Taste buds are aggregates of 50–100 cells, only a fraction of which express genes for taste receptors and intracellular signaling proteins. We combined functional calcium imaging with single-cell molecular profiling to demonstrate the existence of two distinct cell types in mouse taste buds. Calcium imaging revealed that isolated taste cells responded with a transient elevation of cytoplasmic Ca2+ to either tastants or depolarization with KCl, but never both. Using single-cell reverse transcription (RT)-PCR, we show that individual taste cells express either phospholipase C β2 (PLCβ2) (an essential taste transduction effector) or synaptosomal-associated protein 25 (SNAP25) (a key component of calcium-triggered transmitter exocytosis). The two functional classes revealed by calcium imaging mapped onto the two gene expression classes determined by single-cell RT-PCR. Specifically, cells responding to tastants expressed PLCβ2, whereas cells responding to KCl depolarization expressed SNAP25. We demonstrate this by two methods: first, through sequential calcium imaging and single-cell RT-PCR; second, by performing calcium imaging on taste buds in slices from transgenic mice in which PLCβ2-expressing taste cells are labeled with green fluorescent protein. To evaluate the significance of the SNAP25-expressing cells, we used RNA amplification from single cells, followed by RT-PCR. We show that SNAP25-positive cells also express typical presynaptic proteins, including a voltage-gated calcium channel (α1A), neural cell adhesion molecule, synapsin-II, and the neurotransmitter-synthesizing enzymes glutamic acid decarboxylase and aromatic amino acid decarboxylase. No synaptic markers were detected in PLCβ2 cells by either amplified RNA profiling or by immunocytochemistry. These data demonstrate the existence of at least two molecularly distinct functional classes of taste cells: receptor cells and synapse-forming cells.
Specialized neuroepithelial cells in taste buds detect chemical stimuli in the oral cavity and send signals to the brain via afferent cranial nerves. Bitter, sweet, and umami tastes are transduced by G-protein-coupled receptors (GPCRs) for taste (Chandrashekar et al., 2000; Chaudhari et al., 2000; Nelson et al., 2001, 2002). A specific form of phospholipase C, PLCβ2, is found in many taste cells (Rossler et al., 1998). In a semi-intact slice preparation, responses to umami tastants were found in taste cells that express PLCβ2 (Maruyama et al., 2006). Knock-out of the PlCb2 gene leads to profound taste deficits (Zhang et al., 2003; Dotson et al., 2005). These and other findings indicate that chemosensory transduction for bitter, sweet, and umami is predominantly mediated through a shared signaling pathway that involves phosphoinositide-mediated release of stored intracellular Ca2+ (Akabas et al., 1988; Spielman et al., 1996; Caicedo and Roper, 2001).
Although many details of these initial events of taste transduction have been explained recently, aspects of signal processing in taste buds and signal transmission to gustatory afferent nerve terminals remain unresolved, especially regarding the functional specialization of the different cell types. Murray (1974) described three subtypes of taste cells (types I–III) based on their ultrastructural characteristics. More recent electron microscopic studies have documented that only type III cells form synapses with primary sensory afferent terminals and that the presynaptic plasma membrane synaptosomal-associated protein 25 (SNAP25) is associated with synaptic junctions between type III taste cells and nerve terminals (Yang et al., 2000a). Conversely, the taste-specific G-protein α-gustducin is expressed in type II cells (Yang et al., 2000b). These and related morphological and immunocytochemical data have led to the suggestion that there are at least three classes of cells within taste buds, one that possesses typical synapses (type III cells), another that expresses chemosensory transduction proteins (type II cells), and a third class that expresses none of these markers (type I cells) (Yee et al., 2001; Clapp et al., 2004).
The functional correlates of these three classes of taste cells are addressed in the present study. Specifically, we asked the following. Are the Snap25 and PLCb2 genes expressed in separate taste cell populations, as inferred from immunocytochemical studies? If so, do the cells expressing these proteins display functional properties expected for presynaptic cells and chemosensitive cells, respectively? Finally, does the expression of additional synapse-related genes support the designation of only one of these cell types as presynaptic to gustatory sensory afferent terminals? We used single-cell reverse transcription (RT)-PCR to examine the expression profiles of individual taste cells and correlated these findings with functional responses using calcium imaging. Our data confirm that there are two very distinct and separate classes of taste cells. We further show that one class has functional properties of gustatory receptor cells, whereas the other class has characteristics of presynaptic cells.
