Ocular dominance (OD) columns, alternating regions of left and right eye input in the visual cortex of higher mammals, have long been thought to develop from an initially intermingled state by an activity-dependent process. While indirect evidence points to potential alternative mechanisms based on molecular cues, direct proof for a molecular difference between left- and right eye columns is missing. Here, we show that heat shock protein 90 alpha (Hsp90α) is expressed in a clustered fashion in the developing cat visual cortex. Clusters of Hsp90α-positive cells are in register with OD columns of the ipsilateral eye as early as postnatal day 16, when OD columns have just appeared. Importantly, a periodic, clustered expression of Hsp90α is already present weeks before OD columns have started to form, suggesting that molecular differences between future left and right eye OD columns may contribute to the segregated termination of eye specific afferents in the developing visual cortex.
The brain contains a number of sensory maps in which the outside world is represented in an orderly fashion. Many of these maps develop under tight control of molecular cues, which guide growing axons toward their correct termination sites in the target region (Sperry 1963; Cheng et al. 1995; Drescher et al. 1995; Mombaerts et al. 1996). More recent studies on the formation of retinotopic maps in the thalamus and midbrain have demonstrated that the initial, molecular-based mapping is complemented by activity-dependent refinement of connections during later stages of development (Grubb et al. 2003; McLaughlin et al. 2003; Chandrasekaran et al. 2005; Mrsic-Flogel et al. 2005). In the cerebral cortex, too, topographic map formation in sensory areas depends on the graded distribution of membrane bound guidance factors (Prakash et al. 2000; Vanderhaeghen et al. 2000; Dufour et al. 2003; Cang, Kaneko et al. 2005). Like in subcortical structures, molecular cues and patterned neuronal activity act in concert to generate the mature fine grain topographic maps (Cang, Renteria et al. 2005; Cang, Niell et al. 2008).
In contrast, other cortical maps have long been thought to develop exclusively under the control of activity-dependent mechanisms. The most prominent example is the ocular dominance (OD) map in the visual cortex of higher mammals, where the segregated termination of eye-specific thalamic afferents in layer 4 forms the structural basis for alternating bands and patches of left and right eye preference (Hubel and Wiesel 1962; LeVay et al. 1975; Shatz and Stryker 1978). OD columns are thought to develop from an initially uniform distribution of left and right eye inputs in layer 4 (Hubel et al. 1977; LeVay et al. 1978) by an activity dependent process, as evidenced by the observation that altering the relative activity levels of both eyes by monocular deprivation (MD) causes expansion of open eye columns at the expense of deprived eye columns (Hubel et al. 1977; Shatz and Stryker 1978; see also Katz and Shatz 1996). However, later experiments have casted doubt on the notion that OD columns develop from an initially intermingled state: Crowley and Katz (2000), using more selective tracing methods than those employed in the original studies, found that OD columns in ferret visual cortex form much earlier than initially thought, and, importantly, without an apparent phase of overlap between left and right eye inputs. This finding speaks against an activity-dependent mechanism being responsible for the initial sorting of eye-specific inputs in the visual cortex, but rather argues for a prespecified termination pattern.
Other experiments, too, have challenged the view that OD columns develop by activity dependent remodeling only: OD column-like domains can form without any input from the eyes (Crowley and Katz 1999), and blockade of early spontaneous activity in the retina influences the development of OD columns, but does not prevent their formation (Huberman et al. 2006). In addition, it has been shown that early in development input to the visual cortex is strongly dominated by the contralateral eye, while inputs from the ipsilateral eye are only strengthened later (Crair et al. 1998), an observation hard to reconcile with an activity-dependent, competition-based mechanism. Together, these findings raised the possibility that molecular cues may contribute to or even govern setting up OD domains (Crowley and Katz 1999, 2000; Huberman et al. 2006), but to date direct evidence for such molecules is lacking.
We therefore screened for genes differentially expressed in OD columns in developing cat visual cortex. We found one candidate gene, cat heat shock protein 90 alpha (Hsp90α), to be specifically associated with OD columns of the ipsilateral eye. Clusters of Hsp90α expression were already present about 2 weeks before OD columns had formed, suggesting that future left and right eye OD columns are molecularly distinct prior to the segregation of thalamic fibers in the developing visual cortex and that such molecular differences may contribute to setting up the initial pattern of OD columns.
