Determination of the binding mode for the cyclopentapeptide CXCR4 antagonist FC131 using a dual approach of ligand modifications and receptor mutagenesis

Abstract

Background and Purpose

The cyclopentapeptide FC131 (cyclo(-L-Arg¹-L-Arg²-L-2-Nal³-Gly⁴-D-Tyr⁵-)) is an antagonist at the CXC chemokine receptor CXCR4, which plays a role in human immunodeficiency virus infection, cancer and stem cell recruitment. Binding modes for FC131 in CXCR4 have previously been suggested based on molecular docking guided by structure–activity relationship (SAR) data; however, none of these have been verified by in vitro experiments.

Experimental Approach

Heterologous ¹²⁵I-12G5-competition binding and functional assays (inhibition of CXCL12-mediated activation) of FC131 and three analogues were performed on wild-type CXCR4 and 25 receptor mutants. Computational modelling was used to rationalize the experimental data.

Key Results

The Arg² and 2-Nal³ side chains of FC131 interact with residues in TM-3 (His¹¹³, Asp¹⁷¹) and TM-5 (hydrophobic pocket) respectively. Arg¹ forms charge-charge interactions with Asp¹⁸⁷ in ECL-2, while D-Tyr⁵ points to the extracellular side of CXCR4. Furthermore, the backbone of FC131 interacts with the chemokine receptor-conserved Glu²⁸⁸ via two water molecules. Intriguingly, Tyr¹¹⁶ and Glu²⁸⁸ form a H-bond in CXCR4 crystal structures and mutation of either residue to Ala abolishes CXCR4 activity.

Conclusions and Implications

Ligand modification, receptor mutagenesis and computational modelling approaches were used to identify the binding mode of FC131 in CXCR4, which was in agreement with binding modes suggested from previous SAR studies. Furthermore, insights into the mechanism for CXCR4 activation by CXCL12 were gained. The combined findings will facilitate future design of novel CXCR4 antagonists.

Tables of Links

TARGETS	LIGANDS
CCR2 chemokine receptor	AMD3100 (plerixafor)
CCR5 chemokine receptor	Aplaviroc
CXCR4 chemokine receptor	CXCL12
EBI2 (GPR183)	T140

This Table lists key protein targets and ligands in this article which are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Pawson et al., 2014) and are permanently archived in the Concise Guide to PHARMACOLOGY 2013/14 (Alexander et al., 2013).

Introduction

The chemokine receptor CXCR4 belongs to the class A 7-transmembrane helix (7TM) receptors, also known as GPCRs. It plays a role in human immunodeficiency virus (HIV) infection by being a cell-entry co-factor for T-cell tropic HIV strains (Berson et al., 1996; Feng et al., 1996), and is together with its endogenous agonist CXCL12, central for stem cell recruitment and cancer development, progression, and metastasis (Murphy et al., 2000; Balkwill, 2004; Bachelerie et al., 2014). These implications have facilitated the development of drug candidates targeting CXCR4 (Choi et al., 2012). One of these, the bicyclam compound AMD3100 (plerixafor), was approved in 2008 for haematopoietic stem cell mobilization in patients with multiple myeloma and non-Hodgkin's lymphoma (DiPersio et al., 2009a,b,; Micallef et al., 2009), although the initial indication was as an anti-HIV compound (De Clercq et al., 1994; Steen et al., 2009).

The polyphemusin II-derived peptides are a large class of CXCR4 antagonists that include the 14-mer T140 (Tamamura et al., 1998) and analogues, for example, CVX15 (Wu et al., 2010), as well as the cyclic pentapeptide FC131 (Figure 1A) and analogues (Fujii et al., 2003). The cyclopentapeptide CXCR4 antagonists were designed by combining the four most important residues of T140 (Arg², 2-Nal³, D-Tyr⁵ and Arg¹⁴) with a Gly residue to give a cyclopentapeptide library (Tamamura et al., 2000; Fujii et al., 2003). The optimal combination of sequence and stereochemistry was shown to be cyclo(-L-Arg¹-L-Arg²-L-2-Nal³-Gly⁴-D-Tyr⁵-), that is, FC131, which displays nanomolar affinity and potency at CXCR4 (Fujii et al., 2003) and serves as lead compound for development of more drug-like peptidomimetic CXCR4 antagonists. Structure–activity relationship (SAR) studies of FC131 have shown that although a positive charge is preferred at position 1, substitution of Arg¹ with the uncharged citrulline (Cit), or even less structurally related amino acids, is tolerated (Tamamura et al., 2005a; Demmer et al., 2011; Mungalpara et al., 2012). In contrast, Arg² is a crucial functionality and minor modifications abolish activity (Tamamura et al., 2005b; Demmer et al., 2011; Mungalpara et al., 2012). The aromatic residues in positions 3 (2-Nal³) and 5 (D-Tyr⁵) are also important for proper function; importantly, position 3 requires conservation as 2-Nal (Tamamura et al., 2005b; Mungalpara et al., 2013) while position 5 allows for some modifications (Tamamura et al., 2005b; Tanaka et al., 2009; Mungalpara et al., 2013).

Figure 1.

Compounds and mutations included in this study. (A) Structure of FC131 and (B) analogues [Cit¹]FC131, [Aib¹]FC131 and [D-Arg¹]FC131, for which only the structure of the modified side chain 1 is shown. (C) Helical wheel diagram of CXCR4 as seen from the extracellular side showing the upper halves of the TMs and parts of ECL-2. Residues with black background are conserved among class A GPCRs and residues on grey background were mutated in this study. (D) Surface expression and response to 0.1 μM CXCL12 in functional assay of each mutant.

Several computational models for binding of peptide antagonists to CXCR4 have been suggested. The first models for T140 (Trent et al., 2003) and FC131 (Våbenø et al., 2006a; Kawatkar et al., 2011) were based on the crystal structure of bovine rhodopsin (Palczewski et al., 2000); however, the subsequent publication of the crystal structures of CXCR4 (Wu et al., 2010) revealed significant structural differences between rhodopsin and CXCR4. The co-crystal complex between CXCR4 and CVX15, a 16-mer analogue of T140, also showed that peptide CXCR4 antagonists bind differently than previously suggested. Based on molecular docking to this crystal structure, guided by SAR data, more reliable binding models have since emerged for the cyclopentapeptide antagonists (Demmer et al., 2011; Kobayashi et al., 2012; Mungalpara et al., 2012; 2013,; Yoshikawa et al., 2012). These models collectively suggest an interaction between Arg¹ of FC131 and Asp¹⁸⁷ (in ECL-2) and Asp⁹⁷ in TM-2, while Arg² interacts with His¹¹³ (TM-3) and Asp¹⁷¹ (TM-4). [The position of residues according to the Baldwin/Schwartz (Schwartz, 1994; Baldwin et al., 1997) and the Ballesteros/Weinstein numbering system (Ballesteros and Weinstein, 1995) is given in the tables.] Furthermore, the 2-Nal³ side chain is located in a hydrophobic pocket facing TM-5, while D-Tyr⁵ is proposed to interact with either Glu³² in the N-terminus, Tyr⁴⁵ in TM-1, or aromatic residues in ECL-2. Moreover, the chemokine receptor-conserved Glu²⁸⁸, a residue often involved in binding of positively charged small-molecule ligands (Rosenkilde and Schwartz, 2006), is suggested to interact indirectly with FC131 via water molecules (Mungalpara et al., 2012; Yoshikawa et al., 2012). Thus, in line with reported SAR for FC131, the crucial Arg² and 2-Nal³ side chains bind deep in the receptor main binding crevice, while the less important Arg¹ and D-Tyr⁵ side chains experience a larger degree of conformational flexibility and are partly solvent exposed, facing the extracellular surface of the receptor (Mungalpara et al., 2013). However, none of these computational models have been accompanied by in vitro experiments that verify the suggested binding modes.

To determine the binding mode for the lead cyclopentapeptide CXCR4 antagonist FC131, we here report experimental studies that involve modifications of both receptor and ligand. Thus, FC131 and the three analogues [Cit¹]FC131 (substitution of the positively charged L-Arg in position 1 with the neutral L-Cit), [Aib¹]FC131 (substitution of Arg¹ with the small hydrophobic 2-aminoisobutyric acid) and [D-Arg¹]FC131 (opposite stereochemistry in position 1) (Figure 1B) were tested in a library of 25 CXCR4 mutations including Ala, Asn or Trp substitutions of residues in TM-1 to -7 and ECL-2 (Figure 1C) in ¹²⁵I-12G5-binding and receptor-activation assays. This combined approach is the first of its kind to directly investigate the binding mode for FC131 in CXCR4 with in vitro experiments. Interestingly, the receptor mutagenesis also revealed residues important for CXCL12-induced receptor activation. The combined findings provide new experimental insight into the molecular mechanisms of CXCR4 antagonism and will facilitate future design of novel CXCR4 antagonists.

