How Do Thalamic Axons Find Their Way to the Cortex?

Z. Molnkr and C. Blakemore - Development of thalamocortical innervation CORTEX How do thalamic axons find their way to the cortex?

Zolthn Molnhr and Colin Blakemore

A cascade of simple mechanisms influences thalamic innervation of the neocortex. The cortex exerts a remote growth-promoting influence on thalamic axons when they start to grow out, becomes growth-permissive when the axons begin to invade, and later expresses a ‘stop signal’, causing termination in layer 4. However, any part of the thalamus will innervate any region of developing cortex in culture, and the precise topographic distribution of thalamic fibres in vivo is unlikely to depend exclusively on regional chemoaffinity. The ‘handshake hypothesis’ proposes that axons from the thalamus and from early-born cortical preplate cells meet and intermingle in the basal telencephalon, whereafter thalamic axons grow over the scaffold of preplate axons, and become ‘captured’ for a waiting period in the subplate layer below the corresponding part of the cortex. The bizarre pattern of development of thalamic innervation in the mutant reeler mouse provides strong evidence that thalamic axons are guided by preplate axons.

Trends Neurosci. (1995) 18, 389-397

NY DEVELOPMENTAL NEUROSCIENTIST with a have small cerebral hemispheres in which adjacent A tendency to masochism is encouraged to contem- neocortical fields generally receive their projections plate the development of thalamocortical projections. from neighbouring thalamic nuclei2,3. Table 1 shows In general, each thalamic relay nucleus projects pre- the chronicle of thalamocortical development for rat cisely and topographically to a particular cortical terri- and mouse. As in all eutherian mammals, the major tory. However, the routes taken by individual fibres in events take place in utero, and are initiated before the the adult brain can be tortuous, with sharp bends and thalamus is innervated by the sensory pathways. even rotations of the array around one axis but not How the cortex is built the orthogonal. Moreover, single thalamic nuclei can project to more than one cortical area; each cortical The cells of the cerebral cortex originate in the pro- field can have several thalamic inputs; and the topo- liferative neuroepithelium that lines the forebrain graphic order of projections from a single nucleus can vesicle. From this mitotic factory, immature neurones change abruptly at the borders of neighbouring corti- migrate along the processes of radial glia towards the cal areas (such as the reversal of the retinotopic map pial surface to generate the layers of the cortex8. The between primary and secondary visual areas). first post-mitotic neurones form the primordial plexi- How could this bewildering complexity and preci- form zone or preplate, which is later split by the sion be generated by the simple mechanisms that are invasion of true cortical neurones that are generated familiar to developmental neuroscience, such as subsequentlyg-‘l. The latter are distributed in an trophism, selective adhesion and competitive interac- inside-out sequence to create layers 6-2 of the cortex, tion? This question assumes special significance which are sandwiched between the superficial and deep because of the growing evidence that the input from components of the original preplate [the marginal each major thalamic nucleus can act as a local ‘extrin- zone (cortical layer 1) and the subplate, respectively sic signal’, influencing the fate of the cortex it inner- (see below)]. vates by setting the boundaries of the cortical field, Neurones of the preplate mature precociously, send- delivering its afferent input, and influencing its ing out axons, and expressing a variety of transmitter regional differentiation’. and receptor systems, long before the cells of the corti- In considering how neural systems form, it is cal plate mature. However, a substantial proportion of important not to infer complexity during develop- these neurones die after the arrival of thalamic axons, ment from the appearance in the mature animal. and the completion of cortical neuronogenesisl’. Many of the perplexing features of thalamocortical These characteristics provide strong circumstantial organization (especially the convoluted paths taken evidence that preplate cells play a role in the devel- by individual axons, the rotations of axon arrays and opment of the cortex. Of particular interest is the Zoltrin Molntir and irregular topology between thalamus and cortex) mutant mouse reeler. In this autosomal-recessive Colin Blakemore are might be produced after the initial establishment of mutant, whose aberrant gene product has been cloned at the Laboratory of connections. The ultimate imbroglio of thalamocorti- very recently 12,the cortex forms in a roughly outside- Physiology, cal organization might be generated by a cascade of in sequence, leaving the entire preplate stranded as a University of individually simple mechanisms. ‘super-plate’ above the inverted layers of the cortical plate; Oxford, Parks Road, In an attempt to simplify the problem, a number of however, the basic topographic interrelations between Oxford, studies have been performed on rodents, since they thalamus and cortex develop essentially normally13. UK OXI 3PT.

0 1995, Elsevier Science Ltd TINS Vol. 18, No. 9, 1995 389 CORTEX Z. Molnir and C. Blakemore - Development of thalamocortical innervation

