Open Access

Ali Karimi

• The central goal of this study was to generate synapse-resolution maps of local and long-range innervation on apical dendrites (AD) in mouse cerebral cortex. We used three-dimensional electron microscopy (3D-EM) to first measure the cell-type specific balance in the excitatory and inhibitory input on ADs. Further, we found two inhibitory axon populations with preference for apical dendrites originating from layer 2 and 3/5. Additionally, we used a combination of large-scale volumetric light and electron microscopy to investigate the innervation preference of long-range cortical projections onto ADs. To generate such large-scale 3D-EM datasets, we also developed a software package to automate aberration adjustment. The balance of excitation and inhibition defines the computational properties of neurons. We, therefore, generated 6 datasets and annotated 26,548 excitatory and inhibitory synapses to map the relative inhibitory strength on the AD of pyramidal neurons in layers 1 and 2 (L1 and 2) of the cortex. We found consistent and cell-type specific patterns of inhibitory strength along the apical dendrite of L2-5 pyramidal neurons in primary somatosensory (S1), secondary visual (V2), posterior parietal (PPC) and anterior cingulate (ACC) cortices. L2 and L5 pyramidal neurons had inhibitory hot-zones at their main bifurcation and distal apical dendrite tuft, respectively. In contrast, L3 neurons had a baseline (~10%) level of inhibition along their apical dendrite. As controls, we quantified the effect of synapse strength (size), dendrite diameter, AD classification and synapse identification methods on the cell-type specific synapse densities. To classify L5 pyramidal subtypes, we performed hierarchical clustering using morphological properties that were described to differentiate slender- and thick-tufted L5 neurons. We also investigated the distance to soma as a predictor of fractional inhibition around the main bifurcation of apical dendrites. Interestingly, we found a strong exponential relationship that was absent in density of either synapse type. This suggests a distance dependent control mechanism designed specifically for the balance (in synapse numbers) of excitation and inhibition. Next, we focused on the inhibitory innervation preference for apical dendrite of pyramidal neuron. We, therefore, annotated 5,448 output synapses of AD-targeting inhibitory axons and found two populations specific for either L2 or L3/5 apical dendrites. Together with previous findings on preferential innervation of sub-cellular structures by inhibitory axons, this suggests two distinct inhibitory circuits for control of AD activity in L2 vs. deep-layer pyramidal neurons. This innervation preference was surprisingly consistent across S1, V2, PPC and ACC cortices. 3D-EM data acquisition is a laborious process that is made easier and more popular everyday by technical progress in the laboratory and industrial settings. To make data acquisition robust using our custom-built 3D-EM microscopes, an automatic aberration software was implemented to adjust the objective lens and the stigmators of the electron microscope. This method was used in multiple month-long experiments across 2 microscopes and 10 datasets. The aberration adjustment used the reduction in image details (high-frequency elements) to estimate the level of deviation from optimal focus and stigmator parameters. However, large objects in EM micrographs such as blood vessel and nuclei cross-sections generated anomalous results. We, therefore, added image processing routines based on edge detection combined with morphological operations to exclude such large objects. Finally, we performed a correlative three-dimensional (3D) light (LM) and electron (EM) microscopy experiment to map the long-range primary visual (V1) and secondary motor (M2) cortical input to ADs in layer 1 of PPC using the “FluoEM” approach. This method allows for identification of the long-range source of projection axons in EM volumes without the need for EM-dense label conversion or heat-induced markings. The long-range source of an axon in EM is identified based on the fluorescent protein that is expressed in its LM counterpart. In comparison to M2 input, Long-range axons from V1 had a higher tendency to target L3 pyramidal neurons in PPC according to our preliminary analysis. In combination with the difference observed in the synapse composition of L2 and L3 apical dendrites, this suggests the need for separate functional and structural analysis of L2 and 3 pyramidal neurons.