Analysis and comparison of morphological reconstructions of hippocampal field CA1 pyramidal cells

Abstract

Morphological reconstructions have become a routine and valuable tool for neuroscientists. The accuracy of reconstructions is a matter of considerable interest given that they are widely used in computational studies of neural function. Despite their wide usage, comparisons of reconstructions obtained using various methodologies are lacking. We reviewed reconstructions of hippocampal CA1 pyramidal cells from five published studies and found marked differences in some of the most basic measurements. For four of the five studies means of total cell length clustered in the 11,479-13,417-microm range. The remaining study had a significantly larger value for this index at 16,992+/-5,788 microm. Surface area means varied more than 4-fold from 16,074 to 67,102 microm2. Volume means varied more than 8-fold from 3,828 to 30,384 microm3. Simulated passive input resistance means varied from 38.0 to 172.1 MOmega, reflecting the variability in cell dimensions. Estimates of the electrotonic length varied from 1.26 to 1.56. In two reconstructions used in previously published studies, simulated somatic excitatory postsynaptic potentials (EPSPs) varied 2-4-fold in amplitude, time to peak and half-width, for synaptic inputs at similar locations. Substantial jitter on the z-axis was identified as one likely source of the discrepancy in total cell length, while substantial differences in diameter measurements across studies, and sometimes within the same study, accounted for the variability in surface area and volume. While some part of the observed variability is surely due to the diversity of CA1 pyramidal cells, our analysis suggests that a substantial portion stemmed from methodological inconsistencies and from technological limitations. Suggestions are made for improving the quality and usefulness of morphological reconstructions. We conclude that reconstructions across studies have substantial variability in measures that are very relevant to neuronal function. Consequently, modelers are advised to use more than just one reconstructed cell in their simulations of neural function.