Primary olfactory neurons are, under normal circumstances, the only part of our central nervous system that is exposed to the outside air. Each neuron carries on its surface one of a large number of different protein receptors. An even larger number of volatile drugs called odorants are capable of turning on these receptors. The result is the sensation of smell. Two things are peculiar about smell. First, the sensation unerringly discriminates among odorants. As far as we know, no two odorants in the pure state smell the same to us. Second, it has turned out to be impossible to predict the smell of a molecule from its molecular structure.
How is the character of an odor, loosely defined by the words used to describe its smell, written into the molecule? What properties of the molecule determine its odor? The most commonly accepted theory is that odor is not written into the molecule, but arises because the odorant happens to fit the shape of its many receptors. If you could by some means change the specificity of each receptor, the perceived odor would be completely different. This makes sense. Lock and key molecular recognition is everywhere in biology; antibodies and enzymes would not work without it.
Two assumptions are at work here. The first is that receptors work entirely by a lock and key mechanism. The free energy of a ligand binding to a receptor alters the receptor’s energy landscape. An on conformation is favored by agonists and an off conformation favored by antagonists. To be potent and specific, the ligand must interact with the receptor via several noncovalent interactions, and must be arranged in a distinctive spatial pattern. The second is that olfaction is combinatorial. If both assumptions are correct, then even if receptors have only two positions, N receptors will be able to distinguish 2N smells. For fruit flies, this would mean 263, or approximately 1019 smells.
There are sixty-three receptors in fruit flies, approximately four hundred in human beings, and close to a thousand in mice. Clearly, any mechanism by which odorants could differentially stimulate receptors would do, provided that the brain has enough processing power to treat the signal; what we know of our other senses does not suggest that processing power is in short supply.
To most biochemists, the notion that a vast number of olfactory receptors results in an effectively infinite range of sensations therefore poses no conceptual problem at all. It is small wonder that we cannot predict odor from molecular structure. The task would be difficult enough if there were one receptor involved, as is the case when designing drugs. When there are dozens of receptors, the problem is simply intractable. Whatever the steps from molecular structure to odor, they cannot be followed in reverse. If this is correct, all we can do is retreat and declare victory, which is what mainstream thinking in the field appears to be doing.
Odorants are not typical drugs, and if they were, they would not be odorants. In order to be volatile, odorants must feature only a few sticky chemical groups, such as hydrogen bond donors and acceptors.1 Otherwise they would stick both to each other and to their substrate. This characteristic gives odorants limited solubility in water. But sticky groups are required for tight and specific binding; odorant affinities for receptors will typically be micromolar, approximately a thousand times lower than those of proper, water soluble, nonvolatile drugs. The specificity of odorants for receptors is correspondingly low.
The Database of Odorant Responses compiled at the University of Konstanz lists identified receptors in the fruit fly and the molecules known to turn them on (or off).2 What is striking is the vast range of odorants to which each receptor responds. If a particular odorant generates a significant receptor response, it would be tempting to call the excited receptor the receptor for that odorant. But that would be foolish, since another as yet untried odorant might always give a still larger response.3 Conversely, some odorants inhibit the receptor. This pattern is true of all fruit fly receptors, and there are strong indications of something similar occurring in vertebrate receptors.
The notion of a combinatorial on-off pattern was therefore incorrect. Most odorants activate most receptors to some extent, and if there is a pattern, it must lie in the relative degree of excitation. It is very hard to estimate the discriminatory ability of such a system. Exciting many receptors promiscuously, certain odors may reduce a system’s discrimination, while a system’s ability to make use of different degrees of excitation may increase it. The balance between the two determines how many odorants can be discriminated. Given our current state of knowledge, we cannot estimate, even approximately, this balance. Color vision, for example, uses just three analog channels to discriminate tens of thousands of colors. The problem of relating structure to odor seems ever more forbidding.
Reviews of aromachemical structure-odor relations are mostly catalogues de curiosités, and emphasize how small modifications to their structure often give rise to large changes in their odors. But these qualifiers are left undefined, since structural distance depends on the relative fit to receptor structures, which are unknown, while odor character is largely unmapped. I remember once smelling a compound that was distinctly ketonic (nail varnish) at high dilution, breadlike at higher concentrations, and very much like curry at the highest concentration. Until then I had never thought of these three characters being adjacent, but one can, it seems, move continuously from one to the other by a change in concentration.
It is not surprising that most fragrance chemists regard the problem of structure-odor relations as unsolvable. Biologists assume it might be solvable in theory, if only we had all the three-dimensional structures of all the receptors, and could calculate how they fit the odorants. Synthetic chemists and biologists seem to agree that this is not an urgent question.4
There are, however, some remarkable regularities in olfaction that have bothered curious researchers for decades.
