2004 Nobel Prize in Physiology or Medicine: Axel and Buck Discover the Combinatorial Code of Smell
What You Will Learn from This Article
The 2004 Nobel Prize in Physiology or Medicine was awarded to Richard Axel and Linda Buck, from Columbia University, for elucidating the molecular and genetic basis of olfaction, i.e., how we distinguish between the scents of coffee and roses. In 1991, the two researchers discovered a family of approximately 1,000 G protein-coupled receptor (GPCR) genes in mouse olfactory epithelium. They then established the principle that each olfactory neuron expresses only one of these receptors and demonstrated the coding principle in which a single odor molecule partially binds to multiple receptors, creating a combinatorial pattern. This discovery is the foundation for the flow that leads to today's sensory neuroscience, the fragrance industry, understanding olfactory loss in COVID-19, and early diagnosis of Alzheimer's disease.
A Story Different from Common Sense: Odors are Vectors
The intuitive common sense that "the nose detects each odor with a single receptor corresponding to that odor" is incorrect. The number of odors that humans can distinguish is estimated to be at least one trillion. However, the number of human olfactory receptor genes is approximately 350 in their active state. How can 350 "hardware" units distinguish between one trillion odors? The answer to this question is the combinatorial code discovered by Axel and Buck.
If we translate this into a CS framework, olfaction works as follows: A single odor molecule partially binds to multiple of the 350 receptors, with varying strengths. Each receptor reports the binding strength to the brain. The result is a 350-dimensional vector. This vector is the fingerprint of a single odor. Different odors have different vectors, and similar odors have similar vectors. The brain's olfactory bulb pattern matches this vector to recognize a specific odor. The arithmetic that 1,000 genes create one trillion combinations is the answer to the exponential explosion of this vector space. Even considering only the binary state of each receptor being either on or off, 2^350 far exceeds the number of all possible odors that humans can imagine.
This principle remarkably resembles feature hashing or Bloom filters in the CS world. A single input goes through multiple hash functions, leaving signals in multiple positions in the vector, and this vector serves as the fingerprint of the item. The fact that evolution discovered this architecture in sensory hardware, and that 21st-century information retrieval and machine learning rediscovered it in software, is strong evidence that the principles of information processing converge in both silicon and neurons.
The Landscape of the Time: Impeachment, Tsunami, and the Restructuring of Internet Services
In 2004, Korea began with an unprecedented event in its constitutional history. On March 12, the National Assembly passed an impeachment motion against President Roh Moo-hyun, and his presidential duties were suspended for two months until the Constitutional Court's decision. On May 14, the Constitutional Court dismissed the impeachment, and in the April 15 general election, the Uri Party created a governing majority, restoring the Participatory Government. It was a politically turbulent year. In September, the Zaytun unit began its deployment to Iraq, and the rapid rise in Seoul apartment prices became a serious social problem. The aftermath of the credit card crisis was coming to an end, and the Korean economy was experiencing its second major adjustment since the IMF crisis.
In the world, George W. Bush was re-elected on November 2, solidifying the long-term nature of the Iraq War. In April, the Abu Ghraib prison abuse scandal was revealed, leaving a significant scar on American moral leadership. And on December 26, the Indian Ocean tsunami occurred, triggered by an earthquake off Sumatra, killing approximately 220,000 people in 14 countries, including Indonesia, Thailand, Sri Lanka, and India. It is recorded as one of the worst natural disasters in human history.
In the internet world, the year saw a restructuring of services. On February 4, Mark Zuckerberg launched Facebook from his Harvard University dorm room. On April 1, Google launched Gmail, and the 1GB email inbox, which was initially thought to be an April Fool's joke, turned out to be a real service. Just before YouTube was launched the following year, these two services were harbingers of the flow that would completely change user and data flows. The fact that the coding of smells and the restructuring of data flows in this era are not unrelated will be revealed laterβit was around this time that feature hashing became a standard tool for web search and spam filters.
Personal Stories: A Master of Molecular Biology and His Young Postdoctoral Researcher
Richard Axel (1946β) was born in Brooklyn, New York. He grew up in a Jewish immigrant family, received his bachelor's degree from Columbia University, and his M.D. from Johns Hopkins University School of Medicine, and then returned to Columbia, where he spent his entire career. He was one of the early pioneers of recombinant DNA technology in the 1970s, and he developed several methodologies for DNA transformation. These methodologies later became the standard for gene therapy and the creation of transgenic mice. Axel's lab addressed various neuroscience problems, and olfaction was one of them. He was known for his hands-off style, allowing his students and postdoctoral researchers to explore freely.
Linda Buck (1947β) was born in Seattle, Washington. She received her undergraduate degree in psychology and microbiology from the University of Washington and her Ph.D. in immunology from the University of Texas. She joined Axel's lab as a postdoctoral researcher in the early 1980s and began working on olfaction. She was particularly attracted to the fact that the molecular basis of olfaction had remained unsolved for over 30 years.
The problem was this: While psychology and electrophysiology had established the existence of olfactory receptors, no one had identified the genes for those receptors. It was expected that the receptors would be expressed in small amounts, only in the olfactory epithelium, and that there would be multiple genes that were slightly different, but a cloning method had not been established. Buck started with three hypotheses. First, olfactory receptors would be in the G protein-coupled receptor (GPCR) family. Second, multiple similar genes would form a superfamily. Third, these genes would be expressed only in the olfactory epithelium.
Based on these three hypotheses, Buck utilized a technique called degenerate PCR. She designed a set of primers that allowed for multiple variations in the conserved sequence of GPCR and amplified a family of similar genes from mouse olfactory epithelium RNA in one go. The results were surprising: in a single experiment, dozens of new GPCR genes were detected, and when screened on a genomic scale, the number of olfactory receptor genes in mice was found to be approximately 1,000. This was the largest family of genes in the mouse genome. This paper, published in Cell in 1991, opened the molecular era of olfaction.
