A Transcription Factor Etv1/Er81 Is Involved in the Differentiation of Sweet, Umami, and Sodium Taste Cells

Abstract Taste cells are maintained by continuous turnover throughout a lifetime, yet the mechanisms of taste cell differentiation, and how taste sensations remain constant despite this continuous turnover, remain poorly understood. Here, we report that a transcription factor Etv1 (also known as Er81) is involved in the differentiation of taste cells responsible for the preference for sweet, umami, and salty tastes. Molecular analyses revealed that Etv1 is expressed by a subset of taste cells that depend on Skn-1a (also known as Pou2f3) for their generation and express T1R genes (responsible for sweet and umami tastes) or Scnn1a (responsible for amiloride-sensitive salty taste). Etv1CreERT2/CreERT2 mice express Etv1 isoform(s) but not Etv1 in putative proprioceptive neurons as comparable to wild-type mice, yet lack expression of Etv1 or an isoform in taste cells. These Etv1CreERT2/CreERT2 mice have the same population of Skn-1a-dependent cells in taste buds as wild-type mice but have altered gene expression in taste cells, with regional differences. They have markedly decreased electrophysiological responses of chorda tympani nerves to sweet and umami tastes and to amiloride-sensitive salty taste evoked by sodium cation, but they have unchanged responses to bitter or sour tastes. Our data thus show that Etv1 is involved in the differentiation of the taste cells responsible for sweet, umami, and salty taste preferences.


Introduction
The sense of taste plays a key role in food choice. Individual taste cells of vertebrate animals primarily respond to only one taste (Yarmolinsky et al., 2009;Matsumoto et al., 2013); thus, the diversity of tastes that an animal can perceive depends on the diversity of taste cells present. Mice can distinguish five basic tastes: sweet by cells co-expressing Tas1r2 and Tas1r3 or expressing Tas1r3 alone, umami (i.e., savory) by cells that co-express Tas1r1 and Tas1r3, bitter by Tas2r-expressing cells, sour by Pkd2l1-expressing cells, and salty by cells that express the amiloride-sensitive Na 1 channel ENaC (Nelson et al., 2001(Nelson et al., , 2002Zhao et al., 2003;Mueller et al., 2005;Huang et al., 2006;Chandrashekar et al., 2010;Ohmoto et al., 2020a). Taste cells can be distinguished from one another by their molecular features, especially by the expression of specific taste receptors, although no unique molecular markers specific for amiloride-sensitive salty taste cells (hereafter referred as sodium-taste cells) have been found, because ENaC subunit genes are also expressed in other taste cells (Chandrashekar et al., 2010;Nomura et al., 2020).
Taste cells emerge from local epithelial stem cells that express Krt5 and Sox2, and their diverse populations are maintained by continuous turnover throughout an animal's life (Beidler and Smallman, 1965;Farbman, 1980;Stone et al., 1995;Ohmoto et al., 2017Ohmoto et al., , 2020b. The transcription factor Skn-1a (also known as Pou2f3) specifies the fate of a cell as a sweet, umami, bitter, or sodium taste cell lineage, rather than a sour lineage, probably at the late precursor stage (Matsumoto et al., 2011;Ohmoto et al., 2020a). However, the molecular and cellular mechanisms underlying how these cells then differentiate into different functional taste cells of the specified lineage remain elusive.
Etv1 (also known as Er81), a transcription factor of the ETS family, is one such candidate for this process. In the nervous system it is involved in the terminal differentiation of peripheral and central neurons, such as proprioceptive sensory neurons and cerebellar granule neurons (Arber et al., 2000;Patel et al., 2003;Abe et al., 2011;Fleming et al., 2016;Okazawa et al., 2016). Etv1 is also required for the survival of proprioceptive sensory neurons with various degrees of dependence among proprioceptive sensory neurons (Patel et al., 2003) but not for the survival of granule cells in cerebellum or their formation of synaptic connections .
Etv1 is significantly enriched in taste buds of the rhesus macaque, and its expression is as high as that of tastebud-specific transcription factor genes in tongue epithelium (Hevezi et al., 2009). Recent RNA sequencing data of taste cells support the presence of Etv1 in Tas1r3 1 sweet taste cells but not in bitter or sour taste cells (Lee et al., 2017;Sukumaran et al., 2017). However, it remains unclear whether sodium-taste cells or nonsensory taste bud cells express Etv1 or whether all Tas1r3 1 taste cells express it.
In the present study, we demonstrate that Etv1 is involved in differentiation of Skn-1a-dependent taste cells (i.e., often referred to as Type II cells): Etv1 is expressed in sweet, umami, and sodium taste cells. Genetic ablation of Etv1 markedly decreases the expression of T1R genes in taste buds and culminates in the loss or reduction of responses to sweet and umami tastes in gustatory nerve recordings in mice. Etv1 ablation also represses Scnn1a expression and amiloride-sensitive salt taste responses in gustatory nerves. Sweet, umami, and sodium taste cells are still present in Etv1-deficient taste buds, but their populations are smaller than in heterozygous control mice.

Tamoxifen administration
Tamoxifen (10 mg/ml in corn oil; Sigma-Aldrich) was injected intraperitoneally in mice (100 mg/kg body weight) once a day for five consecutive days. The mice were euthanized 1 d after the last tamoxifen injection.

Tissue preparation
Oral epithelia were dissected from mice deeply anesthetized by urethane, and sensory ganglia were dissected from mice that had been transcardially perfused with icecold PBS under anesthesia. Freshly dissected tissues were quickly embedded and frozen in optimal cutting temperature (O.C.T.) compound (Sakura Finetech). For tissues fixed with 4% paraformaldehyde (PFA), mice were deeply anesthetized with urethane and transcardially perfused with PBS followed by 4% PFA in PBS. For the developing neurons in the dorsal root ganglia (DRG), embryos were harvested from pregnant mice and fixed with 4% PFA, and lumber spinal columns including spinal cord and DRG were dissected. Dissected oral epithelia and embryonic spinal columns were postfixed, cryoprotected, and frozen as previously described (Ohmoto et al., 2008). Cryosections (8-mm thickness) were prepared using a CM1900 cryostat (Leica Microsystems), mounted on tissue adhesive-coated glass slides (Fisher Scientific), and preserved at À80°C until use.