While frying onions, Enrico Fermi once exclaimed, “Wouldn’t it be nice to know how smell works?” Fermi made that remark about onions, a vegetable whose genus is dominated by sulfur and its smells.5
Perhaps the best known structure-odor equation of all is:6
–SH = rotten eggs
No matter its structure, if a molecule contains an –SH group, a rotten egg character is superimposed upon the molecule’s other smells. If the molecule smells of pine needles, the total will smell of eggy pine needles, such as in the case of pinanethiol, a grapefruit odorant. Conversely, with one notable exception to which we shall return, no molecule devoid of sulfur ever smells of rotten eggs.
Given just the odor, how do we know sulfur is there?7 A physicist friend jokingly offered that we know sulfur is in there because it is yellow. This may not be far from the truth.
If –SH were the only functional group we could recognize by smell, we could legitimately construct an ad hoc explanation, and it would fit many of the known facts. –SH is a weak acid. It readily dissociates to –S- and H+. And the sulfide anion binds very tightly to metal ions. One could therefore imagine a specific receptor containing one such metal ion, dedicated to sensing all sulfides. No more mystery.
But the situation is more tangled than that. Our ability to detect the presence of distinctive groups of atoms—called functional groups because in large part they determine the chemistry of the molecule—extends far beyond –SH. The very language of chemistry is replete with illustrations of this; chemists have long described odors as nitrilic (-C≡N), ethereal (-C-O-C-), aldehydic (-(H)C=O), ammonia-like (-NH2). To these a perfume chemist might add the distinctive odor character of molecules containing an oxime (-NOH) and an amide (-C=O-NH-) group.
No receptor could unerringly detect these functional groups, whatever the molecular context, by conventional molecular recognition. They are too small to have a distinctive shape, and they only interact through one, or perhaps two, hydrogen bonds. Are our noses somehow reading the atomic composition of the molecule?8
How might a chemist identify an unknown molecule? Before the advent of nuclear magnetic resonance spectrometers in the 1970s, the instrument of choice was the infrared spectrometer. Its principle of operation is simple. Even when a molecule has no overall electric charge, its component atoms are not quite electrically neutral. Electrons are pulled this way and that, like a too-short blanket that leaves either your feet or your shoulders out in the cold. A typical HCNO odorant might then have partial charges on its component atoms amounting to a fraction of the electron charge e, typically of the order of a tenth of an electron charge. In a C=O group, for example, the oxygen would carry -0.2 e and the carbon +0.2 e. The C=O could then react to electric fields. In particular, it would absorb energy from oscillating electric fields at the resonant frequency of the C=O stretch vibration. This resonant frequency is determined by the strength of the bond and the masses of the atoms at either end, and is usefully diagnostic of the functional group. For –C=O, it lies typically around 1,700 cm-1, for ‑C≡N, it is around 2,150 cm-1, for –SH, it is around 2,550 cm-1, and so on.
The only other chemical bond that vibrates at the –SH frequency is the –BH bond of boranes. It also smells of rotten eggs, a fact which lay buried in the literature for eighty four years until I rediscovered it.9 –BH does not bind to metals.
–BH = rotten eggs
It was a chemist, Malcolm Dyson, who in 1938 first proposed the radically strange idea that the nose was a vibrational spectroscope, and that this is why it can sense functional groups.10 Just as importantly, vibrational spectroscopy can easily distinguish between two molecules that bear a functional group attached to the same atoms, but that are arranged differently. For example, propanol and iso-propanol are both three-carbon alcohols (–OH functional group) with the –OH attached either at the end or in the middle. They have the same composition (C3H8O), but different smells. The former smells like ethanol, while the latter has the familiar, much more pungent smell of rubbing alcohol. They have the same –OH stretch frequency, but the rest of their vibrational spectrum, namely the part involving movements of the carbon atoms, is very different. This part is governed by the topology of the molecule, and experiments have shown that it is unique to each molecule. The part of the spectrum bearing collective vibrations of the whole molecule is called the fingerprint region.
The entire vibrational spectrum covers the range from 0 to 4,000 cm–1. The fingerprint region is in the lower half of this range, while most of the functional groups are in the upper half. If the nose were indeed a vibrational spectroscope that could read both halves, the mystery of olfaction would be solved.11 Odor character is written into the molecule, and large numbers of receptors are required merely to ensure that every odorant binds to enough of them to be properly analyzed by a spectrometer.
In the eight decades since Dyson first proposed it, this idea has fascinated a few scientists, myself included. It has elicited horror in others. Those who love it are smitten with the idea that evolution could figure out how to do on a nanoscale what usually takes up two square meters of bench space. They also love the idea that three of our five senses analyze vibrations to tell us about the world around us. Neither is a properly scientific argument. Those who hate it had a far more cogent reason to do so. There was for a long time no plausible mechanism to make a biological spectroscope.
