Understanding the molecular mechanisms underlying the formation of selective intracortical circuitry is one of the important questions in neuroscience research. “Barrel nets” are recently identified intracortical axonal trajectories derived from layer 2/3 neurons in layer 4 of the primary somatosensory (barrel) cortex. Axons of layer 2/3 neurons are preferentially distributed in the septal regions of layer 4 of the barrel cortex, where they show whisker-related patterns. Because cadherins have been viewed as potential candidates that mediate the formation of selective neuronal circuits, here we examined the role of cadherins in the formation of barrel nets. We disrupted the function of cadherins by expressing dominant-negative cadherin (dn-cadherin) using in utero electroporation and found that barrel nets were severely disrupted. Confocal microscopic analysis revealed that expression of dn-cadherin reduced the density of axons in septal regions in layer 4 of the barrel cortex. We also found that cadherins were important for the formation, rather than the maintenance, of barrel nets. Our results uncover an important role of cadherins in the formation of local intracortical circuitry in the neocortex.
The cerebral cortex contains complex and sophisticated neuronal circuitries that are the structural bases of higher brain functions. Because the formation of appropriate neuronal circuitry in the cerebral cortex during development is essential for normal brain functions, it is important to elucidate the mechanisms underlying the formation of neuronal circuitry in the cerebral cortex. Although previous pioneering studies have intensively investigated the electrophysiological and anatomical properties of local neuronal circuitry in the cerebral cortex, the mechanisms underlying the formation of local neuronal circuitry in the cerebral cortex are not fully understood.
The primary somatosensory cortex (barrel cortex, S1) has been widely used for investigating neural circuits and the mechanisms underlying their formation (Woolsey 1990; O'Leary et al. 1994; Erzurumlu and Kind 2001; López-Bendito and Molnár 2003; Rebsam and Gaspar 2006; Fox 2008; Toda et al. 2008; Meyer et al. 2010a,b; Erzurumlu and Gaspar 2012; Oberlaender et al. 2012; Toda et al. 2013). Using in utero electroporation as a tool to visualize axonal trajectories of a selective population of cortical neurons, we recently uncovered a novel intracortical circuit, which we named “barrel nets,” in layer 4 of the barrel cortex in mice and rats; axons of layer 2/3 neurons are predominantly located in the septal regions of layer 4 and therefore show a whisker-related pattern (Sehara et al. 2010; Sehara and Kawasaki 2011; Sehara et al. 2012). Our results showing presynaptic structures on barrel nets suggested that barrel nets are involved in the functioning of the neuronal circuitry of the barrel cortex (Sehara et al. 2010; Sehara et al. 2012).
To investigate the mechanisms underlying the formation of intracortical neural circuits, we focused on classic cadherins, which are known as homophilic cell adhesion molecules (Takeichi 1991; Redies and Takeichi 1996; Tepass et al. 2000; Takeichi 2007). This was because it has been proposed that the homophilic binding activity of classic cadherins is well suited to mediate selective circuit formation (Suzuki et al. 1997; Inoue et al. 1998). Furthermore, it has been reported that several classic cadherins are predominantly expressed in whisker-related patterns in the barrel cortex. Cadherin-8 mRNA was detected in septal regions of layer 4, as well as layers 2/3 and 5a (Suzuki et al. 1997; Gil et al. 2002; Dye et al. 2011; Hertel and Redies 2011; Krishna-K et al. 2011; Lefkovics et al. 2012). Previous studies showed that, in the developing barrel cortex, cadherin-6 expression is relatively high in septal regions compared with barrel hollows (Suzuki et al. 1997; Inoue et al. 2008; Krishna-K et al. 2011; Lefkovics et al. 2012; Terakawa et al. 2013). In addition, immunohistochemical studies revealed that R-cadherin and N-cadherin expressions are observed in barrel hollows of layer 4 (Huntley and Benson 1999; Obst-Pernberg et al. 2001). However, the roles of classic cadherins in selective circuit formation in the cerebral cortex still remained largely elusive. Here, using barrel nets as a model, we uncovered a role of classic cadherins in intracortical circuit formation.
