Thread by @harris_kaplan

How does the brain coordinate animal behavior across timescales?

@salazarori and I address this question in our latest from @Zimmer_Lab: “Nested neuronal dynamics orchestrate a behavioral hierarchy across timescales”

Link @NeuroCellPress:
cell.com/neuron/fulltex…

Thread 1/

@salazarori @Zimmer_Lab @NeuroCellPress In a behavioral hierarchy, as proposed by Tinbergen, long-timescale behaviors are composed of shorter-timescale ones. For example, the stickleback’s reproductive instinct (Hierarchy Level 1) includes mating and fighting (Level 2); each in turn consists of several sub-actions. 2/

@salazarori @Zimmer_Lab @NeuroCellPress Many behaviors can be described this way (e.g. recent work by @gordonberman @aexbrown @daniobrain), but are the underlying neuronal mechanisms actually hierarchical? 3/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain First, we described a hierarchy of #celegans behaviors across three timescales. For example, head-casts (Hierarchy Level 3) occur only during specific windows of the propagated-bend (Level 2) phase. We call this “phase-nesting,” with fast behaviors nested within slower ones. 4/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain We then set out to uncover the neuronal control mechanisms. First, we delineated circuits driving each behavior. In our previous work with @neurotheory, we described brain-wide dynamics underlying Hierarchy Level 1, forward/reverse switching. 5/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Here, we used whole-nervous-system calcium imaging and targeted neuronal inhibition to discover local motor neuron circuits driving fast behaviors:

• B-MNs drive prop-bends (Hierarchy Level 2)

• SMDD and SMDV drive dorsal and ventral head-casts (Level 3)

6/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Having identified global (Hierarchy Level 1) and local (Levels 2 and 3) neuronal populations driving each behavior, we next asked how these populations interact to implement the behavioral hierarchy.

To do so, we imaged B-MN DB02, SMDD, and SMDV in moving animals. 7/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Switches between forward and reverse state (Hierarchy Level 1) toggled DB neuron oscillations on and off, and SMDs between different activity states and functional roles.

SMDs acted either as motor neurons or interneurons, depending on the higher-level state! 8/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Finally, we examined Hierarchy Levels 2 and 3. Here, SMDs showed two different activity signatures:

• During prop-bends (Level 2), SMDD and SMDV showed alternating activity peaks.

• During lateralized head-casts (Level 3), only one of the two SMDs oscillated.

9/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Remember the “phase-nesting” concept? Well, we wondered if we could find phase-nesting purely using these SMD activity signatures, given that they occur during phase-nested behaviors. Turns out we could! That was reassuring. However, it begs the question… 10/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory What’s the direction of causality? Is phase-nested SMD activity a cause or a consequence of phase-nested behavior?

Well, we found that SMDs were phase-nested even in immobilized animals! This suggests that SMD phase-nesting drives behavioral phase-nesting, not vice versa.

11/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory This result is especially striking because SMD dynamics are 10x slower in immobilized animals versus moving ones, suggesting that phase-nesting is an extremely robust circuit feature. 12/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Altogether, we mapped the strict hierarchical relationships in behavior to strictly nested neuronal activities.

We conclude that nested neuronal activity patterns are a repeated dynamical motif of the worm’s nervous system, which together implement a behavioral hierarchy. 13/

@salazarori @Zimmer_Lab @NeuroCellPress @gordonberman @aexbrown @daniobrain @neurotheory Thanks to our lab and @IMPVienna @UniVienna colleagues, and many others, including 3 reviewers, for evaluating and improving our work tremendously.

Also, check out this interpretation of our work by Viennese artist Franka Rothaug! See more at frankarothaug.com.

14/14