Lactisole Interacts with the Transmembrane Domains of Human T1R3 to Inhibit Sweet Taste*

The detection of sweet-tasting compounds is mediated in large part by a heterodimeric receptor comprised of T1R2+T1R3. Lactisole, a broad-acting sweet antagonist, suppresses the sweet taste of sugars, protein sweeteners, and artificial sweeteners. Lactisole's inhibitory effect is specific to humans and other primates; lactisole does not affect responses to sweet compounds in rodents. By heterologously expressing interspecies combinations of T1R2+T1R3, we have determined that the target for lactisole's action is human T1R3. From studies with mouse/human chimeras of T1R3, we determined that the molecular basis for sensitivity to lactisole depends on only a few residues within the transmembrane region of human T1R3. Alanine substitution of residues in the transmembrane region of human T1R3 revealed 4 key residues required for sensitivity to lactisole. In our model of T1R3's seven transmembrane helices, lactisole is predicted to dock to a binding pocket within the transmembrane region that includes these 4 key residues.

The detection of sweet-tasting compounds is mediated in large part by a heterodimeric receptor comprised of T1R2؉T1R3. Lactisole, a broad-acting sweet antagonist, suppresses the sweet taste of sugars, protein sweeteners, and artificial sweeteners. Lactisole's inhibitory effect is specific to humans and other primates; lactisole does not affect responses to sweet compounds in rodents. By heterologously expressing interspecies combinations of T1R2؉T1R3, we have determined that the target for lactisole's action is human T1R3. From studies with mouse/ human chimeras of T1R3, we determined that the molecular basis for sensitivity to lactisole depends on only a few residues within the transmembrane region of human T1R3. Alanine substitution of residues in the transmembrane region of human T1R3 revealed 4 key residues required for sensitivity to lactisole. In our model of T1R3's seven transmembrane helices, lactisole is predicted to dock to a binding pocket within the transmembrane region that includes these 4 key residues.
Taste is a primal sense that enables diverse organisms to identify and ingest sweet-tasting nutritious foods and to reject bitter-tasting environmental poisons (1). Taste perception can be categorized into five distinct qualities: salty, sour, bitter, umami (amino acid taste), and sweet (1). Salty and sour depend on the actions of ion channels. Bitter, umami, and sweet depend on G-protein-coupled receptors (GPCRs) 1 and coupled signaling pathways. Sweet taste in large part depends on a heterodimeric receptor comprised of T1R2ϩT1R3 (2)(3)(4)(5).
The T1R taste receptors (T1R1, T1R2, and T1R3) are most closely related to metabotropic glutamate receptors (mGluRs), Ca 2ϩ -sensing receptors (CaSRs), and some pheromone receptors (6 -10). All of these receptors are class-C GPCRs, with the large clam shell-shaped extracellular amino-terminal domain (ATD) characteristic of this family. Following the ATD is a cysteine-rich region that connects the ATD to the heptahelical transmembrane domain (TMD); following the TMD is a short intracellular carboxyl-terminal tail. The solved crystal structure of the ATD of mGluR1 identifies a "Venus flytrap module" (VFTM) involved in ligand binding (11). The canonical agonist glutamate binds within the VFTM in a cleft formed by the two lobes of this module to stabilize a closed active conformation of the mGluR1 ATD. In contrast, several positive and negative allosteric modulators of class-C GPCRs have been identified and shown to act via binding not within the VFTM but instead within the TMD (12)(13)(14)(15)(16).
Over the past few decades, multiple models of the sweet receptor's hypothetical ligand binding site have been generated based on the structures of existing sweeteners but without direct knowledge of the nature of the sweet receptor itself. A consensus feature of these models is the presence of A-H-B groups, in which the AH group is a hydrogen donor and the B group is an electronegative center. These models have explanatory and predictive value for some, but not all sweeteners, suggesting that not all sweet ligands bind in the same manner to the receptor. Based on homology to mGluR1, the canonical ligand binding site of the sweet receptor is likely to reside in the cleft between lobes 1 and 2 of the ATD of T1R2. However, additional binding sites may exist elsewhere in T1R2 or in the T1R3 component of the heterodimeric sweet taste receptor. The cysteine-rich region of human T1R3 (hT1R3), which has been shown to be critical for determining responses to intensely sweet proteins, brazzein and monellin (17), is one such potential binding site. Xu et al. (18) and we (19) have recently shown that mutations in VFTM of human T1R2 (hT1R2) affect the sweet receptor's responses to aspartame, whereas mutations in the TMD affect receptor responses to cyclamate.
