Abstract
Peripheral taste neurons exhibit functional, genetic, and morphological diversity, yet understanding how or if these attributes combine into taste neuron types remains unclear. In this study, we used male and female mice to relate taste bud innervation patterns to the function of a subset of proenkephalin-expressing (Penk+) taste neurons. We found that taste arbors (the portion of the axon within the taste bud) stemming from Penk+ neurons displayed diverse branching patterns and lacked stereotypical endings. The range in complexity observed for individual taste arbors from Penk+ neurons mirrored the entire population, suggesting that taste arbor morphologies are not primarily regulated by the neuron type. Notably, the distinguishing feature of arbors from Penk+ neurons was their propensity to come within 110 nm (in apposition with) different types of taste-transducing cells within the taste bud. This finding is contrary to the expectation of genetically defined taste neuron types that functionally represent a single stimulus. Consistently, further investigation of Penk+ neuron function revealed that they are more likely to respond to innately aversive stimuli—sour, bitter, and high salt concentrations—as compared with the full taste population. Penk+ neurons are less likely to respond to nonaversive stimuli—sucrose, umami, and low salt—compared with the full population. Our data support the presence of a genetically defined neuron type in the geniculate ganglion that is responsive to innately aversive stimuli. This implies that genetic expression might categorize peripheral taste neurons into hedonic groups, rather than simply identifying neurons that respond to a single stimulus.
Significance Statement
Peripheral taste neuron coding has been heavily debated. Our study delves into this issue by leveraging genetic expression in a specific neuron subset to relate peripheral innervation patterns to functional taste responses. We examined a taste neuron type that appears to be in apposition with multiple taste-transducing cell types and responds to innately aversive concentrations of sour, bitter, and high NaCl stimuli. These collective observations suggest that genetic markers can delineate groups of neurons sharing similar hedonic responses rather than categorizing neurons solely based on individual taste qualities.
Introduction
Throughout the nervous system, neurons are categorized into types based on their morphology, function, and genetic expression (Zeng and Sanes, 2017). However, a comprehensive understanding of these characteristics is still lacking for the peripheral taste ganglion neurons responsible for innervating taste buds. Within taste buds, cells transduce chemical information from the oral cavity into electrical signals. These signals are then relayed to taste ganglion neurons, which subsequently transmit them to the brainstem. Historically, taste ganglion neurons have been classified into types according to their physiological response properties (Spector and Travers, 2005; Yarmolinsky et al., 2009). Recent studies have delineated neuron types based on molecular expression (Dvoryanchikov et al., 2017; J. Zhang et al., 2019), and in some instances, variations in molecular expression have been linked to limited functional characterization (J. Zhang et al., 2019). Nevertheless, understanding how these functional and genetic characteristics correlate with the innervation pattern of peripheral taste neurons within taste buds remains elusive.
Taste ganglion neurons are pseudounipolar neurons whose cell bodies reside in the geniculate ganglion. The peripheral axons of taste ganglion neurons can branch a variable number of times before reaching the taste bud, such that the peripheral axon of each taste ganglion neuron has anywhere from 1 to 14 separate taste arbors (T. Huang et al., 2021). These taste arbors also vary in terms of their morphology based on numerous morphological features (Ohman et al., 2023). Because taste-transducing cells die and are replaced, some of these morphological features are altered over time, presumably to connect neurons with new taste-transducing cells (Whiddon et al., 2023). However, it remains a possibility that some features of the taste arbor may reflect the intrinsic features of a neuron type. For example, taste arbors differ based on their apposition to specific types of taste-transducing cells within the taste bud (T. Huang et al., 2021). Some taste arbors are only close to Type 3 cells (transducing sour) or Type 2 cells (transducing sweet/bitter/umami), whereas others come into proximity of both types (Ohman et al., 2023). Given that functional properties of peripheral taste neurons arise from the type(s) of taste-transducing cells innervated, neurons responding to different taste stimuli would be predicted to have differing patterns of innervation.
While it seems reasonable that the functional traits of neurons can predict morphological differences, it is equally plausible that studying neuron morphology could elucidate function. In exploring other peripheral sensory neurons, advances in understanding have been made by investigating the links between peripheral neuronal structure and sensory function (Abraira and Ginty, 2013; Bai et al., 2015). For instance, in touch neurons, the formation of lanceolate endings on one side of a hair follicle led to the hypothesis and subsequent discovery of their directional selectivity (Rutlin et al., 2014). We hypothesize that similar insights might be gleaned by analyzing the innervation patterns of genetically identified gustatory neurons.
To gain genetic access to taste neuron types, RNA-seq analysis was used to divide peripheral taste neurons into groups based on molecular expression (Dvoryanchikov et al., 2017; J. Zhang et al., 2019). In one recent study, each genetically defined cluster mediated a basic taste quality (sweet, sour, etc.; J. Zhang et al., 2019). These findings were important for two reasons: (1) by gaining genetic access to taste ganglion neurons, we can now map the peripheral morphologies and central connections of functionally defined taste neuron groups, and (2) the findings supported a “labeled-lined” concept of taste coding in which neuron types represent a specific taste quality (i.e., sweet, sour, etc.), a notion which has been heavily debated in the literature (Spector and Travers, 2005; Yoshida et al., 2006; Yarmolinsky et al., 2009; Carleton et al., 2010; Ohla et al., 2019).
Our aim was to delineate taste arbor morphologies and innervation patterns of two distinct taste neuron types, which have been characterized both molecularly and functionally (Dvoryanchikov et al., 2017; J. Zhang et al., 2019). Surprisingly, our observations revealed morphological characteristics in these neurons that diverged from what was expected based on prior functional assessments (J. Zhang et al., 2019). This disparity prompted us to reevaluate the function of one of these types of geniculate ganglion neurons. Upon reexamination, we found neurons of this type respond to taste stimuli typically considered aversive (sour and bitter). Our findings suggest that genetic labels might identify groups of taste ganglion neurons that respond to stimuli sharing similar innate hedonic value rather than conforming to the traditional five basic taste categories. Our findings highlight the importance of scrutinizing the anatomical features of genetically defined neuron types as a crucial step in designing functional experiments and interpreting their results.
Materials and Methods
Animals
To genetically label a random subset of peripheral taste neurons, we crossed the TrkBCreER mice (Rutlin et al., 2014; Ntrk2tm3.1(cre/ERT2)Ddg; https://www.jax.org/; catalog #JAX:027214) with Cre-dependent tdTomato mice (Ai14, JAX:007914) to obtain TrkBCreER:tdTomato mice. To label Penk-expressing and Egr2-expressing cells, we bred Penktm2(Cre)Hze mice (JAX:025112) and Egr2tm2(Cre)Pch mice (JAX:025774) with tdTomato (AI14) as described above. In these mice, there is extensive labeling in the tongue. To limit labeling to peripheral taste neurons expressing Penk, Penktm2(Cre)Hze mice were bred with mice containing a Gt(ROSA)26Sor tm65.1(CAG-tdTomato)Hze (Ai65; JAX:021875) construct and a Phox2b-Flpo3276Grds/J (Phox2b-Flpo, JAX:022407) construct. In these mice, tdTomato is both Cre- and Flpo-dependent, which helps to limit the tdTomato expression to Penk+/Phox2b+ neurons. These mice are henceforth referred to as Penk-tdTomato. Lastly, we collected more randomly selected taste arbors from the full population by breeding Cre and Flpo-dependent tdTomato mouse line (Ai65) with an Isl1CreER mouse and a Phox2b-Flpo mouse injected with a low dose of tamoxifen in adulthood. For mice used in calcium imaging experiments, Penktm2(Cre)Hze mice, TrkBCreER mice, and Phox2b-Cre mice (Mutant Mouse Resource and Research Centers; MMRRC strain 034613-UCD, NP91Gsat/Mmcd, RRID: MMRRC_034613-UCD) were bred with a Cre-dependent GCaMP6s mice [Ai96(RCL-GCaMP6s), Jax#028866]. These mice were referred to as Penk-GCaMP6s, TrkB-GCaMP6s, and Phox2b-GCaMP6s, respectively. For live imaging studies, Penk-tdTomato mice were bred with GAD1-GFP (Jax#007677) mice. All experiments used both male and female mice.
Animals were cared for and used in accordance with guidelines of the US Public Health Service Policy on Humane Care and Use of Laboratory Animals and NIH Guide for the Care and Use of Laboratory Animals.
