Intracortical circuits of pyramidal neurons reflect their long-range axonal targets

Abstract

Cortical columns generate separate streams of information that are distributed to numerous cortical and subcortical brain regions. We asked whether local intracortical circuits reflect these different processing streams by testing whether the intracortical connectivity among pyramidal neurons reflects their long-range axonal targets. We recorded simultaneously from up to four retrogradely labelled pyramidal neurons that projected to the superior colliculus, the contralateral striatum or the contralateral cortex to assess their synaptic connectivity. Here we show that the probability of synaptic connection depends on the functional identities of both the presynaptic and postsynaptic neurons. We first found that the frequency of monosynaptic connections among corticostriatal pyramidal neurons is significantly higher than among corticocortical or corticotectal pyramidal neurons. We then show that the probability of feed-forward connections from corticocortical neurons to corticotectal neurons is approximately three- to fourfold higher than the probability of monosynaptic connections among corticocortical or corticotectal cells. Moreover, we found that the average axodendritic overlap of the presynaptic and postsynaptic pyramidal neurons could not fully explain the differences in connection probability that we observed. The selective synaptic interactions we describe demonstrate that the organization of local networks of pyramidal cells reflects the long-range targets of both the presynaptic and postsynaptic neurons.

Figure 1

Different frequencies of monosynaptic connections between corticotectal, corticostriatal or corticocortical neurons. Presynaptic action potentials elicit a synaptic response in a postsynaptic cell during simultaneous recordings from two monosynaptically-connected corticotectal neurons (a), corticostriatal neurons (b) and corticocortical neurons (c). d, The frequency of identified monosynaptic connections among connections tested is shown for corticotectal connections, corticostriatal connections and corticocortical connections; **P < 0.05.

Figure 2

Feedforward synaptic connections between presynaptic corticocortical neurons and neighbouring corticotectal neurons are almost four times more likely than connections among corticocortical neurons. a, An example of a quadruple recording. Three corticocortical neurons (red) and one corticotectal neuron (green) were recorded simultaneously and the 12 possible synaptic connections were tested. The corticocortical neurons, CC1 and CC2, synapsed onto the neighbouring corticotectal neuron. The corticotectal neuron in turn synapsed onto the corticocortical neuron, CC3. b, The frequency of synaptic connections identified among the tested connections is shown for the four possible types of connections among corticocortical and corticotectal neurons; **P < 0.01.

Figure 3

The average axonal and dendritic architecture alone cannot explain differences in the connection probability. The morphology of 15 corticotectal (a), corticostriatal (b) and corticocortical neurons (c; blue: dendrites; red: axons; black: somas). d, The dendritic and axonal length-density maps were used to estimate the spatial overlap of the neuronal processes for the five types of connections tested physiologically. The resulting maps of axodendritic overlap, generated from cells aligned relative to the pial margin and shifted 50 µm relative to each other, are shown. Scale bars, 200 µm. e, The probability of physiological connection and the average axodendritic overlap are plotted for each type of connection tested.