Sweet taste of heavy water

Hydrogen to deuterium isotopic substitution has only a minor effect on physical and chemical properties of water and, as such, is not supposed to influence its neutral taste. Here we conclusively demonstrate that humans are, nevertheless, able to distinguish D2O from H2O by taste. Indeed, highly purified heavy water has a distinctly sweeter taste than same-purity normal water and can add to perceived sweetness of sweeteners. In contrast, mice do not prefer D2O over H2O, indicating that they are not likely to perceive heavy water as sweet. HEK 293T cells transfected with the TAS1R2/TAS1R3 heterodimer and chimeric G-proteins are activated by D2O but not by H2O. Lactisole, which is a known sweetness inhibitor acting via the TAS1R3 monomer of the TAS1R2/TAS1R3, suppresses the sweetness of D2O in human sensory tests, as well as the calcium release elicited by D2O in sweet taste receptor-expressing cells. The present multifaceted experimental study, complemented by homology modelling and molecular dynamics simulations, resolves a long-standing controversy about the taste of heavy water, shows that its sweet taste is mediated by the human TAS1R2/TAS1R3 taste receptor, and opens way to future studies of the detailed mechanism of action. Ben Abu, Mason and colleagues use molecular dynamics, cell-based experiments, mouse models, and human subjects to determine that, unlike ordinary water, heavy water tastes sweet to humans, but not mice. Mechanistically, this effect is mediated by the human TAS1R/TAS1R3 sweet taste receptor.

H eavy water, D 2 O, has fascinated researchers since the discovery of deuterium by Urey in 1931 1,2 . The most notable difference in physical properties between D 2 O and H 2 O is the roughly 10% higher density of the former liquid, which is mostly a trivial consequence of deuterium being about twice as heavy as hydrogen. A more subtle effect of deuteration is the formation of slightly stronger hydrogen (or deuterium) bonds in D 2 O as compared to H 2 O 3,4 . This results in a small increase of the freezing and boiling points by 3.8°C and 1.4°C, respectively, and in a slight increase of 0.44 in pH (or pD) of pure water upon deuteration 5 . In comparison, a mere dissolution of atmospheric CO 2 and subsequent formation of dilute carbonic acid in open containers has a significantly stronger influence on the pH of water, changing it by more than one unit 6 .
Biological effects are observable for high doses of D 2 O. While bacteria or yeasts can function in practically pure D 2 O, albeit with somewhat hindered growth rate [7][8][9] , for higher organisms damaging effects on cell division and general metabolism occur at around 25% deuteration, with lethal conditions for plants and animals typically occurring at~40-50% deuteration of the body water 2,10,11 . Small levels of deuteration are, nevertheless, harmless. This is understandable given the fact that about 1 in every 6400 hydrogens in nature is found in its stable isotope form of deuterium 12 . Oral doses of several milliliters of D 2 O are safe for humans 13 and are used in the isotopic form D 2 18 O for metabolic measurements in clinical praxis (known as "doubly labeled water" technique) 14 . Probably the best-established effect of D 2 O is the increase of the circadian oscillation length upon its administration to both animals and plants. This has been attributed to a general slowdown of metabolism upon deuteration, although the exact mechanism of this effect is unknown 15,16 .
A long-standing unresolved puzzle concerns the taste of heavy water. There is anecdotal evidence from the 1930s that the taste of pure D 2 O is distinct from the neutral one of pure H 2 O, being described mostly as "sweet" 17 . However, Urey and Failla addressed this question in 1935 concluding authoritatively that upon tasting "neither of us could detect the slightest difference between the taste of ordinary distilled water and the taste of pure heavy water" 18 . This had, with a rare exception 19 , an inhibitive effect on further human studies, with research concerning effects of D 2 O focusing primarily on animal or cell models. Experiments in animals indicated that rats developed aversion toward D 2 O when deuteration of their body water reached harmful levels, but there is conflicting evidence regarding their ability to taste heavy water or use other cues to avoid it 20,21 .