TABLE I. Chronology of thalamocortical development in rat and mouse ing period of a few days before invasion of the cortex. Although the insensitivity of their methods led them Age Major events (for occipital cortex and LGN) to overestimate the age by a couple of days at each Rat Mouse stage, this general sequence has been confirmed in recent studies using the acutely sensitive carbocyanine El2 El1 Start of cortical neuronogenesis Early El4 Early E I3 First preplate cells complete their migration dyes’4,‘5 to trace axons in fixed fetal brainslh-” (see El4 El3 Commencement of axon outgrowth from preplate and Table 1). Catalan0 and colleague? also reported that thalamus fibres from the ventral thalamus arrive under the El5 El4 Arrival of first cells of the cortical plate somatosensory cortex early, and enter the cortical Early E I 5 Early E I4 Preplate and thalamic axons meet in basal telencephalon plate before birth, but they demonstrated axons that El6 El5 Thalamic axons arrive under cortex penetrate the lower layers of the cortex immediately El8-I9 El7-I8 Major invasion of the cortical plate begins after their arrival, and questioned the existence of a E211PO E2llPO Birth waiting period in rodents. However, the consensus PI PO Arrival of thalamic axons in layer 4 seems to be that many thalamic fibres accumulate under the true cortical plate but for a shorter time Approximate ages in rat and mouse (including the reeler mutant) are given for the main events in the development of the occipital cortex, and projections from the lateral geniculate nucleus than the prolonged waiting periods that are observed (LGN), the visual relay in the dorsal thalamus. The ‘plug date’ is taken as embryonic day zero (EO) in carnivores” and primatesz3. for the dating of fetuses, and the first postnatal day as PO. Cortical-cell generation finishes before The sequence and topography of initial outgrowth birth but migration into the upper layers continues until the end of the first postnatal week’? In from the dorsal thalamus to the occipital cortex is the tangential direction, across the developing cerebral cortex, there is another distinct neuro- illustrated in Figs 1 and 2. Figure 1 shows two bundles genetic gradient, the anteroventral areas of the hemisphere preceding the caudodorsal in maturity of axons that have been labelled by placement of by more than a day’,‘. small crystals of different dyes in neighbouring regions of the dorsal thalamus of an embryonic day 14.5 (E14.5) rat fetus. Even at this early stage, when The timing and pattern of thalamic-axon outgrowth the major nuclei of the thalamus are just beginning to be distinguishable, and the cortex consists of preplate In an early study, using degeneration techniques, alone, thalamic axons have grown down through the Lund and Mustads described the outgrowth of fibres diencephalon in reasonable topographic order, and from the fetal rat thalamus, and their accumulation out through the primitive internal capsule under the under the appropriate region of the cortex for a wait- anlage of the corpus striatum, and are coursing up into the intermediate zone of the telencephalon. Figure 2 illustrates the subsequent events for both normal and reeler mice. In the normal animal, thalamic axons approach the cortex but then gather in the subplate layer. Very few thalamic axons penetrate the cortical plate until a couple of days before birth, when they suddenly turn and invade radially en masse, branching and terminating mainly in layer 4. The pattern seems very different in the reeler mutant, in which the neurones that are equivalent to the normal subplate remain above the gathering cortical plate. Thalamic fibres stream obliquely upwards in fascicles through the cortical plate, and accumulate for a few days in the superplate layer above. They then plunge down and terminate, presumably on cells that are equivalent to the normal layer 4 (Refs 26 and 27). Could a single developmental programme explain both Fig. 1. Topographic outgrowth of axons from the thalamus. Two small crystals of different carbocyanine dyes were patterns of innervation? inserted 200 pm oport, I, I ‘-diocfadecyl-3,3,3 ‘,3’-tetromethylindocorbocyonine perch/orate (dil; orange) lateral to 4-[4-(dihexadecylamino)styrylJN-methylpyridinium iodide (diA; yellow/green), in the leff dorsal thalomus of an embry- A cascade of signals from the onic day 14.5 fixed fetal rat brain that was then incubated at room temperature for three weeks. Each dye lobe/led a cortex is revealed in vitro discrete bundle of thalamic axons (filed orrow, dil; open arrow, diA), that remain separate as they pass down through Organotypic co-culturez8 pro- the ventral diencephalon (A), out through the internal capsule and under the developing corpus striatum (B), where vides a simplified system in which to growth cones are visible at the advancing front of each bundle at high mognificotion. At this stage, descending axons from the cortical prep/ate are also only just approaching the basal telencephalon and, therefore, could not have guided study the expression of signals that tholomic axons this far. However, there are other axon systems present in the diencephalon and the internal capsule might play a part in the initiation, (from Ihe thalamic reticular and perireticular nuclei2’, the medial gang/ionic eminence” and the ventral cortex), that guidance and termination of thal- might play a port in the ear/y phase of guidance of thalamic fibres. Scale bar, 500 pm. amic innervation of the developing

