in agree-

ment with the fact that the LLE tends to increase with the

increasing number as well as the H-bond capacity of het-

eroatoms in R2 (Figure 3). The N-oxide oxygen in R2

fragment 3 accepts one H-bond from N^5.43but has a higher

H-bonding capacity than the pyridine nitrogen in fragment

7, resulting in a higher LLE value (Figure 3). Fragments 2

(Figure 6A), 4 (Figure 6C), and 8 (Figure 6B) can form two

H-bonds with N^5.39and N^5.43simultaneously (Table S2 of

Supporting Information), explaining the relative increase in

LLE when comparing compounds 4 and 5 or 8 and 10

(fragment 2 versus 7) and compounds 26 and 27 (fragment

7718 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

Figure 7. Plot of hERG potency in electrophysiological experi-

ments versus measured compound lipophilicity for the subset of

25 CCR8 antagonists chosen for hERG site directed mutagenesis

studies. The numbering of compounds is given in Table 1 and Table

S1 of the Supporting Information. Two types of compound classi-

fications are illustrated. The first classification, along the axes X,

hERG

defined as (pIC₅₀

þ Log D), denotes the direction of the

desolvation component of hERG potency. The second classifica-

tion, along the axis Y, defined as -(pIC₅₀^hERG- Log D), signifies

repulsive nondesolvation related interactions with the hERG chan-

nel cavity (eq 4). The average blue line with the slope of -1 divides

the compound set into two classes, hydrophobic (shown as stars)

and polar (shown as spheres). The average red line with the slope of

þ1 splits the compound set into two classes of intrinsic hERG

nonbinders (shown in green) and intrinsic hERG binders (shown in

red). The following six compounds that cannot be unequivocally

classified are excluded in the second classification: 2, 3, 14, 18, 23,

and 26.

Figure 8. Results of two types of PLS-DA for different effects of

added lipophilicity on hERG potency in electrophysiological assays

in the subset of 25 CCR8 antagonists as described by the hERG site-

directed mutagenesis data (Table 2 and Table S1 of the Supporting

Information) as variables. Plot (A) describes the desolvation com-

ponent of hERG potency; plot (B) describes repulsive nondesolva-

tion related interactions within the hERG cavity. The coefficients of

the optimal cross-validated PLS-DA linear equations are presented

as histograms for hERG mutants T623S, S624A, S624T, Y652A,

Y652F, F656M, F656T, and F656W. Positive coefficients indicate

that compounds with added lipophilicity are more potent in the

wild-type hERG than in the particular mutant with respect to the

compounds of the alternative class. The error bars correspond to the

confidence level of 95%. Highly significant coefficients are under-

lined.

lipophilicity on hERG potency and, especially on the mo-

lecular mechanism of the anti-hERG propensity of the

particular lipophilic fragments of CCR8 antagonists, elec-

trophysiological measurements of eight hERG mutants have

been performed using 25 compounds (Table 2). In each of

these mutants, one of the residues that have been previously

shown to be critical for binding of various hERG blockers,

namely T623, S624, Y652, and F656,^84-87was changed to

make the following eight mutants of the channel: T623S,

S624A, S624T, Y652A, Y652F, F656M, F656T, and

F656W. The differences in potencies obtained at the wild-

type hERG and the hERG mutants are presented in Table 2.

Figure 7 presents the plot of hERG potency of the 25 test

compounds, which were chosen from the focused set, mea-

sured in electrophysiological experiments against their ob-

served lipophilicity (Log D_7.4). A cross-validated PLS

discriminant analysis (PLS-DA)⁸⁸was applied to this com-

pound set to understand the molecular basis for the two

alternative effects of lipophilicity on hERG potency. Ac-

cordingly, two alternative classifications of this compound

set were made. In the first classification, compounds were

split into two classes along the axis X. In this particular

classification, the compounds were grouped according to the

The eight variables, i.e., the differences between potencies at

wt-hERG

wild-type hERG and at the particular mutants (pIC₅₀

- pIC₅₀^mutant), were centered but not scaled in order to take

the magnitude of the observed effects of different mutations

into account. This analysis helps to identify places in the

hERG ion channel, where compounds of the different classes

behave in the opposite ways. The cross-validated correla-

tion coefficient obtained for this analysis was q²=0.48. The

coefficients of the optimal linear equation inherent in the

more lipophilic class are presented in Figure 8A. A similar

analysis was performed using the alternative classification of

the set of compounds along axis Y in Figure 7, i.e., in the

direction of the anti-hERG propensity given by eq 4. In this

case, six compounds, namely 2, 3, 14, 18, 23, and 26, were

excluded from classification as the corresponding data

points were located too close to the average hERG lipophi-

licity baseline shown in red in Figure 7. Nine compounds

with higher intrinsic anti-hERG propensities were classified

as intrinsic hERG nonbinders, and the remaining 10 less

lipophilic compounds as intrinsic hERG binders. The result-

ing PLS-DA showed the value of the cross-validated correla-

tion coefficient of 0.34. The coefficients of the optimal linear

equation characteristic of the more lipophilic class are pre-

sented in Figure 8B.

sign of their contribution to the lipophilicity-driven compo-

hERG

nent of hERG potency (pIC₅₀

þ Log D) with respect to

its average value given by the blue line in Figure 7. Twelve

compounds were classified as lipophilic hERG binders and

the remaining 13 compounds as polar hERG binders. The

PLS-DA was performed using the site-directed mutagenesis

data (Table 2 and Table S1 of Supporting Information) as

variables (SIMCA-Pþ version 11; Umetrics, Kinnelon NJ).

Coefficients of the PLS-DA equations shown in Figure 8

indicate where the two alternative types of added lipophilicity

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7719

in the compound set tend to interact with the hERG channel.

