Development of Cortical Circuits: Lessons from Ocular Dominance Columns

Home , Axon guidance, Critical period, Ocular dominance column, Retinal waves, Subplate

REVIEWS

DEVELOPMENT OF CORTICAL CIRCUITS: LESSONS FROM OCULAR DOMINANCE COLUMNS

Lawrence C. Katz and Justin C. Crowley The development of ocular dominance columns has served as a Rosetta stone for understanding the mechanisms that guide the construction of cortical circuits. Traditionally, the emergence of ocular dominance columns was thought to be closely tied to the critical period, during which columnar architecture is highly susceptible to alterations in visual input. However, recent findings in cats, monkeys and ferrets indicate that columns develop far earlier, more rapidly and with considerably greater precision than was previously suspected. These observations indicate that the initial establishment of cortical functional architecture, and its subsequent plasticity during the critical period, are distinct developmental phases that might reflect distinct mechanisms.

ORIENTATION COLUMNS Historically, neuronal development has been divided guided the interpretation of developmental events in Orientation tuning is a property into a sequence of events that leads from the initial spec- many other systems. Despite the powerful appeal of of visual cortical neurons that ification of neuronal cell fate to the eventual emergence this general model, and the experimental support that allows the detection of lines and of adult circuits1. In this formulation, many circuits accumulated over several decades, a number of recent edges within visual scenes by encoding their orientations. undergo a pivotal transition in which precise patterns of findings indicate that some of the assumptions under- Neurons that share the same synaptic connections emerge from an earlier stage of lying the conventional formulation might need to be orientation tuning are grouped more coarsely specified connections. Although the initial revised. Such revisions, in turn, indicate that alterna- into orientation columns. organization of neural circuits relies on a variety of mol- tive explanations for the patterning of connections ecular cues that guide axons to generally appropriate should be considered, and that the definition of and regions, the final specification of patterned connections evidence for ‘activity-dependent refinement’ requires is widely held to depend on patterns of neuronal activity, greater precision. generated either by circuits intrinsic to the developing brain or by early experience2–11. In particular, as postu- History of theories of column development lated by Hebb, correlations in presynaptic and post- Hubel and Wiesel initially described ocular dominance synaptic activity patterns strengthen and retain ‘correct’ columns in the early 1960s15. By making electrophysio- synapses, and eliminate ‘inappropriate’ connections. logical recordings in cat primary visual cortex, they In the central nervous system, and in the mam- noted that the two eyes differentially activated cortical Howard Hughes Medical malian neocortex in particular, much of this prolonged neurons (the physiological property of ocular domi- Institute and Department of Neurobiology, Box 3209, sculpting of neuronal connections is thought to occur nance). Cells with similar eye preference were grouped Duke University Medical during ‘critical periods’,when circuits are particularly together into columns, and eye dominance shifted Center, Durham, North susceptible to external sensory inputs. Such ideas orig- periodically across the cortex. On the basis of a few Carolina 27710, USA. inated from developmental studies of functional archi- recordings in very young, visually inexperienced cats, Correspondence to L.C.K. tecture in the mammalian visual cortex, especially the Hubel and Wiesel originally argued that ‘innate’ mech- e-mail: 12–14 [email protected] formation of ocular dominance columns (FIG. 1). anisms determined the organization of the cortex into DOI: 10.1038/nrn703 This highly influential body of work subsequently ocular dominance columns and ORIENTATION COLUMNS16

34 | JANUARY 2002 | VOLUME 3 www.nature.com/reviews/neuro REVIEWS

Temporal Primary visual cortex layer 4 A substantial alteration in this formulation occurred retina when transneuronal transport of tritiated amino acids Ipsilateral (and later, sugars) — for example, 3H-proline — made it possible to directly visualize ocular dominance columns19–23. In adult monkeys19, injections of one eye Nasal retina Contralateral produced bands of transneuronally transported label that revealed the termination patterns of one eye’s representation in the cortex, alternating with dark bands, Nasal retina Ipsilateral nearly devoid of label, which corresponded to the LGN other eye’s thalamic input (FIG. 2a). Monocular deprivation during the postnatal critical period led to a Contralateral remarkable and satisfying correspondence between the Temporal electrophysiological loss and gain of responsiveness, retina and the shrinkage and expansion of eye-specific bands in layer 4 (REFS 22–24). Figure 1 | Segregation of eye-specific information at the early stages of visual processing. In mammals with binocular vision, the nasal portion of one retina encodes the same part of the The same approach was then applied to the develop- visual world as the temporal portion of the other retina. The axons of retinal ganglion cells from the ing visual system to determine how ocular dominance nasal portion of each retina cross the optic chiasm and project to the same lateral geniculate columns first formed22,23. Unlike in the adult, injections nucleus (LGN) as the axons from the temporal portion of the other eye. These projections form of amino acids into cats’ eyes before the onset of the criti- discrete, eye-specific LGN layers. The projection from the LGN to layer 4 of the primary visual cal period yielded a homogeneous band of label in layer 4, cortex maintains this eye-specific segregation by terminating in eye-specific patches that are the regardless of which eye was injected. Beginning at about anatomical basis for ocular dominance columns. Ocular dominance columns can therefore be considered to correspond to an eye (left or right) or a retinal location (nasal or temporal). 3 weeks after birth, and continuing over the next month, the adult pattern of segregated ocular dominance stripes gradually appeared. The timing of the emergence of the (‘innate’ was used interchangeably with ‘genetic’ in their adult-like pattern overlapped beautifully with the period early writings17). Monocular eye closure during the first when columns were susceptible to alterations of visual few months of life — the critical period — decreased experience13,14,25. Physiological evidence, based on sin- the numbers of cells activated by the closed eye and gle-unit recordings, also indicated a higher proportion markedly increased the number of neurons activated of binocular neurons at early ages, perhaps reflecting the exclusively by the open eye12,13,18. The initial descrip- greater degree of overlap of afferents representing the tions of the effects of eye closure during the critical two eyes23. In contrast to the original Hubel/Wiesel for- period indicated that pre-existing connections had sub- mulation, these observations indicated that the precise sequently been altered, through a competitive process, organization of columns in layer 4 was not innately to cause the loss of cells driven by the closed eye. In specified, but was gradually moulded by the same mech- their interpretations, Hubel and Wiesel clearly distin- anisms that guided columnar rearrangement during guished between the innate mechanisms that guide the the critical period — correlation-based synaptic com- initial formation of cortical functional architecture, and petition. Indeed, computer models based on initially the experience-dependent, competition-based mecha- overlapping inputs and differential activity patterns can nisms responsible for their later modification during produce columns with patterns strikingly similar to the critical period. those observed in vivo 3–5,8,9,26–28.

Figure 2 | Early developmental organization of ocular dominance columns. Ocular dominance segregation in primates and carnivores precedes the onset of the critical period for ocular dominance column plasticity. a | Adult-like ocular dominance segregation occurs in the macaque monkey before birth. A surface view of radioactive proline labelling in layer 4 after injection of one eye shows clear alternating columns (alternating bright and dark bands) in an animal that had received no visual stimulation. b | Ocular dominance column segregation in the ferret occurs by postnatal day 16 (P16). In this coronal section from a P21 ferret (equivalent to a P0 cat), labelled axons form segregated patches in layer 4 of the visual cortex (pia is at the top). Scale bar, 500 µm. c,d | Ocular dominance columns form before the critical period in cat. Intrinsic signal optical imaging (c) and improved transneuronal transport methods (d) show ocular dominance segregation at P14; a reconstruction of the pattern in a P14 cat shows segregated columns at this early time. Part a reproduced with permission from REF.31 © 1996 Society for Neuroscience; part b reproduced with permission from REF.58 © 2000 American Association for the Advancement of Science; parts c and d reproduced with permission from REF.48 © 2001 John Wiley & Sons, Inc.

