Communication between pre- and postsynaptic cells promotes the initial organization of synaptic specializations, but subsequent synaptic stabilization requires transcriptional regulation. Here we show that fibroblast growth factor 22 (FGF22), a target-derived presynaptic organizer in the mouse hippocampus, induces the expression of insulin-like growth factor 2 (IGF2) for the stabilization of presynaptic terminals. FGF22 is released from CA3 pyramidal neurons and organizes the differentiation of excitatory nerve terminals formed onto them. Local application of FGF22 on the axons of dentate granule cells (DGCs), which are presynaptic to CA3 pyramidal neurons, induces IGF2 in the DGCs. IGF2, in turn, localizes to DGC presynaptic terminals and stabilizes them in an activity-dependent manner. IGF2 application rescues presynaptic defects of Fgf22-/- cultures. IGF2 is dispensable for the initial presynaptic differentiation, but is required for the following presynaptic stabilization both in vitro and in vivo. These results reveal a novel feedback signal that is critical for the activity-dependent stabilization of presynaptic terminals in the mammalian hippocampus. DOI: http://dx.doi.org/10.7554/eLife.12151.001, eLife digest Nerve cells in the developing brain must organize themselves into complex networks by forming appropriate connections with one another. These connections are known as synapses, and they assemble via two critical stages. First, a new synapse forms, and then it stabilizes. This first stage is a localized event that involves the contact site between the two nerve cells, while the stabilization of a synapse requires the expression of genes in a nerve cell’s nucleus. Furthermore, only active synapses may be stabilized. Many synapses form in a region of the brain called the hippocampus, which plays a key role in learning and memory. A protein called fibroblast growth factor 22 (or FGF22 for short) helps synapses to initially form within the hippocampus. However, much less is known about the signals that regulate the stabilization of synapses and the genes that are involved. It is also not clear if these genes might be controlled by FGF22 signaling. To address these questions, Terauchi et al. searched the mouse hippocampus for genes with expression that depended on FGF22 signaling. One gene in particular, which encodes a protein called insulin-like growth factor 2 (IGF2), was much less expressed in mice that lack FGF22 compared to normal mice. Further experiments revealed that only active nerve cells transport IGF2 to synapses, and that IGF2 helps to stabilize these structures. By contrast, IGF2 is not required for synapse to initially form. This indicates that FGF22 controls both the formation and stabilization of synapses, and that it controls the first stage directly, and the second stage indirectly via its effects on IGF2 expression. Terauchi et al. also showed that FGF22-IGF2 signaling is not involved in the stabilization of all synapses in the mouse hippocampus. Instead, synapses between different types of nerve cell appear to use distinct signals for synapse formation and stabilization. A key topic for future studies will be to understand these specific signals and how they cooperate in the brain to establish precise networks of nerve cells. DOI: http://dx.doi.org/10.7554/eLife.12151.002