In 1993, I came across electron tunneling spectroscopy.12 When electrons jump from one place to another, energy must be conserved. If they jump from a higher to a lower energy, some energy must be lost in the process, usually in the form of heat or atomic motion. Under the right conditions, it is possible to confine this heating to a single vibrational mode of a molecule.13 This is achieved by placing the molecule between high and low energy sites. If the molecule possesses a vibrational mode at an energy corresponding to the energy the electron must lose in the jump, the electron is forced to jump. The fact that the electron disappears on one side of the molecule and reappears on the other, while exciting a vibration, is a quantum mechanical effect.
Electrons are plentiful in biology.14 They mostly live in mitochondria, formerly free-living bacteria that we have coopted into making electrochemical energy for us. Almost all the oxygen we breathe is turned into water by mitochondria, which add two electrons and two H+ ions to each oxygen atom, making four electrons for an O2 molecule. It is easy to calculate how much electron current is flowing through our cells. Multiply oxygen consumption, in atoms, by four. It turns out the numbers are huge; the resting human being harbors 100 amperes of electric current, and an athlete at full stretch reaches 1,000A.15 The average wall plug is fused at 15A.
Twenty years ago, it seemed to me that it was not beyond the reach of evolution to steal a few microamps of this current to power olfactory receptors. I proposed that receptors were quantum switches; odorants would bind to them, and once bound would be probed by electrons flowing through the receptor. Receptors, like the color-sensitive cells in the retina, would be divided into classes that sensed different parts of the vibrational spectrum by tuning the electron energy gap. If, once bound, the odorant vibrated in the region of the spectrum probed by that receptor, the receptor would tell the brain that a molecule possessing that vibration had reached the nose. Smell would then be a sort of infrared color vision triggered by odorants. If the specificity of each receptor is changed, the perceived smell remains the same, provided on average enough receptors in each spectral class bind the odorant and probe its vibrations. In the shape theory, odor is written into the receptors. In the vibrational theory, odor is written into the odorant.
How might one test the vibrational theory? Simple in principle, tricky in practice.
An obvious way would be to replace some of the atoms in the odorant with heavier isotopes of the same element. For a first approximation, if we think of a molecule as being made up of masses (the nuclei) connected by springs (the two-electron bonds), then isotope substitution would change the masses and leave the springs unaltered. There are 3N-6 vibrational frequencies for a molecule made up of N atoms. All of them shift downwards, but the lowest energy configuration remains unaltered. This trick has been used for decades in biochemistry, with radioactive isotopes. In our case, there is no need to use radioactive elements.
The bigger the fractional change in atomic weight, the bigger the effect on the vibrations. Putting deuterium in place of hydrogen doubles the mass.16 Molecules whose atoms differ in atomic mass are called isotopologues.
The question now becomes: do isotopologues differ in smell?
The answer is yes. Humans easily distinguish a deuterated musk from the normal one.17 By putting in a nitrile group, which has the same stretch vibration, biochemists have been able to fool flies into believing deuterium is present in a molecule.18 Evidence from honeybees indicates that they are sensitive to vibrational differences between deuterated odorants and normal ones; deuterated odorants inhibit receptors, and normal ones excite them.19
Has this convinced everyone that olfaction detects vibrations? Not quite.
The objections come in three flavors. First, it might be the case that there is some small, non-vibrational difference between the deuterated and normal molecules. Second, the difference might be caused by an impurity in the isotopologue. Third, behavioral tests are meaningless anyway.
The first objection is the hardest to answer. There are, indeed, very small physicochemical differences between normal and deuterated molecules, such as their solubility in oil. It is impossible to say that these are too small be responsible for a difference in odor, though they do not account for the honeybee experiments above or for the fly experiments involving nitrile and deuterium. Impurities also cannot explain the musk experiments on humans, which used pure odorants.
The third objection is typical of olfaction as a field. Scientists are not perfumers; many seem to believe that odor is completely subjective and irrelevant to scientific inquiry. This is nonsense. Nobody would dream of suggesting the same about color or sound.
I am simplifying the terms of the often intemperate debate because I have, over the years, come to think that the roots of the problem lie elsewhere. The question cannot be settled as posed.
What bothers opponents of the vibrational theory is that it proposes a novel mechanism for receptor activity.20 This may have implications far beyond olfaction.
