Activity-Dependent Cortical Target Selection by Thalamic Axons

See allHide authors and affiliations

Science  24 Jul 1998:
Vol. 281, Issue 5376, pp. 559-562
DOI: 10.1126/science.281.5376.559


Connections in the developing nervous system are thought to be formed initially by an activity-independent process of axon pathfinding and target selection and subsequently refined by neural activity. Blockade of sodium action potentials by intracranial infusion of tetrodotoxin in cats during the early period when axons from the lateral geniculate nucleus (LGN) were in the process of selecting visual cortex as their target altered the pattern and precision of this thalamocortical projection. The majority of LGN neurons, rather than projecting to visual cortex, elaborated a significant projection within the subplate of cortical areas normally bypassed. Those axons that did project to their correct target were topographically disorganized. Thus, neural activity is required for initial targeting decisions made by thalamic axons as they traverse the subplate.

During the wiring of connections between the thalamus and cortex in mammals, there is an intermediate step in which thalamic axons grow and interact with a special population of neurons—subplate neurons—before they contact their ultimate target neurons within the cortical plate (1,2). For example, LGN axons en route to visual cortex emit transient side branches that extend into the subplate under both target and nontarget cortical areas (3) and form functional synaptic contacts with subplate neurons (4). During this period of development, spontaneous action potential activity generated in the retina and relayed through the LGN likely drives these subplate synapses in vivo (5). Thus, synaptic relations within the subplate could support activity-dependent interactions during the process of thalamocortical axon target selection.

To examine if activity is needed for thalamic axons to form connections with their appropriate cortical target area, we infused tetrodotoxin (TTX, a sodium channel antagonist that blocks action potentials) or vehicle through osmotic minipumps (6) into the brain of cat fetuses between E42 (E42 = 42 days of gestation) and E56. At E42, the first LGN axons have just reached the subplate underneath visual cortex but still have side branches along their trajectory. Between E42 and E50, the majority of LGN axons have arrived in the visual subplate; by E56, many have departed the subplate and reached their ultimate target, layer 4 of the cortical plate (3). To assess the consequences of the treatments on the thalamocortical projection, we injected carbocyanine dyes at E56 to label retrogradely LGN neurons (7) and subsequently counted the numbers of neurons sending axons to the subplate or cortical plate of either visual (the correct target) or auditory (an incorrect target) cortex.

The number of LGN neurons projecting to visual cortex was decreased in TTX-infused animals (Fig. 1), both within the subplate [Fig. 1C; an average of 69 ± 5% SEM fewer neurons than vehicle controls,n = 8 animals; 4 littermate pairs treated with TTX or vehicle and matched for similar 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) injection sizes] and within the cortical plate (Fig. 1C; 94 ± 0.5% SEM, n = 8 animals; 4 littermate pairs treated with TTX or vehicle and matched for similar DiI injection sizes) (8). However, the overall development of the LGN itself was relatively unaffected (compare Fig. 1, A and B). Previous studies have shown that the dendritic and somatic development of LGN neurons also proceeds essentially normally during the entire treatment period (9), although we noted a modest decrease in the size of the LGN [size of TTX-treated LGN is 76 ± 5% of that of vehicle-treated LGN, n = 3 littermate pairs (6 animals)].

Figure 1

The number of LGN neurons projecting to visual cortex is markedly decreased after intraventricular infusion of TTX between E42 and E56 (6). For assessment of the consequences of the treatment, DiI was placed into the cortical plate or subplate of visual or auditory areas at E56 to label neurons retrogradely within the LGN, and the number of labeled LGN neurons was counted (7). All data are from littermate pairs treated with either TTX or vehicle and also matched for similarly sized dye injections in cortex or subplate. LGN neurons from a vehicle-treated (A) or TTX-treated (B) littermate pair were retrogradely labeled after injection of DiI in the visual cortex (green cells in these fluorescence photomicrographs). Despite similarly sized and located DiI injections, the TTX-treated animal has far fewer LGN neurons projecting to the visual cortex than the vehicle-infused littermate. Scale bar (A and B), 500 μm. The number of LGN neurons labeled after dye placements in either the cortical plate (CP, filled dots) or subplate (SP, open dots) of visual cortex (C) or auditory cortex (D) is plotted on a logarithmic scale. Individual littermate pairs treated with either TTX or vehicle and matched for dye injection size are pictured as pairs of connected dots. The range of counts reflects the sizes of the dye injections [for example, in (C), the largest matched injections labeled 4345 LGN neurons in a vehicle-treated fetus but only 1153 in a TTX-treated littermate]; comparisons should be made between littermate pairs [n = 8 littermate pairs for (C) and 4 littermate pairs for (D)]. Note the decrease in numbers of LGN neurons projecting to visual cortex and visual subplate (C) and the increase in numbers projecting to auditory subplate (D).