NATURE REVIEWS | NEUROSCIENCE VOLUME 3 | JANUARY 2002 | 35 REVIEWS

However, the carefully executed work of LeVay et al.23 waves seem to be crucial for the emergence of segregated revealed a complication of transneuronal transport. In retinal projection patterns in the LGN itself 39 (BOX 1). young animals, anterogradely transported tracer Moreover, recordings from both rodents and ferrets injected into one eye could leak into inappropriate eye- show that the spontaneous patterns of activity generated specific layers (‘spillover’) of the lateral geniculate in the two eyes activate thalamic neurons, thereby poten- nucleus (LGN). So, 3H-proline acted as a transneuronal, tially providing signals for driving the segregation of but not necessarily as a trans-synaptic, tracer. Knowing thalamic afferents in the cortex 40–42. that spillover was more severe in younger animals, Multielectrode recordings obtained before eye open- LeVay et al. sought to compensate for the consequent ing from the LGN of awake, behaving ferrets showed that blurring of cortical columns by quantifying the extent of the patterns of activity generated by the two eyes pro- spillover in the LGN. After accounting for spillover in duced correlational patterns, consistent with the idea that each animal individually, the authors concluded that the two eyes could act as independent oscillators. The axons of the eye-specific LGN layers were intermingled patterns of activity were more highly correlated at LGN in young cats, and that segregation of geniculocortical sites within the same eye-specific layer, and less well cor- axons into ocular dominance columns progressed over related between layers42. However, these recordings a period of several weeks, reaching the adult level at uncovered unexpected differences between inputs from about postnatal day 39 (P39). the two eyes (FIG. 3). Eliminating inputs from the eye As noted by these investigators, quantification of IPSILATERAL to the recorded LGN had virtually no effect on spillover could be done only at the end of the experi- the pattern of spontaneous activity in both layers, indicat- ments, leaving open the possibility that more tracer was ing that the CONTRALATERAL retina alone, perhaps in concert present in inappropriate layers during the 1–2 weeks with cortical feedback, could activate the entire circuit. By required for transport of the tracer than at the comple- contrast, elimination of the contralateral input strongly tion of the experiment. So, in young animals, the pres- increased the correlations between eye-specific layers, ence of a continuous band of label in layer 4 might producing the same pattern of activity that was observed represent either the absence of segregation — as subse- in the LGN when all retinal inputs were eliminated. This quent investigations have widely assumed — or result indicates that before eye opening the inputs from the two from extensive spillover in the LGN. On the basis of dif- eyes are not equivalent in their ability to activate thalamic ferent labelling techniques (see below), spillover seems and cortical circuits: the contralateral eye provides much to be the more likely explanation. stronger drive. A similar contralateral bias in responsiveness has been observed in recordings from cat cortex Early development of columns shortly after eye opening and before the onset of the criti- The formation of ocular dominance columns in cats cal period: most neurons were initially activated exclu- initially seemed to coincide with the beginning of the sively by the contralateral eye, and only weeks later did critical period, about 21 days after birth. As this is con- responses to the ipsilateral eye appear43. siderably later than eye opening (around P7), it was These observations in ferrets and cats are not consis- originally supposed that visual experience, in the form tent with a strictly Hebbian-type correlation mecha- of visually evoked patterns of action potentials, drove nism for segregating ocular dominance columns43.If ocular dominance column segregation22–24.However, spontaneous and evoked activities are both initially early work in macaque monkeys strongly indicated that strongly biased to one eye (the contralateral eye), these thalamocortical afferents begin to segregate into stripes inputs should effectively take over the entire cortex and before birth29, and are arranged into functional columns eliminate the much weaker ipsilateral inputs. At the very by birth30. More recent experiments have shown that, least, there should be a substantial discrepancy in the anatomically, ocular dominance column segregation in relative sizes of the two representations, but this is not newborn monkeys is as precise as in adults31. To recon- the case: ipsilateral and contralateral inputs normally cile these findings with previous data implicating activity- occupy roughly equivalent cortical territories. It would dependent competition in column formation, a further, seem, therefore, that there must be some mechanism non-visually driven source of activity was suggested that prevents early imbalances in activity from being to provide the signals for driving segregation in the translated into anatomical rearrangements. prenatal cortex. Even more surprising is the finding that retinal activ- In postnatal animals, local correlations in the firing of ity does not seem to be required for ocular dominance retinal ganglion cells were found even in the dark32. This column formation. If both eyes are removed early in life finding was followed by the discovery that the prenatal (P0 in the ferret), before the layers in the LGN have seg- retina could generate patterned activity before the differ- regated (and well before LGN afferents have reached entiation of photoreceptors33,34. Multielectrode record- layer 4 of area V1), normally segregated columns of

IPSILATERAL ings and calcium-imaging studies revealed the presence layer-specific LGN afferents still form in the cortex. On the same side of the body. of ‘retinal waves’,which are spontaneously generated, These columns faithfully reflect the pattern of connec- correlated patterns of activity that course across substan- tions seen in normal animals: they have the same peri- CONTRALATERAL tial areas of the neonatal retina35–38. Because these waves odicity and consist of thalamocortical projections from On the opposite side of the body. are generated independently in each eye, they could, in what would be the same eye-specific layer in the LGN44. 42 ENUCLEATION theory, provide the patterns of activity necessary to seg- Because ENUCLEATION does not silence the LGN , this Removal of the eyeballs. regate thalamic afferents in the cortex. Indeed, these finding does not rule out a role for correlated activity in

36 | JANUARY 2002 | VOLUME 3 www.nature.com/reviews/neuro REVIEWS

Box 1 | Activity-dependent segregation of retinal axons Current evidence for activity-based competition as a generative mechanism for ocular dominance columns relies largely on analogies with pattern formation in other parts of the central nervous system, rather than direct tests in the developing visual cortex. Compelling evidence for activity-based competition in the formation of segregated patterns came from work on dually innervated optic tecta in goldfish and frogs. In a normal frog, retinal ganglion cells from each eye project to the contralateral tectum. When a third eye primordium is implanted in tadpoles, axons of retinal ganglion cells from the ectopic eye innervate an optic tectum that also receives a normal complement of innervation from its usual source. Activity-dependent competition between the two sets of retinal afferents results in segregated, eye-specific stripes with a striking visual similarity to ocular dominance columns85 (see figure; autoradiographs reproduced with permission from REF.86 © 1981 Massachusetts Institute of Technology). In goldfish, regenerating axons from the two eyes, forced to grow into the same tectum, also form clear stripes87. Blocking retinal activity with 51 tetrodotoxin (TTX) prevents stripes from forming and can desegregate existing stripes . Significantly, blocking the NMDA (N-methyl-D-aspartate) glutamate receptor, which is required in mammals for the induction of long-term potentiation in the hippocampus, also induces desegregation or blocks segregation88,89. This points to an appealing model in which topographic cues intrinsic to retinal axons and the tectum guide axons from the two eyes to similar tectal locations, where activity-based competition sorts the two populations on the basis of correlated activity90. The formation of stripes represents a compromise between chemoaffinity cues guiding axons to the same locale, attractive interactions between axons with similar activity patterns (from the same eye), and repulsive interactions between axons with dissimilar activity patterns (from the other eye). In dually innervated tecta, it is extremely unlikely that an intrinsic stripe-like molecular cue in the tectum presages the segregation of stripes. A strong case has also been made for a role of correlated activity in the specification of eye-specific layers in the cat and ferret lateral geniculate nucleus (LGN; see REF. 91 for a recent review). Early in development, axons from the two eyes form simple, sparsely branched structures that form sparse synapses throughout the undifferentiated LGN. Later, the short, spine-like branches that synapse in the inappropriate Optic layer disappear, and there is a rapid and pronounced tectum proliferation of branches and synapses in the appropriate eye-specific layer, leading to the formation of segregated layers45. This depends on retinal activity: Transplanted eye blockade by either TTX or agents that block retinal waves39,49 prevents axons from developing their layer- specific arborizations. Normal eye A compelling body of anatomical, electrophysiological and pharmacological experiments substantiates all of these findings. However, they are metaphors for ocular dominance column formation, rather than direct tests of the process. It will be important in future experiments to apply some of these paradigms directly to the emergence of ocular dominance columns, now that we have a better idea of when the columns emerge during development.

