Encoding Taste: From Receptors to Perception

Abstract

Taste information is encoded in the gustatory nervous system much as in other sensory systems, with notable exceptions. The concept of adequate stimulus is common to all sensory modalities, from somatosensory to auditory, visual, and so forth. That is, sensory cells normally respond only to one particular form of stimulation, the adequate stimulus, such as photons (photoreceptors in the visual system), odors (olfactory sensory neurons in the olfactory system), noxious heat (nociceptors in the somatosensory system), etc. Peripheral sensory receptors transduce the stimulus into membrane potential changes transmitted to the brain in the form of trains of action potentials. How information concerning different aspects of the stimulus such as quality, intensity, and duration are encoded in the trains of action potentials is hotly debated in the field of taste. At one extreme is the notion of labeled line/spatial coding – information for each different taste quality (sweet, salty, sour, etc.) is transmitted along a parallel but separate series of neurons (a “line”) that project to focal clusters (“spaces”) of neurons in the gustatory cortex. These clusters are distinct for each taste quality. Opposing this are concepts of population/combinatorial coding and temporal coding, where taste information is encrypted by groups of neurons (circuits) and patterns of impulses within these neuronal circuits. Key to population/combinatorial and temporal coding is that impulse activity in an individual neuron does not provide unambiguous information about the taste stimulus. Only populations of neurons and their impulse firing pattern yield that information.

“Taste coding” is often interpreted to mean how the gustatory nervous system discriminates sweet, sour, bitter, salty, umami, and perhaps fat tastes (Roper and Chaudhari 2017; Running et al. 2015). However, gustatory stimuli have other properties/features that the sensory nervous system encodes, including stimulus intensity, duration, and hedonic value (or “valence,” i.e., pleasant vs unpleasant). The following pages attempt to guide the reader how the gustatory nervous system processes and encodes taste signals, beginning with initial sensory transduction at the level of membrane channels and receptors on taste bud sensory cells, and progressing to higher order brain centers in the gustatory cortex. The significance of understanding how the nervous system encodes sensory information in general, and taste in particular, is highlighted, for example, by successful efforts in vision, where images can be elicited by applying electrical pulses with an appropriate “code” to the retina (Brackbill et al. 2020) (http://med.stanford.edu/artificial-retina.html), or perceived images can be decoded and reconstructed from electrical signals recorded from the visual cortex (Tripathy et al. 2021). However, perhaps most impressive are the advances in decoding CNS language circuits and signals. Here, biomedical researchers have constructed brain–computer interfaces to restore the ability of individuals with severe speech impairments to communicate (Moses et al. 2021). All these endeavors have required a fundamental understanding of how and where the brain had encoded the sensory signals.

1. Information Coding in the Peripheral Nervous System

1.1. Exteroreceptors

Before focusing on taste, per se, some basic principles and common themes in the sensory nervous system are in order. The following is a brief overview of how sensory signals are received and transduced in peripheral sensory organs and transmitted to the brain (Fig. 1).

Fig. 1.

Schematic diagram of sensory receptor pathways. (a) Drawing showing receptor cells and their central connections in the somatosensory nervous system (touch, proprioception, pain, itch). Sensory neurons (1) reside in ganglia (red dashed line) that lie alongside the spinal cord and brain. These neurons send sensory afferent fibers to the periphery (to the left). These peripheral processes express molecular receptors for the cell’s adequate stimulus. The central processes of sensory neurons (to the right) enter the CNS and synapse with neurons in the spinal cord and hindbrain (2). These CNS neurons in turn project to higher brain centers (3). (b) In the sensory end organs for hearing, balance, vision, and taste (cochlea, vestibular apparatus, retina, and taste buds, respectively) (blue dashed line), the sensory cells (4) communicate synaptically with sensory ganglion neurons (1). The sensory cells may have cell–cell interactions within the peripheral end organ itself (double arrows in 4). Ganglion neurons (1) for these senses transmit signals into the CNS, as in (a). Olfactory sensory neurons (not depicted) are somewhat a “hybrid” of these two structures (a, b). Olfactory sensory neurons reside in the peripheral tissue (olfactory epithelium in the nasal cavity) and send their axons directly into the brain (olfactory lobe)

Sensory stimuli for the body’s exteroreceptors consist of some form of external energy or external force, such as a photon (vision), or gravitational pull (balance), or the chemical energy released when an odorant binds to its receptor (smell), or a physical distortion of the cell membrane (touch), and so forth. Every sensory receptor cell has a specialized region where membrane proteins designed to capture a particular form of energy are embedded. For example, the distal ends of rod photoreceptors are pancaked with flattened intracellular compartments (“discs”) whose membranes are rich in rhodopsin, a G protein coupled receptor protein (GPCR) that absorbs photons (Fig. 2a). Or, cochlear hair cells possess an apical tuft of specialized, elongate stereocilia that express mechanosensitive channel proteins on their tips (Fig. 2b). These mechanosensitive channels are tugged open/shut as the stereocilia on hair cells are pushed to and fro by acoustical vibrations. Thus, sensory receptor cells for each modality (vision, hearing, touch, taste, smell, etc.) are characterized by the presence of specific membrane receptor proteins located in particular regions of the cell, with these receptor proteins being “tuned” to the appropriate energy source for that modality. This energy source is often termed the adequate stimulus. Interactions between an adequate stimulus and its cognate membrane receptors are converted into generator potentials within the sensory receptor cell, described below.

Fig. 2.

Sensory receptor cells have specialized regions designed to capture their adequate stimulus. (a) rod photoreceptor, showing stacks of specialized structures (intracellular disks) that contain the photosensitive protein, rhodopsin. Reproduced with permission from Wikimedia (DžiugilėMED – Own work, CC BY-SA 4.0, https://commons.wikimedia.org/w/index.php?curid=43488131). (b) cochlear hair cell illustrating the specialized apical stereocilia that transduce mechanical perturbations. Courtesy of J. Hudspeth, with permission

In brief, the “code” for stimulus modality at the level of peripheral sensory receptor cells is the expression of appropriate membrane receptor proteins. This is under transcriptional control by the sensory receptor cell genome – i.e., the code for sensory “modality” is a genetic code.

1.2. Stimulus Intensity

The interaction between an adequate stimulus and its membrane receptor protein in most instances is not an all-or-none event. That is, apart from certain GPCRs such as rhodopsin, ligand binding/receptor activation is not believed to act in the manner of an on/off switch. Rather, as described by Kobilka and Deupi (2007), “GPCRs behave more like rheostats.” Thus, encoding stimulus intensity at the level of single membrane receptor proteins, certainly for GPCRs and perhaps also for ionotropic receptors, is carried out by an increase in the receptor active state. Put simply, the stronger the stimulus energy, the more likely the receptor protein will be in an active state.

At the level of the sensory cell, an increased stimulus intensity will recruit an increased number of membrane receptor proteins, each of which contributes to a varying degree depending on its activation state. In certain cases where the adequate stimulus itself is an ion (e.g., for nociceptors, K⁺ released by damaged cells), an increased stimulus intensity (increased extracellular K⁺ concentration) drives a larger ion influx (larger electrical current) through membrane ion channels. There is no agonist-triggered “active state”; the relevant channels are constitutively open. However, in most cases, an increased stimulus intensity (mechanical, thermal, chemical, etc.) recruits an increasing population of membrane receptors or channels, each of which (especially for GPCRs) is in an elevated state of activity, i.e., their “rheostat” is turned up.

The net effect of stimulating receptor proteins on sensory receptor cells is to initiate a flux of ions across the receptor cell membrane (i.e., electrical current), either directly in the case of ionotropic receptors and ion channels (as just described above for nociceptors) or indirectly as a consequence of activating effector ion channels downstream of GPCRs. This ionic flux produces a change in the voltage across the peripheral sensory receptor cell membrane, termed a generator potential (Fig. 3).

Fig. 3.

