The cell biology of synapse formation

Abstract

In a neural circuit, synapses transfer information rapidly between neurons and transform this information during transfer. The diverse computational properties of synapses are shaped by the interactions between pre- and postsynaptic neurons. How synapses are assembled to form a neural circuit, and how the specificity of synaptic connections is achieved, is largely unknown. Here, I posit that synaptic adhesion molecules (SAMs) organize synapse formation. Diverse SAMs collaborate to achieve the astounding specificity and plasticity of synapses, with each SAM contributing different facets. In orchestrating synapse assembly, SAMs likely act as signal transduction devices. Although many candidate SAMs are known, only a few SAMs appear to have a major impact on synapse formation. Thus, a limited set of collaborating SAMs likely suffices to account for synapse formation. Strikingly, several SAMs are genetically linked to neuropsychiatric disorders, suggesting that impairments in synapse assembly are instrumental in the pathogenesis of neuropsychiatric disorders.

Figure 1.

Synapses are communication nodes that connect neurons into circuits. (A) Electron micrograph of a human synapse with two synaptic junctions to illustrate the canonical features of all synapses: An intercellular junction in which a presynaptic varicosity that is filled with synaptic vesicles contacts a postsynaptic dendrite that contains multiple trafficking organelles as well as ribosomes (image courtesy of Dr. Christopher Patzke). Red arrows indicate synaptic junctions. Most neurons form thousands of input and output synapses. (B) Schematic view of a cortical microcircuit in which two pyramidal neurons both directly excite a postsynaptic pyramidal neuron and indirectly inhibit it via an interneuron. If the presynaptic neurons fire in bursts and trains, as is commonly observed in brain, the postsynaptic pyramidal neuron will exhibit differential increasing or decreasing responses depending on whether the various excitatory and inhibitory synapses are facilitating or depressing. (C) Flowchart of the lifecycle of a synapse. After neurons are born, migrate to their appropriate positions, and extend dendrites and axons, neurons form synapses. Synapses initiate as nascent contacts that mature into functional but plastic synaptic connections and are eliminated under control of unknown signals. Synapse turnover rates vary, but many synapses are continuously renewed. (D) Schematic of nascent synapses (left), mature synapses (center), and synapses being eliminated (right). In nascent synapses, transneuronal interactions mediated by SAMs such as latrophilins are proposed to initiate the intracellular signaling cascades that organize synaptic specializations. Subsequent synapse maturation and shaping of synapse properties (center) is controlled by a different set of SAMs such as neurexins. During synapse elimination, SAM interactions weaken, which may induce separation of synaptic junctions and withdrawal of synaptic processes. (E) Schematic of how SAMs organize synapse formation and synapse elimination. CASK, calcium/calmodulin dependent serine protein kinase; Cblns, cerebellins; GluD, δ-type glutamate receptor; Lphns, latrophilins; Nlgns, neuroligins.

Figure 2.

Synapses are composed of presynaptic specializations containing a canonical neurotransmitter release machinery and postsynaptic specializations constructed of diverse receptors and postsynaptic densities. The molecular composition of the presynaptic specialization is largely independent of the neurotransmitter type, with similar proteins mediating the localized and fast Ca²⁺-dependent fusion of synaptic vesicles (Südhof, 2013). In contrast, postsynaptic specializations are diverse, with little overlap in their molecular components. Four types of receptors are associated with distinct postsynaptic molecular complexes: glutamate receptors (center) account for ∼80% of synapses, pentameric cys-loop receptors (GABA_A, glycine, acetylcholine, and serotonin, left) for ∼20% of synapses, and the remaining two receptor classes (metabotropic GPCRs and P2X receptors, right) for <1% of synapses (note that metabotropic GPCRs and P2X receptors are abundantly present outside of synapses). Whereas the only difference among various presynaptic specializations are the enzymes and vesicular transporters specific for particular neurotransmitters (summarized on the top right), few components of different postsynaptic specializations are currently known to be shared, including neuroligin-3, a SAM that binds to presynaptic neurexins. AcCh, acetylcholine; GluA, AMPA-type glutamate receptor; GluD, δ-type glutamate receptor; GluK, kainate-type glutamate receptor; GluN, NMDA-type glutamate receptor; Nlgn, neuroligin; Rec., receptor; STED, stimulated emission depletion; Syts, synaptotagmins.