Materials and Methods
Experiments were carried out on 1- to 28-day-old cats in accordance with the guidelines of the local authorities and the Society for Neuroscience.
Optical imaging of intrinsic signals was employed to visualize OD maps in the cat visual cortex (Hübener et al. 1997). Anesthesia was induced with an i.m. injection of ketamine and xylazine. Animals were anesthetized throughout the procedure by artificially ventilating them with 1.0–1.5% halothane in a 3:2 mixture of N2O/O2. A craniotomy was performed over the posterior part of one hemisphere and parts of areas 17 and 18 were exposed by careful removal of the dura. The cortex was covered with 2% agarose in saline and sealed with a cover slip. The cortex was illuminated with 707 nm light, and images were captured with a cooled slow-scan CCD camera focused 500 µm below the cortical surface. Visual stimuli consisted of large field moving gratings of different orientations presented independently to the 2 eyes. Differential OD maps were computed by dividing the sum of all images obtained during stimulation of one eye by that of the other eye.
Tissue Sampling from Individual OD Columns
OD column samples for the PCR-based differential screen were collected from postnatal day (P) 16 cats following optical imaging. By overlaying the OD map onto an image of the surface blood vessel pattern, cortical regions corresponding to ipsi- and contralateral OD columns in area 17 were identified. Tissue samples were collected by centering a sharpened cannula with an inner diameter of 0.4 mm on an OD column and slowly lowering it into the visual cortex while gently applying suction. Samples were ejected from the cannula by positive pressure and immediately frozen in RNase-free Eppendorf tubes.
The total RNA from ipsi- or contralateral-eye OD columns was isolated separately using the RNeasy mini kit (Qiagen). The total RNA was resuspended in 10 μL suspension solution (0.5% NP-40, 3U Prime RNase Inhibitor [Eppendorf]) and 0.5 μL was used to generate first-strand cDNA, which was then amplified by PCR to obtain larger quantities of cDNA as described previously (Dulac and Axel 1995; Tanabe et al. 1998). The amplified cDNA was analyzed by Southern blots hybridized with several 32P-labeled DNA probes from test genes expressed at high (GAD67, cannabinoid receptor), middle (otx1) or low levels (brain-derived neurotrophic factor [BDNF], NT3). The relative abundance of these genes in the amplified cDNA samples prepared from individual OD columns was comparable to their levels determined by northern blot of the P16 cat visual cortex, indicating that the amount of a given mRNA was not severely biased during cDNA amplification. Of the cDNA prepared, 1 μg was used to construct cDNA libraries of ipsi- and contralateral OD columns (λZAPII cDNA library construction kit, Stratagene). About 8000 recombinant phage clones from the cDNA library were plated at low density (500 clones/150 mm dish plate), and two identical membranes (Hybond-N+, Amersham) from each plate were differentially screened with 32P-labeled probes derived from the same cDNA of ipsi- and contralateral OD columns that was used to construct the cDNA library. Probes were prepared as described before (Dulac and Axel 1995; Tanabe et al. 1998). After phage clones showing differential expression in ipsi- or contralateral OD columns were identified, their DNA inserts were amplified using PCR of T3 and T7 primers and electrophoresed on agarose gels. Duplicate membranes prepared from the gel were hybridized with the same ipsi- and contralateral OD column-specific probes as used in the differential screen. Southern blot analysis was used to confirm the clones identified in the differential screen. Plasmids containing the cDNA of these clones were obtained from the isolated phages by performing phagemid rescue (Stratagene).
Isolation of Full-Length cDNA
To isolate the full-length cDNA of the positive clones, a cDNA library from the P16 visual cortex containing larger inserts was prepared in the λZAPII vector (Stratagene) and screened with 32P-labeled probes of partial cDNA of the clones identified in the differential screen.
