All animal experiments were approved by the Animal Care Committee of Korea University College of Medicine (KOREA-2019-0132). All mice were maintained under standard animal housing conditions [12 h light-dark cycles with ad libitum access to normal chow diet (SAFE^® A03, France) and water]. The T1r2 knock-in allele was generated by Macrogen (Korea). The original founder was generated under C57BL6/N background. The donor construct sequences, including the homology arms and the Cre sequence used to achieve homology-mediated direct repair, are indicated in Supplementary Figure 1.

To generate T1r2-tdTomato mice, T1r2-Cre mice were bred to ROSA26-LSL-tdTomato mice (JAX007908). To obtain T1r2-tdTomato:Npy-hrGFP or T1r2-tdTomato:POMC-hrGFP mice, T1r2-tdTomato mice were bred to either of Npy-hrGFP (JAX006417) or POMC-hrGFP (JAX006421) mice, respectively. We confirmed the genotype of every mouse by PCR amplification of genomic DNA extracted from their tails (MyTaq Extract-PCR kit, BIO-21127, Bioline, United Kingdom). The primer sequences used for genotyping are indicated in Table 1.

Adult mice were acclimated individually in plastic cages with ad libitum access to food. Drinking water was supplied via two custom sipper tubes for at least 1 week. After acclimation, most of the mice did not show any position bias. A session was composed of two test days. Mice were allowed to choose between sipper tubes containing water and tastants for 2 days, and the tube positions were switched daily. Each session was separated from the next by a 5-day interval to prevent the results of later sessions from being affected by previous sessions. When identical tastants were being tested, they were presented in increasing concentrations. The volume of ingested liquid was measured and a preference index (P.I.) was calculated according to the following equation: P.I. = volume of ingested tastant solution/(volume of ingested pure water + volume of ingested tastant solution). The numbers of animals subjected to two bottle test are four and three for wild-type and T1r2 KO, respectively.

T1r2-tdTomato mice were serially perfused with 0.1 M phosphate buffered saline (PBS) and 4% paraformaldehyde (PFA) in PBS. Their brains were dissected from their skulls, post-fixed overnight, and cryoprotected in 30% sucrose in PBS at 4°C for several days until sinking. They were then embedded in Tissue-Tek OCT (Sakura, Japan). Cryoblocks were stored at −80°C until used. Brain samples were cut into coronal sections with a thickness of 30 μm. Every third free-floating section from the anterior olfactory bulb to the caudal cerebellum was collected. Tongue samples were cut into coronal sections with a thickness of 12 μm and directly attached to slide glass. The sections were then blocked in 5% goat or donkey serum in 0.2% Triton X-100 PBS (PBST) for 30 mins at room temperature. Primary antibodies were diluted in the corresponding blocking buffer and incubated overnight. After washing three times with PBST for 10 mins each, they were incubated in secondary antibodies dissolved in PBST for 2 h at room temperature. DAPI stain (1:5,000; D9542, Sigma) was added after three washes and the sections were mounted with Vectashield (Vector Laboratories, Burlingame, CA, United States) and a cover glass. All the primary and secondary antibodies used are listed in Table 2. Images were acquired with either an LSM 700 or LSM800 confocal microscope (Zeiss, Germany). Total 14 male and 2 female mice were subjected to immunostaining, and we could not find any sexual differences on expression pattern.

Every third brain slice covering the ARC was collected and subjected to a double IHC staining protocol. Co-localization analysis was performed with Imaris (Bitplane, Zurich, Switzerland). To determine whether hrGFP fluorescent signals showed co-localization with tdTomato fluorescent signals, the merged images were analyzed with the Imaris co-localization module. Spots were assigned to specific labels based on peak mean intensities. A co-localization threshold of 0.6 μm was set as the pixel boundary occupied by a single parenchymal cell.

Total 2 mL of FITC-Dextran 70 kDa (5 mg/mL, Sigma, 46945) was injected into the left ventricle of deeply anesthetized T1r2-tdTomato mice at 2 mL/min. Brain was harvested immediately after perfusion, post-fixed in 4% PFA for 24 h, and immersed in 30% sucrose in PBS overnight. Following procedures were conducted similarly with conventional IHC methods.

All data are expressed as means ± S.E.M. Statistical differences among groups were analyzed using two-sample t-tests. Asterisks indicate ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.005.

To genetically label T1r2-expressing cells, we generated T1r2-Cre knock-in mice using the CRISPR/Cas9 system. By substituting the coding region of the first exon of T1r2 with Cre sequence using homology-directed repair, we were able to generate mice that express Cre recombinase under the control of the endogenous promoter and enhancers of the T1r2 gene (Figure 1A). We confirmed the precise genomic exchange for the target region by genomic PCR (Figure 1B). We designed this mutant allele for use, not only as a Cre driver for the T1r2 gene in the heterozygous state, but also as a T1r2 knock-out (KO) in the homozygous state.