Materials and Methods
Physiological buffers, dyes, and reagents
Tyrode’s solution was composed of the following (in mm): 145 NaCl, 5 KCl, 2 CaCl2, 1 MgCl2, 10 HEPES, 5 NaHCO3, 10 Na pyruvate, and 10 glucose, pH 7.2. Low-Na/Ca buffer contained the following (in mm): 290 mannitol, 5 KCl, 3 MgCl2, 10 HEPES, 5 NaHCO3, 10 Na pyruvate, and 10 glucose, pH 7.2. Responses to 50 mm KCl were obtained using Tyrode’s solution with an equimolar substitution of KCl for NaCl. Tastants (cycloheximide and/or saccharin) were dissolved directly in Tyrode’s buffer. All imaging dyes were obtained from Invitrogen (Carlsbad, CA). Other reagents were obtained from Sigma (St. Louis, MO), unless otherwise indicated.
Tissues and cell collection
All procedures were approved by the University of Miami Animal Care and Use Committee. Adult C57BL/6J mice were killed with CO2 and cervical dislocation as recommended by the National Institutes of Health (www.grants.nih.gov/grants/olaw/references/phspol.htm). The tongue was removed, a protease mixture consisting of 3.2 mg/ml collagenase, type A (Roche Products, Indianapolis, IN), 8 mg/ml dispase (Roche Products), and 0.8 U/ml purified elastase (Worthington, Lakewood, NJ) was injected under the circumvallate papilla, and the epithelium was peeled away after 20 min. In some experiments, we injected an alternative enzyme, Protease XXIII (4 mg/ml; Sigma), dissolved in low-Na/Ca buffer (see above), and a small block of the tongue was incubated in the same buffer for 10–15 min before delamination. We did not observe a consistent difference between these enzyme mixtures with respect to the yield or health of isolated cells. The peeled epithelium was then redigested in the enzyme mixture for 2 min, followed by 1000 U/ml DNase I (Sigma) for 5 min. Isolated taste cells were then gently collected with a polished glass pipette (inner diameter, 80 μm) expelled into 5 μl drops of Tyrode’s buffer onto Cell-Tak (BD Biosciences, San Jose, CA)-coated coverslips and were allowed to settle before washing, harvesting, or recording.
To measure Ca2+ responses evoked by taste stimuli and/or by potassium depolarization, we adapted two different approaches to load taste cells with calcium-sensitive fluorescent dyes. In some cases, taste cells in the intact tongue were loaded iontophoretically with Calcium Green-1 dextran (3000 molecular weight) as described by Caicedo et al. (2002) and Richter et al. (2003). Dye-loaded taste cells were then isolated as outlined above. In other experiments, taste cells (not preloaded with dye) were collected and then incubated in Calcium Green-1 AM (10 μm) or fura-2 AM (10 μm) in Tyrode’s solution for 45–60 min before washing and recording. We used these alternative methods to test whether the dye-loading procedures had an impact on the integrity of cellular RNA. Functional data and RT-PCR results from either method for dye loading and for either calcium-sensitive dye did not differ significantly for the purposes of this report. Hence, data were pooled.
Lingual slice recordings.
To record taste cell responses in living isolated slices of lingual epithelium, vallate papillae were prepared and taste cells were loaded with Calcium Orange (CaO), as described fully by Richter et al. (2003). CaO was used to image functional responses because some of the taste cells in these experiments expressed green fluorescent protein (GFP) (see below). The use of appropriate excitation and emission filters eliminated spectral overlap between GFP and CaO, thereby allowing us to image calcium responses from taste cells expressing GFP.
Calcium imaging was conducted using a Fluoview laser scanning confocal microscope and software (Olympus America, Melville, NY) for isolated cells loaded with Calcium Green dextran and for lingual slices (above). For isolated cells loaded with fura-2 AM, we used an imaging system based on an inverted microscope (IX70; Olympus America), cooled-CCD camera, and Imaging Workbench software (Indec Biosystems, Mountain View, CA). We recorded images at 3–5 s intervals. Stimuli were bath applied. The baseline period before stimulation was averaged to calculate Fo. For each time point, the change in fluorescence was calculated as ΔF/Fo and was considered a stimulus-evoked response if the ΔF/Fo was ≥5% for several successive data points. The average baseline signal fluctuation was ≤2%.
To screen for expression of selected genes and to validate each pair of primers (supplemental Table 1, available at www.jneurosci.org as supplemental material), we performed RT-PCR on taste buds or delaminated nontaste lingual epithelium. RNA was purified from tissues using the RNA microprep kit and included a digestion with DNase I (Stratagene, La Jolla, CA). cDNA was synthesized using Superscript III reverse transcriptase (Invitrogen) as described previously (Richter et al., 2004). To validate the specificity of PCRs, we performed parallel reactions on cDNA from taste buds, nontaste lingual epithelium, and on water in place of template. In the case of the genes expressed at low abundance, such as the calcium channel subunits, we also verified the specificity of the single-cell RT-PCR by Southern blot hybridization with a previously sequenced DNA used as probe. Optimum annealing temperature for each primer pair was determined on a gradient in the iCycler (Bio-Rad, Hercules, CA). The template for each RT-PCR was limited to one taste bud equivalent of cDNA for 35 cycles only. PCR products obtained with taste cDNA for every primer pair were sequenced to further validate specificity. Primers were designed to span at least one intron and were positioned as close to the 3′ end as practical, unless there were known splice variants in the region. Primers of a pair were located within a single exon only in the case of synapsin II and calcium channel α1A (set B) because of unusually long final exons and, in the case of Tas2r105, an intronless gene.