Northern Blot Analysis
Total RNA was prepared from the P16 cat visual cortex. Twenty micrograms of total RNA was separated by electrophoresis on 1% formaldehyde gels and blotted on GeneScreen membrane (NEN). Membranes were hybridized with the 32P-labeled probes prepared from the rescued plasmids.
In Situ Hybridization
In situ hybridization using digoxigenin-labeled cRNA probes was carried out on frozen sections (20–50 μm thickness) as described (Tomita et al. 2000). To visualize the hybridized cRNA probes, NBT/BCIP (Roche) was employed as a substrate. As probes we used the rescued plasmid (cat Hsp90α, 1 kb) or cDNAs isolated by reverse transcriptase-PCR (RT-PCR) (cat Arc and BDNF).
Alignment Between In Situ Hybridized Sections and OD Maps
After visualizing OD maps in the visual cortex with intrinsic signal imaging, electrolytic lesions (5 μA, 5 s) were placed at 3 or more positions close to the imaged region. They were subsequently recovered in the histological sections and used to align the positions of Hsp90α-positive cells in 5–8 successive tangential sections with the OD map. At least 2 lesions were recovered in each histological section. Cortical blood vessels provided additional landmarks to facilitate the alignment of sections. Plotting of labeled cells was done blindly without any access to the OD map. The alignment of the cells' positions and the OD maps with the help of the electrolytic lesions was done only after all cells had been plotted. To determine the numbers of positive cells in different regions of the OD map, the map was smoothed (Gaussian; σ = 232 μm), equal areas were assigned to ipsi- and contralateral regions, and the cells were counted by a computer program. Two-dimensional cross-correlations were computed by determining the proportion of labeled cells in ipsilateral regions with increasing offsets in x- and y-direction between the OD map and the plotted cells.
Labeling of OD Columns by Arc/BDNF Induction
In order to label OD columns by Arc/BDNF induction, one eyelid was sutured shut under ketamine/xylazine anesthesia, and kittens were subsequently kept in total darkness overnight. On the following morning, animals were returned to normal light for 30 min in the alert condition, in order to induce expression of Arc and/or BDNF in open-eye OD columns. Animals were sacrificed with an overdose of pentobarbital, the brains were removed, and frozen sections underwent in situ hybridization for Arc or BDNF.
Reconstruction of Surface Views of Hsp90α-Positive Cells from Serial Coronal Sections
For animals younger than P14, we did not need to align the hybridization pattern with the OD map (which is not yet detectable with optical imaging). Therefore, coronal (rather than tangential) sections could be used. Before sectioning, a fine needle was inserted orthogonal to the section plane into the tissue block at 3 positions in order to facilitate later alignment of the sections. At each age, more than 70 serial coronal sections were prepared. They were subjected to in situ hybridization for Hsp90α, and images of the sections were aligned based on both the shape of each section and the needle holes. The positions of Hsp90α-expressing cells were plotted, projected into the tangential plane, and assembled to generate a surface view of their distribution.
One eyelid of kittens was sutured shut, either for 4 days (P14–18) or 8 days (P17–25). Following the last day of MD, OD columns of the non-deprived eye were labeled by light-induced Arc or BDNF expression, as described above. To this end, animals were kept in total darkness for one night, and then exposed to light for 30 min with the deprived eye still closed. After sacrificing the animal, frozen sections were in situ hybridized for Arc or BDNF.
Eye-Specific Gene Expression in the Early Cat Visual Cortex
In order to identify candidate molecules differentially expressed in OD columns serving both eyes in the developing cat visual cortex, we devised a PCR-based differential screen followed by several confirmation steps to eliminate false-positive clones (Fig. 1A) (Dulac and Axel 1995; Tanabe et al. 1998). In a first step, intrinsic signal imaging was used to visualize OD columns in the kitten visual cortex at P16, the age at which these structures first become apparent (Crair et al. 1998, 2001). Subsequently, ipsi- and contralateral column tissue samples were microdissected from the visual cortex and their total RNA was extracted. The corresponding cDNA was amplified and used to construct cDNA libraries of both ipsi- and contralateral OD columns. The duplicate recombinant phage clones from the cDNA libraries were then differentially screened with cDNA probes from ipsi- and contralateral OD columns. The initial screen of more than 8000 phage clones identified about 600 clones which were expressed at different levels in ipsi- and contralateral OD columns. The specificity of these clones was further examined by amplifying their cDNA inserts by PCR and subjecting the PCR products to Southern blots hybridized with the same cDNA probes from ipsi- and contralateral OD columns that were used in the differential screen (Fig. 1B). This sensitive confirmation analysis revealed a total number of 62 differentially expressed clones, 45 with a higher expression in ipsilateral and 17 with a higher expression in contralateral columns.
