Structural Basis for Allosteric Ligand Recognition in the Human CC Chemokine Receptor 7

Abstract

The CC chemokine receptor 7 (CCR7) balances immunity and tolerance by homeostatic trafficking of immune cells. In cancer, CCR7-mediated trafficking leads to lymph node metastasis, suggesting the receptor as a promising therapeutic target. Here, we present the crystal structure of human CCR7 fused to the protein Sialidase NanA by using data up to 2.1 Å resolution. The structure shows the ligand Cmp2105 bound to an intracellular allosteric binding pocket. A sulfonamide group, characteristic for various chemokine receptor ligands, binds to a patch of conserved residues in the Gi protein binding region between transmembrane helix 7 and helix 8. We demonstrate how structural data can be used in combination with a compound repository and automated thermal stability screening to identify and modulate allosteric chemokine receptor antagonists. We detect both novel (CS-1 and CS-2) and clinically relevant (CXCR1-CXCR2 phase-II antagonist Navarixin) CCR7 modulators with implications for multi-target strategies against cancer.

Keywords: CCR7; G protein-coupled receptors; allosteric modulation; cancer; chemokine receptors; crystal structure; lymph node metastasis; membrane proteins; structure-based drug screening.

Graphical abstract

Figure S1

Related to Star Methods Section CCR7 Constructs and Expression Schematic illustration of constructs for CCR7 crystallization screening. Six rounds of construct design and screening were carried out. Expression constructs contained a C-terminal fusion of enhanced green fluorescent protein (eGFP) (Cormack et al., 1996) to determine expression levels followed by a decahistidine-tag (His10) for affinity chromatography purification. Human Rhinovirus 3C protease (Cordingley et al., 1990) recognition sequences were introduced to cleave the N- and C terminus during purification. A replacement of leucine residue at position 145 with tryptophan (Roth et al., 2008) improved thermal stability of the receptor. A series of 17 soluble proteins were selected from the protein data bank for insertion into intracellular loop 3 based on following selection criteria: (1) virtually parallel N- and C-termini with a distance between 6–15 Å. (2) structures available at a resolution better than 2.0 Å. (3) no disulfide bridges or posttranslational modifications. (4) A maximum number of 600 inserted residues v) final manual verification to ensure a maximal diversity of protein folds. Six fusion proteins (shown with name, PDB code, and molecular weight) were chosen for large scale expression of which three crystallized. The Sialidase NanA fusion yielded the best diffracting crystals and was chosen for a final round of optimization resulting in the structure of the CCR7-Sialidase NanA fusion protein.

Figure 1

Structure of CCR7 Bound to Cmp2105 The human CCR7 shares the seven transmembrane (TM1–TM7) helical fold of GPCRs where helices are connected by three intracellular (ICL) and three extracellular (ECL) loops. A short aliphatic helix (H8) anchors CCR7 in the cytoplasmic side of the membrane. CCR7 (light blue) viewed in parallel to (A) and from the extracellular (B) and intracellular (C) sides of the membrane (indicated by gray boundaries). The connecting linkers to the fusion protein are labeled (yellow) with two unresolved residues indicated as dashed lines. The small molecule antagonist Cmp2105 (green sticks) binds at the intracellular side of CCR7, similar as described for binding of Cmpd-15PA to the β2-adrenergic receptor, binding of CCR2-RA-[R] to CCR2 and binding of Vercirnon to CCR9.

Figure S2

Related to Figure 1 (A) CCR7-Sialidase NanA crystal structure with allosteric antagonist Cmp2105. The chemokine receptor CCR7 (blue) is connected by two linkers in TM5 and TM6 (yellow) to the fusion protein Sialidase NanA (orange) which is located 9.2 Å apart from Cmp2105. (B) Crystal lattice arrangement viewed from three different angles. Crystal contacts exist between adjacent Sialidase NanA and the CCR7 loops ECL1 and ECL2. B-factors (red to blue) are higher for the receptor domain (mean B-factor: 100.6 Å²) than for the fusion protein (mean B-factor: 21.7 Å²). The B-factors indicate a greater flexibility of the receptor compared to the fusion protein Sialidase NanA. Because of the few crystal contacts restraining the receptor, future studies on conformational dynamics at room temperature might be feasible. The large size of the fusion protein may further qualify the CCR7-Sialidase NanA construct for single particle analysis using electron microscopy.

