Structure-based exploration and pharmacological evaluation of N-substituted piperidin-4-yl-methanamine CXCR4 chemokine receptor antagonists

Article abstract of DOI:10.1016/j.ejmech.2018.10.060

Using the available structural information of the chemokine receptor CXCR4, we present hit finding and hit exploration studies that make use of virtual fragment screening, design, synthesis and structure-activity relationship (SAR) studies. Fragment 2 was identified as virtual screening hit and used as a starting point for the exploration of 31 N-substituted piperidin-4-yl-methanamine derivatives to investigate and improve the interactions with the CXCR4 binding site. Additionally, subtle structural ligand changes lead to distinct interactions with CXCR4 resulting in a full to partial displacement of CXCL12 binding and competitive and/or non-competitive antagonism. Three-dimensional quantitative structure-activity relationship (3D-QSAR) and binding model studies were used to identify important hydrophobic interactions that determine binding affinity and indicate key ligand-receptor interactions.

Full text of DOI:10.1016/j.ejmech.2018.10.060

Accepted Manuscript
Structure-based exploration and pharmacological evaluation of N-Substituted
piperidin-4-yl-methanamine CXCR4 chemokine receptor antagonists
I. Adlere, S. Sun, A. Zarca, L. Roumen, M. Gozelle, C. Perpiñá Viciano, B. Caspar, M.
Arimont, J.P. Bebelman, S.J. Briddon, C. Hoffmann, S.J. Hill, M.J. Smit, H.F. Vischer,
M. Wijtmans, C. de Graaf, I.J.P. de Esch, R. Leurs
PII:
S0223-5234(18)30937-1
DOI:
https://doi.org/10.1016/j.ejmech.2018.10.060
EJMECH 10846
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 12 September 2018
Revised Date: 23 October 2018
Accepted Date: 27 October 2018
Please cite this article as: I. Adlere, S. Sun, A. Zarca, L. Roumen, M. Gozelle, C.Perpiñá. Viciano, B.
Caspar, M. Arimont, J.P. Bebelman, S.J. Briddon, C. Hoffmann, S.J. Hill, M.J. Smit, H.F. Vischer, M.
Wijtmans, C. de Graaf, I.J.P. de Esch, R. Leurs, Structure-based exploration and pharmacological
evaluation of N-Substituted piperidin-4-yl-methanamine CXCR4 chemokine receptor antagonists,
European Journal of Medicinal Chemistry (2018), doi: https://doi.org/10.1016/j.ejmech.2018.10.060.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Structure-Based Exploration and Pharmacological Evaluation of N-Substituted
Piperidin-4-yl-methanamine CXCR4 Chemokine Receptor Antagonists
†
,a
†,b
†,b
b
b,c
d,e
f
b
I. Adlere, S. Sun, A. Zarca, L. Roumen, M. Gozelle, C. Perpiñá Viciano, B. Caspar, M. Arimont, J.
b
f
d,e
f
b
b
b
P. Bebelman, S. J. Briddon, C. Hoffmann, S. J. Hill, M. J. Smit, H. F. Vischer, M. Wijtmans, C. de
b
a,b
a,b
Graaf, I. J. P. de Esch, R. Leurs*
a
Griffin Discoveries BV, Department of Medicinal Chemistry, De Boelelaan 1083, 1081 HV Amsterdam, The
Netherlands
b
Division of Medicinal Chemistry, Amsterdam Institute for Molecules, Medicines and Systems, Faculty of Science, Vrije
Universiteit Amsterdam, De Boelelaan 1108, 1081 HZ Amsterdam, The Netherlands
c
Department of Pharmaceutical Chemistry, Faculty of Pharmacy, Gazi University, 06560, Ankara, Turkey
d
Institute for Molecular Cell Biology, CMB-Center for Molecular Biomedicine, University Hospital Jena, Friedrich-
Schiller University Jena, Hans-Knöll-Strasse 2, 07745, Jena, Germany
e
Institute of Pharmacology and Toxicology, University of Würzburg, Versbacher Str. 9, 97078, Würzburg, Germany
f
Division of Pharmacology, Physiology and Neuroscience and Centre of Membrane Proteins and Receptors
(
COMPARE), School of Life Sciences, University of Nottingham, Nottingham, U.K. NG7 2UH
*
Corresponding author.
E-mail address: r.leurs@vu.nl (R. Leurs)
†
Author Contributions Ilze Adlere, Shan-Liang Sun and Aurélien Zarca contributed equally to this work.
1

ACCEPTED MANUSCRIPT
ABSTRACT
Using the available structural information of the chemokine receptor CXCR4, we present hit finding and hit
exploration studies that make use of virtual fragment screening, design, synthesis and structure-activity
relationship (SAR) studies. Fragment 2 was identified as virtual screening hit and used as a starting point for
the exploration of 31 N-substituted piperidin-4-yl-methanamine derivatives to investigate and improve the
interactions with the CXCR4 binding site. Additionally, subtle structural ligand changes lead to distinct
interactions with CXCR4 resulting in a full to partial displacement of CXCL12 binding and competitive and
or non-competitive antagonism. Three-dimensional quantitative structure-activity relationship (3D-QSAR)
and binding model studies were used to identify important hydrophobic interactions that determine binding
affinity and indicate key ligand-receptor interactions.
KEYWORDS
G protein-coupled receptors; CXCR4 chemokine receptor; antagonists; structure-based fragment virtual
screening; structure-activity relationship; 3D-QSAR.
2

ACCEPTED MANUSCRIPT
1
. Introduction
Chemokines and G protein-coupled chemokine receptors (GPCRs) play an important role in the immune
defense system by controlling the migration, activation, differentiation, and survival of
leukocytes.[1] Endogeneous chemokine proteins stabilize their cognate chemokine receptors in an active
conformation that facilitates intracellular signal transduction by interactions with G proteins and/or
arrestins.[1, 2] Because of their crucial role in the migration of immune cells, chemokine receptors are
promising drug targets for various immune-related diseases, including chronic obstructive pulmonary disease,
multiple sclerosis, rheumatoid arthritis, HIV-1 infection and cancer.[3, 4] Molecular pharmacology, medicinal
chemistry and molecular modeling studies have provided insights into molecular determinants of chemokine
receptor modulation by proteins, peptides, and small-molecule ligands.[1, 5] In the past few years, the first
high-resolution crystal structures of chemokine receptors have been solved and these have given detailed
structural information on the interaction of chemokine receptors and their ligands.[6] The crystal structures of
vMIP bound CXCR4[7], CCL5 bound CCR5[8], and CX3CL1 bound US28[9] complexes show how
chemokine ligands bind the N-terminal and extracellular loop regions of the receptor with their relatively
conserved C-terminal domains and target the orthosteric seven-transmembrane helical domain (TMD) with
their variable N-terminal regions.[5] Moreover, CCR2, CCR5 and CXCR4 crystal structures show how small-
molecule drug-like ligands (BMS-681, maraviroc, IT1t, Fig. 1) and medium sized peptidomimetic (CVX15)
target the TMD binding site (“ancestral” orthosteric binding site[10]) and block the binding of the chemokine
N-terminus.[6, 11, 12] Recent CCR2 and CCR9 crystal structures reveal that chemokine receptors may also
contain a conserved intracellular allosteric binding site overlapping with the G protein coupling site that can
be targeted by small drug-like ligands (CCR2-RA-[R], Vercirnon).[12-15] Despite the breakthroughs in the
elucidation of crystal structures of chemokine receptors, the computational prediction of receptor-ligand
interactions to guide structure-based ligand discovery is still facing several challenges. The large, open and
solvent accessible orthosteric TMD binding sites of chemokine receptors are challenging targets for structure-
based virtual ligand screening[5] compared to the more druggable, occluded binding sites of e.g. aminergic
3

