Smell and Taste

Most mature olfactory neurons in a mouse express one odorant-receptor gene at high abundance, chosen from roughly a thousand, and transcribe it from only one of its two copies.

Smell and taste both turn chemicals into spikes, yet they solve the coding problem in opposite ways. The nose has a huge receptor family and reads an odour from the pattern of receptors it activates. The tongue has a few receptor types in cells that each report one quality. This lesson takes each in turn and then asks what experiments tell the two strategies apart.

One neuron, one receptor, two glomeruli

Olfactory sensory neurons sit in the olfactory epithelium high in the nasal cavity. Each sends cilia into the mucus, where odorants bind, and an axon through the bone above to the olfactory bulb. There the axons end in glomeruli, round knots of neuropil (tangles of axons and dendrites) where receptor-neuron axons meet the dendrites of the bulb’s output neurons, the mitral and tufted cells.

In 1991 Linda Buck and Richard Axel described 18 members of a gene family from rat olfactory tissue that encoded proteins with seven transmembrane segments, the shape of G-protein-coupled receptors. Their title kept a hedge: the family may encode odorant receptors. The paper did not show what any receptor binds or where its neurons project. Later genome work put the family at roughly a thousand genes in the mouse and a few hundred intact genes in humans; how the family grew and shrank belongs to molecular evolution.

When an odorant binds, the receptor activates the G protein Golf, which drives adenylyl cyclase 3 to make cAMP. cAMP opens cyclic-nucleotide-gated channels that let in sodium and calcium, and a calcium-activated chloride current amplifies the depolarization. The molecular steps are in senses and cGMP, G proteins and cAMP and PKA.

Which receptor does a neuron use? In the mouse neurons Chess and colleagues examined in 1994, a given receptor gene was transcribed from one allele. Single-cell sequencing in 2015 (Hanchate and colleagues) found that most mature mouse olfactory neurons express one receptor gene at high abundance, while many immature neurons express several at low levels. Note what this rule is about. One receptor per neuron is a statement about genes, not about odorants: that one receptor can still bind many odorants.

Where do the axons go? Mombaerts and colleagues (1996) engineered mice so that neurons expressing one chosen receptor also made a marker that filled their axons. Those axons, scattered across the epithelium, converged on two fixed glomeruli in each bulb, out of about 1,800 mouse glomeruli. Swapping in a different receptor’s coding sequence moved the target, which suggested that the receptor helps guide its own axon without being the only cue. Each glomerulus therefore collects mainly from neurons expressing one receptor type, and the glomerular layer is largely a map of receptor identity. How that map is built is covered in partners, layers and self-avoidance; olfactory neurons are also replaced throughout life (adult neurogenesis).

From the bulb, mitral and tufted cells project to the piriform cortex and other olfactory areas without first relaying in the thalamus, unlike vision, hearing and touch, though a route through the thalamus to frontal cortex also exists.

Reading a combinatorial code

If each neuron has one receptor, what does each receptor detect? Malnic, Hirono, Sato and Buck (1999) loaded dissociated mouse olfactory neurons with a calcium indicator, applied a panel of odorants, and then identified the receptor in each responding cell by single-cell RT-PCR. One receptor recognized several odorants. One odorant was recognized by several receptors. And changing an odorant’s concentration could change which receptors it activated. Odour identity, on this view, is a combinatorial code: a pattern across receptors rather than one receptor per odour, a population code of the kind set out in sensory coding.

How much can such a code carry? If you treat each of N receptor types as a switch that is either on or off:

number of distinct non-empty patterns = 2^N − 1

Worked problem: a made-up receptor matrix. Suppose you imaged mouse olfactory neurons and found four receptor types, A to D. These data are invented for illustration. At a low concentration, odorant 1 activates A and B; odorant 2 activates B and C; odorant 3 activates A, B and C; odorant 4 activates D only. Assume each receptor is simply on or off, and that the brain can tell two odorants apart only if their patterns differ.

Step 1: capacity. Four switches give 2^4 = 16 patterns. Removing the all-off pattern leaves 15 that could report an odorant.

Step 2: remove receptor B. Odorant 1 becomes A; odorant 2 becomes C; odorant 3 becomes A+C; odorant 4 stays D. All four patterns are still different, so the model predicts no loss of discrimination.

Step 3: remove receptor A instead. Odorant 1 becomes B; odorant 2 is B+C; odorant 3 is also B+C; odorant 4 is D. Odorants 2 and 3 now give the same pattern, so the model predicts they can no longer be told apart.