Materials and Methods
ICR mice were purchased from SLC (Hamamatsu) and were reared on a normal 12-h light/dark schedule. The day of conception and that of birth were counted as E0 and P0, respectively. All procedures were performed in accordance with protocols approved by the University of Tokyo Animal Care Committee and Kanazawa University Animal Care Committee. All experiments were repeated at least 3 times and gave consistent results.
All genes were expressed under the control of the CAG promoter (Niwa et al. 1991). Plasmids were purified using the EndoFree plasmid maxi kit (QIAGEN). pCAG-GFP, pCAG-mCherry, and pCAG-synaptophysin-GFP were described previously (Sehara et al. 2010). Dominant-negative cadherins, hDNcad and cN390Δ, were generous gifts from Dr. Jonas Frisén (Karolinska Institutet) and Dr. Masatoshi Takeichi (RIKEN CDB), respectively (Fujimori and Takeichi 1993; Nieman et al. 1999; Togashi et al. 2002; Barnabé-Heider et al. 2008). pCAG-hDNcad and pCAG-cN390Δ were made by substituting GFP of pCAG-GFP with hDNcad and cN390Δ, respectively. pCAG-synaptophysin-mCherry was made by substituting GFP of pCAG-synaptophysin-GFP with mCherry. pCAG-ERT2CreERT2 and pCAG-floxedSTOP-GFP were described previously (Ako et al. 2011). pCAG-floxedSTOP-mCherry, pCAG-floxedSTOP-hDNcad, and pCAG-floxedSTOP-cN390Δ were made by replacing GFP of pCAG-floxedSTOP-GFP with mCherry, hDNcad, and cN390Δ, respectively.
In Utero Electroporation
In utero electroporation was performed as described previously with slight modifications (Sehara et al. 2010; Kawasaki et al. 2012; Kawasaki et al. 2013). Briefly, pregnant animals were anesthetized with sodium pentobarbital, and the uterine horns were exposed. Approximately 1–2 μl of DNA solution (1–3 mg/mL) was injected into the lateral ventricle of embryos at the indicated ages using a pulled glass micropipette. Each embryo within the uterus was placed between tweezer-type electrodes (CUY650-P5, NEPA Gene). Square electric pulses (45 V, 50 ms) were passed 5 times at 1-s intervals using an electroporator (ECM830, Harvard Apparatus). The directions of pulses were adjusted so that cortical glutamatergic neurons were transfected. Care was taken to quickly place embryos back into the abdominal cavity to avoid excessive temperature loss. The wall and skin of the abdominal cavity were sutured, and embryos were allowed to develop normally.
4-Hydroxytamoxifen (4-OHT) (>70% Z isomer, Sigma–Aldrich) administration was performed as described previously with slight modifications (Ako et al. 2011). 4-OHT was dissolved in 99.5% EtOH at a concentration of 20 mg/mL as stock solution. The stock solution was diluted with sesame oil at a concentration of 2 mg/mL before use, and then, the EtOH was vaporized. 4-OHT in sesame oil was injected subcutaneously 66 μg/g of body weight at the indicated time points.
Dissociated Primary Cultures of Cortical Neurons
Cultures of dissociated cortical neurons were made as described previously with slight modifications (Yamasaki et al. 2011). In utero electroporation was performed at E15.5, and the cerebral cortex was dissected at E16.5. After the meninges were pulled off, the dissected cortices were incubated for 20 min in papain solution plus DNase I at 37°C and were dissociated mechanically. The dissociated neurons were plated onto poly-D-lysine-coated coverslips in 24-well plate (1 × 106 cells/well) and maintained in 50% Basal Medium Eagle (Invitrogen) containing 1 mml-glutamine (Sigma–Aldrich), 25% Hank's balanced salt solution (Invitrogen), 25% horse serum (Invitrogen), 6.5 mg/mL D-glucose (Sigma–Aldrich), 100 U/mL penicillin/streptomycin (Invitrogen), and 1% HEPES pH 7.4 (Invitrogen) at 37°C in 5% CO2. The day of dissection was counted as day in vitro 0 (DIV0).