390 TINS Vol. 18, No. 9, 1995 Z. Molnk and C. Blakemore - Development of tlnlamocortical innervation CORTEX cortex’7,2v”4. In our own studies, Normal mouse reeler thalamic explants from rat em- bryos (from E14.5 to E21, but mostly E16) were co-cultured in serum-free medium3’ with slices of neocortex from other animals aged between E14.5 and postnatal day 11 (I’ll). These ages were chosen E14.5 to span the natural period of thalamic-fibre outgrowth, arrival, accumulation and ingrowth. For most experiments, thalamic blocks were taken from the region of the putative lateral geniculate nucleus (LGN), and cortical slices were taken from the occipital region, which is assumed to be the pre- El8 cursor of the primary visual cortex. The thalamic explants were pre- incubated with l,l’-dioctadecyl- 3,3,3’3’-tetramethylindocarbocyanine perchlorate (di1) to label outgrowing axons31. Three patterns of axon growth occurred, depend- ing on the age at which the cortical slice was taken. The cortex exerts a remote growth- promoting influence from the earliest P3 stage When isolated thalamic explants are cultured in serum-enriched medium, there is considerable (and variable) outgrowth of neurites. This is not obviously enhanced Fig. 2. Thalamic axons arrive, wait and invade in normal and reeler mice. These fluorescence micrographs of the by the presence of a nearby slice lateral part of the occipital cortex (loterol is up; ventral to the /eft) show the sequence of events as thalamic axons of cortex, which led Bolz and approach, pause and enter the cortical p/ate in the normal mouse (left) and in the mutant reeler mouse (right), in colleagues3’ to conclude that the which the cells of the prep/ate remain OS the ‘superplote”3 (s’p) above the orriving cells of the cortical plate (cp), rather cortex does not produce diffusible than being split into morginol zone (mz) and subplate (sp). Jhalamic axons have been lobe/led with a crystal of 7,7 ‘- dioctodecyl-3,3,3 ‘,3 ‘-tetramethylindocorbocyanine perch/orate (di/) p/aced in the dorsal thalamus. (A and 6) At substances that affect extension of embryonic day 74.5 (E74.5), they ore approaching the subplate in the normal mouse but are growing in oblique fosci- thalamic axons. However, serum c/es through the cortical plate in the reeler mouse. (C and D) At E78 (three days before birth), they ore still concen- contains a variety of factors that trated dense/y in the subplate layer in the normal mouse but have paused in the super-plate of the reeler mouse. Before can stimulate growth of neurites, birth in both phenotypes, they start to leave the waiting comportment ond enter the cortical plate, turning up in the making it difficult to interpret any normal mouse and down in the reeler mouse ‘@‘. (E and F) By postnatal day three (P3), most thalamic axons have additional effect from an added terminated in layer 4 or its equivalent in the reeler mouse. Scale bars, 200~m for A, 6, E and F; 700 pm for C and D. cortical slice. In serum-free medium, there is very little extension of neurites from a thal- enhancing growth towards the primitive internal amic block cultured alone, but if a cortical slice is capsule. placed within about 2 mm when cultured on a collagen- The cortex becomes growth permissive at the end of the coated membrane (or 4 mm on laminin), axons stream natural waiting period out of the thalamic explant on all sides, and a mass of When an El%16 rat thalamic explant is cultured fibres fills the gap between the explants’7,34. This with a cortical slice taken before about E20, thalamic target-dependent outgrowth of axons is seen even for axons grow out across the floor of the culture cham- El5 rat cortical slices, suggesting that the cortex ber at about 1 mm per day. When the axons reach the produces a diffusible substance that has a growth- slice, most of them encircle it, and some run over its promoting influence on thalamic axons, from the surface, but few invade it (Fig. 3), suggesting that beginning of formation of the cortical plate until after thalamic fibres do not find early embryonic cortex an the age at which innervation is complete. attractive environment for growth32,37,38. There is no evidence for an explicit effect on the At about E19-20 in the rat, the tissue of the cortical direction of growth of individual axons that would slice becomes growth permissive, and it is penetrated be indicative of a true tropic influence36. However, if readily by thalamic axons17,31-33,37,3*.With a cortical the culture chambers were left undisturbed, neurites slice taken at E20-P3, flbres grow profusely out of an usually extended further from the explant, and more El6 thalamic block, through the ventricular surface of densely on the side facing the cortical slice”, pre- the slice and onward, in a radial direction, across the sumably because of the establishment of a concen- intermediate zone and subplate, and into the cortical tration gradient. Thus, in viva, the diffusible substance plate itself. Growing at approximately 1 mm per day, might not only initiate outgrowth of axons from the they pass through the superficial layer of newly thalamus but also influence its direction, perhaps arrived immature neurones (the dense cortical plate),

TINS Vol. 18, No. 9, 1995 391 El6 CTX PO CTX P6 CTX

El6 LGN El6 LGN El6 LGN

Fig. 3. Co-culture reveals a sequence of three molecularsignals that is expressed by the cortex. Small explants from the dorsal diencephalon [the region of the lateral geniculate nucleus (LCN)] of embryonic day 76 (E76) fetal rats and s/ices of occipital neocortex, 350-400pm thick, were cultured in a serum-free (NZ) medium35. Such explants survive we// and remain about 750-200pm thick, even after many weeks in culture. The LGN block was pre-incubated with an alcoholic solution of 7,7 ‘- dioctadecyl-3,3,3 ‘,3’-tetramethylindocarbocyanine perch/orate (di/) to label outgrowing axons. Representative floorescence micrographs are superimposed on schematic drawings showing the behaviour of thalamic axons after co-culture for four days with cortical s/ices taken at E76, postnatal day zero (PO) and P6. The plane of focus was adjusted to be near the middle of the thickness of each section. With El 6 cortex (CTX), neurites extend from the thalamic block but they encircle the cortical explant or run over its surface, and do not penetrate it. With a slice of PO cortex, the axons run in through the ventricular surface and grow radio//y through the intermediate zone, subplote and cortical plate. They continue up to the pial surface, and turn to grow tangentially in the cortical plate or penetrate the surface of the s/ice and grow around its margin. In the P6 s/ice, most of the tholomic axons slow down, branch, lose their growth cones, and terminate -300 pm below the piol surface, in what is presum- ah/y /uyer 4. Such experiments have been interpreted to mean that the cortex expresses a diffusible, growth-promoting signal from about E IS but does not become permissive to axon ingrowth until about El 9-20 (Refs 32 ond 37), and that layer 4 expresses a ‘stop signal’ ofter about PZ-3 (Ref. 3 I). Arrows indicate growth cones at the tips of axons. Scale bar, 7 00 pm for El 6, and 250 pm for PO and P6.