Clustered Expression of Hsp90α in the Cat Visual Cortex
We next used in situ hybridization with digoxigenin-labeled probes to examine the spatial expression of these clones in coronal sections from the developing cat visual cortex. Of the 62 differentially expressed cDNA clones, 4 showed a clustered expression pattern. One of them was analyzed in detail, since it showed a clustered expression in the deeper layers of the developing visual cortex resembling the OD column pattern. The full-length cDNA of this clone was isolated from a cDNA library of the P16 cat visual cortex containing larger inserts than used in the initial screen. Sequence analysis revealed that this gene encoded cat Hsp90α. We isolated 3 independent full-length cDNAs of cat Hsp90α and verified that they had identical sequences. The expression of Hsp90α in the kitten visual cortex was examined by northern blot analysis, which revealed expression of Hsp90α at the investigated age P16 (Fig. 1C). The size of the mRNA species (3 kb) corresponds closely to the size of the full length cDNA of cat Hsp90α.
To confirm the clustered expression seen in the initial results we repeated in situ hybridization with a Hsp90α probe on coronal sections of the visual cortex from 2 P16 kittens and indeed found clusters of positive cells in cortical layer 5 (Fig. 1D–F). These clusters have a size and spacing of about 400 μm. Hybridizations on tangential sections cut parallel to the cortical layering confirmed the clustered expression (Fig. 1G–J). Clusters of Hsp90α-positive cells were also present in area 18, where their spacing was generally wider (see Supplementary Fig. S1).
Hsp90α Clusters Coincide with Ipsilateral OD Columns
To test whether the clustered Hsp90α expression indeed coincides with the pattern of OD columns in the developing visual cortex, we visualized the OD pattern with optical imaging of intrinsic signals and subsequently carried out in situ hybridizations on tangential sections obtained from the same cortical region (Fig. 2 depicts one example of the alignment procedure in a P19 cat, see also the Materials and Methods section). The sections (Fig. 2D–F) were analyzed blindly and the positions of positive cells were plotted (Fig. 2G) and subsequently overlaid onto the imaged OD map (Fig. 2H) using the electrolytic lesions (Fig. 2, green arrowheads) as landmarks. The overlay (Fig. 2I) reveals a clear correlation between clusters of Hsp90α-positive cells and ipsilateral (dark) OD columns. Figure 3A–H shows a second example from a P16 cat, again demonstrating the clustered distribution of Hsp90α-positive cells and the alignment of these clusters with ipsilateral eye columns. In contrast, contralateral eye columns are largely devoid of Hsp90α-positive cells (Fig. 3H). Another example from a P18 cat is shown in Supplementary Fig. S2. In total, 6 experiments from both hemispheres, 3 from the right (Fig. 2) and 3 from the left (Fig. 3A–H; see Supplementary Fig. S2), were analyzed and showed similar results.
For all of our experiments, we determined the numbers of positive cells that came to lie in the ipsi- and contralateral OD domains, respectively. We found that the clear majority of cells (64%; range: 58–73%; P < 0.005 or smaller in all experiments; χ2 test; n = 6; best match is shown in Fig. 2 and worst one in Supplementary Fig. S2) was located in ipsilateral OD domains. In order to test to which degree this result was dependent on the exact delineation of the border, we performed the same calculation for 10 regions each containing the first, second, etc. decentile of the pixels in the OD map (Fig. 3I). For all 6 experiments, the plot of this analysis clearly shows that there is a steady increase in the number of Hsp90α-positive cells within the respective decentiles: from low numbers in the contralaterally dominated regions (1st to 5th decentile) to high numbers in ispilateral regions (5th to 10th decentile). As a further test, we quantified the match between the in situ hybridization and the OD map by calculating a 2-dimensional (2D) cross-correlation between the 2 patterns (see the Materials and Methods section). The 2D correlograms from all 6 experimental animals are depicted in Figure 3J. In all correlograms, a distinct local maximum came to lie close to the origin, indicating a good correlation between the 2 patterns. The additional maxima on the edges of some of the correlograms are expected, as they reflect the periodicity of the maps with a period of approximately 1 mm.