Immunohistochemistry and Immunocytochemistry
Immunohistochemistry was performed as described previously (Toda et al. 2008; Hayakawa and Kawasaki 2010; Iwai et al. 2013). Briefly, mice were anesthetized with pentobarbital and transcardially perfused with 4% paraformaldehyde/PBS. To make coronal sections, the cerebral cortices were dissected, cryoprotected by overnight immersion in 30% sucrose/PBS, and embedded in OCT compound. To make tangential sections, the cortical hemispheres were isolated, flattened, and embedded in OCT compound as described previously (Toda et al. 2008; Sehara et al. 2010). Sections were made using a cryostat.
Sections of 50-μm thickness were permeabilized with 0.1–0.5% Triton X-100 in PBS and incubated overnight with primary antibodies. After being incubated with Alexa 488-, Alexa 647-, or Cy3-conjugated secondary antibodies and 1 μg/mL Hoechst 33342, the sections were washed and mounted. The primary antibodies included rabbit anti-VGLUT2 antibody (Synaptic Systems), goat anti-VGLUT2 antibody (Frontier Institute), rabbit anti-GFP antibody (Medical & Biological Laboratories), rat anti-GFP antibody (Nacalai Tesque), rabbit anti-RFP antibody (Medical & Biological Laboratories), rabbit anti-RORβ antibody (Diagenode), goat anti-Brn2 antibody (Santa Cruz), and rat anti-Ctip2 antibody (Abcam).
For immunocytochemistry, dissociated neurons cultured on coverslips were fixed with 4% paraformaldehyde/PBS for 15 min at 37°C. Then, the neurons were incubated in blocking solution (2% BSA/PBS) and were incubated at 4°C overnight with primary antibodies, which included rat anti-GFP antibody (Nacalai Tesque) and sheep anti-N-cadherin antibody (R&D Systems). The neurons were washed with 0.5% Triton X-100/PBS and then incubated with Alexa488- and Cy3-conjugated secondary antibodies and 1 μg/mL Hoechst 33342, followed by washing and mounting.
In Situ Hybridization
In situ hybridization was performed as described previously with slight modifications (Kawasaki et al. 2004; Iwai and Kawasaki 2009). Sections with 25-μm thickness were prepared from fresh-frozen tissues, and those with 14-μm thickness were from fixed tissues. The sections were treated with 4% paraformaldehyde for 10 min and 0.25% acetic anhydride for 10 min. After prehybridization, the sections were incubated overnight at 58°C with digoxigenin-labeled RNA probes for cadherin-6, cadherin-8, or hDNcad diluted in hybridization buffer (50% formamide, 5× SSC, 5× Denhardt's solution, 0.3 mg/mL yeast RNA, 0.1 mg/mL herring sperm DNA, and 1 mm DTT). The sections were washed with SSC, incubated with alkaline phosphatase-conjugated anti-digoxigenin antibody (Roche), and then incubated with NBT/BCIP as substrates.
Epifluorescence microscopy was performed with an AxioImager A1 microscope (Carl Zeiss) or a BIOREVO BZ-9000 (Keyence). Confocal microscopy was performed with an LSM510 microscope (Carl Zeiss) or an FV10i FLUOVIEW microscope (OLYMPUS).
Quantifications of Barrel Net Formation
Quantifying GFP fluorescence of barrel nets was performed as described previously (Sehara et al. 2010). Briefly, coronal sections of 50-μm thickness were prepared at the indicated time points and stained using VGLUT2 antibody. Images were taken under a confocal microscope. Signals in layer 4 were measured using the “Plot Profile” command of ImageJ (NIH). Each barrel unit (which refers to a barrel and its septal rim) was then separated using VGLUT2 signal intensities in layer 4. GFP signals within each barrel unit were normalized by subtracting the minimum value within the barrel unit from the measured values, and dividing the values by the total signal intensity of the barrel unit. Each barrel unit was divided into ten subunits of equal width (hereafter referred to as barrel subunits), and the values of each subunit were plotted against the horizontal position of the subunit within the barrel unit.
