Carla J. Shatz

Developmental mechanisms that generate precise patterns of neuronal connectivity

The functioning of the nervous system depends upon the underlying detailed and highly stereotyped patterns of neuronal connectivity. How these precise patterns of synaptic connections form during development is the subject of this review. The specificity of synaptic connections unfolds in three major steps: pathway selection, target selection, and address selection. First, the growing tips of neurons, the growth cones, traverse long distances to find their correct target region. En route, they are confronted by a series of choice points and yet correctly navigate these pathways in a remarkably unerring way. Once they reach the correct neighborhood, they contact and recognize their correct target, typically a regionally localized set of neurons. In this way, the overall scaffold of projections and synapses is initially established. But these initial patterns of connections are then refined, as axonal terminals retract and expand to select a specific subset of cells from. within the overall target. This remodeling, called address selection, relies on the context of and competition with surrounding inputs and is capable of transforming a coarse-grained and overlapping projection into a refined and highly tuned pattern of connections. Experiments over the last few decades have clarified the issue of how neural specificity is generated by suggesting that two broad mechanisms work in concert to orchestrate the formation of precise patterns of neural connections during development: those that require neuronal activity (activity dependent), and those that do not (activity independent). The initial steps of growth cone guidance typically occur before neurons become functionally active and rely on molecular mechanisms of pathway and target recognition that are largely activity independent. These mechanisms bring together multiple inputs with appropriate targets to form initial patterns of connections. From this point on, the patterns of neuronal activity within these emerging patterns of connections take over as the predominant mechanism that drives the refinement and remodeling of these initial projections into highly tuned and functioning circuits. This process of activity-dependent synaptic plasticity does not stop at birth but continues throughout the lifetime of the organism as the patterns of neural activity driven by input from the external world continue to modify the strength and structure of dendrites, axonal arbors, and synapses (e.g., Purves et al., 1986; Bailey and Chen, 1989). The growing evidence that adults and embryos may use common molecules and mechanisms to modify their synapses (Cline and ConstantinePaton, 1989; Mayford et al., 1992) has linked the once separate fields of developmental neurobiology and learning and memory. In this review, we consider the range of activityindependent and activity-dependent mechanisms that generate precision of neuronal connections. We then focus on two examples at opposite ends of the spectrum-the connections between motoneurons and muscles and between the retina and higher visual centers--to highlight the extent to which different parts of the nervous system use the same mechanisms but in different proportions to achieve the final specificity.

Ocular dominance in layer IV of the cat's visual cortex and the effects of monocular deprivation.

1. The relation between the physiological pattern of ocular dominance and the anatomical distribution of geniculocortical afferents serving each eye was studied in layer IV of the primary visual cortex of normal and monocularly deprived cats. 2. One eye was injected with radioactive label. After allowing sufficient time for transeuronal transport, micro‐electrode recordings were made, and the geniculocoritcal afferents serving the injected eye were located autoradiographically. 3. In layer IV of normal cats, cell were clustered according to eye preference, and fewer cells were binocularly driven than in other layers. Points of transition between groups of cells dominated by one eye and those dominated by the other were marked with electrolytic lesions. A good correspondence was found between the location of cells dominated by the injected eye and the patches of radioactively labelled geniculocortical afferents. 4. Following prolonged early monocular deprivation, the patches of geniculocortical afferents in layer IV serving the deprived eye were smaller, and those serving the non‐deprived eye larger, than normal. Again there was a coincidence between the patches of radioactively labelled afferents and the location of cells dominated by the injected eye. 5. The deprived eye was found to dominate a substantial fraction (22%) of cortical cells in the fourth layer. In other cortical layers, only 7% of the cells were dominated by the deprived eye. 6. These findings suggest that the thalamocortical projection is physically rearranged as a consequence of monocular deprivation, as has been demonstrated for layer IVc of the monkeys visual cortex (Hubel, Wiesel & Le Vay, 1977).

Inhibition of ocular dominance column formation by infusion of NT-4/5 or BDNF.