To confirm its function, we first performed immunohistochemistry (IHC) staining of the tongue. After crossing T1r2-Cre mice to ROSA26-LSL-tdTomato mice, we visualized Cre-labeled cells in their progeny (hereafter, T1r2-tdTomato). In the tongue, we found tdTomato fluorescence restricted to taste bud-containing papillae (Figure 1C). In circumvallate papillae (CVP), T1r2-expressing cells belong to the subset of TRPM5 and PLCβ2-expressing cells. This indicates T1r2-Cre labels type II taste receptor cells (TRCs). The fluorescence we observed, however, also showed only a partial overlap with GNAT3 (Figure 1C). Given that GNAT3 is expressed preferentially in bitter TRCs in CVP (Kim et al., 2003; Kusakabe et al., 2005; Sainz et al., 2007; Shindo et al., 2008; Tizzano et al., 2008), our data suggest T1r2-Cre accurately recapitulates the expression of native T1r2 in sweet TRCs. Next, by performing two-bottle behavioral assays, we found homozygous mice showed impaired choice behaviors toward sucrose but not denatonium or NaCl (Figure 1D). This is consistent with previous studies that used other mutant T1r2 alleles (Zhao et al., 2003). Thus, our data justify the use of this novel strain in further studies.

We inspected the brains of T1r2-tdTomato mice from the olfactory bulb to the brain stem. The overall distribution and intensity of tdTomato fluorescence are indicated in Figure 2 and Table 3, respectively. Every brain section was subjected to double IHC against tdTomato and NeuN to recognize the overall structure. Most tdTomato-expression overlapped with anti-NeuN staining, especially in the olfactory bulb (OB), cerebral cortex, thalamus, nucleus accumbens (NAc), bed nucleus of the stria terminalis (BNST), amygdala, and septal nuclei. This overlap also included several hypothalamic nuclei, including the medial preoptic area (MPA), paraventricular nucleus (PVN), suprachiasmatic nucleus (SCN), ARC, lateral hypothalamic nucleus (LH), and dorsomedial hypothalamic nucleus (DMH) (Figures 2–7 and Table 3).

Although we did observe tdTomato fluorescence scattered throughout the brain, there were significant concentrations in specific regions. The most prominent signals were located near the ventricular system (Figure 3 and Table 3), with intense staining in the circumventricular organs (CVOs)—the subfornical organ (SFO), the median eminence (ME), the area postrema (AP), the organum vasculosum lamina terminalis (OVLT), and the subcommissural organ (SCO) (Figure 3). There was also a morphologically heterogenous population of tdTomato-positive cells in the mediobasal hypothalamus—not only parenchymal cells with a neuron-like morphology, but also cells lining the internal wall of the third ventricle and the external wall of the ME (Figures 3, 4). The tdTomato-positive cells in the SCN and MPA, in contrast, were morphologically homogenous (Figure 4). In addition, while the dorsal hypothalamus, PVN, DMH, and LH showed significant tdTomato fluorescence, other important hypothalamic nuclei, including the anterior hypothalamic nucleus and ventromedial hypothalamic nucleus did not (Figure 4).

We also observed significant tdTomato expression in the hindbrain (Figure 5). The dorsal tegmental nucleus (DTN), median raphe nucleus (MRN), and interpeduncular nucleus (IPN) each showed a bright and densely fibrous arbor of staining arising from a relatively limited number of tdTomato-expressing cells (Figures 5A–C). In addition, we observed some tdTomato labeling in the parabrachial nucleus (PBN), the nucleus of the lateral lemniscus (NLLS), the pontine central grey (PCG), the dorsal motor nucleus of the vagus nerve (DMX), the nucleus of the solitary tract (NST), and the reticular nucleus (RN) (Figures 5D–I). In the midbrain, we observed labeling in a subset of neurons in the superior (SC) and inferior colliculi (IC), as well as in the ventrolateral PAG and DRNs (Figures 5J–L). In the cerebellum, we observed mutually exclusive expression of either tdTomato or NeuN, indicating expression of T1R2 in Purkinje cells rather than granular cells (Figure 5M).

In the OB, the tdTomato and anti-NeuN signals showed overlap outside the olfactory glomeruli and in the mitral and granular layers, indicating that periglomerular cells, mitral cells, and granular cells express T1r2 (Figure 6A). When we examined the cortex, we found most cortical regions showed some scattered fluorescence. Although some tdTomato-expressing cells showed overlap with anti-NeuN staining, most did not (Figures 6B,C). In the hippocampus, neurons in CA1, CA2, CA3, and dentate gyrus (DG) were tdTomato-positive regardless of their rostrocaudal position (Figure 6E). In the anterior thalamus, the tdTomato signal was concentrated in the anterodorsal (AD), anteroventral (AV), and anteromedial (AM) nuclei, as well as in the lateral habenula (Figures 6F,G). In contrast, we observed significant staining scattered throughout the entire posterior thalamus (Figures 6H–J).

In the mesolimbic system, the lateral septum (LS), NAc, caudate putamen (CPu), BNST, central (CEA) and median amygdala (MEA), and septal nuclei expressed tdTomato in some neurons (Figure 7), but the ventral tegmental area (VTA) did not.