the LGN of enucleated animals. However, to induce col- cleates genuine ocular dominance columns? Although umn formation, the patterns of activity in enucleated they have the right size and shape, and seem to reflect animals should carry correlational information that is inputs from different LGN layers, early enucleation can sufficiently similar to that in normal animals. Moreover, severely disrupt the organization of the LGN itself 45–47. that information must exist even when the LGN itself It is possible that the patchy connections observed after has not yet segregated into eye-specific layers. enucleation represent segregation across a modality Recordings from the LGN of young ferrets (at P25, other than ocular dominance. The atrophy of the LGN after columns have already formed) indicate that enu- that is induced by binocular enucleation might also cleation alters the correlational structure of sponta- result in a nonspecific clustering of LGN afferents. neous activity42. After this manipulation, activity in Although intriguing, the results of these experiments the two LGN layers is much more highly correlated, alone provide only indirect insights into the forces leading to degradation (but not elimination) of layer- guiding column formation. specific correlational cues. However, similar recordings have not been obtained at the appropriate ages Activity-based geniculocortical segregation (that is, before P16), so the patterns of spontaneous Despite the proposed central role for correlated neu- activity in these very young animals are unknown. ronal activity in driving the segregation of overlapping The LGN–cortical loop remaining after enucleation thalamic afferents in the primary visual cortex, remark- could generate sufficient correlational information to ably little evidence directly supports this idea. Several drive column formation. experiments indicate that activity is necessary to main- There are other important caveats in interpreting tain the segregated state, but few reveal how that state these findings. Are the columns that are present in enu- was achieved in the first place.

NATURE REVIEWS | NEUROSCIENCE VOLUME 3 | JANUARY 2002 | 37 REVIEWS

Dorsal Control Cut i Cut i + c However, recent work points to a different inter- 100 pretation. Using improved transneuronal autoradio- Anterior Posterior graphic techniques, together with optical imaging of intrinsic signals, it is now clear that, in the cat, Ventral 18 geniculocortical afferents are already segregated by P14 (REFS 43,48; FIG. 2c,d). Taking these newer findings into account, it is evident that Stryker and Harris4 began 80 their activity blockade when LGN afferents were already well segregated. So, rather than preventing segregation of afferents, activity blockade probably desegregated ocular dominance columns that were already present. This could be a consequence of sprout- ing or non-selective growth of axons induced by TTX 60 blockade, which has been observed in this system and others49–52. This interpretation is consistent with a recent demonstration that neural activity is required to maintain segregated, eye-specific axonal termination P Contra Ipsi C (A) (A1) patterns in the retinogeniculate projection. Blocking retinal ganglion cell activity in ferrets after eye-specific 40 On Off On Off segregation has occurred results in desegregation, with axons from both eyes mingled together in the same region of the LGN53. Although the results of Stryker and Harris4 cannot directly support the contention that activity is required 85 for the formation of ocular dominance columns, the 1 mm 20 data do highlight the importance of ongoing activity for normal development. Other manipulations of activity (for example, STROBE REARING, DARK REARING, blocking cortical NMDA (N-methyl-D-aspartate) receptors, or inacti-

Firing rate (Hz) vating the cortex through GABA (γ-aminobutyric acid) receptor agonists) have been carried out during the crit- 0 0 ical period (see REF.54 for a recent review). Therefore, Figure 3 | Contralateral bias of spontaneous activity in ferret LGN. Chronic multielectrode any effects on ocular dominance column segregation as array recordings show the normal pattern of bursting activity in the lateral geniculate nucleus a consequence of these manipulations occur against a (LGN) of a postnatal day 27 (P27) ferret, and the greater influence of the contralateral eye’s afferents. Left: sagittal view of a ferret LGN illustrates the method used to record multiple units in background of pre-existing columns. the awake ferret LGN. An array of eight electrodes spans the main eye-specific layers of the LGN. Contra, contralateral; Ipsi, ipsilateral; P, perigeniculate. Right: the activity recorded at each Critical period and thalamocortical segregation electrode is represented by one of the eight columns of pixels in each sweep; bright points In cats, ocular dominance column segregation is evi- correspond to high levels of activity. A comparison of control activity and the activity pattern in the dent at P14, about a week before the onset of the critical LGN after the ipsilateral optic nerve was cut (cut i) reveals little difference, whereas subsequently period. Transneuronal tracing revealed no evidence of cutting the contralateral optic nerve (cut i + c) markedly increased the correlations across eye- 48 specific layers. Reproduced with permission from REF.42 © 1999 American Association for the segregation a week earlier (P7) . However, as discussed Advancement of Science. above, transneuronal autoradiography is limited in its ability to detect segregation in very young animals. If it reveals segregated columns, they are certainly present, but The landmark experiments of Stryker and Harris4 failure to detect columns does not necessarily confirm TETRODOTOXIN A neurotoxin derived from the were designed to test directly whether activity (either their absence. Fugu, or puffer fish, which spontaneous or evoked by sensory experience) is This is evident when the state of thalamocortical specifically and reversibly blocks required to drive segregation of overlapping afferents in axon segregation is determined by direct injections of voltage-gated sodium channels. the developing cat visual cortex. They used repeated anterograde tracers into the developing LGN. Two binocular injections of TETRODOTOXIN (TTX) to block all recent studies using transneuronal transport concluded STROBE REARING 55,56 An experimental rearing forms of retinal activity from P14 (before segregated that segregation begins at P37 in the ferret . This cor- condition in which the only light columns are visible by transneuronal transport) until responds to the onset of the critical period, which is to which an animal is exposed is P45, when ocular dominance columns are clearly evi- around P35 (REF.57). By contrast, direct injections of the stroboscopic (flashing). This dent in normal animals. In the TTX-treated animals, ferret thalamus showed that columns were clearly segre- provides correlated stimulation 58 of the two eyes. there was no evidence of segregated columns at P45. gated by P16, almost 3 weeks earlier (FIG. 2b and FIG. 4). Instead, the label in layer 4 was continuous, similar to This is roughly equivalent to an embryonic cat 5 days DARK REARING the pattern in P14 animals23. The obvious conclusion, before birth. Recent multielectrode recordings have An experimental condition in consistent with all of the information available at the shown that correlated spontaneous activity in ferret which an animal is reared in time, was that blocking retinal activity prevented the cortex at P22 is organized into periodic patterns that total darkness so that only 59 endogenous activity is present in normal activity-driven competition that should have might reflect the presence of these early columns .As the developing visual system. resulted in segregation by P45. the visual systems of cats and ferrets develop with

