How to count synapses unbiasedly and efficiently at the ultrastructural level: proposal for a standard sampling and counting protocol.

After almost 40 years, there is still no consensus on criteria for identifying different types of synapse seen in electron microscopical thin sections or on methods for counting them unbiasedly in 3D. This review proposes a procedure which meets these aims and could be adopted as a standard best-practice sampling and counting convention. It deals exclusively with unbiased stereological methods for counting particles in 3D space because these are efficient and applicable to arbitrary particles regardless of their size, shape and orientation. Methods based on individual sections are excluded because arbitrary particles cannot be counted unbiasedly with such sections. Model-based methods (e.g. treating synaptic membrane densities as circular disks) are excluded because they are not unbiased in general and now have limited (mainly historical) interest only. For unbiased counting, the absolute minimum requirement is a pair of parallel sections (dissector). The following protocol is recommended for future studies on synapse number: (1) use para(membrane) densities as synaptic counting units, (2) do not qualify definition of the counting unit by reference to a minimum number of synaptic vesicle profiles, (3) sample and count synapses unbiasedly using the dissector, and (4) in preference convert number per volume into absolute number or, in this is not possible, estimate a synapse-to-neuron ratio.