and reach the marginal zone after about three to four Thalamic innervation in vitro is not obviously days in culture, without terminating or even hesitat- regionally selective ing en route (Fig. 3). Some axons then turn and run In principle, chemospecific variation4’ between tangentially through the cortex but many penetrate regions in the cortex might determine the innervation the surface, and run along it, directly below the pia of cortical areas by particular thalamic nuclei. itself, forming a fasciculated band of fibres that is However, it is difficult to imagine how regional differ- similar to the circling axons that are observed at earlier ences in a diffusible growth-promoting substance stages. could establish such ordered patterns of thalamic out- layer 4 expresses a ‘stop signal’ afier about 1’2-3 growth as are observed in vivo (see Fig. 1). Further- In rat cortical explants taken after about P2, thal- more, the growth-permissive property is not expressed amic axons grow less quickly. After only four days in until about E20, three or four days after thalamic culture, most of them have branched locally, lost their axons have distributed themselves in a fairly precise growth cones, and terminated approximately 300 pm pattern under the cortical plate (Fig. 2). below the pial surface, at a depth corresponding to Despite these reasons for scepticism, the possibility layer 4 (Refs 17,29,31,33 and 39) (see Fig. 3). A few of regional chemoaffinity was tested by culturing axons extend up to the marginal zone, as observed thalamic blocks in a choice paradigm in which a cen- in vivo. tral thalamic explant was flanked by two slices, one It has been suggested that around two to three days from the appropriate target area, and the other from after birth in the rat, layer 4 starts to express a ‘stop an inappropriate region31. These procedures were signal’ that causes thalamic fibres to branch and ter- capable of revealing target preferences: when con- minate in vivo31. The failure of most thalamic axons to fronted with a slice of El&P8 cerebellum, small num- grow beyond layer 4 in cortex that is older than P2 is bers of thalamic axons encircled the slice but very few not due merely to the later-forming upper layers of the penetrated it. However, when flanked with two differ- cortex lacking growth-permissive properties, since ent explants of neocortex, one from the correct target, axons grow readily down through the upper layers, and the other from a region that the chosen region of and terminate in the presumptive layer 4 if the thal- thalamus would never approach in tivo, there were no amic block is placed against the pial surface of the obvious differences in the patterns of innervation31 cortical slice31*32. (Fig. 4). This experiment does not exclude entirely the

392 TINS Vol. 18, No. 9, 1995 Z. MolnOr and C. Blakemore - Development of thalamocortical innervation CORTEX possibility of a subtle gradient of signals across the A P6 OCC cortex, nor the expression of regionally specific factors at the tips of early corticofugal axons (see below), but it does suggest that thalamic axons do not identify their particular region of the cortical plate on the basis of chemoaffinity. Although the cascade of cortical properties that are revealed in vitro might play a part in determining the timing of outgrowth, waiting, innervation and termination of thalamic axons in viva, it seems that other explanations must be sought for the topographic pattern of their guidance. Do early corticofugal projections from the preplate guide thalamocortical fibres? Shatz and ctilleagues41-43reported that neurones of the subplate (those first-generated preplate cells that remain below the thickening cortical plate) send pio- neering axons to the diencephalon in cat and ferret, and suggested that these fibres might play a part in the establishment of ascending and other descending projections. If descending fibres are entirely responsible for guiding thalamic axons to their correct cortical target regions, it should be possible to show that axons from Fig. 4. Thalamic axons prefer cortex to cerebellum but show no obvi- the preplate of each cortical region pioneer a route to ous regional preference within the cortex. Results of the ‘choice’para- the corresponding thalamic nucleus, that they form digm in which an explant of embryonic day 16 rat lateral geniculate their guidance projections before axons leave the nucleus (El6 LGN) was cultured next to two different s/ices”3’. Camera thalamus, and that thalamic axons grow over the pre- lucida drawings show 7,7 ‘-dioctadecyl-3,3,3 ‘,3’-tetramethylindocarbo- plate scaffold to their targets. cyanine perch/orate-lobe/led thalamic axons after five days in vitro. (A) The subplate scafold is organized topographically Given the choice between postnatal day 6 occipital cortex (P6 OCC), The outgrowth and topography of the early axon which is the appropriate target, and a slice of P6 cerebellum (CER), projection from preplate cells can be revealed by plac- which (he thalamus does not innervate in vivo, axon invasion was ing small crystals of carbocyanine dye into the cortex restricted largely to the cortical s/ice, in which most fibres terminated in of perfused fetal rats and mice of known gestational what appears to be layer 4, -300-400 pm below the pial surface. Of the few axons that sprouted toward the cerebellor slice, most ran age and, after sufficient incubation to enable diffusion around the border of the s/ice or in fascicles across its surface. of the dye along axons, and then examining sections (B) Confronted with P6 occipital cortex and frontal cortex (FRO), axons by fluorescence and confocal microscopy16-20,26,44~45. from the LGN explant showed no obvious preference, despite the fact Immediately after the first post-mitotic neurones that they do not innervate frontal cortex in vivo. In 398 co-cultures, no have migrated to the outer edge of the cerebral wall, consistent differences were observed in the pattern or density of inner- axons can be observed leaving these preplate cells. vation, whatever region of thalamus and neocortex were combined. From each region of cortex, the axons grow through Scale bar, 1 mm. the intermediate zone towards the basal telencephalon (Fig. 5). After -24 h (for fibres from occipital cortex), they curve medially beneath and through the fibres is not fulfilled. Outgrowth of axons from each developing corpus striatum towards the primitive thalamic nucleus is approximately synchronized with internal capsule. Interestingly, the appearance of the that from the preplate of the corresponding cortical corticofugal projection is initially indistinguishable in area, rather than being delayed until the completion normal rodents and in reeler mice, because the cells of of the preplate scaffold (Table 1). Implantations of dye the true cortical plate (whose bizarre aberrant distri- crystals in the cortex only two days after the initial bution characterizes reeler mice), are not yet present’3,46. outgrowth of preplate axons (El6 for rat occipital cor- Labelling with parasagittal or coronal rows of dye tex) leads to labelling of cell bodies in the matched placements of different colour shows that each region thalamic nucleus, presumably because of retrograde sends a quite tightly organized bundle of axons diffusion along thalamic axons that have already towards the internal capsule, with little or no inter- reached the cortex at that age16-‘0. mixing of axons from neighbouring cortical areas Fibres from different thalamic nuclei project even where the array narrows as it approaches the towards the internal capsule as an ordered array internal capsule16 (Fig. 6B). At this early stage, each (Fig. 1). Even fibres that originate in an individual labelled bundle has uncomplicated geometry, simply nucleus maintain reasonable topography, some appar- curving in an arc without twists or sharp bends and ently making corrective detours, as they stream taking the shortest route through the intermediate towards and through the internal capsule, and into zone16-20. Thus, axons from the preplate do indeed the telencephalon47, without any possibility of seem to establish a reasonably well-ordered topographic guidance from preplate axons. Recent evidence sug- scaffold. gests that, during this initial part of their growth, Outgrowths ofpreplate and thalamic axons are synchronized, thalamic fibres might be associated with other transient not sequential axon systems 24,25.However, the preplate array could Unfortunately, the second expectation for the play a role during the distal part of the trajectory of simple theory of preplate-axon guidance of thalamic thalamic axons.

