Specific Engineered G Protein Coupling to Histamine Receptors Revealed from Cellular Assay Experiments and Accelerated Molecular Dynamics Simulations

G protein-coupled receptors (GPCRs) are targets of extracellular stimuli and hence occupy a key position in drug discovery. By specific and not yet fully elucidated coupling profiles with α subunits of distinct G protein families, they regulate cellular responses. The histamine H2 and H4 receptors (H2R and H4R) are prominent members of Gs- and Gi-coupled GPCRs. Nevertheless, promiscuous G protein and selective Gi signaling have been reported for the H2R and H4R, respectively, the molecular mechanism of which remained unclear. Using a combination of cellular experimental assays and Gaussian accelerated molecular dynamics (GaMD) simulations, we investigated the coupling profiles of the H2R and H4R to engineered mini-G proteins (mG). We obtained coupling profiles of the mGs, mGsi, or mGsq proteins to the H2R and H4R from the mini-G protein recruitment assays using HEK293T cells. Compared to H2R–mGs expressing cells, histamine responses were weaker (pEC50, Emax) for H2R–mGsi and –mGsq. By contrast, the H4R selectively bound to mGsi. Similarly, in all-atom GaMD simulations, we observed a preferential binding of H2R to mGs and H4R to mGsi revealed by the structural flexibility and free energy landscapes of the complexes. Although the mG α5 helices were consistently located within the HR binding cavity, alternative binding orientations were detected in the complexes. Due to the specific residue interactions, all mG α5 helices of the H2R complexes adopted the Gs-like orientation toward the receptor transmembrane (TM) 6 domain, whereas in H4R complexes, only mGsi was in the Gi-like orientation toward TM2, which was in agreement with Gs- and Gi-coupled GPCRs structures resolved by X-ray/cryo-EM. These cellular and molecular insights support (patho)physiological profiles of the histamine receptors, especially the hitherto little studied H2R function in the brain, as well as of the pharmacological potential of H4R selective drugs.


Introduction
In the human body, the neurotransmitter histamine interacts with four subtypes of histamine receptors (H 1-4 R) that are all classified as rhodopsin-like class A G proteincoupled receptors (GPCRs) [1]. As the largest membrane protein superfamily [2,3], GPCRs (5-HT 2A R) [57] receptors so far. Recent years have seen remarkable advances in the structural determination of GPCR-G protein complexes [58,59]. In 2020 alone, new structures of 34 GPCR-Gs complexes, 19 GPCR-Gi/o complexes, and one GPCR-Gq/11 complex were published (www.gpcrdb.org, access date 8 October 2021 [58]), providing valuable atomic-level insights into the binding interface of GPCRs with G proteins from different major families. Another milestone in histamine receptor research was the resolution of the first active structure of the histamine H 1 receptor in complex with Gq in 2021 [60]. Nevertheless, static structures alone do not allow us to map the dynamics of a GPCR-G protein interaction, and only complexes of primary coupling GPCRs and G proteins are available so far. Thus, the dynamic mechanism of specific GPCR-G protein interactions remains poorly understood. To fill this gap, computational approaches have been developed to model the dynamic GPCR-G protein interactions [61,62]. For instance, Flock et al. provided a bioinformatics approach to determine a selectivity barcode (patterns of amino acids) of GPCR−G protein coupling [42]. More commonly, molecular dynamics (MD) simulations have been used to explore the conformational changes and free energy landscapes of GPCR-G protein interactions, ideally combined with complementary experiments [63][64][65]. However, conventional MD (cMD) simulations often suffer from insufficient sampling of GPCR-G protein interactions due to limited simulation timescales. Thus, enhanced sampling methods have been applied to improve the simulations of GPCR-G protein interactions [62,66]. Among these methods, Gaussian accelerated molecular dynamics (GaMD) is a robust method that allows for unconstrained enhanced sampling and free energy calculations of large biomolecules [67][68][69][70]. GaMD has been applied to successfully simulate protein folding [68,71], protein-ligand binding and unbinding [67,68,72], GPCR activation [72], and binding to a G protein mimetic nanobody [73]. In the latter study, the nanobody binding pathway to the muscarinic acetylcholine M 2 (M 2 R) receptor has been investigated, demonstrating that the intracellular loops play a key role in nanobody recognition and binding [73]. Moreover, GaMD has been used to identify the coupling mechanisms of adenosine A 1 (A 1 R) and A 2A (A 2A R) receptors to Gs and Gi proteins [74]. Protein flexibility and complementary residue interactions at the protein interface have revealed that the A 1 R preferred Gi coupling and the A 2A R coupled to Gs and Gi [74].
In the present study, we have combined cellular assay experiments and GaMDenhanced sampling simulations to investigate coupling profiles of the H 2 R-and H 4 R-G protein complexes. Three types of engineered G proteins, mGs, mGsi, and mGsq, were used to characterize the coupling profiles of each receptor. We have explored the functional responses of both receptors upon activation by the endogenous ligand histamine using a recently published mini-G protein recruitment assay [52]. Furthermore, we performed all-atom GaMD simulations on the H 2 R and H 4 R in complex with mGs, mGsi, and mGsq in explicit lipids and solvent. In all six systems, the HRs were bound to the endogenous ligand histamine. The GaMD simulations have allowed us to characterize the structural flexibility and low-energy conformations of the HRs in complex with the mGs, mGsi, and mGsq proteins. The combination of cellular assay experiments and GaMD simulations has provided important mechanistic insights into the selective coupling of engineered G proteins to the HRs.