Lactisole, an arakyl carboxylic acid composed of a carboxyl group and a hydrophobic phenoxyl group, inhibits sweet and umami taste perception by humans but not by rats (20). Here, we report that lactisole inhibits the hT1R2ϩhT1R3 human sweet taste receptor by binding to the TMD of hT1R3. The carboxyl group and aromatic ring of lactisole are necessary for the suppression of sweet taste. In addition, we have identified 7 critical residues within the TMD of T1R3 that form a potential binding pocket for lactisole and are responsible for the sweet receptor's sensitivity to lactisole. We present a model in which lactisole is docked to a potential binding site within the TMD of T1R3.
Residue Numbering-The general numbering of residues in the TMD of T1R3 follows T1R3's primary sequence. Superscripted residue numbers follow the generic numbering system of Ballesteros and Weinstein (21).
Preparation of Chimeras and Point Mutations-Human and mouse T1R2 and T1R3 clones were as described (17). The G␣ 16 -gust44 construct was as described (22). In preliminary studies, the G␣ 16 -gust44 construct gave larger and/or more reliable responses than did G␣ 16 -i3; therefore, G␣ 16 -gust44 was used in all experiments. Human/mouse chimeras of T1R3 were constructed using an overlapping PCR strategy as described (17). The integrity of all DNA constructs was confirmed by sequencing. Point mutations in T1R clones were made by the same overlapping PCR strategy or by site-directed mutagenesis (Stratagene, La Jolla, CA). Combinations of hT1R2 with the following chimeras were not functional (data not shown): mT1R3.h.712-852, mT1R3.h.729 -852, mT1R3.h.751-852, and mT1R3.h.797-852.
Heterologous Expression-HEK293 EBNA (HEK293E) cells were cultured at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and transfected as described (17). Cells for calcium imaging were seeded onto 6-well plates (Corning); cells in each well were co-transfected using Mirus TransIT-293 (Panvera, Madison, WI) with plasmid DNAs encoding T1Rs (0.6 g of each) and G␣ 16 -gust44 (0.5 g). After 24 h, the transfected cells were trypsinized and seeded onto 96-well assay plates (Corning, Corning, NY) at about 40,000 cells/ well in low glucose Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% dialyzed fetal bovine serum (Invitrogen) and 1ϫ GlutaMax-1 (Invitrogen). After an additional 24 h, the cells were washed once with Dulbecco's phosphate-buffered saline (DPBS), loaded with 75 l of 3 M Fluo-4 (Molecular Probe) in DPBS, incubated for 2 h, and then washed twice with DPBS and maintained at 37°C in 50 l of DPBS. The dye-loaded transfected cells in plates were placed into a FlexStation II system (Molecular Devices) to monitor fluorescence (excitation, 488 nm; emission, 525 nm; cutoff, 515 nm) change after the addition of 50 l of DPBS supplemented with 2ϫ tastants. For each trace, tastant or tastant ϩ lactisole was added 30 s after the start of the scan, scanning continued for an additional 150 s, and data were collected every 2 s. Data Analysis-After obtaining a calcium mobilization trace for each sample, calcium mobilization in response to tastants or tastant ϩ lactisole was quantified as the percentage of change (peak fluorescence Ϫ baseline fluorescence level, denoted as ⌬F) from its own baseline fluorescence level (denoted as F). Peak fluorescence intensity occurred about 20 -30 s after the addition of tastants. As controls, buffer alone or compounds that do not taste sweet evoked no change of fluorescence (⌬F/F Ϸ 0, S.E. is about 1%). The data were expressed as the mean Ϯ S.E. of the ⌬F/F value of three independent samples, similarly presented in Jiang et al. (17). The analysis was done automatically by an in-house written SAS program. The bar graph and curving-fitting routines were carried out using Graph-Pad Prism 3.0 (GraphPad Software, Inc.).