In situ hybridization
In situ hybridization using fresh-frozen or PFA-fixed sections was conducted as previously described (Ohmoto et al., 2008(Ohmoto et al., , 2020a. Digoxigenin-labeled and fluorescein-labeled antisense RNAs were synthesized and used as probes after fragmentation to ;150 bases under alkaline conditions. The probe regions are shown in Table 1. For the Tas2r gene probes, riboprobes of Tas2r104, Tas2r105, Tas2r118, and Tas2r126 were mixed. Fresh-frozen sections were fixed with 4% PFA and treated with diethylpyrocarbonate. The PFA-fixed sections were treated with proteinase K (3 mg/ml, Thermo Fisher Scientific), postfixed with 4% PFA, and then acetylated with acetic anhydride. The sections were prehybridized with salmon testis DNA, hybridized with the riboprobes for 40 h, and washed in 0.2Â saline sodium citrate at 58°C, except when using the riboprobes for Calhm1 and Scnn1a, for which sections were hybridized and washed at 65°C. After washing, chromogenic and/or fluorescent signals were developed as follows.
For single-label in situ hybridization, the chromogenic signals of the hybridized probes were detected using alkaline phosphatase-conjugated anti-digoxigenin antibodies (Roche Diagnostics, 11093274910, RRID:AB_514497) followed by 4-nitro blue tetrazolium chloride/5-bromo-4chloro-3-indolyl phosphate as a substrate. The signals were developed overnight except for Calhm1 probes, which were developed for 2 d. Short-term signal development for Tas1r3 was also investigated at 4 h. Stained images were obtained using a Nikon Eclipse 80i microscope equipped with a DXM1200C digital camera (Nikon Instruments).
For double-fluorescence in situ hybridization in oral epithelia, the fluorescent signals of the riboprobes were developed using alkaline phosphatase-conjugated antidigoxigenin antibodies followed by the HNPP Fluorescent Detection Set (Roche Diagnostics) and biotin-conjugated anti-fluorescein antibodies (Vector Laboratories, BA-0601, RRID:AB_2336069) followed by avidin-biotin complex (Vector Laboratories), TSA Biotin Tyramide Reagent (PerkinElmer), and Alexa488-conjugated streptavidin (Thermo Fisher Scientific). Fluorescent images were acquired with a TCS SP2 confocal microscope (Leica Microsystems). Brightness and contrast were linearly adjusted, and all adjustments were globally applied to confocal images and overlaid multiple fluorescence images using Photoshop (Adobe Systems). The presence of fluorescence signals was determined manually by visual inspection.
For double-fluorescence in situ hybridization in geniculate ganglia (GG), fluorescein-labeled probes and digoxigenin-labeled probes were sequentially detected as follows. Sections were incubated with biotin-conjugated anti-fluorescein antibodies, followed by avidin-biotin complex, TSA-plus Biotin solution (PerkinElmer), and Alexa488conjugated streptavidin. After treatment with 3% H 2 O 2 in PBS for 60 min, sections were then incubated with horseradish peroxidase-conjugated anti-digoxigenin antibodies (Roche, 11207733910, RRID:AB_514500) followed by TSA-plus Cyanine 3 solution (PerkinElmer) and 4',6-diamidino-2-phenylindole (DAPI). The fluorescent images were acquired, and optical confocal images were processed as described above. Only neurons whose profiles contain a nucleus with and without fluorescence signals were counted, and the frequencies of expression of Etv1-derived transcripts in P2x2 1 neurons and those of P2x2 1 neurons in GG were quantitatively analyzed using every 10 sections (i.e., 80 mm apart) of the GG of wildtype and Etv1 C/C mice (n = 3 for each genotype).
Double labeling of Scnn1a with other genes with a combination of chromogenic and fluorescent signals was conducted as previously described (Taruno et al., 2013;Ohmoto et al., 2020a). The fluorescent signals were first developed using a biotin-conjugated anti-fluorescein antibody (Vector Laboratories) followed by avidin-biotin complex (Vector Laboratories), TSA Biotin Tyramide Reagent (PerkinElmer), and Alexa488-conjugated streptavidin (Thermo Fisher Scientific). After acquiring the fluorescent images, the chromogenic signals for Scnn1a were detected using an alkaline phosphatase-conjugated anti-digoxigenin antibody and 4-nitroblue tetrazolium chloride/5-bromo-4chloro-3-indolyl-phosphate. The signals were developed for 7 d. Stained images were obtained as described above. Fluorescent and stained images were processed with Photoshop (Adobe Systems) and analyzed on a computer screen. Fluorescence signal and chromogenic stain surrounding a nucleus were counted as positive signal. Sporadic punctate fluorescence was regarded as nonspecific background. The frequencies of expression of Scnn1a in Skn-1a 1 cells (i.e., sodium taste cells) were semiquantitatively analyzed by calculating the ratios of taste bud profiles that harbored at least one cell co-expressing Skn-1a and Scnn1a to all taste bud profiles using every 8 and 10 sections of palate and fungiform papillae (FuP), respectively, of wild-type and Etv1 C/C mice (n = 3 for each genotype). Scnn1a 1 cells devoid of Skn-1a signal, which are likely sour taste cells (Chandrashekar et al., 2010;Ohmoto et al., 2020a), were not counted.
Quantitative PCR Epithelia of CvP, anterior tongue, and soft palate, and DRG and GG were dissected from C57BL/6J mice and Etv1 C/C mice. Total RNA was extracted using TRIzol (Life Technologies, 15596026), and cDNA was synthesized using SuperScript IV VILO Master Mix with ezDNase Enzyme (Life Technologies, 11766050). Quantitative PCR (qPCR) was performed using Fast SYBR Green Master Mix kit (Applied Biosystems,4385612). Genes that did not show single peaks of melt curve in a C57BL/6J tissue were regarded as "not amplified" and removed from further analyses. Gapdh was used as a control to normalize the expression levels of individual genes. The relative gene expression level was calculated as 2 -(CTTarget-CTGapdh) (Livak and Schmittgen, 2001). Details about the primers used are given in Table 2.