During the development of the visual system of higher mammals, axons from the lateral geniculate nucleus (LGN) become segregated into eye-specific patches (the ocular dominance columns) within their target, layer 4 of the primary visual cortex. This occurs as a consequence of activity-dependent synaptic competition between axons representing the two eyes. The possibility that this competition could be mediated through neurotrophin-receptor interactions was tested. Infusion of neurotrophin-4/5 (NT-4/5) or brain-derived neurotrophic factor (BDNF) into cat primary visual cortex inhibited column formation within the immediate vicinity of the infusion site but not elsewhere in the visual cortex. Infusion of nerve growth factor, neurotrophin 3 (NT-3), or vehicle solution did not affect column formation. These observations implicate TrkB, the common receptor for BDNF and NT-4/5, in the segregation of LGN axons into ocular dominance columns in layer 4. Moreover, they suggest that in addition to their better known roles in the prevention of cell death, neurotrophins may also mediate the activity-dependent control of axonal branching during development of the central nervous system.

Sernaphorin III can function as a selective chemorepellent to pattern sensory projections in the spinal cord

Distinct classes of primary sensory neurons in dorsal root ganglia subserve different sensory modalities, terminate in different dorsoventral locations in the spinal cord, and display different neurotrophin response profiles. Large diameter muscle afferents that terminate in the ventral spinal cord are NT-3 responsive, whereas small diameter afferents subserving pain and temperature are NGF responsive and terminate in the dorsal spinal cord. Previous in vitro studies showed that the developing ventral spinal cord secretes a diffusible factor that inhibits the growth of sensory axons. Here we show that this factor repels NGF-responsive axons but has little effect on NT-3-responsive axons. We also provide evidence implicating semaphorin III/collapsin, a diffusible guidance molecule expressed by ventral spinal cord cells, in mediating this effect. These results suggest that semaphorin III functions to pattern sensory projections by selectively repelling axons that normally terminate dorsally.

Transient period of correlated bursting activity during development of the mammalian retina.

The refinement of early connections in the visual pathway requires electrical activity in the retina before the onset of vision. Using a multielectrode array, we have shown that the spontaneous activity of cells in the neonatal ferret retina is correlated by patterns of periodically generated traveling waves. Here, we examine developmental changes in the characteristics of the waves and show that retinal ganglion cells participate in these patterns of activity, which are seen during the same period as synaptic modification in the lateral geniculate nucleus; that the waves subside gradually as the connectivity in the lateral geniculate nucleus stabilizes; and that their spatial structure allows for refinement of the retinotopic map, as well as for eye-specific segregation in the lateral geniculate nucleus.

Requirement for Cholinergic Synaptic Transmission in the Propagation of Spontaneous Retinal Waves

Highly correlated neural activity in the form of spontaneous waves of action potentials is present in the developing retina weeks before vision. Optical imaging revealed that these waves consist of spatially restricted domains of activity that form a mosaic pattern over the entire retinal ganglion cell layer. Whole-cell recordings indicate that wave generation requires synaptic activation of neuronal nicotinic acetylcholine receptors on ganglion cells. The only cholinergic cells in these immature retinas are a uniformly distributed, bistratified population of amacrine cells, as assessed by antibodies to choline acetyltransferase. The results indicate that the major source of synaptic input to retinal ganglion cells is a system of cholinergic amacrine cells, whose activity is required for wave propagation in the developing retina.

Regulation of Class I MHC Gene Expression in the Developing and Mature CNS by Neural Activity

To elucidate molecular mechanisms underlying activity-dependent synaptic remodeling in the developing mammalian visual system, we screened for genes whose expression in the lateral geniculate nucleus (LGN) is regulated by spontaneously generated action potentials present prior to vision. Activity blockade did not alter expression in the LGN of 32 known genes. Differential mRNA display, however, revealed a decrease in mRNAs encoding class I major histocompatibility complex antigens (class I MHC). Postnatally, visually driven activity can regulate class I MHC in the LGN during the final remodeling of retinal ganglion cell axon terminals. Moreover, in the mature hippocampus, class I MHC mRNA levels are increased by kainic acid-induced seizures. Normal expression of class I MHC mRNA is correlated with times and regions of synaptic plasticity, and immunohistochemistry confirms that class I MHC is present in specific subsets of CNS neurons. Finally, beta2-microglobulin, a cosubunit of class I MHC, and CD3zeta, a component of a receptor complex for class I MHC, are also expressed by CNS neurons. These observations indicate that class I MHC molecules, classically thought to mediate cell-cell interactions exclusively in immune function, may play a novel role in neuronal signaling and activity-dependent changes in synaptic connectivity.