Mapping of OD columns with optical imaging is limited to a relatively small, exposed region of the visual cortex. In addition, the proper alignment between imaged functional maps and histological sections cut parallel to the imaging plane and tangential to the cortical surface is complicated by the strong curvature of the cortex. We therefore used a second approach to confirm the coincidence between Hsp90α clusters and ipsilateral OD columns. Induced expression of the immediate early gene Arc (Tagawa et al. 2005) and the neurotrophin BDNF (Lein and Shatz 2000; Tagawa et al. 2005) have been shown to reliably label OD columns in the cat visual cortex after a brief exposure of only eye to light. For this analysis, we used adjacent coronal rather than tangential sections, because they can be more accurately aligned, irrespective of the cortical curvature. Comparing the labeling pattern for Hsp90α and Arc in adjacent sections of the visual cortex ipsilateral to the activated eye reveals that clusters of Hsp90α- and Arc-positive cells coincide in the lower cortical layers (Fig. 4A–F; see also Supplementary Fig. S3). In contrast, in the hemisphere contralateral to the activated eye, the patterns interdigitate (Fig. 4G–L). These results 1) confirm that Hsp90α expression is limited to ipsilateral OD columns and 2) show that our observations are not confounded by rapid, activity-driven upregulation of Hsp90α expression. Similar results were obtained in both hemispheres of a second animal (see Supplementary Fig. S3).
Hsp90α Pattern is Independent from Eye-Specific Changes in Neuronal Activity
At the age of 2 weeks, when functional OD columns become first detectable in the cat, overall cortical responses are strongly dominated by the contralateral eye, while the ipsilateral eye provides a much weaker drive (Crair et al. 1998). One might thus argue that the clustered organization of Hsp90α-positive cells is not causally involved in the segregated termination of eye-specific inputs, but rather, assuming that expression of Hsp90α is activity dependent, results from spatially distinct cortical activity levels. In order to test this idea, we induced MD for 4 days at P14, thereby strongly reducing cortical activity in that eye's OD columns. At P18, the final day of MD, OD columns of the non-deprived eye were labeled by light-induced BDNF expression (Lein and Shatz 2000; Tagawa et al. 2005) and compared with the Hsp90α expression pattern on adjacent sections. As shown in Figure 5A–L, MD did not alter the coincidence of Hsp90α-positive cells with ipsilateral OD columns, irrespective of whether the contralateral (Fig. 5A–F) or the ipsilateral (Fig. 5G–L) eye was deprived. Longer MD of 8 days between P17 and P25, partially overlapping with the critical period, essentially yielded the same results (Fig. 5M–X). Taken together, these results strongly indicate that the differential Hsp90α expression is not caused by the spatial activity distributions generated by the inputs of both eyes in the visual cortex.
Hsp90α Pattern Precedes OD Column Formation
While the above data show a clear coincidence between the Hsp90α expression pattern and ipsilateral OD columns, they do not allow concluding that one is caused by the other. Therefore, we examined the in situ hybridization pattern of Hsp90α in the first 2 postnatal weeks, before OD columns in the cat visual cortex can be detected with functional (Crair et al. 1998, 2001) or anatomical (Crair et al. 2001) techniques (Fig. 6). In this series of experiments, we used coronal sections as shown in Figure 6A,B (see also the Materials and Methods section). At P1, when geniculocortical afferents are starting to form the first branches in their target layer 4 (Shatz and Luskin 1986; Ghosh and Shatz 1992), clusters of Hsp90α expression in cortical layer 5 are clearly apparent (Fig. 6C). A surface view generated from 74 successive coronal sections reveals an irregularly shaped, banded pattern of Hsp90α-expressing cells (Fig. 6D). Similarly, at P5 and P10 positive cells form patches or bands with a variable spacing between 500 and 1000 µm (Fig. 6A,B,E–H). We performed a 2D Fourier analysis to determine the prevalent spatial wavelength of these patterns of labeled cells and found broad peaks at 480, 714, and 636 µm for kittens of age P1, P5, and P10, respectively. These values are smaller than those measured for OD columns in four week old kittens (around 1100 µm; Keil et al. 2010); a difference which might be explained by the substantial increase in area of the visual cortex throughout this period (Duffy et al. 1998; see the Discussion section).
