Thread by @JonCheung6

Tweetorial time!

1. Question: How do mice locate objects with their whiskers? There are many models, but nobody knows for sure.

We investigated six models of whisker-guided object localization to determine the simplest model that fit behavior best.

cell.com/current-biolog…

2. We tried to keep things simple. We trained head-fixed mice to locate an object on one axis along the face using a single whisker. This approach limits the free parameters we need to tease apart to find the answer.

3. We found mice are incredibly good at this. They can discriminate the location of objects to better than half a millimeter spacing (2 degrees of whisker angle), using only one whisker!

But how?🤔

4. We carefully examined how the mice explored their surroundings with their whisker.

Their exploration pattern was intentional, adaptive to touch, directed to the rewarded location range, and noisy.

5. They tried to only search in the locations where object presence indicated a 💦reward would be available (the Go range). But they just couldn't control their whiskers well enough to consistently avoid hitting objects the unrewarded No-go range.

6. If they DID touch the object, they then made many more whisks directed to the go range. This caused there to usually be more touches when the object was in the Go range than the No-go range.

7. The number of touches made on a trial predicted the choice of mice. But the prediction wasn't perfect. In fact, if mice made just a few touches, they could still tell the difference between Go and No-go locations. They must be using some other feature to discriminate!

8. We thought they could be using the angle of the whisker at touch, or how far the object was from the face. To test which, we modified the task to separate these two features.

Mice used whisker angle at touch, not distance to object to decide where the object was.

9. Prior work suggested mice don't know the exact angle of their whisker at touch. But, the phase of whisker motion is represented in primary afferents (Fee, Severson), while the amplitude and midpoint of whisker motion is in M1 (Hill). Could one of these be a location clue?

10. We built statistical models to predict choice from these whisker motion components, plus touch count, and found that midpoint + touch count predicted the choice of mice best.

11. We suspect that since midpoint of the exploration whisking changes more slowly than angle, phase or amplitude, using this feature as a reference for where the object is, helps smooth out the noisy variation in whisker angle from whisk to whisk.

12. Where does object location emerge in 🧠?
Spikes in L4 of S1 are tightly correlated to touch. Both M1 and L4 S1 neurons project to L5B excitatory neurons of S1. We speculate midpoint info from M1 is combined with touch count info from L4 S1 to produce a location code in L5B.

13. That circuit model would be consistent with 2p imaging & ephys results from Ninglong Xu, Jeff Magee, Leopoldo Petreanu, Karel Svoboda & others.

For a deeper dive, please see our full paper in @CurrentBiology published online today.

cell.com/current-biolog…