We believe our results may (partly) reframe how we interpret fMRI connectivity ▶️ fMRI connectivity ≠direct communication strength ▶️ fMRI connectivity is supported by distributed slow neuronal coupling ▶️ Hyper/hypoconnectivity (eg., in brain disorders) may reflect cortical hypo/hyperexcitability 16/n

They may also have implications for brain stimulation. For example: if we increase excitability in a cortical area (with TMS) we may see a decrease in its fMRI connectivity. What we like here is that these are testable hypotheses: and so we will soon see if (any of) this holds in humans! 17/n

Thus our work suggests that 1️⃣ cortical excitability inversely modulates fMRI connectivity 2️⃣ fMRI coupling rests on distributed, slow neuronal fluctuations (i.e. QPPs, CAPs, neuromodulation pulses..) 3️⃣cortical excitability gates local coupling by weakening or facilitating that slow synchrony 15/n

Notably, biophysical modelling supports this framework. Using a simple three-node model we found that local excitability changes are sufficient to reproduce the direction of the low-frequency coherence effects across perturbations. This offers a plausible mechanistic account of our results! 14/n

So these results suggest that ▶️slow, shared LFP fluctuations provide a neuronal scaffold for fMRI connectivity ▶️cortical excitability gates how strongly regions participate on this process: shifts in cortical excitability weaken or facilitate this coupling, leading to hypo/hyperconnectivity 13/n

By comparing fMRI to LFP coherence one thing stood out ▶️ low-frequency coherence (<4 Hz) consistently tracked fMRI effects across all manipulations! Higher frequencies also change (sometimes a lot), but they don’t covary with fMRI. So slow neuronal coupling is the common denominator of fMRI! 12/n

Given that fMRI is a measure of how synchronous fMRI fluctuations are across regions, we thougth measures of interareal electrophysiological (LFP) coherence could shed light into rhythms underlying large-scale fMRI coupling 11/n

This seem to hold across different manipulations, and different cortical areas. Which raises the obvious question: what neural signals/mechanism track the observed fMRI connectivity changes? 10/n

However, fMRI revealed something very consistent ▶️higher excitability →reduced fMRI connectivity ▶️lower excitability →increased fMRI connectivity So in our datasets more local activity ≠ more fMRI connectivity! Rather, fMRI connectivity is inversely related to cortical excitability. 9/n

As expected, these manipulations look very different spectrally. So if fMRI connectivity depended on a complex mix of neuronal rhythms, we should see divergent effects once we map the effect of these manipulations on large-scale fMRI coupling. 8/n

Using these manipulations we sought to push cortical circuits into different firing-rate regimes. ▶️↑Excitation → higher firing ▶️↓Inhibition → higher firing ▶️ Silencing → lower firing So we can now map what happens when we manipulate excitability via different circuit mechanisms 7/n

We considered three manipulations (through novel or existing datasets) 1️⃣ increasing pyramidal-cell excitability (↑Excitation) 2️⃣ reducing PV interneuron activity (↓Inhibition) 3️⃣ pan-neuronal suppression (Silencing) 6/n

To test our hypothesis, we combined (in the mouse PFC) ▶️Chemogenetics to manipulate excitability ▶️Electrophysiology (spikes + LFP) ▶️fMRI connectivity But crucially: we combined the result of multiple manipulations into one unified framework, instead of interpreting each one in isolation 5/n

In our work, cortical excitability = mean population firing. This is because in recurrent cortical networks, excitatory and inhibitory activity change in a coordinated way: if excitatory neurons fire more, inhibitory neurons also fire more (& viceversa - we show this using a simple model) 4/n

The key idea here is 👉Cortical *excitability" might be a key hidden physiological variable controlling fMRI connectivity. This hypothesis stems from the observation that cortical excitability closely regulates rhythmic interareal coupling. So we thought this could apply to fMRI connectivity too 3/n

fMRI connectivity is often interpreted as a proxy for direct interareal communication, but its physiology remains debated. In particular, it is unclear how local circuit activity maps onto fMRI connectivity. 2/n

📢 New preprint from the lab🧠 ▶️doi.org/10.64898/2026.03.12.710517 What does fMRI connectivity actually reflect at the neural level? The natural intuition is: more neural activity = more connectivity! Using cortical perturbations we show this is not necessarily the case: sometimes less is more! 👇🧵

🧠What if increasing neuronal firing could actually reduce fMRI connectivity? 📡What drives long-range fMRI connectivity? I’m very excited to share that my PhD work is now out as a preprint Check this out! 👇 doi.org/10.64898/202...

🐭🧠 Can fUSI truly map canonical mouse resting-state networks — and how does it compare to fMRI? I’m excited to share the work of my PhD in our new preprint!👉 doi.org/10.64898/202... Go check it out — it’s time to expand your neuroimaging toolkit 🧠🚀

Jane Goodall, known for her pioneering work with chimpanzees, has passed away aged 91 go.nature.com/46K10ja