A positive (or negative) coefficient of the PLS-DA equations

indicates that the compounds with the added lipophilicity

demonstrate a higher (or a lower) difference between potency

at hERG and at the corresponding mutant than the less

lipophilic compounds. As is seen in Figure 8A, the most

significant feature that sets apart lipophilic and polar hERG

binders is linked to the higher increase of potencies of

compounds in the lipophilic class at mutants Y652A and

Y652F than at the wild-type hERG channel and the higher

increase of potencies at wild-type hERG than at the mutant

S624T. In all of these three hERG mutants, Y652A, Y652F,

and S624T, the lipophilicity-driven increase of potency is

linked to the increase of the hERG channel vestibule size

upon mutation. This suggests that the desolvation related

component of hERG potency studied in the direction X

(Figure 7) is associated with the added compound lipophili-

city that fills the available space within the hERG channel

vestibule. This result seems to disagree with the current

opinion of the fundamental role of residues Y652 and F656

in hydrophobic interactions with hERG blockers.^42,84-87

However, the obtained data do not contradict with the

importance of these key residues of the hERG cavity for

binding of all known blockers. The PLS-DA focuses on what

is different in the two classes but not on what is common. The

origin of the appeared discrepancy can be understood by

studying structural features of compounds that belong to the

more lipophilic class along the desolvation component of

hERG potency. Analysis of the 25 compounds that have

been studied by site-directed mutagenesis suggests that the 2-

phenoxyphenyl moiety dominates in this class. Four out of

five compounds in the data set that possess this moiety,

namely 2, 3, 10, and 15, belong to this class. This bulky

moiety does not fit to the lower part of the hERG

channel and likely fills the vestibule cavity,⁴⁴which is in

agreement with earlier studies that identified the ortho

topology in hERG blockers as especially predisposed for

hERG binding.⁴²The lipophilic 2-phenoxyphenyl fragment

needs a large room to be properly accommodated, corre-

spondingly the more room that is available for this group in

the channel vestibule, the higher the increase of potency is

caused by the addition of this type of lipophilicity in the

compound set, whether it is hERG or hERG mutant. On the

other hand, most of the molecules of the less lipophilic

class bind the hERG cavity in a classical way, in which

the lipophilic periphery interacts with F656.⁴²Accordingly,

the PLS-DA detects that the lipophilicity-driven component

of hERG potency in the particular data set is associated with

filling the vestibule cavity. This result confirms our earlier

predictions of positioning of the ortho-phenoxyphenyl

peripheral moiety of CCR8 antagonists in the hERG

vestibule.⁴⁴

channel lacks four critical phenylalanine residues, and is

much less hydrophobic at this place. Note that phenylalanine

is roughly 1.4 logarithmic units more hydrophobic than

threonine. Nevertheless, compounds 26, 28, 29, and 30 of

the more lipophilic class are roughly equipotent at the wild-

type hERG and at the F656T mutant (Table 2). These are

the first compounds ever reported that do not drop potency

at the F656T mutant. In addition to the decrease in local

lipophilicity, this particular place of the F656T mutant

cavity displays more space available for blockers. The

fact that compounds of the more lipophilic class demons-

trate extra difficulties in binding the wild-type hERG but

less problems with binding the F656T mutant suggests

that the lipophilic peripheral fragments of these hERG

blockers, which are supposed to bind to this particular

narrow place in the funnel-looking hERG channel,⁴²simply

do not fit between the four residues F656 in hERG. Thus, the

PLS-DA based on site-directed mutagenesis data and per-

formed in the direction of the anti-hERG propensity in

the particular data set suggests that the added lipophilicity

does not fit to the hERG channel cavity in the region of

residue F656.

The fact that the solution of the PLS-DA appears more

blurry for the desolvation-driven component of hERG

potency (Figure 8A) than for the anti-hERG propensity

(Figure 8B) was expected. Desolvation-linked hydrophobic

interactions of lipophilic compounds with lipophilic binding

sites are nonspecific as they are virtually independent of the

chemical nature of lipophilicity. Correspondingly, there are

generally an abundance of lipophilic corners in binding sites

and a vast variety of structural forms of lipophilic motifs in

compounds, where these hydrophobic interactions can be

exploited. On the other hand, the nondesolvation related

direct interactions of compounds with binding sites are

determined by stereochemical complementarity, which is

highly dependent on structure. Whereas it is easier to in-

crease compound potency by increasing lipophilicity, struc-

tural alterations leading to changes in the nondesolvation

related interactions with binding sites, both intended and

unintended, have more value in overcoming selectivity-re-

lated issues.

Results of the PLS-DA provide the interpretation for the

observed geometric features of distinct lipophilic fragments

of CCR8 antagonists that exhibit significant anti-hERG

propensity. Figure 9 presents correlations of the relative

anti-hERG propensity of a number of lipophilic left-hand

side fragments of the CCR8 antagonists and their relative

selectivity toward the F656T mutant versus the wild-type

hERG ion channel. The plot was obtained by two-dimen-

sional fragment-based QSAR analysis. The illustrated data

suggest that the increase of the anti-hERG propensity of

compounds caused by conformational rigidity and bulkiness

of the distinct R1 fragments is indeed linked to difficulties in

fitting compounds to the wild-type hERG with respect to the

more spacious mutant F656T. As is seen, the preference of

the two-oxygen R1 fragments 9, 13, and 12 for the F656T

mutant does not increase as much as that of the one-oxygen

fragments 5, 8, 16, 17, and 14. One possible explanation for

this observation is that the mechanism of the anti-hERG

propensity of the two-oxygen lipophilic R1 fragments in-

volves multiple binding modes in the channel. In this case,

bulky lipophilic fragments of this type could partially avoid

clashes with the F656 residues of hERG channel by position-

ing in the vestibule cavity. Rigid bulky lipophilic peripheral

Within the second type of compound classification

(Figure 8B), the most significant difference between the

two classes is found to be in the F656T mutant. The negative

value of the corresponding coefficient indicates that the more

lipophilic class in the direction of the anti-hERG propensity

(axis Y in Figure 7) demonstrates more difficulties in binding

the wild-type hERG with respect to the F656T mutant than

the less lipophilic class. Hydrophobic interactions of lipo-

philic peripheral fragments of hERG blockers with residue

F656 in wild-type hERG are known to belong to the major

determinants of their binding to the channel.^{42,43,85,87,89}The

F656T mutant is free of this determinant, as this mutant

7720 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

adding lipophilic groups. One of such corners, which is

located in a vicinity of hERG pore helices, has been recently

identified by site-directed mutagenesis studies using ana-

logues of clofilium.⁸⁵

The obtained results suggest a possible way of overcoming

unintentional hERG binding when adding lipophilic frag-

ments in chemical series. One should avoid lipophilic frag-

ments that may fill the available space in the hERG ion

channel cavity. Instead, one has to expand the lipophilic

groups in the peripheral part of the molecules where they can

cause steric clashes within the channel. Figure 10 gives a

schematic representation of the direct repulsive interactions

induced in the hERG channel by bulky lipophilic left-hand

side fragment 12 in compound 28. Consistent with the

fragment-based QSAR studies, this is the second alkyl group

in these fragments, which does not fit to the hERG channel

cavity because the first alkyl likely avoids steric clashes with

F656 by positioning along the channel axis in the equatorial

conformation.