Step 4: restore all four receptors and raise the concentration. At tenfold concentration, odorant 1 also recruits C, a lower-affinity receptor. Its pattern becomes A+B+C, the low-concentration pattern of odorant 3. The model predicts that odorant 1 at high concentration could be mistaken for odorant 3: a change in quality with concentration.

Two cautions. The 2^N figure is the capacity of a toy model, not a count of smells: real receptors respond in graded amounts, responses are noisy, and the bulb and cortex transform the pattern. And predictions about what a mouse can tell apart need a behavioural test; a matrix alone cannot settle them.

Five qualities, several transduction routes

Taste starts in taste buds, clusters of specialized epithelial cells, not neurons, on the tongue and palate. Afferent fibres of cranial nerves VII, IX and X carry their signals to the nucleus of the solitary tract in the medulla, which relays through the thalamus (in rodents by way of the pontine parabrachial nucleus) to gustatory cortex in the insula. The textbook map with sweet at the tip and bitter at the back is discredited: taste buds across the tongue detect every quality, with modest regional differences in sensitivity.

The five qualities taught for mammals are sweet, sour, bitter, salty and umami. They use three kinds of cell.

Type I cells have glial-like supporting properties.

Type II cells carry the GPCRs. Sweet cells express T1R2 with T1R3; umami cells express T1R1 with T1R3; bitter cells express T2Rs. In heterologous cell assays, human T1R2/T1R3 responded to diverse sweeteners and human T1R1/T1R3 to L-glutamate, with 0.2 mM IMP shifting the glutamate EC50 (the concentration giving a half-maximal response) about 30-fold lower (Li and colleagues, 2002). Meyerhof and colleagues (2010) screened 25 human bitter receptors against 104 bitter compounds and found that receptors varied greatly in how many compounds activated them.

Sweet, umami and bitter cells share the steps after the receptor: PLCβ2 makes IP3, IP3R3 channels release calcium from the endoplasmic reticulum, and calcium opens TRPM5 (see PLC, IP3, DAG and PKC). TRPM5 is gated by calcium but passes monovalent cations, so it depolarizes the cell. In mice, TRPM4 contributes too (Dutta Banik and colleagues, 2018). The depolarized cell has no conventional vesicular synapse; instead it opens the voltage-gated CALHM1/3 channel, and ATP flows out through it onto the afferent fibres.

Worked problem: one TRPM5 channel. Prawitt and colleagues measured a unitary conductance of 25 pS for human TRPM5 expressed in HEK-293 cells. Assume a linear current-voltage relation, a membrane potential of −60 mV and a reversal potential of 0 mV, as for a non-selective cation channel. Then I = g × (Vm − Erev) = 25 pS × (−60 mV − 0 mV) = −1,500 fA = −1.5 pA. The minus sign means inward current of positive charge. This is a stipulated example, not a measured taste-cell current.

Type III cells detect sour and form conventional synapses, releasing amine transmitters such as serotonin by calcium-dependent exocytosis. Their acid sensor is OTOP1, a proton-selective channel identified by Tu and colleagues in 2018. In Otop1-null mice, gustatory nerve responses to acids were severely and selectively reduced (Teng and colleagues, 2019). PKD2L1, long used to mark sour cells, is not the primary sour receptor.

Salty is the least settled. In mice, deleting the ENaCα subunit of the epithelial sodium channel from taste cells abolished responses to low sodium concentrations and their block by amiloride (Chandrashekar and colleagues, 2010). High salt, which mice avoid, also recruited sour and bitter cells while the ENaC pathway stayed active (Oka and colleagues, 2013). How human salt taste is transduced is unsettled. Fat and kokumi are candidate qualities, not established ones, and much of what people call flavour is smell from food in the mouth reaching the nose from behind (retronasal smell).

Two answers to one coding problem

Smell meets an open-ended world of molecules with a large, overlapping receptor array: quality is read from a pattern. Taste mostly answers a few questions (eat it or reject it, and what it contains) with cells dedicated to each quality: quality is read from which cell fires, a labeled line at the periphery.

The strongest evidence that the cell, not the receptor or the cascade, sets taste quality comes from mouse genetics. Zhang and colleagues (2003) removed PLCβ2 and lost the tested sweet, amino-acid and bitter responses, while sour and salty were spared. Restoring PLCβ2 only in bitter-receptor cells brought back bitter responses without sweet or amino-acid responses. Zhang and colleagues (2019) went the other way: putting OTOP1 into sweet cells made those cells respond to sour stimuli.

Keep the scope. These are mouse experiments on the periphery; they do not show one exclusive wire for each human percept, and high salt shows that one stimulus can recruit two cell types.

Design problem

Is “bitter” a property of the receptor or of the cell carrying it? Design an experiment in mice that tells the two apart.