Functional Characterization of Histamine at H 2 R-and H 4 R-Mini-G Protein Complexes
Recently, we have developed a dynamic split-luciferase based mini-G protein recruitment assay for the histamine receptor family [52] meeting the requirements of a proximal readout as well as a simple, robust, non-radioactive and homogenous performance [75]. In this assay, HEK293T cells express HR subtypes that are C-terminally fused to the small fragment of the NanoLuc (NlucC) and mini-G proteins N-terminally attached to the large fragment (NlucN). The complementation of the NanoLuc, and thus signal output, is enabled by the recruitment of the mini-G protein to the HR upon receptor activation by a ligand. In the present study, we applied the same concept to investigate the coupling profiles of the H 2 R and H 4 R to engineered G proteins. Therefore, transfectants expressing the H 2 R and H 4 R in combinations with mGs, mGsi, and mGsq were used to probe the functional responses to histamine. We observed that the H 2 R interacted with mGs, mGsi, and mGsq, whereas the H 4 R was selective for binding to mGsi ( Figure 1A,B). Notably, histamine was less potent (pEC 50 ) and effective (E max ) in H 2 R-mGsi (pEC 50 = 5.30 ± 0.06; E max = 26.77 ± 3.47) and H 2 R-mGsq (pEC 50 = 5.48 ± 0.04; E max = 29.56 ± 3.66) systems than in the H 2 R-mGs system (pEC 50 = 6.86 ± 0.04; E max = 100%), which represents the canonical interaction pair (Table 1). In other words, we observed a stronger binding of the H 2 R to mGs than to mGsi and mGsq. These observations were consistent with prior findings for the H 2 R of Okashah et al. using BRET-based heterotrimeric G protein sensors and of Inoue et al. using chimeric G proteins [39,44]. In our cell experiments, the H 4 R bound selectively to mGsi yielding a pEC 50 of 6.60 ± 0.10 and an E max of 100%, (per definition) for the endogenous ligand histamine. Similar results were reported in a previous study using the chimeric G protein approach, in which the H 4 R selectively coupled to members of the Gi family and only slightly to Gα16 of the Gq family [44]. ment assay for the histamine receptor family [52] meeting the requirements of a proximal readout as well as a simple, robust, non-radioactive and homogenous performance [75]. In this assay, HEK293T cells express HR subtypes that are C-terminally fused to the small fragment of the NanoLuc (NlucC) and mini-G proteins N-terminally attached to the large fragment (NlucN). The complementation of the NanoLuc, and thus signal output, is enabled by the recruitment of the mini-G protein to the HR upon receptor activation by a ligand. In the present study, we applied the same concept to investigate the coupling profiles of the H2R and H4R to engineered G proteins. Therefore, transfectants expressing the H2R and H4R in combinations with mGs, mGsi, and mGsq were used to probe the functional responses to histamine. We observed that the H2R interacted with mGs, mGsi, and mGsq, whereas the H4R was selective for binding to mGsi ( Figure 1A,B). Notably, histamine was less potent (pEC50) and effective (Emax) in H2R-mGsi (pEC50 = 5.30 ± 0.06; Emax = 26.77 ± 3.47) and H2R-mGsq (pEC50 = 5.48 ± 0.04; Emax = 29.56 ± 3.66) systems than in the H2R-mGs system (pEC50 = 6.86 ± 0.04; Emax = 100%), which represents the canonical interaction pair (Table 1). In other words, we observed a stronger binding of the H2R to mGs than to mGsi and mGsq. These observations were consistent with prior findings for the H2R of Okashah et al. using BRET-based heterotrimeric G protein sensors and of Inoue et al. using chimeric G proteins [39,44]. In our cell experiments, the H4R bound selectively to mGsi yielding a pEC50 of 6.60 ± 0.10 and an Emax of 100%, (per definition) for the endogenous ligand histamine. Similar results were reported in a previous study using the chimeric G protein approach, in which the H4R selectively coupled to members of the Gi family and only slightly to Gα16 of the Gq family [44]. Concentration-response curves of histamine obtained in a split-Nanoluciferase based mini-G protein recruitment assay using HEK293T cells transiently expressing either the H2R (A) or the H4R (B) in combination with mGs, mGsi, or mGsq. Presented data are of three independent experiments (N = 3), each performed in triplicate. (C) Free energy profiles of GaMD simulations with complexes containing either the H2R or the H4R in combination with mGs, mGsi, or mGsq. Distances of the D 3.32 (Cγ atom) and the amino group of histamine (Nα atom) as well as of the intracellular ends of TM3 and TM6 were used as reaction coordinates. TM3-TM6 distances were calculated using Cα, C, and N atoms of R 3.50 and E/A 6.30 . For each system, three independent GaMD simulations were used for analysis. (Labels: "B" indicate "Bound" lowenergy wells of fully active receptors bound to histamine, "I1" indicates "Intermediate" low-energy wells of receptor conformations with smaller and larger TM3-TM6 spacings. "S1", "S2", and "S3" illustrate low-energy states with different TM3-TM6 distances, in which histamine was separated from the conserved receptor residue D 3.32 ).

Figure 1.
Concentration-response curves of histamine obtained in a split-Nanoluciferase based mini-G protein recruitment assay using HEK293T cells transiently expressing either the H 2 R (A) or the H 4 R (B) in combination with mGs, mGsi, or mGsq. Presented data are of three independent experiments (N = 3), each performed in triplicate. (C) Free energy profiles of GaMD simulations with complexes containing either the H 2 R or the H 4 R in combination with mGs, mGsi, or mGsq. Distances of the D 3.32 (Cγ atom) and the amino group of histamine (Nα atom) as well as of the intracellular ends of TM3 and TM6 were used as reaction coordinates. TM3-TM6 distances were calculated using Cα, C, and N atoms of R 3.50 and E/A 6.30 . For each system, three independent GaMD simulations were used for analysis. (Labels: "B" indicate "Bound" low-energy wells of fully active receptors bound to histamine, "I1" indicates "Intermediate" low-energy wells of receptor conformations with smaller and larger TM3-TM6 spacings. "S1", "S2", and "S3" illustrate low-energy states with different TM3-TM6 distances, in which histamine was separated from the conserved receptor residue D 3.32 ).

Free Energy Profiles of HR-mG Protein Complexes Were Calculated from GaMD Simulations
In the absence of experimental structures, homology models were built in this study to generate initial complexes of the H 2 R and H 4 R with the mGs, mGsi, and mGsq proteins. A model recently published by Conrad et al. [76] and a model provided by GPCRdb (www.gpcrdb.prg, accessed on 10 August 2021 [77]) were used as simulation starting structures for the H 2 R and the H 4 R, respectively. A computational model of GaMD simulation systems is provided in the Supporting Material ( Figure S1). For each of the six systems, we performed three independent 1000 ns GaMD simulations (Table 2). Overall, the H 2 R and H 4 R remained in complex with the ligand histamine and their respective mini-G proteins during the simulations. Considered individually, the same regions were identified as structurally flexible in the complexes. With the exception of the respective mG-α5 helix, which was located in the HR-binding surface, the mini-G proteins mGs, mGsi, and mGsq were structurally more flexible than the membrane-bound H 2 R and H 4 R (see Supporting Material, Figure S2). For the receptors, the largest flexible regions were located at the N-and C-terminal ends and the extracellular loop (ECL) 2 (see Supporting Material, Figure S2A-C). For the H 4 R, a higher flexibility of the intracellular loop (ICL) 3, ECL2, and ECL3 was additionally observed (compare Supporting Material, Figure S2D-F). In all H 2 R and H 4 R complexes, the histamine molecule was located inside the orthosteric binding pocket, but different root-mean-square fluctuations (RMSFs) were observed. Generally, the RMSF of both the proteins and the ligand histamine were the lowest in the systems close to the natural complexes (H 2 R-mGs and H 4 R-mGsi), indicating a favorable complex formation (see Supporting Material, Figure S2). Comparing the H 2 R systems with the H 4 R systems, the replacement of mGs with mGsi and mGsq in the H 2 R systems resulted in a higher structural flexibility of the mG-binding surface (TM5, TM6, H8, and ICLs). In contrast, in H 4 R complexes, the structural variability at the binding surface was comparatively low when mGsi was replaced by mGs or mGsq. Only in H 4 R-mGs were higher fluctuations observed in the intracellular ends of TM5, TM6 (kinked TM6, Figure 2F), ICL2, and ICL3. More characteristic for the H 4 R-mGs and -mGsq complexes was the structural flexibility of the extracellular TM ends, which might be related to the structural flexibility of the histamine ligand. Combined with the cell assay results, we suggest that the more dynamic intracellular surface of H 2 R positively contributes to G protein binding, whereas the increased flexibility of the orthosteric binding pocket in the H 4 R-mGs and -mGsq complexes might be unfavorable for the complex formation.  50 ) and efficacies (E max ) of histamine obtained in the split-Nanoluciferase-based mini-G protein recruitment assay using HEK293T cells transiently expressing either the H 2 R or the H 4 R in combination with mGs, mGsi, or mGsq. Presented data are of three independent experiments (N = 3) each performed in triplicate. Binding energies (kcal/mol) of histamine obtained in the same complexes applying the MM/GBSA method to the GaMD trajectories. For each system, 60,000 frames of the top histamine cluster were analyzed. Averages ± SD are given.