5'-RACE
First strand cDNA to Etv1 gene-derived transcript in the total RNA extracted from DRG of C57BL/6J mice was synthesized after the treatment with ezDNase Enzyme (Life Technologies, 11766050) using a primer 59-TTCCATG TCCGTCTTCAGCAG-39, which is also used to amplify C-terminal coding region of Etv1 by qPCR, and then polyadenylated by terminal deoxynucleotidyl transferase. A 59 end of Etv1 gene-derived transcript was amplified by PCR using a primer 7R1 (59-CGGTCAGGTTTCGGAGTATGA GC-39) and oligo(dT) primer and then a primer 7R2 (59-ATAGTCGACCACTTGTGGCTTCTGATCATAGGC-39) and oligo(dT) primer. Sequence of the amplified cDNA fragment was determined using the primer 7R2.
Fluorescent labeling of KCNQ1 and double-fluorescence labeling of IP 3 R3 and Ddc (also known as AADC) were conducted using every 12, 6, and 6, sections of CvP, FuP, and palate, respectively, of Etv1 C/1 ; R26 Tom/1 (heterozygous control) and Etv1 C/C1 ; R26 Tom/1 mice (n = 3 for each genotype). The sections were counterstained with DAPI, and fluorescent images were acquired with a confocal microscope. KCNQ1 1 cells, Ddc 1 cells, IP 3 R3 1 cells, and IP 3 R3 1 tdTomato 1 cells that had DAPI signals in taste bud profiles were counted, and the ratios of IP 3 R3 1 tdTomato 1 cells to IP 3 R3 1 cells were calculated. Double-fluorescence labeling of P2X2 and KCNQ1 was conducted in gustatory tissues of wild-type and Etv1 C/C mice (n = 2 for each genotype). The sections were counterstained with DAPI, and fluorescent images were acquired with a confocal microscope in a Stellaris 5 WLL AOBS system (Leica Microsystems). Serial zstacks were merged as maximum projection images using LASX software. and Skn-1a À/À (right) mice. n = 3 for CvP, n = 1 for palate and FuP; n ! 9 sections for each experiment. Scale bars: CvP, 50 mm; FuP and palate, 25 mm. B, Double-fluorescence in situ hybridization of Etv1 (magenta) with marker genes (green) in the CvP (top), FuP (middle), and palate (bottom) taste buds in WT mice. White arrowheads indicate cells singly expressing Etv1. For each experiment: CvP, n ! 2, with !7 sections; FuP, n ! 2, with !18 sections; palate, n ! 1, with !9 sections. Scale bars: 25 mm. C, Double-labeling in situ hybridization to study expression of Scnn1a in Etv1 1 taste cells: Scnn1a (dark purple; left), Etv1 (green; middle), and merged (right) images of FuP (top) and palate (bottom) taste buds of WT mice. Yellow arrowheads indicate cells co-expressing Scnn1a and Etv1; red arrowheads indicate cells expressing only Scnn1a. n = 3 for each experiment. Scale bars: 25 mm. Taste buds and basal epithelial-mesenchymal boundaries are indicated by broken lines.

Whole chorda tympani nerve recordings
We investigated the electrophysiological response of the chorda tympani nerve in mice of the wild-type and Etv1 C/C mice at two to four months old of both sexes (n = 8 per genotype). The experimenters were blinded to the genotype of the mice during testing. The mice were anesthetized with an intraperitoneal injection of a mixture of 4.28 mg/ml ketamine, 0.86 mg/ml xylazine, and 0.14 mg/ml acepromazine in saline (5 ml/g body weight). Anesthesia was maintained with additional injections. Each mouse was fixed with a head holder after its trachea was cannulated; then the chorda tympani nerve was dissected free from the lingual nerve near the tympanic bulla, cut, and placed on a platinum wire recording electrode. An indifferent electrode touched the walls of the wound. Taste stimuli were delivered to the tongue with a computer-controlled open flow system under constant flow and temperature (25°C) conditions. Each stimulation lasted for 30 s with a 60-s rinse between stimulations. Care was taken to ensure the flow was directed over the FuP. The nerve impulses were fed into a custom-made amplifier, monitored over a loudspeaker and with an oscilloscope, and recorded (PowerLab/sp4; AD Instruments). The integrated response during stimulation was calculated by subtracting the area of nerve activity preceding the stimulation from that during stimulation. Thus, the data reflect the level of activity during the stimulation period. The responses to all compounds were expressed relative to the response to 0.1 M NH 4 Cl for each mouse. The averages for each animal and group were calculated for the statistical analyses.

Statistical analyses
Data are shown as the mean 6 SD except those of gustatory nerve recordings that are shown as the mean 6 SEM. A Welch's t test (for qPCR, taste bud numbers, taste bud cell populations, and tdTomato 1 cell ratio to IP 3 R3 1 taste cell) or repeated-measures two-way ANOVA (for gustatory nerve recordings) was used to determine the effects of genotype using Prism 6 and 8 software (GraphPad Software). Sidak's multiple comparison tests were conducted to identify significant differences between pairs of mean values.

Expression of Etv1 in taste cells
We examined the expression of Etv1 mRNA in mouse taste buds by in situ hybridization (for probe lengths and regions, see Table 1) and observed signals in a minor subset of taste bud cells in all gustatory areas of wild-type mice but not in Skn-1a-deficient mice, which indicates the restricted expression of Etv1 mRNA in Skn-1a-dependent taste cells (Fig. 1A). Double-fluorescence in situ hybridization analyses showed that Etv1 is expressed in a subset of Skn-1a 1 and Trpm5 1 (i.e., sweet, umami, and bitter taste) cells in the CvP in the wild-type mice, especially in Tas1r3 1 cells, but not in Tas2r 1 cells responsible for bitter taste (Fig. 1B). In taste buds in the FuP and palate, Etv1 was present in the cells lacking Tas1r3, Tas2r, and Trpm5 expression in addition to Tas1r3 1 cells, suggesting its expression in sodium-taste cells. Double-label in situ hybridization of Etv1 with Scnn1a showed partial co-expression in taste cells in the FuP and palate (Fig. 1C). Together with its restricted expression to Skn-1a-dependent taste cells, these results indicate that Etv1 is expressed in Skn-1a 1 Scnn1a 1 sodium-taste cells (Ohmoto et al., 2020a). Therefore, gene expression analyses demonstrate that Etv1 is expressed in sweet, umami, and sodium taste cells, all of which are involved in eliciting taste preferences.