ACCEPTED MANUSCRIPT
GPCRs.[16, 17] To effectively interact with these binding sites, most chemokine receptor ligands are
relatively large and/or hydrophobic, and contain multiple cationic centers to interact with conserved
negatively charged residues in chemokine receptors.
Hallmark chemokine receptor CXCR4 is activated by the endogeneous chemokine CXCL12 (also known as
stromal cell-derived factor-1, SDF-1α) and targeted by the antagonist plerixafor/AMD3100 (Fig. 1), the first
approved drug acting on chemokine receptors and used for stem cell mobilization.[18] The CXCR4 receptor
was the first chemokine receptor to be crystallised with small-molecule, peptide, and chemokine ligands and
provides an ideal system to investigate the possibilities and limitations of structure-based ligand design.[19,
2
0] Chemokine receptor modeling studies, including the community-wide GPCR DOCK 2010 assessment to
predict the three-dimensional coordinates of the IT1t and CVX15 bound CXCR4 crystal structures, have
identified several pitfalls associated with matching the interaction properties of chemokine receptor binding
sites and small molecule ligands.[21] Firstly, the possibilities to translate binding mode hypotheses between
chemokine receptors and/or ligand chemotypes is limited by: i) the symmetric distribution of anionic residues
2.63
4.60
6.58
7.39
in the receptor (e.g. D , D , D , E
in CXCR4) and complementary cationic centers in known tool
compounds (e.g. AMD3100, IT1t), ii) the existence of multiple orthosteric and allosteric small-molecule
binding pockets, and iii) the ligand dependent effects of receptor mutation studies.[5] Secondly, the structure-
based identification and optimization of chemokine receptor ligands is complicated by conformational
sampling of larger, flexible ligands and receptor binding sites as well as by defining effective scoring methods
for the prioritization of potential ligands based on their predicted interactions with solvent accessible receptor
binding sites.[5] Several potent small-molecule ligand classes, such as the ones exemplified by IT1t and
AMD3100, have been identified for CXCR4 (Fig. 1)[18, 22-29] Virtual screening campaigns to discover
novel CXCR4 ligands mostly yielded high micromolar binding affinities (IC , K )[30, 31] or no measurable
5
0
i
binding affinity in radiolabeled chemokine displacement studies[32, 33] and, considering the low ligand
efficiency (delta free energy of binding divided by the number of heavy atoms[34]) of these hits, the potential
for successful optimization was not evident. Considering the low LEs, it is no surprise that fragment-based
4

ACCEPTED MANUSCRIPT
approaches for peptidergic GPCRs such as chemokine receptors have so far been relatively scarce[5],
especially when compared to other GPCRs like adenosine and aminergic GPCRs, for which in silico fragment
screening and hit exploration was very successful [35, 36]. Starting point for our studies was a virtual
screening hit that contains an N-substituted piperidin-4-yl-methanamine core. Several piperidine-containing
CXCR4 ligand classes have been reported[30, 31], including AMD3100 derivatives[37], dual CCR5/CXCR4
inhibitors[38], benzenesulfonamides[39] and N-substituted benzimidazoles[40]. Here we used a fragment-
based approach that makes use of the CXCR4 structural information and molecular modelling studies to
complement the structure-activity relationship (SAR) studies during hit exploration.
Fig.1. Selected CXCR4 reference antagonists
2
. Results and discussion
2
.1. Structure-based virtual screening
We designed a structure-based virtual screening workflow focusing on the identification of small, fragment-
like molecules[41] and customized to experimentally supported[5] CXCR4 ligand interaction features (HB
2
.63
7.39
and ionic interactions with residues D97 and E288 ) (Fig. 2A). In the first step, a focused chemical library
was prepared containing fragment-like molecules (number of heavy atoms
≤ 22, logP < 3, number of H-bond
5

ACCEPTED MANUSCRIPT
donors
≤ 3, number of H-bond acceptors ≤ 3, number of rotatable bonds ≤ 5, number of rings ≥ 1) with two
basic centers, consistent with the conserved cationic pharmacophore features of IT1t and AMD3100 (Fig. 1)
2.63
4.60
6.58
7.39
and complementary to the negatively charged residues D97 , D171 , D262 , E288
that have been
shown to play a role in small-molecule ligand binding to CXCR4.[5] This focused virtual library of 52.500
fragment-like molecules with two cationic centers was docked in the CXCR4 crystal structure (PDB ID:
3
ODU)[6] using GOLD[42] and PLANTS[43] docking algorithms. Molecules that were able to
2.63
7.39
simultaneously form H-bond and ionic interactions with D97 and E288 were ranked according to their
GOLD (503 compounds) and PLANTS (1414 compounds) docking scores, as well as their structural
Interaction FingerPrint (IFP)[44] compared to the co-crystallized IT1t reference (Fig. 2B). The docking poses
of the top 200 ranked molecules were visually inspected, and molecules with polar groups docked in the
2
.60
3.33
previously identified hydrophobic hot spot between W94
novelty filter (ECFP-4< 0.4[46] as compared with any known CXCR4 ligands) resulted in a final selection of
4 fragment-like compounds, of which 23 commercially available compounds (specified in Fig. S1) were
and Y116 were discarded.[45] A structural
3
1
25
purchased and validated in I-CXCL12 binding studies.
125
Tested at 63 µM, four hits (1-4) showed more than 50% inhibition of I-CXCL12 binding to HEK293T cell
membranes transiently overexpressing human CXCR4 (Fig. 2C) and these were selected for further
evaluation. Fragments 2 and 3 share the same benzylpiperidin-4-yl-methanamine scaffold and fragment 3 was
therefore discarded from further validation. Fragment hit 4 holds a chiral center and can potentially form a
reactive quinone moiety and further fragment growing from this fragment was therefore deprioritized. The two
125
remaining hits 1 and 2 were subsequently tested for concentration-dependent inhibition of I-CXCL12
binding to hCXCR4 (IC , Table 1), resulting in a better pIC value (5.0) for fragment 2 than for 1.
5
0
50
6