Dissociated taste cells were individually collected under microscopic examination, either as “naive” cells (i.e., without dye loading and functional imaging) or after Ca2+ imaging. We verified under 200× magnification that only a single cell was collected in a minimum volume (≤20 nl) of Tyrode’s buffer, using 5- to 10-μm-diameter glass pipettes. Cells were expelled into a tube containing 50 μl of cell lysis buffer containing guanidine thiocyanate, β-mercaptoethanol, and 200 ng of polyinosinic acid, and the tip of the collection pipette was broken into the tube. Total cellular RNA was isolated using the Absolutely RNA Nanoprep kit (Stratagene). RNA was eluted from kit-supplied columns in 10 μl of Tris-HCl, pH 7.5, and was immediately denatured for 5 min at 65°C in the presence of oligo-dT(12–18) and dNTP. First-strand cDNA synthesis was then initiated with the addition of 8 μl of RT reaction mix and 200 U of Superscript III and then incubated for 60 min at 50°C. The resulting 20 μl of single-cell cDNA was divided as follows: 2 μl for β-actin and 5 μl each for SNAP25 and PLCβ2. Each PCR was performed in 20 μl for 45 cycles. The remaining 8 μl of each single-cell cDNA were used for preliminary screening PCRs for calcium channels and other genes (see Results, Genes expressed in presynaptic taste cells). Positive control reactions using cDNA from taste buds and negative controls (water substituted for cell sample before RNA purification) were run in parallel from master mixes.
RNA amplification followed by RT-PCR
T7 RNA amplification was performed using the MessageBOOSTER cDNA Synthesis kit for qPCR (MB051224; Epicentre, Madison, WI) essentially according to the instructions of the manufacturer. Briefly, single-cell RNA was purified as described above, and the volume of each eluted RNA sample was adjusted to 3 μl by evaporation in a Savant Speed-Vac (GMI, Ramsey, MN). First-strand cDNA was synthesized using a T7-oligo-dT anchor primer and Superscript III reverse transcriptase for 30 min at 50°C. Then, second-strand cDNA was synthesized with RNase H and DNA polymerase I, and the double-strand cDNA served as template in an in vitro transcription reaction using T7 RNA polymerase. The resulting amplified antisense RNA (aRNA) was treated with DNase I, purified in RNeasy MinElute spin columns (Qiagen, Valencia, CA), and reverse transcribed into first-strand cDNA using Superscript II and random hexamer primers for 1 h at 37°. Diluted cDNA (2–4% for each reaction) then served as a template in PCR analysis with primers for taste-specific markers (supplemental Table 1, available at www.jneurosci.org as supplemental material). The conditions for PCR were 94°C for 2 min, followed by 45 cycles of 94°C for 30 s, 57–60°C for 30 s, and 72°C for 30 s.
We verified that GFP expression was an accurate marker for endogenous PLCβ2 expression in PLCβ2-GFP transgenic mice using immunofluorescence (Kim et al., 2006). Circumvallate papillae, fixed with 4% paraformaldehyde, were cryosectioned at 25 μm, and sections were incubated overnight with rabbit anti-PLCβ2 (1:1000; Santa Cruz Biotechnology, Santa Cruz, CA). We also immunostained tissues with rabbit anti-SNAP25 (1:500; AB1762; Chemicon, Temecula, CA), rabbit anti-neural cell adhesion molecule (NCAM) (1:500; AB5032; Chemicon), or rabbit anti-aromatic amino acid decarboxylase (AADC) (1:500; GTX30448; Genetex, San Antonio, TX). Thorough validation in taste tissue has been demonstrated previously for many of these antibodies, including anti-PLCβ2 (Kim et al., 2006), anti-NCAM (Yee et al., 2001), and anti-SNAP25 (Yang et al., 2000a). After three washes in buffer, sections immunostained for PLCβ2 were incubated in goat anti-rabbit IgG, conjugated to Alexa Fluor 594 (1:1000; Invitrogen). Immunostaining for SNAP25, NCAM, and AADC was amplified using tyramide following instructions provided with the kit [T-20925 (Invitrogen); 1:500 diluted goat anti-rabbit IgG, horseradish peroxidase conjugate, and Alexa Fluor 594–tyramide substrate, diluted 1:100]. Negative controls were processed in parallel in every experiment, with primary antibody omitted. No nonspecific fluorescence was detected (see Fig. 6D). Images were obtained with a Zeiss Microimaging (Thornwood, NY) Axioplan epifluorescence microscope using Axiovision version 3.0 software (for Nomarski differential interference contrast optics) and an Olympus America laser-scanning confocal microscope using Fluoview software (for GFP and immunofluorescence). We estimate a thickness of ∼3 μm for the optical sections taken from the confocal microscope.
