Comparison of the GHSR Dynamics in the Apo and Ghrelin-Bound State—Detector Analysis Identifies Dynamic Hotspots

To first identify dynamically interesting regions in GHSR, we compare the MD trajectories of apo and ghrelin-bound GHSR. As an indicator of local dynamics, we investigate the backbone H−N bond reorientational dynamics using detector analysis, which allows the timescale-specific characterization of motion.^16a-16c Detector analysis returns amplitudes of motion for seven timescale windows from ps to μs (termed ρ₀–ρ₆) as detailed in Figure 1. The final detector of the MD analysis (ρ₆) is effectively an order parameter, increasing in size where less motion occurs, or where motion is too slow to be captured by the 35.5 μs MD trajectories. All other detectors increase when amplitude increases within their window.

From ρ₆ (Figure 1B), we obtain an overview of the total motion of GHSR. Significant dynamic differences occur between apo and ghrelin-bound GHSR, with some more pronounced alterations occurring in the loops and in TM5. We also notice that parts of TM6 and TM7 are more dynamic in the apo form. The oscillating response of ρ₅ and ρ₆ (Figure 1A) observed in this region is characteristic of overall helix motions,¹⁸ where the local influence of the motion depends on the orientations of the individual H−N bonds.

Only considering the total motion, however, underestimates the differences between apo and ligand-bound states. If we view ECL2 in particular, we find that while the total motion is similar (ρ₆), large differences are observed in ρ₀–ρ₅, as better seen in Figure 2A. In the apo form, motion is broadly distributed across ρ₀-ρ₅, whereas in the ghrelin-bound form, dynamics are more concentrated in ρ₅ (correlation time ≈5 μs), indicating similar amplitude of motion but much slower dynamics in ghrelin-bound GHSR. We also observe changes in the ECL3 dynamics, where amplitudes are similar for apo and bound GHSR for ρ₀–ρ₄, but the apo form exhibits significantly higher dynamics for ρ₅.

The differences in TM5–TM7 are also of particular interest to GHSR function, noting that it is well established that TM6 must move outwards for G-protein binding upon ligand binding.^{4, 7, 8, 11b, 19} An examination of detector responses in Figure 1B/C reveals that the motional differences are concentrated in ρ₅, corresponding to ≈5 μs. For a better view, we zoom into this region and timescale in Figure 2B. In the apo form, the extracellular part of TM5 exhibits the most motion outside of loop regions. Although having lower amplitude than TM5, TM6 and TM7 are also more dynamic in the apo form than in ghrelin-bound GHSR. Note that the increased ECL3 dynamics observed in Figure 2A (ρ₅) are almost certainly coupled to this motion, since ECL3 sits between TM6 and TM7. Of particular interest for GHSR function, the intracellular part of TM5 and TM6 exhibit increased dynamics compared to the bound form. As TM6 must open to accommodate G-protein binding, these increased dynamics may eventually allow apo-GHSR to transform into the active state, although on average, the intracellular part of TM6 is located significantly further outward in the bound form, as seen in Figure 2C.

Next, we investigate whether the dynamics changes observed via MD are consistent with experiments, determine what interactions may be relevant for bringing about these dynamic changes via cross-correlation analysis, and investigate the specific structural changes occurring in GHSR via PCA.

Experimental Insights in the Dynamics of GHSR Activation from Three Histidine Residues

Any change in the local environment of a given residue is reflected as a variation of its ¹³C NMR chemical shift.²⁰ Conformation-dependent NMR chemical shifts have been exploited to characterize the activation of several GPCRs.²¹ From the detector analysis of the MD trajectories, ECL2 and TM6 provide hotspots of dynamic alterations observed during GHSR activation. For an experimental validation of the structural dynamics in GHSR, biophysical probes in these receptor segments would be useful. Analyzing the receptor sequence reveals that histidine (His) residues are prominently distributed in these dynamic hotspots. Coincidently, the native GHSR sequence only contains three His residues, putatively resulting in the necessary low spectral complexity without the need for further mutagenesis to simplify the NMR spectra. Thus, His residues may report conformational changes of GHSR upon activation. H258^6.30 and H280^6.52 are found in the lower and upper parts of TM6, respectively. H186^ECL2 is part of the ECL2, which showed different dynamics in the apo and the ligand-bound states.