ACCEPTED MANUSCRIPT
Fig. 2. Overview of the structure-based (SB) ligand virtual screen and design (A-G).
(
A) Overview of the different steps in the SB virtual screening work flow. (B) Compound IT1t (green stick)
binding to CXCR4 (yellow cartoon, PDB ID: 3ODU[6]). Key residues are shown as grey sticks and protein-
based (PB) hydrophobic hot spots are shown in transparent grey surface. (C) Single concentration (63 µM)
binding studies of 23 commercially available SBVS hit analogues and the structures of four hits showing more
1
25
than 50% inhibition of I-CXCL12 binding to human CXCR4. (D-E) Two alternative binding modes of 2
(
magenta stick) binding to CXCR4. IT1t is shown in transparency as a reference. Key residues are shown in
grey stick and PB hot spots are shown in transparent grey surface. (F) Comparative structural interaction
fingerprint (IFP)[44] analysis of binding modes of IT1t and 2. The structural receptor−ligand interaction
patterns are described by IFP bit strings encoding different interaction types between the ligand and receptor
CXCR4 amino acid residues. (G) Schematic illustration of SAR exploration of N-substituted piperidin-4-yl-
methanamines.
7

ACCEPTED MANUSCRIPT
Docking studies of 2 into the X-ray structure of hCXCR4 (PDB ID: 3ODU[6]) suggest two alternative binding
2.63
7.39
modes (Fig. 2D, E), which both include ionic and H-bond interactions with D97 and E288 , consistent
with the binding mode of IT1t in the CXCR4 crystal structure (Fig. 2B).[6] In the binding mode 1, compound
2
accommodates its chlorinated phenyl group in the hydrophobic hot spot of CXCR4 between TM helices 1-3
and 7[5, 45] (Fig. 2D), whereas in binding mode 2 the chlorinated phenyl group is directed towards the major
binding pocket of CXCR4 between TM helices 3-7 (Fig. 2E). Structural Interaction FingerPrint (IFP)
analysis[47] of IT1t and these two poses of compound 2 (Fig. 2F) shows shared interactions with key residues
2
.60
2.63
3.32
7.39
(
W94 , D97 , Y116 and E288 ). The two alternative binding mode hypotheses and structural analyses
were used to guide fragment growing studies to explore structure-activity relationships and improve the virtual
screening hit 2. The ensuing design strategy involved substitutions of varying chemical nature on both amine
moieties of the scaffold (Fig. 2G)
2
.2. Chemistry
The synthesis of the compounds based on 2 is outlined in Scheme 1. Compounds 6a,b,d-l were prepared in a
direct one-pot reductive amination of benzaldehydes and commercially available 4-(Boc-
aminomethyl)piperidine 5a or 4-(Boc-aminoethyl)piperidine 5b (in case of 6f) in the presence of
NaBH(OAc)
3
. Compound 6c was obtained by alkylation of amine 5a with 1-(chloromethyl)-2-methylbenzene
and K CO . Deprotection of 6a-l with HCl in dioxane, followed by a basic workup (except for compounds 7g
2
3
and 7h, which were isolated as hydrochloride salts) provided key building blocks 2 and 7b-l. The final
compound series 8-32 was obtained in a two-step reductive amination of benzaldehydes and 2, 7d,g-l via
1
imine formation (followed by H NMR spectroscopy on isolated aliquots), following by reduction with NaBH4
in MeOH. Compounds 20, 21 and 29 retained traces of the benzylic alcohol (formed from the starting
benzaldehyde during NaBH treatment) even after acid/base workup and crystallization as fumarate salts
4
proved efficient to remove these impurities.
8

ACCEPTED MANUSCRIPT
a
Scheme 1. Synthesis of Small-Molecule CXCR4 ligands
a
1
Reagents and conditions: (a) NaBH(OAc) , DCE, R CHO (2-Cl-C H -CHO for 6f), rt, 17 h–6 d, 33–98%; for 6c: 1-
3
6
4
(
chloromethyl)-2-methylbenzene, K CO , EtOH, reflux, 3h, 80%; (b) (i) 4 M HCl/dioxane, rt, 1–3 h; (ii) basic extraction,
2 3
2
5
8–99% (7g and 7h isolated as dihydrochloride salts); (c) (i) R CHO, anhydrous Na SO , when using 7g and 7h: TEA,
2 4
DCM, rt, 24 h–5 d; (ii) NaBH , MeOH, rt, 3–30 min, 46–96%; for 21, 22 and 27 (iii) fumaric acid, 2-PrOH, rt, 2–24 h,
4
3
8–52%.as fumarate salt.
9

ACCEPTED MANUSCRIPT
2
.3. Structure-activity relationship
We synthesised and evaluated a variety of analogues of hit fragment 2. As depicted in table 1, the left-hand
1
2
ring of the scaffold bearing substituent R and the right-hand ring with R substitution are assigned as the A-
ring and B-ring, respectively. To evaluate the binding affinity, displacement assays were performed in which
125
I-CXCL12 binding to human CXCR4 was displaced by the ligands at multiple concentrations (Table 1). As
partial or no displacement of CXCL12 binding by small-molecule CXCR4-binding ligands is a known
phenomenon,[38] we also monitored the extent of displacement at 100 µM concentration of a ligand (Table 1).
To assess the relative contributions of the different chemical modifications to CXCR4 binding affinity, we
monitored the ligand efficiency (LE) and ligand-lipophilic efficiency (LLE) metrices (Table 1).[48] We first
explored a small series of analogues in which the 2-chlorophenyl moiety of 2 was varied (7b-e) to evaluate the
1
effect of the ring substituent R . Comparing the SBVS hit 2 (pIC = 5.0) and its derivatives 7b-e (pIC < 5),
50
50
the o-chlorophenyl moiety shows the best results. Elongating the chain between the piperidine and the NH
2
group (7f) did not improve binding affinity. Considering ligand binding mode variability associated with the
symmetric di-cationic pharmacophore[49] and chemical elaborations[50] of the central scaffold, we continued
to probe the A-ring while appending several simple benzyl-type B-rings (8-13). Compounds 8 and 9 failed to
show good affinity (pIC50
≤ 5), indicating the possible requirement for a lipophilic substitution on the A-ring.
The o-methoxy analogue 10 (pIC = 5.6) gave a modest increase in affinity with respect to 8, which could be
50
further enhanced by a m-methyl or m-ethyl substituent on the B-ring (11, 12). However, as observed in the
analogues without B-ring, the affinity of o-chlorophenyl analogue 13 (pIC = 6.6) was superior as it was 10-
5
0
fold higher than that for o-methoxy substituted compound 10, indicating a key overall contribution of the o-
125
chlorophenyl substituent to the binding affinity. Compound 13 showed full displacement of I-CXCL12
(
Table 1, Fig. 3). Further exploration kept the o-chlorophenyl group in place and was dedicated to explore the
preferred nature and substitutions of the B-ring. Replacing the phenyl B-ring in 13 with polar rings such as
pyridine (14) or imidazole (15) resulted in reduced affinity (pIC50 = 5.7 and 6.0, respectively). Yet, both
125
compounds displayed relatively high (89 and 96 %) displacement of I-CXCL12. The introduction of a
1
0