Figure S3

Related to Figure 2 Automated serial data collection was performed to identify the location of native sulfur atoms and assist model building of CCR7. Single-wavelength anomalous dispersion of sulfur atoms (S-SAD) is a routine phasing method (Weinert et al., 2015) that does not need derivatization of the target protein. However, the method is challenging in cases where only weakly diffracting microcrystals are available because it requires very accurate measurements of anomalous differences vulnerable to radiation damage. Automated serial data collection based on raster scanning of mounted mesh loops and automated selection of crystals using X-ray diffraction at 6 keV (A) before and (B) after diffraction screening, blue to red indicates diffraction power at a 5x5 μm² grid point) provides a solution to this problem (Wojdyla et al., 2018, Wojdyla et al., 2016) as the required dose can be spread over small wedges of data collected from thousands of crystals with sizes of only a few micrometers. Diffraction data was collected on 2343 crystals within 24 h of beamtime. Data from the best 726 crystals were combined into a high multiplicity dataset (Table S1). The anomalous signal in these data provided the position of sulfur atoms in the CCR7-Sialidase NanA fusion protein (C) (yellow spheres indicate positions with signal above and gray below 3.0 σ) and Cmp2105 (D) (yellow mesh, 3.0 σ). Additional phase information was further used to assist structure determination by molecular replacement. The location of Cmp2105 was confirmed by simulated annealing F_obs-F_calc omit maps (E) (green mesh, 2.5 σ).

Figure S4

Conservation Patterns of the CCR7 Orthosteric and Allosteric Ligand Binding Site Compared to Other Chemokine Receptors, Related to Figure 3 (A) Small molecules can inhibit chemokine receptors through either the extracellular orthosteric binding pocket, the chemokine (e.g., CCL19/21) activation site, or the intracellular allosteric pocket, which interacts with signaling partners like G proteins. The Cα backbone atoms of residues in contact with small molecule ligands in human chemokine receptors are plotted on the structure of CCR7 (spheres, conservation from blue to red). (B) Illustration of the size, shape, and position of the small molecule binding pockets in respect to the cellular membrane (blue bars). Numbers in brackets indicate sequence identity to CCR7 and root means square deviation of Cα atoms in transmembrane helices. (C) Interatomic contacts (defined as pair of atoms with < 4 Å distance) between chemokine receptors with small molecule ligands (PDB: 5T1A, 3ODU, 4MBS, and 5LWE), chemokines (PDB: 4RWS and 4XT1) and a selection of GPCR-effector complexes (PDB: 3SN6, 6CMO, and 5DGY). Overall a higher overlap between contact sites is observed in the tight intracellular allosteric binding pocket overlapping with the GPCR-effector binding site. The gray-scale indicates the number of contacts to the respective ligand. The orthosteric binding pocket has a higher level of sequence variation compared to the allosteric site. It is wide open and highly polar to allow binding of chemokines which are, with approximate sizes of 8 to 12 kDa, much larger compared to a typical class A GPCR ligand. Synthetic small molecule ligands of chemokine receptors, therefore, target a smaller subpocket as observed for IT1t in CXCR4 (Wu et al., 2010) and BMS-681 in CCR2 (Zheng et al., 2016) or several subpockets like the HIV drug Maraviroc in CCR5 (Tan et al., 2013). In the CCR7 structure much of the homologous pocket surface is occupied by a desirable mix of hydrophobic and polar residues as interaction points for small molecules with an average conservation 58% between human chemokine receptors. Of particular interest for the design of orthosteric CCR7 ligands are Trp114^2.60 and Tyr136^3.32 where the corresponding residue is highly conserved and is directly interacting with a ligand in all known structures. Tyr312^7.38, on the other hand, is a ligand-interacting glutamate in CCR2, CXCR4 and CCR5 but hydrogen-bonding to Tyr136^3.32 in case of CCR7. Finding and exploring such variations in the orthosteric binding pocket will help to optimize selective ligands to pharmaceutically target CCR7 once initial hits have been identified.

Figure S5

Related to Figure 2 and 3 Cmp2105 belongs to a set of four CCR7 binding molecules orginating from a large group of thiadiazole-dioxides developed and patented as potent ligands for chemokine receptors. They contain a thiadiazole-dioxide core motif with a characteristic sulfonyl group. Modification of the two amine-linked exit vectors of the central motif allows to fine-tune binding toward CCR7 (dissociation constants [K_d] taken from Taveras et al., 2010).

