El Duvelle 🌍

@ElDuvelle

Maybe you think that hippocampal #PlaceCells encode the value 📈 of spatial goals?

You might also think that their place fields always accumulate at goals 🎯?

*** think again - they do not! ***

Thread on our 2019 #JNeurosci paper:
https://www.jneurosci.org/content/39/13/2522

1/🔽

23/04/2021, 17:07:08

Favs: 142

Retweets: 28

El Duvelle 🌍

@ElDuvelle

Why a thread now, two years later?
because:
- I wasn’t on Twitter yet😞
- People seem to regularly misinterpret it – but really, it’s quite simple 😊
- It's a really cool paradigm (yes, I'm biased)😎
- I can add things not included in the paper😉
- we can discuss!

2/

23/04/2021, 17:07:08

Favs: 7

Retweets: 0

El Duvelle 🌍

@ElDuvelle

Summary of main findings if you are in a hurry:
1. rats can remember the location of 2 uncued spatial goals
2. rats can associate different values to these goals
3. place fields do *not* over-represent goals
4. place cells do *not* encode goal value in their firing rate.

3/

23/04/2021, 17:07:08

Favs: 9

Retweets: 1

El Duvelle 🌍

@ElDuvelle

Hippocampal place cells are a strikingly beautiful example of how a tiny biological entity (a cell!) can represent something that is not directly perceivable: the location of the animal.

[region of high activity = place field]

(video from @rmgrieves: https://figshare.com/articles/media/Place_cell_50_minutes_t12c9/8063324)
4/

23/04/2021, 17:07:13

Favs: 9

Retweets: 0

El Duvelle 🌍

@ElDuvelle

An intact #hippocampus is necessary when rats have to locate a hidden goal and flexibly navigate to it (e.g. watermaze).

So, we asked if hippocampal place cells represent spatial goals as well as their value (e.g. where is your fridge and is it full, half-full, or empty)?

5/

23/04/2021, 17:07:14

Favs: 2

Retweets: 0

El Duvelle 🌍

@ElDuvelle

To test this we adapted a task used in the LNC (Marseille): the 'continuous navigation task'.

We trained rats to choose between 2 hidden spatial goals and alternate waiting there (for 2s) with searching for a reward released elsewhere in the environment.
(see Tweet1)

6/

23/04/2021, 17:07:17

Favs: 3

Retweets: 0

El Duvelle 🌍

@ElDuvelle

Why not delivering food at the goal like in most experiments? Because we are interested in the representation of the spatial goal itself, not of reward!

We alternated sessions with equal goal values and sessions where one of the goals consistently delivered 0 or 3 rewards.

7/

23/04/2021, 17:07:18

Favs: 5

Retweets: 0

El Duvelle 🌍

@ElDuvelle

We were happy to find that the rats differentiated the two goals and adjusted their visits with the changing values, even though goal & reward were dissociated!

Here, >0 numbers indicate preference for the goal with highest value 🔽
(left: example, right: average/sessions)

8/

23/04/2021, 17:07:19

Favs: 4

Retweets: 0

El Duvelle 🌍

@ElDuvelle

(small digression: learning the 2-goals takes a looong time - ~6 weeks 😅 - and was done in steps; we tested the 1-0 condition during learning a bit, but not so much the 1-3, which is why it was a good surprise to see it working!)

9/

23/04/2021, 17:07:20

Favs: 4

Retweets: 0

El Duvelle 🌍

@ElDuvelle

Here are the (speed-filtered) rate maps of some example place cells recorded in the task🔽
Note that we were able to follow these cells for several days (2 days shown here), by matching their waveforms. See any change specific to goal value?

(note, plots are rotated a bit)
10/

23/04/2021, 17:07:21

Favs: 5

Retweets: 0

El Duvelle 🌍

@ElDuvelle

To look for goal-value coding, we analysed place field locations, cumulated firing rates, average firing rates, overdispersion (variability of firing): no significant difference was found in any of these.
Note also: no overrepresentation of the goals, even of high-value!

11/

23/04/2021, 17:07:23

Favs: 6

Retweets: 0

El Duvelle 🌍

@ElDuvelle

In previous studies (e.g. Hok et al., 2007) most place cells were found to express some firing at the goal, *outside* of their place field. We also found this (CA1 &CA3, left).

Could it reflect value coding? Nope! The goal firing did not change with goal value (right).

12/

23/04/2021, 17:07:24

Favs: 2

Retweets: 0

El Duvelle 🌍

@ElDuvelle

(note, in the previous plots, the slightly different temporal profile at the end of the delay period for the 0 value goal, with less of a decrease - this might be related to having no food dispenser sound for that goal, so rat staying for longer?🤔)

13/

23/04/2021, 17:07:24

Favs: 2

Retweets: 0

El Duvelle 🌍

@ElDuvelle

Focusing on individual cells, we tested if their firing at a given goal was significantly modulated by goal value across sessions (compared to a shuffle). This was sometimes the case for a single session, but *0* (zero!) cells consistently showed this for multiple sessions.

14/

23/04/2021, 17:07:25

Favs: 2

Retweets: 0