When neuroscientists David Hubel and Torsten Wiesel wanted to figure out how the brain parses its visual environment, they went as simple as they could go. In a Harvard lab crammed with electrical equipment, they positioned cats in front of a screen and showed them extremely basic images: dots in particular locations, lines at various angles. At the same time, they used implanted electrodes to, quite literally, “listen” to neurons in the areas of the brain devoted to vision. By observing which neurons fired in response to which shapes, they were able to unlock a part of the brain’s “visual code,” the way in which it represents visual information about its environment. For their achievement, Hubel and Wiesel won the Nobel Prize in 1981, and their discoveries kick-started the rich, diverse field of visual neuroscience.
But scientists who want to study our sense of smell don’t have the same advantages. Smell “is much more, in a sense, mysterious,” says Edmund Chong, a graduate student in neuroscience at New York University. While complex images and shapes can be broken down into their constituent lines and angles, it’s not immediately obvious how to decompose smells, which are conveyed by airborne chemicals. When a person inhales these molecules, they travel through their nostrils and attach to receptor cells, which spark a pattern of activity in the olfactory bulb—a tiny, elongated brain structure right above the nasal cavity. The brain eventually recognizes this pattern as a particular scent. This system allows humans to detect as many as a trillion different odors, though we are much less gifted smellers than mice, whose olfactory bulbs take up a full 2 percent of their brain volume, compared to a hundredth of a percent in humans.
Because these odor-carrying chemicals aren’t easily broken up into their constituent parts, they are “hard to directly manipulate,” says Chong. So when he wanted to figure out how the brain represents smell, he couldn’t follow in Hubel and Wiesel’s steps. Instead of presenting his lab animals with real chemicals, he went straight into their brains. Last week, Chong and his colleagues published a study in the journal Science showing that they’d worked out some of the details of just how the olfactory bulb represents odors—by making mice smell scents that don’t actually exist in the real world.
“It’s a spectacular achievement, both from a technical perspective and conceptually,” says Sandeep Robert Datta, an associate professor of neurobiology at Harvard Medical School, who was not involved in the study. “They’ve taken advantage of advanced methods [to] trick the animal into thinking it’s smelling a particular smell.” By avoiding the issues of manipulating odor molecules entirely and instead going directly to the brain, Chong and his colleagues were able to investigate in detail the aspects of brain activity that matter most for our sense of smell.
Though making mice sense impossible odors might sound like something out of science fiction, Chong’s general approach—stimulating a part of the brain to figure out its logic—has been around since before Hubel and Wiesel did their cat experiments. Wilder Penfield, a neurosurgeon active in the middle of the 20th century, often used an electrical current to activate different areas on the surface of his patients’ brains. He soon discovered that he could cause his patients to feel a physical sensation on, say, their forearms by stimulating the correct brain region—even though they were not truly being touched.
To induce his mice to detect artificial scents, Chong had to be much more precise. Researchers who study the olfactory bulb know that distinct patterns of neural activity in the bulb correspond to different scents. So to make the mice smell odors that weren’t actually present, Chong used a technique called optogenetics, which allows scientists to stimulate groups of neurons using only light. Optogenetics experiments require mice who have been genetically engineered to make some of their neurons—olfactory bulb neurons, in this case—responsive to blue light. When researchers shine light on those neurons, the illuminated neurons become active. By shining different spots of light to stimulate clusters of olfactory neurons, Chong could generate an artificial smell—and train the mice to recognize that smell over time.
But what, exactly, were the mice recognizing? While the patterns of brain activity that Chong elicited broadly resembled the ones that odor molecules naturally produce, he didn’t try to recreate the pattern for a specific real-world scent. For this reason, it’s impossible to know for sure what odor the mice perceived. “Our mice cannot speak, and they cannot report their inner worlds in the way that a human subject can,” Chong says.
But his team could ascertain whether the mice had perceived the scent at all: Chong taught his mice to report what they were sensing through licking one of a pair of water spouts. Specifically, he trained the mice to lick one spout when they detected a single artificial scent produced through a specific pattern of brain stimulation—the “target scent”—and another when they detected any other odor. If a mouse licked the right spout, they would get a droplet of water as a reward. Once the mice could perform this task correctly, the researchers inferred that they had learned to recognize the target scent, even though they were perceiving this odor solely through optogenetic stimulation and never inhaling it through their noses.