Tamoxifen injections
TrkB-tdTomato animals received 0.5–4.0 mg, and Isl1CreER:Phox2b-flpo:Ai65 mice received 0.01 mg for 1 d at Postnatal Day (P)40, and TrkB-GCaMP6s mice received 4.0 mg daily for 1–2 weeks. Tamoxifen (T-5648, Sigma-Aldrich) was dissolved in corn oil (C-8267, Sigma-Aldrich) at 20 mg/ml by shaking and heating at 42°C and injected at P40 by intragastric gavage. For embryonic injections of TrkB-tdTomato mice, 4-hydroxytamoxifen (4-HOT; H-7904, Sigma-Aldrich), the active metabolite of tamoxifen, was injected intraperitoneally into pregnant dams at Embryonic Day (E)15.5. 4-HOT was prepared and injected as previously reported (H. Wu et al., 2012). Briefly, 4-HOT was dissolved in ethanol at 20 mg/ml by shaking and incubating at 37°C for 15 min and then stored at −20°C. Before use, the stock solution was dissolved in sunflower seed oil to a final concentration of 10 mg/ml 4-HOT, and the ethanol was evaporated by centrifugation under vacuum.
Fluorescent anterograde labeling
The procedures used to label the chorda tympani with fluorescent tracers were previously described (Sun et al., 2015). Briefly, adult Penk-tdTomato or Egr2-tdTomato mice were anesthetized and placed in the head holder as described above. A water-circulating heating pad was used to maintain body temperature. The chorda tympani nerves within the right tympanic bulla were cut near and immediately distal to the geniculate ganglion, and crystals of 3-kDa fluorescein dextran (D3306; Invitrogen) were applied to the proximal cut end of the chorda tympani. A small amount of Kwik-Sil (World Precision Instruments) was then placed over the cut ends of the nerves to prevent crystals from diffusing from the intended labeling site. Postsurgical treatment was the same as described above. After 24 h, mice were killed and perfused with 4% paraformaldehyde (PFA). The geniculate ganglia were dissected and immediately mounted and imaged by confocal microscopy.
Immunohistochemistry
TrkB-tdTomato, Phox2b-tdTomato, Penk-tdTomato, and Egr2-tdTomato mice were killed by avertin overdose (4 mg/kg) and perfused transcardially with 4% PFA. A razor blade was used to remove the circumvallate papilla; the tongues were then carefully split down the midline with a razor blade under a dissection microscope. Dissected tissues were postfixed in 4% PFA overnight and transferred to 30% sucrose at 4°C overnight. Tongues were then frozen in optimal cutting temperature (OCT) and stored at 80°C before sectioning on a cryostat or processing for whole-mount staining (Ohman and Krimm, 2021a).
Geniculate ganglia were rinsed in 0.1 M phosphate buffer (PB) and incubated in blocking solution (3% donkey serum, 0.5% Triton X-100 in 0.1 M PB) at 4°C overnight and incubated for 4 d at 4°C with primary antibodies goat anti-Phox2b (1:200, R&D Systems, catalog #AF4940, RRID:AB_10889846) and rabbit anti-DsRed [1:500, RRID:AB_10013483; Living Colors DsRed polyclonal; Takara Bio USA in antibody solution (0.5% Triton X-100 in 0.1 M PB)]. Tissues were rinsed four times for 15 min each with 0.1 M PB and incubated with secondary antibodies donkey anti-goat 488 (1:200, Jackson ImmunoResearch Laboratories, catalog #705-545-147, RRID:AB_2336933) in antibody solution for 2 d at 4°C. Tissues were then rinsed again, mounted with Fluoromount-G, and coverslipped (high precision, 0107242; Marienfeld).
Half tongues were sectioned at 70 µm sagittally from the midline to lateral edge. Whole mounts of the lingual epithelium were prepared as described previously (Ohman and Krimm, 2021a). Briefly, the lingual epithelium of half tongues was first separated from the underlying musculature using dissection scissors under a dissection microscope. The epithelium was separated into three pieces using dissection scissors once most of the musculature was removed, leaving only a minimal, even layer of musculature/lamina propria on the underside of the epithelium. In preparation for removing any remaining musculature/lamina propria using the cryostat, occasionally, small cuts were made to facilitate laying the epithelium flat in a tissue mold (muscle side down). The epithelium was covered in OCT and frozen. Thin sections (20 µm) were cut and examined under the fluorescent microscope to assess the proximity to the underside of the epithelium. This process removes as much underlying tissue as possible for antibody penetration without removal of any taste buds. A detailed description of this procedure is available in Ohman and Krimm (2021a).
The isolated lingual epithelium was then thawed and washed for 15 min (three times) in 0.1 M PB. Tissues were incubated in blocking solution (3% donkey serum, 0.5% Triton X-100 in 0.1 M PB) at 4°C overnight and incubated for 5 d at 4°C with primary antibodies [phospholipase Cβ -2 (PLCβ2); 1:500, RRID:AB_2630573; Santa Cruz Biotechnology] in antibody solution (0.5% Triton X-100 in 0.1 M PB). Tissues were rinsed four times for 15 min each with 0.1 M PB, incubated with secondary antibodies (1:500, Alexa Fluor 488 AffiniPure, RRID: AB_2340619; Jackson ImmunoResearch Laboratories), rinsed again (four times for 15 min each with 0.1 M PB), and incubated with 5% normal rabbit serum in antibody solution. Tissues were rinsed and incubated with AffiniPure Fab fragment donkey anti-rabbit IgG (20 mg/ml, RRID:AB_ 2340587; Jackson ImmunoResearch Laboratories) in antibody solution, rinsed, and incubated with Zenon Alexa Fluor 555 rabbit IgG labeling kit [according to the instructions for Zenon complex formation (Z25305; Invitrogen)] using anti-DsRed (1:500, RRID:AB_10013483; Living Colors DsRed polyclonal; Takara Bio USA). Tissues were rinsed, incubated for 5 d at 4°C with carbonic anhydrase 4 (Car4) primary antibody (1:500, RRID:AB_10013483; R&D Systems), rinsed, and incubated with secondary antibodies (1:500, Alexa Fluor 647 AffiniPure, RRID:AB_2340438). Tissues were then rinsed again, mounted with Fluoromount-G, and coverslipped (high precision, 0107242; Marienfeld). In cases where all primary antibodies are from different species, primaries (rat anti-Troma-1 and rabbit anti-dsRed) were added in the first primary antibody incubation, and secondaries (donkey anti-rat 488,1:500, Alexa Fluor 488 AffiniPure, RRID: AB_2340684; Jackson ImmunoResearch Laboratories) and donkey anti-rabbit 555 were added during the first secondary incubation. Then the tissues were rinsed and mounted as described above.
Confocal imaging
Taste bud images were obtained using an Olympus FluoView FV1000 confocal laser-scanning microscope with either a 60× NA1.4 lens (taste buds) or 20× (geniculate ganglia) objective using a zoom of 3, Kalman 2. Image sizes were initially set at 1,024 × 1,024 pixels but were cropped to the size of the taste bud to reduce scanning time and photobleaching. Serial optical sections at intervals of 0.47 µm in the Z dimension were captured, which is the optimal resolution in the z-axis with a 60× 1.4 NA lens. All colors were imaged sequentially in separate channels to avoid bleed-through. Image stacks were then deconvolved using the AutoQuant X3 software (Media Cybernetics) to reduce out-of-focus florescence and improve image quality.
Two-photon in vivo imaging
In vivo live imaging of taste buds was accomplished as described previously (Whiddon et al., 2023). Briefly, two-photon laser–scanning microscopy was performed using a Movable Objective Microscope (Sutter Instruments). A Fidelity-2 1,070 nm laser (Coherent) was used to visualize tdTomato, and a Chameleon tunable laser (Coherent) set to 920 nm was used to visualize GFP. The excitation wavelength at the sample ranged from 20 to 51 mW measured at the exit of the microscope objective. A custom-built tongue holder, based on a published design (Choi et al., 2015; Han and Choi, 2018), was used to stabilize the anterior tongue for imaging of the dorsal surface. Mice were anesthetized using a 0.6% isoflurane (Henry Schein) and O2 mixture. A temperature controller was used to monitor and maintain body temperature at 36°C. Ophthalmic lubricant ointment (Henry Schein) was applied to the eyes. Taste bud maps were acquired using a 10 × 0.3 numerical aperture water immersion objective lens (Carl Zeiss), and images of taste arbors were acquired using a 40 × 1.0 numerical aperture water immersion lens (Carl Zeiss). An early version of ScanImage (MATLAB) was used to collect images (Pologruto et al., 2003). Taste arbors were typically imaged to 70 microns in depth below the surface of the tongue, and 80 × 80 µm optical sections (512 × 512 pixels) were collected at 0.5 or 1 µm increments.