Homology Modeling-The multiple sequence alignment among T1Rs, hT1R2, hT1R3, mT1R2, mT1R3, mGluR1, and bovine rhodopsin was generated by the ClustalW program (23) with some manual adjustments to the non-homologous regions (see Supplemental Fig. 2, multiple sequence alignment). The sequence identities (similarities) for TM helices 1-7 between hT1R3 and bovine rhodopsin are 22.6% (61.3%), 13.3% (40%), 22.6% (48.4), 25.0% (33.3%), 11.1% (40.7%), 11.8% (61.8%), and 16% (60%), respectively. Based on the sequence alignment, the hT1R3 TM homology model was constructed by residue replacement using the InsightII (Accelrys, San Diego, CA) biopolymer module. The non-conserved proline kink within TM helix 6 of the model was built by manually shifting 3 residues according to the template. Extracellular loop 3 between TM helix 6 and TM helix 7 was generated by ab initio loop prediction (24,25). The other intracellular and extracellular loop regions were generated by the modeler program (26,27) based on the bovine rhodopsin template. The side chains of the residues in the 7 TM region of hT1R3 and all the loops were refined by the following procedure: 500 steps of steepest descent followed by 2 ps of heating from 0 to 300 K, and 20 ps equilibration and 100 ps of production run at 300 K to relax possible bad contacts of the model followed by energy minimization 1000 steps of steepest descent and 10,000 steps of adopted basis Newton-Raphson or until the root mean square energy gradient of 0.05 kcal/mol/Å was achieved. During the simulation, the ␣-carbon atoms of the helixes were fixed to avoid possible distortion, and the distance-de-pendent dielectric constant (⑀ ϭ 4R ij ) was used to mimic solvation effects and the membrane environment. All model refinements were carried out by the CHARMM program (28) using the CHARMM27 all-hydrogen force field.
Molecular Docking-The geometry of lactisole (S(-)2-(4-methoxyphenoxy)-propanoic acid) was fully optimized by the ab initio quantum chemistry method at the HF/6 -31G* level and followed by a single point calculation with the polarized continuum model solvation method to obtain the electrostatic potentials using the Gaussian98 package (29). Then, the CHELPG charge-fitting scheme (30) was used to calculate partial charges for lactisole. The missing force field parameters for lactisole were obtained from similar parameters taken from CHARMM27 or Quanta (Accelrys, San Diego, CA).
Lactisole was docked into the pocket of the TM region of hT1R3 using the automatic docking program autodock V3.0.5 (31), which uses a powerful Lamarckian genetic algorithm for conformational sampling and docking. The docked conformations of lactisole were analyzed by the cluster analysis of the autodock. The final docked conformation was selected by comparing the available mutagenesis results followed by some manual adjustments of the positions of lactisole and the side chains of hT1R3 before employing model refinement by MD simulations. A similar MD protocol was used for the docked lactisole-hT1R3 complex structure refinement that except the ␣-carbon atoms of the helixes were restricted by 1.0 kcal/mol/Å 2 harmonic restraint force instead of fixing the ␣-carbon atoms during the simulations.
Lactisole suppresses sweetness in humans but not in mice. To determine whether one or both components of the heterodimeric human sweet receptor is required for sensitivity to lactisole, we tested the responses of human, mouse and human ϩ mouse mismatched heterodimers to D-tryptophan with and without lactisole (Fig. 1B). As expected, the fully human pair (hT1R2ϩhT1R3) was sensitive to lactisole, whereas the fully mouse pair (mT1R2ϩmT1R3) was not. One mismatched pair (hT1R2ϩmT1R3) behaved like the fully mouse heterodimer, showing no sensitivity to lactisole. As reported previously (17), the other mismatched pair (mT1R2ϩhT1R3) does not produce a functional receptor, precluding us from analyzing its sensitivity to lactisole. These results demonstrate that hT1R3 is required for lactisole sensitivity. Below, by using human/mouse chimeras of T1R3, we show that hT1R2 is not required for sensitivity to lactisole (Fig. 2).