PirB is a functional receptor for myelin inhibitors of axonal regeneration.

A major barrier to regenerating axons after injury in the mammalian central nervous system is an unfavorable milieu. Three proteins found in myelin—Nogo, MAG, and OMgp—inhibit axon regeneration in vitro and bind to the glycosylphosphatidylinositol-anchored Nogo receptor (NgR). However, genetic deletion of NgR has only a modest disinhibitory effect, suggesting that other binding receptors for these molecules probably exist. With the use of expression cloning, we have found that paired immunoglobulin-like receptor B (PirB), which has been implicated in nervous system plasticity, is a high-affinity receptor for Nogo, MAG, and OMgp. Interfering with PirB activity, either with antibodies or genetically, partially rescues neurite inhibition by Nogo66, MAG, OMgp, and myelin in cultured neurons. Blocking both PirB and NgR activities leads to near-complete release from myelin inhibition. Our results implicate PirB in mediating regeneration block, identify PirB as a potential target for axon regeneration therapies, and provide an explanation for the similar enhancements of visual system plasticity in PirB and NgR knockout mice.

Unified nomenclature for the semaphorins/collapsins [1]

A new semaphorin is defined as a fully sequenced gene, cDNA, or protein which by sequence homology falls into the semaphorin family of proteins. We have established a Semaphorin Nomenclature web site (http://www.semaphorin-nomenclature.bs.jhmi.edu) that provides details for assigning appropriate names and numbers to new semaphorins, apparent orthologs, gene loci, splice variants, or EST sequences.We recommend that this system of nomenclature be used in all future publications.Semaphorin Nomenclature Committee33441The following scientists have endorsed the use of this nomenclature system: J.A. Bamberg, S. Baumgartner, H. Betz, J. Bolz, A. Chedotal, C.R.L. Christensen, P.M. Comoglio, J.G. Culotti, P. Doherty, H. Drabkin, A. Ensser, M.C. Fishman, B. Fleckenstein, G.J. Freeman, H. Fujisawa, A. Ghosh, D.D. Ginty, C.S. Goodman, S. Guthrie, S. Inagake, R. Keynes, T. Kimura, M. Klagsbrun, A.L. Kolodkin, J.Y. Kuwada, Y. Luo, J.D. Minna, S.L. Naylor, T.P. O’Connor, D.D.M. O’Leary, A. Pini, M.-m. Poo, A.W. Puschel, J.A. Raper, J. Roche, C.J. Shatz, W.D. Snider, E. Soriano, M.K. Spriggs, S.M. Stritmatter, S. Sullivan, L. Tamagnone, M. Tessier-Lavigne, T. Tohyama, J. Verhaagen, F.S. Walsh, and T. Yagi.2Three-letter species designations are based on the first letter of the genus and the first two letters of the species. Additional details about gene nomenclature conventions can be obtained at the Human Gene Nomenclature web site (http://www.gene.ucl.ac.uk/nomenclature/) and the links therein.3The appropriate citation for this letter is: (Semaphorin Nomenclature Committee, 1999).4The committee consists of: please see the list of authors.

Blockade of Endogenous Ligands of TrkB Inhibits Formation of Ocular Dominance Columns

We have examined the hypothesis that the segregation of LGN axon terminals into ocular dominance (OD) patches in layer 4 of the visual cortex requires neurotrophins, acting as signals to modulate the pattern of synaptic connectivity. Neurotrophin receptor antagonists, composed of the extracellular domain of each member of the trk family of neurotrophin receptors fused to a human Fc domain, were infused directly into visual cortex during the peak phase of OD column formation. Infusion of trkB-IgG, which binds BDNF and NT-4/5, inhibited the formation of OD patches within layer 4, while trkA-IgG and trkC-IgG, which preferentially bind NGF and NT-3, respectively, had no effect. The autoradiographic labeling of LGN terminals in cortical layer 4 was reduced by trkB-IgG, in contrast with the increased labeling observed following NT-4/5 infusion. These data suggest that an endogenous ligand of trkB, normally present in limiting amounts within visual cortex, is necessary for the selective growth and remodeling of LGN axons into eye-specific patches.