Figure 2

Cmp2105 Exerts Intracellular Allosteric Inhibition of CCR7 (A) Thermal-shift assays (top) verify binding of Cmp2105 to CCR7. Increasing concentrations of Cmp2105 reveal a strong dose-dependent stabilizing effect on CCR7 of up to 20.1°C (mean ± SEM from 3 independent experiments with 3 measurements each). Cmp2105 allosterically inhibits binding of the native chemokine CCL19 ligand in scintillation proximity assays (bottom) with a half inhibitory concentration (IC₅₀) of 35 nM. (B) Overlay of the CCR7 structure with the position of the Gα_i subunit (Kang et al., 2018) (red) or arrestin (Kang et al., 2015) (purple) in structures of rhodopsin signaling complexes. The comparison places Cmp2105 (green; sticks and spheres) in a position where it interferes with binding of these GPCR effector proteins. (C) A structural comparison with the inactive conformation of CCR2 (Zheng et al., 2016) and the active conformation of the viral US28 with bound chemokine (Burg et al., 2015) suggests Cmp2105 to stabilize an inactive CCR7 conformation with closed intracellular effector binding site. View from the cytoplasmic side with arrows indicating relative positions in the inactive and active GPCR conformation. Our assignment to a deactivated CCR7 is further confirmed by a putative sodium ion in a conserved site between TM2, TM3, TM6, and TM7, which is known to negatively modulate activity in many GPCRs (Liu et al., 2012). Our results thus show how Cmp2105 exerts allosteric antagonism close to the intracellular G protein binding pocket of CCR7.

Figure 3

Binding Mode Comparison for Cmp2105 (CCR7), CCR2-RA-[R] (CCR2), and Vercirnon (CCR9) (A–C) Schematic ligand interaction profiles with protein residues colored according to their chemical nature (hydrophobic in green, polar in pink, acidic in red, and basic in blue). Dotted green lines represent regions with hydrophobic interactions. (D–F) Structural comparison of the CCR7 allosteric binding pocket with those of CCR2 and CCR9 reveals strong similarities in allosteric chemokine receptor ligand binding to a conserved patch of residues in the TM7-H8 turn (CCR7 in blue, CCR2 in yellow, and CCR9 in orange). (G) Summary of interatomic contacts (defined as pair of atoms with < 4 Å distance) in the allosteric binding pocket with CCR2-RA-[R] (PDB: 5T1A), Vercirnon (PDB: 5LWE), and a selection of GPCR-effector complexes (PDB: 3SN6, 6CMO, and 5DGY). Residues are shown in single-letter code with critical sites emphasized with bold letters. The grayscale indicates the number of contacts to the respective ligand or effector protein.

Figure 4

Focused Screening and Structural Diversity of Chemokine Receptor Ligands (A) 293 compounds were selected by virtual screening from Roche’s compound repository and tested experimentally for their ability to thermally stabilize CCR7. The graph plots the difference in thermal stability in presence of 50 μM ligand (black dots mean ± SEM from 3 determinations) to the DMSO control. Potentially stabilizing ligands are colored orange and hits above the combined standard deviation of DMSO and ligand are red. (B) To probe the selectivity of the binding pocket, we tested and analyzed a series of known chemokine receptor ligands and positive hits from virtual and semi-automated thermofluor screening (mean ± SEM from 3 measurements) (Mattle et al., 2018). The dashed vertical line divides measurements from the automated thermofluor assay and separate measurements with selected ligands. (C) Selection of allosteric small molecule ligands against chemokine receptors tested for their effect on CCR7. The central core motif binding the TM7-H8 motif is highlighted. (D and E) Docking of Navarixin and CS-1 into the allosteric CCR7 binding pocket. The TM7-H8 joint is shown in blue and selected key residues are drawn as sticks.

Figure 5

Cellular Arrestin Recruitment Assay and Dose Response Thermofluor Assay (A) Dose response curves with the native agonist CCL19 (green, circles) show arrestin binding with a half effective concentration (EC₅₀) of 0.012 μM. (B) Addition of the allosteric inhibitor Cmp2105 (red, circles) or Navarixin (orange, squares) suppresses arrestin binding in response to activation by CCL19 with half inhibitory concentrations (IC₅₀) of 7.3 μM and 33.9 μM respectively. Data points are the mean of two independent measurements and are normalized to the maximal response. (C) Dose-dependent thermo-stabilizing effect of Cmp2105 on CCR7. Results from 3 independent experiemnts with 3 measurements each (mean ± SEM) were fitted to extract the maximal stabilizing effect and half effective concentration of 0.48 uM for Cmp2105. (D) Dose-dependent thermo-stabilizing effect of Navarixin on CCR7. Results from three independent experiments with three measurements each (mean ± SEM) were fitted to extract the maximal stabilizing effect and half effective concentrations of 13.38 μM for Navarixin.