Probing the mechanisms of fMRI dysconnectivity with chemogenetic-fMRI

Resting-state fMRI (rsfMRI) is widely used to map brain network organization in health and disease. However, the neural underpinnings and significance of interregional coupling as assessed with rsfMRI remain unclear. Neocortical Excitatory/Inhibitory (E/I) balance critically affects local and long-range information processing and as such can conceivably bias large-scale interareal functional connectivity and fMRI coupling. This notion would be consistent with the concomitant presence of neocortical E/I imbalance and atypical rsfMRI connectivity in multiple brain disorders, such as autism, or schizophrenia. Here, we combine chemogenetic manipulations, rsfMRI, electrophysiology in the mouse to causally probe how alteration in regional neocortical E/I balance affects interareal neural and rsfMRI coupling. We used DREADD-based chemogenetics to remotely alter the E/I balance of the mouse medial prefrontal cortex (PFC) by increasing pyramidal neuron excitability (↑Excitation), or by reducing the activity of fast spiking parvalbumin positive (PV+) inhibitory interneurons (↓Inhibition). For each of the employed manipulations, we recorded both locally elicited fMRI network activity, as well as large scale functional connectivity as measured with rsfMRI. To uncover the neural rhythms associated with our manipulations, we carried out corresponding in-vivo multielectrode electrophysiology in a sperate cohort of animals. Chemogenetic activation of pyramidal neurons (↑Excitation) or inhibition of PV+ interneurons (↓Inhibition) resulted in increased neural firing and evoked-BOLD activity in the targeted area. Both manipulations also produced socio-behavioral impairments in a three-chamber social-interaction test. However, the two manipulations were associated with different spectral signatures as probed with local-field potential measurements (LFP). Specifically, ↑Excitation produced largely decreased broad-band oscillatory power, while ↓Inhibition led to a robust increase in local oscillatory activity. Notwithstanding these spectral differences, both manipulations produced analogous patterns of fMRI hypoconnectivity in the mouse default mode network. To understand how these different local spectral activity may produce converging patterns of rsfMRI dysconnectivity, we turned to multielectrode electrophysiology and measured interareal coupling between the manipulated region (PFC) and its anatomical targets (retrosplenial cortex). These investigations revealed that fMRI hypoconnectivity produced by ↑Excitation or ↓Inhibition was associated with dissociable LFP coherency signatures. Specifically, chemogenetic activation of pyramidal neurons (↑Excitation) produced decreased coherence in slow-δ range (0.1-4 Hz), while inhibition of PV+ activity (↓Inhibition) produced a composite response, comprising reductions in slow frequency coherence along with a largely increased coherence in theta to beta bands. By relating fMRI changes to corresponding LFP coherency differences, we found that neural coupling in slow (0.1-4 Hz), but not faster frequency bands, significantly predicted the fMRI connectivity changes produced by all the employed manipulations. These results suggest that large-scale fMRI connectivity is primarily supported by electrophysiological coherence in infraslow/slow rhythms, and it is disproportionately less sensitive to high frequency coupling. Importantly, this relationship also held when recordings obtained upon chemogenetic silencing of the PFC (i.e. producing a reduced ↓Excitation) were included. Collectively, our results shed light on the general principles underlying macro-scale fMRI network organization in the mammalian brain with a number of important implications for the interpretation of rsfMRI connectivity in health and disease. First, they suggest that the relationship between interareal neural activity and fMRI connectivity is critically biased by local Excitatory/Inhibitory ratio. Second, our findings also support a simple framework whereby interareal patterns of hyper- and hypo-connectivity observed in brain disorders may counterintuitively reflect reduced or increased neural excitability of afferent systems, respectively. Third, these observations point at a possible unifying mechanistic link between E/I imbalance and connectivity disruption in brain disorders, suggesting that these two phenotypes may be the result of a unique etiopathological insult. Future extensions of this framework may offer opportunities to model the local contribution of regional E/I balance in affecting fMRI connectivity, and to physiologically decode fMRI dysconnectivity in human disorders