Hacking need states
We found that we could ‘hack’ the hunger system in mice by cell-type-selective activation of a molecularly defined hypothalamic population called AGRP neurons. Cell type-specific AGRP neuron activation induced a virtual starvation state even in well-fed mice, eliciting voracious eating within minutes. The amount eaten was proportional to the level of activity in AGRP neurons, and the motivation was specific for consumption of food over water. This showed that AGRP neurons could serve as an entry point for studying a goal-directed motivated behavior.
Deconstructing hunger circuits
Because AGRP neurons are a single population that is sufficient to elicit intensive food-seeking behaviors, the axon projections of these neurons are well-suited to identify other hunger circuit nodes. We have discovered several circuits that induce robust food consumption. We are developing new methods to explore the neural circuits underlying appetite as well as other need states.
Unpleasantness of need
Using electrical activity manipulations of AGRP neurons and deep-brain calcium imaging, we found a motivational mechanism for hunger that contributes to the difficulty of maintaining weight loss on a low calorie diet. We showed that AGRP neurons mediate food-seeking behavior by signaling the unpleasantness (i.e., negative valence) of hunger. Our imaging and neuron perturbation experiments indicate a new learning rule for homeostatic need states, where starvation-sensitive AGRP neurons motivate behavior by a negative valence signal. Inhibition of this pathway is mediated by food and is rewarding; consequently, food-seeking behavior is selectively reinforced in hunger over behaviors that don’t provide nutrients. AGRP neurons are the first neuron population found to elicit appetite by mediating the unpleasant aspects of hunger. Interestingly, we discovered that a different set of neurons, which elicit thirst, also transmit a negative valence signal. Thus, this mechanism appears to be generalizable to another need state.
Molecular control points for hunger
To bridge the basic science of neural circuits with therapeutic strategies for obesity and eating disorders, we have used methods for cell-type-specific transcriptomics of AGRP neurons and another important population called POMC neurons. We have identified several receptors, ion channels, and neuropeptides that increase or decrease appetite as well as hundreds of other candidates for appetite-regulatory pathways. Using these molecular control points for AGRP neuron activity may ultimately be important for developing non-surgical weight loss strategies, which would have important societal impact to counteract obesity.
Tools for causal neuroscience
Cre-dependent viral vectors
We developed highly compact Cre-dependent viral vectors using the flip-excision (FLEX) switch that is now used for cell type-specific transgene expression by hundreds of labs.This also allowed the first example of long-range functional synaptic circuit mapping from molecularly defined neurons.
Cell type-specific pharmacology systems
PSAM/PSEM ion channels. Cell type-specific tools are crucial for establishing causal relationships between neuron activity and animal behavior. My lab developed a modular toolbox of ion channels for cell-type-selectively activating or inhibiting neurons by combining synthetic chemistry and protein engineering (with Loren Looger). Selective brain-penetrant chemical ligands (PSEMs) were synthesized and tailored for engineered ligand binding domains (Pharmacologically Selective Actuator Modules, PSAMs). These PSAMs and their selective agonists could be mixed-and-matched with different ion channel domains to create designer ligand-gated ion channels (LGICs) that had readily tuned electrical properties for cell-type-selectively activating or inhibiting neurons with cell type-selectivity in vivo.
Synaptic silencers. We examined the mechanism of the hM4Di/CNO DREADD system and found that it was only a modest inhibitor of neuronal electrical activity, but it was actually a powerful silencer of synaptic release. We modified the system to target it selectively to synaptic release sites (hM4Dnrxn) so that the effect on synaptic silencing can be achieved using targeted intracranial CNO injections. This tool is useful for loss-of-function circuit manipulations on molecularly defined cell types and does not appear to affect action potential transmission to downstream areas. For comparison PSAM/PSEM silencer or optogenetic silencers can be used to inhibit axonal projections, but they do so by blocking action potential propagation which also affects every site downstream of the targeted area.
Cell-type selective ion channel pharmacology. We used an enzymatic unmasking strategy to target the use-dependent NMDA-R ion channel antagonist, MK801 to molecularly defined cell types. Rendering MK801 inert with a stable ester modification (to form a reagent called CM-MK801), the masked molecule could be liberated in neurons transgenically expressing pig liver esterase (PLE). We found that we could use this tool in brain slices to block NMDA-Rs selectively in PLE-expressing neurons but not neurons that lacked PLE.