TINS Vol. 18, No. 9, 1995 393 CORTEX Z. Molnt and C. Blakemore - Development of thalamocortical innervation

normally meet (the basal telencephalon or intermediate zone) if, subsequently, thalamic axons are to stop and form temporary synapses within the subplate layer48. This handshake hypothesis can be tested simply by labelling the two axon systems with different dyes, and observing whether they intermingle. Using this test, Miller and colleagues4’ concluded that the two fibre systems grow in adjacent compartments. However, interpretation is not straightforward. If the pathways are labelled too early, they will not have met; if too late, the corticofugal projection will include the axons of true cortical cells of layers 6 and even 5, which cer- tainly follow different paths47,4g.More importantly, it is essential to label precisely matched cortical and thalamic loci. Since the spatial relationships of thalamic nuclei change somewhat during development, it is not easy to know that regions have been labelled that will correspond in the adult. After implanting different carbocyanine dyes into the dorsal thalamus and the occipital cortex in El.515.5 fetal rats, we too have observed labelled bundles running separate and parallel to each other. However, this can probably be attributed to misalignment of the two dye placements because, in many such experiments, intermingling has been observed’6,‘7,44. Although Bicknese and col- leagues4y suggested that thalamic axons tend to lie superficial to preplate axons as the former approach the cortex, Bicknese and Pearlmans have also demonstrated interdigitation of descending and ascending axons in the internal capsule. Figure 6A shows an example of thalamic and subplate axons, running in opposite directions through the intermediate zone but contributing to the same fascicles. The theory of co-fasciculation is also supported by the fact that crystal implantation at any point in the cortex after the arrival of thalamic axons, labelling both descending and ascending projections, stains discrete, tight bundles of fibres rather than separate groups of afferent and efferent axons17 (Fig. 6B). Evidence from the reeler mouse Fig. 5. Outgrowth ofoxons from prep/ate cells. A sing/e crystal of I, I ‘-dioctodecyl-3,3,3 ‘,3’- tetramethylindocarbocyanine perchlorote (di!, was inserted in the dorsoloteral part of the The proposal that thalamic fibres complete their occipital cortex in the left hemisphere of a fixed embvonic day 13.5 mouse embryo and, after journey to the cortex by growing over preplate axons, incubation for four weeks at room temperature, 75pm coronal sections were examined in a and that this intimate association is the essential prel- confocal microscope. The plane of this montage of confocal micrographs (2-pm optical sec- ude to their accumulation within the waiting com- tioning) lies close to the implantation site. Spindle-shaped cells (presumably neuroepithelial partment, is supported strongly by the peculiar pat- cc//s) in the ventricular and subventricular zone (VZ and SVZ), and radial glia, running from tern of innervation in the reeler mouse (Fig. 2B, D and the ventricular to the pial surface, are labeled in the region of the implantation site (large F). The earliest stages of formation of the preplate, the arrow). The first post-mitotic cells have migrated into position to form the prep/ate or primordial plexiform zone (PPZ), as Marin-PadHa original/y co/led it. Axons from preplote cells outgrowth of a topographically ordered array of have already grown down in the intermediate zone, and are turning under the developing descending axons from the preplate cells (Fig. 5), the corpus striatum (ST) with particularly elaborate growth cones at their tips (small arrows). At outgrowth of thalamic axons through the primitive this stage, before the orrival of cortical-plate cells, the appearance is indistinguishable in internal capsule and their tangential distribution in normal and reeler mice. This example is from a reeler mouse. Scale bar, 7 00 pm. association with the scaffold of corticofugal fibres towards the appropriate cortical regions all occur in Tholamic axons grow over the preplate axon scafold: the the reeler mouse, just as in the normal mousez6. The ‘handshake hypothesis’ two genotypes begin to diverge in appearance only We have suggested that thalamic axons meet sub- after cells of the cortical plate itself start to arrive plate axons, and intermingle with them en route to the (about El5 for the occipital cortex). In the reeler subplate layer17p31.Indeed, this ‘handshake’ between mouse, cortical-plate cells gather below the ‘super- the two arrays might be needed not only for the final plate’ cells, whose pioneer axons then come to lie in phase of topographic guidance of thalamic axons but oblique fascicles that are surrounded by the thicken- also for the initiation of the waiting period. The fact ing cortical plate. that thalamic axons in co-culture show no tendency The rat cortex does not become growth permissive to accumulate within the subplate layer suggests that until some days after thalamic axons arrive under the the two axon systems need to establish some sort of cortical plate in v~vo~‘,~~(see Fig. 3). However, when intimate association in the environment where they thalamic axons in the reeler mouse reach the cortex,