Figure 9. Correlations between anti-hERG propensity of lipophilic

R1 fragments and their contribution to the selectivity of compounds

for the F656T hERG mutant against the wild-type hERG. Axes X

and Y are relative contributions of fragments to (pIC₅₀^hERG- Log D)

and ΔF656T=(pIC₅₀^wt-hERG- pIC₅₀^F656T-hERG), respectively, cal-

culated by the Free-Wilson approach. Fragments are numbered

according to Figure 4. Fragments with one oxygen are designated by

squares, those with two oxygens by triangles. Linear regressions

built separately on these two groups are illustrated.

Conclusions

The increase of selectivity of biologically active compounds

against undesirable binding to homologous proteins and

various target-unrelated sites of interaction is very important

in the lead optimization process. Most of the selectivity issues

with respect to target-unrelated binding can be solved by

increasing ligand lipophilicity efficiency (LLE) in the com-

pound series. This is usually done by adding polar groups to a

lead compound that form direct attractive interactions with

the primary binding site. Apart from the important ionic

interaction of the centrally located basic amine in CCR8

antagonists with E^7.39, we found four additional H-bonds

that can be formed with CCR8, three of them associated with

the polar subsite and one with the lipophilic subsite of the

receptor.

Whenpolarityof the leadcompound isalreadyatthe higher

limit, such that the further increase of polarity would jeopar-

dize DMPK properties, the selectivity issues can alternatively

be solved by adding lipophilic fragments to the molecules to

induce direct repulsive interactions with the specific antitar-

get. In this report, we propose that the increase of bulkiness

and rigidity of the lipophilic periphery of compounds in a

series of CCR8 antagonists induces van der Waals clashes

with F656 in the hERG ion channel. This residue forms the

narrow part of the hERG ion channel, such that the bulky

lipophilic peripheral moieties of the CCR8 antagonists do not

fit. Site-directed mutagenesis studies of the hERG ion channel

played a pivotal role in the interpretation of the observed

anti-hERG propensity of distinct fragments of the CCR8

antagonists.

It is not easy to identify places in lipophilic binding sites

of promiscuous antitargets like hERG, albumin, or metabolic

enzymes, where lipophilic motifs of ligands can induce repul-

sive interactions, and the systematic investigation of nonde-

solvation related interactions with antitargets in a particular

chemical class augmented by site-directed mutagenesis, is

an effective pathway to detect these places. The emphasis

on the nondesolvation related components of potencies,

rather than on the overall affinities, illuminates structural

features of molecules, which are important for selectivity

and helps to develop compounds with enhanced safety mar-

gins, thereby increasing their chances to become drugs.

This result is in line with previous studies that highlighted

Figure 10. Interpretation of the intrinsic anti-hERG propensity of

lipophilic peripheral fragments within the subset of CCR8 antago-

nists as it is seen from the results of hERG site directed mutagenesis.

Only two of four subunits of the hERG channel are displayed for

simplicity. Putative binding mode of compound 28 is illustrated; its

bulky lipophilic group does not fit well to the hERG cavity in the

space available between residues F656. The geometry of the channel

was generated from a homology model of the closed hERG channel

using the X-ray crystallographic structure of KcsA potassium ion

channel (PDB entry 1R3J)⁹⁵as it is described in ref 44. The residues

that are known to be critical for binding of hERG ion channel

blockers are illustrated.

fragments of CCR8 antagonists cause nondesolvation re-

lated repulsive interactions within the hERG ion channel

cavity, thereby increasing selectivity against hERG because

they do not fit to the space in the narrow part of the hERG

channel between the four residues F656 but fit to the

primary binding site. There are other narrow corners in the

hERG channel that can be utilized for inducing direct

repulsive interactions in particular structural classes by

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7721

€

ꢀ

the importance of nondesolvation related interactions for

selectivity of drugs.^41,49

(9) Deupi, X.; Dolker, N.; Lopez-Rodrıguez, M. L.; Campillo, M.;

Ballesteros, J. A.; Pardo, L. Structural models of class A G-protein-

coupled receptors as a tool for drug design: insights on transmem-

brane bundle plasticity. Curr. Top. Med. Chem. 2007, 7, 991–998.

Experimental Section

ꢀ

(10) Qu, C.; Edwards, E. W.; Tacke, F.; Angeli, V.; Llodra, J.; Sanchez-

Schmitz, G.; Garin, A.; Haque, N. S.; Peters, W.; van Rooijen, N.;

Sanchez-Torres, C.; Bromberg, J.; Charo, I. F.; Jung, S.; Lira, S.

A.; Randolph, G. J. Role of CCR8 and other chemokine pathways

in the migration of monocyte-derived dendritic cells to lymph

nodes. J. Exp. Med. 2004, 200, 1231–1241.

The potency of binding of the compounds under consideration

to CCR8 was measured in the binding assay described in the

literature.^22,90The potency of binding to the hERG potassium

ion channel was measured in an HEK cell line expressing

recombinant hERG channel.^91,92Measurements of compound

lipophilicity, as described by the logarithm of the apparent

n-octanol/water partition coefficient at pH=7.4, Log D_7.4, are

detailed in the literature.^93,94Purity of compounds was deter-

mined to be more than 95% by HPLC with detection carried out

at 220, 254, and 280 nm. For the electrophysiological measure-

ments, an approach developed by Molecular Devices (Sunnyvale

CA) that records averaged ionic currents from a population of

cells expressing a recombinant hERG ion channel was utilized.

