How does neuronal activity affect the development of neural circuits? Work on the retina shows that blocking activity at the synapses between neurons reduces local synapse assembly without affecting global cellular structure.
Nervous system function depends on the organization of underlying neural circuits — groups of neurons whose ability to perform specific functions depends on an organized pattern of intercellular communication at neuron junctions (synapses). The importance of synaptic activity in the assembly of circuits during development has been a long-standing debate in neurobiology: a dominant model in the field1,2 suggests that the amount of neurotransmission at synapses during maturation of the nervous system affects the large-scale arrangement of neural circuits, and even the structure of their component neurons. On page 1016 of this issue, Kerschensteiner et al.3 present exciting work that revises this model. The authors show that, in the retina of mice, excitatory synaptic activity determines the density of synapses at individual neurons (a local effect), but does not affect cellular structure on a global scale.
Synaptic organization has been studied extensively in the mammalian retina, a tissue comprised of about 15 discrete circuits, which can be distinguished both functionally and anatomically4. Retinal circuits are activated by photoreceptors (rods or cones), which transform light energy into neural signals and make contacts with both excitatory interneurons (bipolar cells) and inhibitory interneurons (horizontal cells) that process information within the retina. Subsets of bipolar cells and another class of inhibitory interneuron (amacrine cells) converge selectively onto one of about 15 types of ganglion cell. Long, thread-like projections, or axons, from these output neurons form the optic nerve and travel into the brain. The circuit of each ganglion-cell type determines its specific sensitivity to contrast, colour, motion and other features of the visual input.
A fundamental organizing principle of the retina is its division into ON and OFF pathways4. Bipolar cells are excited by either increases (ON cells) or decreases (OFF cells) in light; the axons of ON and OFF bipolar cells terminate in different sub-layers of the inner synaptic layer of the retina (Fig. 1a). Here, they release the neurotransmitter glutamate, which crosses synapses and activates excitatory receptors on short processes called dendrites that project from ganglion cell bodies.
Early in postnatal development in the mouse, before eye opening, ganglion cells extend dendrites to make contact with bipolar cells in both the ON and OFF retinal sub-layers. At maturity (three weeks of age), some ganglion cells remain bistratified, with dendrites in both layers, and thereby receive a mixed ON–OFF input4. Others lose their dendrites in either the ON or OFF sub-layer, maturing to become monostratified ganglion cells, which make contact with only one type of bipolar cell (ON or OFF) (Fig. 1a). Hence, the category of a ganglion cell's response depends both on the morphology of its dendrites, which determines potential synaptic partners, and the presence of synapses, which enables actual communication with those partners.
The pruning of dendrites from ganglion cells as they become monostratified is thought to be induced by differences in excitatory synaptic activity, with the elimination of synapses (and eventually dendrites) from less-active inputs2. Kerschensteiner and colleagues3 re-examine this hypothesis by generating mice with imbalances in synaptic activity at retinal ganglion cells. The ON bipolar cells in these mice were induced to express tetanus toxin (TeNT), which cleaves a protein needed to release neurotransmitters at synapses5. So, although the ON bipolar cells are structurally normal, synaptic neurotransmitter release from these cells is markedly reduced, and ON ganglion cells lack almost all excitatory synaptic input, both spontaneous and light-evoked.
Kerschensteiner et al. found that blocking synaptic activity of ON bipolar cells altered local ganglion-cell structure. The authors used confocal microscopy of fluorescently tagged proteins associated with glutamate receptors to visualize synapses in ganglion cells. They found that these receptors were in the correct location — near the silent ON bipolar-cell terminals — but that their number was halved. Furthermore, they found that the reduced density of synapses did not result from increased synapse elimination, but rather from a reduced rate of synapse formation between ON bipolar cells and ON ganglion cells.