The intracortical pathway normally taken by LGN axons en route to visual cortex bypasses many cortical areas, including auditory cortex. Thus, we examined whether, after TTX treatment, LGN neurons send axons to the subplate or cortical plate of auditory cortex rather than to visual cortex (Fig. 1D). In vehicle-treated cases at E56, relatively few LGN neurons could be retrogradely labeled after 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine perchlorate (DiD) injections into the subplate or cortical plate of auditory cortex. The largest dye injection (covering 9.6 mm3 of the cortical plate and subplate of A1) labeled only 354 LGN neurons; a similarly sized dye placement in V1 labeled 4345 LGN neurons. After TTX treatment, fewer LGN axons could be retrogradely labeled by an injection of DiD into auditory cortex proper (Fig. 1D; 75 ± 6% SEM, n = 4 animals; 2 littermate pairs treated with TTX or vehicle and matched for similar DiD injection sizes). However, in TTX-treated animals, there was a substantial increase in the number of LGN neurons retrogradely labeled after tracer injections into the subplate underlying auditory cortex (Fig. 1D; 437 ± 189% SEM; n = 4 animals; 2 littermate pairs treated with TTX or vehicle and matched for similar DiD injection sizes) (10). Thus, blockade of action potential activity apparently alters the spatial patterning of the geniculocortical projection by derailing many LGN axons into the subplate underlying auditory cortex.

To visualize directly the aberrant routing of LGN axons within the subplate, we injected DiI into the LGN to label the axons anterogradely. In vehicle-treated animals (Fig. 2A; n = 2 animals), as in unmanipulated littermates at E56 (n = 2) (3,11), LGN axons were tightly fasciculated in the optic radiations but branched extensively within visual subplate and visual cortex; few, if any, side branches were given off as LGN axons passed by other cortical areas. In contrast, axons in TTX-treated littermates (n = 2) extended many branches in the subplate of nonvisual areas along their entire route to visual cortex (Fig. 2B, arrowheads), and few axons actually made it as far as V1. This experiment confirms the conclusion inferred from the retrograde labeling experiments that TTX treatment alters the spatial patterning of the projection from the LGN to cortex.

Figure 2

After activity blockade, LGN axon growth and arborization in the visual cortex are reduced, and branching en route within the subplate is increased. (A) Anterograde labeling of axons from injection of DiI in LGN reveals that, in vehicle-treated animals, axons navigate normally to primary visual cortex (V1) and selectively branch within visual subplate and cortical plate (arrow). As LGN axons traverse nonvisual areas, they remain fasciculated in the optic radiations. (B) In TTX-treated littermates, LGN axons growing within the optic radiations extend many branches into the subplate of nonvisual areas (arrowheads). Their growth into visual cortex is diminished relative to vehicle-treated controls. Visual cortex (V1) is located left of the vertical bars at the top of each panel. Laminar divisions of neocortex are indicated at the bottom of each panel (OR, optic radiations; SP, subplate; CP, cortical plate).

Two additional aspects of the geniculocortical projection were also perturbed after TTX treatment. Many LGN axons within the cortical plate of V1 fail to arborize within cortical layer 4 and instead send unbranched projections up into the pial surface (12). Here we report that the topography of the projection to visual cortex was also degraded after TTX treatment (Fig. 3). This degraded topography is evident when comparing the spatial distribution of retrogradely labeled neurons after similarly located and sized DiI injections into visual cortex (Fig. 3A; n = 8 TTX-vehicle littermate pairs): Neurons were more dispersed across the LGN in the TTX-treated cases than in the vehicle-treated cases (13). The portion of total LGN covered by labeled neurons was consistently larger in the TTX-infused brains than in those of vehicle-treated controls (Fig. 3B) (average 52 ± 8% SEM in TTX brains versus 28 ± 2% in control brains;n = 4 TTX-vehicle littermate pairs; P< 0.01, Student's t test). There was also more variability in the location of the peak number of retrogradely labeled neurons within the LGN in TTX-treated animals (14). These results suggest that the topographic map of those LGN neurons that did manage to project to visual cortex is significantly degraded; regions of the LGN that would normally not project to a given cortical location at E56 did so in TTX-treated animals. Thus, thalamocortical connections, like retinotectal connections (15), must use both activity-independent and activity-dependent mechanisms in the refinement of topographic projections within their target area.