Mini-G Protein Recruitment Assay
MM/GBSA pEC 50  In the starting structures of the simulations, histamine formed a salt bridge by its charged amino group with the conserved residue D 3.32 in the orthosteric binding pocket, similarly as postulated in the literature for the entire histamine receptor family [78]. During the simulations, this interaction performed different dynamics in the H 2 R and H 4 R systems. Whereas histamine was permanently bound to D 3.32 in the H 4 R-mGsi and -mGsq complexes, we were able to sample the disengagement of histamine in the H 2 R complexes and the H 4 R-mGs complex, however, to distinct extent (cf. schematic illustration of the reaction coordinates and detailed time courses in Supporting Material, Figures S3-S5, respectively). The lower abundance of the salt bridge in the H 2 R complexes was in good agreement with the generally lower binding affinity of histamine to H 2 R than to the H 4 R [6]. To investigate whether the presence of the salt bridge was related to the activation state of the receptors, we used the distance between the charge centers of the amino group in histamine and the conserved receptor aspartate D 3.32 (denoted as D 3.32 -HSM distance) as well as the receptor TM3-TM6 distance (measured between the Cα, C, and N atoms of residues D 3.50 and E/A 6.30 ) as reaction coordinates to calculate the free energy profiles ( Figure 1C). The outward movement of TM6 and thus a larger TM3-TM6 spacing is known as one of the key features of GPCR activation and is also suggested to determine G protein selectivity of Gs-and Gi-coupling GPCRs [79,80]. In all systems, the global minima were found for the fully histamine-bound states (D 3.32 -HSM distance of ≈3 Å). In the H 2 R-mGs system, the global minimum was at a TM3-TM6 distance of ≈19 Å (denoted as "B") ( Figures 1C and 2E). In the H 2 R-mGsi and -mGsq complexes, energetic minima at similar TM3-TM6 spacings were present (20 and 19 Å, respectively) ( Figures 1C and 2E). However, the global minima of these complexes were located at a slightly smaller TM3-TM6 distance of~17 Å (denoted as "I1") ( Figures 1C and 2F). The lower capacity of mGsi and mGsq to stabilize the fully active H 2 R conformation was in concordance with the lower E max values (~25-30%) in the mini-G recruitment assay ( Table 1). The global minimum of the canonical H 4 R system, H 4 R-mGsi as well as of H 4 R-mGsq were observed at a TM3-TM6 distance of 18 Å (Figures 1C and 3E). Smaller TM3-TM6 distances in Gi-compared to Gs-coupled GPCRs are related to the smaller α5 helix volume of Gi that requires less three-dimensional space [81]. Interestingly, in the H 4 R-mGs system, the TM3-TM6 distances was even smaller ("I1" state in Figure 2E). Strikingly, this could be attributed to a kinked TM6 helix indicating an unfavorable interaction between the H 4 R and mGs ( Figure 2F). Table 2. Summary of the boost potentials applied in the Gaussian accelerated molecular dynamics (GaMD) simulations. In all GaMD simulations, the total potential energy (E pot ) and the dihedral energy (E dihedral ) were boosted ("dual-boost"). Average ± SD of the corresponding boost potentials (∆V pot and ∆V dihedral ) are given for the different simulations of the H 2 R and H 4 R systems.

Different Affinities and Binding Conformations of Histamine in H 2 R and H 4 R Complexes
As described, we determined the efficacy (E max ) and potency (pEC 50 ) of histamine at the different complexes in the applied mini-G protein recruitment assay. It is worth noting that complicated relationships might exist between ligand affinity and efficiency [82], and we were interested in whether the experimentally determined pEC 50 values of histamine would correlate with its binding affinities in the respective HR complexes. Therefore, we calculated the binding energy of histamine in all six complexes using the molecular mechanics generalized Born surface area (MM/GBSA) approach [83][84][85]. Since we observed different dynamics in histamine binding (Figure 1), only frames of the top histamine cluster obtained from the GaMD simulations were used to determine the binding energies. In the H 2 R-mGs complex, histamine was bound ≈1.0 kcal/mol stronger than in the H 2 R-mGsi and H 2 R-mGsq systems (Table 1). This trend was in concordance with the pEC 50 values of the live cell assay (Table 1). In the H 4 R systems, histamine was most effectively bound to the H 4 R-mGsi complex with a difference of 7.88 kcal/mol or 7.42 kcal/mol compared to the systems containing mGs or mGsq, respectively (Table 1). Although the computational approach gave more pronounced differences in the binding energies of histamine at the H 2 R-mGs and H 4 R-mGsi complexes than the pEC 50 values obtained in the live cell assay, the calculated binding energies seemed to be consistent with the general dynamics of the systems. In the four complexes with weaker histamine binding (H 2 R-mGs, H 2 R-mGsi, H 2 R-mGsq, and H 4 R-mGs), the disengagement of histamine from D 3.32 could be detected ( Figure 1C). Combined with the lacking interaction in the live cell assay, the more pronounced difference in histamine binding energies might serve as an indicator that the H 4 R is difficult to couple with Gs and Gq. to Gs-coupled GPCRs are related to the smaller α5 helix volume of Gi that requires less three-dimensional space [81]. Interestingly, in the H4R-mGs system, the TM3-TM6 distances was even smaller ("I1" state in Figure 2E). Strikingly, this could be attributed to a kinked TM6 helix indicating an unfavorable interaction between the H4R and mGs ( Figure  2F).