38 | JANUARY 2002 | VOLUME 3 www.nature.com/reviews/neuro REVIEWS

Postnatal day 10 20 30 40 50 60 Overlap and segregation of thalamic afferents Regardless of when columns form, it is important to determine how they form. In the classical view, the Ocular dominance MD critical period anatomical basis for the homogeneous labelling patterns LGN injection observed after transneuronal transport was overlap- OD patches ping terminal fields of axons representing the two Transneuronal OD patches eyes22–24,29,62. In this model, well-developed but exuberant arborizations of thalamocortical axons were pruned LGN axons Eye opening to eliminate branches (and synapses) located in the Arrive, synapse Cortex visually responsive 63–65 in layer 4 wrong column. An alternative view is that axons are initially simple, and that a selective outgrowth of axon Figure 4 | Timeline of ferret ocular dominance column development. The emergence of terminals in appropriate columns, together with elimi- ocular dominance (OD) columns as revealed by direct lateral geniculate nucleus (LGN) injections precedes the critical period for monocular deprivation (MD), the appearance of segregation by nation of a small number of aberrant processes, pro- transneuronal transport, the opening of the eyes and the onset of visual responses in the cortex. duces mature circuits. Recent evidence supports the The appearance of segregated columns occurs while LGN axons are arriving and forming latter formulation: axons initially grow to their correct synapses in layer 4 of the primary visual cortex. The sequence of events in the developing cat locations and generate increasingly dense arborizations, cortex is the same. The equivalent ages for the cat can be roughly determined by subtracting 21 with little evidence of overlap between adjacent days from the ferret (for example, postnatal day 21 (P21) in ferret is approximately P0 in cat; P0 is columns58. Early emerging columns seem to be no less equivalent to cat embryonic day 44). Modified with permission from REF.58 © 2000 American Association for the Advancement of Science. segregated initially than later, implying that once they are established, little further refinement occurs (BOX 2). Without evidence from real-time imaging or finer- almost identical time courses57, this strongly indicates scale labelling, it is possible that there is ongoing elimi- that in cats, columns are present by birth, about 3 weeks nation of errant branches or collaterals. The ideal before the onset of the critical period. experiment for addressing this issue would be to visual- In monkeys, the separability of the critical period ize individual thalamocortical axons in vivo, and to and thalamocortical axon segregation is perhaps even determine their relationship with the emerging colum- more clear-cut. As discussed above, adult-like ocular nar architecture (perhaps as assessed by optical imag- dominance segregation occurs before birth in the ing). This is difficult for two reasons: first, it is difficult macaque monkey29–31, yet critical period plasticity is, to label individual axons, and second, it requires an by definition, a postnatal event. It is unclear whether independent method to visualize the overall structure activity-dependent remodelling of the macaque of the nascent column. There have been heroic attempts geniculocortical projection could occur before birth. to relate the morphology of individual axons to the However, some reports have indicated a change in V1 emergence of columns50, but in the light of recent find- physiological responses and gene expression associ- ings on the timing of column formation, it seems that ated with the initial exposure of the visual system to these studies were done after the columns had formed. light after dark rearing60,61. This indicates that opening Although these earlier studies attempted to find evi- the door to critical period plasticity might require dence for segregation at the level of individual arboriza- genuine visual stimulation, rather than spontaneous tions, the predominant change between P19 and P39 in activity alone. the cat is that arborizations become more elaborate. Regardless of the mechanism(s) driving their initial There are indications that arborizations change the lat- formation, ocular dominance columns clearly develop eral extent of their innervation: axons might initially considerably earlier than was believed when the provide input to two same-eye columns, and subse- hypothesis of segregation on the basis of activity-based quently reduce this to a single column. Rearrangements competition was first formulated. As a consequence, no that were observed might also reflect normal variation pharmacological manipulations of activity have been in arborizations, and perhaps non-homogeneous carried out sufficiently early to test whether activity is elaboration within individual columns. required for column establishment. To test the role of Further evidence that initially supported the view activity in column establishment, experiments would that overlapping thalamocortical axons gradually segre- have to begin no later than P10 in the ferret, or embry- gate into discrete domains was provided by the presence onic day 52 in the cat. Although monocular enucleation of a greater-than-expected number of binocularly acti- shortly after column formation in the ferret does not vated neurons in the cortex. This was consistent with the induce changes in the pattern of segregated input44,58, anatomical observations that inputs from the two eyes this approach alone does not directly test whether pat- were overlapping in layer 4. However, both older record- terned activity is involved. However, these results do ings and more recent work have shown that the initial indicate that a gross imbalance in retinal input is not, at state of the cortex is, if anything, highly monocular. these early stages, translated into morphological Most neurons in the cat visual cortex after eye opening changes as it would be during the subsequent critical are driven monocularly, rather than binocularly66. period. It remains an open question whether more sub- Moreover, in kittens, the cortical responses before P21 tle manipulations (such as silencing, rather than are strongly dominated by the contralateral eye. Binocular removing an eye) can cause shifts in the patterns of responses develop considerably later, long after ocular these early columns. dominance columns have formed43.