Data on the expression of individual genes in single-cell aRNA/RT-PCR was compared between two cell populations using either a χ2 test or the more stringent two-tailed Fisher’s exact test. The same tests were also used on data comparing the occurrence of depolarization-triggered Ca2+ responses in GFP-expressing and GFP-lacking cells in lingual slices from transgenic mice. All statistics were calculated using Prism version 4.0 (GraphPad Software, San Diego, CA).
Taste cells respond to either tastants or depolarization
We hypothesize that taste buds contain separate populations of chemosensory receptor cells and presynaptic cells (i.e., cells that form synapses with gustatory afferent nerve terminals). As a first step toward testing this hypothesis functionally, we applied Ca2+ imaging to analyze stimulus-evoked responses in individual isolated taste cells (Fig. 1A). We recorded responses to a mixture of prototypic taste stimuli and to potassium depolarization. From an estimated 1032 isolated, imaged cells, 53 responded to depolarization with 50 mm KCl with an increase in intracellular calcium and 34 to stimulation with a mixture of a sweet-tasting (2 mm saccharin) and a bitter-tasting (100 μm cycloheximide) compound (Fig. 1B,C). [Saccharin and cycloheximide were selected because they are effective in eliciting Ca2+ responses in taste cells in a semi-intact slice preparation (Caicedo et al., 2002) and in isolated taste buds and cells (Bernhardt et al., 1996; Huang et al., 2005).] None of the isolated cells responded to both KCl depolarization and taste stimulation (Fig. 1D). As noted, the majority of cells tested failed to generate responses to either stimuli. Because the C57BL/6J strain is known to be highly sensitive to both cycloheximide and saccharin (Bachmanov et al., 2001; Boughter et al., 2005), we considered what other factors might contribute to the low incidence of responses. First, a large fraction of cells in the taste bud are thought to function as glial-like or supporting cells and lack depolarization-activated Ca2+ fluxes [e.g., type I cells by Yee et al. (2001); Medler et al., (2003)]. Additionally, cells in our analysis may have not responded because our stimulus mixture was limited to only two tastants (saccharin and cycloheximide), the cell was unhealthy, or a combination of these factors. We interpreted that cells responding to the tastant mixture represent gustatory receptor cells, whereas cells responding to depolarization are putative presynaptic cells. This interpretation was tested further by gene expression profiling (below).
Taste cells express either PLCβ2 or SNAP25
Because the functional responses of isolated taste cells fell into two distinct classes, we asked whether key molecular markers might also be expressed in two categories of cells. For the postulated presynaptic (i.e., KCl-responsive) cells, we selected SNAP25 because it is known to be associated with synapses in general and with taste cell afferent synapses in particular (Yang et al., 2000a). For receptor cells, we selected PLCβ2 because it is a downstream signaling enzyme shared by known taste GPCRs (Rossler et al., 1998; Zhang et al., 2003). We collected individual taste cells and processed the extracted RNA for single-cell RT-PCR for β-actin, SNAP25, and PLCβ2. Examples of harvested cells and the resulting RT-PCR data are shown in Figure 2, A and B. The presence of β-actin RT-PCR product from 10% of cell cDNA served as an indication that RNA of sufficient quality and quantity was recovered from individual cells. Of 51 β-actin-positive cells, 16 expressed SNAP25 (31%), whereas 10 expressed PLCβ2 (20%). Only one cell expressed both SNAP25 and PLCβ2 (Fig. 2C). These data demonstrate that most taste cells express SNAP25, PLCβ2, or neither. Coexpression of these markers is rare (∼2% of cells tested).
Physiological responses correlate with molecular expression
Our next step was to test whether the two classes of cells determined by functional imaging mapped onto the two categories determined by expression of SNAP25 and PLCβ2. We recorded calcium responses to 50 mm KCl and the tastant mix (2 mm saccharin plus 100 μm cycloheximide) from individual taste cells as described above, followed by single-cell RT-PCR for β-actin, SNAP25, and PLCβ2 from the same cells. Although the sample size was limited because of a low incidence of functional responses (as noted above), combined with the difficulty with degradation of RNA from taste cells during calcium imaging, the results nonetheless were clear-cut. From a total of 22 cells analyzed, every cell that expressed SNAP25 responded to KCl depolarization (n = 7), and every cell that expressed PLCβ2 responded to tastant stimulation (n = 3). [We noted that four KCl-responsive cells and eight tastant-responsive cells expressed neither SNAP25 nor PLCβ2. We expect that many of these apparent nulls were attributable to degradation of RNA during dye loading and functional imaging; however, we cannot rule out the possibility that these functionally identified cells may represent additional molecular classes of taste cells.] The results confirmed that there was no overlap in functional responses between KCl depolarization and taste stimulation, as observed in the first series of experiments and no overlap in expression of SNAP25 and PLCβ2 as in the second series. Furthermore, SNAP25 appeared to be expressed only in depolarization-responsive cells, whereas PLCβ2 expression appeared only in tastant-responsive cells.
