High-affinity agonist binding to C5aR results from a cooperative two-site binding mechanism

A current challenge in the field of life sciences is to decipher, in their native environment, the functional activation of cell surface receptors upon binding of complex ligands. Lack of suitable nanoscopic methods has hampered our ability to meet this challenge in an experimental manner. Here, we use for the first time the interplay between atomic force microscopy, steered molecular dynamics and functional assays to elucidate the complex ligand-binding mechanism of C5a with the human G protein-coupled C5a receptor (C5aR). We have identified two independent binding sites acting in concert where the N-terminal C5aR serves as kinetic trap and the transmembrane domain as functional site. Our results corroborate the two-site binding model and clearly identify a cooperative effect between two binding sites within the C5aR. We anticipate that our methodology could be used for development and design of new therapeutic agents to negatively modulate C5aR activity.

interact with receptors having a protruding height centered at 3.1 ± 1.0 nm together with forces 138 of 125 ± 50 pN, while C5a tips were found to interact specifically with receptors more buried 139 into the membrane (height of 1.7 ± 0.5 nm) and interacting with forces of 110 ± 45 pN. 140 Together, these results confirm that the receptor orientation can be determined using only 141 their protrusion heights as those are significantly different ( Figure 2K). An overlay of the AFM 142 topography and the specific adhesion events recorded (colored pixels) on the same area with 143 both functionalized tips (C5a tip in red and tris-NTA-Ni 2+ tip in green) reveals the identity of the 144 side exposed to the tip ( Figure 2L). Receptors having a protrusion height under a threshold of 145 1.75 nm were encircled. Together, these data confirm the possibility to identify with a high-146 probability (> 95%) the receptors oriented in their native state. Therefore, this height criterion 147 will be used in the following force spectroscopy experiments to validate our measurements and 148 to only probe native state receptors.

149
C5a is a high-affinity agonist of C5aR 150 Next, we wanted to characterize C5a ligand binding to C5aR ( Figure 3A). To this end, we found. The dissociation constant was calculated using the relation ΔGbu = kBT x ln(0.018 Kd) with 178 0.018 l mol -1 being the partial molar volume of water. The Kd estimated using our single-179 molecule method is the same order of magnitude as previously reported values, ranging from 180 1-10 nM, based on radioactive ligand binding assays (Gerber, Meng et al., 2001;Robertson, 181 Rappas et al., 2018). These results highlight that FD-based AFM was suitable to quantify the 182 kinetic and thermodynamic binding of a large ligand interacting with its receptor via multiple 183 orthosteric sites.

185
Although it is thought that C5a binds C5aR through two orthosteric sites, a crystal structure of 186 the C5a-C5aR complex is missing. To better investigate the key residues responsible for the high 187 affinity interaction, we turned our attention to PMX53, a well-known C5aR full antagonist that

203
The Y258-R6PMX53 cation-π interaction was broken (R6PMX53 CZ and Y258 ring-centroid distance 204 > 6.0 Å) halfway through the simulation but the two residues remained close to each other 205 ( Figure 4G). Disruption of the cation-π interaction allowed R6PMX53 to interact with D282 in a 206 head-on manner ( Figure 4G and Figure S3C). W5PMX53 and R6PMX53 saddled Y258 but did not 207 interact directly during the production run ( Figure 4G and Figure S3C). profile of the pulling simulation is presented in Figure 4B. The application of force on the PMX53 219 molecule led to a gradual build-up of force until a critical point was reached that was sufficient 220 to break the key intermolecular interactions to allow the dissociation of the bound molecule.
The plot showed two such critical points, a minor drop in force around t = 308 ps and a major   C5aR D282 is critical for binding thermodynamics 290 Finally, we studied the PMX53 binding to the C5aR D282A mutant by FD-based AFM ( Figure S5C).  C5a C-terminus weakly binds C5aR effector site 299 As PMX53 is a peptide that mimics the structure of the C-terminal segment of C5a, we C5aR∆Tyr was measured and the dependence of the rupture force with the loading rate was 306 plotted in the DFS graph in Figure 5B. A non-linear dependency of the rupture force with the 307 loading rate was again observed and the FNdY model was used to fit the data. We extracted an 308 equilibrium force Feq of 32 ± 14 pN, a binding equilibrium free energy ΔGbu of -4.7 ± 3.4 kcal 309 mol -1 and a receptor-ligand half-residence time, τ0.5 of 0.5 ms. The calculated ΔGbu corresponds 310 to a dissociation constant Kd of ≈ 20 mM and corresponds to a surprisingly low-affinity, in 311 contrast to the high-affinity binding of C5aR WT . To further increase the accuracy of the 312 extracted parameters, we also recorded specific binding events at lower LRs by oscillating the cantilever at frequencies of 1-10 Hz ( Figure 5C). The measured forces over the low LRs regime 314 (10 2 -10 5 pN . s -1 ) align well with binding forces obtained at higher LRs (10 5 -10 7 pN . s -1 ). The  The rigid core of C5a has a low affinity for the C5aR binding site 321 As the binding of C5a to the C5aR ES alone fails to explain the high-affinity interaction, we also    Our study addressed the complex binding process of a large ligand to a GPCR. C5a, a 74-aa 424 glycoprotein binds to C5aR through two distinct and physically separated binding sites, namely 425 the effector and binding site (Siciliano, Rollins et al., 1994). While the existence of the two-site 426 binding motif has been previously reported (Siciliano, Rollins et al., 1994), the functional 427 relationship between the two sites was missing until now. Our method enabled, for the first 428 time, to probe multiple ligand binding sites at the sub-site level in order to study their 429 respective contribution to the overall binding. We demonstrated that both orthosteric ligand 430 binding sites interact with the ligand with a low affinity when working independently.