NATURE REVIEWS | NEUROSCIENCE VOLUME 3 | JANUARY 2002 | 39 REVIEWS

Box 2 | Activity-dependent segregation in other sensory systems mechanisms. But if the mechanisms are the same, they must operate in substantially different cellular environ- The idea that thalamocortical connections are initially highly precise, rather than ments. For example, optical imaging experiments have initially crude and only gradually refined, is supported by the analysis of development indicated that the response to monocular deprivation in other sensory systems, most notably the representations of whiskers in the during the critical period occurs first in the upper layers somatosensory barrel cortex and the glomeruli of the olfactory bulb. In both cases, the of the cortex (layer 2/3), and is then imposed, through initial projections laid down during the establishment phase are remarkably precise. intrinsic circuits, on the organization of geniculate axons Although some controversy remains, considerable evidence indicates that the initial in layer 4 (REF. 68). Columns in ferrets emerge by P16, patterning in these segregated systems does not depend on correlated activity patterns, before most upper layer neurons have migrated into whereas maintenance of the segregated state requires activity. position or extended axons. So, an intrinsic circuit impli- In the barrel cortex, ingrowing axons form precise termination patterns in layer 4 cated in critical period plasticity is simply absent when (REFS 92–94;but see REF.95). These patterns undergo little subsequent refinement, although activity blockade can reduce their subsequent specificity, and, as in the visual cortex during columns first form. the critical period, reduces their ability to undergo rearrangement96. Cortical activity In the ferret, the mechanism(s) that initially form 97,98 columns must be present by P16. Between E27 and P10, blockade does not prevent BARREL formation. Even if NMDA (N-methyl-D-aspartate) receptors are disrupted in postsynaptic cortical neurons, thalamocortical axons segregate LGN axons are in the cortical subplate, but have not yet into clusters, although the formation of the cellular aggregates that are characteristic of invaded layer 4 (REFS 69,70). Retinal waves are present barrels is disrupted99. throughout this time71, and correlation-based informa- Perhaps the most rigorous tests of the role of activity in generating modular tion could reach developing axons in the subplate. organization have been accomplished in the mouse olfactory bulb. Axons of olfactory Geniculate axons form synapses in the subplate well over sensory neurons bearing the same odorant receptor, which are widely distributed in the a week before they reach layer 4. Through interactions olfactory epithelium, converge onto a few distinct glomeruli in the olfactory bulb100–102. with the postsynaptic neurons in the subplate, these This is a remarkable feat of axon sorting, given that there are about 1,000 different axons could acquire information about their respective receptors and a correspondingly large number of distinct axon populations. At first eyes of origin. In this model, specific ingrowth into layer glance, this would seem to be an ideal case in which correlation-based sorting could be 4 might reflect the outcome of competitive events that involved in segregating axons into discrete glomeruli, as all the axons bearing the same take place earlier in the subplate. Ablation of the sub- receptor would presumably show highly correlated activity. However, several elegant plate prevents the formation of ocular dominance genetic manipulations have conclusively shown that neither spontaneous nor odorant- columns72,73, although these experiments do not explic- evoked activity is required for the initial specification of glomeruli. Disruption of the itly address the relative roles of activity-dependent 103,104 peripheral transduction apparatus silences the sensory neurons, but glomeruli or -independent cues that might be present on subplate form normally. Even elimination of postsynaptic neurons fails to disrupt glomerular neurons. The presence of columnar, patterned, sponta- specificity105. Although activity is not required to form the map, there is considerable neous activity in the cortex at early ages could also indi- evidence that its maintenance requires activity103,106, and that, as for ocular dominance 107 cate a role for local cortical circuits in the development columns, competitive interactions can occur after map formation . 59 The mechanisms underlying the development of visual cortical columns, barrels in of columns . It is not yet clear whether these patterns of somatosensory cortex, and glomeruli in the olfactory bulb, share the same overall activity are involved in constructing columns or reflect features: precise, rapid establishment of initial connections that is relatively immune to the presence of already segregated afferents. manipulations of activity, and a subsequent period of plasticity to manipulations of Even after axons reach the cortex, it is not yet possi- activity or the sensory periphery. ble to determine how the early columns form, or how precise (adult-like) they are. Axons are detectable as early as P10 in layer 4, but they are so sparse that it is Taken together, a picture emerges in which columns unclear whether they are organized into columns58. form rapidly, well before the critical period and with lim- During the days after their initial ingrowth into layer 4, ited production of exuberant projections. Furthermore, competition between axons might be required to during this initial stage of formation, ocular dominance establish appropriate territories. columns do not seem to respond to changes in activity as The rapid, early and specific emergence of columns, predicted by simple Hebbian rules. These findings re- and their resistance to activity imbalances or retinal inforce the idea that the critical period has both an onset removal, indicate that molecular cues could also guide and a termination57,67, and that it occurs against the the initial formation of columns44,58,74. In the decades background of an already differentiated system of after the original descriptions of ocular dominance col- columns. So, activity during the critical period does not umn development, knowledge of the molecular cues instruct the formation of columns from a blank slate. responsible for axon guidance and map formation has Rather, abnormal activity can compromise the normal exploded, providing a rich palette of plausible molecular pattern. In the absence of experimental manipulations, mechanisms that could generate the relatively simple the main role of visual experience during the critical striped patterns of ocular dominance columns. In con- BARREL period might be to reinforce and augment an already sidering whether molecular cues might be involved, it is A cylindrical column of neurons found in the rodent neocortex. appropriately situated set of basic connections, rather important to recognize that the distinction between ‘left’ Each barrel receives sensory than to instruct their de novo formation. and ‘right’ eye could be irrelevant. In each LGN, the eye- input from a single whisker specific layers receive retinal input from the nasal retina follicle, and the topographical What guides the establishment of columns? on the contralateral side of the brain, and the temporal organization of the barrels The observations that column establishment and the retina from the ipsilateral side. The distinction between corresponds precisely to the arrangement of whisker follicles critical period are separable developmental events do not nasal and temporal retina might be a critical feature of on the face. necessarily imply that these phenomena rely on different column development, as ocular dominance columns can

40 | JANUARY 2002 | VOLUME 3 www.nature.com/reviews/neuro REVIEWS

RECEPTOR TYROSINE KINASES be viewed as the cortical representation of this peripheral powerful tools for molecular analysis — transgenic and A family of membrane distinction. The distinction between nasal and temporal knockout animals and commercially available gene chips receptors, the intracellular retina is specified early in development, and retinal axons — cannot be used to directly approach this issue. domains of which catalyse the reaching the chiasm can choose to project either ipsilat- The mechanisms underlying ocular dominance col- phosphorylation, by ATP,of specific tyrosine residues on erally or contralaterally, on the basis of molecular cues on umn segregation cannot be uncovered solely by the their target proteins. their growth cones and at the chiasm75. This information standard experimental approaches of blocking neu- can be retained at the level of the LGN (see REFS 76–78). In ronal action potentials or postsynaptic receptors. Such achiasmatic sheepdogs, for example, the normally manipulations cannot distinguish between instructive crossed nasal axons innervate the appropriate layers in roles of activity (such as that envisioned by Hebbian the LGN on the same side of the brain, indicating an models) and permissive roles (for example, neurons affinity between nasal and temporal axons and their might need to be electrically active to differentiate nor- respective LGN layers79. mally). A more appropriate test for the role of activity is However, as there has previously been little motivation to artificially change the pattern of activity while leav- to search for molecular correlates of ocular dominance ing the relative levels unchanged. These are extraordi- column formation, any hypothesis at this point is simply narily difficult experiments to carry out, particularly in speculation. Many of the molecules implicated in attrac- very young animals. However, in the case of orientation tive and repulsive axon guidance are found in the cortex tuning in the visual cortex, such experiments show that at appropriate ages, but there is no evidence that any of the development of overall structure and pattern in ori- them are involved in column formation. Members of the entation maps is unchanged by alterations in the corre- ephrin family of RECEPTOR TYROSINE KINASES are widely dis- lational structure of retinal input, although changes are tributed in the prenatal monkey visual cortex 80, but they evident in detailed receptive field properties84. do not form any obvious stripe-like patterns (although To unravel how, or whether, activity cues and molec- obvious patterns are not an absolute requisite for the ular patterning information interact to drive column potential involvement of a molecule). There are interest- formation will require a leap of faith that such patterning reports of patchy distributions of various neurotrans- ing information actually exists. If it does, then a number mitter system components early in cortical develop- of approaches that have successfully identified axon ment81,82, but no evidence to directly implicate any system guidance and topographic cues should yield some hints in column formation. The most tractable mammalian as to their identity. Some 40 years after Hubel and Wiesel system for studying molecular or genetic cues — the suggested innate mechanisms for the development of mouse — shows critical period plasticity83, but lacks seg- cortical functional architecture, an intriguing system of regated ocular dominance columns. So, some of the most specification remains to be fully elucidated.