To calculate the periphery/center ratio (p/c ratio) for GFP, the mean signal intensity of subunits 1 and 10 was divided by that of subunits 5 and 6. The center/periphery ratio (c/p ratio) of VGLUT2 was the inverse number of the p/c ratio. For statistical analyses of the p/c and c/p ratio, the average value was calculated for each pup and compared using Welch's t-test.
Quantification of the Morphology of Dissociated Neurons
Dissociated primary cultures were fixed at DIV4 and were stained with anti-GFP antibody. Images were taken under an epifluorescence microscope, and axons from individual neurons were traced using the NeuronJ plugin (Meijering et al. 2004; Meijering 2010) of ImageJ software. We used the “Add tracings” command for tracing the axons and the “Measure tracings” command for measuring the length of axon branches in NeuronJ. Branch patterns were annotated manually using NeuronJ, and then, the analysis and statistics were performed using the R environment (R Core Team 2014). Five GFP-positive neurons on each coverslip were traced and analyzed. Seven coverslips per each experimental condition were used for statistical analyses using Welch's t-test.
Quantifications of the Expression of Layer Markers and the Density of GFP-Positive Neurons
Z-stack images of immunostained sections were taken using a confocal microscope, and the expression of layer markers was examined in GFP-positive neurons with Hoechst-positive nuclei. GFP-positive neurons without Hoechst-positive nuclei were excluded from the analyses.
To determine the transfection efficiency, the density of GFP-positive neurons in layer 2/3 of coronal sections with 50-μm thickness was calculated. Statistical analyses were performed using Mann–Whitney's U-test.
Quantifications of the Density and the Size of Synaptophysin-mCherry-Positive Puncta
Coronal sections prepared at P15 were immunostained, and Z-stack confocal images were taken. Average-projection images with 3.5-μm optical thickness were smoothened (mean filter, 2-μm radius) and then used for further analyses. Synaptophysin-mCherry-positive puncta were detected using the “Find maxima” command of ImageJ, in which mean plus standard deviation values were used as the tolerance level. GFP-positive axons were traced, and their lengths were measured using the original images and NeuronJ. Then, mCherry-positive puncta were counted only if they overlapped with GFP-positive axons. The radius of mCherry-positive puncta was calculated as the distance from the maxima to the first deflection point of the mCherry fluorescence. Mann–Whitney's U-test was used for statistical analyses.
ERT2Cre-Mediated Postnatal Expression of Dominant-Negative Cadherin in Layer 2/3 Neurons
Because it has been reported that the multiple classic cadherins are expressed in whisker-related patterns in the somatosensory cortex (see Supplementary Fig. 1) (Huntley and Benson 1999; Obst-Pernberg et al. 2001; Huntley 2002; Dye et al. 2011; Krishna-K et al. 2011), we simultaneously suppressed functions of multiple classic cadherins using dominant-negative cadherin (dn-cadherin). Consistent with a previous report (Nieman et al. 1999), our experiments using dissociated primary cultures of the cerebral cortex demonstrated that the dn-cadherin hDNcad (Nieman et al. 1999; Barnabé-Heider et al. 2008) strongly inhibited the expression of classic cadherins in cortical layer 2/3 neurons (Supplementary Fig. 2). We then co-expressed GFP and hDNcad in layer 2/3 neurons in vivo using in utero electroporation at E15.5 under the control of the CAG promoter. Consistent with a previous report (Kawauchi et al. 2010), however, constitutive expression of hDNcad disrupted migration of the GFP-positive transfected layer 2/3 neurons (Fig. 1A).