Stimulus intensity and duration are encoded by the graded amplitude and variable duration of generator potentials in somatosensory ganglion neurons (a) and in receptor cells of sensory end organs (b). (a) increasing the stimulus () strength (stim) produces larger generator potentials which, if they reach threshold, elicit action potentials (inset, above left). Note that although the stimulus might be maintained (inset), the generator potential declines, or adapts. Action potentials are propagated (inset, right) to the neuronal soma in a sensory ganglion and into the central nervous system. (b) similarly, increasing the stimulus strength to a receptor cell in a sensory end organ recruits additional membrane receptors and produces a larger generator potential. This graded generator potential, in turn, causes graded release of synaptic transmitter (vesicles at base of sensory receptor cell, right). Note that the generator potential here also adapts, despite a maintained stimulus

All this is to say that the cumulative activity of activated GPCRs and their downstream ion channel effectors, or the activity of ionotropic receptors leads to a generator potential (sometimes called “receptor potential”). The generator potential is graded and encodes increased stimulus intensity by an increased membrane potential change; the greater the number of individual receptor events, the larger the ion flux, and hence the larger the change in membrane potential.¹

1.3. Stimulus Duration

Generally speaking, the precise duration of a sensory stimulus is only approximately encoded by the time course of the generator potential it produces. Passive electrical properties of sensory receptor cells (capacitance, conductance) determine how quickly the generator potential tracks the ion fluxes (transmembrane currents) initiated by the sensory stimulus. Importantly, adaptation limits the duration of the generator potential in most sensory receptor cells. Adaptation is the dissipation of the generator potential even as the sensory stimulus remains constant (Fig. 3, insets).

Cellular and molecular events underlying adaptation of the generator potential are complex and vary greatly among the different types of receptor cells. Adaptation of the generator potential can be produced by: (a) relaxation of the proximate stimulus due to properties of accessory tissues surrounding the receptor cell. For example, in mechanosensitive Pacinian corpuscles, non-sensory cells that surround and encapsulate the sensory neuron terminal dissipate/filter the physical deformations of the sensory organ (Mendelson and Lowenstein 1964). This is a classic example of how accessory tissues mediate adaptation; (b) receptor protein desensitization caused by intracellular biochemical processes, such as phosphorylation of GPCRs followed by binding of inhibitory, “arrestin,” proteins (Gurevich and Gurevich 2020); (c) inhibition of effector ion channels downstream of GPCRs, caused by Ca²⁺ influx (Nakatani et al. 2002); and (d) actions of Ca²⁺ upon key enzymes in the transduction pathway, resulting in inhibitory feedback (Fain et al. 2001). Finally, generator potentials are shaped by the intrinsic biophysical properties of the sensory receptor cell itself, such opening/closing of membrane ion channels triggered by membrane depolarization and hyperpolarization. There is no one-size-fits-all explanation for sensory receptor cell adaptation.

In brief, the time course of a generator potential does not necessarily accurately depict the precise time course of the sensory stimulus.² In most instances the generator potential fades before the sensory stimulus disappears (adaptation). Interestingly, in certain sensory receptor cells, specifically nociceptors, the opposite occurs; positive feedback mechanisms can prolong and even amplify the generator potential (Woolf and Salter 2000).

1.4. Transmitting Sensory Information from the Periphery to the CNS

Once the sensory stimulus produces a generator potential in the sensory receptor cell, there are two possible outcomes for encoding this information. In primary sensory afferent fibers of the somatosensory system (i.e., touch, temperature, pain, itch, proprioception) and sensory receptor neurons in the olfactory system, generator potentials elicit action potentials that are propagated directly from the periphery into the CNS. By contrast, in peripheral sensory organs such as the cochlea, vestibular apparatus, retina, and taste buds, generator potentials elicit synaptic neurotransmitter release from local sensory receptor cells onto sensory neuron terminals and these (secondary) sensory neurons transmit the “code” into the CNS.

1.5. Stimulus Modality

In peripheral sensory afferent fibers, information about the sensory modality (i.e., touch, temperature, taste, itch, pain, etc.) is encoded by the activation of discrete neural pathways. That is, touch receptors are part of a specific neuroanatomical pathway that ultimately terminates within a particular brain cortical region and activates neuronal circuits in the CNS that are dedicated to touch; thermoreceptors participate in a separate pathway that terminates in somewhat different cortical regions and activates different neural circuits, and so forth. For the somatosensory nervous system, topographical maps of these distinct regions in the brain form miniature representations of the body projected onto the cortical surface to form an homunculus (Fig. 4). Other sensory systems (auditory, visual, gustatory, olfactory) have varying degrees of topographical or other systematic mapping from the periphery onto the cortical surface, from highly ordered (auditory, visual) to lack of a precise mapping (olfactory).

Fig. 4.

The sensory homunculus. Somatosensory information is encoded as a topographical map of the body onto the primary sensory cortex (below, red). Cortical neurons here occupy an area proportional to their sensory field (receptive field) in the periphery. Thus, body regions having a dense somatosensory innervation such as the face and lips take up an enlarged cortical surface relative to less-densely innervated regions (e.g., neck, trunk). The result is a distorted topographical body map, or homunculus, on the primary sensory cortex (Penfield and Rasmussen 1950). Reproduced and modified, with permission, from https://upload.wikimedia.org/wikipedia/commons/e/e6/BA312_-_Primary_Somatosensory_Cortex_-_lateral_view.png and https://commons.wikimedia.org/wiki/File:Homunculus-ja.png

1.6. How Do Afferent Fibers Encode Intensity?

The intensity of sensory stimulation is encoded by the firing rate of action potentials in an individual sensory afferent fiber (frequency coding) and by the number of afferent fibers carrying action potentials (population coding). In brief, the more intense the stimulation, the higher the frequency of action potentials transmitted by a sensory afferent fiber and the greater the number of fibers activated.

1.7. Sensory Afferent Fibers Only Partially Encode Stimulus Duration

The duration of a sensory stimulus is only approximately coded by the evoked action potentials in many afferent fibers. Often there is an initial burst of activity (“on response”) followed by a decrease or even cessation of impulses even in the face of a maintained stimulus, particularly for mechanosensors.³ Adaptation of the receptor potential (above) largely explains the decline of impulse traffic in afferent fibers even when the sensory stimulus is maintained. The rate of adaptation varies among different types of sensory afferent fibers, with some being rapidly-adapting and others being slowly-adapting. Furthermore, for some afferents, particularly rapidly-adapting somatosensory afferents, there is a second burst of impulses at the offset of the stimulus (“off response”) that encodes termination of stimulation (Fig. 5). Biophysical mechanisms underlying off responses in sensory afferent fibers are not well described. Presumably, off responses in mechanoreceptor sensory organs involve displacement of the surrounding tissues when the physical stimulus is withdrawn (Iggo and Ogawa 1977). Overall, adaptation of impulse activity in sensory afferent fibers limits the ability of these fibers to precisely encode the kinetics of a stimulus.⁴

Fig. 5.

Sensory afferent fiber “on response” and “off response.” Action potentials in a single afferent fiber from a rapidly adapting receptor in the isolated trachea of an anesthetized cat. Upper record, tracheal pressure; lower record, action potentials. N.B. how this rapidly adapting sensory fiber encodes the onset (“on response”) and offset (“off response”) of stimulation, but not the entire duration. Redrawn, from Widdicombe (1954)

1.8. How Do Sensory Afferent Fibers Signal the Location of Stimulation?

The location of a stimulus is encoded according to the notion of receptive fields. The receptive field for somatosensations (touch, temperature, pain, itch) refers to the specific surface area on the body where an afferent fiber terminates. This area is the field of skin (for exteroceptors) where the stimulus is effective (e.g., a minute area of skin on the hand for touch receptors there) (Fig. 6). Each afferent fiber has its own “personal” receptive field. Receptive fields of neighboring afferents of the same modality overlap such that adjacent fibers share portions of their receptive fields. By comparing signals coming from two or more adjacent afferent fibers, the nervous system can achieve higher spatial acuity for discriminating the position of a stimulus.

Fig. 6.

Receptive fields on a monkey fingertip. The stimulus is a field of raised dots on the surface of a rotating drum. Illustrated here is a typical area on the finger where tactile stimulation would excite an underlying somatosensory receptor terminal, i.e., the terminal’s receptive field. Modified from DiCarlo et al. (1998), with permission

As mentioned above, somatosensory receptive fields are mapped topographically onto brain structures such as the somatosensory cortex. Every region of this cortex responds to a specific part of the body, producing a map of the body on the cortex, the “homunculus.” This map is a distorted representation of the body according to the relative densities of the receptive fields. Thus, regions such as the hands and fingers that are densely populated with sensory afferents and their receptive fields command a larger cortical area relative to input from other body regions.