The ¹³Cα-¹³Cβ region of the DARR NMR spectrum of apo GHSR recorded under magic-angle spinning (MAS) conditions is shown in Figure 3A. ¹³C/¹⁵N-His labeled GHSR was produced using cell-free (CF) expression and reconstituted into DMPC membranes in the apo and the ghrelin-bound state.²² Obviously, more than the expected three NMR signals are detected. The complication arises from the C-terminal 6× His-tag attached for purification, which is also labeled in CF synthesis. Nonetheless, we are able to assign the spectrum, using a strategy based on three pillars: first, the chemical shift index (CSI).²³ This empirical analysis suggests that signals with ¹³Cα chemical shifts from 56–60 ppm represent α-helical structures and could belong to H258^6.30 and H280^6.52. Similarly, the NMR signals with ¹³Cα chemical shifts below 56 ppm, coinciding to random coil structures, could correspond to H186^ECL2 and the His from the purification tag. Second, ShiftX²⁴ and SHIFTX2²⁵ predicted the chemical shifts of the His residues based on the crystal structure of apo GHSR and is shown as crosshairs in Figure 3A. Third, a mutagenesis approach was adopted and three GHSR mutants (H186A, H280Q, and H258Y) were expressed, reconstituted and investigated by NMR (Figure 3B). H186A and H280Q were reported to be functional before,²⁶ while H258Y is a known natural mutation of GHSR.²⁷ All mutants showed good activity in vitro (Supporting Information Figure S1). While each mutation leads to some chemical shift change, areas lacking spectral intensity are identified yielding the peak assignment in good agreement with the chemical shift prediction and the general secondary structure assignment. Additional spectral intensity is assigned to the His-tag, which does not superimpose with the signals from the native His residues. With this assignment in hand, the spectral signature of activation-induced alterations in GHSR are characterized. Figure 3C compares the DARR NMR spectra of GHSR in the apo and ghrelin-bound states. Upon activation, both the Cα NMR peaks of H258^6.30 and H280^6.52 move downfield by ≈1.5 ppm, while the NMR signal of H186^ECL2 moves slightly upfield by 0.4 ppm. It was not possible to discern changes in the chemical shifts of sidechain carbons due to line broadening and lack of sidechain assignments.

The chemical shift changes of H258^6.30 and H280^6.52 are consistent with an increase in helical character upon ghrelin binding. While we expect these residues to be helical in both the apo- and bound forms, the increased mobility in the apo form will allow the sampling of more backbone torsional angles, and thus the averaged chemical shift will tend more towards random-coil, i.e., lower, chemical shifts. The slight alteration in peak position of H186^ECL2 is consistent with an involvement of ECL2 in the structural alterations of GHSR upon activation as seen in the crystal/cryo-EM structures,^{4, 7, 8} and is also identified as the strongest detector responses in our analysis (Figure 1). However, for H186^ECL2 we see a broad distribution of dynamics (≈100 s of ps to several μs) for both the apo and ghrelin-bound forms in ECL2, although with notably lower amplitude in the bound form for ρ₀–ρ₄. Interestingly, the change in dynamics in this range is mostly compensated by additional motion near 5 μs in ghrelin-bound GHSR (Figure 2A). Since the differences in detector responses indicate primarily a change in correlation time of motion rather than a large change in amplitude, the populations of torsional angles sampled may not be vastly altered, and thus the chemical shift also only sees minor changes. Note that microsecond motion can also have an influence on linewidths, via T₂ relaxation, although with the signal-to-noise acquired, we were not able to determine linewidth differences to sufficient precision. As sample preparation improves, we hope to measure R_1ρ to experimentally determine the timescale of slow motion, observed near 5 μs via MD.