To avoid the migration defects caused by dn-cadherin, we used the ERT2Cre/loxP system to control the timing of dn-cadherin expression (Fig. 1B) (Matsuda and Cepko 2007). We transfected layer 2/3 neurons with pCAG-ERT2CreERT2, pCAG-floxedSTOP-XFP, and pCAG-floxedSTOP-hDNcad using in utero electroporation at E15.5 and examined the time point after which 4-OHT injection does not lead to the migration defect (Fig. 1B,C). While injection of 4-OHT at P0 caused the evident migration defect (Fig. 1C, arrowheads), injection at or after P2 did not (Fig. 1C). We therefore injected 4-OHT at P2 in the following experiments.
Next, we examined how many times of 4-OHT injection are necessary for sufficient ERT2Cre-mediated gene expression (Fig. 2). Expression levels of GFP and the number of GFP-positive cells were markedly increased in samples that received 4-OHT injection daily for 3 consecutive days than in those that received injection only once (Fig. 2B). Therefore, to sufficiently express dn-cadherin only after migration of layer 2/3 neurons had occurred, we injected 4-OHT daily from P2 to P4 in the following experiments.
Postnatal Expression of dn-cadherin in Layer 2/3 Neurons Disrupted Barrel Nets in the Barrel Cortex
We then examined whether classic cadherins are required for the formation of barrel nets by expressing dn-cadherin and GFP in layer 2/3 neurons (Fig. 3A). Although the whisker-related pattern of barrel nets was clearly visible in tangential sections of control samples, we found that the pattern was almost absent in samples expressing hDNcad (Fig. 3B). To make sure that this effect of hDNcad is mediated by suppressing cadherin functions, we employed another dn-cadherin, cN390Δ (Fujimori and Takeichi 1993; Togashi et al. 2002). Consistently, we found that the expression of cN390Δ also resulted in disruption of barrel nets (Fig. 3C). Examination using coronal sections also showed that GFP-positive axons failed to accumulate in septal regions of layer 4 when one of the dn-cadherin constructs was expressed in layer 2/3 (Fig. 4A, hDNcad and cN390Δ). These results suggest that inhibition of classic cadherin functions in layer 2/3 neurons disrupts the formation of the barrel net pattern.
We then quantified the effect of dn-cadherin on barrel net formation with a method we used previously (Sehara et al. 2010). We quantified the distribution of GFP-positive axons around individual barrels in coronal sections using confocal microscopic images (Supplementary Fig. 3). In control samples, GFP-positive axons were abundant in peripheral regions around thalamocortical axon patches visualized by anti-VGLUT2 antibody (Supplementary Fig. 3, control; compare green and gray) (Fremeau et al. 2001; Fujiyama et al. 2001). In contrast, in dn-cadherin samples, such preferential distribution of GFP-positive axons was inhibited (Supplementary Fig. 3, hDNcad, green). We calculated the ratio of GFP signal intensities in peripheral regions to those in central regions of individual VGLUT2-positive patches (hereafter referred to as “the periphery/center ratio” or “the p/c ratio,” see Materials and methods for details) (Fig. 4B, green). We found that the p/c ratio of GFP was significantly higher in control samples than that in dn-cadherin samples (Fig. 4B, green), whereas the c/p ratio of VGLUT2 signals was similar in both control and dn-cadherin samples (n = 3 pups for each condition from 3 pregnant mice). However, the density of GFP-positive layer 2/3 neurons in hDNcad samples was not different from that in control samples (Fig. 4D) (n = 3 pups for each condition from 3 pregnant mice). Taken together, these results indicate that classic cadherins play a crucial role in the formation of barrel nets.
Classic Cadherins are Required for the Accumulation of Fine Axons in Septal Regions
Because our previous confocal microscopic studies showed that there is a much higher density of GFP-positive fine axon branches in septa than in barrel hollows (Sehara et al. 2010; Sehara et al. 2012), it seemed plausible that the effect of dn-cadherin on barrel nets resulted from the reduction of fine axons in the septal regions of layer 4. We therefore examined the phenotype of dn-cadherin overexpression in more detail using a confocal microscope (Fig. 4C). We found that the number of fine axon branches markedly decreased in septal regions when hDNcad was expressed (Fig. 4C). Although hDNcad samples may have more axons in the barrel hollows, it was difficult to conclude whether hDNcad affected the number of axons in the barrel hollows. These results indicate that classic cadherins are required for the accumulation of fine axons in the septal regions of layer 4.