To confirm this interpretation for a larger number of cells and with an independent methodology, we took advantage of transgenic mice in which cells expressing PLCβ2 were genetically tagged with GFP (Kim et al., 2006). Green fluorescent protein was expressed under the control of 2.9 kb of the mouse PlCb2 gene promoter. Immunocytochemical analysis of taste papillae from the 5288 line of transgenic mice (Kim et al., 2006) showed that PLCβ2 immunoreactivity and GFP expression showed near-perfect overlap (Fig. 3A) (i.e., the transgene was expressed identically to the endogenous gene). We then examined calcium responses in taste buds from the PLCβ2-GFP transgenic mice. By using a well established slice preparation in which taste cells are maintained in a more native environment (Caicedo et al., 2002), we avoided problems with cell viability after isolation. We loaded taste cells in circumvallate papillae with Calcium Orange (a calcium-sensitive fluorescent dye whose spectral characteristics allow imaging in cells expressing GFP), prepared 100 μm vibratome sections of taste papillae, and imaged taste cells with confocal microscopy (Fig. 3B). Bath-applied 50 mm KCl elicited a transient elevation of intracellular Ca2+ in 22% of CaO-loaded taste cells (Fig. 3C) (Caicedo and Roper, 2001; Richter et al., 2003). Importantly, responses to potassium depolarization were detected only in taste cells lacking GFP (31 of 100 cells), that is, in cells that do not express PLCβ2. No cells expressing GFP (0 of 31) responded to potassium depolarization, yielding a highly significant difference between the two populations (Fig. 3C) (p < 0.0001; two-tailed Fisher’s exact test). Cells expressing GFP did respond to stimulation with the bitter tastant cycloheximide (Fig. 3C), indicating that the presence of GFP, per se, does not occlude Ca2+ signals. These data confirm and extend the above studies on isolated single cells and show that potassium depolarization activates cells that express SNAP25 and lack PLCβ2. A separate population of taste cells that expresses PLCβ2 is sensitive to tastants.
Genes expressed in presynaptic taste cells
Considerable recent effort has focused on molecular characterization of tastant-responsive cells and the transduction pathways from molecular receptors to the production of tastant-evoked Ca2+ signals (Huang et al., 1999; Perez et al., 2002). However, much less is known about the cells in taste buds that possess synapses (i.e., those expressing SNAP25). To understand the significance of these presynaptic cells, we sought to identify additional genes expressed in them. To select appropriate candidates for gene expression profiling in single cells, we first screened a number of candidate genes using RT-PCR on whole taste buds as described next.
Voltage-gated calcium channels
The first class of genes that we examined consisted of the voltage-gated calcium channels because these typically are essential for neurotransmitter exocytosis and for producing the depolarization-evoked calcium transients that we and others have observed in taste cells. There are 10 known genes for the channel-forming (α1) subunits of voltage-gated calcium channels (Catterall, 2000; Yu and Catterall, 2004). We focused on the seven high-threshold activated channels (α1A–α1F, α1S) because low-threshold T-type channels (α1G–α1I) with their rapidly inactivating currents are less likely to be responsible for the major presynaptic calcium signal. We designed primer pairs for each of these and used them for RT-PCR on RNA extracted from taste buds and adjacent nontaste lingual epithelium. This initial screening indicated that sequences corresponding to α1A and α1B (typical presynaptic P/Q- and N-type channels, respectively) and α1C (a widespread L-type channel) were preferentially expressed in taste buds relative to nontaste lingual epithelium (Fig. 4A). The α1D subunit (a neuroendocrine L-type channel) was expressed in both taste and nontaste lingual epithelium. In contrast, α1E (the R-type channel), α1F (a retinal L-type channel), and α1S (the skeletal L-type channel) were expressed at very low levels or not at all (results not shown).
Next, we conducted a preliminary series of RT-PCRs on single-cell cDNAs (as in Fig. 2). One of the high-threshold calcium channels, α1A, was found to be expressed in many of the same cells as SNAP25. In contrast, α1B was detected in only 1 of 33 cells, and this cell expressed neither SNAP25 nor PLCβ2 (results not shown). We were unable to detect α1C in single-cell cDNA, suggesting that its expression level per cell may be low. Based on these findings, α1A was selected for the detailed analyses of amplified RNA from single taste cells (see below).