Agonist-GHSR interactions via cross-correlation analysis

In the previous sections, we have seen significant alteration of dynamics in ECL2 and TM5–7. In particular, we see a shift in timescale of the motion of ECL2 and also the rigidification of TM5–TM7 upon ghrelin binding. This is accompanied by increased helical character of H258^6.30 and H280^6.52, seen via NMR spectroscopy. In order to better understand the change in dynamics, we first investigate interactions between ghrelin and GHSR via iRED cross-correlation analysis²⁸ for motions around 5 μs correlation time (ρ₅). This analysis yields correlation coefficients for motion falling within the ρ₅ detector window. The results are shown in Figure 4 with site resolution for ghrelin. Note the timescale filter supplied by detector analysis is critical to obtaining the cross-correlation coefficients, since the total cross-correlation coefficients (motion from all timescales) are contaminated by significant contributions from fast, local dynamics (see Supporting Information Figure S2).

To understand the large change in ECL2 dynamics (Figure 2B), we first consider that the available cryo-EM data allows one to attribute changes in ECL2 dynamics to two single direct contacts between ghrelin and GHSR. In the ghrelin/GHSR/Gq complex, a hydrogen bond is identified between S6^Gh and E197^ECL2,⁷ whereas in the ghrelin/GHSR/Go complex, a hydrogen bond between R15^Gh and E187^ECL2 is described.⁹ Interestingly, our correlation analysis does not find an overall strong correlation between the motions of ghrelin and ECL2, but the specific correlation between S6^Gh and E197^ECL2 is noticeably higher than the correlation of S6^Gh to other nearby residues (Figure 4).

We are furthermore interested in understanding the rigidification of TM5–TM7. We find correlation of ghrelin motion with the motion of GHSR, especially TM2 and TM3, and to a lesser extend ECL2, TM6, and TM7. A weaker correlation between S6 and TM5 can also be observed. For quantification, we counted the number of correlation coefficients exceeding 0.5 for the N-terminal and C-terminal segments of ghrelin with residues from GHSR confirming that the motions of the TM helices are predominantly correlated with the 7N-terminal residues of ghrelin (Supporting Information Figure S3). Interestingly, some of the strongest correlation with ghrelin residues 9–15 occur for the intracellular side of TM6 and the middle of TM1 and TM7. The strong correlation between ghrelin motion and TM5–TM7 motion indicates somewhat rigid connections between these regions, and can explain some of the reduction in dynamics in the helices. Assuming that TM6 dynamics and translocation is crucial for GHSR activation and its basal activity, this analysis also suggests that the sites where ghrelin residues S6^Gh, H9^Gh, Q10^Gh, and Q13^Gh interact with GHSR could be critical and may be targeted in rational drug development.

Accompanying the change in TM5–TM7 dynamics is the opening of the lower part of TM6 (Figure 2C); note that while TM6 is highly dynamic in the apo form, the simulation does not sample structures that would allow G-protein binding. This is in contrast to the bound form, where TM6 remains fixed in the open position, allowing space for G-protein binding. In addition to opening of TM6 in the bound form, TM5 partially unwinds near P224^5.50, accompanied by a significant increase in dynamics around that residue (Figure 2B, right). The cryo-EM structure reveals that the first 7 residues of ghrelin penetrate the GHSR binding pocket and form molecular contacts with key residues in TM2, TM4, TM6, and TM7.^{7, 9} One assumes that the reduced dynamics of TM6 is caused by contacts described for TM6 (G1^Gh-F279^6.51and S3^Gh-F286^6.58), and indeed our analysis shows stronger correlation between ghrelin and TM6 in between these positions (Figure 4, see S3^Gh), while also exhibiting increased correlation away from these interactions, on the intracellular side of TM6. Note that while we observe correlation to TM6, there are actually more significant correlations to TM2 and TM3 (Supporting Information Figure S3); although the motions of TM2 and TM3 are correlated with ghrelin dynamics, there are no noteworthy alterations in the amplitude of motions of these segments between apo and ghrelin-bound states.