Cells were plated into a 384-well PatchPlate substrate in which

each well contained multiple recording sites, and the ionic

currents were recorded using IonWorks HT medium-throughput

electrophysiology device.

(11) Soler, D.; Chapman, T. R.; Poisson, L. R.; Wang, L.; Cote-Sierra,

J.; Ryan, M.; McDonald, A.; Badola, S.; Fedyk, E.; Coyle, A. J.;

Hodge, M. R.; Kolbeck, R. CCR8 Expression identifies CD4

memory T cells enriched for FOXP3þ regulatory and Th2 effector

lymphocytes. J. Immunol. 2006, 177, 6940–6951.

(12) Panina-Bordignon, P.; Papi, A.; Mariani, M.; Di Lucia, P.; Casoni,

G.; Bellettato, C.; Buonsanti, C.; Miotto, D.; Mapp, C.; Villa, A.;

Arrigoni, G.; Fabbri, L. M.; Sinigaglia, F. The C-C chemokine

receptors CCR4 and CCR8 identify airway T cells of allergen-

challenged atopic asthmatics. J. Clin. Invest. 2001, 107, 1357–1364.

ꢁ

(13) Sebastiani, S.; Albanesi, C.; De Pita, O.; Puddu, P.; Cavani, A.;

Girolomoni, G. The role of chemokines in allergic contact derma-

titis. Arch. Dermatol. Res. 2002, 293, 552–559.

(14) Peroni, D. G.; Panina-Bordignon, P.; Piacentini, G. L.; Bodini, A.;

Ress, M.; Mariani, M.; Sinigaglia, F.; Boner, A. L. CC chemokine

receptor expression in childhood asthma is influenced by natural

allergen exposure. Pediatr. Allergy Immunol. 2006, 17, 495–500.

(15) Chensue, S. W.; Lukacs, N. W.; Yang, T. Y.; Shang, X.; Frait, K.

A.; Kunkel, S. L.; Kung, T.; Wiekowski, M. T.; Hedrick, J. A.;

Cook, D. N.; Zingoni, A.; Narula, S. K.; Zlotnik, A.; Barrat, F. J.;

O’Garra, A.; Napolitano, M.; Lira, S. A. Aberrant in vivo T helper

type 2 cell response and impaired eosinophil recruitment in CC

chemokine receptor 8 knockout mice. J. Exp. Med. 2001, 193, 573–

584.

(16) Buckland, K. F.; O’Connor, E. C.; Coleman, E. M.; Lira, S. A.;

Lukacs, N. W.; Hogaboam, C. M. Remission of chronic fungal

asthma in the absence of CCR8. J. Allergy Clin. Immunol. 2007,

119, 997–1004.

(17) Norman, P. CCR8 antagonists. Expert Opin. Ther. Patents 2007,

17, 465–469.

(18) Pease, J. E.; Horuk, R. Chemokine receptor antagonists: Part 2.

Acknowledgment. We gratefully acknowledge the support

of many people at AstraZeneca R&D who have contributed

to some aspects of the work presented in this paper. We

thank Tim Hammond, Department of Safety Pharmacology,

Safety Assessment UK, AstraZeneca R&D Alderley Park, for

his sponsorship of electrophysiological studies of hERG

mutants. We express our gratitude to Nicholas Tomkinson

(AstraZeneca R&D Charnwood, UK) for writing a script for

the QSAR analysis presented in the paper. C.d.G. gratefully

acknowledges financial support from AstraZeneca R&D

through a postdoctoral scholarship.

Expert Opin. Ther. Patents 2009, 19, 199–221.

(19) Pease, J. E.; Horuk, R. Chemokine receptor antagonists: Part 1.

Expert Opin. Ther. Patents 2009, 19, 39–58.

Supporting Information Available: Experimental procedures

and characterization data for compounds 1, 3, 5, 8, 10-13, 15,

17, 18, 25, and 28-30. Full structures and experimental results

obtained for the focused set of CCR8 antagonists as well as

details of homology modeling and docking. This material is

available free of charge via the Internet at http://pubs.acs.org.

(20) Dragic, T.; Trkola, A.; Thompson, D. A. D.; Cormier, E. G.;

Kajumo, F. A.; Maxwell, E.; Lin, S. W.; Ying, W.; Smith, S. O.;

Sakmar, T. P.; Moore, J. P. A binding pocket for a small molecule

inhibitor of HIV-1 entry within the transmembrane helices of

CCR5. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 5639–5644.

(21) Jensen, P. C.; Nygaard, R.; Thiele, S.; Elder, A.; Zhu, G.; Kolbeck,

R.; Ghosh, S.; Schwartz, T. W.; Rosenkilde, M. M. Molecular

interaction of a potent nonpeptide agonist with the chemokine

receptor CCR8. Mol. Pharmacol. 2007, 72, 327–340.

References

(22) (a) Bladh, H.; Connolly, S.; Dyke, H. J.; Lisius, A.; Price, S.;

Shamovsky, I.; van den Heuvel, M. Preparation of novel diazas-

piroalkanes and their use for treatment of chemokine receptor

CCR8 mediated diseases. PCT Int. Appl. WO2005040167, 2005,

136 pp. (b) Connolly, S.; Skrinjar, M. Novel diazaspiroalkanes and their

use for treatment of CCR8 mediated diseases. PCT Int. Appl.

WO2006107252, 2006, 47 pp. (c) Connolly, S.; Linnanen, T.; Skrinjar,

M. Novel diazaspiroalkanes and their use for treatment of CCR8

mediated diseases. PCT Int. Appl. WO2006107253, 2006, 55 pp. (d)

Connolly, S.; Skrinjar, M. Novel diazaspiroalkanes and their use for

treatment of CCR8 mediated diseases. PCT Int. Appl. WO2006107254,

(1) Gerard, C.; Rollins, B. J. Chemokines and disease. Nat. Immunol.

2001, 2, 108–115.

(2) Palczewski, K.; Kumasaka, T.; Hori, T.; Behnke, C. A.; Motoshima,

H.; Fox, B. A.; Le Trong, I.; Teller, D. C.; Okada, T.; Stenkamp, R.

