Summary: Researchers identify a hidden order in what appeared to be random synaptic connection strengths between neurons.
Source: University Hospital Bonn
The brain constructs perception through vast networks of neurons linked by synapses, where both the number and strength of those links can differ across cell types and over time.
A collaborative team from University Hospital Bonn (UKB), the University Medical Center Mainz, Ludwig-Maximilians-University Munich (LMU) and the Max Planck Institute for Brain Research in Frankfurt, working within the DFG-funded Priority Program “Computational Connectomics” (SPP2041), has uncovered that the apparently irregular distribution of synaptic strengths conceals a reproducible order. This hidden structure helps maintain the stability and function of cortical networks.
The study appears in the journal PNAS.
A decade ago, creating a comprehensive map of the brain’s connections — the connectome — was identified as a major scientific goal. Each neuron can form thousands of synapses, and the strengths of these connections play a central role in learning and cognition. Yet direct measurements have shown wide variability: the same synapse type measured in the same brain region can yield different strength values, complicating efforts to identify general principles of network function.
“Individual synapses are unique and their strengths change with time. That experimental variability has made it hard to see organizing principles that ensure robust neural computation,” explains Prof. Tatjana Tchumatchenko, who leads research groups at the Institute of Experimental Epileptology and Cognitive Research at UKB and at the Institute of Physiological Chemistry at the University Medical Center Mainz. This challenge motivated the combined experimental and theoretical investigation.
Bringing mathematics and experiment together
The team focused on the primary visual cortex (V1), the cortical area that first represents visual signals relayed from the eye through the thalamus. They examined the connections of neurons activated by visual stimuli, combining in vivo recordings with computational modeling.
In mice, the researchers recorded the joint responses of two main neuronal classes to a range of visual inputs while using a nonlinear theoretical framework to infer synaptic connection strengths that could produce the observed activity. Their modeling approach relied on the stabilized supralinear network (SSN) model, a nonlinear network model that allows direct comparison between simulated and experimentally observed neural activity.
“SSN is one of the few nonlinear models that can be grounded against real neuronal response data,” says Prof. Laura Busse, head of an LMU Neurobiology group. “By integrating SSN with recordings from the thalamus and V1, we could infer multiple possible sets of synaptic weights that reproduce the recorded visual responses.”
A consistent sequence of connection strengths is the key
Across the range of inferred configurations, a consistent ordering of connection strengths emerged. Although absolute synaptic magnitudes varied — as in previous experimental reports — their relative ranking remained stable. For instance, excitatory-to-inhibitory connections were consistently among the strongest, while inhibitory-to-excitatory links were typically weaker in the primary visual cortex.
This finding indicates that the network’s response profile depends more on the relative ratios of synaptic weights than on their absolute values. “It’s notable that earlier direct measurements of cortical synapses follow the same ordering our method predicts from neural responses alone,” says Simon Renner, Ph.D., of LMU Neurobiology, whose recordings helped characterize cortical and thalamic activity patterns.
The investigators emphasize that neural activity itself carries rich information about underlying connectivity that may not be obvious from isolated synapse measurements. “Our approach offers a practical route to infer network motifs in brain regions where direct access to all synaptic connections is limited,” says Nataliya Kraynyukova, Ph.D., from UKB and the Max Planck Institute.
This study reflects an interdisciplinary collaboration that combined the computational strengths of Prof. Busse’s lab with the experimental expertise of Prof. Tchumatchenko’s teams, enabling complementary analyses that linked measured responses to plausible network architectures.
About this neuroscience research news
Author: Inka Väth
Source: University Hospital Bonn
Contact: Inka Väth – University Hospital Bonn
Image: The image is in the public domain
Original Research: Open access. “In vivo extracellular recordings of thalamic and cortical visual responses reveal V1 connectivity rules” by Simon Renner et al., PNAS
Abstract
In vivo extracellular recordings of thalamic and cortical visual responses reveal V1 connectivity rules
The brain’s connectome frames canonical neural computations, yet measurements of connectivity in mouse primary visual cortex (V1) have reported variable numbers and strengths of connections between defined neuron types. This raises the question of whether such variability can coexist with stable, canonical computations in V1.
Using a theory-driven approach, the authors inferred V1 network connectivity from visual responses recorded in mouse V1 and the visual thalamus (dLGN). Their method showed that identical recorded responses can be explained by multiple connectivity configurations. Crucially, across most inferred configurations, the magnitudes and selectivity of connectivity weights followed a consistent order.
The authors propose that this ordered relationship between weights arises from characteristic shapes of the recorded contrast response functions and from contrast-invariant orientation tuning. Furthermore, connectivity weights derived from previously published connectivity studies conform to the same order, suggesting that relative relations among synaptic weights—rather than absolute magnitudes—constitute a connectivity motif that supports canonical computations in V1.