Etv1 deficiency in taste buds of Etv1 C/C mice Mutant mouse strains with a null or loss-of-function mutation for Etv1 are useful models to study the function of Etv1 in the taste system. Etv1 C/C mice are designed to encode a tamoxifen-inducible Cre recombinase (CreER) instead of Etv1 and carry a CreERT2-frt sequence in exon 2 in place of the initial ATG of the Etv1 coding sequence (Extended Data Fig. 2-1A; Taniguchi et al., 2011). Therefore, they were expected to be conventional knock-outs like Etv1 nlslacZ mice (Extended Data Fig. 2-1A; Arber et al., 2000). Etv1 C/C mice, however, are viable and fertile and do not present any overt abnormalities, although other Etv1null mutant strains (Etv1 ETS/ETS and Etv1 nlslacZ/nlslacZ mice; for their allele structures, see Extended Data Fig. 2-1A) exhibit limb ataxia or abnormal postures attributable to defects in proprioceptive function and die three to five weeks after birth (Arber et al., 2000;Patel et al., 2003). Therefore, we first continued FuP and palate; !9 sections for each experiment. Scale bars: CvP, 50 mm; FuP and palate, 25 mm. B, Expression of mRNA (top) and protein (middle and bottom) of Etv1 isoform(s) in the sensory neurons in the dorsal root ganglia of wild-type (WT; left) and Etv1 C/C (right) embryos at embryonic day 15.5. Etv1-immunoreactive signals (magenta) were observed in the parvalbumin (PV, green)-expressing proprioceptive neurons (bottom). n = 3 for each experiment, 10 sections for each genotype. Scale bars: 100 mm (chromogenic images) and 50 mm (fluorescent images). C, Schematic structure of Etv1 gene and its CreERT2 knock-in allele. Boxes indicate exons. Protein coding regions are shaded. Primers used for PCR analyses are indicated by triangles (see Table 2). The reverse primer of the primer set C and primers 7R1 and 7R2 were used for 5'-RACE. CreERT2 sequence is inserted in front of the forward primer of the primer set N in the second exon of Etv1 (Extended Data Fig. 2-1). D, Gel image of agarose electrophoresis of qualitative PCR products from circumvallate papillae (CvP) and dorsal root ganglia (DRG) of wild-type (WT) and Etv1 C/C mice. Predicted amplicon sizes by primer sets N and C are 172 and 198 bp, spanning from exon 2 to exon 5 and from exon 12 to exon 13, respectively. Lane M, 100-bp DNA ladder marker. E, F, qPCR analyses to detect the expression of Etv1 and/or Etv1-derived transcripts in DRG (E) and CvP (F) in WT and Etv1 C/C mice. Open circles, individual data; white and red bars are the mean 6 SD of WT and Etv1 C/C mouse tissues, respectively. Significance was assessed by Welch's t test. Primers used are sets N and C to DRG (E) and set C to CvP (F). n = 5 for WT mice; n = 4 for DRG and n = 5 for CvP for Etv1 C/C mice.  Fig. 3-1) of wild-type (WT) and Etv1 C/C mice. A, C, In situ hybridization analyses. Left, WT mice (n = 3; !9 sections for CvP, !27 sections for FuP). Right, Etv1 C/C mice (n = 3; !9 sections for CvP, !27 sections for FuP). Scale bars: CvP, 50 mm and FuP, 25 mm. B, D, qPCR analyses. n = 5 for each genotype. Open circles, individual data; white and red bars are the mean 6 SD of WT and Etv1 C/C tissues, respectively. Significance was assessed by Welch's t test. evaluated whether Etv1 is knocked out in Etv1 C/C mice. Etv1 mRNA was completely absent in taste buds in all gustatory areas examined in Etv1 C/C mice, while taste buds of wild-type mice showed Etv1 mRNA signals ( Fig. 2A). Intriguingly, however, we observed the signals of Etv1 mRNA and protein in a subset of dorsal root ganglia (DRG) neurons of wild-type and Etv1 C/C mice (Fig. 2B); the frequency and intensity of these signals in Etv1 C/C mice were qualitatively comparable to those in wild-type mice. Etv1 immunoreactivity was observed in most parvalbumin 1 neurons (i.e., proprioceptive sensory neurons) in both wild-type and Etv1 C/C mice (Fig. 2B). These results indicate that the Etv1 C/C mice are Etv1 null mutant in taste buds and suggest that Etv1 variant(s) are expressed in DRG neurons.
To determine whether Etv1 is knocked out and whether an Etv1 variant is expressed in DRG of Etv1 C/C mice, we performed qualitative PCR and qPCR using primers corresponding to N-terminal and C-terminal coding sequences. cDNA fragments corresponding to N-terminal coding sequence were not amplified from CvP or DRG of Etv1 C/C mice, indicating the disruption of Etv1 by the insertion of CreERT2 in DRG as well as CvP (Fig. 2C-F; Extended Data Fig. 2-1). However, cDNA fragments corresponding to the ETS domain in the C-terminal region were amplified from DRG but not from CvP of Etv1 C/C mice (Fig. Skn-1a or Pkd2l1 (green; middle), and overlaid (right) images of taste buds in fungiform papilla of wild-type (WT) and Etv1 C/C mice. Yellow arrowheads indicate cells co-expressing Scnn1a with Skn-1a (A) or Pkd2l1 (B), and red arrowheads indicate cells expressing only Scnn1a. n = 3; 27 sections for each genotype (A); n = 2; 18 sections (B). Scale bars: 25 mm. C, qPCR analysis of Scnn1a in taste buds in fungiform papillae of WT and Etv1 C/C mice. n = 5 for each genotype. Open circles, individual data; white and red bars are the mean 6 SD of WT and Etv1 C/C tissues, respectively. Significance was assessed by Welch's t test. Fig. 2-1A). 5'-RACE revealed that a major transcript in DRG is Etv1 variant 2 (Etv1-v2, NM_001163154), which has little impact of its expression by CreERT2 insertion (Fig. 2C; Extended Data Fig.  2-1B,C). These results indicate that the Etv1 C/C mice are Etv1 null mutant in taste buds and that the major transcript from the Etv1 allele in taste tissues is Etv1. In DRG, Etv1-v2 and its encoded protein, Etv1 isoform b, are expressed in DRG neurons, in addition to Etv1 mRNA and protein, in proprioceptive neurons. It remains unclear whether the Etv1 isoform b is as functional as Etv1, although it can seemingly compensate for the loss of Etv1 in the proprioceptive neurons.

