Some forms of synaptic plasticity require rapid, local activation of protein synthesis. Although this is thought to reflect recruitment of mRNAs to free ribosomes, this would limit the speed and magnitude of translational activation. Here we provide compelling in situ evidence supporting an alternative model in which synaptic mRNAs are transported as stably paused polyribosomes. Remarkably, we show that metabotropic glutamate receptor activation allows the synthesis of proteins that lead to a functional long-term depression phenotype even when translation initiation has been greatly reduced. Thus, neurons evolved a unique mechanism to swiftly translate synaptic mRNAs into functional protein upon synaptic signaling using stalled polyribosomes to bypass the ratelimiting step of translation initiation. Because dysregulated plasticity is implicated in neurodevelopmental and psychiatric disorders such as fragile X syndrome, this work uncovers a unique translational target for therapies.M ost studies of translational control focus on initiation, the process where mRNAs recruit ribosomes and catalyze the first step of translation (1). This highly regulated and normally rate-limiting step of translation is followed by elongation and termination, resulting in completed proteins. Although multiple ribosomes on a given mRNA (a polyribosome) imply active peptide synthesis, we and others identified neuronal RNA granulesmotile aggregates of nontranslating ribosomes (2, 3). These electron-dense bodies contain single copies of synaptic mRNAs that are translationally silenced during their transport from soma to synapse (1, 4).Many models assume that neuronally transported mRNAs are translationally paused before completion of the initiation step of translation during transport. An appropriate synaptic signal would then activate translation (initiation/elongation/termination) of the granule mRNA. However, it is not clear how many free ribosomal subunits are present at synapses to support translation initiation. Further, at a typical translation elongation rate of six amino acids per s (5, 6), synthesis of larger synaptic proteins (e.g., microtubule-associated protein 1b; MAP1b) would take over 5 min even if initiation were immediate. These two factors constrain the speed and magnitude of synaptic translation and, thus, plasticity. As some forms of synaptic plasticity require rapid (<10 min) and localized activation of protein synthesis, an alternative model is wanting (7-9).We have previously proposed the concept of a neuronal RNA granule as a stalled polyribosome (10, 11). Ribosomal stalling has been shown to occur in lysates from a mouse neuroblastoma cell line and in an in vitro rabbit reticulocyte lysate translation assay programmed with brain homogenate (12). Whether neuronal ribosome stalling occurs in vivo is uncertain. We hypothesized that neuronal RNA granules contain paused ribosomes with incomplete proteins initiated in the soma before their packaging and transport to dendrites, where translation can be rapidly and loca...