An enormous advantage of this strategy is that it allowed the researchers to interrogate which aspects of the pattern of brain activity mattered most to the mouse’s ability to recognize the target scent. “The real power of using artificial stimuli is that you generate these patterns from scratch, and you have full control over them,” Chong says. To evaluate which parts of the target-scent pattern were important to the mice and which were not, he and his team slightly adjusted the pattern in a variety of ways. They then stimulated the mice with these new patterns and observed which spouts the mice licked in response. If they chose the target spout, that meant that they recognized this new pattern as equivalent to the target scent. If they chose the other spout, they were no longer perceiving the same smell—and that meant the change mattered for the brain’s olfactory code.
Using this strategy, Chong and his colleagues were able to answer questions that had long stymied researchers in their field. For example, many scientists who study smell already believed that the timing of activity in the olfactory bulb was important to how the brain processes smell. But they disagreed about the details. By adjusting the timing of his patterns of brain stimulation—making one spot of light come a little bit earlier, or another spot a little later—Chong and his colleagues were able to demonstrate that specific timing is much more important early in the pattern of brain activity. When spots early in the target sequence came just a bit earlier or a bit later than the mouse had become used to, the mice often licked the non-target water spout. On the other hand, they usually got it right when the researchers delayed or accelerated later parts of the sequence.
Without this artificial smell technique, cracking this piece of the olfactory code would have been extraordinarily difficult. This study “shows things that everybody was thinking but nobody had ever been able to experimentally prove,” says Christiane Linster, a professor of neurobiology and behavior at Cornell University, who also was not involved in the research. “It really advances the field by consolidating that the way we think is right.”
Still, it’s hard to be certain that what we learn from fake smells applies to how the brain processes real odors. That’s because the spout-licking strategy has some shortcomings. Although it’s one of the only ways for mice to communicate with the scientists who study them, by rewarding the mice over and over for responding to a particular pattern of activity in their olfactory bulbs, Chong and his colleagues may have trained the mice to respond to those patterns in atypical ways. “Most animals interact with odors in a manner that’s unrewarded,” Datta notes. “Mostly, when you and I walk around the world, we’re smelling smells and we’re not getting hit over the head or getting little sugar pellets in response to that.”
Linster agrees that this discrepancy seriously limits what Chong’s study can tell us about how smell works in the real world. “What [the study] shows us is that animals will learn what you teach them,” Linster says. “It teaches us less about odor coding than the fact that you can use any stimulus to teach an animal, and then they will learn it.”
Chong thinks there is reason to believe that the mice are treating these artificial smells like real ones. “We find that animals learn our task with comparable speed, and perform with similar accuracy, compared to if they were learning to discriminate real odors,” he wrote in an email. But he agrees that his team’s strategy cannot be the be-all and end-all of olfaction research. “Ultimately, we view our artificial approach as being complementary … to conventional approaches using more ‘real-world’ odors,” he wrote.
Chong’s colleagues are already hard at work bringing this artificial approach closer to the real world of smells that mice and humans experience. One strategy they are pursuing is to try to replicate patterns of activity in the olfactory bulb that represent real odors. To do so, they must expose mice to a real chemical scent—like isoamyl acetate, which smells strongly of banana— observe the sequence of neural activity that the odor elicits, and replicate that sequence with light stimulation. If they succeed, they will be able to make the mice smell banana without any chemicals. But reproducing a real pattern of brain activity using light is an exceptionally difficult task. “I foresee it [happening] maybe in 5 to 10 years,” Chong says. “Maybe faster.”
Even though there’s still a great deal of research to be done, Chong’s success in generating smells using brain stimulation opens some tantalizing possibilities for medical applications. Vision scientists have been at work for years building artificial retinas that would allow blind people to see by sending signals directly to their brains, bypassing any nonfunctional parts of the eye or optic nerve. Olfaction researchers might now have a chance to work toward curing anosmia, or loss of smell, which can be caused by chemotherapy, Parkinson’s, and even Covid-19, among a host of other diseases. In theory, one day people might regain their sense of smell through an implanted device that interfaces directly with their brains.
“Imagine you build an artificial nose,” Datta suggests. “What pattern of activity would you have that artificial nose emit in order to most convincingly replicate an odor?” To Datta, Chong’s work takes a first step toward answering that question.
Although bionic noses may still lie far in the future, Chong’s work still represents major progress for the field, Datta says. “We really don’t understand basic principles that underlie how the brain codes information about smells,” he says. “This work really takes advantage of the bleeding edge of technology to get close to some sort of answer about how the brain might code information about smells we encounter in the world.”