Comparable receptive fields and their topological mapping in vision also occur in the retina, thalamus, and primary visual cortex.

The concept of receptive fields only loosely applies to olfaction. Although olfactory sensory neurons are distributed in broad stripes or zones along the olfactory epithelium inside the nasal cavity (Ressler et al. 1993), these are not receptive fields, per se. Spatial localization of odor stimuli has more to do with behavioral responses such as sniffing and turning one’s head than it has to do with olfactory “receptive fields.” There is no topographical map or homunculus in the brain for odors⁵; indeed, how odors, let alone their localization in space, are represented in the brain appears to be as “dispersed ensembles” of neurons without any obvious topological pattern (Stettler and Axel 2009).

In taste, gustatory receptive fields resemble somatosensory receptive fields and describe a spatial localization of the stimulus on the tongue. Indeed, localizing taste stimuli is assisted by tactile (somatosensory) receptive fields.⁶ That is, a sapid stimulus (e.g., food particle, crystal of salt) is often accompanied by tactile stimulation. Nonetheless, when somatosensations are carefully controlled, human subjects can identify the location of taste stimuli on the tongue (Lim and Green 2008). There are bona fide taste receptive fields. Curiously, the tongue has higher “acuity” for some tastes (e.g., sweet, salty) compared to others (e.g., bitter) (Lim and Green 2008).

2. Information Coding in the Central Nervous System

2.1. Sensory Modalities

Every sensory modality (vision, touch, hearing, smell, taste, etc.) has specific features or qualities that can be deciphered. Thus, visual stimuli consist of different colors, shapes, and contrasts. Our somatosensory system readily discriminates light touch versus pressure versus stretch. Humans can distinguish different tones within the range of ~20 Hz to ~20 kHz. There are five or more basic tastes and perhaps more than a trillion distinctive odors (Bushdid et al. 2014, but see Gerkin and Castro 2015). How the brain identifies and discriminates among different qualities within a given modality is a major question in sensory coding.

Impulses from the peripheral sensory afferent fibers for a given modality (discussed above) synapse with specific clusters of neurons (nuclei) embedded in the spinal cord or hindbrain. Those neurons then send axons to the thalamus, a central sensory processing structure in the brain (Fig. 7). The thalamus is subdivided into different anatomical regions for each modality – vision, touch, audition, somatosensations (i.e., touch, temperature, itch, pain, others) and so forth.

Fig. 7.

The thalamus collects information from peripheral sensory receptors and distributes it to overlying cortical structures. The thalamus is a large, bilateral ovoid structure in the middle of the brain (dark shaded area). The overlying somatosensory cortex is shown here in green. For clarity, other sensory cortices (e.g., auditory, gustatory, visual) are not depicted. Reproduced and modified, with permission, from https://www.kenhub.com/en/library/anatomy/parietal-lobe, with permission

From the thalamus, sensory information is radiated to the appropriate, overlying primary sensory cortex (Fig. 7), distinct for each modality. For instance, visual information is routed from the retina to the thalamus and then to the primary visual cortex in the occipital lobe; auditory information from the cochlea to hindbrain, then thalamus, and on to primary auditory cortex in the temporal lobe; taste information from taste buds to the hindbrain, then thalamus, then primary gustatory cortex in the insula and operculum.

A striking exception to this general pattern of sensory nervous system organization is olfaction. Here, sensory information travels from the olfactory epithelium in the nose to its first relay in the CNS, the olfactory bulb. From the olfactory bulb, olfactory signals are transmitted directly to the primary olfactory (piriform) cortex and other CNS regions, bypassing the thalamus.

As a general principle, stimulus identity, intensity, and duration are encoded by ensembles of neurons (neural circuits) in the CNS, and not by individual neurons per se. Ensembles of CNS neurons can be relatively compact and highly organized, as in the cortical columns found in the somatosensory, auditory, and visual cortices (Linden and Schreiner 2003; Mountcastle 1997). Or, the neurons that process sensory information in the CNS can be widely dispersed with no obvious anatomical relationships, as in the olfactory (piriform) cortex (Stettler and Axel 2009).

The point is that in the brain, the neural code for identifying a stimulus, its intensity, and duration is no longer a simple construct of a generator potential and series of action potentials in one neuron or one nerve fiber. Consequently, notwith-standing the tremendous advances made with single channel microelectrode recordings in the somatosensory, auditory and visual systems, to understand CNS coding of sensory signals, especially in taste and olfaction, requires technologies for recording from large ensembles of neurons simultaneously, either with multi-electrodes, high-resolution fMRI, or optical methods.

The reader is referred to any number of modern neuroscience textbooks for our current understanding of how the brain encodes sensory signals in the auditory, visual, vestibular, somatosensory, and olfactory nervous systems. Such an overview is beyond the reach of this chapter; the remainder will focus on sensory coding in the gustatory nervous system.

3. Sensory Coding in the Gustatory Nervous System: From Taste Buds to Cortex

Food and beverages contain compounds (“tastants”) that either (a) bind to membrane receptors on the apical tips of taste bud cells, (b) permeate ion channels and generate ion flux across the taste cell membrane, or (c) penetrate taste bud cells and modulate the intracellular (cytoplasmic) face of membrane ion channels. The “taste code” at this initial stage of signal generation is which of these membrane proteins is/are involved, and of course, which taste bud cell expresses that receptor/ion channel.

3.1. How Do Gustatory Membrane Receptors Identify Taste Stimuli?

Sugars, artificial sweeteners, and sweet-tasting proteins are agonists for two class C GPCRs – TAS1R2 and TAS1R3 (abbreviated hereafter as T1R2 and T1R3). These GPCRs form a heterodimer having several different and somewhat independent binding sites for different sweet-tasting agonists. These sites are illustrated in Fig. 8. Thus, the “code” can be somewhat ambiguous even at the earliest stage of signal transduction. How different sweet-tasting agonists are differentiated is still not well understood, and may involve differences in duration of ligand occupancy, interactions of the ligands with receptors other than T1R2 + T1R3, and ligand interactions with sugar transporters on the taste cell surface (which themselves may contribute to conscious perceptions, although the jury is still out on this notion) (Glendinning et al. 2017; Yee et al. 2011). For example, (a) some sugars such as fructose activate T1R2 + T1R3 and elicit sweetness; (b) others such as glucose activate T1R2 + T1R3 and are taken up by taste cells (via sugar transporters) to activate K_ATP mechanisms and cephalic phase insulin release (CPIR) (Glendinning et al. 2017); and finally (c), some compounds (e.g., starch) elicit CPIR but without sweet taste. Yet, sugar transporters and T1R2 + T1R3 are present in the same taste cell (Yee et al. 2011). Thus, what is the taste receptor cell “code” for sweetness versus the “code” for CPIR? The signals seem to arise from the same taste bud cells.

Fig. 8.

The G protein-coupled sweet taste receptor heterodimer, T1R2 + T1R3, has multiple sites where ligands can bind. Approximate sites where several sweet taste compounds bind the receptor are shown. Reproduced here with permission from Roper (2020)

T1R2 + T1R3 dimers are expressed by a specific set of taste bud cells belonging to the Type II category (Fig. 9). Ligand binding to T1R2 + T1R3 initiates a cascading series of intracellular events that culminates in specific downstream ion channels opening, thereby producing an inward (depolarizing) current (generator potential) and triggering action potentials (Roper and Chaudhari 2017). This represents the conversion of the chemical signal (tastant) into an electrical signal (generator potential), a key step in sensory coding, as previously discussed.

Fig. 9.

Taste buds are populated by several types of cells. (a) Electron micrograph of a rabbit foliate taste bud showing cells with dark or light cytoplasm, and nerve profiles (arrows). Asterisks mark Type II (receptor) cells. (b) Electron micrograph of cross section through a rat vallate taste bud, illustrating distribution and proportion of Type I (here, “D” for dark) and Type II (here, “L” for light) cells. [In this early study it was not possible to distinguish Type III cells, which requires identifying synapses]. Note that the dark Type I cells enwrap surrounding taste cells with lamellar processes. (c) schematic diagram of taste bud depicting relative proportion of Type I (red), Type II (cyan), Type III (blue), and Type IV (grey) cells. Rectangle through middle of taste bud shows approximate plane of cross section in (b). Reproduced here with permission from Roper (2020)

A different combination of T1R receptors, namely TASA1R1 + TAS1R3 (T1R1 + T1R3) is activated by savory (“umami”) tastants (e.g., amino acids, notably glutamate). Umami tastants also activate other Class C taste GPCRs, namely truncated forms of metabotropic glutamate receptors, mGluR1 and mGluR4 (Chaudhari et al. 2000; Nakashima et al. 2012; San Gabriel et al. 2009; Yasumatsu et al. 2015). Little is known about multiple binding sites in all these umami taste receptors. However, T1R1, T1R2, and T1R3 are co-expressed in some taste bud cells, generating possible ambiguity in the encoding of sugars (“sweet”) versus amino acids (“umami”). In fact, under certain conditions, rodents confuse sweet and umami tastants (Saites et al. 2015; Stapleton et al. 1999).