1. Cowan, W. M. in International Review of Physiology, 14. Hubel, D. H. & Wiesel, T. N. The period of susceptibility to Soc. Lond. B 278, 377–409 (1977). Neurophysiology III (ed. Porter, R.) 149–191 (Univ. Park the physiological effects of unilateral eye closure in kittens. 25. Wiesel, T. N. & Hubel, D. H. Single cell responses in striate Press, Baltimore, 1978). J. Physiol. (Lond.) 206, 419–436 (1970). cortex of kittens deprived of vision in one eye. 2. Katz, L. C. & Shatz, C. J. Synaptic activity and the construction 15. Hubel, D. H. & Wiesel, T. N. Receptive fields, binocular J. Neurophysiol. 26, 1003–1017 (1963). of cortical circuits. Science 274, 1133–1138 (1996). interaction and functional architecture in the cat’s visual 26. Von der Malsberg, C. & Willshaw, D. J. A mechanism for 3. Swindale, N. V. The development of topography in the visual cortex. J. Physiol. (Lond.) 160, 106–154 (1962). producing continuous neural mappings: ocularity cortex — a review of models. Netw. Comput. Neural Syst. 7, 16. Hubel, D. H. & Wiesel, T. N. Receptive fields of cells in striate dominance stripes and ordered retino-tectal projections. 161–247 (1996). cortex of very young, visually inexperienced kittens. Exp. Brain Res. (Suppl. 1), 463–469 (1976). 4. Stryker, M. P. & Harris, W. A. Binocular impulse blockade J. Neurophysiol. 26, 994–1002 (1963). 27. Swindale, N. V. A model for the formation of ocular dominance prevents the formation of ocular dominance columns in cat In this classic paper, Hubel and Wiesel describe the stripes. Proc. R. Soc. Lond. B 208, 243–264 (1980). visual cortex. J. Neurosci. 6, 2117–2133 (1986). initial state of columnar systems in the immature 28. Jones, D. G., Van Sluyters, R. C. & Murphy, K. M. The first (and only) experiments to directly test the striate cortex, and conclude that functional A computational model for the overall pattern of ocular role of activity in the initial formation of ocular architecture is innate. dominance. J. Neurosci. 11, 3794–3808 (1991). dominance columns. 17. Wiesel, T. N. & Hubel, D. H. Ordered arrangement of 29. Rakic, P. Prenatal genesis of connections subserving ocular 5. Erwin, E. & Miller, K. D. Correlation-based development of orientation columns in monkeys lacking visual experience. dominance in the rhesus monkey. Nature 261, 467–471 ocularly matched orientation and ocular dominance maps: J. Comp. Neurol. 158, 307–318 (1974). (1976). determination of required input activities. J. Neurosci. 18, 18. Wiesel, T. N. & Hubel, D. H. Extent of recovery from the 30. Des Rosiers, M. H. et al. Functional plasticity in the immature 9870–9895 (1998). effects of visual deprivation in kittens. J. Neurophysiol. 28, striate cortex of the monkey shown by the 6. Sur, M. & Leamey, C. A. Development and plasticity of 1060–1072 (1965). [14C]deoxyglucose method. Science 200, 447–449 (1978). cortical areas and networks. Nature Rev. Neurosci. 2, 19. Wiesel, T. N., Hubel, D. H. & Lam, D. M. K. Autoradiographic Using improved autoradiographic techniques, this 251–262 (2001). demonstration of ocular-dominance columns in the monkey paper conclusively showed that ocular dominance 7. Weliky, M. Correlated neuronal activity and visual cortical striate cortex by means of transneuronal transport. Brain columns can emerge fully without visual experience. development. Neuron 27, 427–430 (2000). Res. 79, 273–279 (1974). 31. Horton, J. C. & Hocking, D. R. An adult-like pattern of ocular 8. Miller, K. D. in Self-Organizing Brain: from Growth Cones to 20. LeVay, S., Hubel, D. H. & Wiesel, T. N. The pattern of ocular dominance columns in striate cortex of newborn monkeys Functional Networks (eds Vanpelt, J. et al.) 303–318 dominance columns in macaque visual cortex revealed by a prior to visual experience. J. Neurosci. 16, 1791–1807 (1996). (Elsevier Science, Amsterdam, The Netherlands, 1994). reduced silver stain. J. Comp. Neurol. 159, 559–576 (1975). 32. Mastronarde, D. N. Correlated firing of cat retinal ganglion 9. Miller, K. D., Keller, J. B. & Stryker, M. P. Ocular dominance 21. Hubel, D. H. & Wiesel, T. N. Ferrier lecture. Functional cells. I. Spontaneously active inputs to X- and Y-cells. column development: analysis and simulation. Science 245, architecture of the macaque monkey visual cortex. Proc. R. J. Neurophysiol. 49, 303–324 (1983). 605–615 (1989). Soc. Lond. B 198, 1–59 (1977). 33. Maffei, L. & Galli-Resta, L. Correlation in the discharges of 10. Constantine-Paton, M., Cline, H. T. & Debski, E. Patterned 22. LeVay, S., Wiesel, T. N. & Hubel, D. The development of neighboring rat retinal ganglion cells during prenatal life. activity, synaptic convergence, and the NMDA receptor in ocular dominance columns in normal and visually deprived Proc. Natl Acad. Sci. USA 87, 2861–2864 (1990). developing visual pathways. Annu. Rev. Neurosci. 13, monkeys. J. Comp. Neurol. 191, 1–51 (1980). 34. Galli, L. & Maffei, L. Spontaneous impulse activity of rat retinal 129–154 (1990). 23. LeVay, S., Stryker, M. P. & Shatz, C. J. Ocular dominance ganglion cells in prenatal life. Science 242, 90–91 (1988). 11. Wong, R. O. The role of spatio-temporal firing patterns in columns and their development in layer IV of the cat’s visual 35. Meister, M., Wong, R. O., Baylor, D. A. & Shatz, C. J. neuronal development of sensory systems. Curr. Opin. cortex: a quantitative study. J. Comp. Neurol. 179, 223–244 Synchronous bursts of action potentials in ganglion cells of Neurobiol. 3, 595–601 (1993). (1978). the developing mammalian retina. Science 252, 939–943 12. Wiesel, T. N. & Hubel, D. H. Single cell responses in striate A seminal paper that provided the first ever view of (1991). cortex of kittens deprived of vision in one eye. the pattern of developing ocular dominance columns, Using a technically advanced microelectrode array, J. Neurophysiol. 26, 1003–1017 (1963). and provided the experimental underpinning of most this paper revealed that spontaneous action potentials 13. Wiesel, T. N. & Hubel, D. H. Comparison of the effects of subsequent theories of column formation. in the previsual retina are organized into wave-like unilateral and bilateral eye closure on cortical unit responses 24. Hubel, D. H., Wiesel, T. N. & LeVay, S. Plasticity of ocular patterns. These patterns could provide cues for in kittens. J. Neurophysiol. 28, 1029–1040 (1965). dominance columns in monkey striate cortex. Phil. Trans. R. competitive interactions in the thalamus and cortex.

