Steviol rebaudiosides bind to four different sites of the human sweet taste receptor (T1R2/T1R3) complex explaining confusing experiments
In this section, we provide a comprehensive analysis of sweet ligand binding to the T1R2/T1R3 receptors, organized into three main parts. First, we present results from radioligand competitive binding assays, which offer insights into the binding affinities and dissociation constants of Steviol Rebaudiosides. Next, we discuss the findings from the FLOWER experiments, which enabled us to directly observe the binding responses of sweet ligands to stabilized heterodimer receptors, including the impact of G protein coupling. Finally, we demonstrate the outcomes of our silico analyses, which involved predicting ligand binding energies and identifying potential binding sites through detailed docking simulations. By combining these experimental and computational approaches, we provide a comprehensive insight into the binding dynamics and interactions of sweet taste receptors with Steviol Rebaudiosides.
Radioligand binding studies: taste cell membrane assays
Kinetics of [3H]-Rebaudioside B binding in the presence or absence of Ligands
Results of kinetic binding experiments using [3H]-Rebaudioside B as radioligand presumably bound at VFD2 are shown in Fig. S3. Association was followed by dissociation which was initiated by adding excess (1000-fold) of cold Rebaudioside B at 60 min (time 0) after the start of incubation. [3H]-Rebaudioside showed a kon = 91.82 M−1 min−1, koff = 0.5321 min−1 and Kd = 0.005795 M. Addition of both GDP and GTP resulted in inhibition of Rebaudioside B binding. GDP had an IC50 = 2.152 mM (Fig. S4) while GTP had an IC50 = 9.222 mM (Fig. S5).
Competition binding experiments with various ligands with [3H]-Rebaudioside B
RebM showed ambiguous binding inhibition with RebB while RebC showed low competitive inhibition (Table S1). This was not expected since all the rebaudiosides are presumed to bind in the same VFD2 site. This could indicate that the RebC inhibition is generated at a binding site outside of the VFD2 site. The Ace-K and S819 results were expected. Ace-K was inhibited by RebB, presumably at the VFD2 site. S819 did not show strong inhibition with RebB and it has not been reported as binding in VFD2.
The positive inhibition results for NHDC and amiloride indicate that RebB may have the ability to bind in both TMD regions.
Kinetics of [3H]-Lactisole binding in the presence or absence of ligands
Association was followed by dissociation which was initiated by adding excess (1000-fold) of cold lactisole at 60 min (time 0) after the start of incubation. Lactisole was presumably bound on TMD3. [3H]-Lactisole binding studies showed kon = 217.7 M−1 min−1, koff = 0.01456 min−1 and Kd = 0.0000669 M (Fig. S6). No inhibition was seen with GDP (Fig. S7) or GTP (Fig. S8).
Competition binding experiments using [3H]-lactisole as radioligand
Expected results were seen for NHDC (inhibition), Ace-K and S819. NHDC is known to bind TMD3 like lactisole while Ace-K and S819 do not. The positive inhibition seen with RebM and RebC was unexpected. This could indicate that either lactisole binds at other sites, such as the VFD2 or that RebC and RebM also bind at the TMD3 and are inhibited by lactisole. Since the lactisole site is known to be a negative allosteric binding site, the binding of these Rebs at this location could explain the low potency of these sweeteners where they bind at the TMD3 at high concentrations for a negative allosteric effect after the activation of the receptor by orthosteric binding at the VFD2 at lower concentrations. Relative binding energies are needed to determine if this is possible.
Kinetics of [3H]-Perrilartine binding in the presence or absence of non-labeled test ligands
Association was followed by dissociation which was initiated by adding excess of cold Perillartine at 60 min (time 0) after the start of incubation. Perillartine presumably binds on TMD2. [3H]-Perrilartine had a kon = 184288 M-1 min-1, koff = 0.5337 min-1 and a Kd = 0.000002896 M (2.896 µM) (Fig. S9). Both GDP and GTP had an inhibitory effect on the binding of perillaratine (GDP IC50 = 12.65 mM, Fig. S10; GTP IC50 = 13.167 mM, Fig. S11).
Competition binding experiments using [3H]-Perillartine as radioligand
RebM and RebC showed on competitive inhibition with perillartine as expected since there are no reports of these binding in TMD2. However, amiloride and S819 are known to bind at TMD2 but did not show inhibition with perillartine. This was unexpected. Also unexpected were the high levels of inhibition seen with Ace-K and NHDC which are not known to bind at TMD2 and thus should not show such inhibition. The inhibition of NHDC could indicate that either NHDC can bind at TMD2 or that perillartine can bind at the TMD3 site where NHDC also binds.
Kinetics of [14C]-sucrose binding in the presence or absence of ligands
Radio-labeled sucrose binding was performed as mentioned above. This was to observe the target sweetener. Association was followed by dissociation which was initiated by adding excess of cold sucrose at 60 min after the start of incubation. [14C]-sucrose showed kon = 15.06 M−1 min−1, koff = 0.1863 min−1 and Kd = 0.0123 M (Fig. S12). This was expectedly higher than the previously mentioned high intensity sweeteners.
Competition binding experiments using [14C]-sucrose as radioligand against NHDC
As a method check, the binding competition between sucrose, a VFD2 and 3 binder and NHDC, a TMD3 binder, was performed. Inhibition was observed (Ki = 0.01362 M, Fig. S13).
During this study, very high Kd values have been observed that were in the millimolar range. These values were calculated from the kinetic binding experiments from the ratio of koff /kon. One earlier report has investigated the binding of sweet molecules to bovine taste bud cells63. Kd values reported were of the order of millimolar, which corroborates our findings. For instance, the Kd value for sucrose was 1.1 mM and 3.4 mM for glucose. Such high Kd values may drastically affect the stability of radioligand-receptor complex during washings to separate bound from free ligand. While these dissociation constants and those observed during our study are significantly higher than the usually observed, they might represent a system with different physiological requirements. In taste system, loose binding may be of advantage as a tight binding will result in persistent taste sensation due to the stability of taste receptor-ligand complex and cause taste confusions. A weak taste receptor-tastant complex will subsequently allow the processing of new incoming taste information.
It is known that the presence of nucleotides decrease the binding of agonists but not antagonists with receptors64. We observed during the current study that binding of radioligands, i.e., [3H]-Perillartine and [3H]-RebB, decreased with increasing concentrations of GTP and GDP but not for [3H]-Lactisole. This observation is consistent with the known agonist and antagonist activity of these molecules and indicates that the binding of these molecules depends on the availability of receptors that are coupled to a G protein65. Future studies should use labeled [35S]-GTPγS could give a better understanding of the efficacy and effect on the binding of ligands and provide information about the possible allosteric modulation by other ligands65.
FLOWER binding studies
Binding with and without C-20
Upon binding to the T1R2/T1R3 receptor in FLOWER experiment, sweet ligands induce a red shift in the microtoroid’s resonance. This shift arises from the interaction between receptor-ligand binding and the microtoroid’s evanescent electrical field. Figure 2 shows the binding response of Rubu and RebM. The binding response curve consistently exhibits a peak at approximately 150 s after sweet ligand injection, followed by a decrease and stabilization within the subsequent 400 s to a sustained response, representing around two-thirds to zero of the peak response magnitude. With increasing concentration, both the final stabilized binding response and peak value also increase. Previous work on DMR assays for the human sweet taste receptor has revealed that measuring the peak DMR response offers a more robust assay window compared to measuring the final and sustained DMR response42.Hence, we choose the peak response in the FLOWER resonance shift signal as the binding response in our human sweet taste receptor assays.
Binding curves were constructed by plotting the extracted resonance shift peak value as a function of ligand concentration. In some binding response curves, after reaching saturation, the binding response continues to increase linearly with the sweet ligand concentration. This continued increase after the saturation point is considered nonspecific binding occurring at high ligand concentrations. To determine the binding affinity, we employed a one-site specific binding model66 that takes into account the background response and non-specific binding. The model is fit to the experimental data to assess the binding affinity and the Hill slope. The binding constant Kd and Hill slope nH re determined through the curve fitting process according to the following equation:
$$\triangle \lambda =\fracB_\max * [L]^n_HK_d^n_H+[L]^n_H+\rmNS* \left[\rmL\right]+B_background$$
(1)
where Bmax is the maximum specific binding, ∆λ is the resonance shift at response peak, NS is the non-specific binding parameter, [L] represents the concentration of the sweet ligand solution, and Bbackground is the background response. To create a normalized binding curve, we eliminated the background signal and non-specific binding response from the experimental data. Following this, the data points were normalized by dividing each point by the maximum observed resonance shift in the dataset,
$$\triangle \lambda _norm=\frac\triangle \lambda -B_backgroundB_\max -\frac{\rmNS* \left[\rmL\right]}B_\max =\frac[L]^n_HK_d^n_H+[L]^n_H$$
(2)
This normalization process enhances the clarity of the specific binding response curve, facilitating a more precise evaluation of the binding constant Kd and Hill slope nH. The normalized binding curve and experiment data points are shown in Fig. S2. The binding constant Kd is listed in Table 1 and the Hill slope nH is listed in Table 2. The R2 of the binding response curves are shown in Table S6. When the Hill slope, nH is less than 1, suggesting negative cooperativity, the binding site number (N) satisfies the condition \(N\ge \frac1n_H\)67. Table 2 might contain errors in the Hill slope due to the errors inherent in experimental measurements and the curve fitting process. When the Hill slope nH is significantly less than 1, it is considered indicative of multiple binding sites. In Table 2, all four sweet ligands, RebC, Rubu, RebD, and RebM, exhibit instances where the Hill slope is significantly less than 1. Following the experimental principle \(N\ge \frac1n_H\), this suggests that these four sweet ligands possess multiple binding sites on the human T1R2/T1R3 sweet receptor.
Considering the sweetness levels of the ligands in Table 1, RebC is categorized as a low-sweet ligand, while RebM and RebD are categorized as high-sweet ligands, with RebM slightly surpassing RebD in sweetness. Rubu falls in the intermediate sweetness category, positioned between RebD and the low-sweet RebC. In Table 1, across all scenarios involving human sweet T1R2/T1R3 receptors, the binding constant Kd consistently shows an inverse correlation with sweetness, except for RebD. Despite having similar sweetness to RebM, RebD deviates from this pattern, suggesting lower efficiency in transmitting the sweetness biological signal. The stronger binding affinity of RebD results in a lower sweetness signal compared to RebM. In all cases where G protein C20 is introduced to the microtoroid functionalized with sweet T1R2/T1R3 receptors before sweet ligand injection, a significant reduction in the binding constant kd is observed. This implies that the coupling of G protein C20 to the TMD of the T1R2/T1R3 heterodimer enhances the heterodimer sweet receptor’s binding affinity upon its first activation.
The interaction with antibodies elicits distinct effects on the T1R2/T1R3 sweet receptor. Attaching antibodies to the Helix 8 C-terminus of the GPCR places them at a significant distance from the extracellular TMD binding site, indicating minimal impact on sweet ligand binding at the TMD binding site. When the antibody is attached to the TMD Helix 8 site, potentially involved in G protein coupling to TMD, it may impede C20 binding to TMD. In the absence of C20, RebM and RebC exhibit a lower binding constant Kd in the TMD2 open case than the TMD3 open case. According to the Hill slope nH for the multiple binding of negative cooperativity binding in the supplementary information, the enhanced binding affinity at the orthosteric binding site leads to a decrease in the binding constant Kd with an increase in Hill slope nH. The elevated Hill slope in the TMD2 open case suggests increased binding affinity at the orthosteric binding site (VFD2), as indicated by docking simulations for RebM and RebC. The anti-flag antibody coupled to TMD2 may reorient TMD2 to stabilize the VFD2 conformation through CRD2, enhancing VFD2 binding affinity for RebM and RebC. Notably, this antibody’s enhancement effect is relatively weaker compared to that of G protein C20 coupling to TMD. However, for Rubu, whose orthosteric binding site is VDF3 as indicated by docking simulations, the binding constant Kd remains largely unchanged between the TMD2 open case and TMD3 open case in the absence of C20. The TMD3 open case’s Hill slope even shows a slight decrease, suggesting that the anti-rho antibody coupled to TMD3 may not enhance the VFD3 binding affinity for Rubu. It appears that the anti-rho antibody coupled to TMD3 does not induce reorientation of TMD3 or stabilize the VFD3 conformation but don’t enhance the VFD3 binding affinity for Rubu.
In the presence of G protein C20, RebM and RebC exhibit a lower binding constant Kd transitioning from the TMD3 open case to the TMD2 open case, especially for the high-sweet RebM. Simultaneously, the Hill slope of the TMD2 open case is higher than the TMD3 open case, indicating enhanced binding affinity at the orthosteric binding site (VFD2). For RebM and RebC in the TMD3 open case, the presence of C20 results in a lower binding constant Kd and Hill slope nH compared to the absence of C20. As per the Hill slope nH for the multiple binding case for the negative cooperativity binding in the supplementary information, the orthostatic binding site’s binding affinity enhancement, causes the binding constant Kd to decrease with a decrease in Hill slope nH. Therefore, G protein coupling to TMD3 enhances the orthostatic binding site of RebM and RebC, which could be VFD3, TMD2, or TMD3. For Rubu, whose orthosteric binding site is VFD3, in the TMD3 open case, the binding constant Kd decreases with C20 coupling to TMD3. Simultaneously, the presence of G protein C20 increases the Hill slope nH close to 1, signifying enhancement at the orthosteric binding site VFD3. The experimental results suggest that G protein might bind to both TMD3 and TMD2 intracellular regions and enhance the binding of the corresponding binding sites VFD2 and VFD3 separately for the Steviol Rebaudiosides.
In silico binding energies
Docking results of various Rebs at multiple binding sites, VFD2, VFD3, TMD2, and TMD3
All docking results of various steviol glycosides with or without C20 at VFD2, VFD3, TMD2, and TMD3 are summarized in Table 3. Based on UCavE, most of steviol glycosides without C20 prefers to bind at the orthosteric binding site, cVFD2, except RebB and Rubu.
Compared to the orthosteric binding site (UCav E: –64.66 kcal/mol), Rubu which is the smallest steviol glycoside shows more stable interaction at VFD3 (UCav E: -82.41 kcal/mol), where natural sugars, sucrose or fructose, can also bind.
RebB is the only ligand with a charged group among the steviol glycosides, which was used as a radiolabeled ligand for this binding study. Because of the presence of the charged group, RebB showed a different binding preference compared to other steviol glycosides. It has much favorable interactions at TMDs (UCav E: –98.53 kcal/mol for TMD3, -96.11 kcal/mol for TMD2) rather than VFD2 (UCav E: –72.84 kcal/mol). As shown in Fig. 3, the carboxylate group at the R1 site of RebB can form a salt bridge with R725 (5.37) at TMD2 and with the protonated H734 (5.44) at TMD3. RebB also has several H-bonds at the backbone of L719 (EC2) /S726 (EC2) and the side chains of R790 (7.34) at TMD3. The best binding site for RebB is the allosteric binding site, TMD3. Thus, this preferred binding of RebB at TMD3 and not VFD2 results in ambiguous data for RebM at VFD2 from the radio-labeled ligand binding study in Table 3.
When C20 is present at TMD3, all four cases of RebM, RebD, RebC, and Rubu prefer to bind to the orthosteric binding site, VFD2. However, when C20 is present at TMD2, RebM and RebD with high sweetness do not affect the binding preference, while Rubu and RebC with low sweetness differs in binding preference (Table 3). The order of UCav binding energy differs in the low sweet cases. Rubu first binds at oVFD3, while RebC first binds at TMD3.
Docking result of various ligands
T1R2/T1R3 TMD3 binders
Cyclamate allosteric agonist, Neohesperidin dihydrochalcone (NHDC) artificial sweetener, and lactisole negative allosteric modulator are well-known TMD3 binders to the T1R2/T1R3 heterodimer. When we used lactosole as a radiolabeled ligand, the VFD2 binder Ace-K (artificial sweetener) and the TMD2 binder S819 (positive allosteric modulator) do not compete with lactisole (Table 4). However, TMD3 binder NHDC shows the highest inhibition with Ki of 333.7 µM among the 6 ligands (RebM, RebC, Ace-K, S819, NHDC, and amiloride). Our docking study also found that NHDC has the lowest UCav binding E (the second lowest BE), consistent with the highest inhibition (Ki) among six ligands (Table 4). From the experimental observation, the NHDC binding site at the human sweet taste receptor overlaps with those for the sweetener cyclamate and the sweet taste inhibitor lactisole68. The observed binding site of cyclamate and/or lactisole are Q636 (3.28), Q637 (3.29), S640 (3.32), H641 (3.33), H721 (EC2), R723 (5.36), S729 (5.42), F730 (5.43), A733 (5.46), F778 (6.51), V779 (6.52), L782 (6.55), R790 (7.34), and L798 (7.36) of hT1R3. Additional important amino acids for NHDC binding are Y699 (4.60), W775 (6.48), and C801 (7.39) from point-mutation experiments. From our docking study, NHDC forms multiple H-bonds at the backbone of F624 (2.60) and V720 (EC2) and with the side chains of Q636 (3.28), Q637 (3.29), S640 (3.32), C722 (EC2), S726 (EC2), R790 (7.34), and Q794 (7.38), as shown in Fig. 4. Steviol glycosides, RebC and RebM, also show mM level binding affinities, 6.98 and 11.75, respectively. RebC has better scoring energies than RebM does, which also agrees with experimental observations. Ligands that bind to other sites, Ace-K, S-819, and amiloride, reveal higher binding energy at TMD3, which also have fewer H-bonds. As shown in Fig. 4, S819 has H-bonds at H641 (3.33) and N737 (5.47). Amiloride forms H-bonds at S640 (3.32), C722 (EC2), R723 (EC2), R790 (7.34), and Q794 (7.38). Ace-K has H-bonds at Q637 (3.29) and R790 (7.34).
T1R2/T1R3 VFD2 binders
The artificial sweeteners saccharin and aspartame as well as the natural sweeteners sugar and stevioside can bind to the VFD2 of th eT1R2/T1R3 heterodimer. From docking studies, we found RebM, which has high sweetness, shows the lowest binding energy in all scoring energies (UCav, BE, SolvE), although RebM gave an ambiguous result in experiment because of its low solubility (Table 5). The second lowest energy is from RebC, which has lower sweetness than does RebM. Ace-K has the highest Kd value (bad binder to VFD2), resulting in the highest binding energy. However, the best binder to VFD2 was amiloride (TMD2 binder), although the binding energy shows it to be unfavorable. In addition, several ligands known to bind TMD2 or TMD3 were found at a detectable binding affinity. The third lowest one by UCav and BE is the TMD3 binder NHDC, which displays a slightly lower binding affinity. One possibility of these mixed data is the multiple binding sites of RebB used for the radiolabeling. Based on the experiment (Table 4), since RebC can bind to TMD3, RebB without sugars at the R1 position which is smaller than RebC is also predicted to bind to TMD3. As shown in Fig. 5, S819 and Ace-K have two hydrogen bonds at D142 and R383. Amiloride forms multiple hydrogen bonds at Y103, N143, S165, D278, and R383. NHDC has hydrogen bonds at N44, N143, D213, and R383. RebM made hydrogen bonds at S40, D142, D278, D307, R339, and R383. Our docking pose of RebB at the binding site of TMD3 shows the possibility of a salt-bridge at the protonated histidine, H734 (5.44), as shown in Fig. 3 (right). Based on our docking study, RebB can also bind to TMD2. Supporting this, RebB at the binding site of TMD2 formed a salt-bridge at R725 (5.37) as well as H-bonds at N731 (5. 43) with the terminal carboxylate (Fig. 3, left). So, the high binding affinity data of amiloride might arise from TMD2 binding. The other possibility derives from the multiple binding sites of the tested ligands. For example, amiloride (TMD2 binder) could bind to multiple binding sites because of its small size and its ability to form multiple hydrogen bonds. Thus, this discrepancy arises from mixed data of multiple binding sites of tested ligands as well as the radio labeled ligand.
T1R2/T1R3 TMD2 binders
Allosteric agonist perillatine, antagonist amiloride, and positive allosteric modulator S819 are reported as TMD2 binders of the T1R2/T1R3 heterodimer. From experiments for testing the binding affinity at TMD2 (Table 6) using perillartine for radiolabeling, there is much ambiguous data. Even TMD2 binders such as S819 and amiloride result in ambiguities. Unexpectedly, Ace-K (VFD2 binder) and NHDC (TMD3 binder) reveal a detectable level of binding affinity.
From docking, we have good agreement between theory and experiment for three cases (S819, NHDC, amiloride). NHDC has a better binding energy than does S-819 or amiloride. However, Ace-K is less favorable than NHDC. Since perillartine (agonist) has a structure very similar to amiloride (antagonist), we suspect this mixed data arises from multiple binding sites of the radiolabeled and the test ligand. As shown in Fig. 6, NHDC has hydrogen bonds at R3.28, S5.51, R5.37, R7.34, and D7.38. RebM made hydrogen bonds at R3.28, K4.53, D5.47, S5.51, and N7.45. S819 has a hydrogen bond at D7.38. Amiloride forms multiple hydrogen bonds at N5.43, T5.44, S6.48 and N7.45. Ace-K has a salt bridge at R5.37 with one hydrogen bond at N5.43.
Docking result of radio-labeled ligands, Perillatine and Lactisole
From the docking study, perillatine, an agonist of the human sweet taste receptor, interacts with N799 (7.37) at TMD2 of T1R2/T1R3 heterodimer with UCav E of -25.92 kcal/mol and BE of -35.28 kcal/mol, as shown in Fig. 7 (Left). Lactisole as a negative allosteric modulator can bind to TMD3 of T1R2/T1R3 heterodimer. From our docking study, lactisole, an inhibitor of the human sweet taste receptor, interacts with S640 (3.32), the protonated H641 (3.33), H734 (5. 44) and Q794 (7.38) at TMD3 of T1R2/T1R3 heterodimer as shown in Fig. 7 (Right). This docking result of lactisole agrees with the inhibitory activity of (±)-lactisole, which was not determined by H641A and Q794N mutation at T1R3 TMD30.
In silico binding site discovery
We discovered nine sites using the SiteMap method. The sweet heterodimer appears to have many potential binding sites that make interpretation of the experimental binding data difficult. Two sites discovered correspond to the canonical VFD2 binding site (in silico site 5, Fig. S14) and VFD3 binding site (in silico site 1, Figure S14).
Two additional sites were found near the interaction of the VFD. Site 4 was near the top intersection of the VFD. Site 6 was near the bottom intersection of the VFD. The VFD2 and VFD3 upper distances, near site 4 have been correlated with binding and the active state of the receptor of some ligands69. Binding in this site could potentially influence binding and activation. Binding in site 6 could influence the transmission of allosteric movements between the heterodimer section.
Site 9 was found at the intersection of the VFD2 and CRD2. Binding at site 9 could influence the canonical binding site of VFD2.
Two sites (8 and 10) were more associated with TMD, 10 was located at the intersection of TMD2 and CRD2. This was near the sweet protein binding site. Presumably binding at this site could activate the receptor. Finally, site 8 is located on the intracellular side of TMD3. Binding at this site could potentially impact the G protein if it is bound on TMD3.
The δG of binding from MM/GBSA (Table S4) are shown below for high sweet (RebB) medium sweet (Rubu) and low sweet (RebC) ligands. Site 5 in Fig. S14 was selected as the canonical VFD2 binding site and MM/GBSA values are shown as the delta from those of site 5 to enable a comparison. Sites 2, 6 and 10 show better relative binding than site 5 (VFD2) from RebB. RebC showed the same but with the addition of site 8. Rubu showed less favorable relative binding at all sites probably due to the smaller size and less sugar attachments.
The SiteMap analysis indicates the potential for several non-canonical binding sites that could influence the interpretation of experimental binding results.
AlphaFold structure comparison
We compared our MD structures of RebM bound TAS1R2/1R3 with C20 at T1R2TMD or T1R3TMD with AlphaFold 3 structures, generated using an AI model developed by Google DeepMind and Isomorphic Labs available at Three FASTA sequences were derived from our MD structures, while AlphaFold 3 predicted a total of five structures. As depicted in Figure S16, all five AlphaFold structures exhibited significant similarity. For the comparison of our MD structure of RebM-bound TAS1R2/1R3 with C20 at T1R3TMD, the RMSD values from PyMOL were 14.05, 13.82, 13.82, 13.97, and 14.51 Å for models 0–4, respectively. Similarly, for our MD structure of RebM-bound TAS1R2/1R3 with C20 at T1R2TMD, the RMSD values were 13.29, 12.48, 12.85, 13.08, and 13.47 Å for models 0–4, respectively. Both structures exhibited identical TMD interfaces such as TM6-6, which is known TMD interface in Class C GPCRs upon activation.
We observed a discrepancy in the disulfide bonding patterns within the VFD region (Table S5). The AlphaFold model indicates inter-disulfide bridges between C359 (1R2) and C129 (1R3), whereas our predicted model shows an intra-disulfide bridge between C359 and C363 in 1R2VFD. For 1R3VFD, the AlphaFold model depicts a disulfide bridge between C370 and C373, while our model predicts two disulfide bridges: one between C351 and C370 and another between C373 and C375. This disparity arises from the use of different templates for VFD model generation. Our VFD model was based on the structure of the rat metabotropic glutamate receptor 1 (mGluR1) (PDB ID: 1ewt). Subsequently, the crystal structure of the medaka fish taste receptor T1R2/1R3 complex with L-glutamine (PDB ID: 5x2m) was reported, showing an inter-disulfide bridge between VFDs.
However. the putative C20 binding site in the AlphaFold model was located in EC2 or CRD2, rather than between TM3 and TM6, which is recognized as the G protein coupling site. The top AlphaFold model (model 0) revealed a salt bridge between K345 (K-11) and D688 (EC2 of 1R2), as illustrated in Figure S17. Conversely, in our MD structure of RebM-bound TAS1R2/1R3 with C20 at T1R2TMD, two salt bridges were observed among R664 (IC2), E346 (E-9), and D341 (D-15), whereas in the structure with C20 at T1R3TMD, one salt bridge was noted between R760 (6.33) and the C-terminal end of F354 (F-1), as shown in Figure S17. Thus, AlphaFold 3 was not able to predict the C20 binding site within the known cytoplasmic end of TM3-6 of GPCRs.
link