Involvement of Etv1 in the expression of taste receptor genes in Etv1 1 cells in taste buds
To understand the role of Etv1 in taste cells, we studied the impact of Etv1 deficiency on taste receptor gene expression by in situ hybridization and qPCR analyses in taste buds of Etv1 C/C mice. The umami taste receptor gene Tas1r1, which is predominantly expressed in the taste buds in FuP and palate but rarely detected in CvP of Figure 5. Limited impact of Etv1 deficiency on expression of taste signaling genes necessary for T1R-mediated tastes. A, Expression of Gnat3, Plcb2, Trpm5, Calhm1, and Calhm3 that are partially co-expressed with Etv1 in circumvallate papillae of wildtype (WT; left) and Etv1 C/C (right) mice by in situ hybridization. n = 1 for each genotype, for each experiment; 3 sections. Scale bars: 50 mm. B, qPCR analyses in taste buds in circumvallate papillae of WT and Etv1 C/C mice. n = 5 for each genotype. Open circles, individual data; white and red bars are the mean 6 SD of WT and Etv1 C/C tissues, respectively. Significance was assessed by Welch's t test. For results in fungiform papillae and soft palate, see Extended Data Figures 5-1 and 5-2, respectively. wild-type mice, as reported previously (Hoon et al., 1999;Nelson et al., 2001), was completely absent from the taste buds of any gustatory area of Etv1 C/C mice (Fig. 3A,C; Extended Data Fig. 3-1A). The sweet taste receptor gene Tas1r2, which is predominantly expressed in the taste buds in CvP and palate but rarely detected in FuP of wild-type mice, as reported previously (Hoon et al., 1999;Nelson et al., 2001), was almost (in CvP) or completely (in FuP and palate) lost in taste buds of Etv1 C/C mice (Fig. 3A,C; Extended Data Fig. 3-1A). qPCR analyses exhibited results consistent with in situ hybridization analyses but found no significant differences where Tas1r1 and Tas1r2 are rarely expressed in wild-type mice (Fig. 3B,D; Extended Data Fig. 3-1B). The sweet and umami taste receptor gene Tas1r3 was less influenced by Etv1 deficiency than were Tas1r1 and Tas1r2: intensity and frequency of signals of Tas1r3 in taste buds were weaker and sparser, respectively, than in Figure 6. Unaltered expression of genes that are not co-expressed with Etv1 by Etv1 deficiency. Expression of T2R genes, Pkd2l1, and Entpd2 was examined in circumvallate papillae of wild-type (WT) and Etv1 C/C mice. A, C, E, In situ hybridization. T2R gene expression was examined using a mixture of probes to four T2R genes (Tas2r104, Tas2r105, Tas2r118, and Tas2r126). Left, WT. Right, Etv1 C/C mice. For each genotype, n = 2, with 5 sections for T2R genes; n = 1, with 3 sections for Pkd2l1 and Entpd2. Scale bars: 50 mm. B, D, F, qPCR analyses. T2R genes used for in situ hybridization were individually measured (B). n = 5 for each genotype. Open circles, individual data; white and red bars are the mean 6 SD of WT and Etv1 C/C tissues, respectively. Significance was assessed by Welch's t test. For results in fungiform papillae and soft palate, see Extended Data Figures 6-1 and 6-2, respectively. Figure 7. Presence of CreER 1 taste cells in Etv1-deficient taste buds. A, Triple-fluorescence labeling to study the impact of an Etv1 deficiency on the generation of Etv1 1 taste cells in circumvallate papillae: Trpm5 (green; left) to identify Skn-1a-dependent taste cells, tdTomato (red; left middle) to identify Etv1 1 taste cells, KCNQ1 (blue; right middle) that marks most taste bud cells, and merged images (right) of circumvallate taste buds of Etv1 C/1 ; R26 Tom/1 (top) and Etv1 C/C ; R26 Tom/1 (bottom) mice. n = 2, with 6 sections for each experiment. B, Triple-fluorescence images for IP 3 R3 (to identify Skn-1a-dependent taste cells, green) and tdTomato (to identify Etv1 1 taste cells, red), and Ddc (to identify sour taste cells, blue) in taste buds in circumvallate (CvP; left) and fungiform (FuP; center) papillae and palate (right) of Etv1 C/1 ; R26 Tom/1 (top) and Etv1 C/C ; R26 Tom/1 (bottom) mice (n = 3 for each genotype). wild-type mice (Fig. 3A,C; Extended Data Fig. 3-1A). qPCR showed drastic decrease in Tas1r3 expression only in CvP of Etv1 C/C mice (Fig. 3B,D; Extended Data Fig. 3-1B). However, Tas1r3 signals were barely detectable in the taste buds in FuP and palate of Etv1 C/C mice when signal was developed for a shorter time (Extended Data Fig. 3-1C,D), suggesting a decrease in Tas1r3 expression by Etv1 deficiency.
Scnn1a, a gene responsible for amiloride-sensitive salty taste, is expressed in all Pkd2l1 1 cells and a subset of Skn-1a 1 cells in taste buds in the FuP in wild-type mice, as reported previously (Chandrashekar et al., 2010;Ohmoto et al., 2020a). In Etv1 C/C mice, Scnn1a was rarely co-expressed with Skn-1a in Etv1 C/C mice (Fig. 4A), while its co-expression with Pkd2l1 1 in sour taste cells was unaltered ( Fig. 4B), indicating that Etv1 is required for the Scnn1a expression in sodium taste cells. However, qPCR analyses exhibited higher relative expression of Scnn1a than of T1R genes in FuP and did not show significant differences between wild-type and Etv1 C/C mice (Fig. 4C). Scnn1a may be expressed in many cells inside and/or outside taste buds at lower levels than in sour and sodium taste cells, and thus the decrease in sodium taste cells by Etv1 deficiency may be too small to affect the total amount of Scnn1a detected by qPCR.
