Abstract
How different is local cortical circuitry from a random network? To answer this question, we probed synaptic connections with several hundred simultaneous quadruple whole-cell recordings from layer 5 pyramidal neurons in the rat visual cortex. Analysis of this dataset revealed several nonrandom features in synaptic connectivity. We confirmed previous reports that bidirectional connections are more common than expected in a random network. We found that several highly clustered three-neuron connectivity patterns are overrepresented, suggesting that connections tend to cluster together. We also analyzed synaptic connection strength as defined by the peak excitatory postsynaptic potential amplitude. We found that the distribution of synaptic connection strength differs significantly from the Poisson distribution and can be fitted by a lognormal distribution. Such a distribution has a heavier tail and implies that synaptic weight is concentrated among few synaptic connections. In addition, the strengths of synaptic connections sharing pre- or postsynaptic neurons are correlated, implying that strong connections are even more clustered than the weak ones. Therefore, the local cortical network structure can be viewed as a skeleton of stronger connections in a sea of weaker ones. Such a skeleton is likely to play an important role in network dynamics and should be investigated further.
Introduction
A fundamental question in neuroscience is whether the fine-scale wiring of the cortex resembles a random "spaghetti" of connections or a precisely engineered circuit. While the sparsity of cortical connectivity is well-known, statistical sampling has often given the impression of randomness. This study by Song et al. (2005) challenges this view, providing evidence that local circuits are organized into highly structured, non-random motifs.
Methods
The researchers utilized quadruple whole-cell recordings in slices of rat visual cortex (P12–P20). This technique involves placing microscopic electrodes into four neighboring "thick-tufted" Layer 5 pyramidal neurons simultaneously. By stimulating each neuron individually and measuring the resulting Excitatory Postsynaptic Potentials (EPSPs) in the other three, the team mapped every possible directed connection within the four-cell group.
The study analyzed a dataset of 816 quadruple recording attempts, probing 8,050 potential connections. Of these, 931 actual connections were detected, resulting in an overall connectivity rate of 11.6%. Observed patterns were compared against a null model of independent connection probabilities using rigorous statistical corrections.
Results
The analysis revealed several non-random features that define the architecture of local cortical circuits.
Lognormal Distribution of Synaptic Strength
Synaptic strengths were found to follow a heavy-tailed lognormal distribution rather than a normal or Poisson distribution. This results in a "skeleton" of strong connections: just 17% of connections contribute to half of the total synaptic weight in the circuit. EPSP amplitudes ranged over several orders of magnitude, from less than 0.1 mV to over 10 mV, with an average of 0.77 mV.
Reciprocity and Clustering
Bidirectional connections were found to be approximately four times more likely than predicted by a random model ($p < 0.0001$). Furthermore, the strengths of these reciprocal connections were correlated ($R = 0.36$). The network also exhibited significant clustering; for instance, if neuron A connects to B and B connects to C, the probability of a connection between A and C is significantly higher than average. Out of 16 possible three-neuron motifs, several highly connected patterns were significantly overrepresented.
Discussion
The local cortical network structure is best described as a skeleton of strong connections embedded within a "sea" of much weaker ones. This non-random organization suggests that specific motifs, rather than random connectivity, drive cortical computation. The lognormal distribution of weights may be a natural consequence of multiplicative plasticity rules (e.g., LTP/LTD). While this study focused on Layer 5 pyramidal neurons within a 100µm radius, it provides a foundational framework for understanding how structural motifs influence network dynamics and information flow in the brain.
Summary updated