Methods

Compounds

Complete details of the synthesis and characterization of the cyclopentapeptide ligands FC131, [Cit¹]FC131, [Aib¹]FC131 and [D-Arg¹]FC131 have been reported earlier (Mungalpara et al., 2012).

Site-directed mutagenesis

Receptor mutations were introduced by the polymerase chain reaction overlap extension technique or the QuikChange technique (Agilent Technologies, Santa Clara, CA, USA) using wild-type (WT) CXCR4 as template. All reactions were carried out using Pfu polymerase (Stratagene, La Jolla, CA, USA) under conditions recommended by the manufacturer. The mutations were cloned into the eukaryotic expression vector pcDNA3.1+ (Invitrogen, Carlsbad, CA, USA) and verified by restriction endonuclease digestion and DNA sequencing (Eurofins MWG Operon, Ebersberg, Germany).

Transfections and tissue culture

COS-7 cells were grown in DMEM with Glutamax (Invitrogen) supplemented with 10% FBS, 180 U·mL^–1 penicillin and 45 μg·mL⁻¹ streptomycin at 37°C in a 10% CO₂/90% air-humidified atmosphere. Transfection of cells was carried out by the calcium phosphate precipitation method (Rosenkilde et al., 1994; Kissow et al., 2012). Briefly, plasmid DNA (20 μg of receptor cDNA and 30 μg of the chimeric G-protein Gα_qi4myr for inositol phosphate (IP) assays, or 40 μg receptor cDNA for ¹²⁵I-12G5-binding assays) was mixed with TE buffer (10 mM Tris-HCl, 2 mM EDTA-Na₂, pH 7.5) and 30 μL calcium chloride (2 M) to a total volume of 480 μL, and was then added to the same amount of HEPES buffered saline (280 mM NaCl, 50 mM HEPES, 1.5 mM Na₂HPO₄, pH 7.2). Precipitation was allowed for 45 min at room temperature, after which the precipitate, together with 300 μL chloroquine (2 mg·mL⁻¹) in 10 mL culture media, was added to the 6 × 10⁶ COS-7 cells seeded the day before. Transfection was stopped after 5 h by replacing media and cells were incubated overnight.

Functional assay

The potency was measured using IP accumulation assays. In brief, 1 day after transfection, COS-7 cells (1.5 × 10⁵ cells/well) were incubated for 24 h with 2 μCi of [³H]-myo-inositol in 0.3 mL of growth medium per well in a 24-well plate. The following day, cells were washed twice in PBS and were incubated in 0.2 mL of Hank's balanced salt solution (Invitrogen) supplemented with 10 mM LiCl at 37°C in the presence of various concentrations of ligands for 90 min. All mutations were tested for their ability to become activated by CXCL12 using concentrations from 10 pM to 0.1 μM. The IC₅₀ of the antagonists was determined in cells activated by CXCL12 to approximately 80% of CXCL12 E_max. The antagonists (in a range from 1 nM to 100 μM) were added 10 min prior to addition of CXCL12, and co-incubated with CXCL12 for 90 min. Assay medium was then removed, and cells were extracted by addition of 1 mL of 10 mM formic acid to each well, followed by incubation on ice for 30–60 min. The generated [³H]IPs were purified on an AG 1-X8 anion exchange resin. After addition of multipurpose liquid scintillation cocktail (Gold Star, Triskem-International, Bruz, France), radiation was counted in a Beckman Coulter counter LS6500 (Beckman Coulter Danmark ApS c/o OptiNordic ApS, Copenhagen, Denmark). As an alternative assay measuring IP accumulation, the SPA-IP was sometimes used. In brief, 1 day after transfection, COS-7 cells (35.000 cells/well) were incubated with [³H]myo-inositol (5 μL·mL^–1, 2 μCi·mL^–1) in 0.1 mL of media overnight in a 96-well plate. The next day, cells were treated as mentioned above with volumes adjusted as follows: 100 μL of reaction solution with LiCl and 50 μL ice-cold formic acid (10 mM, 50 μL/well). The [³H]IPs in the formic acid cell lysates were thereafter quantified by Ysi-poly-D-Lys-coated SPA beads. Briefly, 20 μL of cell extract was mixed with 80 μL of SPA bead suspension in H₂O (12.5 μg·μL^–1) to give a final volume of 100 μL in a PicoPlate-96 white plate. Plates were sealed, agitated for at least 30 min and centrifuged (5 min, 402 rcf). SPA beads were allowed to settle and react with the extract for 8 h before radioactivity was determined using a Packard Top Count NXT™ scintillation counter (Perkin Elmer). All determinations were made in duplicate. These overall readouts have earlier been used effectively for CXCR4 and other chemokine receptors and were found to give comparable results (Brandish et al., 2003; Mungalpara et al., 2012; Thiele et al., 2012).

Binding experiments

Cells were transfected as described above. The number of cells seeded per well was determined by the apparent receptor expression efficiency and was aimed at obtaining 5–10% specific binding of the added radioactive ligand. Two days after transfection, cells were assayed by competition binding for 3 h at 4°C using 10–15 pM ¹²⁵I-12G5 plus unlabelled ligand in 0.2 mL (in 24-well plates, up to 150.000 cells/well) or 0.1 mL (96-well plates, up to 35.000 cells/well) of 50 mM HEPES buffer, pH 7.4, supplemented with 1 mM CaCl₂, 5 mM MgCl₂ and 0.5% (w/v) BSA in 24-well plates. The binding of ¹²⁵I-12G5 was competed for with increasing concentrations of the unlabelled ligand ranging from 10 pM to 100 nM (12G5) or from 1 nM to 100 μM (FC131, [Cit¹]FC131, [Aib¹]FC131 or [D-Arg¹]FC131). After incubation, cells were washed quickly two times in 4°C binding buffer supplemented with 0.5 M NaCl. Cells were lysed by addition of 0.5 mL carbamide solution (18% acetic acid, 8 M urea, 2% v/v P-40) and radioactivity was counted in a WALLAC Wizard Gamma Counter (Perkin Elmer). Non-specific binding was determined in the presence of 0.1 μM unlabelled 12G5. Determinations were made in duplicate.

elisa

COS-7 cells were seeded in 96-well plates (6 × 10³ cells/well) and transfected with 12.5 ng/well N-terminally FLAG-tagged receptor DNA using lipofectamine transfection according to manufacturer's instructions (Invitrogen). Two days after transfection, cells were washed in Tris-buffered saline (TBS; 0.05 M Tris Base, 0.9% NaCl, pH 7.6), fixed in 3.7% formaldehyde for 15 min at room temperature, washed three times in TBS, and incubated in TBS with 2% BSA for 30 min. The cells were then incubated for 2 h with anti-FLAG M1-antibody (Sigma-Aldrich, St. Louis, MO, USA) at 2 μg·mL⁻¹ in TBS with 1 mM CaCl₂ and 1% BSA. After three washes with TBS supplemented with 1 mM CaCl₂, the cells were incubated with goat anti-mouse HRP-conjugated antibody at 0.8 μg·mL⁻¹ (Thermo Fisher Scientific, Waltham, MA, USA) for 1 h. The immunoreactivity was revealed by addition of TMB Plus substrate (Kem-En-Tec, Taastrup, Denmark) after three additional washes, and the reaction was stopped with 0.2 M H₂SO₄. Absorbance was measured at 450 nm on a Wallac Envision 2104 Multilabel Reader (Perkin Elmer).

Molecular docking

Docking of the cyclopentapeptide ligands to the CXCR4 receptor was performed as described earlier (Mungalpara et al., 2012). Briefly, the X-ray structure of human CXCR4 (bound to CVX15, PDB code 3OE0) (Wu et al., 2010) was prepared with the Protein Preparation Wizard workflow (Schrödinger Suite 2011 Protein Preparation Wizard; Epik version 2.2, Schrödinger, LLC, New York, NY, USA; Impact version 5.7, Schrödinger, LLC; Prime version 2.3, Schrödinger, LLC), and our previously reported bioactive backbone conformation for FC131 (Våbenø et al., 2006b) was used to build the structure of the cyclopentapeptide ligands. The ligands were docked using Schrödingers induced-fit docking workflow (Schrödinger Suite 2012 Induced Fit Docking protocol; Glide version 5.8, Schrödinger, LLC; Prime version 3.1, Schrödinger, LLC), which takes the conformational flexibility of both ligand and receptor residues into account. Asp¹⁸⁷ was used as centroid for the docking box (a cube with 26 Å length) and a H-bond constraint was applied to the carboxylate oxygen atoms of Asp¹⁷¹. As the side chain of Arg¹⁸⁸ partly restricted access to Asp¹⁷¹, the ‘trim’ option was used for Arg¹⁸⁸, that is, the side chain is replaced with a methyl group (alanine) in the initial docking step and then placed back in the final redocking step. For all four ligands, 100 initial poses were generated, and the top 10 optimized poses were retrieved.