Next, we studied the impact of Etv1 deficiency on the expression of genes encoding the downstream signaling molecules commonly required for sweet, umami, and bitter tastes. In situ hybridization analyses showed no obvious differences in the expression of any genes for taste signaling molecules between wild-type and Etv1 C/C mice ( Fig. 5A; Extended Data Figs. 5-1A, 5-2A). However, Calhm3 expression was significantly decreased by Etv1 deficiency in all gustatory areas examined, and the expression of Calhm1 and Plcb2 was also significantly decreased in one or two gustatory area ( Fig. 5B; Extended Data Figs. 5-1B, 5-2B). Although it seems to be largely dispensable for the expression of taste signaling molecule genes, Etv1 may participate in their expression.
We then examined the expression of genes that in wildtype mice are expressed in Etv1-negative taste bud cells. Frequency of signals to T2R genes in taste buds of Etv1 C/C mice were comparable to those in wild-type mice by in situ hybridization ( Fig. 6A; Extended Data Figs. 6-1A, 6-2A). Consistent with this, qPCR analyses showed no significant difference between wild-type and Etv1 C/C mice in CvP and FuP ( Fig. 6B; Extended Data Fig. 6-1B). Intriguingly, however, the expression of three out of four T2R genes examined was significantly increased in palate (Extended Data Fig. 6-2B). The expression of Pkd2l1 and the nonsensory taste bud cell marker Entpd2 were also unaltered in the taste buds of Etv1 C/C mice compared with wild-type mice (Fig. 6C-F Collectively, these results suggest that Etv1 is involved in the differentiation of taste cells by regulating expression of taste receptor and taste signaling molecule genes in sweet, umami, and sodium taste cells.
Differential impact of Etv1 deficiency on the composition of CreER 1 taste cells The residual expression of T1R genes in the Etv1-deficient mice indicates that Etv1 is not indispensable for the generation of taste cells. To evaluate whether Etv1 deficiency influences Etv1 1 taste cell populations, we generated Etv1 C/C ; R26 Tom/1 mice (in which tdTomato expression is activated in Etv1 1 cells by tamoxifen injection) and examined whether tdTomato expression was observed in taste buds. In all gustatory areas examined, tdTomato expression was observed in a subset of taste bud cells of Etv1 C/C ; R26 Tom/1 mice (Fig.  7A,B). Presence of tdTomato in taste buds of Etv1 C/C ; R26 Tom/1 mice suggests that Etv1 deficiency is not crucial for generation or maintenance of taste cells (Fig. 7A,B). We then identified Skn-1a-dependent taste cells using specific marker IP 3 R3; distinguished Etv1 1 and Etv1taste cells in Skn-1a-dependent cells as tdTomato 1 and tdTomatocells, respectively; identified sour taste cells by the expression of sour taste cell marker Ddc; and calculated their number in single taste buds. The population of tdTomato 1 cells was significantly decreased to about half in CvP and palate of Etv1 C/C ; R26 Tom/1 mice compared with that in the Etv1 C/1 ; R26 Tom/1 heterozygous control mice, while the populations of KCNQ1 1 , Ddc 1 , and IP 3 R3 1 cells were unaltered in any gustatory areas (Fig. 7C,D; Extended Data Fig. 7-1). Intriguingly, no significant difference in the population of tdTomato 1 cells was observed in FuP between Etv1 C/C ; R26 Tom/1 and Etv1 C/1 ; R26 Tom/1 heterozygous control mice (Fig. 7B,C). These results suggest that Etv1 is involved in generation and/or maintenance of taste cells. In CvP, the impact on the population of tdTomato 1 cells by Etv1 deficiency is less pronounced than that on gene expression, and the expression of T2R genes is not increased (Fig. 6). Although we cannot completely preclude the increase of bitter taste cells and the involvement of Etv1 in cell fate choice toward sweet, umami, or sodium taste cells over bitter taste cells, the decrease of tdTomato 1 cells (and thus the increase of tdTomatocells) seems to be simply explained by the decrease of CreER expression from Etv1 gene: Etv1 may be involved in the positive-feedback regulation of Etv1 gene expression in taste cells.