ACCEPTED MANUSCRIPT
cyclohexyl ring (16, pIC = 5.5) resulted in 12-fold decrease in affinity compared to 13. A 2,3-dichlorophenyl
50
substituent (17) displayed lower affinity (pIC = 5.6) and a remarkable loss of maximal displacement (13%)
50
125
of
I-CXCL12 binding to hCXCR4. Derivatives with oxygen-based groups such as p-methoxy, m,p-
methylenedioxy or p-hydroxy (18-20) showed moderate affinity and displacement, presenting no
125
improvement with respect to 13. Interestingly, the results for 20 (pIC50 = 6.5 and 98% of I-CXCL12
displacement) contrast sharply to those of 9 (with a p-OH on the A-ring), indicating possible favourable
interactions involving hydrogen bonding in the B-ring.
We also explored the impact of the size of the B-ring moiety by introducing a bulky naphthyl (21) or biphenyl
(
22) moiety. Both compounds showed similar affinity for CXCR4 (pIC = 6.3). It is noted that the biphenyl
50
125
analogue 22 fully displaces I-CXCL12 binding to CXCR4 (Table 1, Fig. 3). Interestingly, only a selection
of the compounds in table 1 show a full displacement of the chemokine radioligand, most notably 13, 15, 20
and 22. These ligands show reasonable diversity in the B-ring while other close analogues do not fully
displace the radioligand. This shows that the very subtle pharmacological differences cannot be explained by
SAR or by molecular modelling (vide infra). A focused positional scan of the B-ring with either a Cl- or
methyl-moiety was undertaken (23-28). All six analogues showed slightly lower level of displacement (69-80
%
) compared to the unsubstituted analogue 13 (95 %). The p-chloro (23) and p-methyl (24) analogues show a
decrease in binding affinity and LLE. The o-chloro (25), o-methyl (26) and m-chloro (27) substituted
analogues possess comparable affinities (pIC = 6.7, 6.6 and 6.5, respectively) to 13. Encouragingly, the m-
50
125
methyl analogue (28) shows a pIC50 value of 6.8 with, however, a partial displacement (63%) of I-CXCL12
binding (Table 1, Fig. 3). Substitution on the meta position on the B-ring was deemed preferred within the o-
chlorosubstituted A-ring series. To re-examine the role of the position of the chlorine substituent on the A-ring
with a meta-methyl substituted B-ring, we synthesised positional analogues of 28 (29, 30) as well as selected
dichloro derivatives (31, 32). The loss of affinity for the both m-chloro (29) and p-chloro (30) substituted
analogues (pIC50 5.6 and 5.0, respectively) confirms an important role for the ortho substitution of
chlorophenyl group. The results also revealed that a 2,3-disubstituted dichloro analogue (31) is less potent
1
1

ACCEPTED MANUSCRIPT
(
pIC = 6.1) compared to 28, whereas the 2,6-disubstituted isomer (32) is equipotent to 28. However, both
50
disubstituted analogues possess lower LLE (0.28 and 0.67) compared to 28 (LLE = 1.59) due to the increased
lipophilicity.
Table 1
125
Binding affinity, level of inhibition of I-CXCL12 binding and efficiency metrics for SBVS fragment hits
and improved ligands
125
I-CXCL12
1
2
a
c
d
Compound
R (A-ring)
R (B-ring)
pIC50
displacement,
clogP LE LLE
b
%
CXCL12
AMD3100
IT1t
-
-
-
-
-
-
-
-
9.3 ± 0.1
6.7 ± 0.1
8.0 ± 0.0
97 ± 0
-
-
-
e
98 ± 3
-0.25 0.25 6.78
5.39 0.42 2.61
3.48 0.36 0.74
100 ± 2
e
e
1
< 5
70 ± 3
e
e
2
H
H
5.0 ± 0.0
69 ± 2
2.51 0.43 2.49
3.11 0.38 1.60
f
7
b
< 5
67 ± 6
f
7
c
H
H
< 5
< 5
81 ± 4
2.25 0.40 2.39
1.72 0.39 3.11
f
7
d
67 ± 4
1
2

ACCEPTED MANUSCRIPT
f
f
7
e
H
H
< 5
59 ± 3
56 ± 6
2.49 0.40 1.94
3.04 0.35 1.28
3.98 0.31 1.02
3.31 0.27 1.17
3.89 0.32 1.74
4.39 0.33 1.67
4.92 0.32 1.14
4.69 0.39 1.81
3.19 0.34 2.50
7
f
< 5
8
5.0 ± 0.1
< 5
72 ± 2
f
9
65 ± 2
85 ± 1
84 ± 3
81 ± 1
87 ± 4
89 ± 1
1
0
1
2
3
4
5.6 ± 0.1
6.1 ± 0.1
6.1 ± 0.0
1
1
1
1
g
6.5 ± 0.1
5.7 ± 0.2
1
5
6.0 ± 0.1
5.5 ± 0.1
96 ± 1
77 ± 4
2.52 0.38 3.53
5.31 0.33 0.21
16
1
7
5.6 ± 0.3
13 ± 11
5.99 0.31 -0.40
1
1
8
9
5.9 ± 0.1
6.6 ± 0.2
88 ± 2
72 ± 3
4.61 0.32 1.29
4.65 0.35 1.97
1
3

ACCEPTED MANUSCRIPT
2
0
6.5 ± 0.2
6.3 ± 0.1
98 ± 1
4.02 0.37 2.46
5.86 0.32 0.40
h
2
1
2
78 ± 1
98 ± 1
h
2
6.3 ± 0.0
6.58 0.30 -0.27
2
3
6.2 ± 0.2
6.0 ± 0.2
80 ± 4
62 ± 1
5.40 0.35 0.81
5.19 0.34 0.82
2
2
4
5
6.7 ± 0.2
6.6 ± 0.1
80 ± 4
72 ± 5
5.40 0.38 1.31
5.14 0.38 1.50
2
6
h
2
7
6.5 ± 0.1
6.8 ± 0.1
65 ± 7
63 ± 1
5.40 0.37 1.07
5.19 0.39 1.59
2
2
3
3
8
9
0
1
5.6 ± 0.1
5.0 ± 0.1
6.1 ± 0.1
86 ± 2
86 ± 1
74 ± 2
5.19 0.32 0.38
5.19 0.28 -0.19
5.78 0.33 0.28
3
2
6.6 ± 0.1
60 ± 3
5.90 0.36 0.67
1
4