A Cluster of Aromatic Sidechains Mediates Dynamic Changes of GHSR Initiated by the Toggle Switch W276^6.48

While correlated motion between ghrelin and GHSR likely play an important role in the dynamic alterations upon ghrelin binding, disruption of specific interactions within GHSR may also have an impact. In particular, GHSR activation is initiated by interaction of G1^Gh with F279^6.51 as part of an aromatic cascade formed by W276^6.48, F279^6.51, and F312^7.42 (referred to as WFF cluster).⁹ The swing of the toggle switch, W276^6.48, leads to a structural change of F279^6.51 initiating a cascade of relocations of other highly conserved residues. These are connected to the polar network associated with the interaction of ghrelin with F279^6.51 which is connected to R283^6.55 via cation-π interaction, and is located directly above F279^6.51.⁹ For bovine rhodopsin, a cluster of closely packed aromatic sidechains mediate activation, and structural constraints of W276^6.48 were found to be weakened upon activation.²⁹ Aromatic residues are also important for the activation of other peptide-ligand receptors.³⁰ In our analysis, we have both observed significant correlation of motion between ghrelin and GHSR, especially in TM2, TM3, ECL2, TM6, and TM7, primarily towards the center of the protein, with some correlation to TM6 on the intracellular side (Figure 4, Supporting Information Figure S3), and also the reduction of motion in TM5–TM7, near W276^6.48. Then, we wonder if these dynamics changes are related to the closely packed aromatic sidechains of TM3 and TM5–7.

As a first step, we analyze the amplitude of motion of the W276^6.48 sidechain in Figure 5A (left axis), where indeed, W276^6.48 dynamics increase significantly upon ghrelin binding, especially for longer correlation times (ρ₅). Then, we perform a correlation analysis of motion of the W276^6.48 sidechain and nearby aromatic sidechains via iRED correlation analysis.²⁸ We see motion of several aromatic sidechains surrounding W276^6.48 become more correlated after ghrelin binding, most notably at H280^6.52. However, other correlations are significantly reduced in the ligand-bound state. Similar trends in correlation are also observed at shorter correlation times. The change in correlation appears to primarily be an effect of breaking the interaction of W276^6.48 with F279^6.51, which correspondingly shows a loss in correlation with W276^6.48.⁹ The breaking of this interaction is then accompanied by a decrease in correlation to some residues on the slowest timescale (ρ₅, ≈5 μs), such as such as F221^5.47, F222^5.48, and F226^5.52. On the other hand, especially for faster motion, correlation to many other residues increases, potentially because the broken interaction with F279^6.51 allows W276^6.48 to move more freely with other neighbors (Figure 5B).

While we have previously noted that rigidification of TM6 can lead to decreased sampling of torsional angles in the backbone, and therefore greater helical character at H280^6.52, as observed via NMR chemical shift, the significant change in interactions in the vicinity of W276^6.48 may also have an impact on chemical shift. An interesting change is the hydrogen bond between H280^6.52 and S217^5.43 in the active cryo-EM structure⁷ that it is not present in the inactive structure.⁴ S217^5.43 is an important residue as it stabilizes R283^6.55 by a hydrogen bond. R283^6.55 forms a salt bridge with E124^3.33 in order to separate the ligand-binding pocket in two cavities.⁹ Like the interaction between H280^6.52 and the S217^5.43, the interaction between S217^5.43 and R283^6.55 was found only in the agonist-bound structure, suggesting that H280^6.52 could also be important for the rearrangement of the polar network crucial for the contraction of the agonist-binding pocket of the GHSR.⁹