Within the last two decades, the heterodimer of the taste receptor of the TAS1Rs type of G protein-coupled receptors (GPCRs), denoted as TAS1R2/TAS1R3, was established as the main receptor for sweet taste 22 . The human TAS1R2/ TAS1R3 heterodimer recognizes diverse natural and synthetic sweeteners 23 . The binding sites of the different types of sweeteners include an orthosteric site (a sugar-binding site in the extracellular Venus flytrap domain of TAS1R2) and several allosteric sites, including sites in the extracellular regions of the TAS1R2 and TAS1R3 subunits and in the transmembrane domain of TAS1R3 24,25 (Fig. 1). Additional pathways for sweet taste recognition have also been suggested, involving glucose transporters and ATP-gated K + channel 26,27 .
Interestingly, not all artificial sweeteners are recognized by rodents 28 . Differences in human and rodent responses to tastants, as well as sweetness inhibitors such as lactisole, have been useful for delineating the molecular recognition of sweet compoundsusing human-mouse chimeric receptors, it was shown that the transmembrane domain (TMD) of human TAS1R3 is required for the activating effects of cyclamate 29 and for the inhibitory effect of lactisole 30 .
A combination of TAS1R3 with another member of the TAS1R family, TAS1R1, results in a dimer that mediates umami taste, elicited by molecules such as glutamate or, in case of the rodent umami receptor, other L-amino acids 31 . Bitter taste is mediated by the taste 2 receptor (TAS2R) gene family 32 , a branch of Family A GPCRs 24 . The human genome has 25 TAS2R subtypes and over a thousand of bitter compounds are currently known 33 , with numerous additional bitter tastants predicted 34 .
In this study, we systematically address the question of sweet taste of heavy water by a combination of sensory experiments in humans, behavioral experiments in mice, tests on sweet taste receptor-transfected cell lines, and computational modeling including molecular dynamics (MD) simulations. This combined approach consistently leads to the conclusion that the sweet taste of pure D 2 O is a real effect for human subjects due to activation of the TAS1R2/TAS1R3 sweet taste receptor. While present simulations show, in accord with previous experiments 35 , that proteins are systematically slightly more rigid and compact in D 2 O than in H 2 O, the specific molecular mechanism of the heavy water effect on the TAS1R2/TAS1R3 receptor remains to be established.

Results and discussion
Water purity. We have paid great attention to the purity of the water samples, further degassing and redistilling under vacuum the purest commercially available D 2 O and H 2 O. The lack of nonnegligible amounts of organic impurities was subsequently confirmed by gas chromatography with mass spectrometry analysis and by experiments with water samples at different levels of purification, see Supplementary Information (SI), Figures S1 and S2. This is extremely important-note in this context that "the vibrational theory of olfaction", which suggested distinct perception of deuterium isotopes of odorants due to difference in their vibrational spectra 36 , has been refuted with some of the observed effects turning out to be due to impurities 37,38 .  Figure S3 in SI.
Next, the perceived sweetness of D 2 O in increasing proportion to H 2 O was reported using a 9-point scale, labeled also with verbal descriptions of perceived intensity (1 = no sensation, 3 = slight, 5 = moderate, 7 = very much, 9 = extreme sensation). Sweetness was shown to increase in a D 2 O-dose-dependent manner, reaching average 3.3 ± 0.4 sweetness ("slight" sweetness) (Fig. 2a). The perceived sweetness of low concentrations of caloric D-glucose (Fig. 2b), sucrose (Fig. 2c), and an artificial sweetener cyclamate ( Fig. 2d) was tested when dissolved in H 2 O or in D 2 O, in order to check whether the slight sweetness of D 2 O adds on top of slight sweetness of known sweeteners. As expected 39 , Dglucose was perceived as less sweet compared to sucrose at the same concentration ( Fig. 2b and c). D 2 O added to the perceived sweetness of all tested concentrations of D-glucose and cyclamate (see Figure S4 in the SI for cyclamate results excluding two outliers). The sweetness of the two lowest concentrations of sucrose was significantly higher when dissolved in D 2 O compared to H 2 O.