Physiological analyses have suggested that taste cells synthesize, take up, and release biogenic amine neurotransmitters, including serotonin and norepinephrine (Nagai et al., 1996; Herness et al., 2002; Kaya et al., 2004; Huang et al., 2005). Although the precise role of these neurotransmitters in taste signaling remains to be established, serotonin for one has been localized to synapses in taste buds (Takeda and Kitao, 1980). AADC (also called DOPA decarboxylase), is a biosynthetic enzyme common to the pathways for serotonin, dopamine, norepinephrine, and epinephrine. mRNA for AADC is expressed as neuronal and non-neuronal isoforms by transcription from two alternate promoters (Jahng et al., 1996). We screened whole taste buds by RT-PCR and found robust, taste-selective expression of the neuronal form. No RT-PCR product was detected for the non-neuronal form (Fig. 4C). Hence, AADC was included in the single-cell profiling from amplified RNA below.
Glutamic acid decarboxylase
Although the evidence is less strong than for the biogenic amines, the inhibitory neurotransmitter GABA has also been implicated in taste bud function (Nagai et al., 1998; Cao et al., 2005; Eram and Michel, 2005). GABA is synthesized through the action of one of the two isoforms of glutamic acid decarboxylase (GAD1 and GAD2). By RT-PCR on whole taste buds, we detected expression of one of these genes, GAD1, in a taste bud-selective manner (Fig. 4B) and thus this gene was also included in our profiling of amplified RNA from single cells.
NCAM and synapsin
Many neuronal cells and some taste cells (Nolte and Martini, 1992) express NCAM, a surface glycoprotein involved in homophilic adhesion. By RT-PCR, we detected robust expression of NCAM in taste buds, although it was not seen in nontaste lingual epithelium (Fig. 4B). Finally, we also selected one additional synapse-related protein, synapsin II, that is associated with the membrane of synaptic vesicles (Südhof, 2004). By RT-PCR, we detected robust expression of synapsin II in taste buds (Fig. 4G).
Taste transduction proteins
Finally, to confirm that cells expressing PLCβ2 were indeed taste receptor cells, as postulated, we tested for expression of several well characterized components of the taste transduction cascade. These included the following: T2R5, a bitter taste receptor (Chandrashekar et al., 2000), T1R3, a component of some sweet and umami receptors (Nelson et al., 2001, 2002), IP3R3, a calcium-release channel/inositol triphosphate receptor (Miyoshi et al., 2001), and TRPM5, a cell surface ion channel (Perez et al., 2002). Primers for each of these were validated with taste bud and nontaste cDNA (data not shown) and were then included in the profiling below.
Expression profiles of individual cells
We next wanted to determine whether the expression of any of the above synapse-related genes was associated specifically with the SNAP25-expressing cells that we defined functionally in this study. We used a strategy of single-cell isolation followed by RNA amplification (Kacharmina et al., 1999). This allowed us to analyze the expression of a much larger number of genes than was possible by the direct RT-PCR method used above. We isolated single taste cells, purified cellular RNA, converted it to double-strand cDNA, transcribed antisense RNA, and then again converted this to single-strand cDNA. The resulting cDNA was used as template in PCRs for β-actin, PLCβ2, and SNAP25. Of 53 individual cells that tested positive for β-actin, we selected 10 cells that were positive for PLCβ2 and another 10 that were positive for SNAP25. Only one cell in this series (1 of 53) expressed both PLCβ2 and SNAP25. We included this cell in the detailed analysis below. The aRNA from these 21 cells was then subjected to RT-PCR for the genes selected above. Figure 5A shows an example of aRNA/RT-PCR results from one cell, and a full compilation of results from the 21 cells is shown in Figure 5B.
As expected, PLCβ2-positive cells expressed IP3R3 (11 of 11 cells) and TRPM5 (10 of 11 cells). Furthermore, 6 of 11 of these cells also expressed a known taste receptor, either T1R3 or T2R5. Thus, we conclude that most or all PLCβ2-expressing cells can be designated “taste receptor cells” as postulated. None of the critical transduction genes, Trpm5, Tas1r3, or Tas2r105, was expressed in any of the SNAP25-positive/PLCβ2-negative cells. Conversely, IP3R3 was found in half of the SNAP25-positive cells.