Lactisole Acts on the TMD of hT1R3-To identify the portion of hT1R3 required for the sweet receptor's sensitivity to lactisole, we examined the responses of heterodimeric receptors comprised of hT1R2 plus human/mouse chimeras in which we combined varying portions of hT1R3 with the complementary portion of mT1R3 ( Fig. 2A). Heterodimers of hT1R2 plus T1R3 chimeras containing most or all of the extracellular region of hT1R3 coupled to the TMD and carboxyl-terminal tail of mT1R3 (i.e. h.1-547.mT1R3 and h.1-567.mT1R3) responded to D-tryptophan but were not inhibited by lactisole (Fig. 2B). In contrast, heterodimers of hT1R2 plus T1R3 chimeras containing most or all of the extracellular region of mT1R3 coupled to the TMD and carboxyl-terminal tail of hT1R3 (i.e. mT1R3.h.548 -852 and mT1R3.h.568 -852) responded to D-tryptophan and were inhibited by lactisole. These results indicate that the inhibitory effect of lactisole on sweet receptor responses to D-tryptophan requires the TMD and/or carboxylterminal tail of hT1R3 but not the ATD of hT1R3. To determine whether hT1R2 is required for sensitivity to lactisole, we examined the responses to D-tryptophan and D-tryptophan ϩ lactisole of the mT1R3.h.568 -852 chimera paired with hT1R2 ( Fig. 2B) or with mT1R2 (Fig. 2C). Unlike the non-functional pairing of mT1R2ϩhT1R3 (Fig. 1B), mT1R2 can function in combination with the mT1R3.h.568 -852 chimera. Both mT1R2ϩmT1R3.h.568 -852 and hT1R2ϩmT1R3.h.568 -852 respond to D-tryptophan, and these responses are fully sup-pressed by the addition of lactisole. Thus, hT1R2 is not required for sensitivity to lactisole, indicating that the hT1R3 component of the heterodimeric receptor is necessary and sufficient for sensitivity to lactisole.
T1R3 chimeras in which we replaced TM helices 5 and 6 or TM helix 5 alone of hT1R3 with the corresponding TM helices of mT1R3 (i.e. hT1R3(m734 -792) and hT1R3(m734 -756), respectively) functioned in combination with hT1R2 to respond to D-tryptophan but displayed no sensitivity to lactisole (Fig. 3C). Conversely, T1R3 chimeras in which we replaced TM helices 5 and 6 or TM5 alone of mT1R3 with the corresponding TM helices of hT1R3 (i.e. mT1R3(h729 -787) and mT1R3(h729 -751), respectively) functioned in combination with hT1R2 to respond to D-tryptophan and were sensitive to lactisole (Fig.  3C). Thus, at least some residues required for the humanspecific sensitivity to lactisole lie within or adjacent to TM helix 5 of hT1R3.
Alanine 733 5.46 in hT1R3 TM Helix 5 Is Required for Sensitivity to Lactisole-In the TM helix 5 region of hT1R3 (amino acids 729 -751), only 4 residues (Phe-730 5.43 , Ala-733 5.46 , Ala-735 5.48 , and Thr-739 5.52 ) differ between the human and mouse forms of T1R3. To identify which residue(s) within this region are required for sensitivity to lactisole, we individually substituted each of these 4 residues with the corresponding mouse T1R3 residues (hT1R3(F730 5.43 L), hT1R3(A733 5.46 V), hT1R3(A735 5.48 I), and hT1R3(T739 5.52 M)). We then examined the responses of heterodimers of hT1R2 with these substituted forms of hT1R3. All mutants responded to D-tryptophan; however, only the A733 5.46 V mutant showed diminished sensitivity to lactisole (Fig. 4A). In comparison with wild type hT1R3, the A733 5.46 V mutant showed a right shift in its dose-response curve to lactisole (A733 5.46 V mutant IC50 of 1.9 ϫ 10 Ϫ3 M versus wild type IC50 of 4.1 ϫ 10 Ϫ5 M) (Fig. 4D) but was unchanged in its dose-response curve to D-tryptophan (A733 5.46 V mutant EC 50 of 3.5 mM versus wild type EC 50 of 4.3 mM) (Fig. 4C). Thus, valine substitution at position 733 5.46 selectively affects the receptor's sensitivity to lactisole.