E.; Yamamoto, M.; Miyano, M. Crystal structure of rhodopsin: A

G-protein-coupled receptor. Science 2000, 289, 739–745.

(3) Cherezov, V.; Rosenbaum, D. M.; Hanson, M. A.; Rasmussen, S.

G.; Thian, F. S.; Kobilka, T. S.; Choi, H. J.; Kuhn, P.; Weis, W. I.;

Kobilka, B. K.; Stevens, R. C. High-resolution crystal structure of

an engineered human beta2-adrenergic G protein-coupled recep-

tor. Science 2007, 318, 1258–1265.

(4) Warne, T.; Serrano-Vega, M. J.; Baker, J. G.; Moukhametzianov,

R.; Edwards, P. C.; Henderson, R.; Leslie, A. G.; Tate, C. G.;

Schertler, G. F. Structure of a β₁-adrenergic G-protein-coupled

receptor. Nature 2008, 454, 486–491.

€

2006, 44 pp. (e) Borjesson, L.; Connolly, S.; Johansson, H.; Kristof-

fersson, A.; Linnanen, T.; Shamovsky, I.; Skrinjar, M. Preparation of

novel diazaspiroalkanes for treatment of CCR8 mediated diseases. PCT

Int. Appl. WO2007030061, 2007, 257 pp.

(23) Jin, J.; Wang, Y.; Wang, F.; Kerns, J. K.; Vinader, V. M.; Hancock,

A. P.; Lindon, M. J.; Stevenson, G. I.; Morrow, D. M.; Rao, P.;

Nguyen, C.; Barrett, V. J.; Browning, C.; Hartmann, G.; Andrew,

D. P.; Sarau, H. M.; Foley, J. J.; Jurewicz, A. J.; Fornwald, J. A.;

Harker, A. J.; Moore, M. L.; Rivero, R. A.; Belmonte, K. E.;

Connor, H. E. Oxazolidinones as novel human CCR8 antagonists.

Bioorg. Med. Chem. Lett. 2007, 17, 1722–1725.

(5) Jaakola, V. P.; Griffith, M. T.; Hanson, M. A.; Cherezov, V.;

Chien, E. Y.; Lane, J. R.; Ijzerman, A. P.; Stevens, R. C. The 2.6

angstrom crystal atructure of a human A_2Aadenosine receptor

bound to an antagonist. Science 2008, 322, 1211–1217.

(6) Ballesteros, J.; Palczewski, K. G-protein-coupled receptor drug

discovery: implications from crystal structure of rhodopsin. Curr.

Opin. Drug Discovery Dev. 2001, 4, 561–574.

(24) Jenkins, T. J.; Guan, B.; Dai, M.; Li, G.; Lightburn, T. E.; Huang,

S.; Freeze, B. S.; Burdi, D. F.; Jacutin-Porte, S.; Bennett, R.; Chen,

W.; Minor, C.; Ghosh, S.; Blackburn, C.; Gigstad, K. M.; Jones,

M.; Kolbeck, R.; Yin, W.; Smith, S.; Cardillo, D.; Ocain, T. D.;

Harriman, G. C. Design, synthesis, and evaluation of naphthalene-

sulfonamide antagonists of human CCR8. J. Med. Chem. 2007, 50,

566–584.

(7) Ballesteros, J. A.; Shi, L.; Javitch, J. A. Structural mimicry in

G-protein-coupled receptors: implications of the high-resolution

structure of rhodopsin for structure-function analysis of rhodop-

sin-like receptors. Mol. Pharmacol. 2001, 60, 1–19.

(8) de Graaf, C.; Rognan, D. Customizing G-protein-coupled receptor

models for structure-based virtual screening. Curr. Pharm. Des.

2009, in press.

7722 Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23

Shamovsky et al.

(25) Morgan, T. K., Jr.; Sullivan, M .E. An overview of Class III

electrophysiological agents: a new generation of antiarrhythmic

therapy. Prog. Med. Chem. 1992, 29, 65–108.

(51) Kellenberger, E.; Springael, J. Y.; Parmentier, M.; Hachet-Haas,

M.; Galzi, J. L.; Rognan, D. Identification of nonpeptide CCR5

receptor agonists by structure-based virtual screening. J. Med.

Chem. 2007, 50, 1294–1303.

(26) Aronov, A. M. Ligand structural aspects of hERG channel block-

ade. Curr. Top. Med. Chem. 2008, 8, 1113–1127.

(52) Bissantz, C.; Logean, A.; Rognan, D. High-throughput modeling

of human G-protein-coupled receptors: amino acid sequence align-

ment, three-dimensional model building, and receptor library

screening. J. Chem. Inf. Comput. Sci. 2004, 44, 1162–1176.

(53) Ballesteros, J. A.; Weinstein, H. Integrated methods for the con-

struction of three-dimensional models and computational probing

of structure-function relations in G-protein-coupled receptors.

Methods Neurosci. 1995, 25, 366–428.

(54) Brady, G. P.; Pieter, F. W. S. Fast prediction and visualization of

protein binding pockets with PASS. J. Comput.-Aided Mol. Des.

2000, 14, 383–401.

(55) de Graaf, C.; Rognan, D. Selective structure-based virtual screen-

ing for full and partial agonists of the β₂adrenergic receptor. J.

Med. Chem. 2008, 51, 4978–4985.

(56) Marcou, G.; Rognan, D. Optimizing fragment and scaffold dock-

ing by use of molecular interaction fingerprints. J. Chem. Inf.

Model. 2007, 47, 195–207.

(57) Schroeder, K.; Neagle, B.; Trezise, D. J.; Worley, J.; IonWorks, H.

T. A new high-throughput electrophysiology measurement plat-

form. J. Biomol. Screen. 2003, 8, 50–64.

(58) Bridgland-Taylor, M. H.; Hargreaves, A. C.; Easter, A.; Orme, A.;

Henthorn, D. C.; Ding, M.; Davis, A. M.; Small, B. G.; Heapy, C.