NATURE REVIEWS | NEUROSCIENCE VOLUME 3 | JANUARY 2002 | 41 REVIEWS

36. Wong, R. O., Meister, M. & Shatz, C. J. Transient period of 55. Finney, E. M. & Shatz, C. J. Establishment of patterned 83. Gordon, J. A. & Stryker, M. P. Experience-dependent correlated bursting activity during development of the thalamocortical connections does not require nitric oxide plasticity of binocular responses in the primary visual cortex mammalian retina. Neuron 11, 923–938 (1993). synthase. J. Neurosci. 18, 8826–8838 (1998). of the mouse. J. Neurosci. 16, 3274–3286 (1996). 37. Wong, R. O. L., Chernjavsky, A., Smith, S. J. & Shatz, C. J. 56. Ruthazer, E. S., Baker, G. E. & Stryker, M. P. Development 84. Weliky, M. & Katz, L. C. Disruption of orientation tuning in Early functional neural networks in the developing retina. and organization of ocular dominance bands in primary visual cortex by artificially correlated neuronal activity. Nature Nature 374, 716–718 (1995). visual cortex of the sable ferret. J. Comp. Neurol. 407, 386, 680–685 (1997). 38. Wong, R. O. L. & Oakley, D. M. Changing patterns of 151–165 (1999). 85. Constantine-Paton, M. & Law, M. I. Eye-specific termination spontaneous bursting activity of on and off retinal ganglion 57. Issa, N. P., Trachtenberg, J. T., Chapman, B., Zahs, K. R. & bands in tecta of three-eyed frogs. Science 202, 639–641 cells during development. Neuron 16, 1087–1095 (1996). Stryker, M. P. The critical period for ocular dominance (1978). 39. Penn, A. A., Riquelme, P. A., Feller, M. B. & Shatz, C. J. plasticity in ferret visual cortex. J. Neurosci. 19, 6965–6978 86. Constantine-Paton, M. in The Organization of the Cerebral Competition in retinogeniculate patterning driven by (1999). Cortex: Proceedings of a Neurosciences Research Program spontaneous activity. Science 279, 2108–2112 (1998). 58. Crowley, J. C. & Katz, L. C. Early development of ocular Colloquium (eds Schmitt, F. O., Worden, F. G., Adelman, G. Blocking retinal waves (see reference 35) in one retina dominance columns. Science 290, 1321–1324 (2000). & Dennis, S. G.) 47–67 (MIT Press, Cambridge, leads to laminar rearrangements in the LGN that are By using direct injections of tracers into the Massachusetts, 1981). reminiscent of the effects of activity imbalances in developing LGN, the authors show that ocular 87. Boss, V. C. & Schmidt, J. T. Activity and the formation of ocular dominance column formation. dominance columns develop much earlier and much ocular dominance patches in dually innervated tectum of 40. Mooney, R., Penn, A. A., Gallego, R. & Shatz, C. J. Thalamic more rapidly than was previously believed, and with a goldfish. J. Neurosci. 4, 2891–2905 (1984). relay of spontaneous retinal activity prior to vision. Neuron surprising degree of specificity. 88. Cline, H. T., Debski, E. A. & Constantine-Paton, M. N-methyl- 17, 863–874 (1996). 59. Chiu, C. & Weliky, M. Spontaneous activity in developing D-aspartate receptor antagonist desegregates eye-specific 41. Mooney, R., Madison, D. V. & Shatz, C. J. Enhancement of ferret visual cortex in vivo. J. Neurosci. 21, 8906–8914 (2001). stripes. Proc. Natl Acad. Sci. USA 84, 4342–4345 (1987). transmission at the developing retinogeniculate synapse. 60. Quinlan, E. M., Philpot, B. D., Huganir, R. L. & Bear, M. F. 89. Cline, H. T. & Constantine-Paton, M. NMDA receptor Neuron 10, 815–825 (1993). Rapid, experience-dependent expression of synaptic NMDA antagonists disrupt the retinotectal topographic map. 42. Weliky, M. & Katz, L. C. Correlational structure of receptors in visual cortex in vivo. Nature Neurosci. 2, Neuron 3, 413–426 (1989). spontaneous neuronal activity in the developing lateral 352–357 (1999). 90. Law, M. I. & Constantine-Paton, M. Anatomy and physiology geniculate nucleus in vivo. Science 285, 599–604 (1999). 61. Bear, M. F. & Rittenhouse, C. D. Molecular basis for induction of experimentally produced striped tecta. J. Neurosci. 1, Recordings from multielectrode arrays in the LGN of of ocular dominance plasticity. J. Neurobiol. 41, 83–91 (1999). 741–759 (1981). awake ferrets revealed patterns of activity that were 62. Wiesel, T. N. Postnatal development of the visual cortex and 91. Shatz, C. J. Emergence of order in visual system better correlated within than between eyes, consistent the influence of environment. Nature 299, 583–591 (1982). development. Proc. Natl Acad. Sci. USA 93, 602–608 with the predictions of retinal recordings (see reference 63. Purves, D. & LaMantia, A. S. Construction of modular (1996). 35). However, significant differences were noted in the circuits in the mammalian brain. Cold Spring Harb. Symp. 92. Catalano, S. M., Robertson, R. T. & Killackey, H. P. Rapid strength of ipsilateral and contralateral retinal inputs. Quant. Biol. 55, 445–452 (1990). alteration of thalamocortical axon morphology follows 43. Crair, M. C., Gillespie, D. C. & Stryker, M. P. The role of visual 64. Purves, D., Riddle, D. R. & LaMantia, A. S. Iterated patterns peripheral damage in the neonatal rat. Proc. Natl Acad. Sci. experience in the development of columns in cat visual of brain circuitry (or how the cortex gets its spots). Trends USA 92, 2549–2552 (1995). cortex. Science 279, 566–570 (1998). Neurosci. 15, 362–368 (1992). 93. Catalano, S. M., Robertson, R. T. & Killackey, H. P. Individual Using optical and single-unit recording techniques, 65. Purves, D., White, L. E. & Riddle, D. R. Is neural development axon morphology and thalamocortical topography in the authors show that neurons in the developing Darwinian? Trends Neurosci. 19, 460–464 (1996). developing rat somatosensory cortex. J. Comp. Neurol. cortex are initially activated almost exclusively by the 66. Fregnac, Y. & Imbert, M. Early development of visual cortical 367, 36–53 (1996). contralateral eye, with ipsilateral eye responses cells in normal and dark-reared kittens: relationship between 94. Agmon, A., Yang, L. T., O’Dowd, D. K. & Jones, E. G. developing much later and only with visual experience. orientation selectivity and ocular dominance. J. Physiol. Organized growth of thalamocortical axons from the deep They conclude that the intercalation of ipsilateral (Lond.) 278, 27–44 (1978). tier of terminations into layer IV of developing mouse barrel responses is unlikely to reflect Hebbian competition. 67. Fagiolini, M. & Hensch, T. K. Inhibitory threshold for critical- cortex. J. Neurosci. 13, 5365–5382 (1993). 44. Crowley, J. C. & Katz, L. C. Development of ocular period activation in primary visual cortex. Nature 404, 95. Senft, S. L. & Woolsey, T. A. Growth of thalamic afferents dominance columns in the absence of retinal input. 183–186 (2000). into mouse barrel cortex. Cereb. Cortex 1, 308–335 (1991). Nature Neurosci. 2, 1125–1130 (1999). 68. Trachtenberg, J. T., Trepel, C. & Stryker, M. P. Rapid 96. O’Leary, D. M., Borngasser, D. J., Fox, K. & Schlagger, B. L. Early eye removal does not seem to prevent the extragranular plasticity in the absence of thalamocortical in Symposium on the Development of the Cerebral Cortex formation of segregated patterns of columns in the plasticity in the developing primary visual cortex. Science No. 193 (eds Bock, G. R. & Cardew, G.) 214–222 cortex. The authors argue that retinal activity might 287, 2029–2032 (2000). (John Wiley & Sons, Inc., New York, 1995). not be necessary for segregation, and propose that 69. Johnson, J. K. & Casagrande, V. A. Prenatal development of 97. Chiaia, N. L. et al. Postnatal blockade of cortical activity by activity-independent cues might be involved. axon outgrowth and connectivity in the ferret visual system. tetrodotoxin does not disrupt the formation of vibrissa- 45. Sretavan, D. W. & Shatz, C. J. Prenatal development of Vis. Neurosci. 10, 117–130 (1993). related patterns in the rat’s somatosensory cortex. Brain retinal ganglion cell axons: segregation into eye-specific 70. Herrmann, K., Antonini, A. & Shatz, C. J. Ultrastructural Res. Dev. Brain Res. 66, 244–250 (1992). layers within the cat’s lateral geniculate nucleus. J. Neurosci. evidence for synaptic interactions between thalamocortical 98. Chiaia, N. L. et al. Effects of postnatal blockade of cortical 6, 234–251 (1986). axons and subplate neurons. Eur. J. Neurosci. 6, activity with tetrodotoxin upon the development and 46. Garraghty, P. E., Shatz, C. J., Sretavan, D. W. & Sur, M. 1729–1742 (1994). plasticity of vibrissa-related patterns in the somatosensory Axon arbors of X and Y retinal ganglion cells are differentially 71. Wong, R. O. L. Retinal waves and visual system cortex of hamsters. Somatosens. Mot. Res. 11, 219–228 affected by prenatal disruption of binocular inputs. Proc. Natl development. Annu. Rev. Neurosci. 22, 29–47 (1999). (1994). Acad. Sci. USA 85, 7361–7365 (1988). 72. Ghosh, A. Subplate neurons and the patterning of 99. Iwasato, T. et al. Cortex-restricted disruption of NMDAR1 47. Garraghty, P. E., Shatz, C. J. & Sur, M. Prenatal disruption of thalamocortial connections. Ciba Found. Symp. 193, impairs neuronal patterns in the barrel cortex. Nature 406, binocular interactions creates novel lamination in the cat’s 150–172; discussion 192–199 (1995). 726–731 (2000). lateral geniculate nucleus. Vis. Neurosci. 1, 93–102 (1988). 73. Ghosh, A. & Shatz, C. J. Involvement of subplate neurons in 100. Vassar, R. et al. Topographic organization of sensory 48. Crair, M. C., Horton, J. C., Antonini, A. & Stryker, M. P. the formation of ocular dominance columns. Science 255, projections to the olfactory bulb. Cell 79, 981–991 (1994). Emergence of ocular dominance columns in cat visual cortex 1441–1443 (1992). 101. Mombaerts, P. et al. Visualizing an olfactory sensory map. by 2 weeks of age. J. Comp. Neurol. 430, 235–249 (2001). 74. Katz, L. C., Weliky, M. & Crowley, J. C. in The New Cognitive Cell 87, 675–686 (1996). Revising earlier work (for example, see reference 23), Neurosciences (ed. Gazzaniga, M. S.) 199–212 (MIT Press, 102. Ressler, K. J., Sullivan, S. L. & Buck, L. B. Information this paper shows that ocular dominance columns in Cambridge, Massachusetts, 1999). coding in the olfactory system: evidence for a stereotyped cats emerge before the critical period, consistent with 75. Mason, C. A. & Sretavan, D. W. Glia, neurons, and axon and highly organized epitope map in the olfactory bulb. results in primates and ferrets. pathfinding during optic chiasm development. Curr. Opin. Cell 79, 1245–1255 (1994). 49. Sretavan, D. W., Shatz, C. J. & Stryker, M. P. Modification of Neurobiol. 7, 647–653 (1997). 103. Belluscio, L., Gold, G. H., Nemes, A. & Axel, R. Mice