ACCEPTED MANUSCRIPT
a
125
Measured as competition of
I-CXCL12 (50 pM) binding to hCXCR4 expressed in membranes of transiently
transfected HEK293T cells. pIC values are means ± SEM (N = 3 with each experiment performed in triplicate).
5
125
0
b
Percentage displacement of I-CXCL12 (50 pM) in a presence of the ligand (100 µM) relative to IT1t (100 µM, 100
).
Ligand efficiency LE = ∆G/HA = (– RT ln(IC50))/HA, where R = 8.31447215 J/(K mol), T = 298.15 K, 1 kcal = 4184 J,
%
c
HA = number of non-hydrogen atoms in molecule.
d
Ligand-lipophilicity efficiency LLE = pIC50 – clogP, where clogP is calculated logP value of a compound and logP is
the logarithm of the partition coefficient of the compound between n-octanol and water log(coctanol/cwater).[51]
e
125
Measured as competition of
I-CXCL12 (40 pM) binding to hCXCR4 expressed in membranes of transiently
transfected HEK293T cells. pIC50 values are means ± SEM (N = 3 with each experiment performed in triplicate).
Percentage displacement calculated in a presence of the ligand (63 µM) relative to IT1t (63 µM, 100 %).
f
Full inhibition could not be achieved due to pIC < 5. The shown value is the percentage of inhibition detected at 100
5
0
µM.
g
pIC50 value is mean ± SEM (N = 9 with each experiment performed in triplicate).
Isolated and tested as fumarate salts
h
125
Fig. 3. (A) Inhibition of I-CXCL12 binding to hCXCR4 expressed in HEK293T membranes by compounds
3, 22 and 28, and reference ligands IT1t and CXCL12. Representative curves are shown. Experiments were
performed N 3 with each experiment performed in triplicate and mean values ± SEM are shown in Table 1.
B) The concentration-response curves for displacement of CXCL12-red binding to NLuc-tagged CXCR4 by
1
≥
(
selected ligands 13, 22 and 28. Curves are normalized to buffer (0%) and IT1t (100%). Experiments were
performed N = 3 with each experiment performed in triplicate and mean values ± SEM are shown in Table S1.
1
5

ACCEPTED MANUSCRIPT
2
.4. Pharmacology of key compounds
A concise set of key compounds (13, 22 and 28) was selected for further pharmacological analysis. Compound
3 displays the highest ligand efficiency (LE = 0.40) together with a good affinity (pIC = 6.6) and a full
1
50
125
displacement of I-CXCL12 binding to hCXCR4. o-Chloro substitution on the A-ring together with m-
phenyl (22) or m-methyl (28) substitution on the B-ring showed a positive effect on binding affinity (pIC =
5
0
1
25
6
6
.8 and 6.3, respectively) but a remarkably different level of maximal I-CXCL12 displacement (98 and
3%, respectively). Within this key set of three, the radioligand displacement results were found to correlate
with the results obtained from complementary NanoBRET binding measurements for the displacement of the
binding of fluorescently labelled CXCL12-red (25 nM) to NLuc-tagged CXCR4 by the key ligands (Fig. 3B).
The binding affinities and the displacement (%) values are combined in Table S1.
125
The different levels of I-CXCL12 displacement as observed for 22 and 28 indicate distinct interactions of
the two small molecules with CXCR4. Therefore, we assessed the antagonistic properties of the three ligands
(
13, 22 and 28) and the reference antagonist AMD3100 against CXCL12-induced CXCR4 activation. In the
presence of multiple (0-100 µM) concentrations of the ligand, AMD3100 and 13 (Fig. 4A, B) inhibit the
CXCL12-induced G protein activation by CXCR4 in a competitive manner, most likely indicating orthosteric
interaction with CXCL12. In contrast, compounds 22 and 28 both show non-competitive antagonistic effects
on CXCL12-induced CXCR4 activation (Fig. 4C, D). Interestingly, in the binding study (Fig. 3A) their effect
125
125
on the inhibition of I-CXCL12 binding to CXCR4 differ: compound 22 fully inhibits (98%) I-CXCL12
binding (related to IT1t = 100%), whereas 28 is a partial displacer showing 63 % inhibition (Table 1). Thus,
amongst the series of CXCR4 ligands, we have found both competitive and non-competitive antagonists
including full and partial displacers of CXCL12 binding.
1
6

ACCEPTED MANUSCRIPT
Fig. 4. Evaluation of the effect of key ligands on CXCR4-mediated G protein activation following CXCL12
binding. The concentration-response curves for CXCL12 were determined in the presence of various
concentrations of the ligands. G protein activation was measured by pre-incubation of HEK293T cells with
increasing concentration of a compound for 30 min followed by addition of CXCL12. Experiments were
performed N
≥ 3 with each experiment performed in quadruplicate. (A, B) Competitive behavior by reference
antagonist AMD3100 and compound 13. (C, D) Non-competitive behavior of compounds 22 and 28.
The set of key ligands together with positive control IT1t and the low-affinity ligand 9 as negative control
were evaluated in additional functional assays (
β-arrestin 2 and Inositol phosphate accumulation). CXCR4-
mediated G signalling in response to 10 nM CXCL12 was redirected to the phospholipase C – inositol
i
triphosphate (InsP ) pathway by co-expression of the chimeric G
protein (Fig. 5A), as previously
αq/i5
3
3
described.[52] Key compounds 13, 22 and 28 completely inhibited this CXCL12 induced InsP formation in
1
7

ACCEPTED MANUSCRIPT
concentration-dependent manner with comparable pIC values (Table 2). As expected, compound 9 did not
50
significantly inhibit CXCL12-induced signalling in this assay. In line, key compounds 13, 22 and 28 displayed
similar pIC50 values in inhibiting -arrestin 2 recruitment to hCXCR4 in response to 10 nM CXCL12 as
measured in a BRET-based assay (Fig. 5B and Table 2). Compound 9 had >10-fold lower pIC value, which
β
5
0
125
is in line with its lower ability to inhibit I-CXCL12 binding as compared to compounds 13, 22 and 28.
Taken altogether, these results demonstrate that despite the distinct displacement of CXC12 binding to
CXCR4 (Fig. 3A, B) and being either competitive or non-competitive antagonists (Fig. 4) of CXCL12
signalling via CXCR4, compounds 13, 22 and 28 can be functionally considered full antagonists of CXCR4
chemokine mediated signalling via both G proteins and β-arrestin2.
i
Fig. 5. Inhibition of CXCL12-induced CXCR4 activation by selected compounds. (A) Inhibition of CXCL12-
induced InsP accumulation in HEK293T cells co-expressing CXCR4 and chimeric Gαq/i5 proteins by
increasing concentration compounds. (B) Inhibition of -arrestin2 recruitment to CXCR4 in HEK293T cells in
3
β
response to 10 nM CXCL12 in the presence of increasing concentration compounds or reference IT1t. All
experiments were performed N = 3 with each experiment performed in triplicate and mean values ± SEM are
shown in Table 2.
1
8