We then checked whether the stand-alone and additive effect of D 2 O is sweetness-specific or general, whereby D 2 O might elicit other tastes, or add to their intensity. Savory (umami) taste of monosodium glutamate (MSG) and bitter taste of quinine, which are also taste modalities mediated by GPCR receptors expressed in taste cells, were tested in regular and in heavy water. The intensity of savory taste of MSG in D 2 O did not differ from that in H 2 O (Fig. 2e), while the perceived bitterness of quinine was in fact slightly reduced in D 2 O compared to quinine in H 2 O (Fig. 2f). This is in agreement with the known effect of sweeteners as maskers of bitter taste, that may be due to both local interactions and sensory integration effects [40][41][42] . Thus, we have ascertained that D 2 O is sweet and adds to the sweetness of other sweet molecules, but not to the intensity of other GPCR-mediated taste modalities.
Experiments with mice. Next, we addressed the question whether the sweetness of D 2 O is perceived also by rodents. Lean mice of the C57BL/6J strain were drinking pure H 2 O, D 2 O, or a 43 mmol/l H 2 O sucrose solution for 16 h during a night period. Namely, each of the three groups of mice had a choice from two bottles containing (i) H 2 O and D 2 O, (ii) H 2 O and sucrose solution, or (iii) H 2 O and H 2 O (as a control). The food intake was unaffected in all groups (see SI, Figure S5 and Table S1).
The results of the drinking experiments are presented in Fig. 3a-c, with a snapshot of the experimental setup shown in Fig. 3d. In cages where mice were offered both normal water and heavy water (Fig. 3a) consumption of D 2 O was within statistical error the same as that of H 2 O. Previous reports have shown that on longer timescales than those reported here mice learned to avoid D 2 O, as it is poisonous to them in larger quantities 10 . It is not clear what is the cue that enables the avoidance learning, but it is evident that the early response to D 2 O is not attractive, suggesting that it is not eliciting sweet taste in mice.
By contrast, mice exhibit a strong preference for sucrose solution over H 2 O. Indeed, the consumed volume was significantly increased in line with the predilection of mice for sucrose solutions (Fig. 3b). The amount of H 2 O consumed by the control group from either of the two bottles, both containing H 2 O, is depicted in Fig. 3c. Overall, the data shows that in all three experiments mice consumed comparable amounts of H 2 O and D 2 O, with significant increase of consumption of the sucrose solution.
Assessing involvement of TAS1R2/TAS1R3 receptor using human sensory panel. The chemical dissimilarity of D 2 O from sugars and other sweeteners raises the question whether the effect we observed in human subjects is mediated by TAS1R2/TAS1R3, which is the major receptor for sweet taste 22 . This was first explored by combining water samples with lactisole as an established TAS1R2/TAS1R3 inhibitor 30 . Using the two-alternative forcedchoice (2AFC) method, in which the participant must choose between two samples, 18 out of 25 panelists chose pure D 2 O as sweeter than D 2 O + 0.9 mM lactisole solution (p < 0.05, Fig. 4a). In an additional experiment, the sweetness of pure D 2 O was scored significantly higher than that of D 2 O + 0.9 mM lactisole solution (p = 0.0003), while the same amount of lactisole had no effect on the perception of sweetness of H 2 O that served as control (Fig. 4b).
These results suggest that D 2 O elicits sweetness via the TAS1R2/ TAS1R3 sweet taste receptor.
Cell-based experiments for establishing the role of TAS1R2/ TAS1R3. To confirm the involvement of the sweet taste receptor TAS1R2/TAS1R3 in D 2 O signaling we performed functional calcium mobilization assays using HEK 293 FlpIn T-Rex cells heterologously expressing both required TAS1R subunits as well as the chimeric G protein Gα15Gi3 43,44 . As seen in Fig 45,46 , while lactisole exposure had no effect on cells treated with pure H 2 O water, as expected (Fig. 5c). As a control, 960 mM D-glucose elicited increase in IP1 levels in TAS1R2/TAS1R3 expressing cell, which was inhibited in the presence of lactisole.