G.; Abi-Gerges, N.; Persson, F.; Jacobson, I.; Sullivan, M.;

Albertson, N.; Hammond, T. G.; Sullivan, E.; Valentin, J. P.;

Pollard, C. E. Optimization and validation of a medium-through-

put electrophysiology-based hERG assay using IonWorks HT. J.

Pharmacol. Toxicol. Methods 2006, 54, 189–199.

(59) Obach, R. S.; Lombardo, F.; Waters, N. J. Trend analysis of a

database of intravenous pharmacokinetic parameters of humans

for 670 drug compounds. Drug Metab. Dispos. 2008, 36, 1385–

1405.

(27) Wang, X.-J.; Yang, Q.; Yin, D.-L.; Chen, Y.-D.; You, Q.-D. A

pharmacophore modeling study of drugs inducing cardiotoxic side

effects. Chin. J. Chem. 2008, 26, 2125–2132.

(28) de Graaf, C.; Vermeulen, N. P.; Feenstra, K. A. Cytochrome P450

in silico: an integrative modeling approach. J. Med. Chem. 2005, 48,

2725–2755.

(29) Westby, M.; van der Ryst, E. CCR5 antagonists: host-targeted

antivirals for the treatment of HIV infection. Antiviral Chem.

Chemother. 2005, 16, 339–354.

(30) Horuk, R. BX471: A CCR1 antagonist with anti-inflammatory

activity in man. Mini-Rev. Med. Chem. 2005, 5, 791–804.

(31) Fox, D. J.; Reckless, J.; Wilbert, S. M.; Greig, I.; Warren, S.;

David, J.; Grainger, D. J. Identification of 3-(acylamino)azepan-2-

ones as stable broad-spectrum chemokine inhibitors resistant to

metabolism in vivo. J. Med. Chem. 2005, 48, 867–874.

(32) De Lucca, G. V. Recent developments in CCR3 antagonists. Curr.

Opin. Drug Discovery Dev. 2006, 9, 516–524.

(33) Purandare, A. V.; Wan, H.; Somerville, J. E.; Burke, C.; Vaccaro,

W.; Yang, X.; McIntyre, K. W.; Poss, M. A. Core exploration in

optimization of chemokine receptor CCR4 antagonists. Bioorg.

Med. Chem. Lett. 2007, 17, 679–682.

(34) Sato, I.; Morihira, K.; Inami, H.; Kubota, H.; Morokata, T.;

Suzuki, K.; Iura, Y.; Nitta, A.; Imaoka, T.; Takahashi, T.; Takeu-

chi, M.; Ohta, M.; Tsukamoto, S. Design and synthesis of 6-fluoro-

2-naphthyl derivatives as novel CCR3 antagonists with reduced

Cyp2D6 inhibition. Bioorg. Med. Chem. 2008, 16, 8607–8618.

(35) Leeson, P. D.; Springthorpe, B. The influence of drug-like concepts

on decision-making in medicinal chemistry. Nat. Drug Discovery

2007, 6, 881–890.

(36) Surgand, J. S.; Rodrigo, J.; Kellenberger, E.; Rognan, D. A

chemogenomic analysis of the transmembrane binding cavity of

human G-protein-coupled receptors. Proteins 2006, 62, 509–538.

(37) Lewis, D. F. V.; Dickins, M. Baseline lipophilicity relationships in

human cytochromes P450 associated with drug metabolism. Drug

Metab. Rev. 2003, 35, 1–18.

(60) Hughes, J. D.; Blagg, J.; Price, D. A.; Bailey, S.; DeCrescenzo, G.

A.; Devraj, R. V.; Ellsworth, E.; Fobian, Y. M.; Gibbs, M. E.;

Gilles, R. W.; Greene, N.; Huang, E.; Krieger-Burke, T.; Loesel, J.;

Wager, T.; Whiteley, L.; Zhang, Y. Physiochemical drug properties

associated with in vivo toxicological outcomes. Bioorg. Med.

Chem. Lett. 2008, 18, 4872–4875.

(61) Dey, A.; Okamura, T.; Ueyama, N.; Hedman, B.; Hodgson, K. O.;

Solomon, E. I. Sulfur K-edge XAS and DFT calculations on P450

model complexes: effects of hydrogen bonding on electronic struc-

ture and redox potentials. J. Am. Chem. Soc. 2005, 127, 12046–

12053.

(38) Lewis, D. F. V. Quantitative structure-activity relationships

(QSARs) for substrates of human cytochromes P450 CYP2 family

enzymes. Toxicol. In Vitro 2004, 18, 89–97.

(39) Lewis, D. F. V. Hydrogen bonding in human P450-substrate

interactions: a major contribution to binding affinity. The Scien-

tific World 2004, 4, 1074–1082.

(40) Lewis, D. F. V.; Jacobs, M. N.; Dickins, M. Compound lipophi-

licity for substructure binding to human P450s in drug metabolism.

Drug Discovery Today 2004, 9, 530–537.

(41) Ruben, A. J.; Kiso, Y.; Freire, E. Overcoming roadblocks in lead

optimization: a thermodynamic perspective. Chem. Biol. Drug Des.

2006, 67, 2–4.

(62) Lewis, D. F. V.; Lake, B. G.; Ito, Y.; Dickins, M. Lipophilicity

relationships in inhibitors of CYP2C9 and CYP2C19 enzymes. J.

Enzyme Inhib. Med. Chem. 2006, 21, 385–389.

(63) de Graaf, C.; Oostenbrink, C.; Keizers, P. H. J.; van Vugt-Lussen-

burg, B. M. A.; Commandeur, J. N. M.; Vermeulen, N. P. E. Free

energies of binding of R- and S-propranolol to wild-type and

F483A mutant cytochrome P450-2D6 from molecular dynamics

simulations. Eur. Biophys. J. 2007, 36, 589–599.

(42) Aronov, A. M. Predictive in silico modeling for hERG channel

blockers. Drug Discovery Today 2005, 10, 149–155.

(43) Jamieson, C.; Moir, E. M.; Rankovic, Z.; Wishart, G. Medicinal

chemistry of hERG optimizations: highlights and hang-ups. J.

Med. Chem. 2006, 49, 5029–5046.