Because previous studies reported that classic cadherins are involved in axonal elongation of retinal ganglion cells (Riehl et al. 1996), it was possible that the effect of dn-cadherin on barrel nets was just a result of a general defect in the axonal elongation of transfected layer 2/3 neurons. In order to address whether dn-cadherin expression affects axon elongation in general, we examined the extent of callosal axon projection from layer 2/3 neurons (Fig. 5). Notably, layer 2/3 neurons, regardless of whether they were transfected with hDNcad or control vectors, sent callosal axons to a similar extent in the contralateral cortex, suggesting that the defect in the formation of barrel nets in dn-cadherin samples is not simply because of the inhibition of axonal elongation in general.
It seemed intriguing to examine the effects of dn-cadherin on the morphology of the axons of layer 2/3 neurons at the single-cell level. We therefore transfected layer 2/3 neurons with pCAG-GFP plus either pCAG-hDNcad or a control plasmid and made dissociated primary cultures of the cerebral cortex (Fig. 6). Although we quantified several parameters of the morphology of axons such as the number of branches per an axon (Fig. 6B), the total length of each axon (Fig. 6C), and the length of individual axonal branches (Fig. 6D,E), none of these parameters were significantly different between control and hDNcad-transfected neurons (n = 7 coverslips for each condition). These results imply that dn-cadherin affects the location of axonal branches but not the branching or elongation of axons.
Although dn-cadherin-expressing layer 2/3 neurons were located in the appropriate layer (i.e., layer 2/3), it seemed still possible that the expression of dn-cadherin affected the layer identities of transfected layer 2/3 neurons, and as a result, the pattern of barrel nets was disrupted. To address this point, we performed immunohistochemistry using antibodies against layer marker proteins (Fig. 7). We used Brn2, which is predominantly expressed in layers 2/3 and 5, and Ctip2, which is a marker for layers 5–6 (Hevner et al. 2003; Arlotta et al. 2005). We also used RORβ, which is mainly expressed in layer 4 and sparsely in layer 5 (Schaeren-Wiemers et al. 1997; Nakagawa and O'Leary 2003). Higher magnification images showed that Brn2, RORβ, and Ctip2 immunoreactivities were located in the nucleus (Fig. 7D,E). We found that expression patterns of the layer markers in hDNcad samples were indistinguishable from those in the control samples (Fig. 7A–C). In addition, our quantification also revealed that GFP-positive cells in the hDNcad samples had similar expression profiles to those in control samples (Fig. 7C; n = 3 pups for each condition from 5 pregnant mice). These data suggest that postnatal overexpression of dn-cadherin does not affect the layer identities of transfected layer 2/3 neurons but specifically alters the regional selectivity of axon distribution in layer 4 of the barrel cortex.
Postnatal dn-cadherin Expression Decreased Presynaptic Structures in Septal Regions
Because our previous studies have suggested that barrel nets contain presynaptic structures (Sehara et al. 2010; Sehara et al. 2012), we next examined whether the expression of dn-cadherin affects the number of synaptic structures on barrel nets. We examined the distribution of presynaptic structures on barrel nets, using XFP-tagged synaptophysin as a presynaptic marker (Sehara et al. 2010; Sehara et al. 2012).
