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

1A key feature of the human immune system is the ability to protect against pathogens without targeting healthy cells or tissues. Chemotactic trafficking of the immune response is orchestrated by 20 G protein-coupled receptors (GPCRs) and over 40 chemokines that guide patrolling immune cells to the right place at the right time. Inflammatory chemokines and their receptors are induced upon inflammatory stimuli, whereas homeostatic chemokines are expressed constitutively and create “cellular highways” that constantly navigate cells to specific organs. Important homeostatic CC motif chemokine ligands are CCL19 and CCL21, which bind the CC chemokine receptor 7 (CCR7) to guide B cells, T cells, and antigen-presenting dendritic cells to lymph nodes throughout the body.

2CCR7 and its ligands are essential components in an autoimmune model of rheumatoid arthritis. Moreover, pathogenic bacteria can use CCR7-mediated migration of dendritic cells to transfer to draining lymph nodes, from where they can spread to other organs. CCR7 is further associated with a wide variety of cancers, where chemotactic trafficking allows cancer cells to spread by lymph node metastasis. CCR7 expression in colorectal carcinoma, the second most common malignant tumor worldwide, is linked to lymphovascular invasion and decreased survival rates. Small molecule ligands designed to silence CCR7 thus have great potential to address lymph node metastasis, a major cause for cancer-associated mortality.

3So far it has been difficult to find a suitable variety of small molecule ligands as starting points to develop CCR7-targeting drugs. Furthermore, other chemokine receptors, including CXCR1, CXCR2, CCR5, and CXCR4, have also been implicated with cancer metastasis to specific organs and thus might compensate for each other if only one of them is targeted. On the other hand, small molecules need to have a certain level of selectivity to minimize side effects during treatment. Structural information is needed to develop drugs that can fulfill these potentially conflicting requirements and much effort has been devoted to determining the structures of chemokine receptors in complex with either synthetic ligands or native chemokines.

4Despite several clinical trials, only a few drugs have been approved for the large and highly polar orthosteric chemokine binding pocket. Spatially distinct allosteric sites in chemokine receptors and other GPCRs are more compact and promise higher sub-type selectivity than orthosteric sites and are thus widely discussed as exciting possibilities in modern drug discovery. Here, we present the crystal structure of CCR7 bound to the allosteric antagonist Cmp2105. This small molecule contains a thiadiazole-dioxide scaffold, which was developed and patented as a potent motif to target CXC- and CC-chemokine receptors. We conclude from new ligands found by 3D shape similarity screenings and verified by automated thermofluor screening that thiadiazole-dioxide, thiazole-dioxide, or cyclobutene-dione motifs are key for interacting with a conserved transmembrane helix 7-helix 8 (TM7-H8) protein patch in the Gi protein binding site in CCR7 and other chemokine receptors. Our results suggest this conserved motif as a promising hotspot for targeting chemokine receptors with pharmaceuticals.

5Increasing the hydrophilic surface area by using fusion proteins such as T4 lysozyme is one of the most successful crystallization strategies for GPCRs. Because T4 lysozyme fusions failed to produce crystals in this study and to extend the toolbox for membrane protein crystallization, we identified 17 soluble proteins with optimal properties to facilitate crystallization of CCR7 (Figure S1). Out of five novel fusion constructs expressing at high amounts, three yielded crystals. The construct with the largest fusion protein Sialidase NanA (52.8 kDa) resulted in the most promising diffraction patterns. Further optimization allowed us to collect X-ray crystallographic data to a resolution up to 2.1 Å (Table S1). We employed molecular replacement combined with native single-wavelength anomalous dispersion (SAD) data obtained in a serial-scanning approach to determine the structure of the human CCR7 (Figures 1, S2, and S3). Superposition of the CCR7 seven transmembrane helical bundle with that of other human chemokine receptors (CCR2, CCR5, CCR9, and CXCR4) yields Cα root mean square deviations between 1.28 Å and 1.97 Å in accordance with sequence identities between 29.9% and 39.1% (Figure S4). Sequence and structural differences between the receptors are higher in the orthosteric chemokine binding pocket, on the extracellular half, than in the intracellular side, which opens upon activation to accommodate arrestin and G protein, the two major GPCR signaling proteins.