(64) Franklin, T. J.; Jacobs, V. N.; Jones, G.; Ple, P. Human colorectal

carcinoma cells in vitro as a means to assess the metabolism of

analogs of mycophenolic acid. Drug Metab. Dispos. 1997, 25, 367–

370.

ꢀ

(44) Shamovsky, I. L.; Connolly, S.; David, L.; Ivanova, S.; Norden, B.;

Springthorpe, B.; Urbahns, K. Overcoming undesirable hERG

potency of chemokine receptor antagonists using baseline lipophi-

licity relationships. J. Med. Chem. 2008, 51, 1162–1178.

(45) Kubinyi, H. Lipophilicity and drug activity. Prog. Drug. Res. 1979,

23, 97–198.

(46) Lewis, D. F. V.; Lake, B. G.; Dickins, M. Quantitative structure-

activity relationships (QSARs) in inhibitors of various cyto-

chromes P450: The importance of compound lipophilicity. J.

Enzyme Inhib. Med. Chem. 2007, 22, 1–6.

(65) Free, S. M., Jr.; Wilson, J. W. A mathematical contribution to

structure-activity studies. J. Med. Chem. 1964, 7, 395–399.

(66) Kubinyi, H. Hydrogen bonding, the last mystery in drug design? In

Pharmacokinetic Optimization in Drug Research. Biological, Phys-

icochemical, and Computational Strategies; Testa, B., van de Water-

beemd, H., Folkers, G., Guy, R., , Eds.; Helvetica Chimica Acta and

Wiley-VCH: Zurich, 2001, pp. 513-524.

(67) Ferguson, J. The use of chemical potentials as indices of toxicity.

Proc. R. Soc. London, Ser. B 1939, 127, 387–404.

(47) Davis, A. M.; Teague, S. J. Hydrogen bonding, hydrophobic

interactions, and failure of the rigid receptor hypothesis. Angew.

Chem., Int. Ed. 1999, 38, 736–749.

(48) DeLano, W. L.; Ultsch, M. H.; de Vos, A. M.; Wells, J. A.

Convergent solutions to binding at a protein-protein interface.

Science 2000, 287, 1279–1283.

(49) Davis, A. M.; Dixon, J.; Logan, C. J.; Payling, D. W. Accelerating

the progress of drug discovery. In Pharmacokinetic Challenges in

Drug Discovery; Pelkonen, O., Baumann, A., Reichel, A., Eds.;

Springer: New York, 2002; pp 1-32, E. Schering Research Foundation,

Workshop 37 .

(50) de Graaf, C.; Foata, N.; Engkvist, O.; Rognan, D. Molecular

modeling of the second extracellular loop of G-protein-coupled

receptors and its implication on structure-based virtual screening.

Proteins 2008, 71, 599–620.

(68) Bigge, C. F.; Nikam, S. S. AMPA receptor agonists, antagonists

and modulators: their potential for clinical utility. Exp. Opin. Ther.

Patents 1997, 7, 1099–1114.

(69) Zhi, L.; Tegley, C. M.; Marschke, K. B.; Jones, T. K. Switching

androgen receptor antagonists to agonists by modifying C-ring

substituents on piperidino[3,2-g] quinolinone. Bioorg. Med. Chem.

Lett. 1999, 9, 1009–1012.

(70) De Lucca, G. V.; Kim, U. T.; Johnson, C.; Vargo, B. J.; Welch, P.

K.; Covington, M.; Davies, P.; Solomon, K. A.; Newton, R. C.;

Trainor, G. L.; Decicco, C. P.; Ko, S. S. Discovery and structure-

activity relationship of N-(ureidoalkyl)-benzyl-piperidines as po-

tent small molecule CC chemokine receptor-3 (CCR3) antagonists.

J. Med. Chem. 2002, 45, 3794–3804.

(71) Møller, C.; Plesset, M. S. Note on an approximation treatment for

many-electron systems. Phys. Rev. 1934, 46, 618–622.

Article

Journal of Medicinal Chemistry, 2009, Vol. 52, No. 23 7723

(72) Sæbø, S.; Pulay, P. Local treatment of electron correlation. Annu.

Rev. Phys. Chem. 1993, 44, 213–236.

(73) Sæbø, S.; Tong, W.; Pulay, P. Efficient elimination of basis-set-

superposition errors by the local correlation method: accurate ab

initio studies of the water dimer. J. Chem. Phys. 1993, 98, 2170–

2175.

(74) Castonguay, L. A.; Weng, Y.; Adolfsen, W.; Di Salvo, J.; Kilburn,

R.; Caldwell, C. G.; Daugherty, B. L.; Finke, P. E.; Hale, J. J.;

Lynch, C. L.; Mills, S. G.; MacCoss, M.; Springer, M. S.; DeMar-

tino, J. A. Binding of 2-aryl-4-(piperidin-1-yl)butanamines and

1,3,4-trisubstituted pyrrolidines to human CCR5: a molecular

modeling-guided mutagenesis study of the binding pocket. Bio-

chemistry 2003, 42, 1544–1550.

knowledge and strategies for the early prediction during drug

development. Med. Res. Rev. 2005, 25, 133–166.

(83) Waring, M. J.; Johnstone, C. A quantitative assessment of hERG

liability as a function of lipophilicity. Bioorg. Med. Chem. Lett.

2007, 17, 1759–1764.

(84) Fernandez, D.; Ghanta, A.; Kauffman, G. W.; Sanguinetti, M. C.

Physicochemical features of the hERG channel drug binding site. J.

Biol. Chem. 2004, 279, 10120–10127.

(85) Perry, M.; Stansfeld, P. J.; Leaney, J.; Wood, C.; de Groot, M. J.;

Leishman, D.; Sutcliffe, M. J.; Mitcheson, J. S. Drug binding

interactions in the inner cavity of HERG channels: molecular

insights from structure-activity relationships of clofilium and

ibutilide analogs. Mol. Pharmacol. 2006, 69, 509–519.

