State-dependent neural manifolds, dimensionality, and top-down communication in the visual cortex
1 Institute for Advanced Simulation (IAS-6), Jülich Research Centre, Jülich, Germany
2 Institut de Neurosciences de la Timone (INT), CNRS & Aix-Marseille Université, Marseille, France
Brain activity depends on ongoing behavioral and arousal states, such as attention, eye closure, or sleep. These states display measurable differences in neural manifolds [1], dimensionality [2], noise correlations [3], and autocorrelation timescales [4]. State-dependent changes in the primary visual cortex (V1) are potentially induced by top-down signals, such as those from V4 to V1, known to mediate visual attention in macaques [5]. However, relationships between such top-down signals and the aforementioned state-dependent changes have not been demonstrated directly.
To investigate state-dependent dynamics, we studied the spiking activity in V1 of macaque (N=3, Macaca mulatta), recorded using Utah arrays [6]. We found two distinct neural attractor manifolds, strongly correlated with eyes-open and eyes-closed states, even though the macaques were sitting in a dark room. The neural dynamics during eyes-open periods had a significantly higher dimensionality, primarily due to lower noise correlations.
We hypothesized that the mechanism mediating these state-dependent changes could be top-down communication. We estimated the strength and direction of cortico-cortical communication from LFP coherence and Granger causality, and found that top-down signals from V4 to V1 are significantly stronger during the eyes-open periods. Spectral analysis revealed reduced alpha power during the eyes-open condition, aligning with alpha blocking in EEG studies.
Finally, we constructed a spiking neuron model demonstrating that top-down signals could generate multiple neural manifolds. These manifolds do not reflect modulations in the mean firing rate of the network, but changes in the rates of individual neurons, leading to distinct population activity.
Taken together, our in vivo and in silico analysis suggests that top-down signals can actively modulate the neural manifolds of neural networks. In the case of macaque V1, we postulate that top-down modulation from V4 induces these state-dependent changes in V1 to prepare the visual cortex for fast and efficient visual responses, resulting in a visual stand-by state when the eyes are open.
References: [1] Morales-Gregorio et al. Cell Reports (Accepted in principle), 2024. [2] Stringer et al. Nature 571, 361-365, 2020. [3] Doiron et al. Nat Neurosci 19, 383-393, 2016. [4] Zeraati et al. Nat Comms 14, 1858, 2023. [5] Poort et al. 2012. Neuron 75 (1), 143-156 [6] Chen et al. Sci Data 9, 77, 2022.