Thus, as early as P1, Hsp90α-expressing cells are clustered in the visual cortex. At this time point, the correlation with OD maps cannot be determined as they only form about 2 weeks later (Crair et al. 1998, 2001). Nevertheless, the structural similarity between the 2 patterns is very suggestive of one giving rise to the other. Interestingly, the periodic expression of Hsp90α continues at least until P28, the oldest age tested, and thus into the critical period for OD plasticity (Hubel and Wiesel 1970), although the number of Hsp90α-expressing cells increases with age (see Supplementary Fig. S4). We also used in situ hybridization to test for the expression of Hsp90α in areas outside the visual cortex. Again, we detected labeled cells in cortical layer 5, but there was no indication of a clustered distribution (see Supplementary Fig. S5; n = 3 cats).
Our results represent the first demonstration of a molecular difference between contra- and ipsilateral OD columns in the developing visual cortex. Using 2 different experimental approaches to visualize OD columns, we could show that the distinction is indeed between contra- and ipsilateral domains rather than left and right eye columns. We performed experiments in both hemispheres, and these data unequivocally show that around P16, when OD columns have just formed in the cat (Crair et al. 1998, 2001), Hsp90α is expressed predominantly in ipsi- but not contralateral OD columns. While the differential distribution of Hsp90α-positive cells is highly significant, not all of these cells are located in ipsilateral OD columns. This might be partially due to technical imperfections, for example, during the alignment procedure between the imaged OD maps and the histological sections containing the plotted cells (Fig. 2). One should note though that even in cortical layer 4, the segregation into ipsi- and contralateral OD columns is not strict in the cat visual cortex (Shatz and Stryker 1978) and one would therefore not expect to find a perfect match between these columns and the Hsp90α expression pattern.
In principle, the clustered Hsp90α expression at this age (P16) could be the consequence rather than the cause of differences between ipsi- and contralateral OD columns, for example, in overall neuronal activity levels (Crair et al. 1998), or other factors. Our MD experiments speak against a prominent role for neuronal activity, since the coincidence between Hsp90α clusters and ipsilateral OD columns was resistant to manipulations of activity levels in either eye. What further argues against the view that OD columns come first and then somehow drive the patchy Hsp90α expression is that the clusters are already apparent 2 weeks before eye-specific geniculocortical afferents form the anatomical basis for the OD column pattern (Crair et al. 1998, 2001).
While it is hard to obtain definitive proof for the continuity of the earliest Hsp90α clusters shortly after birth with the later ones, which align with ipsilateral OD columns, several observations support this view: 2-dimensional reconstructions of Hsp90α expression over large cortical regions (Fig. 6) show that the clusters clearly have a banded appearance, which is characteristic for OD columns in the cat visual cortex. Importantly, a recent quantitative study of the layout of OD columns in the developing cat visual cortex found that the “bandedness” of the columns is particularly strong in very young animals of 4 weeks of age (Keil et al. 2010). Our estimates of the cluster spacing in early postnatal animals (480, 714, and 636 µm at P1, P5, and P10, respectively) are smaller than the values reported for P28 kittens (around 1100 µm; Keil et al. 2010). This discrepancy might be explained by the expansion in area of the visual cortex, which is substantial during this period (Duffy et al. 1998). While the growth of the visual cortex at ages above P28 is not reflected in an increase in OD column spacing (Keil et al. 2010), this might well be different at younger ages. Lastly, it seems unlikely that there would be a highly organized spatial expression pattern in the early visual cortex not matching the later OD pattern, which would then in a matter of days reorganize into one perfectly correlating with OD columns. We thus conclude that the early Hsp90α clusters are tentative precursors of the later OD columns.
