Thread by @Gozzi_Ale

Wondering whether axonal input *drives* fMRI-based functional connectivity? Us too!

Here is what we found using #chemogenetic deconstruction of rsfMRI in the mouse🐁🐭

➡️Cortical silencing results in paradoxical fMRI overconnectivity tinyurl.com/yy269j92

Thread below 1/n

Computational and empirical evidence suggest that structural and functional connectivity are robustly related

➡️ this recent review from @richardfbetzel @misicbata summarizes it all tinyurl.com/yyy5u3zm

2/n

⚠️A key prediction of structurally based models of fMRI coupling is that *inactivation* of a brain node would result in reduced rsfMRI connectivity with its targets⚠️

But is that really the case?

3/n

➡️Seminal computational modelling supports this hypothesis tinyurl.com/y4rqucmn

➡️Chemo-fMRI work in primates does too, as inactivation of the amygdala leads to reduce rsfMRI connectivity as predicted
tinyurl.com/y5zv57vf

So this settles the question, right?

4/n

Well, as often the case in science, things might be more complicated than they seem

For example intact bilateral networks have been observed in acallosal people tinyurl.com/yxtwvd7w

5/n

Plus *over-connectivity* is often observed in disorders characterized by loss of cortical function Alzheimer's Stroke.. tinyurl.com/y2pnfxuh

So there seems to be also evidence arguing AGAINST a purely dyadic relationship between structural and functional connectivity..

6/n

...which led us to investigate how inactivation of a cortical region *causally* affects brain-wide fMRI connectivity

➡️We first chronically silenced the 🐭prefrontal cortex via overexpression of a K+ channel. ePhys below shows that this leads to reduced spontaneous firing

7/n

Surprisingly, we found that chronic PFC silencing results in *increased* rsfMRI connectivity between the silenced area and its targets within the mouse default mode network (midline thalamus, posterior cingulate - retrosplenial in rodents)

8/n

By comparing overconnected regions against a voxel-resolution model of the structural connectome tinyurl.com/y4n5mtl8 we found that all the affected regions are robustly innervated by the PFC

➡️ these are areas that are directly "communicating" with the PFC

9/n

Could this effect reflect homeostatic/neuroadaptive adaptations? What happens if we acutely silence the PFC instead?

To address these questions we designed a chemo-fMRI study in which DREADDs were used to acutely silence the mouse PFC

10/n

We found that acute DREADD inhibition of PFC reproduces the same pattern of rsfMRI *overconnectivity*

This means
➡️effect is not manipulation specific
➡️nor a neuroadaptive response to protracted inhibition

Plus, with DREADDs we can now investigate neural mechanism

11/n

Using in vivo ePHYS we found that DREADDs robustly and specifically attenuates firing in PFC

This lead to
➡️*reduced* ƴ activity (expected)
➡️*increased* δ power (intriguing!)

Could δ band coupling explain the observed overconnectivity?

12/n

Multielectrode recordings suggest so!

➡️ All hyperconnected areas show robustly increased δ power & phase coherence
➡️ δ power is linearly related to mean rsfMRI connectivity in all areas

So δ band coherence is a plausible driver of the observed fMRI over-synchronization

13/n

Interestingly, our recordings also showed increased δ coupling between polymodal thalamus and posterior cingulate (retrosplenial in mouse)

➡️If delta is a prominent correlate of rsfMRI over-synchronization, these two regions should then exhibit rsfMRI overconnectivity too

14/n

This may be important mechanistically as polymodal thalamus generates and propagates δ activity to cortex
(see below)

If this were the case
➡️ polymodal/unimodal rsfMRI profiles should be segregable
➡️ polymodal thalamus would be "overconnected" with large cortical areas

15/n

Using a clustering analysis we show this might be the case

➡️we found distinct poly/unimodal rsfMRI profiles
➡️polymodal (but not unimodal) thalamus exhibits rsfMRI overconnectivity with cortical areas
➡️ these include retrosplenial Ctx as predicted by ePHYS recordings

16/n

This leads to a putative model whereby silencing of a cortical region leads to
➡️ reduced "direct" (e.g. ƴ) communication with its targets
➡️ & increased rsfMRI coupling mediated by co-varying slow oscillations (from polymodal thalamus and/or other remote rhythm generators)

17/n

These results may have important implications e.g.

for modelling ➡️rsfMRI connectivity is not always a monotonic indicator of interareal communication

for brain disorders ➡️overconnectivity in degenerative disorders may be indirect result of lower cortical output

18/n

This is it. Many thanks to twitterless
& freshly graduated 🍾 #Carola_Canella
#Federico_Rocchi, #Stefano_Panzeri, @danielgb_87 @GIurilli and all the collaborators at @IITalk
And to @BBRFoundation and @ERC_Research for generous funding. Comments/suggestions are welcome

19/n