Data analysis

Statistical analyses were performed using the GraphPad Prism 5 software (GraphPad Software, San Diego, CA, USA). The EC₅₀ and IC₅₀ values represent the mean of at least three independent experiments (except for [Cit¹]FC131 and [Aib¹]FC131 in 125G binding to Y116A) performed in duplicate (for exact number of experiments, see tables). In cases of incomplete sigmoidal curves (plateau not reached), the curves were extrapolated to baselines (see below) to predict an EC₅₀ or IC₅₀. If this seemed unjustified, logEC₅₀/IC₅₀ values were indicated as >−4 or >−7 in the tables. P-values were calculated using unpaired two-tailed t-test with 95% confidence intervals. Dose–response curves represent averaged, normalized curves. The normalizations were done as follows. (i) In the functional assay, cells were activated to approximately 80% by addition of an appropriate concentration of CXCL12 (see section on functional assay). This activation was set to 100% in each individual experiment, while the background response observed for transfected cells in the absence of ligand was set to 0%. The average of these normalized curves for each assay (‘row means’ in Prism 5) was then calculated. (ii) In the ¹²⁵I-12G5-binding assays, 100% equals maximum ¹²⁵I-12G5 binding to receptor-expressing cells in the absence of unlabelled ligand, while 0% equals the unspecific binding observed with 0.1 μM 12G5.

Materials

All reagents and solvents were purchased and used as received without further purification. The human chemokine CXCL12 was purchased from PeproTech (Rocky Hill, NJ, USA). Human CXCR4 receptor cDNA was kindly provided by Timothy NC Wells (GSK, Brentford, UK). [³H]-myo-inositol (PT6-271), scintillation proximity assay (SPA) beads and ¹²⁵I-Bolton-Hunter reagent were purchased from Perkin Elmer (Waltham, MA, USA). A 12G5 antibody was kindly provided by Jim Hoxie (University of Pennsylvania, Philadelphia, PA, USA) and was iodinated in house as described previously (Rosenkilde et al., 2004). cDNA for the promiscuous chimeric G-protein G_Δ6qi4myr (abbreviated G_qi4myr) was kindly provided by Evi Kostenis (University of Bonn, Germany) (Kostenis et al., 1998). Primers for mutations were bought from TAG Copenhagen (Denmark). The stock solution and dilutions of peptide antagonists were made in water. Stock solution and all dilutions of CXCL12 were made in buffer (1 mM acetic acid + 0.1% BSA).

Results

Expression and functionality of mutant receptors

A library of 25 CXCR4 mutations with Ala, Asn and steric hindrance substitutions of residues located in TM-1 to -7 and ECL-2 (Figure 1C) was created based on previously suggested binding modes of FC131 (Demmer et al., 2011; Mungalpara et al., 2012; Yoshikawa et al., 2012) and the ability of the residues to engage in H-bond, charge-charge and hydrophobic interactions. Thus, these residues constituted likely interaction sites for the main functionalities of FC131, that is, the positively charged side chains of Arg¹ and Arg², the aromatic side chains of 2-Nal³ and D-Tyr⁵, and the peptide backbone.

Initially, WT CXCR4 and all mutant receptors were tested for their surface expression using elisa, and for their functional response towards the endogenous chemokine CXCL12 using COS-7 cells transiently transfected with receptor and the Gα_i- to Gα_q-signal-converting chimeric G_α subunit G_qi4myr, thus measuring accumulation of intracellular IP (Table 2013, Figure 1D). The majority of receptors displayed expression levels from 47 to 111% of WT. R183A, I259W and W94A ranged at the lower end of the scale with 18, 34 and 39% of WT expression, respectively, while two receptors (R188A and H281A) showed expression levels higher than 130% compared with WT (Table 2013). I259A and Q200A displayed the lowest expression with 4.7 and 3.0% of WT level; however, both receptors showed good responses towards CXCL12, demonstrating that they were functional and correctly folded. Likewise, the majority of the receptors showed good responses towards CXCL12 (Table 2013, Figure 1D). Only W94A, D97A and D187A resulted in 8.6- to 14-fold decreased potencies compared with WT CXCR4, and no response was observed for Y116A and E288A (discussed below), despite good surface expression (Table 2013, Figure 1D).

Table 1.

Functional analysis of the interaction between CXCR4 WT and mutants with CXCL12, FC131 and analogues

Receptor			Surface expression		CXCL12					FC131
Helix	Position	Mutation	% ± SEM	(n)	EC50 ± SEM (log)	EC50 (nM)	Fmut	Response at 0.1 μM % of WT ± SEM	(n)	EC50 ± SEM (log)	EC50 (μM)	Fmut	(n)
WT	WT	WTa	100 ± 0.0	(5)	−8.8 ± 0.04	1.5	1.0	94 ± 1.3	(69)	−6.4 ± 0.04	0.40	1.0	(31)
TM-1	I:07/1.39	Y45A	90 ± 7.0	(3)	−8.5 ± 0.20	3.4	2.3**	40 ± 5.5	(10)	−5.8 ± 0.09	1.6	4.1***	(6)
TM-2	II:20/2.60	W94Aa	38 ± 6.4	(3)	−7.7 ± 0.10	18	13***	31 ± 11	(17)	−8.0 ± 0.30	0.01	0.03***	(8)
	II:23/2.63	D97Aa	99 ± 6.3	(4)	−7.7 ± 0.07	20	14***	34 ± 3.5	(10)	−7.4 ± 0.09	0.04	0.11***	(5)
TM-3	III:05/3.29	H113Aa	80 ± 9.0	(3)	−9.1 ± 0.08	0.84	0.58**	83 ± 8.1	(21)	−4.3 ± 0.10	48	119***	(11)
	III:08/3.32	Y116Aa	69 ± 12	(3)	No activation			−0.8 ± 5.9	(5)	Not determined
	III:09/3.33	T117A	74 ± 5.5	(3)	−8.8 ± 0.09	1.7	1.2	57 ± 13	(4)	−6.9 ± 0.05	0.14	0.34**	(3)
TM-4	IV:20/4.60	D171Na	67 ± 11	(3)	−8.5 ± 0.08	3.2	2.2***	38 ± 4.7	(25)	−5.3 ± 0.12	4.6	12***	(13)
	ECL-2/Cys-3	R183A	17 ± 1.8	(3)	−10 ± 0.20	0.10	0.07***	24 ± 1.4	(3)	−6.3 ± 0.18	0.49	1.2	(3)
ECL-2	ECL-2/Cys+1	D187Aa	49 ± 8.7	(4)	−7.9 ± 0.06	13	8.6***	41 ± 5.2	(3)	−6.9 ± 0.02	0.14	0.35***	(3)
	ECL-2/Cys+2	R188A	174 ± 21	(3)	−9.3 ± 0.08	0.53	0.36**	44 ± 3.5	(4)	−6.1 ± 0.17	0.72	1.8	(3)
	ECL-2/Cys+3	F189A	97 ± 8.7	(3)	−8.7 ± 0.11	1.8	1.2	73 ± 9.5	(9)	−7.0 ± 0.20	0.10	0.25***	(6)
	ECL-2/Cys+4	Y190A	105 ± 19	(3)	−8.9 ± 0.18	1.2	0.82	69 ± 7.0	(8)	−6.3 ± 0.24	0.47	1.2	(5)
TM-5	V:01/5.35	V196A	101 ± 13	(3)	−8.9 ± 0.17	1.4	0.96	67 ± 15	(8)	−6.3 ± 0.04	0.47	1.2	(4)
	V:04/5.38	F199A	79 ± 7.2	(3)	−8.8 ± 0.06	1.4	0.98	76 ± 16	(4)	−6.7 ± 0.11	0.21	0.52*	(4)
	V:05/5.39	Q200A	4.7 ± 2.8	(3)	−8.9 ± 0.07	1.4	0.95	71 ± 5.1	(10)	−6.4 ± 0.13	0.36	0.89	(5)
	V:05/5.39	Q200W	75 ± 5.2	(3)	−8.7 ± 0.08	1.8	1.2	34 ± 3.4	(11)	−6.5 ± 0.19	0.33	0.83	(4)
	V:08/5.42	H203A	111 ± 4.6	(3)	−8.9 ± 0.17	1.4	1.0	108 ± 24	(5)	−6.2 ± 0.08	0.58	1.4	(3)
TM-6	VI:13/6.48	W252A	51 ± 4.3	(3)	−9.1 ± 0.06	0.78	0.53**	75 ± 6.2	(11)	−6.1 ± 0.11	0.71	1.8*	(5)
	VI:16/6.51	Y255A	47 ± 3.5	(3)	−8.9 ± 0.11	1.1	0.77	32 ± 9.7	(7)	−6.6 ± 0.36	0.27	0.68	(5)
	VI:20/6.55	I259A	3.0 ± 0.3	(3)	−8.7 ± 0.09	2.1	1.4	59 ± 4.6	(7)	−7.1 ± 0.16	0.08	0.21***	(5)
	VI:20/6.55	I259W	34 ± 4.3	(3)	−8.9 ± 0.06	1.3	0.91	28 ± 4.2	(6)	−6.8 ± 0.20	0.15	0.38**	(5)
	VI:23/6.58	D262Na	54 ± 4.1	(3)	−8.2 ± 0.04	5.8	4.0***	63 ± 7.0	(23)	−5.2 ± 0.09	6.1	15***	(11)
TM-7	VII:−02/7.32	H281Aa	169 ± 29	(3)	−8.7 ± 0.13	1.8	1.2	33 ± 9.5	(18)	−6.1 ± 0.19	0.80	2.0*	(12)
	VII:02/7.35	I284A	56 ± 9.7	(4)	−8.6 ± 0.05	2.3	1.6*	38 ± 4.3	(13)	−6.6 ± 0.10	0.27	0.68	(5)
	VII:06/7.39	E288Aa	77 ± 16	(5)	>−7	>100	>68	11 ± 4.0	(10)	Not determined