continued Taste buds and basal epithelial-mesenchymal boundaries are indicated by broken lines. Scale bars: 25 mm. C, D, Quantitative analyses of taste cell populations. The number of Skn-1a-dependent taste cells (C) and ratios of tdTomato 1 cells to Skn-1a-dependent taste cells (D) in single taste bud profiles in the CvP, FuP, and palate were compared between Etv1 C/1 ; R26 Tom/1 (white bar) and Etv1 C/C ; R26 Tom/1 (blue bar) mice. Data are expressed as the mean 6 SD. Significance was assessed by Welch's t test. For the populations of total taste bud cells and sour taste cells in single taste bud profiles, see Extended Data Figure 7-1. Figure 8. Expression of Etv1-related transcripts in geniculate ganglion neurons. A, In situ hybridization to show the expression of Etv1 and/or Etv1-related transcripts in geniculate ganglia of wild-type mice. Black arrow indicates nonspecific staining in the trunk of facial nerve. Scale bar: 100 mm. B, Double-label in situ hybridization to study expression of P2x2 in Etv1 1 neurons: Etv1 (green; left), P2x2 (magenta; middle), and merged (right) images of geniculate ganglia in wild-type (WT; top) and Etv1 C/C (bottom) mice. White arrowheads, Etv1 1 P2x2 1 neurons; white arrows, Etv1 À P2x2 1 neurons; and yellow arrows, Etv1 1 P2x2 À neurons. Scale bars: 100 mm. C, qPCR analyses to detect the expression of Etv1 and/or Etv1-related transcripts in geniculate ganglia. Primers used are sets N and C (for details, see Fig. 2C; Table 2). n = 5 for WT; n = 4 for Etv1 C/C mice. D, E, Ratio of Etv1 1 P2x2 1 neurons to P2x2 1 neurons (D) and ratio of P2x2 1 neurons to total neurons (E) in geniculate ganglia in WT and Etv1 C/C mice. n = 3 for each genotype. Data are expressed as the mean 6 SD. Significance was assessed by Welch's t test. F, P2X2-immunoreactive fiber projection into taste buds in fungiform papillae: bright field (left), P2X2 (green; left middle), KCNQ1 (magenta; right middle), and merged images To evaluate whether the decrease of Etv1 1 taste cells in Etv1 C/C mice is a cell-autonomous effect or a noncellautonomous effect caused by the loss of gustatory neurons, we interrogated whether Etv1 is expressed in sensory neurons in geniculate ganglia (GG) and, if it is expressed, whether Etv1 is required for their generation and/or innervation. In situ hybridization yielded moderate signals in GG neurons in wild-type mice (Fig. 8A), predominantly in a minor subset of gustatory neurons expressing P2x2 (Finger et al., 2005;Dvoryanchikov et al., 2017;Fig. 8B). Although qPCR showed little expression of Etv1, Etv1-derived transcripts putatively encoding Etv1 isoforms were amplified in GG in Etv1 C/C mice (Fig. 8C). Consistently, signals of Etv1derived transcripts were observed in P2x2-expressing GG neurons in Etv1 C/C mice by in situ hybridization (Fig. 8B). Importantly, the ratios of Etv1 1 P2x2 1 neurons to P2x2 1 neurons and P2x2 1 neurons to all GG neurons are comparable between Etv1 C/C and wild-type mice (Fig. 8D,E). P2X2 1 fibers also comparably projected to taste buds, but not to surrounding epithelium devoid of anti-KCNQ1 antibody immunoreactivity, from mesenchymal tissues in both wild-type and Etv1 C/C mice ( Fig. 8F; Extended Data Fig. 8-1). Etv1 and its isoform(s) appear to be dispensable for the generation and survival of putative gustatory neurons.

Taste responsiveness of gustatory nerves by Etv1 deficiency
Lastly, we examined electrophysiological responses to various taste substances from chorda tympani nerves that innervate taste cells in the FuP (Fig. 9A). Chorda tympani nerve responses to sucrose were significantly reduced but still observed in the Etv1 C/C mice, and a greater reduction in responses to artificial sweeteners was observed ( Fig. 9B; Table 3). Etv1 C/C mice exhibited decreased chorda tympani nerve responses to inosine monophosphate and monosodium glutamate, but their synergistic effects were not observed: the responses of MSG1IMP are comparable to the sums of individual responses of MSG and IMP in Etv1 C/C mice ( Fig. 9C; Table 3). Etv1 C/C mice showed severely reduced chorda tympani nerve responses to sodium chloride (NaCl) and no amiloride-sensitive components of the nerve responses to NaCl, while the amiloride-insensitive components remained almost normal ( Fig. 9D; Table 3). We observed no significant changes in chorda tympani nerve responses to chemicals with a bitter or sour taste in the Etv1 C/C mice (Fig. 9E,F; Table 3). These results of chorda tympani nerve recordings are well consistent with those of gene expression analyses in FuP. Together with the unaltered populations of putative gustatory neurons in GG and Etv1 1 taste cells in FuP, it is likely that gustatory neurons innervate taste cells in FuP in Etv1 C/C mice and that the altered gene expression in Etv1 1 taste cells in Etv1 C/C mice is a cell-autonomous effect of Etv1 deficiency in taste cells. The decreased responses of chorda tympani nerves to sweet, umami, and sodium taste substances in Etv1 C/C mice should thus be functional consequences of the diminished expression of T1R genes and Scnn1a because of Etv1 deficiency in taste cells.

Discussion
Continuous normal turnover of taste cells underlies homeostasis of the sense of taste. Single stem cells can give rise to any type of taste bud cell (Ren et al., 2014), but the molecular mechanisms that generate specific individual taste bud cells remain poorly understood. This study revealed that Etv1 participates in the differentiation of Skn-1a-dependent sweet, umami, and sodium taste cells by regulating the expression of taste receptor genes in them. Etv1 may be a part of the mechanism that, among Skn-1a-dependent cells, distinguishes sweet, umami, and sodium taste cells from bitter taste cells.

Differentiation of taste cells
Etv1 is expressed in sweet, umami, and sodium taste cells, and its deficiency impaired expression of genes encoding taste receptors and intracellular signal transduction molecules necessary for sweet, umami, and sodium tastes. Therefore, Etv1 appears to be involved in the differentiation of these taste cells in multiple ways. However, Etv1 deficiency does not completely abolish the expression of taste receptor genes. There may be other molecules that are functionally redundant to Etv1 and may have partially compensated for the loss of Etv1 in taste cells in Etv1 C/C mice; for example, Etv6 is expressed in Tas1r3 1 taste cells (Lee et al., 2017;Sukumaran et al., 2017). We presume that transcription factors specific to sweet, umami, and sodium taste cells, respectively, are expressed to characterize individual taste cell subsets and may compensate for Etv1 function to some extent in Etv1 C/C mice, which explains the differential impacts of Etv1 deficiency on gene expression. However, this hypothesis cannot fully explain regional differences in gene expression in specific taste cell subsets (e.g., Tas1r2 expression in sweet taste cells in FuP and palate but not in CvP) or the effects on the composition of Skn-1a-dependent taste cells by Etv1 deficiency. There may be more varieties of transcription factors in taste cells and more molecular mechanisms to generate such varieties than we understand today.
