Molecular mechanism of the sweet taste enhancers

Results and Discussion

Functional Mapping of Sweet Enhancer Interaction Sites.

To understand the mechanism of these sweet taste enhancers, it is important to determine where these molecules bind to the receptor. To identify the functional domain required for interaction with enhancer molecules, we used sweet–umami (10) and human–rat chimeric T1R receptors (12) to map the binding site. The sweet taste receptor and the umami taste receptor share a common subunit, T1R3. Because the sweet taste enhancers have no effect on the umami taste receptor, they most likely bind to the T1R2 subunit, either at the N-terminal extracellular domain (NED; composed of the VFT domain and a small Cys-rich region) or the transmembrane domain (TMD).

We swapped the NED of the sweet and umami receptors to generate sweet–umami chimeras. T1R2-1 consists of the T1R2 NED and T1R1 TMD, and T1R1-2 consists of the T1R1 NED and T1R2 TMD. Each hybrid receptor was coexpressed with T1R3 and assayed for enhancement activity of the sweet enhancers (Fig. 1A ). SE-2 was used to study the mechanism of sucralose enhancement, and SE-3 was used for sucrose. To elevate the hybrid receptor activity to measurable levels, we coexpressed human T1R1-2 with rat T1R3 instead of human T1R3. T1R2-1/T1R3 responded to sucralose and sucrose, and the activities could be enhanced by SE-2 and SE-3, respectively. In contrast, T1R1-2/T1R3 responded to glutamate and the activity was not enhanced by either SE-2 or SE-3. These observations indicate that the NED of T1R2 is critical for interaction with SE-2 and SE-3, whereas the TMD of T1R2 is not required.

SE-2 and SE-3 are selective for human T1R2/T1R3 and have no effect on the rat receptor. This species selectivity can be exploited to map the interaction sites using human–rat chimeric receptors (Fig. 1B ). Replacing the human T1R2 NED with its rat counterpart gave rise to a chimeric sweet taste receptor that was not responsive to SE-2 or SE-3. Conversely, replacing the rat T1R2 NED with the human counterpart made the receptor responsive to the enhancers. The results obtained using human-rat chimeric receptors confirmed the role of T1R2 NED in interaction with the SE-2 and SE-3.

Mutagenesis Studies.

To further define the binding region for sweet taste enhancers, we carried out mutagenesis studies on the T1R2 VFT domain. Twenty-three residues within the T1R2 VFT domain were selected for mutagenesis based on sequence alignment with mGluRs and a molecular model of the T1R2 VFT domain (Molecular Modeling). Among the 23 mutants, 7 were found to have significantly reduced responses to sucralose or sucrose, whereas the enhancement activities of SE-2 and SE-3 were essentially not affected (Table 1 and Figs. S1 and S2). Y103A is a representative mutant of this class (Fig. 2A ). When coexpressed with hT1R3, Y103A did not respond to sucralose up to 10 mM, >100 times higher than the EC₅₀ for the wild-type receptor. Similarly, the mutant receptor did not respond to sucrose at 100 mM (Fig. 2B ). SWT819 (12), a sweet receptor agonist that interacts with the T1R2 TM domain, was used as a control. Y103A did not affect SWT819 activity (Fig. 2E ), indicating that the loss of sucralose and sucrose responses was not due to altered protein expression, misfolding, or poor membrane targeting. The enhancement activities of SE-2 and SE-3 were intact. The response to sucralose could be rescued by SE-2 (Fig. 2A ) and that of sucrose by SE-3 (Fig. 2B ).

Mutations of three residues affected the enhancement activities of SE-2 and SE-3. L279A is a typical example of this class of mutants. The dose–response curve of this mutant receptor to sucralose was identical in the presence or absence of 25 μM SE-2, which was sufficient to shift the EC₅₀ of the wild-type receptor by 6-fold (Fig. 2C ). Similarly, the response of L279A to sucrose was no longer potentiated by SE-3 (Fig. 2D ). It should be noted that L279A also had a partial effect on the sucralose and sucrose activities (Molecular Modeling), shifting the dose–response curves to the right by about 5-fold, whereas the response to SWT819 was not changed (Fig. 2E ).

Molecular Modeling.

Homology models of the T1R2 VFT domain (Fig. 3 and Figs. S3 and S4) in open and closed forms were built using crystal structures of mGluR1, mGluR3, and mGluR7 (5, 6, 13) from the Protein Data Bank. Despite low overall sequence identity (∼30%) of T1R2 to mGluR1, residues close to the hinge whose side chains point inward toward the active site show significant similarity (Fig. S5). The model showed similarities and differences both with mGluRs and with our previous model of T1R1.

The upper lobe of T1R2 has a large cavity that can explain the existence of a variety of large sweeteners that interact with the T1R2 VFT domain (Fig. 3A ). In our model, sucrose and sucralose molecules interact with the backbone nitrogens of I167 and S144 and to the hydroxyl of S144 (Fig. 3B and Figs. S3 and S4). These residues are close to the hinge region of the VFT domain. All hydroxyl groups of sucrose and sucralose made hydrogen bonds with adjacent hydrophilic residues such as D142 and E302, whereas the chlorines of sucralose found hydrophobic contacts with residues Y103 and P277. Similar active-site agonist binding residues were described in a previous article on T1R2 modeling with aspartame (14). The positioning of sucrose and sucralose follows the evolutionarily conserved mechanism of agonist–hinge interaction found in other VFT domains and is supported by site-directed mutagenesis. In particular, mutations of D142 and E302 to alanine abolished the activities of sucrose and sucralose. Mutation of S144 or I167 had no obvious effect, probably because of the interaction being mediated largely by the backbone nitrogens. The diminished responses of the seven mutants (S40, Y103, D142, D278, E302, P277, and R383) to sucralose and sucrose could all be rescued by SE-2 and SE-3, respectively, suggesting that these residues are not making direct contact with the enhancers and that the overall binding mode of the agonists in the presence of enhancers is not altered because of these mutations.

In our model, the newly discovered sweet taste enhancers bind adjacent to their agonists, with van der Walls and hydrogen bonding interactions possible between enhancers and sweeteners (Fig. 3 and Figs. S3 and S4). In the closed conformation with sucrose and SE-3, four residues (K65, L279, D278, and D307) surround and interact with the enhancer. Sucrose and SE-3 form direct contact through a hydrogen bond. K65 and D278 form direct electrostatic interactions. The ring nitrogen of bound SE-3 closest to K65 may be unprotonated (with a measured solution pK_a of 5.98), enhancing the compound's interaction with K65. The binding mode of sucralose/SE-2 follows a similar pattern (Fig. S4). Sucralose and SE-2 are in direct contact, and both form extensive hydrogen bonding patterns to residues in the active site. In our model, the two critical chlorines for SE-2 enhancement activity, 4-Cl and 6′-Cl, are not in direct contact with SE-2, but engage in hydrophobic interactions that may help orient the two rings of sucralose to more effectively interact with SE-2.

We introduced the term “pincer residues” to describe amino acids near the lips of the lobes involved in lobe-to-lobe interactions or lobe–enhancer interactions to help stabilize the closed conformation of the VFT domain. The extensive nature of interlobe connections in mGluRs was also noted previously (15). Direct interlobe interactions found in the closed conformation of mGluRs manifest as both electrostatic and hydrophobic interactions in the model of T1R2. Based on our model, the potential electrostatic pincer residues include R383 and K65 of the upper lobe and D278 and D307 of the lower lobe. The potential hydrophobic pincer residues include A43, V64, I67, Y103, and K65 of the upper lobe, and P277, L279, and V309 of the lower lobe. K65 appears to be involved in both electrostatic and hydrophobic interactions. In our model, these residues located on the opposite lobes do not contact each other in the open conformation but are positioned to interact upon lobe closure (Fig. 4B ). It is worth noting that the hydrophobic pincer interaction was absent in our umami receptor model. The role of K65 and D278 as electrostatic pincer residues was supported by the activities of the D278K/K65D double mutant (see details in the next section). Even though these residues are not in direct contact with sucrose or sucralose in our model, mutations of some of these residues (K65, Y103, L279, D307, and R383) still resulted in a diminished response to the sweeteners (Figs. S1 and S2). This is probably due to the critical role of these residues in stabilizing the closed conformation.

We hypothesize that the activation of the T1R2 VFT domain follows a two-step mechanism as proposed in the case of mGluRs: attachment of ligands to the upper lobe (docking) followed by lobe closure (locking) (16 –18). Based on the dynamics of the closure of the iGluR2 VFT domain, the two steps proceed on two different time scales—i.e., approximately microseconds for docking and approximately milliseconds for locking. Consequently, in this dynamic model the opening between the lobes is exposed for a long (approximately milliseconds) time frame, making cooperative attachment of both sweeteners and modulators to the upper lobe possible. The difference in the timing of docking and locking reflects the existence of an entropic penalty that must be compensated by enthalpic free energy contributions during lobe closure. The free energy entropic penalties come from (i) restriction of backbone motions of the hinge through interaction with the sweeteners, (ii) restriction of side chain motions of contact residues to sweeteners and enhancers, and (iii) restriction of relative lobe movement due to sweetener and enhancer binding.

Many large sweeteners, with much higher affinity than sucrose for the sweet taste receptor, interact with the T1R2 VFT domain (unpublished data). A main conjecture of our model is that the newly discovered sweet taste enhancers share the same binding pocket as those of large sweeteners, such as stevioside, superaspartame, and SC-45647.

We used stevioside as an example to illustrate the point (Fig. 3C ). In our model, the sucrose–enhancer complexes spatially overlap with stevioside and fit in the same upper lobe cavities. The single glucose moiety of stevioside is positioned to occupy the same space as sucrose/sucralose, whereas the steviol backbone overlaps with the sweet enhancer molecules. The double glucose moiety extends out of the VFT domain. The combined effect of sucrose/sucralose and enhancers results in a binding mode that mimics the binding mode of stevioside. The model of stevioside binding is supported by the mutagenesis data. The critical residues in Table 1 can be divided into two groups, with one group located near the hinge region adjacent to sucrose/sucralose and the other located close to the enhancers. The majority of the residues from both groups are required for stevioside activity (Table 1). One notable exception is the L279A mutation that does not affect stevioside. This can be explained by the size of stevioside, which already makes extensive hydrophobic interactions within the active site. A small reduction in hydrophobic interactions will not noticeably affect the binding free energy.

Electrostatic Interaction Between K65 and D278.

D278 is one of the pincer residues required to stabilize the closed conformation of the T1R2 VFT domain. D278A had no detectable response to sucralose or sucrose, but the responses could be rescued by low concentrations of SE-2 and SE-3, respectively (Figs. S1 and S2). Surprisingly, SE-2 and SE-3 elicited straight agonist responses on D278A at higher concentrations in the absence of sucralose or sucrose (Fig. 4A ). No response was observed on the wild-type receptor at the same concentration of SE-2 and SE-3. We speculate that the D278A mutant may provide additional hydrophobic contact between the enhancers and lower lobe, enabling the enhancers to activate the receptor without the sweeteners.

The model suggests that D278 stabilizes the closed conformation of the T1R2 VFT domain through interaction with K65 on the opposite lobe. To test this hypothesis, we generated mutants that reverse the charges of D278 and K65, individually and in combination (K65D, D278K, and K65D/D278K double mutant; Fig. 4B ). Reversing the charge on either K65 or D278 individually abolished the response of the receptor to sweeteners including sucrose, D-tryptophan, and even cyclamate. However, the responses to those sweeteners were rescued by the double mutant, indicating that the opposite charges of the pincer residues on the two lobes are important for the receptor activity. This observation argues that the electrostatic interactions between K65 and D278 are critical for the sweet taste receptor activities, providing strong support for our hypothesis that these two residues form a salt bridge upon the closure of the T1R2 VFT domain to stabilize the active conformation. It is somewhat surprising to us that even cyclamate required the opposite charges of the opposite pincer residues, because this is a sweetener that interacts with the TM domain of T1R3. It is likely that activation of the sweet taste receptor by cyclamate still requires conformational changes in the T1R2 VFT domain. In contrast, SWT819 probably activates the receptor through a different mechanism, as the compound can activate all three charge reversing mutants.

In summary, our data generated using chimeric T1R receptors indicate that the two sweet taste enhancers interact with the VFT domain of T1R2. Results from molecular modeling and mutagenesis studies strongly argue that the sweet taste enhancers follow a similar mechanism to that of the umami taste enhancers IMP and GMP. Sweeteners bind near the hinge region and induce initial closure of the VFT domain, whereas the enhancer molecules bind near the opening of the pocket and further stabilize the closed conformation by strengthening the hydrophobic interactions between the two lobes and lowering the entropic penalties of lobe closure. Through cooperative binding with agonists in the VFT domains, this group of positive allosteric modulators for T1R taste receptors represents a unique mechanism of class C GPCR modulation.