Bitter-tastants activate yet another family of GPCRs, named TAS2Rs (hereafter, T2Rs), with 25 members in humans to date (Meyerhof 2005). T2Rs most closely resemble Class A GPCRs (Di Pizio et al. 2016). Each T2R in this family has a ligand binding site buried in the transmembrane domain of the GPCRs. The binding pockets of the different T2Rs vary in their breadth of ligand selectivity. In experiments designed to explore the specificity of different T2R receptors, Meyerhof et al. (2010) challenged each of the family of 25 human T2Rs with a catalog of 58 natural and 46 synthetic bitter compounds. Certain of the T2Rs were rather broadly selective yet others were “tuned” to a much narrower spectrum of bitter tastants. Although one might argue that this narrow tuning of certain T2Rs is a product of the limited number (N = 104) of compounds tested (Palmer 2019), these studies nonetheless suggest that there are differences in the relative selectivity among T2Rs.

Parenthetically, recent cryo-electron microscopy and functional investigations of a unique insect odorant receptor that has broad chemical tuning, carried out on its bound and unbound states, may be illuminating here. These studies revealed that the ligand binding pocket for the odorant receptor is not a tightly organized, “lock-and-key” site shaped to fit specific chemical or molecular features of a given ligand. Rather, the pocket is a more flexible “promiscuous binding site that recognizes the overall physicochemical properties” of multiple ligands (del Mármol et al. 2021). Whether binding sites in taste T2Rs have similar features awaits comparable detailed structural analyses.

Further, individual taste bud cells express multiple different T2Rs (Behrens et al. 2007). The combination of individual T2Rs being somewhat promiscuous regarding their ligand selectivity (above), along with the fact that individual taste bud cells express multiple members of the T2R family, results in a rather broad range of bitter tastants that can activate any given bitter-sensing taste receptor cell. Indeed, when it has been tested, rodents cannot easily discriminate different bitter tastants (Spector and Kopka 2002). Nonetheless, taste bud cells in mice distinguish among different bitter compounds (Caicedo and Roper, 2001) – bitter-sensing taste bud cells do not form a uniform class that responds identically to all bitter compounds. When human subjects distinguish among bitter compounds, this likely also involves other sensory input such as olfaction and chemesthesis.

T2Rs are also expressed by Type II taste bud cells. However, the T1 and T2 classes of taste GPCRs are not often found in one and the same Type II taste bud cell. Sweet/umami taste receptors (T1Rs) and bitter taste receptors (T2Rs) only very occasionally co-localize in Type II cells (Dando et al. 2012; Sukumaran et al. 2017; Yamada et al. 2021), if at all (Adler et al. 2000).

Acids stimulate sour taste. Acid molecules (e.g., acetic acid, citric acid) permeate cell membranes and acidify the cytosol (Lyall et al. 2001; Richter et al. 2003). In Type III taste bud cells, cytosolic acidification blocks potassium channels (K_IR2.1) that establish the resting potential. By doing so, they depolarize the cell (Ye et al. 2016). Protons in solution also enter Type III cells via specific ion channels (OTOP1), generating an inward proton current and depolarizing the membrane (Teng et al. 2019; Zhang et al. 2019) (Fig. 10). These two mechanisms are specific to acidic stimuli but cannot readily discriminate among different acids. Thus, at the membrane receptor level, the code for sour taste appears to be generalized across all acids.

Fig. 10.

Transduction pathways for sour (acid) taste in Type III taste bud cells. Protons permeate the apical tips of Type III taste cells through OTOP1 channels. This generates a small depolarizing (inward) current. Undissociated acid molecules (HA, e.g., acetic acid) penetrate apical junctional complexes (stippled bar) and permeate cell membranes to acidify the cytosol. Intracellular H⁺ from HA permeation and from influx via OTOP1 bind to and block potassium channels (Kir2.1), thereby depolarizing Type III cells. Reproduced here with permission from Roper (2020)

Far less is known about salt (NaCl) taste transduction, particularly in humans. Thus, the code for salty is still somewhat obscure at the level of membrane mechanisms (Roper and Chaudhari 2017). Intriguingly, a recent study concluded that coincident activation of both Na⁺ and Cl⁻ receptor pathways “encodes” salt taste, reinforcing a longstanding view that sodium and chloride ions alike contribute to the taste of NaCl (Roebber et al. 2019). Further, an important new study identified the taste cells and ion channels that contribute to salt taste preference in rodents (Nomura et al. 2020). These authors reported that NaCl-sensing cells secrete the neurotransmitter ATP via CALHM1/3 channels.

In summary, at the initial event in taste reception, gustatory signals are encoded by specific membrane receptor proteins, transporters, or ion channels for the 5 basic tastes (sweet, umami, bitter, sour, salty) and perhaps fat (Roper and Chaudhari 2017; Running et al. 2015). A small minority of taste bud cells co-express receptors for more than one taste quality.

3.2. Gustatory Stimulus Intensity

For GPCR-mediated tastes, the intensity of the stimulus is encoded by the number of receptors occupied, which in turn determines the number of downstream effectors activated and thus the amplitude of the eventual membrane depolarization (generator potential). Similarly, for tastes transduced by ion channels, the stimulus intensity (tastant concentration) is encoded by the magnitude of the ionic flux across the membrane, and in the case of acid stimuli, by the proportion of K_IR2.1 channels that are blocked by intracellular acidification.

3.3. Gustatory Stimulus Duration

Unlike other senses, and especially hearing, precise timing of the stimulus signal is not as critical in taste. Although under certain experimental conditions, rats can recognize a tastant within 250 ms of stimulation (Halpern and Tapper 1971), in most situations, taste response latency and duration are quite variable. For instance, onset of the stimulus depends on several factors, such as enzymatic digestion in the oral cavity (e.g., lipase-mediated release of fatty acids from triglycerides), mastication, solubilization of food chemicals, and so forth. Further, some compounds, such as quinine, are lipophilic and remain active long after the initial stimulus has been rinsed away. Lastly, little information is available on taste GPCR desensitization or inactivation. In short, apart from its obviously important role in food science and for understanding lingering tastes, little attention has been paid to molecular mechanisms of stimulus duration and adaptation in taste.

3.4. How Do Taste Bud Receptor Cells Discriminate Gustatory Stimuli?

Considerable controversy surrounds taste coding at the level of receptor cells and their connections with the CNS. At one extreme, proponents of labeled line coding propose that individual taste bud cells are “tuned” to specific qualities (sweet, salty, sour, etc.) and transmit these signals to the brain via dedicated primary afferent fibers (i.e., the peripheral processes of gustatory ganglion neurons in the petrosal and geniculate ganglia) (Barretto et al. 2015; Yarmolinsky et al. 2009). Countering this, proponents of combinatorial/population coding propose that the identity of taste quality emerges from signal processing among taste bud cells and multiple connections with primary afferent sensory fibers (Erickson 1963; Ohla et al. 2019; Wu et al. 2015).

It is generally accepted that certain cells in the taste bud, namely Type II taste cells, express specific GPCR taste receptors and are tuned to respond either to sweet, bitter, or umami taste stimuli (see above). Nonetheless, as state previously, there is some “noise” in the expression of taste-specific GPCRs. Although it is not pronounced, single cell analyses have shown mouse Type II taste bud cells can express multiple diverse taste GPCRs (i.e., sweet, umami, and bitter) (Dando et al. 2012; Sukumaran et al. 2017; Yamada et al. 2021). Further, recordings from mouse taste buds show that although most Type II taste bud cells are tuned to one tastant, some respond to two or more tastes (i.e., they are broadly tuned) (Tomchik et al. 2007; Yoshida et al. 2009). Importantly, those same studies showed that although isolated Type III taste bud cells may specifically be tuned to acid taste stimuli, in intact taste buds Type III taste cells respond to several (Type II cell) taste stimuli, in addition to sour. These data have led to the postulate that there is cell–cell communication between taste bud cells (specifically, between Types II and III cells) (Roper 2021; Roper and Chaudhari 2018; Tomchik et al. 2007) (Fig. 11). In sum, the notion that individual gustatory cells are strictly dedicated to a single taste stimulus (i.e., the basis for labeled line coding) is questionable even at the level of the taste bud.