We further used an IP1 assay 45,46 on non-transfected HEK 293T cells, where we observed that dose-dependent curves of carbachol-an agonist of the endogenous muscarinic receptor 3 (M3) 47 -did not show any difference between H 2 O and D 2 O-based media (Fig. 6a) and that cell medium that had either 10% or 100% D 2 O, did not activate basal IP1 accumulation (Fig. 6b). Next, TAS1R2/TAS1R3 receptor along with the chimeric Gα16gust44 subunit 44,48 were transiently expressed, and the functionality was illustrated by dose-dependent response to D-glucose (Fig. 6c). Finally, and in agreement with calcium imaging, we found that 10% D 2 O activated these cells. Activation by 100% D 2 O was even more pronounced (Fig. 6d).
Molecular modeling. The cellular response results further support the hypothesis that the sweet taste of D 2 O is mediated via the TAS1R2/TAS1R3 receptor. Various mechanisms governing this effect can be envisioned. As a potential suspect, we focus on a direct effect on the sweet taste receptor, narrowing on the TAS1R3 TMD (see Fig. 1), as it is already known to be a modulation site with functional differences between humans and rodents 23,29,30 . Furthermore, water-binding sites were discovered at the TMD of many GPCRs 49,50 , suggesting a potential target for D 2 O binding. We modeled the human TAS1R3 TMD using the I-TASSER server 51 . Positions of H 2 O molecules were compared among mGluR5 structures (PDB: 4OO9, 5CGC, and 5CGD) and two conserved positions were found. The H 2 O molecules in these two positions were merged with the TAS1R3 model and minimized (Fig. 7a). The water mapping protocol from OpenEye 52 enables mapping of water positions based on the energetics of water, and~40 water molecules were predicted in the binding site using this protocol (Fig. 7a). Water densities of H 2 O and D 2 O in the TMD of the TAS1R2/TAS1R3 receptor were calculated from MD simulations as described below. Overall, all three methods suggest the possibility for at least some internal molecules (trapped in the TMD bundle) in addition to water that surrounds the extracellular and intracellular loops (Fig. 7a).
Next, we carried out microsecond MD simulations of the TMD embedded in a phospatidylcholine (POPC) bilayer in either H 2   , and MSG (f) taste-specific intensity. Asterisks indicate a significant (p < 0.05) difference between water types using the twoway analysis of variance (ANOVA) with a preplanned comparison t-test. All data are presented as the mean ± the Standard Error of Measurement (SEM); n = 15-30 (4-12 males). The y axis shows the response for individual modalities, while the x axis is labeled with different water samples. Scale for each modality is labeled as 1 = no sensation, 3 = slight, 5 = moderate, 7 = very much, and 9 = extreme sensation.   Figure 7b shows the time evolution of the radius of gyration of the TMD domain, while Fig. 7c and d presents the root mean square fluctuations (RMSF) of individual residues of the proteins superimposed on its structure and plotted in a graph together with the mean value of RMSF. A small but significant difference is apparent in the behavior of the protein in H 2 O vs D 2 O. Namely, structural fluctuations of most residues (particularly those directly exposed to the aqueous environment) and of the protein as a whole are slightly attenuated in D 2 O, in which environment the protein is also somewhat more compact than in H 2 O (Fig. 7b). Additional simulations on other representative systems show that the rigidifying effect of heavy water is apparent also in small soluble proteins (see SI, Figs. S6-S8).
Summary and outlook. Sweet taste never ceases to surprise. Over a decade ago, water was shown to elicit sweet taste by rinsing away inhibitors of sweet taste receptors, both in human sensory experiments and in cell-based studies. This effect was explained in terms of a two-state model, where the receptor shifts to its activated state when released from inhibition by rinsing with water 43 . Here, we have studied the taste of D 2 O and H 2 O per se, not related to washing away of sweet taste inhibitors. Using psychophysics protocols, we show that humans differentiate between D 2 O and H 2 O based on taste alone. Importantly, by employing gas chromatography/mass spectrometry analysis we demonstrate that the sweet taste of deuterated water is not due to impurities. Being only isotopically different from H 2 O, in principle, D 2 O should be indistinguishable from H 2 O with regard to taste, namely it should have no taste of its own. Yet, we illustrate that human subjects consistently perceive D 2 O as being slightly sweet and significantly sweeter than H 2 O. Furthermore, D 2 O added to perceived sweetness of low concentrations of other sweeteners. In contrast, it did not elicit umami or bitter taste on its own, nor did it add to the umami taste perception of MSG. D 2 O did not add to the bitterness of quinine, and reduced the perceived bitterness of 0.1 mM quinine, in agreement with the known effect of bitterness suppression by sweet molecules.