As noted above, the results with chimeras h.1-787.mT1R3 and h.1-812.mT1R3 (Fig. 3B) 4. Sensitivity to lactisole depends on several human-specific residues within the TMD of hT1R3. A, upper panel, alignment of human and mouse T1R3 sequences shows that in TM helix 5, only 4 residues differ between these two species. Lower panel, HEK293E cells were transfected with G␣ 16 -gust44, hT1R2, and hT1R3 mutants in which species-variant positions within TM helix 5 were substituted with the mouse sequence. Only receptors containing hT1R3(A733 5.46 V) showed a pronounced insensitivity to lactisole. Receptors containing any of the other hT1R3 substitutions (F730 5.43 L, A7351 5.48 I, T739 5.52 M) responded normally to lactisole. B, upper panel, alignment of human and mouse T1R3 sequences shows that in TM helix 7 and the adjacent third extracellular loop (EX3), 8 residues differ between these two species. Lower panel, transfection and assays of responses to D-tryptophan with or without lactisole are as for panel A except that these hT1R3 mutants were substituted with the mouse sequence at species-variant positions within TM helix 7 and EX3. Receptors containing hT1R3(R790 ex3 Q) or hT1R3(L798 7.36 I) are less sensitive to the lower concentration of lactisole (0.125 mM) than were wild type hT1R3-containing receptors; however, the higher concentration of lactisole (1.25 mM) suppresses the responses of these mutant receptors. All other mutant hT1R3-containing receptors (i.e. V788 ex3. A, L789 ex3. Y, L800 7.38 V, V802 7.40 A, A807 7.45 V, A808 7.46 T) responded normally to both concentrations of lactisole. C, left panel, dose-response profiles toward D-tryptophan of receptors containing hT1R2 plus selected hT1R3 TM helix 5 or TM helix 7 mutants. All receptors responded similarly to D-tryptophan. Right panel, dose-response tivity to lactisole. In the extracellular loop 3 and TM helix 7 region of hT1R3 (amino acids 787-812), 8 residues (Val-788 ex3 , Leu-789 ex3 , Arg-790 ex3 , Leu-798 7.36 , Leu-800 7.38 , Val-802 7.40 , Ala-807 7.45 , and Ala-808 7.46 ) differ between the human and mouse forms of T1R3. To identify which residue(s) within this region are required for sensitivity to lactisole, we individually substituted each of these 8 residues with the corresponding mouse T1R3 residues (hT1R3(V788 ex3 A), hT1R3(L789 ex3 Y), hT1R3(R790 ex3 Q), hT1R3(L798 7.36 I), hT1R3(L800 7.38 V), hT1R3(V802 7.40 A), hT1R3(A807 7.45 V), and hT1R3(A808 7.46 T)) and then examined the responses of these substituted forms of hT1R3 heterodimerized with hT1R2. All of these mutants responded to D-tryptophan; only the R790 ex3 Q and L798 7.36 I mutants showed diminished sensitivity to lactisole (Fig. 4B). Both of these mutants were unchanged in their dose-response curves to D-tryptophan (Fig. 4C) and right-shifted in their dose-response curves to lactisole (Fig. 4C). The effect on lactisole sensitivity of mutating either of these residues was much less pronounced than that seen with A733 5.46 V (R790 ex3 Q mutant IC50 of 1. Ala-733 5.46 Substitutions Affect hT1R3 Responses to Lactisole-To investigate the physicochemical effects of residue Ala-733 5.46 within TM helix 5 of hT1R3 on receptor activity toward lactisole, we made various substitutions at this position (Fig.  4D). Substitution of Ala-733 5.46 by glutamate or phenylalanine abolished the receptor's sensitivity to lactisole. Substitution of Ala-733 5.46 by glycine or serine did not affect the receptor's sensitivity toward lactisole. The threonine substitution (A733 5.46 T) led to slightly reduced receptor sensitivity to lactisole.
hT1R3 Sensitivity to Lactisole Also Involves Residues in TM Helices 3 and 6 -The solved structure of rhodopsin has been used as a template, after sequence alignment, to model the structures of the TMDs of certain family C receptors (mGluR1, mGluR5, and CaSR (13)(14)(15)34)). Such models can be validated by mutating residues predicted to be within binding pockets followed by characterization of the mutant receptor's properties. Rhodopsin's retinol binding pocket contains 17 residues, all of which are within 6.0 Å of retinol in the rhodopsin crystal structure. Based on the alignment of T1R3 with rhodopsin, we picked 17 residues lining a potential binding pocket of T1R3. Each of these residues was substituted with alanine, and the mutants were tested for responsiveness to D-tryptophan and lactisole (Fig. 5A).
Receptors with alanine substitutions at any of five positions (Leu-644 3.36 , Thr-645 3.37 , Tyr-771 6.44 , Gln-794 7.32 , and Ile-805 7.43 ) showed no response to D-tryptophan (Fig. 5A) or to other sweeteners such as aspartame (data not shown). These positions might be required for generating the active state of the receptor or for the structural integrity of the receptor.