Fig. 11.

Schematic diagram summarizing feedforward and feedback signaling in mammalian taste buds. The diagram shows cell–cell interactions between Type II and Type III taste bud cells. Type II cells express G protein–coupled taste receptors for sweet, bitter, or umami taste compounds. Taste stimulation evokes adenosine triphosphate (ATP) secretion from Type II cells. ATP excites (a) gustatory primary afferent fibers (shown at bottom), (b) neighboring Type III taste bud cells, and (c) via autocrine feedback, Type II cells, as shown here in red. ATP released during taste stimulation is degraded to ADP and adenosine (Ado), both of which, along with ATP, serve as autocrine positive feedback signals. Type III cells make synaptic contacts with nerve fibers and secrete serotonin (5-HT) and norepinephrine (not shown). Type III cells also release γ-aminobutyric acid (GABA) when stimulated by acids (sour tastants). GABA and 5-HT from Type III cells, shown here in blue, inhibit Type II cells. Receptors for ATP, adenosine diphosphate (ADP), adenosine, GABA, and 5-HT are identified at their respective target sites. Reproduced here with permission from Roper and Chaudhari (2018)

Parenthetically, the notion of multiply-responsive, broadly tuned taste bud cells is linked to the concept of entropy and the information content in transmitted signals. The concept of entropy in signal transmission is derived on the classic studies of Shannon and Weaver (1949). Taste researchers quantify the breadth of tuning in taste bud receptor cells and taste neurons in terms of H, or signal entropy. Quantification of entropy in studies on taste yields H values that vary between 0 and 1 (Smith and Travers 1979). The greater the number of tastants to which of a cell responds (i.e., the greater the breadth of tuning), the greater the entropy and the higher the value of H (up to a max of 1.0). Conversely, a cell that is tuned to a single taste compound has no entropy and H = 0. Counterintuitively, in signal transmission, the greater the entropy, the greater the information content in the signal (e.g., see https://machinelearningmastery.com/what-is-information-entropy/). That is, a cell “tuned” to a single taste quality (i.e., H = 0) encodes less information than a neuron that responds to multiple taste stimuli (H > 0). Tomchik et al. (2007) reported that the average H for mouse Type II taste bud cells was low, 0.07; Yoshida et al. (2009) also found entropy was similarly low in mouse Type II taste cells, mean H = 0.09. These findings are both consistent with Type II cells being relatively narrowly (but not completely) tuned to single taste qualities. By contrast, Type III taste cells have much higher entropy, mean H = 0.12 (Yoshida et al. 2009) to 0.47 (Tomchik et al. 2007) in the mouse.

In summary, Type II taste bud cells fairly accurately, but not without some ambiguity or “noise,” encode sweet, bitter, salty, or umami taste stimuli. These cells are not all tuned to a single taste stimulus. Type III taste bud cells encode acid tastants as well as other taste stimuli and consequently respond more broadly to multiple tastes. Cell–cell communication between Types II and III taste bud cells may underlie the ability of Type III cells to encode multiple taste stimuli. Thus, a strict labeled line does not accurately describe taste coding at the level of the taste bud. More realistically, it is likely that while some taste cells respond only to one taste quality and appear to be “labeled,” others have a dominant, though not exclusive, tuning. Thus, a given prototypic basic taste stimulus (sweet, salty, sour, bitter, or umami) activates an ensemble of taste cells, some highly tuned to that taste and others less so, with the combinatorial effect being to signal the dominant taste.

3.5. Stimulus Intensity and Duration Coding in Taste Bud Receptor Cells

Where it has been measured, the concentration/response relationship for taste stimuli in taste bud receptor cells shows a steep, monotonically increasing plot that fits a conventional sigmoidal ligand binding curve (Caicedo et al. 2002; Caicedo and Roper 2001; Roebber et al. 2019). Intensity appears to be encoded at the level of taste cells simply as a function of the amplitude of the response (e.g., generator potential).

Measurements of response duration in taste bud cells using taste-evoked Δ[Ca²⁺]-_{intracellular} as a surrogate for generator potentials approximately track the time course of brief presentations of taste stimuli, although this has not been examined extensively. Where it has been tested, NaCl-evoked responses display marked adaptation to prolonged stimulation (Caicedo et al. 2002; Roebber et al. 2019).

3.6. Do Gustatory Sensory Ganglion Neurons Encode Taste?

Taste bud receptor cells communicate with the peripheral processes (sensory afferent terminals) of gustatory sensory neurons located in the geniculate, petrosal, and nodose ganglia and transmit their signals to these processes for propagation into the hindbrain. Although labeled line coding had long been discussed as one possibility for taste coding in these gustatory neurons, this notion was renewed and given new impetus by in vivo functional imaging studies carried out by Barretto et al. (2015) in mice. These investigators stimulated the animal’s tongue with sapid solutions and recorded geniculate ganglion neuron activity. They reported that 2/3 of the sensory neurons in the ganglion were tuned to only one of the five basic taste stimuli (i.e., had low entropy, H = 0). On this basis, they concluded that the coding of taste signals transmitted from taste buds was via labeled lines. Perhaps supporting this notion were earlier findings on human subjects carried out by Von Bekesy (1964, 1966). Von Bekesy (ibid.) reported that stimulating single taste buds electrically with a fine metal probe, or chemically with microscopic droplets of taste solutions elicited singular taste qualities (sweet, salty, bitter, or sour). Subsequent work by Mueller et al. (2005) purported to test this concept by genetically engineering mice in which taste bud cells expressing T1R2 sweet taste receptors were redirected to express T2R16 bitter receptors. The notion was that if T1R-expressing taste bud cells synapse with dedicated afferent sensory terminals and form a “labeled line” for sweet, then a T2R16 ligand (phenyl-β, PTC-d-glucopyranoside, PTC) which normally is “bitter” should elicit “sweet” in the genetically engineered mice. Indeed, transgenic mice preferred the bitter compound PTC in marked contrast to wild type mice, reinforcing their premise. The authors concluded “Together, these results substantiate the coding of both sweet and bitter pathways by dedicated (that is, labeled) lines” (Mueller et al. 2005).

Yet, there is strong evidence that labeled line coding in gustatory sensory neurons is insufficient. Notably, ever since recordings were made from single afferent gustatory sensory fibers, researchers have noted that individual afferent nerve fibers often responded to multiple (different) taste stimuli (i.e., entropy, H > 0). This is antithetical to a dedicated labeled line coding for taste quality. Frank (1973) attempted to resolve the problem by classifying sensory afferent fiber responses as “taste-best.” That is, a single gustatory afferent fiber may respond to multiple tastants, but robustly only to one taste compound and less strongly to other(s). For instance, there were “sucrose-best,” “citric acid-best,” or “NaCl-best” single fibers, and so forth (Fig. 12). Secondary, weaker responses were considered as “noise” or “side-band.” Thus, a “sucrose-best” fiber would still be considered a dedicated (labeled) line for sweet.

Fig. 12.