A further important finding is that lactisole, which is an established blocker of the TAS1R2/TAS1R3 sweet taste receptor that acts at the TAS1R3 transmembrane domain 30 , suppresses both the sweet perception of D 2 O in sensory tests, as well as the activation of TAS1R/TAS1R3 in calcium imaging and in IP1 cell-based assays. In support of these observations we have demonstrated that HEK 293T cells transfected with TAS1R2/TAS1R3 and Gα16gust44 chimera, but not the non-transfected cells, are activated by D 2 O, as measured by IP1 accumulation compared to control values. Finally, taste experiments on mice show that these animals do not prefer D 2 O over H 2 O.
Our findings point to the human sweet taste receptor TAS1R2/ TAS1R3 as being essential for sweetness of D 2 O. Molecular dynamics simulations show, in agreement with experiment 35 , that proteins in general are slightly more rigid and compact in D 2 O than in H 2 O. At a molecular level, this general behavior may be traced back to the slightly stronger hydrogen bonding in D 2 O vs H 2 O, which is due to a nuclear quantum effect, namely difference in zero-point energy 3,4 . Biologically relevant situations where one may expect strong nuclear quantum effects as implications of H/D substitution directly involve proton or deuteron transfer 9 . Unless a yet unknown indirect mechanism is involved, this is not the case for the TAS1R2/TAS1R3 sweet taste receptor, thus the nuclear quantum effect is probably weak in the present case. Future studies should be able to elucidate the precise sites and mechanisms of action, as well as the reason why D 2 O activates TAS1R2/TAS1R3 in particular, resulting in sweet (but not other) taste. To this end, site-directed mutagenesis as well as determination of the precise structure of the TAS1R2/TAS1R3 receptor will be of key importance.
The finding that deuterated water elicits sweet taste via activation of TAS1R/TAS1R2 receptor is of fundamental interest. The difference between hydrogen isotopes is the largest possible isotope effect (doubling of mass in case of deuterium, while tripling in case of tritium), yet deuteration effects on water are generally mild. Nevertheless, water deuteration leads to activation of a GPCR heterodimer to a level that is perceived by humans as sweet taste. While clearly not a practical sweetener, heavy water provides a glimpse into the wide-open chemical space of sweet molecules. Since heavy water may be used in medical procedures, our finding that it can elicit responses of the sweet taste receptor, which is not only located on the tongue but also in other tissues of the human body, represents an important information for clinicians and their patients. Moreover, due to wide application of D 2 O in chemical structure determination by NMR, chemists will benefit from being aware of the present observations.

Methods
Sensory evaluation experiments. A human sensory panel was used to resolve the gustatory effect in perception of D 2 O taste. Subjects between the ages 20 and 43 years were recruited. The study included 10 experiments with different groups of participants (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) subjects; between 4 and 12 males). The perception was tested by sensory tests as detailed below. Either sterile syringes with solutions (0.3 ml) or identical cups with solutions (7 ml), were presented in randomized order, unless otherwise noted. Participants were required to taste each solution using either 'tip of the tongue' or 'sip and spit' procedures, rinse their mouth with water after each solution and to wait for 30 s before moving to the next taste sample. All research  43,53 and preliminary experiments in our lab, which showed that 0.9 mM decreases the sweetness of 100 mM sucrose. Sweeteners concentrations were selected to be in low intensity of sweetness. All solutions were prepared in the morning of the day of the experiment and were stored in individual plastic syringes (1 ml) for each participant.
Data were first analyzed employing ANOVA with participants as a random effect 54 . Thereafter, the Tukey Kramer test was used to compare mean sweetness between all samples 54 . Significance was set at p < 0.05, and preplanned comparison t-tests were used where relevant.
Details concerning the Two-Alternative Forced Choice test 55 were as follows: Participants were presented with two blind coded water samples of D 2 O and For the data analysis, the highest number of responses for one sample was compared to a statistical table 55 which states the minimum number of responses required for a significant difference.