Molecular Modeling of hT1R3's TMD-We molecularly modeled the TM region of wild type hT1R3 and automatically docked lactisole to the predicted binding pocket (see "Experimental Procedures"). The optimal structure of hT1R3 was obtained after building the structure using rhodopsin as a template (35), manual adjustment of the potential kink introduced by Pro-780 6.53 in TM helix 6, and energy minimization. Lactisole is a racemic mixture, with only the S-form active as a sweetness suppressor. 2 Therefore, we used the S-form of lactisole in our docking experiment. After running the AUTODOCK program, we found that lactisole can take on a few conformational states when automatically docked into the TMD binding pocket. By comparing the possible models with our mutagenesis data, we found that only one of the docked structures could satisfy the constraints placed on it by the mutagenesis data (Fig. 6). In this model, His-641 3.33 is 3 Å from the carboxylic group of S-lactisole, Phe-778 6.51 is close to the phenoxyl ring of S-lactisole, and the methoxyl group of S-lactisole is pointing toward Ala-733 5.46 . From this model, we also see that Gln-637 is close to the carboxylic group of S-lactisole. To further test our model, we mutated Gln-637 3.29 to glutamate to test whether the negatively charged group has a detrimental effect on lactisole's activity. Indeed, this mutant was devoid of lactisoleinduced suppression of D-tryptophan activity (Supplemental Fig. 4).

DISCUSSION
Lactisole Broadly Inhibits Human Sweet Taste-Lactisole, an arakyl carboxylic acid, inhibits the perception of sweet and umami taste in humans but not in rodents (20). In humans, 2 Gordon Birch, personal communication.
Lactisole's inhibition of both sweet and umami taste suggests that it might target a shared component of these two signaling pathways, e.g. T1R3. Consistent with this inference, preliminary in vitro studies have shown that lactisole inhibits the sucrose-induced responses of cells heterologously expressing hT1R2ϩhT1R3 (2). More recently, and independently of the present study, Xu et al. (18) have determined that lactisole's effects require the human form of T1R3's TMD and/or cytoplasmic domain but do not require human-specific elements within T1R3's ATD. These authors did not narrow down the specific region within the TMD/cytoplasmic domain that serves as lactisole's binding pocket.
Lactisole Acts on hT1R3's TMD-In the present study, we set out to molecularly identify the target of lactisole's broadly acting inhibitory effects on human sweet taste, to determine why humans but not mice are susceptible to lactisole's effects, and to understand mechanistically how lactisole inhibits sweet taste. First, we used a cell-based assay to determine that lactisole acts to inhibit heterologously expressed human sweet taste receptor (hT1R2ϩhT1R3) responses to all members of a panel of chemically diverse sweeteners. Using this same assay, we determined that lactisole does not inhibit the responses of the mouse sweet taste receptor (mT1R2ϩmT1R3), consistent with its in vivo ineffectiveness in mice. Next, by comparing differences in sensitivity toward lactisole among interspecies (mouse ϩ human) combinations of T1R2ϩT1R3, we demonstrated that lactisole's target is hT1R3; either hT1R2 or mT1R2 in combination with the appropriate T1R3 partner (hT1R3 or its chimeras) equivalently supports sensitivity to lactisole. Then, using human/mouse chimeras of T1R3 in combination with T1R2, we determined that it is the TM domain of hT1R3 that specifies sensitivity to lactisole; the species of origin of the extracellular ATD and the C-rich region of T1R3 were irrelevant to sensitivity to lactisole. Using additional interspecies chimeras, we identified which human TM helices were required for sensitivity to lactisole, and then, using substitution mutants, we determined which specific human versus mouse variant residues within these TMs mattered. Specifically, Ala-733 5.34 in TM helix 5, Leu-798 7.36 in TM helix 7, and Arg-790 ex3 FIG. 5. Alanine scan of candidate lactisole binding pocket residues in hT1R3's TMD. A, residues of hT1R3 identified by homology to rhodopsin as potentially contributing to the lactisole binding pocket were replaced by alanine. hT1R2 and hT1R3 mutants along with G␣ 16 -gust44 were co-transfected into HEK293E cells. The responses of the