Figure 3

Precision of topography in the LGN projection to visual cortex is degraded after activity blockade. (A) The location of LGN neurons retrogradely labeled from comparably sized and located DiI injections into visual cortex is plotted within an outline of the LGN of vehicle-treated (1 to 3) or TTX-treated (4 to 6) animals in comparable sections. The altered distribution of labeled neurons within the LGN of TTX-treated animals is apparent (see also Fig. 1, A and B). Orientation of sections marked by arrows; M, medial; L, lateral; D, dorsal; V, ventral. (B) Neurons projecting to a given location in V1 are distributed over a small percentage of the total width of LGN in vehicle-treated animals (28 ± 2% SEM; open circles); however, in TTX-treated animals (closed circles), neurons covering a much larger percentage of the total LGN width (52 ± 8% SEM; P < 0.01) project to a comparable location in V1. Sections shown in (A) were some of those used to calculate values for the littermate pairs “A” ofFig. 3B. See (13).

The results of this study argue for a requirement for action potential activity in the decision-making that occurs as growing thalamic axons traverse the subplate en route to their appropriate cortical target areas. How might activity operate in the targeting of LGN axons? We think it unlikely that TTX acts simply to “freeze” LGN axons into the projection pattern present at the time of onset of the TTX infusion. At E56, even after TTX treatments, the number of LGN neurons projecting to visual cortex increased threefold over the projection at E42 (n =3 untreated animals at E42 compared withn = 3 TTX-treated animals at E56) (16). In addition, at E42 the side branches present along the route of LGN axons were few and short (3), whereas in the TTX-treated cases at E56 branching was extensive. There was also extensive growth of the dendrites and somata of LGN neurons after similar TTX treatment (9). These considerations indicate that LGN axons are not “frozen” in the immature state present at E42.

The results of our experiments, on the contrary, argue that activity blockade alters the pattern of axonal branching, and therefore the targeting decisions, of LGN axons rather than growth per se. The decrease in the number of LGN neurons projecting to visual cortex and subplate and the accompanying increase in the number projecting to nonvisual subplate, including that of auditory cortex, can be explained if side branches that would normally be eliminated between E42 and E56 instead grow when action potentials are blocked. This suggestion is consistent with the known effects of similar treatments on the growth and branching of axons in many other developing systems, including the mammalian retinogeniculate and retinocollicular pathways and the retinotectal system of fish and frogs (17, 18).

Few LGN axons managed to grow all the way back to the visual cortex when activity was blocked between E42 and E56. Indeed, the subplate along the entire intracortical trajectory of LGN axons may be available as a valid intermediate target for LGN axons in the absence of action potential activity. It may be that once LGN axons arborize within the subplate underlying inappropriate cortical regions, they are no longer able to respond to cues along the pathway that would promote continued growth toward visual cortex (19). Despite the targeting errors, the directionality of LGN axon growth—toward the posterior pole of the cortex—is not perturbed by TTX treatment, consistent with the hypothesis that positional information guiding axon pathfinding can be read by thalamocortical axons in an activity-independent manner. Cues may include the Eph family of receptors and ligands, which are known to subserve such functions in the developing retinotectal system of vertebrates (20).

The subplate appears to be a crucial decision-making compartment in which dynamic interactions between growing thalamic axons and subplate cells ultimately result in the selection of the correct cortical target area (21). Our results argue for a requirement for activity-dependent interactions in the initial steps of cortical target selection by thalamic axons. It is unclear at present whether these interactions enable growing axons to select directly the appropriate target (instructive) or simply to read specific target-derived molecular cues (permissive). Whatever the mechanism, the formation of connections between thalamus and neocortex in mammals may be a special exception to the general rule that target selection by developing axons is independent of neural activity.

  • * Present address: Division of Biology, 216-76, California Institute of Technology, Pasadena, CA 91125, USA. E-mail: scatalan{at}

  • To whom correspondence should be addressed. E-mail: cshatz{at}


Stay Connected to Science

Navigate This Article