6GPCRs are intrinsically allosteric proteins that tightly couple the intracellular and extracellular halves of their 7TM bundle to fulfill their function in cellular signaling. Such a cooperative effect is highly desirable for cancer therapies targeting CCR7, given that it allows for more opportunities to inactivate the receptor to prevent trafficking by chemokines and thus block metastasis through the lymphatic system.

7To date, very few small molecule ligands for CCR7 have been described. However, a large group of patented thiadiazole-dioxides contain four very similar small molecule compounds that bind CCR7 with nanomolar affinity (Figure S5). We purified and crystallized CCR7 in the presence of the thiadiazole-dioxide ligand Cmp2105. The addition of this small molecule ligand results in a strong stabilizing effect of up to 20.1°C on CCR7 (Figure 2A) in thermofluor experiments. Such a strong stabilizing effect is not untypical for a high-affinity binder. For example, the cellular entry inhibitor Maraviroc exerts a similar stabilizing effect of up to 18.9°C on the related CCR5 receptor. Accordingly, we measured a half maximal inhibitory concentration (IC50) of 35 nM for Cmp215 in membrane-based competition experiments with radioactively labeled CCL19. Because Cmp2105 outcompeted the native protein ligand CCL19 and was classified as an antagonist to G protein activation, we initially expected it to bind the orthosteric chemokine binding pocket. However, our structure showed Cmp2105 bound to a pocket at the intracellular part of CCR7 between the ends of TM1, TM2, TM3, and TM6 and the loop between TM7 and H8; this was similar to what was seen for CCR2-RA-[R] in CCR2, Vercirnon in CCR9, and Cmpd-15PA in the β2-adrenergic receptor. Cmp2105 stabilizes an inactive CCR7 conformation as evidenced by comparisons of our structure with the activated conformation of the CX3CL1-chemokine-bound viral US28 receptor and the inactive conformation of the CCR2 receptor, with orthosteric and allosteric antagonists (Figure 2). Based on an overlay with the rhodopsin-Gi protein complex and the rhodopsin-arrestin complex, the intracellular binding pocket spatially overlaps with the intracellular effector binding site of CCR7. Together with the inhibitory effect on CCL19 binding, the structural comparisons identify Cmp2105 as an intracellular allosteric antagonist. The binding site thus represents an attractive opportunity to silence CCR7 with small molecule ligands.

8Cmp2105 is composed of a thiadiazole-dioxide core motif with two amine-linked substituents that can be exchanged to modulate binding affinity to CCR7 (Figure S5). The substituents form interactions to several residues in TM2 (including hydrogen bonds to Thr912.37 and Thr932.39) and TM1 (mainly hydrophobic Val791.53, Thr821.56, and Phe861.60). They further bridge well-conserved residues including Arg1543.50 of the ERY motif in TM3 and Tyr3267.53 of the NPxxY motif in TM7 (Figures 3A and 3D; Table S2), both part of the cytoplasmic cleft that opens upon GPCR activation. The partial overlap with the G protein binding site, in addition to the Cmp2105 receptor interactions, hinder large conformational changes that are required for receptor activation.

9The two amino groups of Cmp2105 form hydrogen bonds with Asp942.40 indicating a central rolein positioning the central core motif. The sulfonyl group interacts with a conserved patch of residues at the turn between TM7 and H8 (Tyr3267.53, Gly3308.47, Val3318.48, Lys3328.49, and Phe3338.50). Gly3308.47 at the very end of TM8 is conserved among most human chemokine receptors (except CXCR6 and CXCR7) and allows the tight interhelical joint of the pocket to form. Mutations of residues Tyr7.53 and Phe8.50 in rhodopsin substantially reduce binding to Gαt, and the recent cryo-EM structures of Gi/o-receptor complexes show how G protein selectivity is mediated by how their C termini approach the conserved motifs within the intracellular cleft including selective interactions with TM7-H8. In the structure of CCR9, the TM7-H8 motif binds a sulfonamide group of the ligand Vercirnon (Figures 3C and 3F), whereas the hydroxy-pyrrolinone ring of CCR2-RA-[R] occupies the equivalent position in CCR2 (Figures 3B and 3E). The conserved TM7-H8 patch (Figure 3G) within the allosteric binding pocket thus seems a promising hotspot for targeting Gαi binding chemokine receptors in agreement with the large variety of potent molecules in the thiadiazole-dioxide series of CXC- and CC-chemokine receptor ligands.