Expression of Etv1 and Eya1 seems to be reversely correlated in Skn-1a 1 cells in taste buds: Eya1 is expressed in the putative precursor and bitter taste cells (Ohmoto et al., 2021), while Etv1 is expressed in the remaining Skn-1a 1 cell types. In addition to its functions as a positive regulator of gene expression, Etv1 may negatively regulate T2R genes as a repressor in sweet, umami, and sodium taste cells. Partially consistent with this possibility, the expression of three out of four T2R genes examined was increased only in palate of the Etv1 C/C mice. This continued (right) of fungiform papillae of WT (top) and Etv1 C/C (bottom) mice. Fluorescent images of P2X2 1 signals were merged images of serial z-stacks. n = 2, with 3-4 sections for each genotype. Scale bars: 20 mm. For results in circumvallate papillae and soft palate, see Extended Data Figure 8-1. Figure 9. Etv1 deficiency in taste cells impairs neural responses to sweet, umami, and sodium tastes. A, Representative charts of chorda tympani nerve responses of wild-type (WT; top) and Etv1 C/C (bottom) mice to taste solutions: 100 mM NH 4 Cl, 300 mM sucrose, 0.5 mM inositol monophosphate (IMP), 100 mM monosodium glutamate (MSG), 100 mM MSG with 0.5 mM IMP, 100 mM sodium chloride (NaCl), 100 mM NaCl with 100 mM amiloride (Ami), 30 mM denatonium, and 10 mM citric acid. Responses of 0.5 mM IMP (orange) and 100 mM MSG (purple) are depicted as shaded rectangles to clarify the synergy (yellow) in the response of 100 mM MSG with 0.5 mM IMP. Responses to NaCl are depicted as shaded rectangles for amiloride-sensitive (blue) and amiloride-insensitive regional discrepancy may be explained by the different degrees of compensation by distinct, yet-to-be-identified transcription regulators that cooperate with Etv1 in sweet, umami, and sodium taste cells, as discussed above. Because these taste cells differ in their spatial distributions (Hoon et al., 1999;Chandrashekar et al., 2010;Ohmoto et al., 2020a), regional differences in impacts of Etv1 deficiency are perhaps unsurprising.
Many genes that encode transcription factors involved in the regulation of gene expression are expressed in taste buds (Hevezi et al., 2009;Lee et al., 2017;Sukumaran et al., 2017). Some of them may play crucial roles in specifying respective taste cells and maintaining their specific identities. Like Sox2 in stem cells and Skn-1a in precursor cells, several factors would act as selectors at several crossroads in the differentiation process of taste bud cells. It would be very interesting to identify which transcription factor genes are expressed in which taste cells by single-cell RNAseq and to analyze which genes' 5'-upstream regions are accessible  WT (circles) and Etv1 C/C (triangles) mice to bitter (E) and sour (F) taste solutions that are received by receptors other than T1Rs (n = 8 for each genotype). Significance was assessed by repeated-measures two-way ANOVA and the Sidak's multiple comparison test. *Significantly different (p , 0.05) responses between genotypes. Data are expressed as the mean 6 SEM; where error bars are not visible, they are smaller than the symbol depicting the mean. See Table 3 for statistical values.
for transcription machinery in individual taste cells by single-cell ATAC-seq.
Tissue-specific or cell-specific isoforms of Etv1 The Etv1 mutant strain used in this study, Etv1 C/C , is designed to encode a tamoxifen-inducible Cre recombinase CreER instead of Etv1, analogous to another Etv1 mutant strain, Etv1 nlslacZ/nlslacZ , which is designed to encode a nuclear-localized b-galactosidase (Extended Data Fig. 2-1A; Arber et al., 2000;Taniguchi et al., 2011). Therefore, these two strains were expected to exhibit similar phenotypes: impaired locomotion because of deficits in proprioceptive neuronal circuits, and a short life span (up to several weeks), as observed in Etv1 nlslacZ/nlslacZ mice, in which Etv1-like immunoreactivity was remarkably reduced in DRG neurons (Arber et al., 2000). However, Etv1 C/C mice exhibit no overt abnormality in locomotion, and this strain can be maintained as homozygous. More notable, Etv1-like immunoreactivity in DRG in Etv1 C/C mice is observed in the proprioceptive sensory neurons, comparable to wild-type mice (Fig. 2). Also, qualitative and quantitative PCR analyses demonstrated no significant change of Etv1-derived transcript in DRG in Etv1 C/C mice, consistent with qualitative observations by in situ hybridization analyses. However, no cDNA fragment was amplified from N-terminal coding exons, probably because transcription from the first exon of Etv1 is stopped by the insertion of CreERT2 in the second exon (Extended Data Fig. 2-1C). The fact that Etv1-v2 detected as a major transcript in DRG by 5'-RACE starts its transcription from the sequence in the fourth intron of Etv1 (Extended Data Fig. 2-1B,C) can account for little, if any, impact on its expression by CreERT2 insertion. The discrepancy of Etv1-v2 expression level between Etv1 nlslacZ/nlslacZ and Etv1 C/C mice may be attributed to the remaining pgk-Neo sequence in the Etv1 locus of Etv1 nlslacZ/nlslacZ mice (Extended Data Fig. 2-1A), as reported in other alleles (Fernex et al., 1997;Nagy et al., 1998;Kist et al., 2005;Joo et al., 2007;Andersen et al., 2014).
Intriguingly, Etv1-derived transcripts are also expressed in putative gustatory neurons in GG in wild-type and Etv1 C/C mice, but unlike in proprioceptive neurons in DRG and in taste cells, in GG in Etv1 C/C mice the expression of Etv1-derived transcripts is significantly decreased but not lost. Species and amounts of Etv1 isoforms seem to vary among cells and/or tissues. Identifying isoforms and examining the phenotypes in Etv1 1 cells in Etv1 C/C mice may provide insights into structure-function relationships of Etv1. For example, because Etv1 is absent in DRG in Etv1 C/C mice, which do not display impaired locomotion or proprioception, such phenotypes observed in Etv1 nlslacZ/nlslacZ and Etv1 ETS/ETS mice can be attributable to the loss of Etv1 isoform b encoded by Etv1-v2 in proprioceptive neurons but not to the loss of Etv1, although they may be functionally redundant. It is suggested that Etv1 is involved in different processes in multiple malignant tissues, such as carcinogenesis of prostate cancer, cell survival of gastrointestinal stromal tumor, and metastasis of pancreatic cancer (Tomlins et al., 2005;Chi et al., 2010;Heeg et al., 2016). It may be not Etv1 but an isoform that is involved in those pathologic conditions. Etv1 C/C mice, together with Etv1 nlslacZ/nlslacZ and Etv1 EST/EST mice, will be very valuable models to understand how Etv1 and its isoforms are involved in those processes in cancer studies, in addition to studying their roles in normal adult tissues.