Individual afferent fibers that innervate taste buds often show stronger responses to one tastant – the “taste best” response. Shown here are sequential recordings from a NaCl-best chorda tympani fiber from a rhesus monkey when the tongue was stimulated with NaCl, citric acid, sucrose, and quinine hydrochloride (QHCl). Bar at top indicates stimulation interval. Redrawn, from Hellekant et al. (1997a)

Complicating the matter, however, is that the strength of responses, as stated above, is a function of the stimulus concentration. Thus, this modification of “labeled line” coding is unsatisfactory because activity in a single fiber, taken individually, could not be decoded to discriminate between two (or more) taste qualities, let alone between taste intensities. As an example, consider a “sucrose-best” fiber. Intense stimulation (high concentration) by a “side-band”, non-sweet tastant could activate a “sucrose-best” fiber as strongly as a low concentration of sucrose. What information would this fiber, taken individually, thus convey? Importantly, in vivo functional imaging of geniculate ganglion neurons by Wu et al. (2015) conducted in parallel with and published shortly after Barretto et al. (2015), obtained results that contradicted Barretto et al. (2015) and provided further evidence for combinatorial/population coding of peripheral taste signals. Namely, Wu et al. (2015) (and replicated independently by Leijon et al. 2019) reported that about half the gustatory ganglion neurons were selectively responsive to a single quality; the remaining half responded to multiple tastants (H > 0) (Fig. 13). Most importantly, the proportion of selectively tuned ganglion neurons depended on the stimulus concentration. Wu et al. (2015) found that with increasing taste stimulus concentration, neurons became increasingly more broadly tuned, that is, responsive to multiple taste qualities. This echoed the findings obtained from single fiber recordings of hamster taste nerves obtained years earlier (Hanamori et al. 1988). To a lesser extent (although not studied in such detail), Hellekant et al. (1997b) reported the same finding – increased breadth of sensitivity – in chimpanzee single fiber chorda tympani responses when the tongue was presented with an increased concentration of NaCl.⁷ These findings would not result if taste was encoded as a dedicated labeled line; the “label” would stay consistent throughout a range of stimulus concentrations. Further, if taste was encoded by labeled lines, these data would suggest that tastes would be more difficult to identify confidently (i.e., the signal would become “noisy”) with increasing concentration, the opposite of what is observed. Instead, the data are more consistent with a combinatorial/population coding of signals generated by several individual ganglion neurons. In sum, the geniculate ganglion neurons that respond to multiple taste stimuli (32–51%, Barretto et al. 2015; Wu et al. 2015)⁸ do not represent “noise” but instead are intrinsic to combinatorial/population coding for taste quality.

Fig. 13.

Sensory ganglion neurons that innervate taste buds respond to single or multiple taste stimuli. Representative examples of Ca²⁺ imaging (GCaMP3) signals recorded from mouse geniculate ganglion neurons in response to prototypic sweet, umami, salty, sour, and bitter taste stimuli. Responses from six different neurons from two mice are shown. The panel of taste stimuli (top) was presented twice in succession. (a) these neurons responded only to one taste stimulus (sucrose, citric acid). (b) these neurons responded to two or more taste stimuli. Stimuli were sucrose (suc), 300 mM; MSG, 100 mM (with 1 mM IMP); NaCl, 250 mM; citric acid, 10 mM; cycloheximide (Cyx), 1 mM, plus quinine•HCl (Q), 0.3 mM. Reproduced here with permission from Wu et al. (2015)

Additional, though indirect, support of combinatorial coding of the output from taste buds is the finding that there is some, yet ill-defined signal processing taking place between taste bud cells during taste excitation. Paracrine cross talk and autocrine feedback –both excitatory and inhibitory – take place among cells within the taste bud during taste stimulation, mentioned above vis-à-vis Type III taste bud cells and summarized in Roper and Chaudhari (2018). Cell–cell interactions in the taste bud are difficult to reconcile with a singular, dedicated, labeled line signal processing.

Detailed single cell RNAseq analyses on geniculate ganglia from mice may help resolve the question of “taste-labeled” sensory neurons (Anderson and Larson 2020; Dvoryanchikov et al. 2017; Zhang et al. 2019).⁹ Transcriptomic profiling revealed three broad classes of sensory neurons that innervate taste buds, a population of neurons totally separate from those that innervate the ear (the geniculate ganglion receives sensory input from two totally separate regions – the oral cavity and the pinna). Intriguingly, Dvoryanchikov et al. (2017) found that one of the classes of gustatory neurons selectively expresses Piezo2, suggesting that these “taste” sensory neurons respond to tactile stimuli. Zhang et al. (2019) purported to identify specific sensory neurons (“labeled neurons”) for each of the basic taste qualities (sweet, sour, salty, bitter, umami) (but see caveat in footnote 9). These reports clearly await functional confirmation.

Another possible factor for encoding taste identity by peripheral neurons is the element of impulse timing. Specifically, there may be significant information in the patterns of action potentials in individual gustatory sensory ganglion neurons. Early recordings from single chorda tympani fibers in the hamster revealed different patterns of rhythmic firing in response to different taste stimuli, but the authors did not emphasize or discuss this observation at length (Fishman 1957). Others have made similar observations in rats (Hallock and Di Lorenzo 2006; Ninomiya and Funakoshi 1981; Ogawa et al. 1973; Ogawa et al. 1974) and even attempted to mimic gustatory coding by stimulating the chorda tympani nerve with taste-specific patterns of excitation (Covey and Erickson 1980). Interestingly, in experiments on human subjects, changing the frequency of electrical stimulation applied to individual taste buds did not affect taste quality sensation (Von Bekesy 1964). Unfortunately, these early observations on taste-evoked response patterns in peripheral gustatory neurons and their implication for temporal coding in the CNS have not been systematically followed up.

Parenthetically, the above discussion of stimulus identification and taste coding in sensory ganglion neurons does not take into consideration that gustatory sensory neurons in the different cranial ganglia (nodose, petrosal, geniculate) and that innervate different regions of the tongue (posterior, anterior) may convey different taste quality information, at least in rodents, reviewed in Spector and Travers (2005). This is not to say, however, that there is a “taste map” for the different regions on the tongue. This lingual taste map was derived from misinterpretations of original psychophysical measurements. The notion of a taste map on the tongue has long since been discarded (Bartoshuk and Pangborn 1993; Lindemann 1999).

3.7. Stimulus Intensity Coding in Afferent Fibers and Gustatory Sensory Ganglion Neurons

In his classic study on salt taste, Beidler (1953) examined the relationship between responses versus concentration for a number of salts for intensity. The resulting monotonically increasing curve (as well as certain other factors) led him to propose the ground-breaking concept at that time that there is a membrane-bound taste receptor, especially for salt (Beidler 1954). Unsurprisingly, others have since showed similar concentration/response relations for other basic tastes (e.g., Arai et al. 2010; Damak et al. 2003; Danilova and Hellekant 2003; Ganchrow and Erickson 1970) (Fig. 14). In addition to increasing response amplitude with increasing taste stimulus concentration in peripheral neurons, sensory ganglion neurons in mice, studied in vivo with confocal Ca²⁺ imaging, responded to increasing numbers of (different) taste qualities, that is, entropy (H) increases (Wu et al. 2015). As mentioned above, this was also observed in electrophysiological recordings from hamster taste nerve fibers (chorda tympani) (Hanamori et al. 1988) and to some extent in recordings from chimpanzee gustatory afferent fibers (Hellekant et al. 1997b).

Fig. 14.

Geniculate ganglion neurons respond to oral taste stimulation in a concentration-dependent manner. Stimuli were presented at increasing concentrations for each of six test compounds representing prototypic sweet, umami, salty, sour, and bitter tastes. All responses (ΔF/F₀) from a given neuron were normalized to the peak response for that neuron. Symbols show means±s.d. Lines are best-fit sigmoidal curves with variable slope. Reproduced here with permission from Wu et al. (2015)

3.8. Gustatory Sensory Ganglion Neurons: Adaptation and Coding Stimulus Duration

Quantifying the precise duration of tastant is complicated by the nature of the tissue; determining the exact moment a stimulus arrives/disappears at taste buds distributed throughout the oral cavity is an inexact science. Nonetheless, a common observation in recordings from the chorda tympani and glossopharyngeal nerves (i.e., the peripheral processes of gustatory sensory ganglion neurons) is that taste-evoked responses decline during prolonged stimulation (adaptation). Little attention has been paid to adaptation of taste signals in the gustatory nerves. Smith et al. (1978) derived a quantitative model for the decline of responses during NaCl stimulation in the rat chorda tympani nerve but could not explain the cellular/molecular events underlying this adaptation. Lyall et al. (2004) studied adaptation of rat chorda tympani nerve responses to sour taste stimuli. They concluded that the ability of the Na⁺-H⁺ exchanger NHE-1 to restore intracellular pH after acidic stimulation explained adaptation during sour taste. Conceivably, adaptation of taste signals transmitted to the CNS by gustatory sensory ganglion neurons might be mediated by some form of propagated impulse filtering at the T-junction that peripheral axons make with the short process that connects them to their parent neuronal soma (Gemes et al. 2013) (Fig. 15). This phenomenon would not, of course, explain adaptation of responses in the chorda tympani or glossopharyngeal nerves, recorded distal to that T-junction.