Rugged Energy Landscapes Describe the Apo and Ghrelin-Bound State of GHSR

In order to connect the above results describing reorientational dynamics to the sampling of distinct structures of GHSR, we performed a PCA of the GHSR backbone (N, C′, Cα). PCA separates translational motion in the MD trajectory into statistically independent modes, and sorts them from largest absolute variance to smallest, allowing one to determine the largest modes of motion for the apo and ghrelin-bound MD simulations. Histograms of the largest principal components often reveal that a protein jumps between a few distinct configurations. In Figure 6, we show 2D histograms as functions of the first four principal components, and select several of the most populated configurations of apo and ghrelin-bound GHSR (markers in Figure 6, left). To the right, we show the corresponding structures. Note that some of these structures occur late in the 35.5 μs trajectories (Supporting Information Figure S4), justifying long MD simulations. The largest difference between apo and bound dynamics is the rigidification of TM5–TM7 on the extracellular side upon ghrelin binding (Figure 6B). In apo GHSR, PCA indicates that this motion in apo GHSR is an outward motion of TM5 on the extracellular side accompanied by smaller either outwards or sideways motions of TM6 and TM7 on the extracellular side (Figure 6A, top, left). Smaller motions of these helices are also observed on the intracellular side. Multiple local maxima in the histograms indicate that several distinct configurations exist for TM5–TM7. This motion manifests as the reorientational dynamics in Figure 2B. It is critical to note that PCA does not reveal displacement of TM6 large enough to accommodate G-protein binding in the apo form. With a long enough trajectory, the protein should eventually transition to the active state, but even with 3×35 μs of simulation, the transition probability for activation remains low. In the ghrelin-bound form, TM6 is fixed in the open position. In contrast to the apo form, the largest structural changes in ghrelin-bound GHSR involve significant reorientation of ECL2, in addition to some small re-positioning of TM6 and of intracellular regions of TM5, TM7, and helix 8 (Figure 6B). Note that significant motion for ECL2 is also observed in apo GHSR (Figure 6A).

Thus far, we have characterized site-specific dynamics in apo and ghrelin-bound GHSR, including evaluation of the degree of cross-correlation of motion. We have also looked at overall structural changes of the GHSR backbone using PCA. Of particular importance appears to be the rearrangement of helices TM5–TM7 in apo GHSR, observed both as high-amplitude dynamics using detector analysis (Figure 2) and as significant structural differences using PCA (Figure 6). By observing the values of PC0 and PC1 as a function of time in the three apo GHSR trajectories (Supporting Information Figure S4), we may estimate an approximate path through the PC0/PC1 histogram (Figure 6A), shown in Figure 7A. Figure 7B then estimates an energy landscape based on density of states at each point along the path (details found in the Supporting Information). Note that the MD trajectories are not long enough to fully equilibrate, so that the energy landscape should not be regarded as quantitative; we can identify the location and corresponding structures of local minima along the path, but the depths of the energy wells will not be accurate beyond a rough estimate (≈5 kJ mol⁻¹). In particular, the population will be biased towards the starting state, thus underestimating its energy (indicated as grey dashed lines in Figure 7B).

Then, we can observe how apo GHSR progresses through a series of different configurations. In the starting configuration of the MD simulations (approximately located at the grey “X” in Figure 7A/B), TM5 is closed on the extracellular side of the protein, similar to the average configuration of the bound state. For comparison, see Figure 7C (grey: apo, blue: ghrelin-bound). From that initial state, TM5 opens on the extracellular side, either in coordination with an outward motion of TM6/TM7 (Figure 7D, top, orange), or a more sideways motion of TM6/TM7 (leftwards in the top view in Figure 7D, bottom, green). The opening of TM5 is significantly larger when occurring in concert with this sideways motion, and samples a number of metastable states (e.g. Figure 7D, middle plots, purple and red). Furthermore, it is accompanied by a small outward motion of TM6 on the intracellular side, although the extent of the motion is far from the amount that would be required for G-protein binding. In general, we see a very rugged energy landscape, characterized by many intermediate plateaus or wells in the free energy.

Finally, we may perform a detector analysis of the first two principal components of apo GHSR (Figure 7E), where we observe that these principal components impact primarily ρ₅, i.e. motion with correlation times near ≈5 μs. PC2-PC4 also make significant contributions in this range, whereas smaller principal components contribute at shorter correlation times (see Supporting Information Figure S6). This analysis then confirms that much of the motion observed near 5 μs with detector analysis (ρ₅, Figure 2) is due to exchange between the structures observed in the histograms of PC0 and PC1.