In the following years, Axel and Buck collaborated and worked independently, each demonstrating the one receptor-one neuron rule and the combinatorial coding principle. Buck moved to the Fred Hutchinson Cancer Research Center in the late 1990s and established her own lab, while Axel continued his research on fruit fly olfaction at Columbia. The Nobel Prize was awarded to these two individuals togetherβa mentor and a mentee, each of whom led the field in their own way.
Key Achievements: Combinatorial Coding in a CS Framework
If you draw the information processing pipeline of the olfactory system step by step, it looks like this:
- Input: Odor molecules (e.g., caffeol in coffee, geraniol in roses, etc.) enter the nose and dissolve in the mucus of the olfactory epithelium.
- Receptor Level: Each olfactory neuron's surface expresses only one type of olfactory receptor (one receptor-one neuron rule). The odor molecule partially binds to this receptor. The binding strength is determined by the molecular structure and the complementarity of the receptor's binding site.
- Signal Transduction: When binding occurs, the receptor, through the standard GPCR pathway, increases cAMP, exciting the neuron.
- Spatial Encoding (Bulb Level): A remarkable principle: Neurons that express the same type of receptor send their axons to the same glomerulus in the olfactory bulb. As a result, the surface of the olfactory bulb is reconstructed as a map of 350 (in humans) glomeruli. Each glomerulus acts as a pixel representing a type of receptor.
- Pattern Matching (Cortical Level): When an odor molecule activates multiple receptors, multiple glomeruli in the olfactory bulb are activated with varying strengths. This activation pattern is the vector fingerprint of the odor. The olfactory cortex and amygdala pattern match this vector to link it to a specific odor, memory, and emotion.
The essence of this pipeline is combinatorial coding + spatial mapping. Each neuron acts like a hardware node, detecting only one type of feature, and these nodes are mapped to specific locations spatially, so that the brain can read the map of "which nodes are activated and how much." In the CS world, this is similar to the first convolutional layer of a CNN. Each filter (neuron) responds to a specific feature (receptor), and the activation pattern of multiple filters is combined in the next layer to create higher-level features.
The limitations of this analogy must also be stated. The pattern matching of olfaction, unlike the pure convolution of CNNs, is directly connected to strong memory and emotion circuits. Olfactory information is the only sensory pathway that does not pass through the thalamus and reaches the amygdala and hippocampus directly. This is why a single smell can evoke entire memories from decades agoβthe "Proust phenomenon." If you try to understand it only as a pure signal processing architecture, you cannot explain this connection with emotions and memories. The intertwining of sensory hardware and emotional systems is the unique feature of olfaction, and it is also the reason why the fragrance industry exists.
Why It Matters: The Sense of Smell Revived by the Pandemic
As a new principle of sensory neuroscience, combinatorial coding has expanded beyond olfaction. Taste also operates in a similar manner, and most pheromone receptors and the chemosensory system independent of pheromones belong to the GPCR superfamily. Axel and Buck's approach β cloning the vast GPCR superfamily at once using degenerate PCR β became a standard technique for discovering genes in various sensory systems.
The perfume, cosmetics, and food industries represent the commercial conclusion of this story. After understanding the principle of combinatorial coding, designing new fragrances involves a molecular-level approach to determine which receptor combinations should be activated to create the desired impression. AI-based fragrance design has also been rapidly growing recently. Neural networks that learn to predict the binding strength between olfactory receptors and odor molecules are used to predict the scent of new molecules in advance without experimentation. In the 21st century, olfactory research is progressing by adding deep learning on top of the experiments conducted by Axel and Buck.
The COVID-19 pandemic has revived the importance of the sense of smell. Even asymptomatic or mildly infected individuals experienced sudden loss of smell (anosmia), which became an early diagnostic indicator. Understanding why the coronavirus attacks the sense of smell was based on the olfactory epithelium structure revealed by Axel and Buck. It was discovered that ACE2 receptors are expressed not in the olfactory neurons themselves, but in the supporting cells (sustentacular cells) that support them, and that these cells are infected, indirectly damaging the olfactory neurons. The regeneration of the sense of smell after recovery from infection is possible because olfactory neurons are one of the few types of neurons that continue to be newly created even in adulthood.
The re-emergence of olfactory tests for early diagnosis of Alzheimer's disease is also an extension of this story. Observations have accumulated that the early pathology of Alzheimer's disease begins in the olfactory nervous system, and research has been conducted on whether subtle changes in the sense of smell can serve as an early warning sign before cognitive symptoms appear. The perspective is that losing the sense of smell may be a window into the health of the aging brain.
To conclude with a story of evolution, humans are a species that has lost more olfactory receptor genes compared to other mammals. Mice have approximately 1200 active olfactory receptor genes, dogs have approximately 900, and humans have approximately 350. The rest are pseudogenes β fossil genes that have lost their function due to mutations and remain in our genome. The evolution of vision has taken away from the olfactory budget. As our primate ancestors developed trichromatic vision, the investment in olfactory genes decreased. Nevertheless, we can still distinguish between a trillion different scents, and this hardware operates on a combination vector of 350 receptors.
The fourth Nobel Prize in Physiology or Medicine of the new century honored the principle of coding for the senses. The subsequent stories β RNA interference, iPS cells, telomeres, and grid cells in the brain β are not far away.
β Previous: 2003 Nobel Prize in Physiology or Medicine β Next: 2005 Nobel Prize in Physiology or Medicine