Fig. 15.

The “T junction” of sensory ganglion cells is a site where action potentials can be filtered en route to the CNS. (a) drawing of histological section through a dorsal root ganglion (Cajal 1899). One ganglion neuron is highlighted for emphasis (blue). (b) highlighted neuron from above, illustrating the T-junction (dashed circle) and propagation of action potentials (arrows) from the periphery (left), into the soma, and to the CNS (right)

3.9. Gustatory Stimulus Discrimination in the CNS

There is a rich and complex literature surrounding how taste signals are processed in the CNS, from gustatory centers in the hindbrain, the Nucleus of the Solitary Tract, to higher centers in the primary gustatory cortex and in secondary, associated cortical areas (amygdala, orbitofrontal cortex, etc.). It is beyond this chapter to present a detailed analysis of taste coding in these areas. However, certain generalities are important to understand.

First, it must be recognized that until recently, many studies were based on recording CNS neuronal activity in deeply anesthetized animals. This is a major caveat insofar as it is widely accepted that anesthetics significantly affect neuronal firing patterns in the CNS (Sorrenti et al. 2021) and to a lesser degree, also in the PNS (depending on the anesthetic, see Larson et al. 2015).

Second, as mentioned at the outset, the analysis of neural circuits, not of individual neurons, becomes paramount in information processing in the CNS. Although there is some indication that cell–cell interactions influence signal processing and information coding at the level of the gustatory end organs in the periphery (i.e., taste buds, vide supra), there is no question that neural circuits, not neurons taken separately, encode taste signals in the CNS.

Third, there are several major approaches that have been used to measure neural activity in the CNS – (a) electrophysiological recordings with microelectrodes, or more recently, microelectrode arrays; (b) functional imaging using Ca-sensitive probes; (c) functional imaging using magnetic resonance imaging and blood oxygen level dependent (BOLD) imaging, (d) electroencephalogram (EEG) recordings, and magnetoencephalography (MEG). These approaches can be and have been carried out on live, awake animals, and certain of them on human subjects, circumventing the problems of anesthesia.

The advantage of microelectrode studies is that they provide a msec by msec record of neuronal activity, and if carried out with multiple microelectrodes, of neural circuit behavior. As will be discussed below, detailed rhythmic activity in groups of neurons (circuits) appears to encode key aspects of taste. Thus, microelectrode studies are critically important in decoding taste signals in the CNS.

A powerful advantage of Ca²⁺ imaging is the ability to monitor the activity in large ensembles of neurons that respond to taste stimulation. This spatial localization helps guide microelectrode placement, among other things. It also provides information about the extent of activation by a given stimulus. With recent developments, fast-responding Ca-sensitive probes are now able to resolve action potentials in individual neurons, adding temporal to spatial resolution (Chen et al. 2013). Moreover, development of voltage-sensitive probes promises to yield even higher temporal and spatial resolution to functional imaging.

Finally, functional magnetic resonance imaging (fMRI), EEG, and MEG allow taste coding in human subjects to be studied and correlated with data from experimental animals. These methods lack the spatial (and for fMRI, temporal) resolution of microelectrode recordings and Ca²⁺ imaging, but they are non-invasive procedures and can be applied to human subjects. The ability to obtain a window into how human brains process taste is a powerful advantage of these methods.

Given these brief descriptions of methodologies, it might be possible to understand and explain the ongoing heated controversy regarding how the CNS encodes taste signals. This controversy pits “labeled line” (also referred to as a “topographic” or “spatial” coding) versus some form of combinatorial or distributed coding in neural circuits.

As explained above, labeled line taste coding originated from findings describing how peripheral receptor cells and gustatory sensory ganglion cells respond to taste stimuli. It has already been mentioned that some investigators were drawn to how many receptor cells and ganglion neurons appeared to respond to a single taste quality (were highly “tuned”), e.g., sweet compounds. The notion of labeled line taste coding in the CNS was broadly promoted by an influential report published in 2011 (Chen et al. 2011). That publication described results from Ca²⁺ imaging studies on deeply anesthetized mice wherein distinct and different “hot spots” of activity in gustatory cortex were found to be associated with sweet, or bitter, or salty or umami tastes. Parenthetically, no sour-selective hot spot was identified. Because neurons activated by a given taste stimulus formed a localized cluster, this is also referred to as spatial or topographical coding.

The report of Chen et al. (2011) was followed by a study where channelrhodopsin was expressed in neurons belonging to either the sweet or the bitter “hot spots” in cortices of mice. These neurons were then stimulated optogenetically in unanesthetized, freely behaving animals while monitoring taste behavior (Peng et al. 2015). When the “bitter spot” was excited, mice exhibited aversive taste responses (stopping licking from a water spout, began gagging and attempted to clean their mouth). In contrast, when the “sweet spot” was optogenetically excited, mice avidly licked at the water spout even if it delivered a dilute quinine solution that mice normally avoid. In short, mice exhibited preferred taste behaviors when the “sweet spot” was optogenetically excited and conversely, showed aversive behaviors when the “bitter spot” was stimulated. The authors interpreted their findings to indicate “individual basic tastes are represented in the (brain) by finely tuned cells organized in a precise and spatially ordered gustotopic map” (Chen et al. 2011). That is, this report posits an extension of labeled line coding wherein neurons that are finely-tuned to one of the basic tastes (i.e., “labeled”) are clustered into discrete hot spots. A recent report from the same group extended these earlier findings and purported to show that bitter or sweet cortical hotspots exert descending positive and negative feedback onto ascending brainstem taste signals (Jin et al. 2021). These reports have propelled labeled line/topographic coding of taste into the spotlight. Many modern textbooks cite these references and strongly promulgate labeled line/topographical coding as how the brain encodes taste signals.

Yet, historically, microelectrode studies in a wide variety of species have shown that CNS neurons recorded individually are not selectively tuned to a given taste quality and that there are no focal regions of selectivity for sweet, bitter, salty, etc. Indeed, gustatory cortical neurons respond broadly to different taste stimuli – their entropy (H) is high (Spector and Travers 2005; Yamamoto et al. 1984). This is not consistent with the presence of cortical “hot spots” where only one taste quality is represented. A possible “workaround” for this conundrum is that although cortical neurons may respond broadly to several tastes, they have a preferred (“best”) taste stimulus to which they respond most robustly. Other tastes only weakly excite the neuron. That is, even though they are multiply-responsive (have high signal entropy), the ratio of “best taste” response to other taste responses (signal-to-noise ratio) (Spector and Travers 2005) is high. In such a scenario, tasting a sweet compound (e.g., sucrose) might robustly activate a specific patch of neurons in the cortex (“hot spot,” labeled line/topographic coding) even though bitter, salty, umami, or sour taste compounds also activate those same neurons to a lesser extent. Calcium imaging, especially in anesthetized animals, might not faithfully report these differences and only reveal the one, most robust signal.

But more challenging to the report of taste-specific gustatory cortical “hot spots” and spatial (topographic) coding of taste has been the failure of other laboratories to replicate the original calcium imaging findings. On the contrary, other studies which also used Ca²⁺ imaging have not found distinct taste-specific regions in the gustatory cortex or rats or mice, anesthetized or behaving (Accolla et al. 2007; Carleton et al. 2010; Fletcher et al. 2017). Indeed, the most recent study – an exceptionally careful and detailed study using sophisticated two photon calcium imaging on awake, behaving mice – revealed a sparse, distributed representation of taste responses – including finely- and broadly tuned neurons – with no evidence for localized and segregated “hot spots” (Chen et al. 2021) (Fig. 16). This discrepancy is striking and strongly supports the notion that taste coding in the CNS is more complex than a topographic, labeled line.

Fig. 16.