Calcium imaging
Mice were sedated with a 0.32 mg/kg injection of Domitor (medetomidine hydrochloride) intramuscularly and anesthetized with 40 mg/kg Ketaset (ketamine hydrochloride). The heart rate and blood oxygen saturation were monitored continuously with a pulse oximeter. Puralube was applied to the eyes to prevent drying while maintaining optical clarity. A rectal temperature probe was gently inserted to monitor and control the mouse temperature core of ∼35–36°C, via a heating pad that is automatically adjusted based on the animal's core temperature throughout the surgical procedures and recording sessions. The mouse's trachea was cannulated to expedite oral stimulation during the recording and/or to deliver isoflurane. The esophagus was exposed with blunt dissection, and a small-diameter cannula was inserted into a small slit in the esophagus and fed the cannula forward to the oral cavity. This cannula serves as the delivery system for orosensory stimuli as reported previously (A. Wu et al., 2015). Once the trachea and esophageal tubes are placed in the animal, the animal was flipped over to surgically access the geniculate ganglion. At that point, anesthesia was switched to isoflurane inhalation anesthesia 1–3% for maintenance, delivered by a vaporizer.
The geniculate ganglion was exposed with a dorsal–lateral approach. Briefly, with the skull of the mouse firmly attached to a stereotactic head holder bar, the musculature overlying the zygomatic arch and the coronoid process of the mandible is removed with a cautery tool. We then drilled and removed the zygomatic arch and the coronoid process of the mandible to expose the cranium. After cleaning the cranial surface of the adhering muscle and connective tissue, we carefully cut a rectangular window with a microdrill, we removed the cranial piece, and we aspirated a portion of the cortex overlying the geniculate ganglion. This approach has been used previously for the trigeminal ganglion (Leijon et al., 2019). By tilting the head at an angle with the nose of the mouse more elevated the back of the head, the geniculate ganglion and the greater superficial petrosal (GSP) nerve are exposed. Once the geniculate ganglion was exposed, the preparation was transferred to our custom epifluorescence microscope and positioned under a 10×, 0.5 NA long-working-distance objective. The cell bodies of peripheral taste neurons located in the geniculate ganglion were imaged for Ca2+ responses during the automated stimulation of the tongue. Artificial saliva was perfused constantly over the tongue surface. For 10 s of every minute, a taste stimulus was infused over the tongue surface. The stimuli used were selected based on J. Zhang et al. (2019) and included low salt (60 mM NaCl), high salt (500 mM NaCl), sour (50 mM citric acid), bitter (5 mM quinine and 30 µM cyclohexamide), sweet (500 mM sucrose), umami (300 mM monosodium glutamate + IMP), and AceK (30 mM).
Video recordings were performed during all trials, using a scientific complementary metal–oxide–semiconductor video camera (pco.panda 4.2) at 2 fps. This system is like that used in other published imaging studies of the geniculate ganglion (Lee et al., 2017; J. Zhang et al., 2019).
Experimental design and statistical analysis
Quantification of geniculate ganglion neurons
To quantify the number of geniculate ganglion neurons, serial optical sections of whole geniculate ganglia of Penk-tdTomato mice combined with either Phox2b+ labeling (n = 3) or chorda tympani nerve labeling (n = 3) were captured using a confocal microscope (FV1200, Olympus). High-resolution images of whole-mount geniculate ganglia were obtained by stitching multiple fields under a 40× lens, yielding a high-magnification image including the entire ganglion in one z-stack. Each labeled signal was collected individually with specific wavelength excitation. Images of whole geniculate ganglia were analyzed using Neurolucida 10 (MBF Bioscience). All labeled neurons were counted by examining each cell through multiple optical sections; ganglia were imaged with a step size of 1 μm so that each cell could be followed through several sections. To ensure that each neuron was counted only once, we added markers to the center of each counted cell in the z-stack. Data are presented as mean ± SD.
Quantification of innervated taste buds
Thick (70 µm) sagittal floating sections of Penk-tdTomato (n = 3) and Egr2-tdTomato (n = 4) tongues were cut on the cryostat. Sections were stained with Troma1 (Keratin-8) and dsRed as described above. Sections were mounted, and all taste buds were examined at 20× under the fluorescent microscope to determine the number of fungiform taste buds that were innervated by tdTomato+ arbors and the number that were not. Percentage values were calculated for each animal.
Quantification of Car4+ cells and Penk-tdTomato arbors in all fungiform taste buds
All fungiform taste buds in Penk-tdTomato (n = 3) half-tongue whole mounts were examined systematically under the confocal microscope using a 60× objective. Each taste bud was examined using the FluoView software to determine the number of Car4+ cells and the number of tdTomato+ arbors entering each taste bud.
Confocal image analysis
Taste arbors were reconstructed using Neurolucida, and the proximity of taste arbors to taste-transducing cells was measured in Imaris. A detailed explanation of these methods has been provided previously (T. Huang et al., 2021; Ohman and Krimm, 2021a). Briefly, taste arbors were reconstructed in Neurolucida 360 starting at the base of the taste bud. Reconstructions were completed in Isl1CreER:Phox2b-flpo:Ai65 mice (nine arbors in three mice), TrkBCreER:Ai14 injected at P40 (138 arbors from 22 mice, some of which were reported previously; Ohman and Krimm, 2021a), TrkBCreER:Ai14 injected at E15.5 (10 arbors from four mice), and Penk-Cre:Phox2b-flpo:Ai65 (42 arbors from five mice). Once the reconstruction was complete, it was exported to Neurolucida Explorer to obtain quantitative measurements of taste arbor features. To obtain proximity information for the same taste arbors, we imported the deconvoluted image stacks into Imaris. Proximity was determined using a distance transformation algorithm that calculates the distance between an object and any/all points in a defined 3D space. This was accomplished using automated thresholding, which identified the surface of labeled objects (taste-transducing cells and taste arbors) and determined the distance between them in voxel increments. When the distance was zero, then the nerve fiber was in direct apposition to the cell at the light level (within 110 nM). To display the relationship between a single taste arbor and labeled taste-transducing cell, we segmented the individual taste-transducing cells and taste arbors in Imaris, and we only duplicated the fluorescent channel within the selected region. Segmentation was completed after analysis for illustrative purposes only.
Calcium image analysis
Image stacks were corrected for movement artifacts using the TurboReg image plugin in ImageJ (A. Wu et al., 2015; Leijon et al., 2019). ROIs were selected in ImageJ using a magic wand (A. Wu et al., 2015; Ghitani et al., 2017; Leijon et al., 2019). Custom MATLAB scripts (MathWorks) made publicly available (https://github.com/Dalston817/calciumImagingROITools) were used to calculate the relative change in fluorescence (ΔF/F0; A. Wu et al., 2015; Leijon et al., 2019). The fluorescence signals of a C-shaped area around most sides of each ROI was subtracted from the cell response to remove the potential contributions of out of focus signals for each ROI frame (Yarmolinsky et al., 2016; Ghitani et al., 2017). Peripheral taste neurons were considered responsive to a given stimulus if ΔF/F0 is greater than five times the standard deviation (SD; Leijon et al., 2019) of the 20 s baseline preceding the stimulus. The camera recorded all responses for the duration of the experiment. Since the absolute level of fluorescence can be influenced by the location of the cell in the ganglion or level of GCaMP6s expression for the neuron, the largest response for each neuron was set to 1.00, and other responses were reported relative to the maximal response. ROIs for individual neurons were analyzed from Penk-GCamP6s (87 from 17 mice), TrkB-GCaMP6s mice (21 from two mice), and Phox2b-GCaMP6s mice (194 from four mice).
There were no differences in the responses between TrkB+ neurons and Phox2b neurons for any stimulus (Mann–Whitney: quinine, U = 1,610; p = 0.11; citric acid, U = 1,592; p = 0.10; 60 mM NaCl, U = 2,009; p = 0.92; 500 mM NaCl, U = 1,983; p = 0.84; sucrose, U = 1,511; p = 0.052; umami, U = 1,616; p = 0.12), so data from these animals were combined to examine the full population. In Phox2b-GCaMP6s mice where the datasets from each animal were large enough to be representative of the pooled data, there were neurons from each mouse that responded maximally to each stimulus used in the study (except for 60 mM NaCl where the responses were consistently lower than for 500 mM NaCl).