10The similarity in how chemokine receptor ligands approach the Gi protein binding site in CCR2, CCR7, and CCR9 is striking. All ligands interact with the conserved TM7-H8 motif in their respective receptors. Nevertheless, CCR2-RA-[R] and Vercirnon did not have a thermo-stabilizing effect on CCR7, despite their similar interactions to the TM7-H8 motif in other receptors. Neither did the CXCR2 antagonist Danirixin nor the CXCR1 and CXCR2 antagonist Reparixin stabilize CCR7, although both molecules contain the characteristic sulfonamide group found in a number of chemokine ligands. Moreover, orthosteric ligands such as the CCR5 antagonist Maraviroc and the CXCR4 antagonist IT1t did not affect CCR7 (Figure 4B). This raises the question of what defines selectivity for allosteric ligands in chemokine receptors?

11To address this question, we conducted a 3D shape similarity search among the 2.3 million compounds of the Roche repository. We used Cmp2105 as a seed in a focused screen to identify ligands with different topologies but a similar 3D pharmacophore. A selection of 293 fitting compounds were choosen for thermal stability assays (Figure 4A) via the automated methodology we developed originally to identify pharmacological chaperones binding to rhodopsin. This combined approach revealed a series of potentially stabilizing molecules above the standard deviation of the dimethyl sulfoxide (DMSO) control, from which we selected the strongest two for further investigation. Molecule CS-1 (Figure 4C) contains the characteristic cyclic sulfone; however, the exit vectors are no longer adjacent and lack the secondary amines found in Cmp2105. Molecular docking places it into the allosteric binding pocket with the sulfonyl group facing the TM7-H8 joint (Figure 4D). No hydrogen bonds are available for interaction with Asp942.40 or other nearby residues. This explains the smaller stabilizing effect of 2.4°C of this ligand but also demonstrates the importance of the interaction to the TM7-H8 core motif to approach the receptor.

12The second hit is of special interest because it is known as Navarixin (or SCH-527123 or MK-7123), a potent and bioavailable antagonist for CXCR1 and CXCR2. The left- and right-hand substituents are almost identical to Cmp2105, but the central thiadiazole-dioxide core of Navarixin is replaced by a cyclobutene-dione. Molecular docking experiments place Navarixin into a similar position as Cmp2105, where it interacts with the TM7-H8 motif as well as with Asp942.40. The only significant difference is the interaction of the dione in Navarixin versus the sulfonyl group in Cmp2105 (Figure 4E). The second carbonyl group of Navarixin cannot form a hydrogen bond with the protein. This could explain why the overall stabilizing effect is lower in thermal stability experiments (4°C compared with 20.1°C) and the half maximal inhibitory concentration (IC50) in cellular CCL19 competition assays is reduced (33.9 μM for Navarixin in comparison with 7.3 μM for Cmp2105) (Figures 5A and 5B). Overall, docking results agree well both with dose-response thermal-shift (Figures 5C and 5D) assays and the allosteric modulation of signaling in a cellular environment. Navarixin is thus validated as a biologically active allosteric antagonist for CCR7.

13Clearly this part of the allosteric pocket can be targeted with a large chemical variety of small molecules to prevent binding of intracellular effector proteins. The selectivity of ligands can be tuned by changing the core motif by using a thiadiazole-dioxide (CCR7-Cmp2105), thiazole-dioxide (CCR7-CS-1), a cyclobutene-dione (CXCR1,CXCR2, and CCR7-Navarixin), a hydroxy-pyrrolinone (CCR2-CCR2-RA-[R]), or just the sulfonamide group alone (CCR9-Vercirnon). Once binding to the receptor with a chosen core-motif is established, near endless variations of the ligands can be achieved by modifying the substituents to fine-tune selectivity to a single or a subset of chemokine receptors.

