191 - Revolutionizing our understanding of mental illness with optogenetics

Deisseroth describes ECT as stunning in its effectiveness — being able to take someone in horrific psychological distress and restore them to a functional state is a remarkable clinical achievement. But he frames it as a prime example of psychiatry's core problem: a treatment that definitively works but whose mechanism is completely opaque. The same critique applies to every other psychiatric treatment — SSRIs, lithium, antipsychotics. Understanding why they work requires knowing which cells are actually doing what, which is what optogenetics is now beginning to reveal. ECT is physically refined today — the body is paralyzed, so what looks dramatic is only the brain-level event.

Deisseroth describes TMS as one of several brain stimulation modalities he uses clinically in his office at Stanford (alongside VNS and ECT), highlighting that despite FDA approval, the science of why and for whom it works is still unclear. The knowledge gap is the same as ECT: you can stimulate a region, but you cannot say which cell type is being beneficially altered. Optogenetics is beginning to provide the cellular maps that could one day guide TMS targeting to precisely the axon terminals or cell bodies of identified beneficial populations.

Deisseroth mentions having a VNS therapy radio frequency controller in his office, indicating active clinical use. He frames VNS as an example of psychiatry's pattern: approval based on statistically detectable but clinically modest effects, without mechanistic understanding sufficient to predict responders. The vagus nerve projects to the nucleus tractus solitarius in the brainstem, which then connects broadly — making this another non-specific stimulation, albeit via a different anatomical route than TMS or ECT.

Deisseroth notes we don't yet know which specific cell types in which regions are actually the anxiety-relevant GABA-A targets. Once optogenetics maps the cellular specificity of anxiety circuits (as partially done in 2013), it may become possible to design future anxiolytics that target only the anxiety-relevant subpopulation — potentially preserving cognitive sharpness while still relieving symptoms. For now, benzodiazepines are the most reliable tool but carry their full side-effect burden.

The technical details Deisseroth walks through on the podcast include: choosing between TH (dopamine), TPH (serotonin), or GAD (GABA) promoters for cell-type targeting; getting to hundreds-of-thousands of opsin copies per cell via high-titer viral packaging; exploiting the brain's immune privilege (T and B cells cannot enter the brain freely) to avoid immune rejection of the foreign opsin protein; and delivering light via implanted fiber optics while the animal moves freely. The system's key advantage is also its limitation: it works in the brain because of immune privilege, but peripheral optogenetics sees loss of opsin-expressing cells over time due to immune rejection.

Deisseroth uses these paradigms to explain how optogenetics proved that dopamine neuron activation is positively valenced: mice work thousands of lever presses per day to trigger optogenetic activation of VTA dopamine neurons — a demonstration of motivational salience that electrode stimulation studies only incompletely achieved. The sugar-water test is used to validate depression models (unpredictable stress + sleep disruption produces anhedonia) and to test whether a cellular intervention (e.g., reducing prefrontal cortex overactivity) restores normal hedonic tone.

Deisseroth explains that each neuron type is defined by its function, which requires specific proteins (enzymes, receptors, transporters), which are encoded by genes with cell-type-specific promoters. A dopamine neuron must make dopamine, so it must express TH (tyrosine hydroxylase), so it must have active TH-promoter activity. Packaging the channelrhodopsin gene behind the TH promoter produces an opsin only in dopamine-making cells. The same logic applies to serotonin neurons (TPH promoter), GABA interneurons (GAD promoter), and so on. This solved the 2004-2009 generalizable targeting problem and was why the field took off globally after 2009.

The key technical insight was combining two previously existing but unconnected technologies: safe modified viruses that could shuttle DNA into neurons without propagating further, and the promoter/enhancer strategy — borrowing the regulatory DNA flanking a dopamine-specific enzyme gene (tyrosine hydroxylase) and placing it in front of the channelrhodopsin gene. The virus then infects all nearby cells, but the opsin is only expressed in cells that already activate the TH promoter — i.e., dopamine neurons. Getting from the first paper (2005) to truly generalizable optogenetics took until 2009, when cell-type targeting was versatile enough for any lab to use. Within six months of the 2005 paper, several other labs submitted papers — including a lab run by Roger Tsien (Nobel laureate) — demonstrating that the field had been waiting for this moment.

The most philosophically striking finding from the 2013 experiment: optogenetic activation could make mice strongly avoid exposed areas (the behavioral avoidance feature) without inducing any detectable negative emotional state. The mice were acting anxious without feeling anxious. This clean dissociation had never been demonstrated before and shows that what clinicians call 'anxiety' is a composite state assembled from features that evolution may have linked together but that can be unlinked at the cellular level. The same logic applies to depression, parenting behavior, and social interaction — all have been deconstructed into independently controllable sub-features using optogenetics.

The mouse model: stressed animals given unpredictable events and sleep disruption stop preferring sugar water over regular water — a simple readout of anhedonia. Optogenetic overactivation of prefrontal-to-dopamine-region projections reproduces this loss of reward sensitivity even without external stress. The finding cuts both ways: the frontal cortex's role is not simply positive (cognitive control of danger) but also potentially damaging (excessive suppression of the reward system). Clinically, this may explain why cognitive styles associated with rumination and perfectionism — both involving high prefrontal engagement — are risk factors for depression.

Roska's achievement required a decade of safety work after the initial cadaveric proof-of-concept: primate studies, safety IND applications, then carefully staged clinical trials. The patient's improvement was modest — reaching for objects, not reading — but real and verifiable. This path took longer for retinal degeneration than for other optogenetic applications partly because of the challenging anatomy of delivering virus to surviving photoreceptor-layer cells. Deisseroth frames this as an existence proof: optogenetics CAN be a treatment in carefully selected sensory disorders where survival of some target neurons can be confirmed.

Deisseroth's nuanced position: mania is a coherent brain state (not random dysfunction) that produces consistency across patients — elevated mood, decreased sleep need, accelerated goal-directed activity, increased risk tolerance. It becomes problematic primarily because of impaired decision-making and the transition phase when depressive negativity combines with manic energy (highest suicide risk). The bipolar-depression pairing may reflect a resource exhaustion model or a lost homeostatic break, but the mechanism is genuinely unknown.

Deisseroth distinguishes two separable neural circuits for social behavior: dopamine-driven social motivation (drive to interact) and frontal-cortex-mediated social cognition (ability to process the fast information stream). Optogenetics has identified cell types in both circuits that can improve social behavior in autism-mutation mice. The therapeutic implication: a drug or brain stimulation that targeted social-motivation neurons might help one subtype of autism patient, while one targeting the social-cognition bottleneck would help a different subtype.

Deisseroth notes that non-human primates can enter maladaptive grief states after losing a mother — failing to feed and protect themselves — which may be a depressive analog, but this falls short of volitional self-termination. Even head-banging and self-injurious behavior documented in captive primates lacks the cognitive architecture of planned self-ending. He suggests this gap requires humans and a unique combination: our brains plus hands — the two distinguishing features that together may enable suicide as a deliberate act.

Researchers who digitally add or subtract tears from facial images find enormous effects on observers' desire to help — far greater than smiling or grimacing. The involuntary nature is key: it's not easy to cry on command, so tears credibly signal genuine need. Deisseroth's evolutionary argument: once you have axons from emotional regions projecting to brainstem circuits that control breathing rate (as in anxiety), those axons are physically proximate to tear-duct motor neurons. A minimal evolutionary rewiring connects the two.