Leslie Vosshall of The Rockefeller University, the most vocal opponent of vibrational olfaction, has stated it most clearly: “Why should odorant receptors be exceptional?”21 Vertebrate olfactory receptors belong to the vast class of G-protein coupled receptors (GPCRs), which includes many neurotransmitters and hormone receptors.22 They work by binding to a molecule on the outside of a cell and relaying the information to specialized intracellular proteins, the G-proteins, which in turn pass on and amplify the message to intracellular signaling systems. The exact mechanism by which the receptors transmit the message to the G-protein is unknown, but nobody has ever suggested that vibrations or electrons are involved.
Why would olfactory receptors be so different?
One could reply that evolution does what it can. After all, GPCRs derive from an ancient light-powered bacterial ion pump called bacteriorhodopsin; yet nobody is arguing that all GPCRs should be light-powered proton pumps.
But there may be a more interesting answer. Forget vibrations for a moment, and consider what a receptor is trying to do. The sequence of biochemical reactions triggered by a receptor is called the signaling cascade. It is what an electronic engineer would call a series of gain stages. Each amplifies the signal it receives and passes it on to the next stage. An efficient design might incorporate gain in the sensor itself; in chemical field-effect transistors (ChemFETs), the transistor gate is exposed to the outside world and the adsorbed molecules modulate electron transport through the transistor.
Might the same be happening in biology? This is not a completely new idea; electron flow in receptors was suggested in 1968 by a brilliant, but little-known, biophysicist, Freeman Cope.23
There is nothing outlandish about this notion. For a start, all the bits and pieces to make it work are already in the evolutionary parts bin. The cell is awash with soluble electron donors (NADH and NADPH among others). Membrane proteins are capable of transporting electrons over the required multi-nanometer distances. And redox control, the electrochemical control of membrane receptors, is well-documented.
Suppose that evolution has made use of these elements to build an electron receptor, with gain. How might that work in the case of other GPCRs? Over the last couple of years, I have—belatedly—realized that a number of known facts fit this idea nicely. This seems like a good time and place to present them, and to connect the dots with a unifying hypothesis. There is something peculiar about most neurotransmitters and drugs. They are positively charged. Acetylcholine, the indoleamines, such as serotonin, and catecholamines, such as adrenaline, all carry a positively charged nitrogen atom at one end. Most naturally occurring drugs acting on neurotransmitter receptors carry a positive H+ sitting on a nitrogen atom. It is why they are called alkaloids.24
A positive charge is exactly what would provide gain for an electron tunneling receptor. Imagine an electron trying to tunnel through a barrier from a filled site to an empty site, at the same energy. No vibrations are excited. The width and height of the barrier determine the rate of electron flow. It is known from experiment and understood in theory that lowering the barrier height or decreasing its width increases the tunneling rate exponentially. Since the electron has a negative charge, a negative charge in the barrier would block its movement, while a positive charge would greatly facilitate its transfer. A positive charge is exactly what would be expected from an electron tunneling switch. It may account for the evolutionary design of, say, acetylcholine with its permanent, quaternary ammonium charge on the choline.
But there is another twist to this story. Many neurotransmitters are themselves electron donors, notably the catecholamines. Catechols, with their two OH groups attached to a benzene ring, are easily oxidized; they can lose two electrons and two H+ ions to oxygen and become quinones.25 What keeps catecholamines marginally stable against electron loss in solution is their positive charge. Add a drop of concentrated alkali to a solution of dopamine to make the molecule neutral, and it promptly turns pink, then black, as the light-absorbing quinones develop and polymerize.
I propose that Class A (rhodopsin-like) GPCRs come in three versions, each having an electron circuit with a different topology.
Type V (vibration) receptors have a tunneling circuit traversing the receptor with an energy jump across the tunneling gap. To turn on such a receptor, the molecule has to be bound and possess one or more vibrations at the correct energy.
Type T (tunneling) receptors have the same circuit topology, but without an energy jump. The receptor is turned on when a molecule binds to it and includes a feature, such as a positive charge, that lowers the barrier to electron tunneling.
Finally, type R (redox) receptors have only the output half of the electron circuit. The electrons come from the bound molecule itself, which undergoes an oxidation step when bound. This is facilitated if the receptor contains a negatively charged group able to cancel the stabilizing positive charge of the electron donor.
If correct, this idea could have interesting implications. Shape-based theories treat receptors and their ligands as mechanical devices. If it turns out instead that drugs are electronic devices, it might explain why structure-activity relations make so little sense in general, and why drugs are discovered by accident.
Put simply, we may have been looking at a vast catalogue of diodes, resistors, and transistors, all the time wrongly thinking they were spanners, screwdrivers, and keys. Testing this idea is a challenge. Direct evidence of electron currents in receptors will be required. I and my colleagues have some ideas for experiments that would test this possibility. They are too tentative to be aired at this point, but I hope to be able to report on progress.