Different Affinities and Binding Conformations of Histamine in H2R and H4R Complexes
As described, we determined the efficacy (Emax) and potency (pEC50) of histamine at the different complexes in the applied mini-G protein recruitment assay. It is worth noting that complicated relationships might exist between ligand affinity and efficiency [82], and we were interested in whether the experimentally determined pEC50 values of histamine would correlate with its binding affinities in the respective HR complexes. Therefore, we calculated the binding energy of histamine in all six complexes using the molecular mechanics generalized Born surface area (MM/GBSA) approach [83][84][85]. Since we observed different dynamics in histamine binding (Figure 1), only frames of the top histamine cluster obtained from the GaMD simulations were used to determine the binding energies. In the H2R-mGs complex, histamine was bound ≈1.0 kcal/mol stronger than in the H2R-mGsi and H2R-mGsq systems (Table 1). This trend was in concordance with the pEC50 values of the live cell assay (Table 1). In the H4R systems, histamine was most effectively bound to the H4R-mGsi complex with a difference of 7.88 kcal/mol or 7.42 kcal/mol compared to the systems containing mGs or mGsq, respectively (Table 1). Although the computational approach gave more pronounced differences in the binding energies of histamine at the H2R-mGs and H4R-mGsi complexes than the pEC50 values obtained in the live cell assay, the calculated binding energies seemed to be consistent with the general dynamics of the systems. In the four complexes with weaker histamine binding (H2R-mGs, H2R-mGsi, H2R-mGsq, and H4R-mGs), the disengagement of histamine from D 3.32 could be detected Recently, the first structure of an HR subtype in an active conformation, the H 1 R, was resolved using cryo-EM, providing insights at the molecular level into the binding mode of the endogenous agonist histamine [60]. The polar interactions with D 3.32 , Y 6.51 , N 5.46 , and T 3.37 postulated for the entire HR family were confirmed in this structure. Generally, we observed similar binding modes, with only slight differences, for histamine in the representative structures of the "Bound" low-energy states of the H 2 R and H 4 R complexes as in the H 1 R structure. In the H 2 R systems, the orthosteric binding pocket was formed by polar residues D 3.32 , C 3.36 , T 3.37 , D 5.43 , T 5.461 , Y 6.51 , and Y 7.42 and hydrophobic residues V 3.33 , F 5.47 , F 6.42 , W 6.48 , and L 7.40 ( Figure 3A-C). In all three H 2 R complexes, a salt bridge of the histamine primary amine group with D 3.32 was formed. The imidazole ring was stabilized by the polar residues D 5.43 and T 5.461 . In addition, T 3.37 is further involved in the latter in the H 2 R-mGs system ( Figure 3A). Unlike in the H 1 R structure, the sidechain of histamine in the H 2 R-mGs and H 2 R-mGsq complexes is directed toward the hydrophobic residues W 6.48 and L 7.40 . In the H 2 R-mGsi complex, the histamine sidechain was more likely to be attracted by the backbone of ECL2 residues (V 45.52 , Q177, and V178) ( Figure 3B).
In the H 4 R systems, the orthosteric binding pocket was surrounded by the polar residues D 3.32 , Y 3.33 , T 3.37 , E 5.461 , Y 5.51 , Q 7.41 , and the hydrophobic residues W 6.48 , F 7.38 , and W 7.42 (Figure 2A-C). As in the H 2 R complexes, D 3.32 formed a salt bridge with the primary amine of histamine. The imidazole ring was bound by polar residues Y 3.33 , T 3.37 , and E 5.461 . The structural details of the separated histamine states "S1" and "S2" of the H 2 R complexes and the H 4 R-mGs complex (cf. Figure 1C) were provided in the Supporting Material ( Figure S4). the H1R structure. In the H2R systems, the orthosteric binding pocket was formed by polar residues D 3.32 , C 3.36 , T 3.37 , D 5.43 , T 5.461 , Y 6.51 , and Y 7.42 and hydrophobic residues V 3.33 , F 5.47 , F 6.42 , W 6.48 , and L 7.40 ( Figure 3A-C). In all three H2R complexes, a salt bridge of the histamine primary amine group with D 3.32 was formed. The imidazole ring was stabilized by the polar residues D 5.43 and T 5.461 . In addition, T 3.37 is further involved in the latter in the H2R-mGs system ( Figure 3A). Unlike in the H1R structure, the sidechain of histamine in the H2R-mGs and H2R-mGsq complexes is directed toward the hydrophobic residues W 6.48 and L 7.40 . In the H2R-mGsi complex, the histamine sidechain was more likely to be attracted by the backbone of ECL2 residues (V 45.52 , Q177, and V178) ( Figure 3B). In the H4R systems, the orthosteric binding pocket was surrounded by the polar residues D 3.32 , Y 3.33 , T 3.37 , E 5.461 , Y 5.51 , Q 7.41 , and the hydrophobic residues W 6.48 , F 7.38 , and W 7.42 (Figure 2A-C). As in the H2R complexes, D 3.32 formed a salt bridge with the primary amine of histamine. The imidazole ring was bound by polar residues Y 3.33 , T 3.37 , and E 5.461 . The structural details of the separated histamine states "S1" and "S2" of the H2R complexes and the H4R-mGs complex (cf. Figure 1C) were provided in the Supporting Material ( Figure S4).  Figure 1). Contact residues within 4 Å of the ligand are highlighted as sticks (dark gray). The conserved salt bridge between the amino group of histamine and D 3.32 is given as a red, dashed line. (D) Side view of the H2R (pink). (E) TM3-TM6 distance of the "B" (fully active histamine-bound receptor) states of the H2R-mGs (salmon), H2R-mGsi (purple), and H2R-mGsq (green). (F) TM3-TM6 distance of the "B" state of the H2R-mGs (salmon) and the "I1" intermediate states of the H2R-mGsi (dark green), and H2R-mGsq (blue). Residues R 3.50 and E 6.30 were used to calculate the TM3-TM6 distance and are given as red spheres.  Figure 1). Contact residues within 4 Å of the ligand are highlighted as sticks (dark gray). The conserved salt bridge between the amino group of histamine and D 3.32 is given as a red, dashed line. (D) Side view of the H 2 R (pink). (E) TM3-TM6 distance of the "B" (fully active histamine-bound receptor) states of the H 2 R-mGs (salmon), H 2 R-mGsi (purple), and H 2 R-mGsq (green). (F) TM3-TM6 distance of the "B" state of the H 2 R-mGs (salmon) and the "I1" intermediate states of the H 2 R-mGsi (dark green), and H 2 R-mGsq (blue). Residues R 3.50 and E 6.30 were used to calculate the TM3-TM6 distance and are given as red spheres.