394 TINS Vol. 18, No. 9, 1995 Z. Molnb and C. Blakemore - Development of thalamocortical innervation CORTEX

Fig. 6. (Right.) The ‘handshake’behveen thalamic and prep/ate axons and precise topography of the ear/y thaiamocorticai projection. (A) In an embryonic day 15.5 (E/5.5) fetal rat brain, a crystal of 7,7 ‘- dioctadecyl-3,3,3 ‘,3’-tetramethyiindocarbocyanine perch/orate (dii) was p/aced into the dorsal thalamus [putative lateral genicuiate nucleus (LCN)] and a 4-[4-(dihexadecy/amino)styryl]-N-methyipyridinium iodide (diA) crystal into what was presumed to be the ‘matching’region of the occipital cortex (putative area 7 7), so that both tholamocortical and corticofugalpreplate tibres were lobe/led. At this stage, very few if any thalamic fibres have reached the occipital cortex (no cells back-labeled with diA were observed in the thalamus), and the dil-labeled fibre population was exclusively thalamocortical (no cortical cells were back- labelled with dil). In this coronal section, many diA-lobelied (ye//ow/green) and dil-labeled fibres (orange) are associated closely in the same plane of focus, indicating that the prep/ate and thaiamo- cortical fibres become associated intimately when they meet each other. Scale bar, 7 00 pm. (B) This fluorescence photomicrograph was taken from a I OO-pm-thick horizontal section of an E20 rat brain after placement of three crystals of different carbocyanine dyes, diA, dil and 4-[4- (didecy/amino)sty@]-N-methylpyridinium iodide (diAsp), in a parasagittal row along the right hemisphere”. Since thalamic fibres have a/ready arrived at the cortex at El 6, each placement labeled a mixed bundle of thalamocortical and corticofugal axons. Six weeks’ incubation at room temperature was used to enable full anterograde and retrograde diffusion. Three distinct bundles are c/ear/y visible passing through the primitive internal capsule without substantial mixing or crossing. Abbreviations: dien, diencephalon; med, media/; post, posterior; and telen, telencephalon. Scale bar, 500 g.m.

they do not wait below but run up diagonally in fascicles through the cortical plate. They then gather over the superplate for some days before growing downwards and arborizing towards the bottom of the plate among cells that presumably would have been destined normally for layer 4 (Refs 26 and 27) (Fig. 2). It seems very likely that thalamic fibres grow through the hostile environment of the cortical plate by following the oblique fascicles of preplate axons laid down earlier. The superficially very unusual pattern of growth of thalamic axons in the reeler mouse, and their aberrant appearance in the adult’l, could be explained by the same algorithm of interaction and guidance that we propose for the normal animal. Conclusions and questions The timing of outgrowth, accumulation, invasion and termination of thalamic axons, and the establishment of the waiting period, are at least partly regu- lated by a cascade of simple signals that is expressed by the cortex. The ordered outgrowths from both pre- on the preplate and thalamic-axon arrays, or merely plate and thalamus might depend on interaction with on the spatial separation of parts of the growing extracellular guidance cues or other fibre systems, but scaffolds. the sequence in which fibres leave the cortex and the It must be emphasized that the basic topography thalamus might also play a part in establishing fibre that is established by the early guidance of thalamic order. Even in rodents, where the whole process of axons is likely to be refined, both during and after cortical neuronogenesis is complete in a week or less, the waiting period. This could be achieved through the anteroventral areas of the hemisphere are more the selective retraction of side-branches within the than a day ahead of the caudodorsal areas in matur- subplate layers2,s3,and by local refinement within the ity‘Ur7, and a similar temporal gradient probably exists cortex through axonal rearrangement, competitive in the developing thalamus. These gradients might interaction, dendritic restructuring and synaptic modifi- impose ‘chronotopic’ patterns on the outgrowth of cation, much of which is likely to depend on afferent axons, each wave of axons being laid down on the activity54. surface of the pre-existing axon array, and thus deter- It would be wrong to disguise the many complex- mine the sequence in which the two fibre arrays arrive ities of thalamocortical innervation that remain to be and interact in the basal telencephalon (Fig. 7). Of explained, especially in animals with larger and more course, time is uni-dimensional, and cannot deter- complex cerebral hemispheres. However, we begin to mine the entire two-dimensional pattern of thalamo- see how the basic pattern of thalamocortical inner- cortical innervation. The other axis of the map might vation might be generated by the successive combi- depend on some sort of gradient of molecular signals nation of a number of individually simple mechanisms.

