Critical assessment of the involvement of perforations, spinules, and spine branching in hippocampal synapse formation.

Several studies propose that long-term enhancement of synaptic transmission between neurons results from the enlargement, perforation, and splitting of synapses and dendritic spines. Unbiased analyses through serial electron microscopy were used to assess the morphological basis for synapse spilitting in hippocampal area CA1. Few perforated synapses and almost no split (i.e., branched) spines occurred at postnatal day 15, an age of high synaptogenesis; thus, synapse splitting is unlikely to be important during development. The synapse splitting hypothesis predicts an intermediate stage of branched spines with both heads sharing the same presynaptic bouton. Ninety-one branched dendritic spines were traced through serial sections, and the different branches never synapsed with the same presynaptic bouton. Projections from spines, called "spinules," have been thought to extend from perforations in the postsynaptic density (PSD), thereby dividing the presynaptic bouton. Forty-six spinules were traced, and only 13% emerged from perforations in the PSD. Most spinules emerged from the edges of nonperforated PSDs, or from spine necks, where they extended into boutons that were not presynaptic to the spine. In summary, these morphological characteristics are inconsistent with synapse and spine splitting. An alternative is discussed whereby perforated synapses and spinules are transient components of synaptic activation, and branched spines appear from synapses forming in close proximity to one another.