Figure 3.

Synapses, monitored via spines as proxies, are continuously replaced under physiological conditions, with different turnover rates in various brain regions. (A) 2-Photon stimulation emission depletion (STED) images of spines on basal dendrites of CA1 pyramidal neurons in vivo at three time points separated by 2 d, illustrating rapid turnover of spines (blue arrowheads, stable spines; red arrowheads, lost spines; green arrowheads, new spines; white arrowhead, axonal bouton [AB]). Data in A–E are from Pfeiffer et al. (2018). (B and C) Quantification of the density (B) and survival fraction (C) of dendritic spines over 4 d (n = 14 dendrites, 3 mice). (D and E) Fraction of lost spines (D) and new spines (E) measured over the first or second 2-d period. Thin gray lines represent the measurements of single dendrites. (F) Summary of the relative turnover rates of dendritic spines in the hippocampus and cortex as determined by Attardo et al. (2015). Rec., receptor.

Figure 4.

Schematic diagram of candidate trans-synaptic SAM complexes governing synapse assembly. Data were assembled from the literature and are represented graphically similar to Südhof (2018). Note that two families of presynaptic SAMs, neurexins and LAR-type receptor phosphotyrosine phosphatases (PTPRD, PTPRF, and PTPRS), are hub molecules that interact with a series of postsynaptic SAM families and also bind to each other in cis (Han et al., 2020). Most candidate SAMs perform additional functions outside of synapses. Lines and arrows indicate interactions, with cis-interactions shown as dotted lines and less validated trans-interactions shown as dashed lines. DCC, deleted in colorectal cancer; EphB, Ephrin B; FLRT, fibronectin leucine-rich transmembrane; LRRTM, leucine-rich repeat transmembrane; Rec., receptor; RTN, reticulon; SALMs, synaptic adhesion-like molecules; SliTrks, Slit- and Trk-like proteins; SynCAM, synaptic cell adhesion molecule; TrkC, tropomyosin receptor kinase C.

Figure 5.

Synapse numbers and properties are shaped by multiple independent molecular mechanisms: Example of the contributions of neuroligins and SPARCL1 (Hevin). (A–E) Exemplary immunocytochemistry (A and B) and electrophysiology (C–E) experiments with cultured hippocampal neurons demonstrating that SPARCL1 and neuroligins differentially and independently control synapses. The immunocytochemistry data (A and B) show that SPARCL1 increases excitatory but not inhibitory synapse numbers, whereas deletion of all neuroligins has no effect on synapse numbers and does not impair the SPARCL1-induced increase in synapse numbers. The electrophysiology results (C–E) show that SPARCL1 increases, whereas the pan-neuroligin deletion decreases, NMDAR-mediated synaptic strength significantly more than AMPAR-mediated synaptic strength. Although these two manipulations act similarly but in opposite directions, they do not depend on each other (D). Only the neuroligin but not the SPARCL1 manipulation affects inhibitory synapse (E). Data are adapted from Gan and Südhof (2020). (F and G) Phase diagram of the effect of SPARCL1, neuroligins, and latrophilin-3 manipulations on excitatory (F) and inhibitory (G) synapses, as analyzed in cultured hippocampal neurons. Values were computed from Gan and Südhof (2020) and Sando et al. (2019). Numerical data in B, D, and E are means ± SEM. Statistical significance was assessed by two-way ANOVA followed by post hoc corrections. Ctrl, control; EPSC, excitation postsynaptic current; IPSC, inhibition postsynaptic current; KO, knockout. In B, D, and E, asterisks indicate statistical significance as calculated by two-way ANOVA (*, P < 0.05; **, P < 0.01; ***, P < 0.001).