ACCEPTED MANUSCRIPT
Table 2
Affinity and functional characterization of selected compounds
3
[
H]-Inositol phosphate
125
I-CXCL12 binding
β-arrestin 2 (BRET)
accumulation (IPx)
a
b
a
b
a
c
b
Compounds
pIC50
% displacement
65 ± 2
pIC50
< 4.5
% inhibition
pIC50
% inhibition
c
c
c
9
4.5 ± 0.3
6.5 ± 0.1
6.3 ± 0.0
6.8 ± 0.1
8.0 ± 0.0
N/A
< 4.5
N/A
1
2
2
3
2
8
87 ± 4
5.4 ± 0.0
5.6 ± 0.0
5.5 ± 0.0
7.3 ± 0.0
103 ± 1
102 ± 6
97 ± 5
5.7 ± 0.2
5.9 ± 0.1
6.0 ± 0.2
7.3 ± 0.0
93 ± 3
94 ± 3
87 ± 5
96 ± 4
98 ± 1
63 ± 1
IT1t
100 ± 2
100 ± 0
a
b
Results are means ± SEM (N ≥ 3 with each experiment performed in triplicate).
Results are expressed as percentage of inhibition of CXCL12 binding (50 pM)/signaling (10 nM) by ligand (100 µM)
with IT1t as reference (100 % inhibition).
c
pIC50 and percentage of inhibition could not be determined.
2
.5. CXCR4 structure-based SAR map
The experimentally determined pIC50 values were used to construct 3D-QSAR models in order to identify
ligand-based interaction hot spots and prioritize CXCR4-ligand binding mode models (Fig. 6). CXCR4
binding mode models of 28, based on the two initial binding modes proposed for the experimentally validated
virtual screening hit 2 (Fig. 2D,E) were refined by MD simulations, yielding two distinct ligand conformations
(
Fig. S2) that were used to build the 3D-QSAR models. Both reference ligand conformations provide
templates to construct predictive 3D-QSAR models with similar regression and predictive squared correlation
2
2
2
2
coefficients for model 1 (R =0.81, q = 0.76, Fig. 6A) and model 2 (R =0.80, q = 0.71, Fig. 6D). Figures 6B
and 6E show that both models are based on three hydrophobic hotspots defined by the GRID C1= probe[53,
5
4], including one LB interaction hotspot associated with chemical variations around the A-ring of 28 (LB hot
1
9

ACCEPTED MANUSCRIPT
spots 1.1 and 2.1), and two hotspots associated with variations around the B-ring of 28 (LB hot spots 1.2/1.3
and 2.2/2.3). We used the consistency between ligand-based and protein-based interaction models[49] as a
complementary criterion to compare ligand binding mode models 1 and 2 (Figs. 6C,F). The 3D-QSAR model
based on binding mode 2 provided a better match between the ligand-based (LB) interaction hot spots 2.2 and
2
.3 identified by the 3D-QSAR model (Fig. 6E) and the hydrophobic interaction hot spots identified in the
2
.60
3.28
3.29
3.32
receptor binding site, composed of hydrophobic residues W94 , V112 , H113 and Y116 (Fig. 6F).
This druggable binding site has indeed been postulated to involve binding of small-molecule ligands to
CXCR4 and other chemokine receptors.[5, 45] Two exemplary compounds 13 and 22 were selected for
binding mode comparison with co-crystallized ligand IT1t. This analysis shows that both compounds can form
2.63
7.39
ionic and hydrogen bond interactions with key residues D97 and E288 , and can target the hydrophobic
area. Compound 13 (Fig. 6G) lacks a methyl moiety which would be located around hot spot 2.2 and 2.3,
explaining the lower binding affinity of 13 compared to 28. However, compound 22 (Fig. 6H) with a
hydrophobic phenyl group also shows lower affinity, which might be explained by steric hindrance. The
described modeling method, matching ligand and protein interaction hotspots derived from experimentally
determined SAR data and molecular interaction field analyses, has previously been successfully applied to the
elucidation of experimentally validated structural protein-ligand interactions for histamine receptors.[49] The
current study demonstrates its applicability in structure-based ligand refinement for less druggable chemokine
receptors binding sites.
2
0

ACCEPTED MANUSCRIPT
Fig. 6. Details of three-dimensional quantitative structure-activity relationships (3D-QSAR) for Model 1 and
Model 2. (A and D) Plot of predicted versus experimental values (pIC ) of Model 1 and Model 2. (B)
5
0
Alignment of 31 compounds in model 1. Compound 28 is shown in cyan stick, while the others are shown in
grey line. The three ligand-based (LB) hot spots are shown in sphere. (C and F) LB 3D-QSAR model aligned
with protein-based (PB) hot spots and some key residues (grey stick). Compound 28 is shown in (C) cyan and
green (F) stick. Important binding residues are depicted as sticks with grey carbon atoms. Oxygen, nitrogen,
and hydrogen atoms are coloured red, blue and white, respectively. H-Bonds described in the text are depicted
by dashed lines. (E) Alignment of 31 compounds in model 2. Compound 28 is shown in green stick, while the
others are shown in grey line. The three LB hot spots are shown in spheres. (G, H) Plausible binding modes of
2
1

ACCEPTED MANUSCRIPT
compounds 13 (dark blue stick) and 22 (orange stick), respectively. Co-crystallized ligand IT1t is shown in
transparent magenta stick.
3
. Conclusions
The current studies explore a fragment-like CXCR4 hit that was identified by virtual fragment screening.
Ligand-based SAR studies were complemented by molecular modelling experiments, including docking and
3
D-QSAR studies. This resulted in models that indicate key ligand-receptor interactions. While the models
help to explain the affinity and antagonism of the ligands, the observed level of displacement of chemokine
CXCL12 binding can so far not be explained by the developed ligand-receptor models, indicating the
limitations of fragment-based ligand design to peptidergic GPCRs.
4
. Experimental
4
.1. Computational Methods
4
.1.1. Residue Numbering and Nomenclature. The Ballesteros–Weinstein residue numbering scheme[55] was
used throughout this manuscript. For explicitly numbered residues in specific receptors, the UniProt residue
7.39
number is given before the Ballesteros–Weinstein residue number in superscript (e.g., E288 in CXCR4).
4
.1.2. Preparation of the Virtual Screening Database. We downloaded commercially available compounds
from 8 trusted vendors from the ZINC8 database[56] in SMILES format and selected di-cationic 52.500
fragment-like compounds (number of heavy atoms 22, logP < 3, number of H-bond donors 3, number of
H-bond acceptors 3, number of rotatable bonds 5, number of rings ≥ 1) from this set[57, 58] using
Openeye’s filter tool[59]. We selected di-cationic compounds based on the experimentally supported binding
≤
≤
≤
≤
2
2

ACCEPTED MANUSCRIPT
2.63
7.39
mode hypothesis that ionic interactions with residues D97 and E288 play an important role in CXCR4
binding. The major protonation states of small molecules were computed with ChemAxon Calculators[60] at
pH 7.4 and converted to Mol2 format with Molecular Networks’ CORINA.[61]
4
.1.3. Automated Docking. CXCR4 crystal structure (PDB.: 3ODU) was prepared for docking simulations
using the MOE[62] Protonate3D module in order to ensure a plausible ionization state for each residue,
followed by visual inspection. Docking experiments were performed with the programs GOLD[42] and
PLANTS,[43] using the crystal structure of CXCR4 (3ODU).[6] PLANTS combines an ant colony
optimization algorithm with an empirical scoring function[63] for the prediction and scoring of binding poses
in a protein structure. GOLD is an automated ligand docking program that uses a genetic algorithm to explore
the full range of ligand conformational flexibilities with partial flexibility of the protein. For each compound,
1
5 poses were calculated, and scored by the ChemPLP scoring function at speed setting 2 in PLANTS. All
other options of PLANTS were left at their default setting. We performed 15 GA runs for each ligand in
GOLD and the population size was set to 100 (selection pressure 1.1, number of islands 3, maximum number
of operation per ligands 3000 and niches size 2); For flags, internal H-bonds and planar trigonal nitrogen
flipping were enabled, and restricted ligand conformational space by torsion angle distributions from CSD.
The genetic operators (pt_crosswt = 95, allele_mutatewt = 95, migratewt = 10) and other options were kept as
default. The docking poses were sorted by GoldScore fitness function. The binding pocket of CXCR4 was
defined by the coordinates of the center of co-crystallized IT1t in the 3ODU structure and a radius of 5 Å
(
which is the maximum distance from the center defined by a 5 Å radius around IT1t).
4
.1.4. IFP Post-processing. Structural interaction fingerprint analysis[44, 64, 65] was used for post-processing
of docking poses in structure-based virtual screening studies. The IT1t binding mode in the original CXCR4
X-ray structure[6] (PDB code 3ODU) was used to generate reference structural interaction fingerprints (IFPs)
as previously described.[44] Seven different interaction types (hydrophobic, aromatic face-to-edge, aromatic
face-to-face, H-bond acceptor, H-bond donor, negatively charged, and positively charged interactions) were
2
3