Receptor			[Cit1]FC131				[Aib1]FC131				[D-Arg1]FC131
Helix	Position	Mutation	EC50 ± SEM (log)	EC50 (μM)	Fmut	(n)	EC50 ± SEM (log)	EC50 (μM)	Fmut	(n)	EC50 ± SEM (log)	EC50 (μM)	Fmut	(n)
WT	WT	WTa	−5.5 ± 0.11	3.0	1.0	(22)	−5.5 ± 0.09	2.9	1.0	(24)	−6.3 ± 0.09	0.52	1.0	(19)
TM-1	I:07/1.39	Y45A	−4.8 ± 0.12	17	5.7**	(4)	−4.3 ± 0.07	46	16***	(3)	−5.8 ± 0.10	1.6	3.0*	(4)
TM-2	II:20/2.60	W94Aa	−6.8 ± 0.41	0.17	0.06***	(9)	−7.7 ± 0.14	0.02	0.01***	(3)	−7.3 ± 0.67	0.05	0.09**	(3)
	II:23/2.63	D97Aa	−6.4 ± 0.02	0.44	0.15*	(3)	−6.5 ± 0.07	0.35	0.12**	(3)	−7.2 ± 0.05	0.06	0.12**	(3)
TM-3	III:05/3.29	H113Aa	>−4	>100	>33	(10)	>−4	>100	>35	(5)	> −4	>100	>170	(3)
	III:08/3.32	Y116Aa	Not determined				Not determined				Not determined
	III:09/3.33	T117A	−5.4 ± 0.10	4.2	1.4	(3)	−6.0 ± 0.14	1.0	0.34	(3)	−6.1 ± 0.12	0.74	1.4	(3)
TM-4	IV:20/4.60	D171Na	−4.6 ± 0.12	25	8.3***	(12)	−5.2 ± 0.18	6.2	2.2	(5)	−4.5 ± 0.27	28	54***	(4)
	ECL-2/Cys-3	R183A	Not determined				Not determined				Not determined
ECL-2	ECL-2/Cys+1	D187Aa	−6.1 ± 0.03	0.76	0.25	(3)	−6.3 ± 0.12	0.49	0.17**	(3)	−6.5 ± 0.15	0.35	0.66	(3)
	ECL-2/Cys+2	R188A	−4.4 ± 0.09	44	15***	(3)	−5.0 ± 0.25	10	3.5	(3)	−5.7 ± 0.10	2.2	4.2*	(3)
	ECL-2/Cys+3	F189A	−6.1 ± 0.19	0.79	0.27*	(5)	−6.5 ± 0.30	0.31	0.11***	(6)	−6.8 ± 0.02	0.15	0.28*	(4)
	ECL-2/Cys+4	Y190A	−5.6 ± 0.54	2.6	0.87	(4)	−5.4 ± 0.13	4.0	1.4	(6)	−6.2 ± 0.24	0.66	1.3	(3)
TM-5	V:01/5.35	V196A	−5.4 ± 0.22	3.6	1.2	(3)	−5.3 ± 0.27	4.6	1.6	(3)	−5.8 ± 0.27	1.6	3.0	(3)
	V:04/5.38	F199A	−5.6 ± 0.17	2.3	0.77	(3)	−5.5 ± 0.06	2.9	1.0	(3)	−6.5 ± 0.16	0.31	0.59	(3)
	V:05/5.39	Q200A	−6.2 ± 0.12	0.61	0.20*	(3)	−5.2 ± 0.19	6.9	2.4	(3)	−6.2 ± 0.05	0.57	1.1	(3)
	V:05/5.39	Q200W	−5.6 ± 0.15	2.7	0.89	(3)	−5.5 ± 0.16	2.8	1.0	(4)	−6.1 ± 0.24	0.86	1.6	(3)
	V:08/5.42	H203A	−5.1 ± 0.24	8.0	2.7	(3)	−4.7 ± 0.12	20	6.8**	(3)	−6.0 ± 0.15	0.98	1.9	(3)
TM-6	VI:13/6.48	W252A	−4.9 ± 0.19	13	4.3*	(5)	−4.8 ± 0.18	16	5.8**	(5)	−5.5 ± 0.01	3.1	6.0**	(3)
	VI:16/6.51	Y255A	−5.5 ± 0.24	3.2	1.1	(4)	−5.1 ± 0.30	8.5	3.0	(3)	−6.1 ± 0.55	0.8	1.6	(3)
	VI:20/6.55	I259A	−6.4 ± 0.49	0.41	0.14*	(3)	−6.3 ± 0.32	0.53	0.18*	(3)	−6.4 ± 0.53	0.4	0.84	(3)
	VI:20/6.55	I259W	−5.7 ± 0.23	2.2	0.74	(3)	−6.2 ± 0.19	0.68	0.24*	(3)	−6.2 ± 0.18	0.66	1.3	(3)
	VI:23/6.58	D262Na	−5.3 ± 0.13	5.5	1.8	(12)	−5.4 ± 0.13	4.3	1.5	(6)	−5.0 ± 0.18	10	20***	(4)
TM-7	VII:−02/7.32	H281Aa	−5.3 ± 0.17	5.2	1.7	(7)	−6.4 ± 0.53	0.36	0.12**	(4)	−6.2 ± 0.29	0.60	1.1	(4)
	VII:02/7.35	I284A	−4.8 ± 0.42	15	4.9*	(3)	−5.4 ± 0.13	3.8	1.3	(4)	−5.6 ± 0.16	2.6	5.0**	(4)
	VII:06/7.39	E288Aa	Not determined				Not determined				Not determined

IP turnover was measured in COS-7 cells co-transfected with CXCR4 receptor constructs and the promiscuous G-protein G_qi4myr. Residue positions are given according to the numbering systems of Baldwin/Schwartz and Ballesteros/Weinstein. The number of experiments is shown in parentheses, and F_mut indicates the fold-difference (ratio) between the potency on WT CXCR4 compared with mutant CXCR4 with codes as follows: bold >50; bold and italic >15, italic >5, underline <0.2. ^***P < 0.001, ^**P < 0.01, ^*P < 0.05.

^aMutant also tested in binding assay (Table 2007).

Nine mutations were also assessed in ¹²⁵I-12G5-competition binding experiments in transiently transfected COS-7 cells ( Table 2007). This assay has earlier been shown to correlate better with HIV-1 antiviral potency of CXCR4 antagonists than functional assays measuring CXCR4 signalling, and also displays a larger dynamic range (Gerlach et al., 2003; Rosenkilde et al., 2007). These selected receptors were able to bind 12G5 with WT-like affinities (1.9–16 nM) (Table 2007). The B_max values were slightly, yet significantly, reduced for H113A, D171N, H281A and E288A (Table 2007), which however did not correlate to their WT-like surface expression (Table 2013).

Table 2.