First, we examined the developmental time course of the formation of presynaptic structures on barrel nets. We expressed mCherry and synaptophysin-GFP in layer 2/3 neurons and examined the distribution patterns of GFP-positive puncta on mCherry-positive barrel nets during development (Fig. 8A). Synaptophysin-GFP-positive puncta markedly increased between P5 and P15, corresponding to the increase of mCherry-positive axons in septal regions (Fig. 8A). We found that the number of synaptophysin-mCherry puncta was markedly reduced by hDNcad, suggesting that dn-cadherin expression affects not only the distribution of axons but also the number of presynaptic structures on barrel nets (Fig. 8B). Our quantification of mCherry-positive puncta on GFP-positive axons showed that neither their density (i.e., the number of mCherry-positive puncta per unit axon length) nor the size of individual mCherry-positive puncta was affected by hDNcad expression (Fig. 8C–E; n = 3 pups for each condition from 4 pregnant mice). This result suggests that although the number of mCherry-positive puncta was reduced by hDNcad, this reduction is due to the decrease in the number of axons in the septal region.
Classic Cadherins are Required for the Formation, but not for the Maintenance, of Barrel Nets
So far, our results were consistent with the idea that classic cadherins are required for the formation of barrel nets. On the other hand, it still remained possible that classic cadherins are involved in the maintenance, rather than the formation, of the pattern of barrel nets. However, we believe that this is unlikely because of the following 2 reasons. The first reason is that barrel nets did not appear from the beginning when dn-cadherin was expressed. We examined the effect of hDNcad sequentially at P5, P10, and P15 (Fig. 9A). In control animals, barrel nets were invisible at P5 and were formed gradually thereafter as we reported previously (Sehara et al. 2010). Interestingly, in hDNcad-transfected animals, barrel nets never appeared (Fig. 9B), suggesting that the formation of barrel nets was inhibited.
The second reason is the fact that expression of dn-cadherin after the formation of barrel nets did not disrupt barrel net patterns. We expressed hDNcad after the formation of barrel nets by late administration of 4-OHT and examined whether the late expression of dn-cadherin results in the degradation of barrel nets (Fig. 9C). Interestingly, the expression of hDNcad after the formation of barrel nets did not disrupt barrel nets (Fig. 9D, late 4-OHT), whereas hDNcad expression before the formation of barrel nets effectively disrupted barrel nets (Fig. 9D, early 4-OHT). Our quantification using confocal microscopic images also supported this conclusion (early 4-OHT, control, n = 3 pups; early 4-OHT, hDNcad, n = 3 pups; late 4-OHT, control, n = 4 pups; late 4-OHT, hDNcad, n = 3 pups; from 4 pregnant mice) (Fig. 9E). Because it seemed possible that late 4-OHT treatment failed to induce hDNcad, we examined whether hDNcad expression was indeed induced by late 4-OHT treatment by using in situ hybridization. We made an in situ hybridization probe that detects hDNcad but does not detect endogenous mouse cadherins in the barrel cortex (Supplementary Fig. 4A,B). Our late 4-OHT treatment led to strong hDNcad expression, as in the case of early 4-OHT treatment (Supplementary Fig. 4B). Collectively, our results suggest that classic cadherins are required for the formation, rather than the maintenance, of barrel nets.
Here, we have shown that classic cadherins are required for the formation, rather than the maintenance, of local circuits in the cerebral cortex during development. Elongation of callosal axons derived from layer 2/3 neurons was not affected by dn-cadherin, suggesting that dn-cadherin selectively inhibited barrel net formation. Our results suggest that classic cadherins are also important for the development of presynaptic structures on barrel net axons. Our findings uncovered an important role of classic cadherins in local intracortical circuit formation.