TINS Vol. 18, No. 9. 1995 395 Fig. 7. How ‘chronotopy’ and the ‘handshake’ could determine thalamocortical topography. These diagrams summarize observations on the relationship between fibres from the prep/ate and the tholamus ond show the way in which temporal grodients of maturation could contribute to the establishment of ear/y topography. They represent coronal sections through the left hemisphere at the /eve/ of the internal capsule (K) of the normal rodent shortly after the appearance of the prep/ate (A), ofier the start of formation of the cortical plate in the lotera/ port of the hemisphere (B), and near the start of the waiting period (C). The prep/ate is shown OS pink, the cortical plate is shown as green. Efferent projections from the cortical plate are not shown. Prep/&e cells (red) send out their axons towards the internal capsule in order, reflecting the ventral-to-dorsal gradient of moturotion. If oxons tend to grow on the surface of pre-existing axons, without crossing them, this chronological sequence of outgrowth would establish one axis of a topographic array. If there is a matched tempo& gradient of outgrowth from the thalamus (light blue), this could determine the order in which tholumic fibres (dark blue) and axons from the prep/ate of the corresponding area of the cortex meet and asso- ciate with each other in the handshake region of the basal telencepholon. This sequential establishment of contact with fibres of the prep/ate scaffold might oid the construction of one axis of the topographic distribution of thalamic axons to the cortex.

Selected references 22 Shatz, C.J. and Luskin, M.B. (1986) ]. Neurosci. 6,3655-3668 Acknowledgements 1 O’Leary, D.D.M. (1989) Trends Neurosci. 12, 400-406 23 Rakic, P. (1977) Philos. Trans. R. Sot. London Ser. B 278, 245-260 The authors thank 2 Caviness, V.S., Jr and Frost, D.O. (1980) \. Comp. Neural. 194, 24 Mitrofanis, J. and Guillery, R.W. (1993) Trends Neurosci. 16, Andri M. Gofj’inet 3.5.5-367 240-245 3 Caviness, V.S., Jr (1988) in Cellular Thalamic Mechanisms 25 M&in, C. and Godement, P. (1994) Sot. Neurosci. Abstr. 20, for providing (Bentivoglio, M. and Spreafico, R., eds), pp. 489-499, Elsevier 1683 reeler specimens, 4 Berry, M. and Rogers, A.W. (1965) J. Anat. 99, 691-709 26 MolnBr, Z. and Blakemore, C. (1992) Sot. Neurosci. Abstr. 18, Richard Adams for 5 Lund, R.D. and Mustari, MJ. (1977) I. Comp. Neural. 173, 778 289-305 27 Yuasa, S., Kitoh, J. and Kawamura, K. (1994) Anat. Embryol. help with confocal 6 Miller, M.W. (1988) in Cerebral Cortex (Vol. 7) (Peters, A. and 190, 137-154 microscopy, and Jones, E.G., eds), pp. 133-175, Plenum Press 28 GBlhwiler, B.H. (1988) Trends Neurosci. 11, 484-489 Pat Cordery and 7 Bayer, S.A. and Altman, J. (1991) Neocortical Development, 29 Yamamoto, N., Kurotani, T. and Toyama, K. (1989) Science Raven Press 245, 192-194 William Hinkes for 8 Rakic, P. (1981) in The Organization of the Cerebral Cortex 30 Bolz, J. et al. (1990) Nature 346, 359-362 technical assistance. (Schmitt, F.O., ed.), pp. 7-28, MIT Press 31 MolnBr, Z. and Blakemore, C. (1991) Nature 351, 475-477 Our work was 9 Marin-Padilla, M. (1971) Z. Anat. Entwicklungsgesch. 134, 32 Bolz, J., Novak, N. and Staiger, V. (1992) J. Neurosci. 12, supported by grants 117-145 3054-3070 10 Angevine, J.B., Jr et al. (1970) Anat. Rec. 166, 257-262 33 Yamamoto, N. et al. (1992) Neuron 9, 217-228 from the Wellcome 11 Shatz, C.G., Chun, J.J.M. and Luskin, M.B. (1988) in Cerebral 34 Rennie, S., Lotto, R.B. and Price, D.J. (1995) Neuroscience 61, Trust, Medical Cortex, Vol. 7 (Peters, A. and Jones, E.G., eds), pp. 35-58, 547-564 Research Council, Plenum Press 35 Romijn, H.J. (1988) Biol. Cell 63, 263-268 36 Lumsden, A. (1991) in The Nerve Growth Cone (Letourneau, Soros-Hungarian 12 D’Archangelo, G. et al. (1995) Nature 374, 719-723 13 Caviness, V.S., Jr, Crandall, J.E. and Edwards, M.A. (1988) in P.C., Kater, S.B. and Macagno, E.R., eds), pp. 167-180, Raven Academy of Sciences Cerebral Cortex, Vol. 7 (Peters, A. and Jones, E.G., eds), pp. Press Foundation and 489-499, Plenum Press 37 Tuttle, R. et al. (1995) J. Neurosci. 15, 3039-3052 Human Frontier 14 Honig, M.G. and Hume, R.I. (1989) Trends Neurosci. 12, 38 Giitz, M. et al. (1992) Development 116, 507-519 333-335 39 Yamamoto, N. et al. (1992) Sot. Neurosci. Abstr. 18, 224 Science Program. 1.5 Godement, P. et al. (1987) Development 101, 697-713 40 Sperry, R.W. (1963) Proc. Nat1 Acad. Sci. USA 50, 703-709 It forms part of the 16 MolnBr, Z. and Blakemore, C. (1990) J. PhysioL 430, 104P 41 McConnell, S.K., Ghosh, A. and Shatz, C.J. (1989) Science 245, work of the Oxford 17 Blakemore, C. and MolnBr, Z. (1990) Cold Spring Harbor Synp. 978-982 Quant. Biol. 55, 491-504 42 Shatz, C.J. et al. (1990) Cold Spring Harbor Symp. Quant. Biol. 55, McDonnell-Pew 18 DeCarlos, J.A. and O’Leary, D.D.M. (1990) Sot. Neurosci. Abstr. 469-480 Centre for Cognitive 16, 311 43 Allendoerfer, K.L. and Shatz, C.J. (1994) Annu. Rev. Neurosci. Neuroscience and 19 DeCarlos, J.A. and O’Leary, D.D.M. (1992) J. Neurosci. 12, 17, 185-218 44 MolnBr, Z., Adams, R. and Blakemore, C. (1993) Sot. Neurosci. the MRC Research 1194-1211 20 Erzurumlu, R.S. and Jhaveri, S. (1992) Cereb. Cortex 2, 336-352 Abstr. 19, 1088 Centre in Brain and 21 Catalano, S., Robertson, R.T. and Killackey, H.P. (1991) Proc. 4.5 Miller, B., Chou, L. and Finlay, B.L. (1993) J. Comp. Neural. Behaviour. Nat1 Acad. Sci. USA 88, 2999-3003 335, 16-41