Statistical analysis
Proportional data were compared between the full population of gustatory neurons and Penk+ neurons using either a χ2 test (more than two categories) or a Fisher’s exact test (two categories). Anatomical data were initially examined to determine if they were normally distributed using a Shapiro–Wilk test. Most of our measures were not normally distributed, so statistical comparisons across multiple groups were compared using a Kruskal–Wallis test, while two groups were compared using a Mann–Whitney test. When multiple Mann–Whitney tests were performed to compare specific anatomical features of the taste arbor, a Bonferroni's correction was used to adjust the alpha level. The base alpha level was set at p = 0.05, and actual p values are reported. Because most of our measures were not normally distributed, Spearman’s correlations were used to analyze the relationship between variables. Differences between groups in the normalized ΔF/F were compared for each stimulus using a Mann–Whitney test, with the number of comparisons controlled using a Bonferroni’s correction (p < 0.0083 for six tests); actual p values were reported.
Results
Penk-tdTomato expression is limited to a subset of taste neurons that innervate the tongue and palate
Penk expression identifies a subset of taste neurons in the geniculate ganglion (Dvoryanchikov et al., 2017; J. Zhang et al., 2019). To characterize the peripheral anatomy of this neuron subset, we needed a transgenic animal that would provide genetic access to this neuron population. Initially, we examined sections of the tongue from a PenkCre mouse bred with a Cre-dependent reporter line (Ai14) and determined that Penk expression is not unique to peripheral taste neurons in the tongue but was also expressed in many non-neuronal cells (Fig. 1A). Therefore, to limit tdTomato expression to taste neurons, we employed intersectional genetics (Madisen et al., 2015). Because Phox2b expression is specific to visceral sensory and parasympathetic neurons and absent from non-neuronal cell types (Ohman-Gault et al., 2017), we bred Phox2b-Flpo mice with PenkCre mice followed by tdTomato reporter line that is both Flpo and Cre dependent (henceforth referred to as Penk-tdTomato). In the tongues of these mice, tdTomato labeling of non-neuronal cells in the tongue was eliminated with labeling limited to a subset of peripheral neurons (Fig. 1B). To determine the number of Penk-tdTomato geniculate neurons, we quantified tdTomato+ and oral cavity innervating neurons (Phox2b+), in the geniculate ganglion of Penk-tdTomato mice (Fig. 1C–G). We found that 16% of the Phox2b+ geniculate ganglion neurons were tdTomato+ confirming that Penk expression is limited to a subset of Phox2b+ neurons of roughly the same size as reported previously (Dvoryanchikov et al., 2017). No neurons were found to be Penk-tdTomato and Phox2b− confirming that Penk expression in the geniculate ganglion is restricted to oral cavity innervating geniculate neurons by the Phox2b-Flpo construct.
Clear visualization of Penk+ taste neurons requires an intersectional approach. A, A sagittal section of the tongue from a Penk-tdTomato mouse, showing extensive tdTomato labeling (magenta). Taste buds are labeled with anti-keratin-8 (green). B, A sagittal section of the tongue from a mouse in which tdTomato expression is both Flpo and Cre dependent, limiting tdTomato to neurons expressing both Penk and Flpo (abbreviated Penk-tdTomato). Because tdTomato is no longer expressed in many cells within the tongue muscle and lamina propria, tdTomato+ innervation entering taste buds is more easily seen. C, Whole-mount geniculate ganglion from a Penk-tdTomato mouse (magenta), immunostained for Phox2b (green). D, A single optical section of the geniculate ganglion shown in C. All tdTomato+ neurons are also Phox2b+, but many Phox2b+ neurons are tdTomato−. E–G, Higher magnification of the white box in D, showing Phox2b+ alone (E), tdTomato alone (F), and both together (G). Phox2b+/tdTomato− (arrowheads) and Phox2b+/tdTomato+ (arrows) were quantified separately, revealing that 16% of geniculate neurons innervating the tongue are tdTomato+. Scale bars: A, 40 µm; B, 50 µm; C, 30 µm (also applies to D); E, 10 µm (also applies to F, G).
Peripheral taste neurons in the geniculate ganglion innervate taste buds in the tongue via the chorda tympani nerve or the soft palate via the GSP nerve. We sought to determine whether Penk-tdTomato neurons innervate both the tongue and the soft palate or just one of these regions. To compare the number of neurons innervating the tongue with those that innervating the palate, we used a chorda tympani nerve label to specifically identify neurons that innervate the tongue (Fig. 2). We found that 37% of Penk-tdTomato neurons innervate the tongue via the chorda tympani nerve such that the remaining 63% should innervate the soft palate (Fig. 2A–E). Given that some Penk+ neurons innervate the tongue and others innervate the soft palate, it appears that genetic types are not defined by which region of the oral cavity they innervate. We then asked whether the taste system is organized such that each taste bud is innervated by neurons of every genetically defined type (Dvoryanchikov et al., 2017; J. Zhang et al., 2019). To address this question, we first quantified the number of fungiform taste buds (located in the front two-thirds of the tongue) that had Penk-tdTomato taste arbors (Fig. 2F,G). Of the 189 taste buds examined in three mice, 61% (117 taste buds) had Penk-tdTomato arbors. This indicates that not all fungiform taste buds are innervated by every genetically defined neuron type.
More Penk+ neurons project to the palate than to the tongue. A, Whole-mount geniculate ganglion from a Penk-tdTomato mouse (magenta), with the chorda tympani nerve labeled (fluorescein, green). B, Single optical section from the geniculate ganglion in A. C–E, Higher magnification of the white box in B, illustrating the quantification of Penk-tdTomato neurons that project to the tongue and palate. Some Penk+ taste neurons are double-labeled with fluorescein from the chorda tympani nerve (arrows), indicating they project to the tongue. Penk+ taste neurons that lack fluorescein label project to the palate (arrowheads). F, G, Sagittal sections of Penk-tdTomato tongues revealed fungiform taste buds (keratin-8, green) innervated by tdTomato+ taste arbors (F, magenta) and those without innervation (G). H, Whole mount of the soft palate with a taste bud containing Penk-tdTomato innervation and one without. Scale bars: A, 30 µm (also applies to B); C, 10 µm (also applies to D, E); F, 4 µm (also applies to G); H, 10 µm.
Since a larger portion of Penk+ neurons innervated the palate than the tongue, we hypothesized that a greater percentage of palatal taste buds would be innervated compared with those on the tongue. To examine this possibility, we quantified the number of taste buds innervated by Penk-tdTomato taste arbors in three whole-mount preparations of the soft palate (Fig. 2H). Of the 316 palatal taste buds examined in three mice, 308 (97%) were innervated by Penk-tdTomato neurons. Indeed, a greater proportion of taste buds on the palate are innervated by Penk-tdTomato neurons compared with those on the tongue (Fisher's exact test, p < 0.0001).
Egr2+ neurons are present in the brainstem, but not in the geniculate ganglion
We aimed to compare the innervation patterns of Penk-tdTomato neurons with those of a second neuron type. Egr2+ neurons were selected because they had been described as salt sensitive (J. Zhang et al., 2019). Because amiloride-sensitive salt responses are primarily transduced by taste buds in the anterior tongue, we hypothesized that in contrast to Penk-tdTomato neurons, more Egr2+ neurons would innervate the tongue than the palate. However, although we used the same Cre+ mouse line used by J. Zhang et al. (2019), we observed no neurons in the geniculate ganglion that underwent gene recombination in Egr2Cre mice bred with a Cre-dependent tdTomato (Ai14, Fig. 3A). Instead, we found many tdTomato+ cells running along the nerve which could be Schwann cells or fibroblasts or both (Fig. 3B). We next examined the tongue and observed abundant tdTomato+ expression in the lamina propria and muscle (Fig. 3C). Much of this labeling was found in locations where axon bundles would be located (Fig. 3D), consistent with the finding that Egr2 is expressed in neural crest boundary cap cells that contribute glia to the peripheral nervous system (Maro et al., 2004). However, after examining all taste buds in three Egr2-tdTomato mice, we did not observe any tdTomato+ taste arbors penetrating the taste bud (Fig. 3E,F). We conclude that gene recombination does not occur in the peripheral taste neurons of Egr2Cre mice, making them unsuitable for examining peripheral taste neuron types.
Egr2Cre mice do not show gene recombination in geniculate neurons. A, A geniculate ganglion from an Egr2-tdTomato mouse, also labeled with Phox2b (green). No tdTomato+ geniculate ganglion neurons were present; only nerve bundles containing tdTomato-positive cells could be seen. B, Higher magnification of a single optical section from the ganglion in A. Arrows indicate the GSP nerve, which contains tdTomato+ cells. C, A tongue section showing many tdTomato+ cells within the tongue. Arrows indicate keratin-8+ taste buds (green). D, Higher magnification of fungiform papillae illustrates that tdTomato+ cell density in the connective tissue core of the fungiform papilla. E, F, These fusiform cells extend a process into the bottom of the taste bud; however, no tdTomato+ taste arbors are seen within the taste bud (dashed line). G, A horizontal section through the brainstem illustrates two stripes of tdTomato+ cells rostral to the NTS. H, Chorda tympani nerve labeling reveals the location of the chorda tympani terminal field (green). Egr2+ neurons are scattered throughout the brainstem, with many labeled neurons in the lateral NTS (arrows). I, Some labeled Egr2+ neurons were present in the densely labeled chorda tympani terminal field (arrowheads). Directions for rostral (R) and medial (M) are indicated with arrows. However, labeled neurons are also found in the high-density (taste) region of the NTS (arrowheads). Scale bars: A, 50 µm; B, 20 µm; C, 50 µm; D, 20 µm; E, 20 µm (also applies to F); G, 1 mm; H, 200 µm; I, 100 µm.
Egr2 was identified as a marker for salt-sensitive neurons by genetically silencing activity in these neurons with a Cre-dependent viral construct expressing tetanus toxin (J. Zhang et al., 2019). Since viral injections were targeted to the first synaptic relay for taste information in the brainstem, the rostral nucleus of the solitary tract (rNTS), the function of Egr2+ neurons in the rNTS could also have been disrupted. Since we found no Egr2-tdTomato geniculate ganglion neurons, we speculated that maybe the brainstem contained some Egr2-tdTomato neurons. To determine the proximity between Egr2+ neurons in the brainstem and the gustatory neuron receptive field, we labeled the brainstem terminal fields with anti-P2X2 or with fluorescein dextran (via chorda tympani nerve label) in Egr2-tdTomato mice. We found two stripes of Egr2-tdTomato cells in the brainstem (Fig. 3G). This was not unexpected, considering that Egr2 is a transcription factor, also known as Krox-20. This hox gene plays a crucial role in regulating the development of rhombomeres 3 and 5 in the brainstem (Schneider-Maunoury et al., 1993). Additional Egr2-tdTomato neurons were scattered throughout the brainstem, including in the rNTS (Fig. 3H,I). While there were a few neurons within the dense label of the chorda tympani terminal field (Fig. 3I, green label, arrowheads), most Egr2-tdTomato cells in the rNTS were just lateral to the dense chorda tympani label, corresponding to the lateral rNTS (Fig. 3H, arrows). All these neurons would likely be within the injection site of any virus injected in the gustatory brainstem potentially explaining the previous behavioral findings using the Egr2Cre line (J. Zhang et al., 2019). We concluded that Erg2Cre mice cannot be used to genetically identify salt-specific taste–responsive neurons in the geniculate ganglion. Therefore, we did not examine Erg2Cre mice further.
Penk-tdTomato taste arbors are not morphologically distinct
We aimed to investigate if the peripheral innervation patterns of the taste axon segment that innervates the taste bud, known as the taste arbor (T. Huang et al., 2021; Ohman et al., 2023), differ among various molecularly defined neuron types (Fig. 4A). Since only one of the two molecular types selected for examination was expressed in geniculate ganglion neurons, we compared Penk-tdTomato taste arbors to a randomly selected subset of taste arbors from the full population. For Penk-tdTomato arbors, we took advantage of the relatively sparse innervation to fungiform taste buds as compared with the soft palate, allowing us to trace individual arbors. The full-population arbors were selected through sparse labeling primarily using TrkB-tdTomato mice (T. Huang et al., 2021). It should be noted that although all geniculate neurons express either the full-length TrkB receptor or the truncated versions (Tang et al., 2017), only 75% of Phox2b+ geniculate neurons undergo gene recombination in adult TrkBCreER mice with a high dose of tamoxifen (T. Huang et al., 2021), so some gustatory neuron subpopulations could be missing from this group. However, TrkB-tdTomato neurons would include neurons not present when only Penk+ neurons are labeled (16%). If taste arbor morphology differs based on neuron type, Penk-tdTomato taste neurons would not be expected to exhibit the full range of possibilities represented in the full population. Visual examination of taste arbors from the full population compared with Penk-tdTomato taste arbors did not reveal any obvious anatomical differences (Fig. 4B,C). They did not differ in complexity (terminal branch number; Fig. 4D; Mann–Whitney; U = 3,080; p = 0.28) and taste arbor size [taste arbor length (Fig. 4E; Mann–Whitney; U = 3,025; p = 0.22) or convex hull (Fig. 4F; Mann–Whitney; U = 2,972; p = 0.3173)]. Unlike the peripheral endings of somatosensory neurons (H. Wu et al., 2012), taste neuron types cannot be readily classified based on the morphology of their terminal endings (i.e., taste arbors).
Penk-tdTomato taste arbors are not distinct from the full population in complexity or size. A, A schematic illustrating the quantification of taste arbor complexity (number of terminal branches, blue arrowheads) and size (arbor length, blue lines, and convex hull, yellow polygon). Each measurement was compared between Penk-tdTomato taste arbors and the full population. B, Example of a taste arbor from the full population. C, Example of a taste arbor from the Penk+ population. The distribution of these measurements for Penk-tdTomato taste arbors (n = 42) was compared with the distribution of arbors from the full population (TrkB-tdTomato, n = 148; Phoxb-tdTomato arbors, n = 9). D–F, Penk-tdTomato taste arbors were not different from the full population in terms of complexity (D) or size (E, F). Scale bars: B, 4 µm; C, 3 µm.
Penk-tdTomato taste arbors are in apposition with multiple types of taste-transducing cells
Taste arbor structure might be expected to be dictated more by plasticity than neuron type; however, stimulus sensitivity is determined by the type of taste-transducing cell with which a neuron connects. Since neuron types are thought to correspond to the five basic taste qualities (J. Zhang et al., 2019), neuron type might be expected to determine which type of taste-transducing cell a given neuron innervates. Sour is transduced by cells that express Car4 (Chandrashekar et al., 2009) and not by cells expressing PLCβ2 (Clapp et al., 2006). The number of Car4+ cells in a fungiform taste bud varies from 0 to 6, resulting in an average of 2.8 Car4+ cells per fungiform taste bud (Ohtubo and Yoshii, 2011; Meng et al., 2015; Biggs et al., 2016). Given that Penk-tdTomato neurons respond to sour stimuli (J. Zhang et al., 2019), we hypothesized that these neurons would preferentially innervate taste buds containing Car4+ cells. To test this hypothesis, we quantified the number of Penk-tdTomato taste arbors and Car4+ cells within 140 taste buds from three Penk-tdTomato mice (Fig. 5A–C, Ohman and Krimm, 2021b). Consistent with our findings from sections, 60% of fungiform taste buds are innervated by Penk-tdTomato taste arbors. We found that there was no correlation between the number of Car4+ cells in a taste bud and the number of Penk-tdTomato taste arbors entering the taste bud (Fig. 5F; r = 0.038; p = 0.66). Surprisingly, the taste buds innervated by Penk-tdTomato taste arbors included both taste buds with and without Car4+ (sour) cells (Fig. 5D,E). Only 59% of taste buds with Car4+ cells were innervated by Penk-tdTomato taste arbors, and 41% of taste buds with Car4+ cells were not innervated by Penk-tdTomato taste arbors (Fig. 5D). Taste buds without Car4+ cells were also innervated by Penk-tdTomato taste arbors (62%; Fig. 5E). Contrary to our hypothesis, Penk-tdTomato taste arbors innervate taste buds lacking Car4+ (sour) taste-transducing cells, and the number of Penk-tdTomato taste arbors and Car4+ cells within a taste bud was not correlated (Fig. 5F).
Penk-tdTomato taste arbors innervate taste buds with and without Type 3 (sour-transducing) cells. A, Representative image of a taste bud showing how the number of Car4+ cells (blue) and the number of Penk-tdTomato taste arbors (magenta, white arrows) were quantified in whole-mount preparations of the lingual epithelium. B, An individual optical section through the taste bud in A (yellow line) showing two Car4+ cells (blue, white arrowheads). C, Optical section through the taste bud in A (white line) illustrates the two Penk-tdTomato taste arbors entering the taste bud (white arrows). D, Example of a taste bud containing Car4+ cells but lacking Penk-tdTomato taste arbors. E, Example of a taste bud lacking Car4+ cells but innervated by Penk-tdTomato taste arbors. F, All taste buds on one-half of the tongue from three mice were included in the quantification, totaling 139 taste buds. No correlation was found between the number of tdTomato+ taste arbors in a taste bud and the number of Car4+ cells. Scale bars: A, B, 4 µm; C, 2 µm; D, 3 µm (also applies to E).
The lack of any relationship between the sour-transducing cells in the taste bud and the number of taste arbors from sour-responsive neurons (Penk-tdTomato) was surprising. We speculated that this might be explained by the plasticity within the taste system. Taste bud cells die and are replaced (Beidler and Smallman, 1965; Perea-Martinez et al., 2013; Barlow, 2015), and each taste bud represents a snapshot in time in a fixed animal. It is possible that taste buds containing sour-transducing cells but no Penk-tdTomato taste arbors were about to acquire new Penk-tdTomato taste arbors or had recently lost them. Similarly, taste buds lacking sour-transducing cells, but containing Penk-tdTomato taste arbors, might add a new sour transducing cell to connect with the available Penk-tdTomato taste arbors. To determine if this was the case, we performed an intravital imaging experiment in Penk-tdTomato mice, where the sour transducing cells, Type 3 cells, were labeled with GFP in GAD67GFP mice (Tomchik et al., 2007). The markers for Type 3 (sour-transducing) cells within the taste bud, Car4 and GAD67GFP, colocalize in mouse fungiform taste buds (Lossow et al., 2017). We imaged a total of 75 taste buds in five mice over 10 d (Fig. 6), which is the average half-life of a taste cell (Beidler and Smallman, 1965). We found that 45 (60%) contained both Penk-tdTomato taste arbors and GFP-labeled cells for all 10 d (Fig. 6E–H). We observed that 12 taste buds contained GFP-labeled cells but lacked Penk-tdTomato arbors (Fig. 6A–D); none of these taste buds added a Penk-tdTomato taste arbor over 10 d. We also found that 18 taste buds had Penk-tdTomato taste arbors on Day 1 but lacked GAD67+ taste-transducing cells (Fig. 6I,J,M,N), of these, half added at least one GFP-positive cell within 10 d (Fig. 6O,P), while the others did not (Fig. 6J,L).
Intravital imaging reveals that the poor relationship between GAD67+ (Type 3) cells and Penk+ innervation persists over time. A–D, A taste bud with two GAD67+ taste-transducing cells and no Penk-tdTomato taste arbors on the first day of imaging; 264 h later, one GAD67+ taste-transducing cell was lost, but no taste arbors were added to the taste bud. E–H, A taste bud with Penk+ innervation and a GAD67+ taste-transducing cell (arrow) on the first day of imaging (T0) and 240 h later. I–L, A taste bud with Penk+ innervation and no GAD67+ taste-transducing cells at T0, with none added over 240 h. M, N, A taste bud with Penk-tdTomato taste arbors with no GAD67+ cells at T0, with one cell added by 240 h later. Scale bar: A–P, 5 µm.
In summary, taste buds containing GAD67+ cells but lacking Penk-tdTomato taste arbors demonstrate that there are GAD67+ cells that either form connections with neurons not expressing tdTomato or lack neural connections altogether. Conversely, Penk-tdTomato taste arbors found in taste buds devoid of GAD67+ cells either establish connections with cells of a different type or persist for extended periods without receiving input from any taste-transducing cell. Collectively these findings suggest no correlation between GAD67+ Type 3 cells, known for transducing sour stimuli, and the innervation by Penk-tdTomato taste arbors, which have been shown to respond to sour taste stimuli (J. Zhang et al., 2019).
We next sought to examine the proximity between sour-transducing cells and Penk-tdTomato taste arbors in taste buds containing both, so we analyzed the proximity between the fungiform arbors reconstructed in Figure 4 and PLCβ2+ and Car4+ taste bud cells. For neurons to form a synapse or synapse-like connection with a taste-transducing cell, the taste arbor must come sufficiently close, resulting in overlapping fluorescence at the light level (Fig. 7A); we have termed these locations “appositions” (T. Huang et al., 2021). However, not all appositions are indicative of a synapse. Consequently, 28% of taste arbors are in apposition to taste-transducing cells of multiple types, even though only a few have synapses with multiple types (4%; Wilson et al., 2022). As such, arbors from all neuron types would be expected to be in apposition with cells of multiple types. However, since it was reported that Penk-tdTomato neurons only respond to sour stimuli (J. Zhang et al., 2019) and many taste arbors are only in apposition to a single taste-transducing cell (Ohman et al., 2023), we hypothesized that more Penk-tdTomato taste arbors would be in apposition to a Car4+ (Type 3, sour) taste-transducing cells compared with the full population. To test this idea, we analyzed the number and type of cells within 110 nm of 45 Penk-tdTomato taste arbors and 157 taste arbors from the full population defined by sparse cell labeling (TrkBCreER; tdTomato mice and Islet1CreER:Phox2bFlpo:tdTomato mice). Contrary to our hypothesis, there was no significant difference in the number of Car4+ cells (Fig. 7B; Mann–Whitney; U = 3,052; p = 0.14) or PLCβ2+ cells (Fig. 7C; Mann–Whitney; U = 3,418; p = 0.73) apposed to Penk-tdTomato taste arbors compared with taste arbors from the full population.
Penk-tdTomato arbors contact both Car4 (Type 3) and PLCβ2+ (Type 2) taste-transducing cell types more frequently than the full population of taste arbors. A, A TrkB-tdTomato arbor in a fungiform taste bud contacting a single PLCβ2+ cell (green). A Penk-tdTomato taste arbor in a fungiform taste bud contacting a single Car4+ cell (blue). White areas define regions of overlapping fluorescence, indicating that the two cell types are within 180 nm of each other. There is no significant difference in the number of Car4+ (B) or PLCβ2+ (C) taste-transducing cells contacted by Penk-tdTomato taste arbors compared with taste arbors from the full population. D, The distance between arbors adjacent to Car4+ cells and (E) PLCβ2+ cells is measured (arrows in A and B). F, The percentage of Penk-tdTomato taste arbors that contact no cells, Car4+ cells, PLCβ2+ cells, and both types of taste-transducing cells compared with taste arbors from the full population of peripheral taste neurons (χ2 = 13.62; p = 0.0035). *p < 0.05.
Some taste arbors come into proximity with a taste-transducing cell for only a short distance (Fig. 7A, left, arrows), while others project alongside a taste-transducing cell for a considerable distance (Fig. 7A, right, arrows). One possibility we considered was that Penk-tdTomato taste arbors might extend a significantly greater distance alongside Car4+ (sour) taste-transducing cells compared with the full population. However, our findings showed no significant difference in this measure (Mann–Whitney; U = 1,500; p = 0.49; Fig. 7D). We did find that Penk-tdTomato taste arbors followed PLCβ2 (sweet, bitter, and umami) taste-transducing cells for a shorter distance compared with the full population (Mann–Whitney; U = 924; p = 0.002; Fig. 7E). This finding might indicate less input from PLCβ2 cells to Penk+ neurons compared with the full population.
While quantifying the number of taste arbors in close proximity to each cell type, we noticed that many Penk-tdTomato taste arbors appeared to be in apposition to both PLCβ2+ cells and Car4+ cells. To examine this more closely, we calculated the proportion of taste arbors that appose only Car4+ cells, only PLCβ2+ cells, and those that appose both cell types (Fig. 7F). We found that this pattern of cellular innervation differed between Penk+ neurons and the full population (χ2 = 13.62; p = 0.0035). Specifically, more Penk+ arbors are in apposition to both cell types and fewer to only one cell type compared with the full population of arbors (Fisher's exact test; p = 0.0084).
Penk-GCaMP6s gustatory neurons respond preferentially to innately aversive stimuli
Our anatomical findings were surprising given the previous report that Penk-tdTomato neurons only respond to sour stimuli (J. Zhang et al., 2019). Instead, our results suggest that Penk neurons may respond to sour and another stimulus transduced by the PLCβ2 cell type (either sweet, bitter, or umami). To examine this possibility, we sought to repeat the study examining Penk-GCaMP6s response characteristics (J. Zhang et al., 2019) but with some important modifications to the stimulus array to better test for responses to bitter, sweet, and umami stimuli. We used identical concentrations of citric acid (50 mM), the artificial sweetener, AceK (30 mM), and salt (60 mM NaCl). However, because mice have 35 different bitter receptors and not all bitter responsive taste-transducing cells respond to every bitter stimulus (Lossow et al., 2016; Yoshida et al., 2018), we used a bitter mixture (5 mM quinine and 30 µM cycloheximide; A. Wu et al., 2015) instead of 5 mM quinine (Y. V. Zhang et al., 2016). Furthermore, we added an umami stimulus (300 mM monosodium glutamate + IMP) and a sugar (500 mM sucrose), neither of which were examined previously. Lastly, because 60 mM is a low concentration of NaCl, we also included high concentration of NaCl (500 mM). Since the citric acid concentration used by Y. V. Zhang et al. (2016) was the same as the “high concentration” of citric acid used by (50 mM) A. Wu et al. (2015), we selected all additional stimuli from the “high concentrations” of A. Wu et al. (2015) for comparison.
Responses were measured in a similar manner as previous studies (Fig. 8A,B; Barretto et al., 2015; Ghitani et al., 2017; J. Zhang et al., 2019) and were consistent across repeated stimulations (Fig. 8C). Most neurons that responded to AceK also responded to sucrose, and all neurons that responded to a low concentration of NaCl (60 mM) also responded to a high concentration of NaCl (500 mM). Therefore, neurons that responded to both sweet stimuli or both NaCl concentrations were categorized as responding to a single taste quality. Of the 66 neurons responding to a single taste quality from the full population, 30 responded to sweet, 18 to salt, 12 to bitter, 3 to umami, and 3 to citric acid. Of the 12 neurons from the Penk+ population that responded to a single quality, 6 responded to bitter, 4 to NaCl, and 2 to citric acid. The distribution in number of taste qualities that the Penk-GCaMP6s population of neurons responded to differed from the full population [χ2 (df = 4) = 10.93; p = 0.027; Fig. 8D,E]. Specifically, a smaller proportion of Penk-GCaMP6s neurons respond to only one taste quality as compared with the full population of taste neurons (Fisher's exact test; p = 0.0014). This finding aligns with a large body of evidence that describe neurons responding to sour as broadly tuned (Pfaffmann, 1941; Frank, 2000; Frank et al., 2008; Breza and Contreras, 2012; Barretto et al., 2015). The proportion of neurons responding to each stimulus was also different between the full population and Penk-GCaMP6s neurons [χ2 (df = 6) = 87.7; p = 0.0001; Fig. 9A]. We observed that more Penk neurons responded to citric acid, quinine, and a high concentration of NaCl; these stimuli are all innately aversive when compared with water. All Penk+ neurons responded to at least one of these three stimuli. We compared the proportion of neurons that responded to innately aversive stimuli (citric acid, quinine, high salt) versus nonaversive stimuli (sweet, umami, and low salt) and found this to be different between Penk+ neurons compared with the full population [Fisher's exact test (p = 0.0001); Fig. 9B].
Penk-GCaMP6s neurons tended to be more broadly tuned than the full population of peripheral taste neurons. A, B, Images showing raw calcium responses of two neurons that respond to citric acid before (A) and during (B) the stimulus. Blue circles illustrate the regions of interest (ROIs) used for collecting calcium responses, while red half circles illustrate ROIs used to collect background responses, which were subtracted for each neuron at each time point. C, The responses of a single neuron to three separate presentations of the same concentration of citric acid, illustrating response consistency. D, Example responses for five individual Penk+ neurons and five individual neurons from the full population (E). Both genotypes included neurons that were narrowly (responded to one taste quality) or more broadly tuned (responded to more than one taste quality). No neurons responded to >5 stimuli of the seven tested. The total number of neurons (n) responding to a single quality or multiple qualities is indicated in the top-left corner for each example neuron. However, the specific stimuli to which they responded varied. The stimuli used were low salt (60 mM NaCl), high salt (500 mM NaCl), sour (50 mM citric acid), bitter (5 mM quinine and 30 µM cycloheximide), sweet (500 mM sucrose and 30 mM AceK), and umami (300 mM monosodium glutamate + IMP).
Penk-GCaMP6s neurons more frequently respond to sour and bitter stimuli and are equally responsive to salt compared with the full population. A, The percentage of Penk-GCaMP6s neurons (n = 87; magenta) with a significant response to each stimulus, compared with the full population (gray; TrkB-GCaMP6s, n = 21; Phox2b-GCaMP6s, n = 194). B, The normalized response for all neurons to each stimulus is shown, regardless of whether they meet the requirement for a significant response. Values accumulating at 1.0 indicate that a particular stimulus was the most effective for that neuron. Note that while many neurons in the full population have response values of one for sucrose and umami stimuli, few to no Penk-GCaMP6s neurons respond best to these stimuli. The stimuli used were sweet (500 mM sucrose and 30 mM AceK), umami (300 mM monosodium glutamate + IMP), low salt (60 mM NaCl), high salt (500 mM NaCl), sour (50 mM citric acid), and bitter (5 mM quinine and 30 µM cycloheximide). **p < 0.01; ***p < 0.001; ****p < 0.0001.
In addition to examining the proportion of neurons that responded to each stimulus, we compared the relative responses (normalized score of 1.0) across neurons for each stimulus, regardless of whether the response was significantly different from the baseline. Penk-GCaMP6s neurons responded poorly to sweet (sucrose; Mann–Whitney; U = 4,479; p = 0.0001; Fig. 9C) and umami stimuli (Mann–Whitney; U = 4,360; p = 0.0008; Fig. 9D), with virtually none of these neurons showing a strong response to these stimuli. Although Penk-GCaMP6s neurons retained some responsiveness to NaCl, these response levels were not significantly different from the full population for either 60 mM NaCl (Mann–Whitney; U = 7,663; p = 0.014; Fig. 9E) or 500 mM NaCl (Mann–Whitney; U = 7,918; p = 0.036; Fig. 9F). Penk-GCaMP6s neurons had a greater median relative response to both citric acid (Mann–Whitney; U = 3,319; p = 0.0001; Fig. 9G) and the bitter mixture (Mann–Whitney; U = 7,984; p = 0.0.007; Fig. 9H) as compared with the full population. We conclude that although Penk-GCaMP6s neurons are sour-sensitive, they are not sour-specific, but instead they respond robustly to multiple taste stimuli that mice find innately aversive including sour, bitter, and high concentrations of NaCl.
Discussion
An advantage of gaining genetic access to a subset of neurons is that the anatomical characteristics of this neuronal subset can then be related to their function (Masland, 2004; Li et al., 2011; Abraira and Ginty, 2013; Zeng and Sanes, 2017). This approach can provide a deeper understanding of how and why neurons exhibit functional characteristics distinct from those of the full population. Such an understanding has not yet been achieved for the peripheral neurons that carry gustatory information from taste buds to the brainstem. Here, we sought to use recently characterized genetic identities to relate the structure of taste neurons to their function (Dvoryanchikov et al., 2011; J. Zhang et al., 2019; Anderson and Larson, 2020).
The structure of individual gustatory neurons has only recently been characterized (T. Huang et al., 2021), and the significance of variation in morphological features remains unclear. In contrast, somatosensory neurons often exhibit stereotypical morphological endings, which vary considerably across neuron types and are related to their function (H. Wu et al., 2012; Rutlin et al., 2014). Similarly, the portion of the taste axon within the taste bud (i.e., the taste arbor) varies tremendously (Ohman and Krimm, 2021b). These variations might signify differences in neuron type or represent snapshots of ongoing plasticity. Unlike the stereotypical endings of somatosensory neurons (Li et al., 2011; H. Wu et al., 2012; Bai et al., 2015; Handler et al., 2023), gustatory neurons connect to taste-transducing cells that have limited lifespans, necessitating connections with new taste-transducing cells over time. Recent studies have shown significant variations in the structure of individual taste arbors over time (Whiddon et al., 2023). Consistently, we found that Penk-tdTomato taste arbors share similar size and complexity variations with the full population of taste arbors, suggesting that anatomical differences in the taste arbor portion of taste axon reflect inherent plasticity rather than distinct neuron types.