14Due to their implications with cancer metastasis and their roles in the pathogenesis of autoimmune diseases, inflammation, and viral infections, chemokine receptors are important targets for pharmaceutical intervention. Despite considerable efforts, only two small molecule drugs targeting chemokine receptors are on the market today: Plerixafor (CXCR4) for stem-cell mobilization and Maraviroc (CCR5) for HIV infection. Clinical failures for drugs that target chemokine receptors often result from low efficacy, low selectivity, as well as redundancies in chemokine signaling, in which the inhibition of one receptor has no significant effect because of the cell’s ability to compensate for that signaling pathway. These efforts should be reignited, considering recent insights from molecular structures and the emerging concept of using allosteric ligands to modulate GPCR signaling. The well-conserved allosteric Gi protein binding pocket on the intracellular side is a particularly interesting pharmacological target and offers advantages in comparison to the orthosteric chemokine binding pocket, as ligand selectivity can be tuned to target a single subtype or a selection of chemokine receptors. The structure elucidation of CCR7 and the validation of the allosteric binding pocket opens a new avenue to a structure-based design approach for finding novel small molecule ligands to silence chemokine receptors. Our study demonstrates this approach and presents a surprising result that Navarixin binds an allosteric pocket in CCR7. Navarixin is currently tested in phase II clinical trials (http//:clinicaltrials.gov) for its anti-metastatic effect on colorectal and other aggressive cancers. Considering the role of CCR7 in metastasis, it is not difficult to imagine that part of the Navarixin anticancer effects are due to silencing CCR7 instead of it acting solely via CXCR1/CXCR2. Drugs binding multiple targets have long been flagged as undesirable, as it was assumed this inherently leads to adverse side effects. However, a multi-target drug can also be beneficial in scenarios where redundant biological pathways lead to a compensation and resistance to single-target therapies. The identification of ligands for the relatively conserved and compact allosteric binding pocket in a subset of homeostatic chemokine receptors might thus be the best approach for alleviating multifaceted maladies such as cancer.

15Further information and requests for reagents should be directed to and will be fulfilled by the Lead Contact Joerg Standfuss (joerg.standfuss@psi.ch).

16The Bac-to-Bac baculovirus expression system (Invitrogen) was used to generate high-titer recombinant baculovirus. Sf9 cells at a density of 2 × 106 cells/mL in SF-4 Baculo Express ICM medium (BioConcept) were infected at a multiplicity of infection of 0.01%–5% (v/v) depending on the virus strength. The cells were shaken in culture flasks (800 mL per 2 L Erlenmeyer flask) for 72 h at 27°C and 120 rpm. The cell pellet was harvested by centrifugation (3000 × g, 20 min, 4°C) and stored at −80°C.

17Cellular CCR7 G protein activation assays were performed using the cAMP Hunter CHO-K1 CCR7 Gi Cell Line (Eurofins). Cellular CCR7 arrestin recruitment assays were done using the PathHunter® eXpress CCR7 CHO-K1 β-Arrestin GPCR Assay (Eurofins). Both assays were performed by Eurofins using standard protocols and relying on CCL19 as activating agonist.

18The wild type human CCR7 DNA sequence was optimized for insect cell expression and cloned into a pFastBac vector (Invitrogen). The receptor sequence (residues 1–348) was fused with enhanced green fluorescent protein for monitoring expression, followed by a C-terminal decahistidine-tag for purification. A tryptophan point mutation (L145W) was introduced to improve the thermal stability of the receptor. In the crystallization construct, two cleavage sites for human Rhinovirus 3C protease have been introduced to cut the N- and C terminus between residue 36-43 and 352-359, respectively. In addition, residues Arg2485.63-Phe256ICL3 have been deleted to introduce residues 1–470 of Sialidase NanA fusion protein (PDB: 2ya4) flanked by a linker (Ser248-Lys-Leu-His-IP9-Ser-Lys-Gly-His256) on each side (Figure S1).

