Postdoctoral research: Neural networks that encode the value of scent

After obtaining my PhD, I sought to apply my knowledge of molecular biology to address problems in systems neuroscience. I joined Richard Axel’s lab where I worked closely with Peter Wang, a graduate student. We used optogenetics and two-photon imaging to investigate representations of odors in four brain regions in a learning task. Our work provides an answer to the question of where value is imposed on the sensory experience of smell.

In mice, olfactory sensory neurons expressing one of a thousand different receptor genes converge onto discrete glomeruli in the olfactory bulb. However, axonal projections from individual glomeruli discard the spatial organization of the bulb and innervate the next olfactory station, piriform cortex, diffusely and without apparent topography. Individual piriform neurons receive input from a random collection of glomeruli and, as a consequence, the piriform representation cannot inherently encode value. Thus, the olfactory cortex presents a puzzle: How does the brain create order from disorder? How is meaning imposed on the unstructured sensory representation in piriform cortex?  We reasoned that meaning or value must be imposed on odor through learning and experience.

We examined the representation of odors in piriform as well as in two downstream stations, orbitofrontal (OFC) and medial prefrontal cortex (mPFC), in a task in which select odors (CS+) predicted water reward and others did not (CS-). We demonstrated that associative learning reveals no significant change in the piriform odor representation suggesting that value is not encoded in piriform. Rather, we observed dramatic changes in the OFC and mPFC. One third of the OFC neurons acquired robust responses to CS+ after learning and interestingly these responses are modulated by context and state, strongly supporting a representation of value. We observed that once the task has been learned, the OFC responses dissipate, even though the learned behavior persists. This highlights yet another interesting conclusion, representations in one brain area may be unstable, and transfer and consolidation in a second area may be necessary to maintain a stable representation. This is reminiscent of how episodic memories are initially represented in the hippocampus but ultimately consolidated as a stable representation in the neocortex. Finally, the mPFC representation revealed yet another novel finding, the emergence of a stable representation not only of the CS+ but CS- odors as well, a finding with important mechanistic implications for such behavioral phenomenon as extinction.

These findings could one day lead to better treatments for various disorders of cognitive impairment, such as intellectual disability, as well as neuropsychiatric diseases, including addiction and PTSD, in which stimuli acquire exaggerated and inflexible value which drives maladaptive behaviors. My postdoctoral work has shown that the value of an odor is encoded in multiple higher-order brain areas. In my own lab, I plan to study how the value of a sensory stimulus is adjusted to reflect changes in the external and internal world and ultimately generate appropriate behavior.

Doctoral research: DNA repair

During my graduate studies in the laboratory of Fred Alt at Harvard University, I identified a novel pathway that joins DNA breaks, which we named Alternative End-Joining (A-EJ). A-EJ is completely distinct from the two canonical pathways (Homologous Recombination and Non-Homologous End-Joining). Moreover, A-EJ relies on a specific joining mechanism that favors short sequences of homology at the DNA ends. My colleagues and I detected these micro-homologies in the DNA repair products generated during immunoglobulin class switch recombination, suggesting that A-EJ plays a role in a physiological process that creates antibody diversity. In separate studies, I generated genetically modified animals deficient in one or more components of the classical DNA repair pathways to examine the contribution of A-EJ to chromosomal translocations. Finally, I genetically eliminated two of the three known mammalian DNA ligases or their cofactors in cell lines, to screen for A-EJ pathway members. In the DNA repair field, it was widely assumed that A-EJ requires DNA Ligase 3 and its Xrcc1 co-factor. Using the cell line system, I defined an A-EJ pathway that is independent of Xrcc-1.