431
Interestingly, when acting in a concerted manner, the interaction rises into a high-affinity 432 interaction, suggesting a cooperativity between both orthosteric binding sites ( Figure 6G). This 433 cooperativity effect resulting from multiple binding site is supported by the theory that predicts  Through our experimental approach combining AFM and simulations, we were able to capture 437 the "cryptic" binding pockets of C5a into C5aR and to reconstruct the binding free-energy 438 landscape for this complex binding mechanism. The importance of the D282 at the extracellular 439 face of TM7 was also put in evidence. Although already predicted to form an important 440 interaction with R74 of C5a, this interaction remained so far a mystery since some D282 441 mutants were showed to be relatively unresponsive to C5a but sensitive to C5a des-Arg and

567
In FD-based AFM measurements, the AFM cantilever is oscillated in a sinusoidal manner well 568 below its resonance frequency, while the sample surface is contoured pixel-by-pixel. For each 569 approach-retraction cycle of the oscillating cantilever, a force-distance curve is recorded.

570
Multiparametric FD-based AFM height, Young's modulus and adhesion maps are then obtained 571 from a pixel-by-pixel reconstruction of the acquired data. Overview FD-based AFM maps were 572 acquired by scanning the sample at 1 Hz and a resolution of 512 x 512 pixels, using a force 573 setpoint of ≈150 pN, a 2 kHz oscillation frequency and a peak-to-peak oscillation amplitude of 574 100 nm. Adhesion maps were recorded using a scan rate of 0.2 Hz and 256 x 256 pixels. The 575 functionalized AFM cantilever was oscillated at 0.25 kHz with peak-to-peak oscillation 576 amplitudes of 100 nm. To vertically oscillate the AFM tip at 1-10 Hz, FD-based AFM was 577 conducted in the ramp mode with a force setpoint of 200 pN, an approach velocity of 1 μm . s -1 , 578 retraction velocities of 0.5-2 μm . s -1 , a ramp size of 150 nm and no surface delay.

581
Several control experiments were designed to ensure that the measured interactions were 582 indeed specific and the functionalization of the AFM tip successful. Adhesion maps of C5aR 583 reconstituted samples were imaged with unmodified or ethanolamine-coated AFM tips ( Figure   584 S1A, D). We also tested C5a ligand binding before and after injection of free C5a on the sample 585 surface ( Figure S1C,D). In another approach, tris-NTA binding to C-terminal of C5aR was tested 586 in the presence of 10 mM EDTA ( Figure S1E,F).     how the orientation of single C5aR particles can be identified using our multiplex probing method. Data are 885 representative of at least three independent experiments. Data in K is displayed as mean ± S.D. and the ANOVA 886 OneWay Tukey test was used to report the statistical significance: *** p<0.001. Data are representative of three 887 independent experiments. the C5a-C5aR free energy landscape. A ligand-receptor bond can be described using a simple two-state model, 899 where the bound state resides in an energy valley and is separated from the unbound state by an energy barrier. 900 The transition state must be overcome to separate ligand and receptor. τ -1 (F) and τ -1 (0) are residence times linked 901 to the transition rates for crossing the energy barrier under an applied force F and at zero force, respectively. ΔGbu 902 is the free-energy difference between bound and unbound state. (F) Force-volume (FV)-AFM and FD-based AFM 903 were used to explore low LRs and high LRs, respectively. For each pixel of the topography the tip is approached 904 and retracted using a linear (FV-AFM) or oscillating movement (FD-based AFM). (G) A force-distance curve can be 905 displayed as a force-time curve, from which the loading rate can be extracted via the slope of the curve just before 906 bond rupture. (H) DFS plot showing the loading rate-dependent interaction forces of the C5a ligand with C5aR. 907 Data combines rupture forces obtained at lower LRs (10 2 -10 5 pN . s -1 ) and higher LRs (10 5 -10 7 pN . s -1 ) Fitting the data 908 using the Friddle-Noy-de Yoreo model (thin grey line) provides average Feq, ΔGbu and residence time (τ0.5) values 909 with errors representing the s.e.m. Each circle represents one measurement. Darker shaded areas represent 99% 910 confidence intervals, and lighter shaded areas represent 99% of prediction intervals. For each condition, data are 911 representative of at least three independent experiments. 912