19The CCR7 was expressed in 10 l scale in 20 l Cell-bag Disposal Bioreactors (Wave Biotech/GE life sciences) at 27°C and 19 rocks/min in a 40% oxygen atmosphere. Typically, 2 × 106 Sf9 cells/mL Sf900-III medium (Invitrogen) at a viability of > 97% and an average diameter of 17–17.5 μm were infected with a volume of infection (VOI) of 0.1%–5% depending on the virus strength. The protein production was aborted at cell counts of 3–3.5 × 106 cells/mL, a cell viability of ∼80% and an average diameter of 20.5–21.5 μm. This led to a fermentation of about 72 h. Cells were pelleted and then stored at −80°C until purification.

20All purification steps were performed at 4°C unless stated otherwise. For cell lysis, frozen cell pellets were thawed in a hypotonic buffer containing 10 mM HEPES/NaOH pH 7.5, 10 mM MgCl2, 20 mM KCl, Complete protease inhibitor (Roche) (1 tablet/100 mL) while stirring at 300 rpm. Cell membranes were disrupted by dounce homogenization and isolated using ultracentrifugation (234788 x g, 45 min, 4°C, Beckmann Optima-80 XE-100, rotor Ti45). Extensive washing of the membranes was performed by repeated resuspension and centrifugation with low salt buffer (10 mM HEPES/NaOH pH 7.5, 10 mM MgCl2, 20 mM KCl, Roche protease inhibitor tablets (1/100 mL) (1x), a high osmotic high salt buffer (10 mM HEPES/NaOH pH 7.5, 10 mM MgCl2, 20 mM KCl, 1 M NaCl) (2x) and low salt buffer (1x) to remove the high salt content. Purified membranes were resuspended in low salt buffer to a concentration of 1 g/mL, flash frozen in liquid nitrogen and stored at −80°C until further use.

21At the purification day, membranes were thawed in water at room temperature and treated with 23 μM Cmp2105, 2 mg/mL iodoacetamide and Complete protease inhibitor (Roche) (1 tablet/50 mL). The membranes were incubated at 4°C for 1 h and solubilized in 50 mM HEPES/NaOH pH 7.5, 300 mM NaCl, 20 mM imidazole/HCl pH 7.5, 1% (w/v) DDM, 0.2% (w/v) CHS, 23 μM Cmp2105 and Complete protease inhibitor (1 tablet/50 mL) keeping a 1:1.75 ratio (amount of biomass to final volume). The mixture was stirred for 1 h at 4°C and 600 rpm and the supernatant was isolated by ultracentrifugation (234788 x g, 1 h, 4°C, Beckmann Optima-80 XE-100, rotor Ti45). The supernatant containing the solubilized receptor was incubated with pre-equilibrated Talon Superflow resin (TaKaRa, approximately 0.3 mL of resin was used for 1 g original cell pellet) and the mixture was stirred for 60 min at 4°C and 600 rpm. After binding, the resin was poured into a Falcon tube, centrifuged (500 x g, 2 min, 4°C, Eppendorf 5801 R) and washed with 10 × 2 column volumes (CV) wash buffer (50 mM HEPES/NaOH pH 7.5, 300 mM NaCl, 20 mM imidazole/HCl pH 7.5, 0.03/0.006% (w/v) DDM/CHS, 2 μM Cmp2105) in batch mode. The resin was packed into a XK16 column (GE Healthcare) column connected to a FPLC system (ÄKTA Prime, GE Healthcare) in a cooling cabinet. The resin was washed with 1 CV 20 mM imidazole and 1 CV 40 mM imidazole and then eluted in 1.5 CV 50 mM HEPES/NaOH pH 7.5, 300 mM NaCl, 200 mM imidazole/HCl pH 7.5, 0.03/0.006% (w/v) DDM/CHS, 2 μM Cmp2105. The eluted receptor was incubated with His-tagged human Rhinovirus 3C protease (HRV 3C) (in-house, 1:10 molar ratio protease:receptor) overnight in order to cleave off the N terminus and C-terminal GFP fusion and decahistidine-tag. The receptor was concentrated to a volume of < 2.5 mL using a 100 kDa molecular weight cut-off Vivaspin concentrator (Sartorius) and subsequently exchanged into a buffer containing 25 mM HEPES/NaOH pH 7.5, 150 mM NaCl, 0.03/0.006% (w/v) DDM/CHS, 2 μM Cmp2105 using a PD10 desalting column (GE Healthcare). The receptor was further purified by removing the C-terminal His-tagged GFP and the 3C HRV protease using NiNTA Sepharose resin (Iba lifesciences, 1 mL of resin per 12 mg of receptor). The pure receptor was collected as the Ni-NTA column flow-through and concentrated to 25 mg/mL. In between, the receptor was supplemented with Cmp2105 to a final concentration of 100 μM when it had reached a volume of about 1 mL. The pure, highly concentrated receptor was flash frozen in liquid nitrogen and stored at −80°C.