Because previous reports showed that multiple classic cadherins are expressed in whisker-related patterns (Huntley and Benson 1999; Obst-Pernberg et al. 2001; Huntley 2002; Dye et al. 2011; Krishna-K et al. 2011), we utilized dn-cadherin to suppress these classic cadherins simultaneously and uncovered that classic cadherins are indeed required for barrel net formation. In order to fully understand the cadherin-mediated mechanisms underlying barrel net formation, there are several important questions to be addressed. First, it would be important to investigate which of the classic cadherins mediate barrel net formation. Combining cadherin-knockout mice and in utero electroporation to reveal barrel nets would be appropriate experiments to address this point. Second, because classic cadherins mediate homophilic binding, it seems possible that cadherin molecules on barrel net axons bind to cadherin molecules expressed on other cells in the septa. It would be intriguing to investigate which neurons in which layers are responsible for this. Alternatively, it seems also possible that cadherins on axons in barrel nets bind to cadherins on neighboring axons in the same barrel net. Third, it would be intriguing to examine the effect of dn-cadherin at the level of single axons. When GFP is expressed with conventional in utero electroporation, numerous axons in barrel nets are labeled with GFP, and as a result, it is difficult to examine what happens to individual GFP-positive axons, such as branching, elimination, and elongation. Our experiments using dissociated primary cultures showed that axon elongation and branching were not affected by dn-cadherin in vitro. Therefore, it seems reasonable to speculate that GFP-positive axons in septal regions were shifted into barrel hollows by hDNcad expression, although this should be addressed by future detailed morphological analyses of single neurons in vivo. Future experiments would be required for obtaining a complete understanding of the mechanisms underlying barrel net formation.
Because of their homophilic binding activity and intriguing expression patterns, classic cadherins have been proposed to be key molecules in the regulation of selective neuronal circuit formation (Suzuki et al. 1997; Inoue et al. 1998). A recent report showed that cadherin-6 mediates axon-target matching of retinal ganglion cell axons in the visual thalamus (Osterhout et al. 2011). Another report demonstrated that cadherin-9 regulates synapse-specific differentiation in the hippocampus (Williams et al. 2011). However, involvement of classic cadherins in selective local circuit formation in the neocortex has not been well understood. Our results indicate that classic cadherins indeed mediate selective local circuit formation in the neocortex. It remains unclear how cadherins actually regulate selective circuit formation in barrel nets. As discussed earlier, it seems likely that the homophilic binding activity of cadherins is involved in axon guidance. On the other hand, it seems also possible that cadherins primarily regulate synaptogenesis to stabilize the distribution of axons because previous studies showed that cadherins are important for synaptogenesis (Togashi et al. 2002; Bozdagi et al. 2004; Williams et al. 2011). However, we think that the latter is unlikely for barrel nets because we showed that the density of synapses along individual axons was not affected by expressing dn-cadherin. It would be intriguing to examine the differential roles of cadherins in axon formation and synaptogenesis.
Uncovering the mechanisms underlying the formation of local intracortical circuits would lead to a better understanding of the development and function of the cerebral cortex. Although we have uncovered anatomical structures and developmental processes of barrel nets (Sehara et al. 2010; Sehara et al. 2012), the function of barrel nets still remains unknown. It seems appropriate to speculate that the function of barrel nets is suppressed in dn-cadherin-expressing animals because of the severe reduction in the number of presynaptic structures by dn-cadherin. Further understanding of the mechanisms underlying barrel net formation would lead to selective loss-of-function experiments involving barrel nets such as the generation of animals without barrel nets, which would reveal their roles in the functioning of the cerebral cortex. Barrel nets seem to be useful for addressing important questions about intracortical local circuits.
This work was supported by Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology-Japan (25290013, 25640032, 25123702, and 26111708); Takeda Science Foundation; Takeda Medical Research Foundation; Astellas Foundation for Research on Metabolic Disorders; the Kurata Memorial Hitachi Science and Technology Foundation; Mitsubishi Foundation; and Research Foundation for Opto-Science and Technology. This work was also supported by a Research Fellowship for Young Scientists from Japan Society for the Promotion of Science (to KS).
We thank Drs. M. Takeichi (RIKEN CDB) and J. Frisén (Karolinska Institutet) for plasmids. We are grateful to Drs. S. Tsuji, T. Kadowaki, H. Bito (The University of Tokyo), E. Nishida (Kyoto University), Y. Sasai (RIKEN CDB), and S. Nakanishi (Osaka Bioscience Institute) for their continuous support and warm encouragement. We express our gratitude to Drs. T. Sakurai and M. Sato (Kanazawa University) for their technical support. We thank Kawasaki lab members and Z. Blalock for critical discussions and comments on this manuscript, and K. Tanno for technical assistance. Conflict of Interest: None declared.