Affinity of 12G5, FC131, [Cit¹]FC131, [Aib¹]FC131 and [D-Arg¹]FC131 for WT CXCR4 and various CXCR4 mutations

	Receptor		12G5				FC131
Helix	Position	Mutation	IC50 ± SEM (log)	IC50 (nM)	Fmut	Bmax ± SEM (fmol/105 cells)	(n)	P (Bmax)	IC50 ± SEM (log)	IC50 (μM)	Fmut	P	(n)
WT	WT	WT	−8.3 ± 0.13	4.7	1.0	0.096 ± 0.018	(12)		−6.1 ± 0.09	0.76	1.0		(12)
TM-2	II:20/2.60	W94A	−8.7 ± 0.15	1.9	0.40	0.053 ± 0.022	(8)		−7.5 ± 0.07	0.03	0.04	***	(9)
	II:23/2.63	D97A	−8.0 ± 0.12	9.4	2.0	0.086 ± 0.013	(3)		−6.8 ± 0.15	0.17	0.22	**	(3)
TM-3	III:05/3.29	H113A	−8.6 ± 0.15	2.7	0.58	0.037 ± 0.006	(7)	*	−4.3 ± 0.18	48	63	***	(8)
	III:08/3.32	Y116A	−8.1 ± 0.08	8.7	1.9	0.036 ± 0.016	(5)		> −4	>100	>132		(3)
TM-4	IV:20/4.60	D171N	−8.7 ± 0.18	2.2	0.47	0.034 ± 0.017	(7)	*	−4.3 ± 0.12	55	72	***	(8)
ECL-2	ECL-2 / Cys+1	D187A	−7.8 ± 0.04	16	3.4	0.161 ± 0.012	(3)		−5.1 ± 0.18	7.7	10	***	(3)
TM-6	VI:23/6.58	D262N	−8.4 ± 0.15	3.7	0.79	0.101 ± 0.020	(8)		−4.1 ± 0.18	73	96	***	(9)
TM-7	VII:−02/7.32	H281A	−8.6 ± 0.15	2.4	0.52	0.028 ± 0.008	(7)	*	−4.9 ± 0.11	13	18	***	(8)
	VII:06/7.39	E288A	−8.7 ± 0.16	2.1	0.45	0.035 ± 0.011	(7)	*	−5.4 ± 0.12	4.1	5.5	***	(8)

	Receptor		[Cit1]FC131				[Aib1]FC131					[D-Arg1]FC131
Helix	Position	Mutation	IC50 ± SEM (log)	IC50 (μM)	Fmut	P	(n)	IC50 ± SEM (log)	IC50 (μM)	Fmut	P	(n)	IC50 ± SEM (log)	IC50 (μM)	Fmut	P	(n)
WT	WT	WT	−5.3 ± 0.12	4.9	1.0		(11)	−5.5 ± 0.16	2.9	1.0		(5)	−6.4 ± 0.14	0.38	1.0		(6)
TM-2	II:20/2.60	W94A	−6.6 ± 0.16	0.27	0.06	***	(9)	−6.9 ± 0.33	0.14	0.05	**	(3)	−7.5 ± 0.09	0.03	0.08	**	(3)
	II:23/2.63	D97A	−6.1 ± 0.05	0.87	0.18	**	(3)	−5.9 ± 0.09	1.2	0.40	ns	(3)	No displacement				(3)
TM-3	III:05/3.29	H113A	>−4	>100	>20	***	(8)	No displacement				(3)	No displacement				(3)
	III:08/3.32	Y116A	>−4	>100	>20	***	(2)	No displacement				(2)	>−4	>100	>261		(3)
TM-4	IV:20/4.60	D171N	−4.3 ± 0.11	47	9.6	***	(8)	>−4	>100	>34	***	(3)	>−4	>100	>261		(3)
ECL-2	ECL-2 / Cys+1	D187A	−5.4 ± 0.20	4.3	0.87	ns	(3)	−5.6 ± 0.19	2.6	0.9	ns	(3)	−4.9 ± 0.17	12	31	***	(3)
TM-6	VI:23/6.58	D262N	−4.2 ± 0.10	61	12	***	(9)	>−4	>100	>34	***	(3)	>−4	>100	>261		(3)
TM-7	VII:−02/7.32	H281A	−4.5 ± 0.17	31	6.3	***	(8)	−4.0 ± 0.10	91	31	***	(3)	−4.0 ± 0.21	107	280	***	(3)
	VII:06/7.39	E288A	−5.7 ± 0.19	2.1	0.44	ns	(6)	−5.4 ± 0.10	4.2	1.4	ns	(3)	>−4	>100	>261		(3)

The data were obtained from competition binding with ¹²⁵I-labelled antibody 12G5 as radioligand on transiently transfected COS-7 cells. Values in parentheses represent number of experiments (n), and F_mut indicates the fold-difference (ratio) between the affinities on mutant receptor compared with WT receptor with codes as follows: bold >100 or no displacement at all, bold and italic >25, italic >5, underline <0.2. ^***P < 0.001, ^**P < 0.01, ^*P < 0.05. Residue nomenclature is given in Table 2013.

While secondary/global effects of the mutations on receptor structure and function cannot be excluded, the created set of receptor mutants was deemed suitable for mapping the binding site of FC131 by assessing its ability to inhibit CXCL12-mediated activation or to displace ¹²⁵I-12G5.

FC131-mediated inhibition of CXCL12-induced receptor activation

The entire mutant library was tested in a functional assay determining the ability of the cyclopentapeptide antagonist FC131 to inhibit CXCL12-induced accumulation of intracellular IP. H113A, D171N and D262N in the major binding pocket resulted in 12- to 119-fold reduced FC131 potencies (Figure 2A), while no effects were observed for mutations in ECL-2 (D187A) and the top of TM-7 (H281A) (Figure 2B). Ala substitution of TM-2 residues Trp⁹⁴ and Asp⁹⁷, pointing towards the minor binding pocket (defined by TM-1, -2, -3, -7), improved the potency of FC131 (Figure 2C). CXCL12-induced activity was highly impaired in Y116A and E288A, both pointing into the major binding pocket (delimited by TM-3 to -7, Figure 1C), and FC131 was consequently not tested further here. A large number of mutations in TM-3 (Thr¹¹⁷), ECL-2 (Arg¹⁸³, Arg¹⁸⁸, Phe¹⁸⁹, Tyr¹⁹⁰), TM-5 (Val¹⁹⁶, Phe¹⁹⁹, Gln²⁰⁰, His²⁰³), TM-6 (Trp²⁵², Tyr²⁵⁵, Ile²⁵⁹) and TM-7 (Ile²⁸⁴) did not impair the antagonistic potency of FC131 (Table 2013). However, a small decrease (4.1-fold) was observed for Ala substitution of Tyr⁴⁵ in TM-1.

Figure 2.

Mutational analysis of FC131 in CXCL12 inhibition and ¹²⁵I-12G5-binding studies. The ability of FC131 to inhibit CXCL12-mediated activation (A–C) or to displace ¹²⁵I-12G5 (D–F) from WT CXCR4 (stippled line) or mutants in the TM area (H113A, Y116A, D171N, D262N) (A and D), the exterior receptor parts (H281A, D187A) and E288A (B and E), or the minor binding pocket (W94A, D97A) (C and F) was assessed (see Methods for details). Y116A and E288A were not activated by CXCL12 and could therefore not be assessed in functional studies of FC131 (A and B); n ≥ 3.

The binding site of FC131 is located in the major binding pocket of CXCR4

The effect of the nine selected mutations on the affinity of FC131 was assessed in the ¹²⁵I-12G5 heterologous competition binding assay (Table 2007). Here, similar results were obtained, yet with the expected generally larger changes in affinity (Table 2007) as compared with changes in potency (Table 2013). Thus, FC131 displayed high affinity to WT CXCR4 (IC₅₀ of 0.74 μM), whereas the H113A, Y116A, D171N and D262N mutants resulted in 63- to >132-fold decreased affinities (Figure 2D). H281A and D187A resulted in a lower though significant decrease (18- and 10-fold respectively). A minor decrease in affinity was also observed for E288A (5.5-fold) (Figure 2E). Finally, in analogy to the functional assay results, W94A and D97A led to 25- and 4.6-fold increased affinities respectively (Figure 2F).