Surprisingly, the reduced number of synapses in the TeNT mouse did not alter the stratification of ganglion cells. Indeed, in mature mice, dendrites of ON ganglion cells made contact with axon terminals of the silent ON bipolar cells — somehow, the dendrites of ON ganglion cells found the correct target layer despite the lack of ON bipolar-cell synaptic activity. Moreover, bistratified ganglion cells also had normal dendrite morphology despite the imbalance in synaptic activity (active neurotransmitter release from OFF bipolar-cell synapses, but not from ON synapses) (Fig. 1b). Thus, ON and OFF bipolar cells that converge on the same ganglion cell apparently do not coordinate their synapse number so that synapse density is balanced. They also don't seem to compete for 'postsynaptic real estate' — if they did, the density of synapses at dendrites in contact with the active OFF bipolar cells would have increased in TeNT mice compared with normal mice.
Kerschensteiner and colleagues' results3 raise an interesting question: why have previous studies shown that manipulations of synapse activity alter dendrite stratification whereas the TeNT manipulation did not? In cats and ferrets, intraocular injections of a drug known to hyperpolarize, and thereby functionally inactivate, ON bipolar cells disrupted stratification of ganglion-cell dendrites2,6. In mice, the normal stratification of ganglion-cell dendrites was delayed by altering spontaneous activity during development and was disrupted by depriving the retina of normal light input2.
One possible explanation is that the TeNT mouse has retained several forms of synaptic activity. For example, OFF bipolar-cell synapses were apparently functioning normally, and some of their neurotransmitter could have reached ON ganglion cells through synaptic 'spill-over'7. Furthermore, the inhibitory interneurons of the TeNT mouse were presumably functional, and their synaptic release might have influenced dendritic refinement during development.
A second possibility is that manipulations in previous studies2,6 had effects beyond reducing signalling through the ON bipolar-cell pathway. For instance, these previous manipulations may have unexpectedly affected the expression of molecules that guide the development of global cellular structure. Indeed, the layered arrangement of axons and dendritic processes in zebrafish8 and chick9 retina depend mostly on expression of adhesion molecules and relatively little on neural activity. However, it is difficult to exclude a role for synaptic activity in mammalian retinal development because the effects of previous experimental manipulations2,6 on activity have not been fully explored. Determining whether these manipulations2,6 acted by altering a conventional synaptic process (for example, amacrine-cell activity) or levels of protein expression (for example, temporal expression of adhesion molecules) awaits further investigation.
Kerschensteiner and colleagues' findings in the retina3 contrast with findings in other parts of the nervous system1,10,11 in which synaptic transmission is essential for determining neuronal structure. During the development of the neuromuscular junction, muscles are initially in contact with multiple motor neurons. But activity-dependent synaptic refinement leads to pruning of axons so that, at maturity, a single axon innervates the muscle10. Furthermore, the dendritic structure of frog tectal midbrain neurons, which receive input from retinal ganglion-cell axons, is strongly dependent on the activity-dependent formation of glutamate synapses1, as is the dendritic structure of Purkinje neurons in the mammalian cerebellum11. It will be interesting to learn whether certain features of a neural circuit — such as the level of plasticity at maturity — determine the relative importance of activity-dependent versus molecular-guidance cues for circuit assembly during development.
Cline, H. & Haas, K. J. Physiol. (Lond.) 586, 1509–1517 (2008).
Tian, N. J. Physiol. (Lond.) 586, 4347–4355 (2008).
Kerschensteiner, D. et al. Nature 460, 1016–1020 (2009).
Wässle, H. Nature Rev. Neurosci. 5, 747–757 (2004).
Schiavo, G. et al. Nature 359, 832–835 (1992).
Bodnarenko, S. R. & Chalupa, L. M. Nature 364, 144–146 (1993).
Blankenship, A. G. et al. Neuron 62, 230–241 (2009).
Nevin, L. M., Taylor, M. R. & Baier, H. Neural Dev. 3, 36 (2008).
Sanes, J. R. & Yamagata, M. Annu. Rev. Cell Dev. Biol. doi:10.1146/annurev.cellbio.24.110707.175402 (2009).
Sanes, J. R. & Lichtman, J. W. Annu. Rev. Neurosci. 22, 389–442 (1999).
Bosman, L. W. & Konnerth, A. Neuroscience 162, 612–623 (2009).