22Prior to crystallization, the receptor was treated with 2 mM Cmp2105, incubated for 30 min on ice and reconstituted into lipidic cubic phase (LCP) using premixed molten monoolein (90% v/v)/ cholesterol (10% v/v) and two coupled Hamilton syringes. The lipid reconstitution and following steps were performed at room temperature (19-22°C). The final mixture contained 40% (v/v) protein solution, 54% (v/v) monoolein, and 6% (v/v) cholesterol. Crystallization trials were set up as 15 nL of protein-laden LCP and 800 nL of precipitant solution per well in 96-well Laminex sandwich glass or plastic plates (Laminex, Molecular Dimensions) using a Mosquito LCP dispensing robot (TTP Labtech) with a humidity chamber set to 70% humidity. Plates were incubated and imaged at 20°C using an automated imager (RockImager 1000, Formulatrix).

23Diffraction quality crystals grew in 23%–27% (v/v) PEG 500 MME, 100 mM ammonium tartrate dibasic, 200 mM MES pH 6.0 or 100 mM Bis-tris pH 5.7, and with the addition of either 2.5 mM GSH (glutathione), 1-2.5 mM GSSG (glutathione disulfide) or 1-5 mM GSH/GSSG. Crystals of near cubic shape usually grew to a maximum size of 10-15 × 10-20 × 10-25 μm3 within 7-10 days at 20°C. Crystals were harvested from the LCP matrix using MiTeGen micromounts and flash frozen in liquid nitrogen ready for data collection.

24Crystallographic data were collected at the X06SA (PXI) beamline at the Swiss Light Source using a 5 × 5 μm2 collimated X-ray beam. Rastering for crystal centering was done using 0.03 s exposure, 90% transmission, 0.0° oscillation and a detector distance of 300 mm. Individual crystals were centered for data collection and data were collected with an exposure time of 0.025 s at 40% transmission with 0.05° oscillation and a detector distance of 300 mm. Data from 11 crystals were processed and integrated with XDS and subject to STARANISO (http://staraniso.globalphasing.org/cgi-bin/staraniso.cgi) analysis for scaling and merging of the data. Data anisotropy appears to be a direct consequence of the rigid insertion protein packing along the a- and b- axes and the conformationally less restrained receptor, packing along the c-axis (Figure S2). Phaser was used to find the initial phase information using a homology model of CCR7 and a structure of the Sialidase NanA fusion protein (pdb code: 2YA4) as separate search models. The electron density was readily interpreted for the complete Sialidase NanA fusion, whereas parts of the linker and CCR7 (N-terminal residues 1-53, ICL2 residues 160-164 and ECL3 residues 289-298) were disordered. The model was refined with Phenix, using Rosetta refinement in initial stages, followed by visual examination and rebuilding of the refined coordinates in COOT using anomalous differences as a guideline for placing the CCR7 sequence.

25Using Cmp2105 as a seed compound, a substructure and 2D similarity search within the Roche compound repository was performed in Pipeline Pilot (Dassault Systèmes BIOVIA) with the extended connectivity fingerprints ECFP-6 and a Tanimoto threshold of 0.4. In addition, a 3D shape similarity search with FastROCS was carried out (OpenEye Scientific Software, Santa Fe, NM, USA. http://www.eyesopen.com). After experimental verification using thermofluor assays (see below) docking experiments were performed with the software GOLD from CCDC with default settings. The best 10 docking poses were examined visually to select the most reasonable docking mode with respect to molecular interactions and internal conformational strain.