Heat shock proteins are molecular chaperones which regulate the folding, intracellular disposition, and turnover of many cellular proteins. Among the heat shock proteins, Hsp90α is unique, since most of its client proteins are involved in cellular signaling pathways such as protein kinases and transcription factors (for a review, see Picard 2002). Therefore, although Hsp90α might initially seem to be an unlikely candidate to be involved in determining the functional architecture of the visual cortex, it is located in a key position to modify proteins and thereby control their functional status. Axon guidance molecules for instance can be critically controlled by Hsp90α (for reviews, see Webber et al. 2002; Eustace and Jay 2004). In addition, Hsp90α might also contribute directly to the patterning of thalamic fiber ingrowth: it can be expressed at the cell surface (Thomaidou and Patsavoudi 1993) and has been shown to be involved in the migration of cerebellar neurons (Sidera et al. 2004). Moreover, a direct neurite-promoting activity of Hsp90α in the developing telencephalon has also been reported (Ishimoto et al. 1998).
We found Hsp90α to be differentially expressed in layer 5 of the developing visual cortex. Based on the distribution of labeled cells and on the size of their nuclei, we conclude that these cells are pyramidal neurons. Irrespective of the identity of the Hsp90α-expressing cells in layer 5, one may wonder how the differential expression of a molecular cue in cortical layer 5 could contribute to the segregation of ipsi- and contralateral eye inputs in layer 4, which is the main target of axons from the LGN. However, growing axons frequently navigate by using intermediate targets, which might attract and/or repel growth cones by substrate bound or diffusible molecules (Dickson and Zou 2010). Thalamocortical axons in particular have been shown to be guided and sorted by molecules of the ephrin family expressed in an intermediate target, the ventral telencephalon (Dufour et al. 2003; Seibt et al. 2003). Moreover, thalamocortical axons form temporary synapses with a population of transient neurons in the developing cortex, which also serve as an intermediate target, the subplate cells (Friauf et al. 1990; Herrmann et al. 1994). Thus, an interaction between thalamic axons and another type of neuron in layer 5 shortly before they reach their final targets in cortical layer 4 is not surprising. Nonetheless, at this point, direct proof for a causal role for Hsp90α in thalamic axon guidance and OD column formation in the cat visual cortex is missing and has to await a detailed analysis of both Hsp90α and other molecular players involved in this process.
Irrespective of whether and how Hsp90α might be involved in the patterning of eye-specific thalamocortical projections, the predominant expression of Hsp90α in one set of domains suggests that OD columns attain a molecular identity long before thalamic fibers start to segregate in the cortex. Do these data now prove that OD column formation is exclusively controlled by molecular factors? Certainly not. Activity-dependent factors might play an important role, and in fact, there are strong indications that they do, one particularly compelling example being the evidence that visually evoked activity influences the micropattern of OD columns (Adams and Horton 2002). A plausible scenario might be that their initial coarse pattern is defined by molecular cues, whereas subsequent fine-tuning is achieved by neuronal activity. This developmental sequence of events would be analogous to the steps known to be involved in the formation of retinotopic maps in the visual system, where molecular cues are responsible for the initial formation of a retinotopic map (Cheng et al. 1995; Drescher et al. 1995) which is later refined by activity-dependent mechanisms (Grubb et al. 2003; McLaughlin et al. 2003; Mrsic-Flogel et al. 2005; Cang, Niell et al. 2008; Cang, Wang et al. 2008).
This work was supported by the Max Planck Society, a postgraduate fellowship from the Japan Society for the Promotion of Science (K.T.) and the European Molecular Biology Organization (K.T.).
We thank J. Sanes for comments on an earlier version of the manuscript, Y. Tanabe, Y. Tagawa, H. Gotoh, H. Hirata, and T. Kobayashi for discussions, and F. Voss for help with data analysis. Conflict of Interest: None declared.