Residue Contacts at the Protein Binding Interface in the H 2 R and H 4 R Systems
To examine the coupling profiles of the H 2 R and H 4 R, we extracted residue contact networks for the representative structures of the receptor-mG complexes from GaMD simulations (Figures 4 and 5), noting that the quantitative number of contacts at the protein interfaces is not necessarily decisive due to the different volumes and thus available contact areas of the mG α5 helices. In general, the mGsi α5 helix formed a smaller number of contacts than mGs and mGsq, even in the canonical H 4 R-mGsi system.
The protein contact areas were very similar in the H 2 R complexes but rather different in the H 4 R complexes. The main contacts of the H 2 R with mGs, mGsi, and mGsq involved the receptors ICL2, TM3, TM5, TM6, and H8 ( Figure 4). In particular, the negatively charged residues of the mG α5 helices (D213 mG , D216 mG , D225 mGsi /E225 mGsq , and E227 mGs ) formed salt bridges with similar H 2 R residues leading to comparable mG binding positions. In all H 2 R systems, R 5.71 (TM5) formed salt bridges with the conserved residues D213 mG and D216 mG of the α5 helices in the mGs, mGsi, and mGsq ( Figure 4). In addition, the α5 hook of the mG proteins was also stabilized by salt bridges in the binding pocket. Accordingly, E227 mGs formed three salt bridges with receptor R 6.29 (TM6), R 8.48 , and R 8.51 (H8), and E225 mGsq formed a single salt bridge with receptor R 8.48 (Figure 4A,C). In addition to the salt bridge between D225 mGsi and R 12.51 (ICL1), the α5 hook in mGsi was stabilized by hydrogen bonds (R224 mGsi -R 8.48 and F229 mGsi -R 6.29 ) in a similar position as in the mGs and mGsq ( Figure 4B).
In contrast to the H 2 R systems, the contact surface in the H 4 R complexes were more versatile. In the canonical system, H 4 R-mGsi, a different contact network, was present compared with that in the H 2 R complexes ( Figure 5B). For the H 4 R, residues of ICL2 and TM5 contributed the main interactions with the mGsi. Similar to the H 2 R systems, mGsi residues D213 mGsi and D216 mGsi formed salt bridges with a receptor residue in TM5, R 5. 68 . In addition, D225 mGsi of the α5 hook formed a salt bridge with R 34.54 of ICL2.  Hydrophobic, polar, basic, and acidic residues are colored in yellow, green, light blue, and pink, respectively. Important amino acids responsible for polar contacts at the binding interface were highlighted by black circles. For the H2R, the BW numbers are used for TM helices and ICL1/2. For ICL3, original index numbers are given. For the mG proteins, the original index numbers are given.
Unlike in the H2R complexes, the α5 hook formed few contacts. Notably, it lacked strong interactions with TM6 and H8. In the H4R-mGsq complex, we detected a comparable contact network as in the H2R-mGsq complex with the main contacts from ICL2, TM5, TM6, and H8 ( Figure 5C). In this system, the conserved mG residue D213 mGsq formed a salt bridge with R 5.68 of the H4R. The α5 hook was further closely bound to the H4R via the two salt bridges E225 mGsq -R 34.54 (ICL2) and E225 mGsq -R 8.49 (H8) and by means of a hydrogen bond between V229 mGsq and R 6.32 . By contrast, the extremely high number of (hydrophobic) contacts in the H4R-mGs complex was particularly striking ( Figure 5A). In addition to the few specific interactions of the α5 hook (salt bridge of E227 mGs with R 6.32 (TM6) and K 8.48 (H8)), these occurred mainly with TM3, ICL2, TM5, and ICL3. Although the H4R comprised a larger ICL3 than the H2R, the ICL3 was little involved in the protein interface in the canonical system, H4R-mGsi. Moreover, after visual analysis of the trajectories, we assumed that mGs stuck to this region due to the large hydrophobic contacts. The interaction contacts were calculated using the representative low-energy conformation of the systems. Van der Waals, hydrogen bond, and salt bridge interactions are colored in gray, blue, and red, respectively. The line thickness is proportional to the number of contacts of the residue pairs. Hydrophobic, polar, basic, and acidic residues are colored in yellow, green, light blue, and pink, respectively. Important amino acids responsible for polar contacts at the binding interface were highlighted by black circles. For the H 2 R, the BW numbers are used for TM helices and ICL1/2. For ICL3, original index numbers are given. For the mG proteins, the original index numbers are given. The interaction contacts were calculated using the representative low-energy conformation of the systems. Van der Waals, hydrogen bond, and salt bridge interactions are colored in gray, blue, and red, respectively. The line thickness is proportional to the number of contacts of the residue pairs. Hydrophobic, polar, basic, and acidic residues are colored in yellow, green, light blue, and pink, respectively. Important amino acids responsible for polar contacts at the binding interface in the canonical H4R-mGsi complex were highlighted by black circles. For the H4R, the BW numbers are used for TM helices and ICL1/2. For ICL3, original index numbers are given. For the mini-G proteins, the original index numbers are given.

α5 Hook Orientation within the Binding Interface of H2R and H4R Complexes
To further characterize the mG binding in the H2R and H4R complexes, specific α5 hook positions within the receptor binding cavity were analyzed by the α5-NPxxY distance (cf. Supporting Material Figure S7) and the sideward orientation of the α5 helix (Figure 6). In the literature, it has been described that the penetration depth of G protein α5 helices is characteristic for the G protein families, which is often described by the distance of the α5 to the conserved NPxxY motif (N 7.49 , P 7.50 , Y 7.53 ). Commonly, larger outward movements of TM6 in Gs-and Gq-coupled GPCRs move the α5 helix further away from TM7 and thus NPxxY [81]. Accordingly, the mGsi protein was located ~1-2 Å closer to the H2R and H4R-NPxxY motifs than mGs and mGsq (cf. Supporting Material, Figure S7).
More decisively, we detected differences in the sideward orientation of the α5 helices in the H2R and H4R systems. In the H2R systems, the last five residues of the α5 helices (hook) of mGs, mGsi, and mGsq were in equal position, which we identified as a Gs-like position considering the available GPCR-Gs complexes as the reference (cf. Supporting Material, Figure S8A,D). Notably, only the α5 hook of mGsi adopted a Gi-like orientation in the H4R systems (cf. Supporting Material, Figure S8B,E,F). In all systems, we identified a single low-energy well from the free energy profiles calculated using the D 3.32 -histamine The interaction contacts were calculated using the representative low-energy conformation of the systems. Van der Waals, hydrogen bond, and salt bridge interactions are colored in gray, blue, and red, respectively. The line thickness is proportional to the number of contacts of the residue pairs. Hydrophobic, polar, basic, and acidic residues are colored in yellow, green, light blue, and pink, respectively. Important amino acids responsible for polar contacts at the binding interface in the canonical H 4 R-mGsi complex were highlighted by black circles. For the H 4 R, the BW numbers are used for TM helices and ICL1/2. For ICL3, original index numbers are given. For the mini-G proteins, the original index numbers are given.
Unlike in the H 2 R complexes, the α5 hook formed few contacts. Notably, it lacked strong interactions with TM6 and H8. In the H 4 R-mGsq complex, we detected a comparable contact network as in the H 2 R-mGsq complex with the main contacts from ICL2, TM5, TM6, and H8 ( Figure 5C). In this system, the conserved mG residue D213 mGsq formed a salt bridge with R 5.68 of the H 4 R. The α5 hook was further closely bound to the H 4 R via the two salt bridges E225 mGsq -R 34.54 (ICL2) and E225 mGsq -R 8.49 (H8) and by means of a hydrogen bond between V229 mGsq and R 6.32 . By contrast, the extremely high number of (hydrophobic) contacts in the H 4 R-mGs complex was particularly striking ( Figure 5A). In addition to the few specific interactions of the α5 hook (salt bridge of E227 mGs with R 6.32 (TM6) and K 8.48 (H8)), these occurred mainly with TM3, ICL2, TM5, and ICL3. Although the H 4 R comprised a larger ICL3 than the H 2 R, the ICL3 was little involved in the protein interface in the canonical system, H 4 R-mGsi. Moreover, after visual analysis of the trajectories, we assumed that mGs stuck to this region due to the large hydrophobic contacts. In combination with the live cell experiments, we therefore suggest an unfavorable interaction for the H 4 R with mGs that would not occur naturally.

α5 Hook Orientation within the Binding Interface of H 2 R and H 4 R Complexes
To further characterize the mG binding in the H 2 R and H 4 R complexes, specific α5 hook positions within the receptor binding cavity were analyzed by the α5-NPxxY distance (cf. Supporting Material Figure S7) and the sideward orientation of the α5 helix ( Figure 6). In the literature, it has been described that the penetration depth of G protein α5 helices is characteristic for the G protein families, which is often described by the distance of the α5 to the conserved NPxxY motif (N 7.49 , P 7.50 , Y 7.53 ). Commonly, larger outward movements of TM6 in Gs-and Gq-coupled GPCRs move the α5 helix further away from TM7 and thus NPxxY [81]. Accordingly, the mGsi protein was located~1-2 Å closer to the H 2 R and H 4 R-NPxxY motifs than mGs and mGsq (cf. Supporting Material, Figure S7). distance and α5-TM2 distance as reaction coordinates ( Figure 6). The α5 helix in all H2R systems as well as inH4R-mGs and H4R-mGsq systems was oriented toward TM6, giving α5-TM2 distances in the range of ~14 Å . In the H4R-mGsi complex, the global minimum was located at an essentially lower α5-TM2 distance of ~8 Å . Thus, the simulation findings suggested that the ability of the H2R or rather the inability of the H4R to direct the α5 helix to a precise Gs-or Gi-like position contributed to the promiscuous coupling of the H2R to Gs, Gi, and Gq and the selective H4R binding to Gi. The last five residues of α5 helix and residue T/S 2.39 were used to determine the α5-TM2 distance in the H2R and H4R complexes, respectively. T/S 2.39 is shown as red spheres. The H2R is depicted in salmon, the H4R is depicted in purple, mGs is depicted in pink, mGsi in is depicted dark green, and mGsq is depicted in green, respectively. (C) Free energy profiles of GaMD simulations with complexes of either the H2R or the H4R in combination with mGs, mGsi, or mGsq. Distances of D 3.32 (CG atom) and the amino group of histamine (Nα atom) as well as of the α5 helix and TM2 were used as reaction coordinates. The α5-TM2 distances were calculated using the geometric center of Cα, C, and N atoms of the last five residues of the mG α5 helix and T/S 2.39 . For each system, three independent GaMD simulations were used for analysis. (Labels: "B" indicates representative low-energy wells of fully active receptors bound to histamine, "I1" indicates low-energy wells of intermediate receptor conformation bound to histamine. "S1" and "S2" indicate low-energy wells containing conformations with histamine separated from D 3.32 , cf. Figure 1). The last five residues of α5 helix and residue T/S 2.39 were used to determine the α5-TM2 distance in the H 2 R and H 4 R complexes, respectively. T/S 2.39 is shown as red spheres. The H 2 R is depicted in salmon, the H 4 R is depicted in purple, mGs is depicted in pink, mGsi in is depicted dark green, and mGsq is depicted in green, respectively. (C) Free energy profiles of GaMD simulations with complexes of either the H 2 R or the H 4 R in combination with mGs, mGsi, or mGsq. Distances of D 3.32 (CG atom) and the amino group of histamine (Nα atom) as well as of the α5 helix and TM2 were used as reaction coordinates. The α5-TM2 distances were calculated using the geometric center of Cα, C, and N atoms of the last five residues of the mG α5 helix and T/S 2.39 . For each system, three independent GaMD simulations were used for analysis. (Labels: "B" indicates representative low-energy wells of fully active receptors bound to histamine, "I1" indicates low-energy wells of intermediate receptor conformation bound to histamine. "S1" and "S2" indicate low-energy wells containing conformations with histamine separated from D 3.32 , cf. Figure 1).
More decisively, we detected differences in the sideward orientation of the α5 helices in the H 2 R and H 4 R systems. In the H 2 R systems, the last five residues of the α5 helices (hook) of mGs, mGsi, and mGsq were in equal position, which we identified as a Gs-like position considering the available GPCR-Gs complexes as the reference (cf. Supporting Material, Figure S8A,D). Notably, only the α5 hook of mGsi adopted a Gi-like orientation in the H 4 R systems (cf. Supporting Material, Figure S8B,E,F). In all systems, we identified a single low-energy well from the free energy profiles calculated using the D 3.32 -histamine distance and α5-TM2 distance as reaction coordinates ( Figure 6). The α5 helix in all H 2 R systems as well as inH 4 R-mGs and H 4 R-mGsq systems was oriented toward TM6, giving α5-TM2 distances in the range of~14 Å. In the H 4 R-mGsi complex, the global minimum was located at an essentially lower α5-TM2 distance of~8 Å. Thus, the simulation findings suggested that the ability of the H 2 R or rather the inability of the H 4 R to direct the α5 helix to a precise Gs-or Gi-like position contributed to the promiscuous coupling of the H 2 R to Gs, Gi, and Gq and the selective H 4 R binding to Gi.

Mini-G Protein Recruitment Assay
The assay protocol has been essentially described before [52,86,87] and was applied with minor modifications as follows. The day prior to the experiment, the cells were detached by trypsin treatment (0.05% trypsin, 0.02% EDTA in PBS), centrifuged at 700× g for 5 min, and subsequently resuspended in Leibovitz's L-15 (L-15) cell culture medium supplemented with 10 mM HEPES (Serva, Heidelberg, Germany). Then, 80 µL of the transfected as well as of the wild-type HEK293T cells (control) were seeded to a white flat-bottom 96-well plate (Brand GmbH + CoKG, Wertheim, Germany) and incubated overnight at 37 • C in a water-saturated atmosphere. On the day of the experiment, the substrate furimazine (Promega, Mannheim, Germany) was 1:100 diluted in L-15 shortly before the experiment, and 10 µL were added to each well. After measuring the baseline luminescence for 15 min at 37 • C using an EnSpire plate reader (Perkin Elmer Inc., Rodgau, Germany), 10 µL of histamine serial dilutions were added to the cells. Finally, luminescent traces were recorded for a further 45 min. Data were analyzed using GraphPad Prism9 software (San Diego, CA, USA). The relative luminescence units (RLU) were corrected for the baseline drift caused by autooxidation of the substrate by dividing all data by the mean luminescence intensity of the HEK293T wild-type control. Thereafter, the baseline luminescence of the respective L-15 control was subtracted. The area under the curve (AUC) of each concentration was normalized to the AUC of 100 µM histamine (100%) obtained in the canonical system (H 2 R-mGs or H 4 R-mGsi, respectively) and L-15 (0%). The logarithmic histamine concentrations were fitted against the normalized AUCs with variable slope (log(c) vs. response-variable slope (four parameters) yielding pEC 50 and E max values.

Preparation of the H 2 R and H 4 R Complexes
To obtain the H 2 R complexes, the H 2 R-Gs structure that has been published by Conrad et al. was used as template comprising H 2 R residues 15-304 [76]. To obtain the mini-G protein structures mGs, mGsi, and mGsq (according to mGs393, mGsi43, and mGsq71, [46]), homology modeling was performed on the Gs structure using Modeller 9.16 [88]. To avoid artificial charges at the termini of the H 2 R, N-terminal acetyl and C-terminal N-methyl capping groups were added to the structure using Sybyl7.3 (Tripos International, St. Louis, MO, USA (2006) Sybyl 7.3). Coordinates of the ligand histamine were extracted from the template structure and added to the H 2 R-mG complexes. The setup of the complexes for subsequent Gaussian accelerated MD simulations was performed as described previously [89]. During the setup, ff99SB force field parameters [90] were assigned to the complexes, missing hydrogen atoms were added, and the H 2 R disulfide bond C91-C174 was created using tleap. The energy minimization and equilibration steps were essentially performed as described before [89] using Amber17 [91] and Gromacs 2016.5 [92]. In order to embed the H 2 R into a membrane, the structure was overlaid with a pre-equilibrated dioleylphosphatidylcholine (DOPC) bilayer (gaff force field, [93]) and solvated with SPC water [94]. Therefore, the pseudo-atom entries in the 3SN6 OPM entry containing the position of extracellular and intracellular membrane layer have been used.
The H 4 R systems were obtained using the homology model of the receptor provided by GPCRdb (www.gpcrdb.org (accessed on 10 August 2021), Version: 3 September 2019, mainly based on M 2 R structure; pdb-id: 4MQT) [77] as template. The utilized model comprised residues 6-380. Of note, the ICL3 was truncated (69 total missing amino acids) and only consisted of the first 8 (-CQSHPGLT-) and last 5 residues (-LHQRE-) instead of the native, 82 amino acid-sized ICL3. Amber coordinates and topology files were generated for the downloaded pdb file using ambpdb, and the disulfide bond C87-C164 was created using tleap. Initially, the structure was solvated in a TIP3 waterbox with Cl − as counter ions and minimized as performed for the H 2 R structures. To set up the H 4 R-mG complexes, a loop refinement of the ICL3 has been performed using the ModLoop web server (www.modbase.compbio.ucsf.edu/modloop, accessed on 10 August 2021 [95,96] to avoid clashes between the receptor and mG protein structures. The refined H 4 R structure and the mG homology models that have been generated in complex with the H 2 R were aligned to the M 2 R structure in complex with Gi (pdb-id.: 6OIK) to extract appropriate coordinates. Thereafter, similar protocols [89] as for the H 2 R-mG complexes were performed to prepare the H 4 R systems (H 4 R-mGs, H 4 R-mGsi and H 4 R-mGsq) bound to the ligand histamine. A computational model of GaMD simulation systems and a sequence alignment of the chimeric mini-G proteins and the "parental" G protein subunit Gαs are provided in the Supporting Material ( Figures S1 and S9, respectively).

Gaussian Accelerated Molecular Dynamics (GaMD) Simulations
In GaMD, a harmonic, Gaussian-distributed boost potential is applied to biomolecules to smooth the potential energy surface and reduce energy barriers [67]. When the system potential V( r ) is lower than a reference energy E, the modified potential V * ( r ) is calculated as: where k is the harmonic force constant. The two adjustable parameters E and k are automatically determined on three enhanced sampling principles as described before [67]. In summary, E needs to be in the range: where V min and V max are the minimum and maximum potential energies of the system. To ensure that Equation (2) is valid, k is defined as k = k 0 ·1/(V max − V min ) and thus 0 < k 0 ≤ 1.
To enable an accurate energetic reweighting using cumulant expansion to the second order, the third standard deviation of ∆V needs to be small enough: where V avg and σ V are the average and standard deviation of the system potential energies and σ ∆V is the standard deviation of ∆V with σ 0 as a user-specified upper limit for accurate reweighting. When E is set to the lower bound E = V max , according to Equation (2), k 0 can be calculated as: Alternatively, when the threshold energy E is set to its upper bound E = V min + 1/k, k 0 is set to: if k 0 is found between 0 and 1. Otherwise, k 0 is calculated using Equation (3).

Energetic Reweighting of GaMD Simulations
To calculate the potential of the mean force (PMF) and energetically reweight GaMD simulations, the probability distribution along a reaction coordinate is written as p * (A). Given the boost potential ∆V(r) of each frame, p * (A) can be reweighted to recover the canonical ensemble distribution p(A), as: where M is the number of bins, β = k B T, and e β∆V(r) j is the ensemble-averaged Boltzmann factor of ∆V(r) for simulation frames found in the jth bin. The ensemble-averaged reweighting factor can be approximated using cumulant expansion: where the first two cumulants are given by: The boost potential obtained from GaMD simulations usually follows near-Gaussian distribution [69,70]. Thus, the cumulant expansion to the second order provides a good approximation for computing the reweighting factor [67,97]. The reweighted free energy F(A) = −k B T ln p(A) is calculated as: where F * (A) = −k B T ln p * (A) is the modified free energy obtained from GaMD simulation and F c is a constant.

Simulation Protocol
GaMD has been implemented in the AMBER software package, so that all molecular dynamics (MD) simulations could be performed using Amber18 [98]. During the simulations, periodic boundary conditions were given. The SHAKE algorithm was applied to remove the bond stretching freedom of all hydrogen-containing bonds. The GaMD simulations were preceded by a short conventional MD (cMD) simulation of 10.4 ns for the statistical collection of boost parameters (V max , V min , V avg , and σ V ), which was followed by 32 ns "dual-boosted" MD simulations using the calculated parameters (2 fs time steps), in which boost parameters were imposed on the total potential energy and the dihedral energy terms. The reference energy was set to the lower bound (E = V max ), and the boost parameters were updated every 400,000 steps (800 ps). The upper limit of the standard deviation of both the boost potential, the total potential energy, and the dihedral energy were defined as σ OP = σ OD = 6.0 kcal/mol. After the preparatory stage, 3 × 1000 ns of GaMD production runs were performed for the H 2 R and H 4 R systems. Table 2 gives an overview of the average boost potentials that have been applied for the different systems and simulation runs.

Structural Analysis
For subsequent analysis, periodic boundaries were removed from the trajectories, and the molecules were centered to the receptors' transmembrane domains using cpptraj of Amber18 software (San Franscicso, CA, USA) [98]. Furthermore, cpptraj was used for the structural analysis of all distances and angles. Contacts between the H 2 R/H 4 R and mini-G proteins were assessed using the nativecontacts command with a specified distance of 5 Å. Binding energies of the ligand histamine were determined using the mm_pbsa.pl script using IGB = 2 and further parameters in default option according to the MM/GBSA method [83][84][85]. Plots were created using GraphPad Prism9 (GraphPad Software, San Diego, CA, USA), Cytoscape (Shannon P., 2003), and Origin2021 (OriginLab Corporation, Northampton, MA, USA), and the structural visualization was performed using pymol (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC, New York, NY, USA).
Important reaction coordinates were identified from the simulation trajectories such that they involved system dynamic regions and could be used to differentiate conformational states of the HR-mG protein complexes. The observed dynamic regions included the agonists, the receptor TM6 helix, and the C terminus of the mG α5 helix. Therefore, the distance between the charged amine group of histamine and the highly conserved D 3.32 and the receptor TM3 and TM6 intracellular ends were selected as reaction coordinates. The distance between the conserved NPxxY motif in the TM7 intracellular end of the receptors and the C terminus of the mG α5 helix (α5 hook) was used to characterize the HR-mG protein interactions. Furthermore, the TM2-α5 distance was calculated to estimate the α5 hook orientation inside the binding cavity. Particularly, distances were calculated between the backbone (Cα, C, and N) atoms using the center of mass of the following residues. For the TM3-TM6 distance, residues R 3.50 and E/A 6.30 , and for the NPxxY-α5 distance, residues N 7.49 , P 7.50 , and Y 7.53 as well as the last five residues of the mG α5 helix were used. To calculate the α5-TM2 distance, the last five residues of the mG α5 helix and T/S 2.39 were used. The time courses of these reaction coordinates obtained from the GaMD simulation were plotted in the Supporting Material ( Figures S4 and S5). Root-mean-square fluctuations (RMSFs) were calculated for the protein residues and histamine, averaged over three independent GaMD simulations, and color coded for the schematic representation of each complex system (cf. Supporting Material, Figure S2). The representative low-energy conformations of the HR-mG protein complexes were used to compute their residue contact network at the binding interface of the proteins using the nativecontacts command of cpptraj with a distance cutoff of 4 Å. For two-dimensional visualization, the software Cytoscape [99] was utilized to plot the residue contact network.
To recover the original free energy or potential of mean force (PMF) profiles of the four HR-G protein complex systems, GaMD simulations were reweighted using the PyReweighting toolkit [97]. PMF profiles were computed using the combined trajectories from all three independent GaMD simulations (3 × 1000 ns length) for each system. A bin size of 1.0 Å was used, and the cutoff was set to 500 frames for 2D PMF calculations. The 2D PMF profiles were obtained for each simulation system regarding D 3.32 -histamine distance in combination with the TM3-TM6 distance, the NPxxY-α5 distance, and the TM2-α5 distance.

Conclusions
In this study, we have investigated the H 2 R and H 4 R in complex with three types of mini-G proteins, mGs, mGsi, and mGsq, respectively. The results obtained in cellular experiments and GaMD simulations were in good agreement providing important insights into the mechanism of engineered G protein coupling to both receptors. In the mini-G protein recruitment assay using HEK293T cells transiently expressing combinations of either the H 2 R or the H 4 R and mGs, mGsi, and mGsq, we observed a promiscuous coupling of the H 2 R to mGs, mGsi, and mGsq upon activation by histamine. Regardless, the H 2 R preferred the mGs binding over mGsi or mGsq binding revealed by pEC 50 and E max values ( Figure 1A,B; Table 1). In contrast, the H 4 R selectively bound to mGsi. In GaMD simulations, a similar preference was observed due to the lower structural flexibility of the H 2 R-mGs and H 4 R-mGsi complexes. In the simulations, a salt bridge between Nα of the endogenous ligand histamine and D 3.32 was present, which is conserved among histamine receptor subtypes [60,78]. Nevertheless, the binding energies of histamine were different in the complexes. The binding energy determined by MM/GBSA was stronger in the H 4 R than in H 2 R complexes, which was in concordance with a reported higher binding affinity of histamine to the H 4 R [100,101]. Another important aspect was that the H 2 R systems were more dynamic than the H 4 R complexes, leading to larger fluctuations in the TM3-TM6 distances in the H 2 R-mGsi and H 2 R-mGsq being considered as common indicators for GPCR activation [79,102]. In agreement with the higher pEC 50 and less E max values, the mGsi and mGsq proteins were less able to stabilize the fully active H 2 R conformation (TM3-TM6 distance of~19-20 Å), and the systems also visited conformations with a smaller TM3-TM6 spacing ( Figure 1C). As observed for many Gi-coupling GPCRs, the TM3-TM6 distance in the fully active H 4 R in the H 4 R-mGsi complex was lower (~18 Å) than in the canonically Gs-coupled H 2 R. Strikingly, the overall TM3-TM6 distances in the H 4 R-mGs complex ("I1" state) were different from the H 4 R-mGsi ("B" state).
The detailed inspection of the protein interfaces revealed similar residue contributions and binding orientations of mGs, mGsi, and mGsq in the H 2 R binding cavity. The α5 helices were in Gs-like orientation toward TM6 that has been stabilized by specific α5 hook polar interactions with TM6 and H8. Contrarily, the contact network in the H 4 R complexes was more versatile. Notably, only the mGsi α5 helix was in Gi-like orientation toward TM2, which was potentially enabled by the specific α5 helix interactions with TM5 and ICL2 and fewer contacts of the α5 hook. In the H 4 R-mGs and -mGsq complexes, mGs and mGsq were in a similar position as in the H 2 R complexes. Moreover, excessively strong hydrophobic mGs interactions with TM3, ICL3, and TM5 led to the assumption of an unfavorable interaction.
In summary, a coherent overall picture of the H 2 R and H 4 R coupling profiles to engineered G proteins has emerged by the combination of biochemical as well as computational experiments supporting the following: The H 2 R can bind promiscuously to Gs, Gi, and Gq, whereas the H 4 R binds selectively to Gi. Our results suggested that particularly the structural flexibility as well as the ability of a receptor to locate the α5 hook in a precise position, e.g., Gs-like or Gi-like position (Figure 6), might contribute to the coupling behavior of GPCRs. However, these results may not be generalized to the entire G protein classes but rather provide an indication of the preferred Gα subunit. Mini-G proteins are chimeric G proteins derived from Gαs, Gαi 1 , and Gαq proteins. Thus, it remains unclear whether or in which way H 2 R and H 4 R would form complexes with Gi family members Gαi 3 and Gαo or to Gα14 and Gα15 of the Gq family having diverse α5 helices within the Gα classes. In particular, future studies are required focusing on the H 4 R interactions with the Gq family subtype Gα15, which has been demonstrated by Inoue et al. [44], to explain Ca 2+ signals that have been obtained earlier using cells co-expressing the H 4 R and Gα15 [30] but were missing in HEK cells recombinantly expressing the H 4 R alone [27]. Furthermore, experiments comprising holistic cellular responses [103] or the investigation of G protein-dependent second messengers along the signal cascade will be advantageous to fully unravel H 2 R and H 4 R signaling.