26The CCR7 constructs used for thermofluor stability assays were based on CCR7-L145W without the fusion protein and were expressed and purified as described above. For the generation of dose-response curves, 59 μL of purified CCR7-L145W apo-receptor at a concentration of 0.524 μM were distributed into the wells of a 96 well PCR plate on ice and 1.51 μL of a compound stock solution (at 0.1 μM – 2 mM) was added to each well, resulting in a final concentration of 0.0025 – 200 μM. The plate was sealed and incubated for 30 min on ice. A 1:100 (v/v) working solution of the CPM (N-[4-(7-diethylamino-4-methyl-3-coumarinyl)phenyl]maleimide) dye stock (3 mg/mL in DMSO) was prepared and 5.14 μL of this solution were added to each well and mixed thoroughly. Out of each well, 3 × 20 μL were distributed into 0.1 mL Rotor-Gene Q tubes (QIAGEN). The melting profiles were recorded using a real-time PCR machine (Rotor-Gene Q, QIAGEN) with temperature ramping from 25°C to 95°C in 1°C steps, 4 s pause after each step, an excitation of 365 ± 20 nm and emission of 460 ± 20 nm. The gain set was determined at the beginning of the run. The melting temperature (Tm) was calculated from the point of inflection.

27The automated thermofluor assay was performed as described with 0.54 μM CCR7-L145W and 50 μM ligand selected by virtual screening (described above). Melting curves were obtained applying a temperature gradient from 25 to 95°C and a heating rate of 0.25°C/s. All liquid-handling steps were performed using a Bravo automated liquid handling platform and a 96LT head.

28Scintillation proximity assays (SPA) were carried out in 96 well plates (Optiplate, Perkin Elmer) using ChemiSCREEN CCR7 Membrane preparations (Millipore), PVT-PEI-WGA Type B SPA beads (Perkin Elmer) and a mixture of radioactively labeled (2200 Ci/mmole, R&D Systems or Perkin Elmer) and non-labeled human CCL19 (Prospec). Reactions took place in 50 mM TRIS, 5 mM MgCl2, 1 mM CaCl2, 50 mM NaCl, 0.1% BSA, pH 7.6 supplemented with a CCR7 cell membrane/SPA beads mix (0.5 mg/well) and serially diluted non-labeled CCL19 (0.01 nM to 30 nM final conc.) or serially diluted Cmp2105 (1 nM to 10 μM) together with 0.05 nM labeled CCL19 in all wells. Assays were incubated for 1 h at room temperature before values were read out using a top count scintillation counter.

29Arrestin recruitment was measured using the PathHunter® β-Arrestin assay (contracted to Eurofins). PathHunter® eXpress CCR7 CHO-K1 β-Arrestin cells were expanded from freezer stocks. Cells were seeded in a total volume of 20 μL Assay Complete Cell Plating Reagent (Eurofins) into white walled, 384-well microplates and incubated at 37°C. For agonist dose response curves, cells were incubated with CCL19 dissolved in DMSO (final concentration 1%) for 90 min at 37°C. For the negative allosteric modulation format, cells were pre-incubated with varying concentrations of either Cmp2105 or Navarixin followed by CCL19 challenge at its EC80 concentration (0.057 μM) for 90 min at 37°C. Assay signal was generated through a single addition of 50% v/v PathHunter Detection reagent cocktail, followed by a one h incubation at room temperature. Microplates were read following signal generation with a PerkinElmer Envision™ instrument for chemiluminescent signal detection.

30IC50 values from the scintillation proximity and cellular assays were determined using GraphPad Prism version 7.0d for Mac, GraphPad Software, La Jolla, CA, USA, http//:www.graphpad.com.

31Melting temperatures (Tm) in the thermofluor stability assays were calculated and analyzed by automated scripts based on a fit to the Boltzmann equation. Details on significance levels and number of experiments are given in the corresponding figure legends.

32The accession number for the data reported in this paper is PDB: 6QZH.