retinal ganglion cell axon morphology by prenatal infusion of 76. Meissirel, C., Wikler, K. C., Chalupa, L. M. & Rakic, P. Early deficient in Golf are anosmic. Neuron 20, 69–81 (1998). tetrodotoxin. Nature 336, 468–471 (1988). divergence of magnocellular and parvocellular functional 104. Lin, D. M. et al. Formation of precise connections in the 50. Antonini, A. & Stryker, M. P. Development of individual subsystems in the embryonic primate visual system. Proc. olfactory bulb occurs in the absence of odorant-evoked geniculocortical arbors in cat striate cortex and effects of Natl Acad. Sci. USA 94, 5900–5905 (1997). neuronal activity. Neuron 26, 69–80 (2000). binocular impulse blockade. J. Neurosci. 13, 3549–3573 77. Chalupa, L. M., Meissirel, C. & Lia, B. Specificity of retinal 105. Bulfone, A. et al. An olfactory sensory map develops in the (1993). ganglion cell projections in the embryonic rhesus monkey. absence of normal projection neurons or GABAergic 51. Reh, T. A. & Constantine-Paton, M. Eye-specific segregation Perspect. Dev. Neurobiol. 3, 223–231 (1996). interneurons. Neuron 21, 1273–1282 (1998). requires neural activity in three-eyed Rana pipiens. 78. Casagrande, V. A. & Condo, G. J. Is binocular competition 106. Zheng, C. et al. Peripheral olfactory projections are J. Neurosci. 5, 1132–1143 (1985). essential for layer formation in the lateral geniculate nucleus? differentially affected in mice deficient in a cyclic nucleotide- The striking pattern of ocular dominance stripes in Brain Behav. Evol. 31, 198–208 (1988). gated channel subunit. Neuron 26, 81–91 (2000). frog optic tectum, induced by transplanting a third 79. Williams, R. W., Hogan, D. & Garraghty, P. E. Target 107. Zhao, H. & Reed, R. R. X inactivation of the OCNC1 channel eye, has served as a model for column development recognition and visual maps in the thalamus of achiasmatic gene reveals a role for activity-dependent competition in the in the cortex. This paper shows that retinal activity is dogs. Nature 367, 637–639 (1994). olfactory system. Cell 104, 651–660 (2001). the crucial feature for inducing the segregation of 80. Donoghue, M. J. & Rakic, P. Molecular gradients and eye-specific stripes. compartments in the embryonic primate cerebral cortex. 52. Frank, E. The influence of neuronal activity on patterns of Cereb. Cortex 9, 586–600 (1999). Online links synaptic connections. Trends Neurosci. 10, 188–190 (1989). 81. Dyck, R. H. & Cynader, M. S. An interdigitated columnar 53. Chapman, B. Necessity for afferent activity to maintain eye- mosaic of cytochrome oxidase, zinc, and neurotransmitter- FURTHER INFORMATION specific segregation in ferret lateral geniculate nucleus. related molecules in cat and monkey visual cortex. Proc. Encyclopedia of Life Sciences: http://www.els.net/ Science 287, 2479–2482 (2000). Natl Acad. Sci. USA 90, 9066–9069 (1993). cortical barrels: maps and plasticity | cortical plasticity: use- 54. Berardi, N., Pizzorusso, T. & Maffei, L. Critical periods during 82. Trepel, C., Duffy, K. R., Pegado, V. D. & Murphy, K. M. dependent remodelling | neural activity and the development of sensory development. Curr. Opin. Neurobiol. 10, 138–145 Patchy distribution of NMDAR1 subunit immunoreactivity in brain circuits | visual system development in vertebrates (2000). developing visual cortex. J. Neurosci. 18, 3404–3415 (1998). Access to this interactive links box is free online.

42 | JANUARY 2002 | VOLUME 3 www.nature.com/reviews/neuro