ACCEPTED MANUSCRIPT
used to define the IFP. The cavity used for the IFP analysis consisted of the same binding pocket used for
1.26
1.32
1.35
1.39
2.59
2.60
2.63
2.64
23.50
3.25
docking, including E32 , K38 , L41 , Y45 , F93 , W94 , D97 , A98 , W102 , C109
,
3.28
3.29
3.32
3.36
4.60
45.47
45.49
45.50
45.51
45.52
6.51
V112 , H113 , Y116 , L120 , D171 , R183 , I185 , C186 , D187 , R188 , Y255
,
7.31
7.35
7.39
7.42
H281 , S285 , E288 , F292 . Standard IFP scoring parameters, and a Tanimoto coefficient (Tc-
IFP)[44] measuring IFP similarity with the reference molecule pose (IT1t in the CXCR4 crystal structure
3
ODU, Fig. 2B), was used to filter and rank the docking poses of the 52.500 fragment-like compounds in the
2.63
virtual screening library. Only poses forming an H-bond and ionic interaction with residues D97
and
7.39
E288 were considered.
4
.1.5. Structure-Based Virtual Screening. The screening database was docked with PLANTS and GOLD, and
resulting docking poses were post-processed using IFP analysis and filtered for ionic and H-bond interactions
2
.63
7.39
with D97 and E288 . IFP (Tc
≥ 0.75) and PLANTS (≤ -90) scoring cut-offs derived from previously
GPCR structure-based virtual screening on H R[65] were used to select a total of 1.917 compounds. This set
1
was clustered and compared to known CXCR4 ligands in ChEMBL using ECFP-4 (extended connectivity
fingerprints)[66] descriptors available in KNIME analytics platform[67] and compared using the Tanimoto
coefficient. The docking poses of well-populated chemical clusters of hit molecules were visually inspected in
more detail, and those molecules that targeted the hydrophobic hot spot in the minor binding site were
prioritized. This yielded a final set of 34 hit molecules of which 23 were purchased and experimentally tested.
4
.1.6. MD simulations. Docking studies on compound 28 revealed two alternative binding models and both
2
.63
7.39
can target D97 and E288 simultaneously. The two distinguished models of the hit compound 28 bound
to CXCR4 were energy minimized for 1000 steps and used to run membrane-embedded MD simulations in
GROMACS.[68] Each system was simulated for 100 ns after an equilibration of 5 ns in which positional
restraints were gradually relaxed in order to allow lipids to properly adapt around the protein and to allow
water molecules to fill up receptor cavities. The trajectories were generated unrestrained with the parameters
and conditions described elsewhere[69]. The parameters of the ligands were obtained using the General
2
4

ACCEPTED MANUSCRIPT
Amber Force Field 2 (GAFF2) and AM1-BCC HF/6-31G* ESP fitted atomic charges[70] were used. Potential
energy, RMSD, RMSF, and dihedrals of the simulations were analyzed with GROMACS tools. The major
protonation state of the 31 small molecules were computed with ChemAxon’s Calculators[60] at pH 7.4.
4
.1.7. 3D-QSAR. The two refined 3D structures of compound 28 derived from MD simulations were used as
templates and other molecules were sketched and refined using MOE[71] as previously described. The MIF
probes (DRY and C1=) were then calculated using the GRID package (version 22 from Molecular
Discovery).[72] The probes in a radius of 5 Å around aligned compounds were calculated using a grid
resolution of 0.5 Å. The probes values were normalized, and probes with standard deviation of less than 1.0 or
correlation less than 0.3 were filtered out by employing R statistical package (version 2.7.1).[73] The Genetic
method followed by GreedyStepwise method from Weka 3.8 data-mining software package[74] were
subsequently used to automatically select the important probes and generate QSAR models, with dependent
variables being pIC of CXCR4.
50
4
4
.2. Pharmacology
2
.2.1. Cell Culture. Human embryonic kidney 293T cells (HEK293T) were grown at 37 °C and 5% CO in
Dulbecco’s modified Eagle medium (Gibco) supplemented with 10% fetal bovine serum (Bodinco), penicillin-
streptomycin (Gibco).
4
.2.2. CXCR4 Membrane Preparation. CXCR4-expressing HEK293T cell membranes were prepared as
6
previously described.[75] HEK293T cells (2⋅10 ) were seeded in a 10-cm dish and transfected the next day.
The medium of the cells was refreshed using 8 mL of culture medium. 5 µg of pcDEF -hCXCR4 was
3
combined with 40 µg of PEI in a total volume of 500 µL 150 mM NaCl and incubated for 20 minutes at room
temperature. Subsequently, the DNA/PEI mix was added to the cells. Two days after transfection, cells were
collected in ice-cold PBS and centrifuged at 1500 g for 10 min at 4°C. Subsequently, cells were washed with
2
5

