Self-centred vs. other-centred homeostatic plasticity in inhibitory interneurons

One of the defining hallmarks of life is homeostasis: maintaining a constant state despite shifting conditions, by compensating for perturbations away from a desired set point. Yet this compensation is imperfect, and sometimes might even backfire. For example, after injuring your foot, you might start limping, but the imbalanced gait could then cause back pain. Understanding when and how biological compensation backfires requires a better understanding of compensatory mechanisms.

We apply this general question to the problem of how the brain maintains stable levels of neuronal activity. This stability is important because neurons in the brain are all connected: when a neuron fires an electrical impulse, it sends chemical signals to other neurons that either excite them (make them fire) or inhibit them (stop them from firing). If excitation and inhibition become imbalanced, a neural network can spiral out of control into a seizure (too much excitation) or silence (too much inhibition). Yet our brains are constantly changing as we learn based on sensory experience. To stop these changes from unbalancing excitation and inhibition, the brain readjusts neurons and their connections to compensate for the changes and restore stable activity levels, a process called "homeostatic plasticity". Problems with homeostatic plasticity are thought to contribute to disorders like epilepsy, autism, and schizophrenia.

We address in particular the homeostatic plasticity of inhibitory neurons. How should inhibitory neurons compensate for changes in their activity? Taking the perspective of the whole network, you might expect that if inhibitory neurons are too active, it's because they're overactivated by the excitatory neurons, so the inhibitory neurons should become even more active so as to better silence their excitatory neighbours. We call this "other-centred" plasticity because the inhibitory neurons mainly care about stabilising the activity of excitatory neurons. On the other hand, more "selfish" inhibitory neurons might notice only that they're too active, and therefore reduce their own activity to return to their preferred set point. This "self-centred" compensation would backfire at the network level, as their excitatory neurons would get even less inhibition than usual, so they would become more active, which worsens the original problem. That is, a compensation rule that makes sense at the local level (it restores the normal activity of the inhibitory neuron) is counterproductive and makes things worse at the network level (because the excitatory neurons become even more active).

Surprisingly, both self-centred and other-centred compensation in inhibitory neurons have been observed in different studies. It's not clear in what contexts one or the other occurs, or by what mechanisms. Answering these questions will help us better understand how neural networks maintain normal activity levels.

We'll study this problem using the olfactory system of the fruit fly Drosophila, where some inhibitory neurons compensate in a self-centred way, while others compensate in an other-centred way. This contrast will let us compare how the self-centred and other-centred neurons affect their excitatory partners and what mechanisms control the plasticity.

First, we'll identify which inhibitory neurons use self-centred vs. other-centred compensation, and if they follow different rules in different contexts. Then, we'll test our hypothesis that self-centred inhibitory compensation backfires and increases activity of excitatory neurons. Finally, we'll investigate the different mechanisms underlying self- vs. other-centred compensation. In particular, we hypothesise that self-centred plasticity uses signalling mechanisms entirely internal to the inhibitory neuron, whereas other-centred plasticity relies on interactions between excitatory and inhibitory neurons (e.g., excitatory neurons "ask" inhibitory neurons for more inhibition).

Inhibitory plasticity is important for stabilising neural network activity, but inhibitory neurons can follow two potentially opposing homeostatic rules: "other-centred" compensation (if inhibitory neurons are too active, they increase excitability to better inhibit the excitatory neurons responsible for over-exciting them) vs. "self-centred" compensation (if inhibitory neurons are too active, they reduce excitability to decrease their own activity). Self-centred inhibitory compensation backfires: it counter-intuitively causes anti-homeostatic effects on excitatory neurons. When and how does each type of compensation occur?

We recently discovered that in the Drosophila mushroom body, the inhibitory APL shows self-centred compensation when the excitatory Kenyon cells are overactivated. In contrast, at least some inhibitory local neurons (iLNs) in the antennal lobe show other-centred compensation. We'll use two-photon imaging and genetic manipulation of neural activity to test (1) whether self- vs. other-centred compensation occurs in different iLN subpopulations or in different stimulation protocols, (2) whether self- and other-centred inhibitory compensation in fact *cause* anti-homeostatic and homeostatic plasticity of excitatory neurons, respectively, and (3) whether self- and other-centred compensation occur via different mechanisms (we hypothesise that self-centred occurs by CaMKII signalling in inhibitory neurons, while other-centred occurs by a sensor detecting coincident activity of inhibitory and excitatory neurons, e.g., mAChRs, Ca2+-dependent adenylyl cyclase, or retrograde signalling from excitatory neurons).