In contrast to the above results on PLCβ2-positive taste receptor cells, 10 of 11 of the SNAP25-positive cells expressed the neuronal surface adhesion protein NCAM. Most SNAP25-positive cells also expressed the synthetic enzymes for biogenic amines and GABA neurotransmitters, AADC and GAD1, respectively. One-half or more of the SNAP25-positive cells also expressed the P/Q-type presynaptic voltage-gated calcium channel α1A and/or as the synaptic vesicle protein synapsin II. It is significant that none of these mRNAs was detected in any of the cells expressing PLCβ2. In summary, the gene expression profiles displayed two distinct cell types that we designate “receptor cells” and “presynaptic cells” (Fig. 5C, white and gray bars, respectively). The incidence of cells expressing both SNAP25 and PLCβ2 is low (overall, 2 of 104 cells).
As an additional test of the genetic profiling (presented in Fig. 5), we immunostained sections of circumvallate papillae to test whether protein expression also segregated into groups of taste cells. Specifically, we took advantage of the PLCβ2-GFP transgenic line of mice to test whether the receptor cell marker PLCβ2 was coexpressed with protein markers for presynaptic cells or whether they were truly expressed in separate cells. These mice allowed us to use robust, well characterized primary antibodies (raised in rabbit). The results were unambiguous: expression of GFP (and, by inference, PLCβ2) did not overlap with either SNAP25, NCAM, or AADC (Fig. 6). Antibodies to other protein markers from our genetic profiling results did not yield sufficiently distinct or specific immunostaining in taste buds to ascertain the extent of coexpression with PLCβ2.
Our results demonstrate the existence of at least two separate classes of taste cells. One class expresses neuronal and synaptic proteins, including SNAP25, NCAM, synapsin II, and the presynaptic voltage-gated calcium channel α1A. These cells also express GAD1 and AADC, key enzymes in the synthesis of GABA and aminergic neurotransmitters, respectively. A separate class of taste bud cells expresses chemosensory signaling molecules, including taste receptors, PLCβ2, and TRPM5. In physiological tests using calcium imaging, taste cells positive for synaptic proteins responded to potassium depolarization but not tastants. Conversely, cells positive for PLCβ2 responded to tastant stimulation but not to depolarization. That is, the two classes of cells based on molecular and functional properties were separate and nonoverlapping. In the lexicon of taste cell morphotypes, receptor/PLCβ2-positive cells would be approximately equivalent to type II taste cells, whereas synaptic/SNAP25-positive cells would approximate type III taste cells (Yee et al., 2001). This concept extends a previous interpretation based on immunocytochemical and electron microscopic evidence (Yee et al., 2001, 2003; Clapp et al., 2004; Yang et al., 2004). Notably, the presence of synapses in electron micrographs was one of the identified characteristics of type III cells (Murray, 1974). Taste buds also contain one or more classes of supporting and progenitor cells.
Taste cells positive for PLCβ2 consistently expressed genes that have been tied previously to chemosensory transduction. These include TRPM5, a transient receptor potential ion channel that plays a critical but as yet poorly understood role in taste transduction for bitter, sweet, and umami tastants (Perez et al., 2002; Zhang et al., 2003; Damak et al., 2006). The taste receptors T1R3 and T2R5 were found in separate cells and always were coexpressed with PLCβ2, as suggested by previous immunocytochemical and in situ hybridization analyses (Adler et al., 2000; Miyoshi et al., 2001; Zhang et al., 2003). The intracellular channel IP3R3 links receptor-triggered PLCβ2 activation to calcium release from stores, and we detected it in all PLCβ2-positive cells. Expression of IP3R3 was also seen in half of the SNAP25-positive cells analyzed, suggesting that, in nonreceptor cells, IP3R3 may play additional roles, as suggested previously (Kataoka et al., 2004). In the rat, IP3R3 has been reported to be expressed in NCAM-expressing cells only infrequently (Clapp et al., 2004), whereas we find the incidence of this overlap substantial in the mouse.
We detected the P/Q-type voltage-gated calcium channel subunit α1A in half of the tested presynaptic cells but none of the receptor cells. High-threshold voltage-gated calcium channels, typically of the N, P/Q, or R types, are a major component of calcium-dependent transmitter release at presynaptic sites (Dunlap et al., 1995). Our failure to detect α1A in all presynaptic cells may be attributable to the existence of multiple subtypes of SNAP25-expressing cells. Alternatively, mRNA for α1A or other calcium channel transcripts may be expressed at lower copy number per cell, making them more difficult to detect. Our results are mostly consistent with those of Medler et al. (2003), who showed using patch-clamp analyses that voltage-activated calcium currents were present in cells immunopositive for NCAM. However, we found no molecular parallels with the subset of type II cells (presumably, receptor cells) reported to have voltage-activated calcium currents (Medler et al., 2003). It is possible that the antigen A immunoreactivity that was used in that previous study may have labeled some nonreceptor cells as well.