396 TIh’S Vol. 18, No. 9, 1995 Z. MoMr and C. Blakemore - Development of thalamocortical innervation CORTEX

46 Goffinet, A.M. (1980) Anat. Embryol. 159, 199-210 18, 778 47 Agmon, A. et al. (1995) 1. A’eurosci. 15, 549-561 51 Caviness, V.S., Jr (1976) 1. Camp. Neural. 170, 435-488 48 Herrmann, K., Antonini, A. and Shatz, C.J. (1994) Eur. ]. 52 Naegele, J.R., Jhaveri, S. and Schneider, G.E. (1988) 1. Camp. Neurosci. 6, 1729-l 742 Neural. 277, 593-607 49 Bicknese, A.R. et al. (1994) I. Neurosci. 14, 3500-3510 53 Ghosh, A. and Shatz, C.J. (1992) I. Neurosci. 12, 39-55 50 Bicknese, A.R. and Pearlman, A.L. (1992) Sot. Neurosci. Abstr. 54 Shatz, CJ. (1990) Neuron 5, 745-756

Exuberant development of connections, and its possible permissive role in cortical evolution

Giorgio M. Innocenti

The callosal visual connections of the cat provide a model for studying the phenotypes of cortical axons and their differentiation. The terminal arbor of a callosal axon develops in several successive stages. At each stage, the arbor approximates the adult phenotype more closely. This is achieved through two mechanisms: (I) exuberant, but increasingly constrained, growth and (2) partial deletion of previously generated parts of the arbor. This differentiation is controlled by interactions of the axon with its cellular environment, and by visual experience. It might have played a permissive role in the evolution of the cerebral cortex by enabling adjustments of cortical connectivity to changes in the number, size, internal organization and cellular composition of cortical areas.

Trends Neurosci. (1995) 18, 397402

Two facets of axonal phenotype ORTICAL AREAS - and within each area, layers, C cortical columns and types of neurones - have dif- Two facets of axonal phenotype are important ferent and characteristic connectivity. In the course of because they implement computational strategies that evolution, the number and sizes of areas and layers’,‘, are typical of the CNS. First is the spatial distribution and the types and numbers of columns and neurones, of the terminal arbor, which determines the topo- have changed. The concomitant evolution of the con- graphical relationship that the axon establishes nectivity patterns required flexible developmental between its parent soma and its synaptic targets. This processes. Two of these processes are retained in aspect of the axonal phenotype implements the spa- extant mammals3”. The first is specification of cortical tial transformations (mapping) between interrelated areas by afferents, particularly the thalamic afferents. neuronal assemblies, and has been the main focus of The second is exuberant axonal growth in develop- past and recent theories of the development of neural ment, that is, the fact that axons that originate or connections (for recent reviews, see Refs 13-16). The terminate, or both, in cerebral cortex initially form mapping algorithm is usually complex. An axon rarely excessive branches and synapses; these are distributed establishes ‘point-to-point’ connections. Instead, its somewhat diffusely, and refined by selective pruning terminal arbor diverges more or less widely within a (for review, see Ref. 7). The key feature of these two volume where it distributes selectively to certain neur- processes is that fundamental aspects of the cortical ones, and to certain subcellular compartments of these phenotype are not narrowly pre-programmed but neurones. The second facet of the morphological instead result from interactions between cortical neur- phenotype of an axon includes the diameter of the ones and their cellular environment. axon trunk and the length, diameter and spatial Here we focus on adult cortical organization and relationships of its branches17. These geometrical the development of visual callosal axons. Evidence properties of the axon determine its conduction proper- suggests that several aspects of the development of ties; therefore, they provide the structural basis of this system of connections can be assumed to apply to the temporal transformations between interrelated other connections in the cerebral cortex and else- neuronal assemblies. where6-s. The notion of ‘axonal phenotype’ refers to These two facets of axonal phenotype became clear an open-ended set of features, some of which are con- in recent studies of biocytin-labelled callosal axons, sidered below with respect to their functional signifi- interconnecting visual areas 17 and 18 of the cat. Giorgio M. cance and differentiation. These features have only Computer techniques were used for the three-dimen- ~nnocentiis at the recently become accessible to investigation, with the sional reconstruction of axons, and for the simulation Institut d’Anatomie, introduction of new anterograde tracers that enable of their activity”,“. 9 rue du Bugnon, the visualization of individual axons, including the It was found that most callosal axons terminate in ImSLausanne, long axons (for examples, see Refs 8-12). radially oriented clusters of pre-terminal branches Switzerland.

0 1995, Elsevier Science Ltd TINS Vol. 18, No. 9, 1995 397