We found that the many T1r2-expressing cells in the cortex, thalamus, and striatum are negative for anti-NeuN. Moreover, we noticed that most of the non-neuronal T1r2-expressing cells have a vascular-like luminal structure (Figures 6B,C,F,H,I, 7A,B, white empty triangles). These presumed blood vessels appear to be of various sizes, implying a general T1r2 expression in the cerebrovascular system. Thick vessels run straight, finally branching into several thin ones (Figures 6B,C, 7A,B, white empty triangles). The thinnest vessels comprised single layers of cells, interconnected with one another to form a vascular web (Figures 6B,C,F,H,I, 7A,B, white empty triangles). To confirm whether the tubular structures formed by T1r2-expressing cells are blood vessels, we visualized blood vessels directly by transcardial injection of FITC-dextran (70 kDa) and found the spatial proximity between FITC signals and T1r2-expressing cells in cortex and CPu (Figures 6D, 7C). We also observed pericytic tdTomato expression distributed homogenously throughout the cortex, posterior thalamus, and dorsal striatum, as well as more prominent expression in the anterodorsal (AD), anteroventral (AV), and anteromedial (AM) nuclei of the anterolateral thalamus (Figures 6, 7). These data indicate an enrichment of T1r2 in the neurovascular system.

Across the entire brain, CVOs showed such strong tdTomato fluorescence that we could detect it without immunostaining. When we performed double IHC with anti-NeuN, however, we found that not all tdTomato-expressing cells in the CVOs were neurons (Figure 3). To identify their cellular identities, we next conducted double IHC with the following glial cell markers: Glial Fibrillary Acidic Protein (GFAP) for astrocytes; Ionized Calcium-Binding Adaptor molecule (IBA1) for microglia; and Myelin Basic Protein (MBP) for oligodendrocytes.

In the dorsal periphery of the SFO, a subset of tdTomato-expressing cells were positive for GFAP (Figure 3A). Due to the restricted subcellular localization of GFAP protein in astrocytic processes rather than cell bodies, we found that the cytosolic tdTomato signal showed only partial overlap with anti-GFAP signal in each labeled cell. There was, however, no overlap with anti-IBA1 or anti-MBP (Figure 3A). Similarly, the ventrolateral margin of the AP showed co-expression of tdTomato and anti-GFAP (Figure 3C), indicating T1r2 is expressed in astrocytes in the SFO and AP.

The labeled cells in the ARC-ME had heterogenous morphology owing to a heterogeneity of their cellular identities. In the ARC-ME complex parenchyma, tdTomato-expressing cells showed a neuron-like morphology and expressed NeuN (Figures 3B, 4A). The cells surrounding the wall of the third ventricle, however, expressed neither neuronal nor glial markers (Figure 3B). Instead, we were able to identify these cells as tanycytes by their long processes projecting radially from the ventricular wall to the parenchyma. We also observed strong fluorescence signal in the perivascular cells covering the ventral margin of the ME (Figure 3B). In addition, we observed tdTomato expression in the cells of the OVLT at the anterior end of the third ventricle and in the SCO at the posterior end of the third ventricle (Figures 3D,E).

The primary centers for central metabolic regulation in the ARC-ME complex are bimodal (Sohn et al., 2013). Depletion of energy is detected by NPY/Agouti-related peptide (AgRP)-expressing neurons (Betley et al., 2015; Beutler et al., 2017; Su et al., 2017), whereas storage of energy is detected by POMC-expressing neurons (Beutler et al., 2017). To determine whether the neurons expressing T1r2 are NPY/AgRP neurons, POMC neurons, or both, we generated triple transgenic mice expressing T1r2-Cre and Rosa-LSL-tdTomato, along with either Npy-hrGFP or POMC-hrGFP. Most of the parenchymal tdTomato signal we observed showed overlap with NPY-expressing neurons (Figure 8A). In contrast, only a small number of tdTomato-labeled neurons were also positive for POMC (Figure 8B). Together, we observed tdTomato fluorescence in 75% (1,155 of 1,550 cells) of Npy positive neurons and 34% (181 of 540 cells) of POMC positive neurons (Figure 8C). These data suggest T1r2 is expressed both in hypothalamic NPY/AgRP and POMC neurons.

In taste buds, type II TRCs employ proteins like GNAT3, PLCβ2, and TRPM5 downstream of taste receptor signaling (McLaughlin et al., 1992; Wong et al., 1996; Perez et al., 2002; Liu and Liman, 2003; Zhang et al., 2003; Mueller et al., 2005). We wondered whether such canonical taste signaling molecules are also co-expressed with T1R2 in the brain. Thus, we again conducted double IHC experiments to visualize PLCβ2, GNAT3, or TRPM5 expression alongside tdTomato expression.

Remarkably, we found complete co-localization of PLCβ2 with tdTomato in hypothalamic perivascular cells of the leptomeningeal layer, but not in other areas (Figure 9A). Although anti-GNAT3 marked the median eminence, it did not co-localize with tdTomato (Figure 9A). We also found that none of the brain regions that expressed tdTomato were stained with the antibody against TRPM5 (Figure 9A). Given that the specificity of the antibodies we used was already validated in taste buds (Figure 1C), the relative lack of GNAT3, PLCβ2, and TRPM5 in the central nervous system indicates a true scarcity.