ACCEPTED MANUSCRIPT
PBS and centrifuged at 1500 g for 10 min at 4°C. The pellet was resuspended in ice-cold membrane buffer (15
mM Tris pH 7.5, 1 mM EGTA, 0.3 mM EDTA, 2 mM MgCl ) and homogenized by 10 strokes at 1100–1200
2
rpm using a teflon-glass homogenizer and rotor. The membranes were subjected to two freeze thaw cycles
using liquid nitrogen and centrifuged at 40,000 g for 25 min at 4 °C. The pellet was resuspended in cold Tris-
sucrose buffer (20 mM Tris pH 7.4, 250 mM sucrose), and frozen in liquid nitrogen. Protein concentration
was determined using a BCA protein assay kit (Thermo Fisher).
125
4
.2.3. I-CXCL12 Binding Assay. CXCR4 membranes (5 µg/well) were incubated in 96-well clear plates
Greiner Bio One, PS, U-bottom, clear) in binding buffer (50 mM HEPES, pH 7.4, 1 mM CaCl2, 5 mM
MgCl2, 100 mM NaCl, and 1.0% (w/v) bovine serum albumin (BSA, fraction V)) with approximately 50 pM
(
125
I-CXCL12 (PerkinElmer) in the absence or presence of unlabeled ligands for 2 hours at 25 °C with gentle
agitation. The incubations were terminated by rapid filtration through Unifilter 96-well GF/C plates
PerkinElmer) presoaked with 0.5% PEI using ice-cold wash buffer (binding buffer supplemented with 0.5 M
NaCl) to separate free from bound radioligand. The filter plates were dried at 52 °C and 25 l Microscint-O
(
µ
was added. Bound radioactivity was quantified with a MicroBeta scintillation counter (PerkinElmer). Data
was analyzed using the GraphPad Prism 7 software. Non-linear regression curves were fitted using the
125
“
log(inhibitor) vs. response (three parameters)” equation. Percentage displacement of I-CXCL12 was
-5
calculated with controls present on each plate (10 M IT1t (Tocris) for determining non-specific binding: NS,
vehicle treated for determining total binding: TB) following this equation: (X-NS)/(TB-NS)x100.
4
.2.4. Bioluminescence Resonance Energy Transfer (BRET)
β-arrestin Recruitment Assay. 0.4 µg of pcDEF3-
hCXCR4-RLuc (as previously described)[76] and 1.6 µg pcDEF -β
3
-arrestin-2-mVenus (as previously
described)[77] plasmids were combined to 12 µg of PEI in a total volume of 250 µL 150 mM NaCl and
incubated for 20 minutes at room temperature. 1 million resuspended HEK293T cells were added to the
DNA/PEI mix, and cells were subsequently seeded (30,000 cells per well) on 96-well white plate (Greiner Bio
One, PS, F-bottom, white). Two days after transfection, culture medium was substituted with Hanks’ balanced
2
6

ACCEPTED MANUSCRIPT
salt solution (Gibco). Next, cells were pre-incubated in Hanks’ balanced salt solution with increasing
concentrations of compound for 60 minutes before stimulation with 10 nM CXCL12 and addition of 5 µM
Renilla Luciferase substrate coelenterazine-h (Promega). After 20 minutes, RLuc (480/20 nm) and BRET
(
540/40 nm) signals were measured on the Mithras LB940 (Berthold Technologies). BRET ratios were
calculated as BRET signal over RLuc signal, and fold over vehicle was determined using controls.
6
4
.2.5. Inositol Phosphate (IP) Accumulation Assay. HEK293T cells (2⋅10 ) were seeded in a 10-cm dish and
transfected the next day. The medium of the cells was refreshed using 8 mL of culture medium. 5 µg of DNA
including pcDEF3-CXCR4 and pcDNA1-HA-Gαq/i5[52] was combined with 40 µg of PEI in a total volume of
5
00 µL 150 mM NaCl and incubated for 20 minutes at room temperature. Subsequently, the DNA/PEI mix
3
was added to the cells. The next day, cells were transferred to (120⋅10 /well) a poly-L-lysine (Sigma) coated
4
8-wells plate and were incubated overnight in DMEM inositol-free medium (MP) supplemented with 1
3
µ
Ci/mL [ H]-myo-inositol (PerkinElmer). Cells were then treated with or without a dilution range of
4 2
antagonist in buffer (20 mM HEPES, 140 mM NaCl, 5 mM KCl, 1 mM MgSO , 1 mM CaCl , 10 mM D-(+)-
Glucose, pH 7.4) with 10 nM CXCL12 and 10 mM LiCl and 0.05% BSA for 1.5h at 37 °C. Cells were lysed
and the accumulated inositol phosphates (InsP3) were isolated using affinity purification columns (Bio-Rad).
The amount of radiolabeled IP was determined after the addition of a scintillation fluid (PerkinElmer) on a
Tri-Carb 2800TR (PerkinElmer).
4
.2.6. Fluorescence Resonance Energy Transfer (FRET) G protein Activation Assay. To test G protein
activation, the previously described Gαi1 FRET-based sensor and the untagged human CXCR4 receptor in
pcDEF3 was used.[78] The G protein sensor contains all three subunits of the G protein in a single plasmid:
the αi1 subunit fused to mTurquoise-
cp173Venus-G 2-IRES-G i1-mTurquoise2-
Wuerzburg (Wuerzburg, Germany) using Dulbecco’s Modified Eagle Medium (DMEM) supplemented with
.5 g/l glucose, 10% (v/v) fetal calf serum, 100 U/mL penicillin G, and 100ꢀµg/mL streptomycin sulphate and
∆
9, the
β
1 subunit and the
γ2 subunit fused to cp173Venus (pGβ1-2A-
γ
α
∆9). HEK293T cells were cultured at the University of
4
ꢀ
ꢀ
2
7

ACCEPTED MANUSCRIPT
L-glutamine (2 mM) at 37 °C and 7% CO . To investigate G protein activation, HEK293T cells were seeded
2
in 100 mm plates and allowed to grow until the cells reached 60-65% confluency. At this stage, cells were
transiently transfected with the Effectene transfection reagent (Qiagen), according to the manufacturer’s
instructions. For transfection, the following DNA amounts were used per plate: 1.4
µg of CXCR4 receptor and
3
µ
g of G i1 sensor. As a control, empty vector plasmid was used. 24 h after transfection, black 96 well
α
BRAND-plates (flat bottom) were coated with 90 µL poly-D-lysine (1 mg/mL) for 30 minutes. Next, poly-D-
lysine was aspirated and each well was washed once with 200 µL of PBS. Transfected HEK293T cells were
harvested by 2 min treatment with 1 mL trypsin solution and cells were resuspended in culture media and
counted. Cells were seeded at a density of 30,000 cells per well. On the day of the measurement, the medium
of the cells was removed and 90 µL of measuring buffer (140 mM NaCl, 5.4 mM KCl, 2 mM CaCl , 1 mM
2
MgCl , 10 mM HEPES, pH 7.3) was added to the cells and incubated at 37 °C during 30 min. Analysis of the
2
TM
cells was done 24h after seeding the cells in the 96-well plates using Synergy Neo2 Multi-Mode Microplate
TM
Reader (Biotek) with Gen5 Data Analysis Software. During the measurement, cells were excited at 420/50
nm (Biotek CFP-YFP Filter; 1035013) and emission was monitored at 485/20 nm and 540/25 nm (Biotek
CFP-YFP Filter; 1035043). The fluorescence was read during 5 minutes to determine the pre-read signal.
Following the pre-read measurement, 10 µL of increasing concentrations of CXCL12 was added to the wells
for a total assay volume of 100 µL. Fluorescence was read again during 20 minutes to determine the post-read
signal. During measurement, cells were kept at 37 °C. Data were analysed using the software GraphPad Prism
6
. To study the effect of the antagonists on G protein activation, the same procedure was applied, but modified
in the following way. Before the measurement, the test compounds, initially dissolved in DMSO, were diluted
in measuring buffer to reach a final assay concentration of 100 µM, 10 µM, 1 µM or 0.1 µM. Cells were pre-
incubated at 37 °C during 30 min with 90 µL of buffer containing the corresponding antagonist concentration.
After 5 min of reading, G protein activation was then stimulated as described above by adding an additional
1
0 µL solution of increasing concentrations of CXCL12 and measuring for additional 20 min. For each
antagonist, 3 to 5 repetitions were performed. To confirm that the different concentrations of DMSO do not
2
8