Panelists were presented with two identical and one different water samples. All three samples were presented to the subjects at once, and the panelists were instructed to taste or smell the samples from left to right and identify the odd sample. Triangle tests were used to examine the difference in taste, as well as in smell, between H 2 O and D 2 O.
Heterologous expression. TAS1R2/TAS1R3 stimulated activation of the Gprotein-mediated pathway was measured applying the IP-One HTRF assay (Cisbio) based on the manufacturer's protocol. In brief, HEK 293T cells (ATCC) were grown to a confluency of~85-90% and transiently transfected with 6 μg/plate DNA (TAS1R2, TAS1R3, Gα16gust44) by applying LipofectamineTM 2000 (Invitrogen, USA, 30 μl/plate) transfection reagent, according to the manufacturer's protocol. The next day, cells were suspended with fresh Dulbecco's modified Eagle's medium (DMEM), containing 10% fetal bovine serum (FBS), 1% Lglutamine amino acid and 1% penicillin streptomycin (10% DMEM), seeded (0.5 ml cells per well) into 24-well culture plate, and maintained for 8-12 h at 37°C. Then cells were "starved" overnight by changing the medium to 0.1% DMEM (containing 0.1% FBS), in order to reduce the basal activity of the cells. Cells exposure was performed by addition of 0.5 ml tested compound (pH = 7.4) dissolved in 0.1% DMEM with 50 mM lithium chloride (LiCl) for 5 min directly into the wells. The presence of LiCl in this step is crucial because LiCl leads to IP1 accumulation 45 . At the end of exposure time, tastant solution was replaced with fresh medium (0.1% DMEM) containing 50 mM LiCl for another 55 min. Later, wells were washed with 100 μl cold phosphate-buffered saline (PBS) + Triton X-100, and kept at −80°C for a few hours, in order to dissolve the cell membrane. For the IP-One HTRF assay, cell lysate was mixed with the detection reagents (IP1-d2 conjugate and Anti-IP1 cryptate TB conjugate, each dissolved in lysis buffer), and added to each well in a 384-well plate for 60 min incubation at room temperature. Finally, the plate was read using Clariostar plate reader (BMG, Germany) equipped with 620 ± 10 nm and 670 ± 10 nm filters. IP1 levels were measured by calculating the 665 nm/620 nm emission ratio.
All responses are presented as the means ± SEM of IP1 accumulation (%). Dose-response curves were fitted by non-linear regression using the algorithms of PRISM 7 (GraphPad Software, San Diego, CA, USA). Column figures were analyzed using one-way ANOVA with a Dunnett's 54 . Each compound was tested in triplicate in three individual experiments in comparison to the reference (carbachol dissolved in H 2 O or basal levels) 45 .
In the case of D 2 O, in order to test its specific effect, we used a powder DMEM medium ( For the functional assays with the human sweet taste receptor, we used a cell line (HEK 293 FlpIn T-Rex), which constitutively expresses the sweet taste receptor subunit TAS1R2 as well as the chimeric G protein Gα15gi3, whereas the sweet taste receptor subunit TAS1R3 can be induced by tetracycline 43,44 . Gαi3 is one of the major species of Gα-subunits capable of coupling to taste receptors, along with Gα-gustducin, GαS, and Gαi2 56,57 and TAS1R2-TAS1R3-Gαi3 cell system was successfully used in a previous study 43 . The functional experiments were done as described before 44 . Briefly, the cells were grown in DMEM supplemented with 10% fetal bovine serum, 100 U Penicillin/ml, 0.1 mg/ml Streptomycin, 2 mM Lglutamine, at 37°C and 5%-CO 2 , 100% air humidity. The day before the experiment, cells were seeded to a density of 50-60% onto 96-well plates coated with 10 µg/ml poly-D-lysine and 0.5 µg/ml tetracycline was added. Next, cells were loaded with Fluo-4 AM in the presence of 2.5 mM probenecid for 1 h. After this, cells were washed twice with C1-buffer (130 mM NaCl, 5 mM KCl, 10 mM HEPES, 1 mM sodium pyruvate, and 2 mM CaCl 2 , pH 7.4) before placing them in a fluorometric imaging plate reader (FLIPR tetra , Molecular Devices) for measurements.
