Synapse distribution suggests a two-stage model of dendritic integration in CA1 pyramidal neurons

Abstract

Competing models have been proposed to explain how neurons integrate the thousands of inputs distributed throughout their dendritic trees. In a simple global integration model, inputs from all locations sum in the axon. In a two-stage integration model, inputs contribute directly to dendritic spikes, and outputs from multiple branches sum in the axon. These two models yield opposite predictions of how synapses at different dendritic locations should be scaled if they are to contribute equally to neuronal output. We used serial-section electron microscopy to reconstruct individual apical oblique dendritic branches of CA1 pyramidal neurons and observe a synapse distribution consistent with the two-stage integration model. Computational modeling suggests that the observed synapse distribution enhances the contribution of each dendritic branch to neuronal output.

Figure 1. Contrasting Models of Synaptic Integration / Spines Near and Far from the Primary Apical Dendrite

(A) Schematic showing the predicted distribution of synaptic conductance along an individual apical oblique dendritic branch if synapses are scaled to normalize somatic EPSP amplitude (left) or the probability of initiating a dendritic spike (right). (B) Serial-section electron microscopic view of a segment of dendrite with spines and synapses. (C) Three-dimensional reconstruction of the segment shown in (B). Dendrite is shown in grey, spines are shown in purple, and postsynaptic densities (PSDs) are shown in blue. Top panel is the reconstruction in its orientation as shown in the electron micrographs. Bottom panel is the reconstruction rotated 180°. (D) Higher magnification serial electron micrographs of the boxed regions in (B). Electron micrographs show two dendritic spines (sp1 and sp2) making synapses with presynaptic axon terminals (at1 and at2). The borders of the PSDs are marked by white arrowheads and the full arrow indicates a perforation. (E) Three-dimensional reconstructions of segments of the same dendrite (bottom three segments) and different dendrites (top two segments). (F) Scatter plot showing spine densities in dendritic segments near branch origins, near branch ends, and near branch centers. Open symbols represent segments from the same branch, filled symbols represent segments from different branches, and lines show the means. (G) Histogram showing the relative and cumulative frequencies of spine volumes in dendritic segments near branch origins, near branch ends, and near branch centers.

Figure 2. Synapses near and far from the primary apical dendrite

(A) Reconstructions of several spines and synapses. (B) Histogram showing the relative and cumulative frequencies of PSD areas in dendritic segments near branch origins, near branch ends, and near branch centers. (C) Electron micrographs of serial sections showing postembedding immunogold labeling of AMPA receptors. (D) Three-dimensional reconstruction of synapses (blue), their spines (purple), and the synaptic immunogold particles (black) shown in C. (E) Correlation of particle number with spine volume. (F) Correlation of particle number with PSD area.

Figure 3. Functional consequences of synapse distribution

(A) The reconstructed CA1 pyramidal neuron morphology used for the models and the apical oblique dendrites used for the simulations (region between dashed lines enlarged). (B) Sample voltage traces from a simulation in which fifteen synapses were randomly selected and activated on branch 4 in the two-stage integration model (188 μm from the soma). Voltage is indicated at the branch end (black trace), center of the branch (red trace), branch origin (green trace), and soma (blue trace). Synapses were distributed along each branch shown in (A) with decreasing density and strength from branch origin to branch end as observed experimentally (two-stage integration model) and the reverse (global integration model). (C) For each branch, synapses were randomly activated until a dendritic spike occurred (200 trials/branch). Probability distribution (bin size = 0.2) showing the location along the dendritic branch of the input triggering the spike (the last synapse selected) for the two models (black and blue lines). The dashed line indicates the uniform distribution expected if the final input contributes equally to the spike at all locations. (D) Somatic depolarization averaged over 500 trials resulting from activation of 5, 10, 15, and 20 synapses in the two models: global integration (black bars) and two-stage integration (blue bars). The white numbers indicate the percentage of trials that resulted in a dendritic spike. (E) Top: The percentage of trials resulting in a dendritic spike for the global (black bars) and two-stage (blue bars) integration models when ten synapses are activated on each branch (500 trials per branch). Bottom: The average somatic depolarization resulting from these simulations shown separately for trials in which a dendritic spike was triggered (dark bars) and trials that did not produce a dendritic spike (light bars). (F) Schematic of the proposed synapse distribution for the CA1 apical dendritic tree based upon previous (Nicholson et al., 2006, Magee and Cook, 2000) and current results.