(75) Tsamis, F.; Gavrilov, S.; Kajumo, F.; Seibert, C.; Kuhmann, S.;

Ketas, T.; Trkola, A.; Palani, A.; Clader, J. W.; Tagat, J. R.;

McCombie, S.; Baroudy, B.; Moore, J. P.; Sakmar, T. P.; Dragic,

T. Analysis of the mechanism by which the small-molecule CCR5

antagonists SCH-351125 and SCH-350581 inhibit human immu-

nodeficiency virus type 1 entry. J. Virol. 2003, 77, 5201–5208.

(76) de Mendonca, F .L.; da Fonseca, P. C.; Phillips, R. M.; Saldanha, J.

W.; Williams, T. J.; Pease, J. E. Site-directed mutagenesis of CC

chemokine receptor 1 reveals the mechanism of action of UCB

35625, a small molecule chemokine receptor antagonist. J. Biol.

Chem. 2005, 280, 4808–4816.

(77) Seibert, C.; Ying, W.; Gavrilov, S.; Tsamis, F.; Kuhmann, S. E.;

Palani, A.; Tagat, J. R.; Clader, J. W.; McCombie, S. W.; Baroudy,

B. M.; Smith, S. O.; Dragic, T.; Moore, J. P.; Sakmar, T. P.

Interaction of small molecule inhibitors of HIV-1 entry with

CCR5. Virology 2006, 349, 41–54.

(78) Maeda, K.; Das, D.; Ogata-Aoki, H.; Nakata, H.; Miyakawa, T.;

Tojo, Y.; Norman, R.; Takaoka, Y.; Ding, J.; Arnold, G. F;

Arnold, E.; Mitsuya, H. Structural and molecular interactions of

CCR5 inhibitors with CCR5. J. Biol. Chem. 2006, 281, 12688–

12698.

(79) Vaidehi, N.; Schlyer, S.; Trabanino, R. J.; Floriano, W. B.; Abrol,

R.; Sharma, S.; Kochanny, M.; Koovakat, S.; Dunning, L.; Liang,

M.; Fox, J. M.; de Mendonca, F. L.; Pease, J. E.; Goddard, W. A.,

III; Horuk, R. Predictions of CCR1 chemokine receptor structure

and BX 471 antagonist binding followed by experimental valida-

tion. J. Biol. Chem. 2006, 281, 27613–27620.

(86) Siebrands, C. C.; Friederich, P. Structural requirements of human

ether-a-go-go-related gene channels for block by bupivacaine.

Anesthesiology 2007, 106, 523–531.

(87) Kamiya, K.; Niwa1, R.; Morishima1, M.; Honjo1, H.; Sanguinetti,

M. C. Molecular determinants of hERG channel block by terfe-

nadine and cisapride. J. Pharmacol. Sci. 2008, 108, 301–307.

ꢀ

(88) Afzelius, L.; Masimirembwa, C. M.; Karlen, A.; Andersson, T. B.;

Zamora, I. Discriminant and quantitative PLS analysis of compe-

titive CYP2C9 inhibitors versus noninhibitors using alignment

independent GRIND descriptors. J. Comput.-Aided Mol. Des.

2002, 16, 443–458.

(89) Stansfeld, P. J.; Sutcliffe, M. J.; Mitcheson, J. S. Molecular

mechanisms for drug interactions with hERG that cause long QT

syndrome. Expert Opin. Drug Metab. Toxicol. 2006, 2, 81–94.

(90) Needham, M.; Sturgess, N.; Cerillo, G.; Green, I.; Warburton, H.;

Wilson, R.; Martin, L.; Barratt, D.; Anderson, M.; Reilly, C.;

Hollis, M. Monocyte chemoattractant protein-1: receptor interac-

tions and calcium signaling mechanisms. J. Leukoc. Biol. 1996, 60,

793–803.

(91) Springthorpe, B.; Strandlund, G. Radiolabeled bispidine com-

pound and assay for I_Krpotassium channel blocking activity.

PCT Int. Appl. WO2005037052, 2005, 22 pp.

(92) Murphy, S. M.; Palmer, M.; Poole, M. F.; Padegimas, L.; Hunady,

K.; Danzig, J.; Gill, S.; Gill, R.; Ting, A.; Sherf, B.; Brunden, K.;

Stricker-Krongrad, A. Evaluation of functional and binding assays

in cells expressing either recombinant or endogenous hERG chan-

nel. J. Pharm. Toxicol. Methods 2006, 54, 42–55.

(80) Pearlstein, R. A.; Vaz, R. J.; Kang, J.; Chen, X.-L.; Preobrazhens-

kaya, M.; Shchekotikhin, A. E.; Korolev, A. M.; Lysenkova, L. N.;

Miroshnikova, O. V.; Hendrix, J.; Rampe, D. Characterization of

HERG potassium channel inhibition using CoMSiA 3D QSAR

and homology modelling approaches. Bioorg. Med. Chem. Lett.

2003, 13, 1829–1835.

(81) Aronov, A. M.; Goldman, B. B. A model for identifying hERG Kþ

channel blockers. Bioorg. Med. Chem. 2004, 12, 2307–2315.

(82) Recanatini, M.; Poluzzi, E.; Masetti, M.; Cavalli, A.; De Ponti, F.

QT prolongation through hERG Kþ channel blockade: current

(93) Donovan, S. F.; Pescatore, M. C. Method for measuring the

logarithm of the octanol-water partition coefficient by using short

octadecyl-poly(vinyl alcohol) high-performance liquid chromatog-

raphy columns. J. Chromatogr. A 2002, 952, 47–61.

(94) Box, K.; Elliott, S.; Cimpan, G. High throughput pKa and logD7.4

measurements. PharmaChem 2003, 2, 55–59.

(95) Zhou, Y.; MacKinnon, R. The occupancy of ions in the K^þ

selectivity filter: Charge balance and coupling of ion binding to a

protein conformational change underlie high conduction rates.

J. Mol. Biol. 2003, 333, 965–975.

Article Doi

Increasing selectivity of CC chemokine receptor 8 antagonists by engineering nondesolvation related interactions with the intended and off-target binding sites

DOI: 10.1021/jm900713y

Source and publish data:

Authors:

Article abstract of DOI:10.1021/jm900713y

Full text of DOI:10.1021/jm900713y

Products guided by the article

R&D Labs maybe for 1179356-28-5

Relevant to this article

Hot Product