C1-buffer prepared with D 2 O was mixed with C1-buffer made with H 2 O to result in the following final D 2 O-concentrations (a further threefold dilution, which occurs upon application of 50 µL stimulus to 100 µL of C1-buffer in the 96-well plates is already included): 18.47, 5.84, 1.85, 0.584, 0.185, 0.058, 0.018, and 0.000 M. Fluorescence changes were monitored after automated application of stimuli. As specificity control C1-D 2 O including 0.9 mM lactisole, a selective inhibitor of the human sweet taste receptor 58 , was applied to identically treated cells. This concentration is the same as used in sensory experiments and close to the 1 mM used in calcium assays in previous work 43 . Experimental results from five biological replicates performed in quadruplicates were used to establish the dose-response relationship using the software SigmaPlot as before 44 . As the highest D 2 O-concentration resulted in fluorescence changes largely resistant to lactisole blocking, the 18.47 M concentration was excluded. Student's t-test was used to confirm that D 2 Oinduced fluorescence changes above baseline were significantly (p ≤ 0.01) different from lactisole-treated controls.
Animal experiments. All animal experiments followed the ethical guidelines for animal experiments and the Act of the Czech Republic Nr. 246/1992 and were approved by the Committee for Experiments with Laboratory Animals of the Czech Academy of Sciences. Three-month-old male C57BL/6J mice (n = 34) from Charles Rivers Laboratories (Sulzfeld, Germany) were housed at a temperature of 23°C with a daily cycle of 12 h light and dark (lights on at 6 am). The mice were placed in groups of two in cages with automatic drinking monitoring system (Developmental Workshops of Institute of Organic Chemistry and Biochemistry of the Czech Academy of Sciences, Prague, Czech Republic). They were given ad libitum water and a standard rodent chow diet (Ssniff Spezialdiäten GmbH, Soest, Germany). On the day of the experiment, during the dark phase of the cycle, freely fed mice were given weighed food pellets and two 30 ml glass bottles. The two bottles contained pure H 2 O and pure D 2 O (n = 12), or H 2 O and sucrose solution (43 mmol/l sucrose solution in H 2 O) (n = 10), see Table S1. Mice drinking H 2 O in both bottles served as a control group (n = 12). Drinking was monitored every 10 min for 16 h (starting from 6 pm) and food intake was determined at the end of the experiment ( Figure S5).
All responses are presented as the means ± SEM. Statistical analysis was performed using ANOVA with a Dunnett's test 54 for food intake. Two-way ANOVA with a Bonferroni's multiple comparisons test was used for analysis of average volume of liquid consumption. Analysis was performed using GraphPad Software, Inc., Prism 8 (GraphPad Software, San Diego, CA, USA). The differences between the control and treated groups were considered significant at p < 0.05.
Modeling and docking. All models were prepared with I-Tasser web server 51 . The templates that were used by I-Tasser for each of the TAS1R2 and TAS1R3 monomers were: 6N51, 5X2M, and 5K5S. To model the full heterodimer, the monomer models were aligned to a Class-C GPCR Cryo-EM structure (PDB: 6N51) and minimized with Schrödinger Maestro 2019-1. To illustrate the orthosteric binding site of sugars D-glucose was prepared (Schrödinger Maestro 2019-1, LigPrep) and docked with Glide XP, to the TAS1R2 VFT domain model that was based on 5X2M in a protocol that was validated in previous work 59 . The TAS1R3-binding site is based on a lactisole molecule docked to TAS1R3 TMD model (Schrödinger Maestro 2018-2, Glide SP, template PDB ID: 4OR2 and 4OO9). The figure was made using ChimeraX (version 0.93) 60 . Water molecules were predicted with a water mapping software (SZMAP 1.5.0.2: OpenEye 52 ) after a successful benchmark over conserved water templates (PDB IDs 5CGC and 5CGD), in which the software was able to identify overall 8 out of 10 crystal water molecules around the ligands of the templates.