In the present study, we focused on receptor cells for sweet, bitter, and umami tastants. We showed previously (Richter et al., 2003, 2004) that Ca2+ responses elicited by sour tastants are consistently found in taste cells with voltage-gated calcium currents (i.e., presynaptic cells according to the present study). Thus, some of the presynaptic cells in taste buds may well be sour sensors. The present distinction between receptor versus presynaptic cells may need to be expanded to accommodate sour (and possibly salty) taste detection once evidence correlating function with molecular expression of diagnostic genes is available at the level of individual cells.
Our molecular profiling of the SNAP25-expressing taste cells provides some insights into their role(s) in taste buds. First, we noted that every one of these cells expressed NCAM, a glycoprotein member of the immunoglobulin superfamily. NCAM is associated with nerve-target recognition, synaptogenesis, and synaptic plasticity, processes that are critical in the continually renewing taste neuroepithelium (Smith et al., 1993). Second, we observed that SNAP25 cells very often express AADC and GAD1, enzymes essential in the biosynthesis of biogenic amine and GABA neurotransmitters, respectively. A considerable body of literature substantiates that small numbers of cells within taste buds selectively take up and accumulate the serotonin precursor 5-hydroxytryptophan and synthesize serotonin under physiological conditions (Takeda and Kitao, 1980; Nagai et al., 1998). Recent physiological studies using biosensors have demonstrated serotonin release from taste buds after stimulation with tastants (Huang et al., 2005). Sensitivity to serotonin may lie in the afferent nerve terminal or on other taste cells or both, as suggested by patch-clamp analyses (Kaya et al., 2004). Norepinephrine too has been implicated in taste bud function (Herness et al., 2002). Although direct evidence for GABA in taste bud function is limited (Cao et al., 2005), recent recordings from taste afferent nerves do suggest the possibility of inhibitory signals (Danilova et al., 2002; Frank et al., 2005). In summary, although many of these neurotransmitters have been implicated in some manner, we present the first evidence that biogenic amines and GABA may be synthesized and exocytosed from the same cells that have structural components of synapses, such as SNAP25 and synapsin II. Recently, it was shown ATP is a neurotransmitter essential for signaling to the taste afferent nerve (Finger et al., 2005). The mechanism for ATP release from taste buds is presently not known, but it will be important to search for molecular markers for this pathway among taste cell types.
Only 2 of 103 cells analyzed by single-cell RT-PCR showed overlapping expression of both SNAP25 and PLCβ2. It is possible that these infrequent cells are an artifact generated by collecting portions of two cells during cell harvesting. Alternatively, they may represent occasionally imprecise gene expression, possibly as a consequence of cell renewal in the taste bud. We cannot rule out the possibility that these rare examples represent cells transforming from one functional class into the other, as suggested previously (for review, see Yee et al., 2001). We conclude that, in the main, there are separate cells for sensing bitter, sweet, and umami taste stimuli (i.e., PLCβ2 receptor cells) and for transmitting signals to gustatory afferent fibers (i.e., SNAP25 synaptic cells), at least via conventional synapses. The existence of these two functional classes of taste cells leads to the question of how taste signals are transmitted to the gustatory afferent nerve. The simplest model would have a taste receptor cell form a synapse with an afferent nerve. However, as we show here, PLCβ2 cells lack synaptic vesicle proteins (SNAP25 and synapsin II), and we found no functional evidence for depolarization-induced Ca2+ influx, a hallmark of typical presynaptic cells. One possible resolution is that gustatory receptor cells directly excite sensory afferent fibers via nonconventional synapses, as proposed recently (Clapp et al., 2004). Such nonconventional synapses remain to be identified functionally and might use very different molecules and functional pathways than the ones we examined here. Alternatively, cells may communicate within the taste bud, for instance, transferring signals from receptor cells to presynaptic cells and then on to the nerve (Herness et al., 2002; Kaya et al., 2004). Mechanisms for such intragemmal communication remain an important but unresolved mystery. If indeed sensory information passes from receptor cells to presynaptic cells within taste buds, this presents the possibility that signals originating in receptor cells of different chemosensitivity (say, umami and bitter) may converge onto and thus be integrated in common presynaptic cells. Such an interpretation could resolve a major conundrum in the literature: whether taste cells are narrowly or broadly tuned (Gilbertson et al., 2001; Caicedo et al., 2002; Zhang et al., 2003). That is, receptor cells could be highly tuned, and presynaptic cells less so.
This work was supported by National Institutes of Health–National Institute on Deafness and Other Communication Disorders Grants 2R01 DC00374 (S.D.R.) and 1R21 DC05500 and 1R01 DC06308 (N.C.). We thank Naomi Rosenkranz and Kristina Trubey for help with early experiments in this project.
↵*R.A.D. and G.D. contributed equally to this work.
- Correspondence should be addressed to Dr. Nirupa Chaudhari, Department of Physiology and Biophysics, University of Miami Miller School of Medicine, 1600 NW 10th Avenue, Rosenstiel Medical Sciences Building 4040, Miami, FL 33136. Email: