To identify novel taste receptors, we developed a multi-step bioinformatics and expression screening strategy. First, since sensory receptors are expected to be membrane proteins, approximately 30,000 mouse open reading frames (ORFs) were scanned for the presence of at least one putative transmembrane segment. Second, because taste receptors are predicted to be very restricted in their expression pattern, ORFs encoding candidate transmembrane proteins were cross-searched against mouse EST databases to eliminate those broadly expressed. Next, to identify the subset specifically enriched in taste tissue, ORFs selected as encoding transcripts infrequently represented in EST databases (~880 candidates) were used in RT-PCR reactions templated with mRNA from TRCs versus control tongue epithelium. Finally, given that our goal was to discover membrane proteins selectively expressed in subsets of TRCs (and ideally not in sweet, bitter or umami sensing cells), we carried out detailed in situ hybridizations against taste papillae. Of 26 cDNAs used in situ studies, five were found to robustly and selectively label subsets of TRCs. Figure 1 shows that one of these candidates, PKD2L1 is expressed in TRCs of all taste papillae, including fungiform, circumvallate, foliate and palate taste buds.

PKD2L1 encodes a polypeptide displaying significant amino acid sequence similarity to PKD2⁴, a gene mutated in many cases of autosomal dominant polycystic kidney disease^17,18. PKD2s are members of the TRP superfamily of ion channels¹⁹, and have been recently shown to function as non-selective cation channels when expressed in heterologous cells^17,18,20. While the exact roles of PKDs remain unknown, they are believed to function as receptor/ion-channel complexes, often localized to ciliated compartments, and implicated in sensing extracellular signals (e.g. in renal epithelial cells^17,18). We reasoned that if PKD2L1 has a specific role in taste it should be expressed in subpopulations of taste receptor cells with unique functional characteristics. To determine which type of TRCs express PKD2L1, we performed double labeling experiments with sweet, umami and bitter taste receptors (T1Rs and T2Rs), as well as TRPM5, the transduction channel of sweet, bitter and umami sensing cells. Our results (Figure 1) established that PKD2L1 is expressed in cells distinct from those mediating sweet, umami and bitter taste (see also ²¹).

Mammalian taste receptor cells project specialized apical microvilli to the taste pore, the site of interaction between tastants and taste receptor proteins. All known taste receptor proteins localize to, and function, in this TRC compartment ^{1,5–7,14,16,22}. Therefore we would expect bona-fide candidate receptors to also be enriched in the taste pore. We generated antibodies to PKD2L1 and used them in immunofluorescence staining of tongue tissue sections. Examination of CV, foliate and fungiform papillae demonstrated that PKD2L1 protein is indeed enriched in the apical surface of taste receptor cells, with the antibodies robustly labeling the taste pore region (Supplementary Figure 1). These results implicate PKD2L1 as part of the taste sensing machinery.

PKD2 isoforms often require PKD1s for functional expression at the cell surface^17,18,20. The mammalian genome contains 4 members of the PKD1 family: PKD1, PKD1L1, PKD1L2 and PKD1L3^17,18. We performed in situ hybridization studies with gene specific probes representing each family member, and determined that PKD1L3 is specifically co-expressed with PKD2L1 in CV and foliate TRCs (Figure 1, see also²¹). We also generated antibodies to PKD1L3 and demonstrated selective co-expression with PKD2L1 in non-TRPM5 expressing cells of the CV and foliate (Figure 1 and Supplementary Figure 1). Surprisingly, PKD1L3 transcript or protein is not detectable in fungiform or palate taste buds, suggesting that a different partner may be expressed in those TRCs.

To functionally dissect the role of PKD2L1-expressing cells in the tongue, we engineered mice where these cells were genetically ablated by targeted expression of attenuated diphtheria toxin²³ (DTA). To validate this approach as a means of uncovering TRC function, we first generated mice where T1R2-regulatory sequences were used to target DTA expression²⁴ (see Supplementary Figure 2). T1R2 is an essential subunit of the sweet receptor heterodimer (T1R2+3), and the selective ablation of these cells should generate animals with a specific loss of sweet taste^6,9,10,16. To investigate the taste responses of the genetically modified mice, we recorded tastant-induced action potentials from nerves innervating taste receptor cells of the tongue; this physiological assay monitors the activity of the taste system at the periphery, and provides an accurate and reliable measure of taste receptor cell function. Indeed, animals expressing DTA in T1R2 cells have an extraordinary loss of sweet, but importantly retain umami, bitter, sour and salty tastes (Figure 2). These results further substantiate the exquisite segregation of taste modalities at the periphery, and demonstrate the utility of using DTA-mediated ablation of TRCs as a strategy for dissecting taste system function. Next, we engineered animals where the PKD2L1 gene was used to target Cre recombinase into PKD2L1-expressing cells (see Supplementary Figures 2 and ​and3).3). These mice were crossed to the conditional DTA lines²⁴, and double-positive progeny were scrutinized both for the specificity and efficiency of killing, as well as the integrity of taste buds. We checked the expression of T1Rs, T2Rs, and TRPM5 ^8,25 in control and DTA-expressing animals, and found no significant differences in the number or distribution of T1R- or T2R-positive cells between wild type and ablated taste tissue (Supplementary Figure 2). In contrast, the DTA-targeted mice had a profound and practically complete loss of PKD2L1-expressing TRCs in the tongue (Supplementary Figure 2). Remarkably, genetic ablation of the PKD2L1-expressing cells produces animals with a devastating loss of sour taste (Figure 2). Responses to all acid tastants, including citric acid, HCl, tartaric acid and acetic acid are completely abolished, with no significant activity over a range of 5 orders magnitude of proton concentrations. However, responses to sweet, umami, bitter or salty tastants remain indistinguishable from wild type control animals. These results firmly establish PKD2L1-expressing cells as the sour taste sensors, and further substantiate a model of coding at the periphery in which individual taste modalities operate independently of each other.

Acid sensing is important not only in the taste system, but also for monitoring the functional state of body fluids, including the internal milieu of the brain. This is particularly well-studied in the central and peripheral control of respiration, where pH sensing is the principal mechanism for monitoring CO₂ levels in the blood and cerebrospinal fluid^11,26,27 (CSF). Thus, we wondered whether PKD2L1 might be expressed in additional cell types, and if so whether such cells may also be involved in pH sensing in other physiological systems.

We carried out in situ hybridization and antibody staining experiments with PKD2L1 on a wide range of other tissues and identified a singular additional domain of expression: a discrete population of neurons surrounding the central canal of the spinal cord, through its entire length, from its origin in the brain stem to its end around the cauda equina (Figure 3). Notably, these neurons send processes into the central canal, suggesting they may function as chemoreceptors sensing the internal state of the CSF (Figure 3b,g, ¹¹). Given their anatomical distribution and cellular morphology, we reasoned these cells might be part of the homeostatic circuitry responsible for monitoring and reporting the pH of the cerebrospinal fluid. This postulate predicts that these neurons should trigger action potentials in response to acid stimulation. Therefore, we engineered mice where a GFP reporter was targeted to PKD2L1-expressing cells, and performed patch clamp recordings from GFP-labeled cells in a spinal cord slice preparation²⁸. A priori, we anticipated some notable differences in the behavior of these cells compared to TRCs; while the taste system is tuned to respond to acid stimulation in the range of multiple pH units (i.e. pH 2–5), we expected the CSF monitor cells to respond to pH changes within a range of a few tenths of deviation from pH 7.4. Indeed, Figure 4 shows that the PKD2L1-expressing neurons display exquisite sensitivity and selectivity to pH stimulation. Exposure to test solutions between pH 6.5 and 7.4 evoked a dramatic, dose dependent and reversible increase in action potential (AP) frequency (Figure 4 and data not shown). In contrast, the same acid stimuli have no significant impact on the response of control (e.g. unlabeled) cells, even after exposure to pH as low as 6.5 (lower pHs triggered irreversible damage to the slice preparation).

Most of the known CSF-contacting neurons in mammals project ciliated dendrites into the CSF, where they are proposed to sense fluid flow, pressure, pH or the composition of the CSF¹¹. Our demonstration that PKD2L1-expressing cells of the spinal cord selectively fire in response to minor changes in proton concentration strongly suggests that they function as sentinels of cerebrospinal and ventricular pH. Collectively, these results uncover an entirely unexpected role for members of the PKD family of proteins, offer a new perspective into the potential significance of PKD2s in health and disease, and bring forth a surprising unity in the cellular basis of pH sensing in very different physiological systems. In the future, it will be of interest to develop an activity assay for PKD2L1 to establish the molecular mechanism of acid activation, to study the phenotype of PKD2L1 knockout animals, and determine whether PKD2L1 functionally associates or interacts with different partners in different cells types. In this regard, it would be worth exploring whether the differences in pH sensitivity between the tongue and spinal cord might be due to differences in PKD2L1-receptor complex composition.

The nature of the mammalian sour taste receptor and sour-sensing TRCs have been fertile ground for speculation over the years. A wide range of cell types, receptors, and even receptor-independent mechanisms, have been proposed to mediate acid detection in the tongue^1–3. The results presented in this paper establish that sour taste, much like our previous findings for sweet, umami and bitter is mediated by a unique cell type, independent of all other taste qualities. In addition, our demonstration that sour-less mice have normal salt responses demonstrates that salt taste is also mediated by independent TRCs. Together, these results impose a considerable revision of the current views of taste representation at the periphery, and make a compelling case for a labeled line mode of coding across all five taste modalities and TRC types.