transfected cells to D-tryptophan (10 mM) were examined in the presence or absence of lactisole (1.25 mM). Mutant-containing receptors that were unresponsive to D-tryptophan (L644 3.36 A, T645 3.37 A, Y771 6.44 A, Q794 7.32 A, and I805 7.43 A) could not be tested for sensitivity to lactisole. H641 3.33 A, H734 5.47 A, W775 6.48 A, F778 6.51 A, and V779 6.52 A mutant-containing receptors had reduced or absent sensitivity to lactisole, although the responses of H734 5.47 A andW775 6.51 A to D-tryptophan were sufficiently impaired that quantifying the effect of lactisole on the D-tryptophan response of each of these mutants was not possible. S640 3.32 A and L782 6.55 A mutantcontaining receptors were more sensitive to lactisole. Q636 3.28 A, H721 Ex2-49 A, R723 Ex2-51 A, S729 5.42 A, and V776 6.52 A mutant-containing receptors had lactisole sensitivity comparable with wild type (WT). B, left panel, dose-response curves toward D-tryptophan of receptors containing hT1R2 plus selected alanine scan mutants. All receptors responded similarly to D-tryptophan. Right panel, dose-response curves toward D-tryptophan (10 mM) plus varying concentrations of lactisole of receptors containing hT1R2 plus selected alanine scan mutants. Receptors containing H641 3.33 A lost all sensitivity to lactisole. Receptors containing F778 6.51 A showed sensitivity toward lactisole that is greatly right-shifted when compared with that of wild type (IC50 1.8 ϫ 10 Ϫ3 M). Receptors containing S640 3.32 A or L782 6.55 A showed enhanced sensitivity toward lactisole (IC50 1.4 ϫ 10 Ϫ5 M and 1.2 ϫ 10 Ϫ5 M, respectively). in extracellular loop 3 were found to be critically important for the human-specific sensitivity to lactisole.
Our results with lactisole sensitivity of expressed hT1R2ϩhT1R3 are generally consistent with the human psychophysical studies. The only notable difference was that in vitro responses of hT1R2ϩhT1R3 to neohesperidin dihydrochalcone and thaumatin were suppressed by lactisole, whereas in vivo, human subjects found the sweetness of these two compounds to be insensitive to lactisole. The reason for the discordance is unknown. One intriguing possibility is that sweet receptors other than, or in addition to, T1R2ϩT1R3 respond to these two compounds and underlie the lactisole-insensitive responses seen in psychophysical tests.
hT1R3's Lactisole Binding Pocket-To obtain a better mechanistic understanding of how lactisole interacts with and inhibits the human sweet taste receptor, we carried out two additional sets of studies, both of which depended on modeling the TMD of hT1R3 based on homology to the known structure of rhodopsin. First, alignment with the retinol binding pocket of rhodopsin was used to identify a potential ligand binding pocket within hT1R3's TMD. Then, alanine-scanning mutagenesis of these potential ligand binding residues in hT1R3 identified 4 residues that had large effects on the sensitivity to lactisole: Ser-640 3.32 and His-641 3.33 in TM helix 3 and Phe-778 6.51 and Leu-782 6.55 in TM helix 6.
In the second study, we constructed a homology model of hT1R3's entire TMD based on the solved crystal structure of rhodopsin and used this model to identify a plausible binding site for lactisole. Our model predicts that lactisole binds to hT1R3 within a pocket formed by TM helices 3, 5, and 6. This model provides a framework to explain the results from our mutagenesis experiments, identify the contributions of residues in each of these three TM helices, and make a number of testable predictions. For this model to be useful, it needs to explain mechanistically the two types of altered responses to lactisole we observed in our mutants: (a) reduced sensitivity to lactisole (seen with mutation of residues His-641 3.33 , Ala-733 5.46 , and Phe-778 6.51 ) and (b) increased sensitivity to lactisole (seen with mutation of residues Ser-640 3.32 and Leu-782 6.55 ).
How might mutations of His-641 3.33 , Ala-733 5.46 , and Phe-778 6.51 reduce lactisole's ability to inhibit receptor activity? At physiological pH, His-641 3.33 within TM3 is likely protonated, and its positive charge is predicted to form a salt bridge with the negatively charged carboxyl group of lactisole. Alanine substitution of His-641 3.33 would eliminate this potential salt bridge, explaining this mutant's total loss of sensitivity to lactisole.