The code for taste is sparsely distributed across the mouse gustatory cortex, with no apparent “hot spots.”(a) schematic showing the prism positioned on the surface of the gustatory cortex with the microscope objective used for two-photon Ca²⁺ (GCaMP6f) imaging. (b) widefield image of the cortex, showing the expression of GCaMP6f (white) and the middle cerebral artery (MCA). (c) two-photon image from the field marked in (b). (d) representative map of neurons with best responses to sucrose (red), NaCl (blue), citric acid (orange), quinine (purple), and water (black). N.B., There is no obvious clustering (“hot spots”) of taste responses. Reproduced here with permission from Chen et al. (2021)

If the code for taste information in the gustatory cortex is not particularly a labeled line, spatial, representation, then what are the alternatives? The leading concept is that central coding of taste signals is based on distinctive rhythmic neuronal firing patterns in dispersed circuits of gustatory neurons. The concept that the pattern of impulses might encode taste was mentioned above vis-à-vis activity in peripheral afferent fibers and was postulated for the CNS decades ago (Erickson 1963; Erickson et al. 1994; Johnson and Covey 1980; Scott and Mark 1986). Momentum for a central gustatory temporal code gained strength with publications by Di Lorenzo’s group (Di Lorenzo and Hecht 1993; Di Lorenzo and Victor 2003). These researchers reported that sweet, sour, salty and bitter stimuli could be discriminated by the temporal patterns of taste-evoked neuronal firing in the rat Nucleus of the Solitary Tract (NTS). This is similar to the finding that had been reported for the peripheral taste system some decades previously (Covey and Erickson 1980). Importantly, merely applying patterned electrical stimulation to the NTS that mimicked neural firing to bitter or sweet taste solutions elicited taste behaviors appropriate for those tastes (Di Lorenzo et al. 2009b; Di Lorenzo and Victor 2003).

Yet, although temporal coding of taste in the brain seemed promising, a major limitation was that for the most part, neuronal activity was measured only from one or a very few neurons at a time. Katz et al. (2001) made a seminal breakthrough by recording taste-evoked activity in ensembles of neurons in awake, behaving rats with multiple recording microelectrodes implanted in the gustatory cortex. Those investigators identified taste-specific temporal patterns, or rhythms, of responses across ensembles of neurons in the 2–3 s following taste stimulus presentation in the oral cavity. Subsequent publications refined the analyses by using sophisticated statistical analyses (Hidden Markov modeling) to demonstrate that taste identification in the gustatory cortex evolves rapidly with shifting rhythms of firing within neuronal ensembles (Miller and Katz 2010; Moran and Katz 2014; Sadacca et al. 2016). The Ca²⁺ imaging findings of Chen et al. (2021), described above – who demonstrated sparse coding of taste in the gustatory cortex of mice – are entirely consistent with this concept of rhythmic firing within neural circuits.

In sum, modern technical advances have allowed researchers to investigate taste-evoked activity in ensembles of gustatory cortical neurons with high spatial and temporal resolution in experimental animals. There is abundant evidence for temporal coding of taste in the gustatory cortex of experimental animals (Stapleton et al. 2006). Studies firmly show that taste identification in the cortex is encoded by changing patterns of rhythmic firing across sparse neural circuits. This is nearly the antithesis of labeled line coding via “hot spots.”

Importantly, recent studies in human subjects tend to come to similar conclusions regarding taste coding in the brain – that taste identification involves temporal coding without well-defined, taste-specific “hot spots” (Avery et al. 2020; Canna et al. 2019; Wallroth and Ohla 2018); but see Porcu et al. (2020) (Fig. 17). Notably, Avery et al. (2020) utilized ultra-high resolution fMRI and reported that their results “suggest that taste quality is not represented topographically, but by a distributed population code.” Excellent topical reviews of studies on how the brain encodes taste identity in experimental animals and in humans have been published (Avery 2021; Boughter and Fletcher 2021; Di Lorenzo et al. 2009a; Hallock and Di Lorenzo 2006; Lin et al. 2021; Ohla 2021; Spector and Travers 2005).

Fig. 17.

The code for taste in the human gustatory cortex is not segregated into distinct, taste-specific “hot spots.” Sucrose and quinine were presented orally to subjects at several different concentrations, ranging from high (red) to low (blue) concentration during functional magnetic resonance imaging (fMRI). At no stimulus intensity (tastant concentration) were distinctly separate cortical regions activated for sucrose versus quinine. Modified from Canna et al. (2019), with permission

3.10. Gustatory Stimulus Intensity Coding in the CNS

As discussed previously, taste intensity coding in the periphery appears to be rather straightforward. A stronger taste stimulus (i.e., more concentrated solution or lower pH for acid/sour) produces a larger signal in the receptor cells and innervating sensory neurons. However, as pointed out by Wu et al. (2015), in addition to the increased signal amplitude, a stronger taste stimulus increases the breadth of tuning, or “noise” in gustatory sensory ganglion neurons. In the CNS, encoding stimulus intensity becomes much more complex. There is some indication in the first central relay station (Nucleus of the Solitary tract), that in addition to a larger response amplitude, increasing the stimulus intensity (i.e., increasing the taste solution concentration) shows an altered temporal response (latency, time-to-peak, and decay; Schwartzbaum and DiLorenzo 1982). Increased stimulus intensity also produces a greater breadth of response in at least some NTS neurons (Geran and Travers 2009).

However, a more enigmatic coding of taste intensity occurs in higher brain centers, namely the insular cortex. Increasing stimulus concentration does not simply increase firing rates of neurons but has a more complex effect. Some neurons decrease firing rate with increased taste stimulus concentration, and conversely, other neurons show a monotonic increase in firing rate with increasing stimulus (MacDonald et al. 2012; Stapleton et al. 2006). Further, neuron action potential timing can convey information about stimulus intensity (Fonseca et al. 2018). Intriguingly, increasing taste intensity even appears to activate different ensembles of neurons, as if different brain circuits are recruited with increasing stimulus intensity (Canna et al. 2019; Porcu et al. 2020).

4. Summary and Caveats

Regrettably, this overview overlooks certain aspects of a taste coding, especially with regard to taste signal processing in the CNS. For instance, the importance of convergent, non-gustatory sensory input (e.g., texture, temperature, astringency, olfaction) that contributes to taste identification is not discussed. Neither is there a mention of the critical importance of learning, attention, and expectation. The effect of active versus passive stimulus presentation on the decoding of taste information in the CNS is ignored. The role of hunger or satiety was not discussed, and how taste hedonics or valence (pleasurable versus aversive taste) might be encoded in the CNS was overlooked. These are all important factors that modulate, regulate, or change how neural ensembles in the CNS respond to a given taste stimulus at a given concentration, that is, how the brain encodes taste. Yet, such a global, comprehensive review is beyond the scope of this outline of taste coding. Excellent reviews that present the history of ideas concerning taste coding can be found in a special issue of Physiology & Behavior (2000), volume 69, issue 1 (https://www.sciencedirect.com/journal/physiology-and-behavior/vol/69/issue/1), and Spector and Travers (2005). More recently, Current Opinion in Physiology (2021), (https://www.sciencedirect.com/journal/current-opinion-in-physiology/special-issue/10ND3QC7M6R) includes a collection of articles on taste coding in the brain that touches upon some of these important issues.

As a final word, it must be acknowledged that there remains today a heated controversy regarding how taste signals are encoded at all levels along the taste axis, from peripheral end organs to the highest levels of cortical processing. This controversy has existed since the first recordings of afferent taste nerve activity (Pfaffmann 1941; Zotterman 1936). Importantly, as noted in this review, a number of prominent reports in the more recent decades adamantly promote the concept that taste can be explained by “labeled line” coding and the existence of tastant-specific “hot spots” (spatial coding) in the CNS (Barretto et al. 2015; Lee et al. 2017; Mueller et al. 2005; Wang et al. 2018; Yarmolinsky et al. 2009; Zhao et al. 2003). Many modern textbooks have adopted these concepts in their chapters on taste (Kandel et al. 2013; Purves et al. 2017; however see Bear et al. 2016). Labeled line and spatial coding are simple and straightforward concepts, hence their attraction for the lay public, students, and scholars. It may be noteworthy that the aforementioned research reports promoting labeled line/spatial coding come mainly from the same group; the findings await validation by other laboratories. Moreover, as this chapter has attempted to show, the vast preponderance of data from researchers studying taste and using multiple different techniques strongly argues against any simple labeled line or spatial coding paradigm at any level of the taste axis. The huge dilemma is that if labeled lines and spatial codes cannot explain taste, how then does the nervous system deal with gustatory signal processing? These pages have argued for some form of “combinatorial” or “population coding,” or rhythmic activation of diffuse neural networks across time and space in the cortex encode taste. But these are vague concepts and difficult to pin down. It remains to be determined in detail how the nervous system analyzes and parses peripheral and central gustatory signals.