The most distinguishing morphological feature of Penk+ neurons is their more frequent apposition with multiple taste-transducing cell types compared with the overall population of geniculate neurons. Taste buds contain two taste-transducing types (Type 2 and Type 3 taste-transducing cells) as defined by electron microscopy (Murray et al., 1969; Murray, 1986). Type 2 cells transduce sweet, umami, and bitter tastes (Delay et al., 1986; Finger, 2005; Yoshida et al., 2009), while Type 3 cells transduce sour stimuli (A. L. Huang et al., 2006; Y. A. Huang et al., 2008; Liang et al., 2023) and nonsodium salts (Oka et al., 2013; Lewandowski et al., 2016; Liang et al., 2023). Penk+ neurons were previously identified as a sour-specific neuron type (J. Zhang et al., 2019), suggesting they receive input solely from a subset of Type 3 taste-transducing cells that only transduce sour taste stimuli (Dutta Banik et al., 2020). However, our results show that some taste buds with Type 3 cells lacked Penk-tdTomato taste arbors and vice versa. If Penk+ arbors were exclusively connected to Type 3 cells, intravital imaging experiments would suggest prolonged periods without any possible connections for Penk-tdTomato taste arbors, which seems unlikely. An alternative explanation could be that Penk+ neurons innervate both Type 2 and Type 3 taste-transducing cells. Consistently, Penk+ neurons show increased sensitivity to bitter stimuli (transduced by Type 2 cells) compared with the full population.
Consistent with the possibility that Penk+ neurons can receive input from multiple cell types, a subset of gustatory neurons innervating the circumvallate taste buds form synapses with both Type 3 and synaptic-like connections with Type 2 taste-transducing cells (Wilson et al., 2022). This demonstrates that taste neurons are capable of synapsing multiple cell types. If the relatively small number of taste arbors connecting to both cell types in circumvallate taste buds also holds true for fungiform taste buds, this implies that only a subset of gustatory neurons has this capability. Here, based on both functional and anatomical characterization, we suggest that Penk+ neurons are one of these neuron types.
It is essential to clarify that observing an individual taste arbor in close proximity (apposition) to both a PLCβ2+ cell and a Car4+ cell at the light level does not confirm the presence of a synapse. We think that Penk+ neurons may establish connections with both Type 2 and Type 3 cell types, based on a combination of their innervation patterns and functional responses. Yet, these response characteristics could also arise if Penk+ neurons specifically connect with a subset of broadly tuned Type 3 cells (Dutta Banik et al., 2020). We consider this possibility less likely because broadly tuned Type 3 taste-transducing cells are known to respond to sucrose and umami as well (Dutta Banik et al., 2020). Therefore, if neurons were primarily connected to broadly tuned Type 3 cells, we would expect them to respond well to sweet, umami, and bitter stimuli, which contradicts our findings. Considering the response characteristics of Penk+ neurons in conjunction with their peripheral innervation patterns, our data suggest that Penk+ neurons likely establish connections with Type 3 cells and a subset of Type 2 cells that are responsive to both bitter and sour stimuli (Barretto et al., 2015). Furthermore, it seems likely that another group of peripheral taste neurons receives input from Type 3 cells, as indicated by the presence of multiple taste buds with Type 3 cells and no Penk-tdTomato taste arbors.
Given that more Penk+ neurons innervate the palate than the tongue, it might be initially surprising that they primarily respond to aversive stimuli. Whole-nerve recordings showed that the taste nerve innervating the palate (GSP) is particularly sensitive to sucrose (Nejad, 1986; Harada and Smith, 1992). However, this high whole-nerve response might result from a few neurons with high-frequency responses to sucrose (Sollar and Hill, 2005). It is also important to note that these earlier studies were not conducted in mice, and there may be differences across rodent species. In mice, the GSP responds robustly to HCl and quinine as well as to sucrose (Tomonari et al., 2014). Additionally, there are more Type 3 (sour-responsive) cells per taste bud in the mouse's soft palate compared with fungiform taste buds (Ohtubo and Yoshii, 2011; Koyanagi-Matsumura et al., 2021). Finally, there may be differences in the responses of Penk+ neurons innervating the palate compared with the tongue, which warrants further investigation.
Anatomical evaluation of genetically defined neuron taste types is essential for accurately interpreting functional responses. For instance, using the same Egr2Cre mouse line that was previously utilized to define salt-sensitive peripheral taste neurons (J. Zhang et al., 2019), we observed no Egr2Cre-mediated gene recombination in the geniculate ganglion neurons. Instead, it is plausible that synaptic inhibition of Egr2+ neurons in the brainstem blocked the increased licking behavior observed in sodium-deprived mice in this earlier study (J. Zhang et al., 2019). This explanation is supported by the extensive gene recombination in brainstem neurons observed in Egr2Cre mice. Thus, understanding the anatomical location of the neurons being manipulated is critical for accurate interpretation of the experimental results. Additionally, we noted discrepancies between the observed patterns of innervation for Penk+ neurons and their previously defined functional characterization as sour-specific neuron types (J. Zhang et al., 2019), prompting a reevaluation of the function of these neurons using additional stimuli processed by Type 2 taste bud cells. Beyond their responses to sour stimuli, we found that Penk+ neurons are more sensitive to bitter stimuli compared with the full population. This is consistent with our anatomical findings that these neurons form appositions with both Type 3 and Type 2 taste bud cells. Our findings emphasize the importance of anatomically characterizing genetically defined neuron types to gain a comprehensive understanding of their functional roles.
Recent RNA-seq analysis of the geniculate ganglion has categorized neurons into several expression-based groups (Dvoryanchikov et al., 2011; J. Zhang et al., 2019; Anderson and Larson, 2020). These groups were initially defined based on taste quality categories, identifying neurons that encode sweet (Spondin1+), salty (Egr2+), sour (Penk+), bitter (Cdh13), and umami (Cdh4; J. Zhang et al., 2019). However, our reexamination contradicts the presence of Egr2+ neurons in the geniculate ganglion and challenges the classification of Penk+ neurons as exclusively sour-specific. Instead, we observed that these neurons exhibited heightened responses to sour and bitter tastes, as well as responses to high salt concentrations—all innately aversive stimuli. Notably, they showed poor responses to nonaversive stimuli, indicating a potential classification of peripheral taste neurons into genetic types based on innate hedonic value rather than taste quality. This is consistent with reports showing that mice show poor discrimination between sour and bitter tastes (Treesukosol et al., 2011), indicating that these two tastes share common perceptual elements. It is also consistent that neural coding in the cortex reflects hedonic value (Raymond et al., 2024). Therefore, the concept that genetic markers neatly correspond to the five basic tastes requires reassessment.
Defining neuron types in other systems has required a continuous splitting and refinement of larger clusters of neurons (Zeng and Sanes, 2017). It is possible that Penk+ neurons could be further subdivided into bitter-, sour-, and salt-responsive neuron types or by morphological differences. Consistently, both terminal branch number and convex hull appear to be bimodally distributed for Penk arbors, so perhaps they can be segregated further. However, we think it unlikely that separation of neuron types will be on strict categories of taste quality, because 86% of Penk neurons respond to multiple qualities, which is consistent with findings that both taste-transducing cells (Tomchik et al., 2007; Dutta Banik et al., 2020) and neurons (A. Wu et al., 2015) can be broadly tuned. Instead, genetic types might cluster groups of neurons responding to similar stimuli (Erickson et al., 1980). Gustatory neurons lack genes required for transduction but must form specific peripheral and central connections. It is more likely that Penk+ neurons are further divided based on differences in connectivity than specific taste qualities. Deciphering these populations will likely require both deeper RNA sequencing of geniculate neurons and intersectional approaches to evaluate combinations of genetic markers.
Footnotes
We thank Dr. Stephen Roper for providing the calcium imaging approach to the Krimm laboratory. We thank David Alston for writing the analysis program used for the calcium imaging data. We thank Kaytee Horn for her help breeding and genotyping the mice and additional technical assistance. This project was supported by National Institute on Deafness and Other Communication Disorders R01 DC007176 to R.F.K and F31 DC017660 to L.C.O.
↵*L.C.O. and T.H. contributed equally to this work.
The authors declare no competing financial interests.
- Correspondence should be addressed to Robin F. Krimm at robin.krimm{at}louisville.edu.