Within TM helix 5, replacement of Ala-733 5.46 by valine (the mouse equivalent), glutamate, or phenylalanine greatly reduced the receptor's sensitivity to lactisole, whereas glycine, serine, or threonine substitutions here had modest or no effect on sensitivity to lactisole. These effects are consistent with our model, which places lactisole's methoxy group in close proximity with the side chain of Ala-733 5.46 . Larger aliphatic or aromatic side chains at this position are likely to reduce sensitivity toward lactisole by their steric effects. Tolerance of serine and threonine substitutions suggests that these residues' hydroxyl groups may have favorable hydrophobic interactions with lactisole's methoxy group. Within TM helix 6, replacement of Phe-778 6.51 (conserved among class-C GPCRs) with alanine greatly reduced the receptor's sensitivity to lactisole. Our model shows sufficient proximity of Phe-778 6.51 's side chain to the phenoxyl ring of lactisole for ainteraction to occur.
Alanine substitution of Ser-640 3.32 or Leu-782 6.55 A markedly increased the sensitivity of the receptor to inhibition by lactisole. In our model, Leu-644 3.36 , Phe-778 6.51 , and Leu-782 6.55 contribute to the lactisole binding pocket via hydrophobic interactions with lactisole. The Leu-782 6.55 A substitution would maintain hydrophobicity and diminish any steric constraints. Replacement of Leu-782 6.55 with asparagine or threonine slightly reduced the sensitivity of the receptor to lactisole (data not shown), consistent with our proposal that decreased hydrophobicity of this pocket would diminish the strength of the interaction between the receptor and lactisole.
The Ser-640 3.32 A substitution may reduce steric constraints by replacing a polar side group with a smaller hydrophobic one. Smaller side chains at either 640 3.32 or 782 6.55 may allow lactisole to better bridge the pocket and lock it into the ground state. Consistent with this proposal is our recent observation that substitution of Ser-640 3.32 with valine (similar in size but hydrophobic) decreased sensitivity to lactisole (data not shown) Out-of-Pocket Expanses-Our model implicates TM helices 3, 5, and 6 of hT1R3 in lactisole binding; however, our chimeric and mutational studies also identified Leu-798 7.36 in TM helix 7 and Arg-790 ex3 in extracellular loop 3 (which connects TM helices 6 and 7) as human-specific residues that affect responsiveness to lactisole. These residues appear to be too far away from the predicted binding pocket to be exerting direct effects on lactisole binding. TM helix 7 and extracellular loop 3 may affect agonist/antagonist activity by acting on TM helix 6. According to our model, Arg-790 ex3 is close enough to Glu-717 ex2 of extracellular loop 2 to form a salt bridge, and this may help to stabilize the top of the TM bundle and "cap" the lactisole binding site.
Common Motifs with Other GPCRs-T1R3's binding pocket for lactisole shares a great deal of similarity with ligand binding pockets of many other GPCRs. For instance, the Asp in position 3.32 is the principal anchoring residue of all monamine receptors (Asp 3.32 ) and CaSR (15,36). Phe-684 3.32 in CaSR is involved in the recognition of allosteric modulators (16). Ala-117 3.32 and Thr-118 3.33 as well as other TM helix 3 residues in rhodopsin make important contacts mainly with the retinylidene group of 11-cis retinal (35). Position 3.33 (Tyr-106 3.33 ) in the thyrotropin-releasing hormone receptor forms a hydrogen-bonding interaction with thyrotropin-releasing hormone (37). In rhodopsin, Thr-118 3.33 , Tyr-268 6.51 , and Ile-189 ex2 determine the position of the C9methyl group of the retinylidene group (35). Position 6.51 is pivotal for recognition of agonist or antagonist of many GPCRs, including CaSR, mGluR1, rhodopsin, and thyrotropin-releasing hormone receptor (13,15,38). Position 6.55 participates in ligand recognition in serotonin 5-HT2B receptor and M3 muscarinic acetylcholine receptor (ACM3) (39,40).
How does the heterodimeric sweet receptor transit into the active form, and how does lactisole block this? In the case of rhodopsin, the activation mechanism occurs as TM helix 6 tilts away from TM helix 3; this results in the cytoplasmic end of TM helix 6 coming closer to the cytoplasmic end of TM helix 5 (41)(42)(43). We speculate that a similar activation mechanism holds for the sweet taste receptor. Lactisole's sweet antagonism could be mediated by its binding to TM helices 3, 5, and 6 of hT1R3 so as to restrict movement required for conversion to the active state. In effect, lactisole locks the receptor into the ground state. This would explain how lactisole binding to the TMD of hT1R3 can "act at a distance" to antagonize sweeteners, which are thought to bind to the ATD of T1R2.