Molecular docking of FC131 in CXCR4

Next, FC131 was docked to the X-ray crystal structure of CXCR4 [PDB code 3OE0 (Wu et al., 2010)] using the induced-fit docking protocol developed by Schrödinger (see Methods). As the binding and functional studies (Tables 2013 and 2007) both showed a dependence on the spatially close residues His¹¹³ (TM-3) and Asp¹⁷¹ (TM-4), a H-bond constraint was set on the carboxylate group of Asp¹⁷¹. The following binding mode, which was among the top 10 optimized poses and supports the experimental data outlined above, is suggested (Figure 3A–C): Arg¹ of FC131 interacts with Asp¹⁸⁷ while Arg² interacts with His¹¹³/Asp¹⁷¹; although also a direct interaction of Asp⁹⁷ with Arg¹ in FC131 is observed in this docking pose (not shown) and in earlier computational studies (Demmer et al., 2011; Mungalpara et al., 2012; Yoshikawa et al., 2012), the observation that the D97A mutation led to an increased affinity and potency of FC131 argues for a different role of Asp⁹⁷ (Figure 2C and F). The aromatic 2-Nal³ side chain is positioned in a tight hydrophobic pocket facing TM-5, and sandwiched between Arg¹⁸⁸ (cation-π interactions) and His²⁰³ (π-π interactions). In most poses, D-Tyr⁵ of FC131 points towards Glu³² in the receptor N-terminus, while in some poses an interaction with Asp²⁶² was observed (not shown). Finally, Glu²⁸⁸ interacts with the backbone of the ligand via a water-mediated H-bond network. Thus, FC131 binds in the major binding pocket of CXCR4, with the Arg² and 2-Nal³ side chains buried deeply, while the Arg¹ and D-Tyr⁵ side chains point outwards.

Figure 3.

The binding mode of FC131 in CXCR4. (A) Two-dimensional representation of the FC131-binding site in CXCR4. Residue colours: red, negative; purple, positive; cyan, polar; green, hydrophobic. Interactions: pink full and stippled arrows, H-bond with main and side chain respectively; green line, π-π stacking; red line, cation-π interaction; grey cloud, solvent-exposed atom. Three-dimensional model of FC131 binding in CXCR4 as seen from top (B) or the side (C). TM-5 and -6 have been removed for clarity.

Overall, His¹¹³, Asp¹⁷¹, Asp¹⁸⁷ and Glu²⁸⁸ are part of the binding site (Glu²⁸⁸ via water molecules), confirming recently suggested binding modes for FC131 (Demmer et al., 2011; Mungalpara et al., 2012; Yoshikawa et al., 2012), while Tyr⁴⁵, Tyr¹¹⁶ and His²⁸¹ do not directly interact with FC131, but nevertheless influence its binding and activity to varying extents. A direct interaction of D-Tyr⁵ of FC131 with Asp²⁶² is only seen in a few poses and is not likely to account for the large effect of the D262N mutation (96- and 12-fold decrease in affinity and potency, respectively, of FC131). However, Asp²⁶² is a central residue in a H-bond network involving Gln²⁰⁰ (TM-5), His²⁸¹ (TM-7), and Arg³⁰ (N-terminus) (not shown), and removal of the charge in Asp²⁶² may disturb this network and thereby indirectly affect FC131 binding and function.

[Cit¹]FC131 without a positively charged side chain in position 1 loses dependency on Asp¹⁸⁷ in ECL-2

Previous SAR studies of FC131 (Figure 4A) have shown that Arg¹ (but not Arg²) can be replaced with the uncharged L-Cit residue (Figure 4B) (Mungalpara et al., 2012). In order to confirm the suggested binding mode of FC131 (Figure 3), we subjected the [Cit¹]FC131 analogue to the same mutational analysis in ¹²⁵I-12G5-binding and CXCL12-functional studies. Consistent with previous data (Mungalpara et al., 2012), [Cit¹]FC131 displayed six- to eightfold lower affinity and potency as compared with FC131 (Tables 2013 and 2007). However, the effect of most mutations on [Cit¹]FC131 was similar to that observed for FC131 (Figure 4D). Thus, mutation of residues facing the major binding pocket (H113A, Y116A, D171N, D262N) resulted in 9.6- to >20-fold decreases in affinity (Figure 4E and Table 2007). H281A, in the top of TM-7, led to a 6.3-fold decrease, while E288A resulted in a partial displacement with unchanged affinity. Furthermore, as for FC131, mutations in TM-2 of the minor binding pocket (W94A and D97A) led to increased affinities (Table 2007). However, contrary to what was observed for FC131 (Figure 4D), D187A in ECL-2 did not affect the binding of [Cit¹]FC131 (Figure 4E). Thus, the affinity of FC131 on D187A (IC₅₀ of 7.7 μM) is similar to the affinity of [Cit¹]FC131 on WT CXCR4 (IC₅₀ of 4.9 μM), pointing towards an interaction of side chain 1 with Asp¹⁸⁷.

Figure 4.

[Cit¹]FC131 and [Aib¹]FC131 confirm the orientation of FC131 in CXCR4, and highlight the role of Arg¹⁸⁸ and His²⁰³ for sandwiching the 2-Nal³ side chain. The structures of the side chains in position 1 of FC131 (A), [Cit¹]FC131 (B) and [Aib¹]FC131 (C) are given. (D–F) ¹²⁵I-12G5-binding assays on WT CXCR4 and mutant receptors (H113A, D171N and D187N, symbols as in Figure 2). (G–I) Effects of mutations D187A, R188A, F189A, Y190A (ECL-2) and H203A (TM-5) in binding (D187A) or functional assays (other mutants) shown as fold decreases. (J and K) Molecular docking of the analogues.

The effect of receptor mutants on the ability of [Cit¹]FC131 to inhibit CXCL12-mediated activation confirmed the picture observed in ¹²⁵I-12G5 binding. Mutation of residues in the major binding pocket, including those located deeply and those located more superficially, resulted in strongly (H113A) or moderately (I284A, D171N) decreased or unchanged (D262N, H281A) potency of [Cit¹]FC131 (Table 2013). The lack of effect for D262N could be due to the generally smaller dynamic window in functional studies compared with ¹²⁵I-12G5 binding and the smaller effect observed in binding for [Cit¹]FC131 (12-fold) versus FC131 (96-fold). Furthermore, the potency of [Cit¹]FC131 was decreased 5.7-fold for Y45A, while mutation of residues in TM-2 led to increased potencies (W94A, D97A) (Table 2013). As expected from ¹²⁵I-12G5-binding experiments, D187A did not impair the antagonistic potency of [Cit¹]FC131 (Figure 4H). Analysis of neighbouring residues in ECL-2 (Figure 4G and H) revealed that Ala substitution of Arg¹⁸⁸ led to a 15-fold reduced potency of [Cit¹]FC131 (Figure 4H), while having no effect on FC131 (Figure 4G). None of the other aromatic residues in ECL-2 (Phe¹⁸⁹, Tyr¹⁹⁰) affected the potency of either [Cit¹]FC131 or FC131 (Figure 4G and H). This highlights the effect of the cation-π interaction between 2-Nal³ and Arg¹⁸⁸ in the absence of Arg¹. Molecular docking of [Cit¹]FC131 into CXCR4 also reveals that Asp¹⁸⁷ is pointing away from side chain Cit¹ (Figure 4J and K). Of the remaining mutations (Thr¹¹⁷, Val¹⁹⁶, Phe¹⁹⁹, Gln²⁰⁰, His²⁰³, Trp²⁵², Tyr²⁵⁵, Ile²⁵⁹), only W252A led to 4.3-fold impaired potency (Table 2013).

[Aib¹]FC131, which lacks a side chain functionality in position 1, displays the same binding mode as FC131 and [Cit¹]FC131

As described above, FC131 tolerates removal of the positive charge from the side chain in position 1. It also tolerates truncation of this side chain to the backbone stabilizing disubstituted residue Aib, that is, [Aib¹]FC131 (Figure 4C) (Mungalpara et al., 2012), which has the same affinity and potency as [Cit¹]FC131 (Tables 2013 and 2007). The mutagenesis study of this compound in ¹²⁵I-12G5 binding revealed a stronger dependence on residues in the major binding pocket (H113, Y116, D171, D262, H281) than for [Cit¹]FC131. However, in analogy with [Cit¹]FC131, no effect was observed for D187A or E288A, while W94A and D97A led to similar increases in affinity (Table 2007, Figure 4F). [Aib¹]FC131 also mirrored [Cit¹]FC131 in the functional studies with a few exceptions (Table 2013): compared with [Cit¹]FC131 it showed decreased dependency on Asp¹⁷¹ and Arg¹⁸⁸ in TM-IV and ECL-2 respectively. In contrast, it depended to a higher degree on His²⁰³, as its potency was decreased 6.8-fold by H203A (Figure 4I), while that of [Cit¹]FC131 was only impaired 2.7-fold (Figure 4H). Thus, the importance of Arg¹⁸⁸ and His²⁰³ in sandwiching the 2-Nal³ side chain, as discussed above, seems to change from Arg¹⁸⁸ for [Cit¹]FC131 to His²⁰³ for [Aib¹]FC131. Furthermore, Y45A resulted in 16-fold decreased potency of [Aib¹]FC131, compared with the smaller impact of 5.7-fold for [Cit¹]FC131. Finally, computational modelling confirms a binding mode that overlaps with those of FC131 and [Cit¹]FC131 for side chains L-Arg², 2-Nal³, Gly⁴ and D-Tyr⁵ (Figure 4J and K). This emphasizes that side chain 1 is not required for achieving this binding mode of cyclopentapeptides in CXCR4, yet plays a role for high potency and affinity, as described earlier (Tamamura et al., 2005a; Demmer et al., 2011; Mungalpara et al., 2012).

The close analogue [D-Arg¹]FC131 behaves differently than FC131 in its ability to displace ¹²⁵I-12G5

[D-Arg¹]FC131 differs from FC131 only in the stereochemistry in position 1 and displays similar potency and two-fold higher affinity; however, this compound interacted differently with CXCR4. In ¹²⁵I-12G5-binding experiments, a stronger dependency was observed on His¹¹³, Tyr¹¹⁶, Asp¹⁷¹, Asp¹⁸⁷, Asp²⁶², His²⁸¹ and Glu²⁸⁸ for [D-Arg¹]FC131 than for FC131, while in functional assays both compounds behaved largely similar on all mutants (Figure 5A, Tables 2013 and 2007). Interestingly, [D-Arg¹]FC131 acted differently on mutations in TM-2 than FC131; while W94A consistently led to increased affinities and potencies, D97A abrogated the ability of [D-Arg¹]FC131 to displace ¹²⁵I-12G5, while it, as for FC131, increased its antagonistic potency (Figure 5B). Molecular docking of [D-Arg¹]FC131 to CXCR4 revealed a larger flexibility of the exteriorly located part of the molecule as compared with FC131. While D-Arg¹ still mainly interacted with Asp¹⁸⁷, D-Tyr⁵ displayed a larger degree of conformational freedom and pointed everywhere from TM-2 and the N-terminus to TM-6 (Figure 5C); however, the crucial ligand side chains Arg² and 2-Nal³ bound to His¹¹³/Asp¹⁷¹ and the hydrophobic pocket around TM-5 in the same way as in FC131.

Figure 5.

Mutational analysis and computational modelling of [D-Arg¹]FC131 binding in CXCR4. (A) Fold decreases of FC131 and [D-Arg¹]FC131 affinity (upper diagram) and antagonistic potency (lower diagram) observed for a range of mutants in comparison to WT CXCR4. (B) The effect of D97A in comparison to WT (stippled line) on the affinity (upper part) and potency (lower part) of FC131 and [D-Arg¹]FC131. (C) Molecular docking of [D-Arg¹]FC131 (white structures) showing multiple binding poses in overlay with the FC131-binding pose (green structure). n.d., not determined.

A H-bond between Tyr¹¹⁶ and Glu²⁸⁸ plays a role in the activation of CXCR4

In the crystal structure of the complex between CXCR4 and the peptide antagonist CVX15 (Figure 6A) (Wu et al., 2010) and in our models of the receptor-bound cyclopentapeptide ligands (Figure 3C), a H-bond is observed between Tyr¹¹⁶ in TM-3 and Glu²⁸⁸ in TM-7. In vitro experiments showed that Ala substitution of Tyr¹¹⁶ or Glu²⁸⁸ abolished the agonistic action of CXCL12 (Figure 6B), despite surface expression levels of 69 and 77% of WT respectively (Table 2013). Both mutant receptors bound ¹²⁵I-12G5 with WT-like affinities, suggesting proper folding of the receptors (Figure 6C). Furthermore, all four cyclopentapeptide ligands were unable to displace ¹²⁵I-12G5 from Y116A-CXCR4, while only [D-Arg¹]FC131 was affected by E288A (Figure 6D, Table 2007). Although Tyr¹¹⁶ was not revealed as a direct interaction partner for the cyclopentapeptide ligands in the docking studies, these mutagenesis data point towards a role of Tyr¹¹⁶ for the ability of the ligands to displace ¹²⁵I-12G5. Furthermore, the H-bond between Tyr¹¹⁶ and Glu²⁸⁸ seems crucial for the activation of CXCR4 by CXCL12. Further studies are needed to fully understand the functional importance of the link between Tyr¹¹⁶ and Glu²⁸⁸ in CXCR4.

Figure 6.

Mutation of the H-bond forming residues Tyr¹¹⁶ and Glu²⁸⁸ abolished CXCL12-induced receptor activation. (A) Tyr¹¹⁶ and Glu²⁸⁸ form a H-bond in the crystal structure of CXCR4 (here bound to CVX15, PDB code 3OE0). (B) Ability of CXCL12 to activate WT CXCR4 and mutants Y116A and E288A. (C) Homologous ¹²⁵I-12G5-competition binding experiments on WT, Y116A and E288A (symbols as in A) and the B_max of 12G5 in fmol/10⁶ cells for each receptor (inset). (D) Ability of FC131 to displace ¹²⁵I-12G5 from WT, Y116A and E288A (symbols as in A).

Discussion and conclusions

We have used a dual approach combining receptor mutational analysis with ligand modifications to determine the binding mode of FC131 in CXCR4: Arg¹, Arg², 2-Nal³ and D-Tyr⁵ of FC131 interact with ECL-2 (Asp¹⁸⁷), TM-3/-4 (His¹¹³, Asp¹⁷¹), TM-5 and the exterior receptor part (Glu³²) respectively. The orientation of FC131 in the pocket was confirmed by [Cit¹]FC131 and [Aib¹]FC131 that both lack the positive charge at position 1 and at the same time do not depend on Asp¹⁸⁷. Overall, our data are consistent with earlier proposed binding modes predicting Arg¹ and D-Tyr⁵ to point outwards, while Arg² and 2-Nal³ interact with residues deep in the major binding pocket (Figure 3) (Demmer et al., 2011; Yoshikawa et al., 2012; Mungalpara et al., 2013).

Comparison of the binding modes of FC131, CVX15 and AMD compounds

The recent crystal structure of CXCR4 with the low MW compound IT1t or the peptide CVX15 (Wu et al., 2010) gave first-hand insights into antagonist interaction with CXCR4. Surprisingly, IT1t interacted with residues Glu²⁸⁸, Asp¹⁸⁷ and Asp⁹⁷ in the minor binding pocket, while the peptide ligand CVX15, a 16-mer analogue of the 14-mer T140 that FC131 was developed from, was located in the major binding pocket and in extracellular receptor regions. Specifically, Arg¹ of CVX15 interacted with Asp¹⁸⁷, Arg² with His¹¹³/Asp¹⁷¹, and Arg¹⁴ with Asp²⁶² (Wu et al., 2010). We find that FC131 mimics the binding of CVX15 as it interacts with Asp¹⁸⁷ via Arg¹, and His¹¹³/Asp¹⁷¹ via Arg². Thus, the two arginine residues in FC131 correspond to Arg¹ and Arg² of CVX15, and not to Arg² and Arg¹⁴ as originally intended (Figure 7). Asp²⁶² was not found to interact with Arg¹ or Arg² of FC131 in any docking pose. Alternatively, the effect of D262N might be attributed to a central role of this residue in a H-bond network involving Arg³⁰, Gln²⁰⁰ and His²⁸¹, as mentioned above. Furthermore, a link between ECL-2 (carrying Asp¹⁸⁷) and Asp²⁶² is established via Gln²⁰⁰ in TM-5, which is directly linked to ECL-2. Removing a conformational constraint between Asp²⁶² in TM-6 and Gln²⁰⁰ in TM-5 might therefore alter the position of ECL-2. Such a scenario would also explain the weakened effect of D262N on analogues [Cit¹]FC131 and [Aib¹]FC131, lacking the positive charge at position 1 and dependency on Asp¹⁸⁷ in ECL-2.

Figure 7.

The binding of FC131 compared with CVX15. Overlay of the binding modes for FC131 (green) identified in the present study and CVX15 (yellow) from the co-crystal structure of CXCR4 and CVX15 (PDB code: 3OE0).

Finally, 2-Nal³ of FC131 and the corresponding 1-Nal³ of CVX15 bind in a hydrophobic sub-pocket at TM-5; however, as previously suggested by comparing SAR data for the cyclopentapeptides and the larger peptide antagonists, the naphthyl groups do not completely overlap (Mungalpara et al., 2013). Clearly, the 2-Nal³ side chain of FC131 goes deeper into the hydrophobic sub-pocket, and presumably contributes more to activity than the 1-Nal³ side chain of CVX15. The tyrosine residue in position 5 of both ligands takes up different positions. While Tyr⁵ of CVX15 faces the upper part of the hydrophobic pocket around TM-5 (Wu et al., 2010), rotation of D-Tyr⁵ to Glu³² (N-terminus) was observed in the FC131-CXCR4 complex, again in agreement with SAR studies on this position suggesting a solvent-exposed, freely rotatable position in CXCR4 (Mungalpara et al., 2013).

The well-described non-peptide AMD-compound series (the bicyclam AMD3100, the monocyclam AMD3465 and the non-cyclam AMD11070) was earlier shown to depend on Asp²⁶²/Glu²⁸⁸ in TM-6/-7 on one side, and Asp¹⁷¹ in TM-4 on the opposite side of the major binding pocket (Gerlach et al., 2003; Rosenkilde et al., 2004; 2007,). Furthermore, mutation of residues in the minor binding pocket was found to impair their action and multiple binding modes were subsequently suggested (Gerlach et al., 2003; Rosenkilde et al., 2004; 2007,; Hatse et al., 2005; Wong et al., 2008; Gudmundsson et al., 2009a,b,; Catalano et al., 2010; Miller et al., 2010; Skerlj et al., 2011). In the present study, FC131 was found to only indirectly interact with Glu²⁸⁸ via a water-mediated H-bond network, and therefore behaves somewhat differently from these reference CXCR4 antagonists and from most other small-molecule antagonists where the chemokine receptor-conserved Glu in position VII:06/7.39 seems to function as a general anchor point for positively charged nitrogens (Rosenkilde and Schwartz, 2006).

The role of ECL-2 in the binding of cyclic pentapeptides in CXCR4

ECL-2 connects TM-4 with TM-5 and is covalently linked to the top of TM-3 via a conserved disulphide bond. Thereby, the C-terminal part of ECL-2 (also called ECL-2b) is being held close to the extracellular surface of the main binding crevice of CXCR4. Asp¹⁸⁷ is located in position Cys+1 in ECL-2b and the D187A mutation resulted in decreased affinities of FC131, but not [Cit¹]FC131, pointing towards an interaction of Asp¹⁸⁷ with Arg¹ of the cyclopentapeptides. The adjacent Arg¹⁸⁸ was earlier suggested to interact with the aromatic 2-Nal³ side chain of FC131 by engaging in a cation-π interaction. This is also observed in our studies (Figure 4H), yet we do not see an effect of R188A on the potency of FC131, whereas the potency of [Cit¹]FC131 decreases by 15-fold and that of [Aib¹]FC131 by 3.5-fold (however, for [Aib¹]FC131 a role of the second suggested interaction partner for 2-Nal³, His²⁰³ in TM-V, becomes visible) (Table 2013). Thus, Asp¹⁸⁷ seems to be the most important residue for FC131 in ECL-2, yet in the absence of the interaction between Arg¹ in FC131 and Asp¹⁸⁷ (in [Cit¹] and [Aib¹]FC131), the impact of Arg¹⁸⁸ becomes visible. Alternatively, Arg¹⁸⁸ of CXCR4 and Arg¹ of FC131 might repel each other. Thus, the R188A mutation would remove the favourable interaction with 2-Nal³, but also the unfavourable electrostatic repulsion of Arg¹, leading to a zero net effect of R188A on FC131. Importantly, a similar direct role of ECL-2b was found for the CCR5 antagonist aplaviroc (Maeda et al., 2006; Thiele et al., 2011). In a broader perspective, binding of small-molecule compounds to extracellular chemokine receptor domains has the potential of overlapping with the binding sites of chemokines, which due to their large size interact with the exterior parts of their receptors (Allen et al., 2007; Scholten et al., 2012). Low MW antagonists, although by default considered allosteric, may therefore become partially overlapping, resulting in competitive behaviour.

The role of Tyr¹¹⁶ for the function of CXCR4 antagonists

According to the two-step model of chemokine receptor activation, the interaction between CXCR4 and CXCL12 involves distinct receptor and chemokine domains in binding and activation (Crump et al., 1997; Gupta et al., 2001). In CXCR4, the initial high-affinity binding is mainly mediated by sulpho-tyrosines in the receptor N-terminus that interact with positively charged residues of CXCL12. In a second step, N-terminal CXCL12 residues interact with transmembrane receptor residues, and presumably also ECL-2 to induce receptor activation (Crump et al., 1997; Ludeman and Stone, 2013). In agreement with this model, and consistent with the data presented here, Ala substitution of transmembrane residues Asp⁹⁷, Tyr¹¹⁶ and Glu²⁸⁸ affects the signalling properties (Table 2013, Figure 6), but not the binding, of CXCL12 (Rosenkilde et al., 2007; Wong et al., 2008). Thus, it can be speculated that the observed H-bond between Tyr¹¹⁶ and Glu²⁸⁸ (Wu et al., 2010) is crucial for CXCL12-mediated receptor activation. Furthermore, the function of all tested cyclopentapeptides depended on Tyr¹¹⁶ (Table 2007, Figure 6D), probably via an indirect mechanism, as no direct interaction with the ligand was observed (Figure 3C). Interestingly, the function of AMD3100 and AMD3465 has also been shown to depend on Tyr¹¹⁶ (Wong et al., 2008); however, it remains to be determined whether this effect is direct or indirect. Finally, as described above, Glu²⁸⁸ was found to be an indirect interaction partner for FC131, but mutation only resulted in minor effects for most cyclopentapeptide ligands, except binding of [D-Arg¹]FC131 (Table 2007). Thus, the Tyr¹¹⁶-Glu²⁸⁸ H-bond at the bottom of the major binding pocket is central not only for the activation but also for the inhibition of CXCR4, and consequently for the activity states of CXCR4.

The entrance to the binding crevice in CXCR4 is covered by H-bond and Cys bridge

While mutations in TM-2 (W94A, D97N) impair the affinity of the AMD compounds (Wong et al., 2008), we observed that W94A and D97A increased the potencies and affinities of the cyclopentapeptide antagonists. In the crystal structure of CXCR4, Asp⁹⁷ forms a salt bridge with Arg¹⁸³ in ECL-2, which together with the chemokine receptor-conserved disulphide bridge between the N-terminus and top of TM-7 results in a partly covered major binding pocket (Wu et al., 2010). This is not seen in the newly released structure of CCR5, which lacks Asp⁹⁷ (or an equivalent thereof) and has a more open entrance to its binding pocket (Tan et al., 2013). Although speculative, it could therefore be argued that breaking the Asp⁹⁷/Arg¹⁸³-salt bridge by Ala substitution of Asp⁹⁷ releases the closed extracellular conformation of CXCR4 and gives FC131 easier access to its binding pocket. Mutation R183A does however not lead to increased potency of FC131 (Table 2013); yet in the absence of the Arg¹⁸³ side chain, it can be speculated that another residue takes over its function in the salt bridge to Asp⁹⁷; a H-bond is indeed observed between Asp⁹⁷ and the backbone of Cys¹⁸⁶. W94A may have a similar effect by providing more space for Asp⁹⁷, thereby breaking the H-bond with Arg¹⁸³, or simply by creating more room for the ligand.

In conclusion, by combining receptor mutagenesis with ligand modifications, we determined the binding site of FC131 in CXCR4. In addition, our studies suggest a H-bond in the centre of the receptor between Tyr¹¹⁶ and Glu²⁸⁸ to be essential for the activation state of CXCR4. Finally, consistent with other studies of class A 7TM receptors, such as EBI2 (GPR183), CCR5 and CCR2, where a central role of the top of TM-2 is identified for the activity state (Alvarez Arias et al., 2003; Benned-Jensen and Rosenkilde, 2008; Rosenkilde et al., 2010), a possible gating function of the top of TM-2 for the entrance into the main binding crevice of CXCR4 is suggested, implying that improved CXCR4 antagonists could be obtained by creating smaller molecules that can easily migrate into the main binding crevice of CXCR4.

Glossary

2-Nal

3-(2-naphthyl)alanine

7TM

7 transmembrane helix

Aib

2-aminoisobutyric acid

Cit

citrulline

ECL

extracellular loop

HIV

human immunodeficiency virus

IP

inositol phosphate

SAR

structure–activity relationship

WT

wild type

Author contributions

J. V., M. M. R. and S. T.: Experimental design. J. M.: Compound synthesis. S. T.: Binding studies. S. T., A. S. and J. M.: Functional studies. S. T.: elisa. S. T. and J. M.: Data analysis. J. M. and J. V.: Computational modelling. S. T., J. M., M. M. R. and J. V.: Manuscript writing.

Conflict of interest

None.