Previous work demonstrated that members of the semaphorin family, Sema3A and Sema3C, act as repulsive and attractive guidance signals, respectively, for cortical axons. During the development of corticofugal projections, these semaphorins are expressed in adjacent cortical zones, but there is a considerable overlap between Sema3A and Sema3C expression in the subventricular zone. We used different in vitro assays to examine the response of cortical axons exposed to defined mixtures of these opposing guidance cues. Results showed that even at very low concentrations, Sema3A overrides the effects of Sema3C. Moreover, experiments with function-blocking antibodies directed against neuropilin provided insights into how cortical axons integrate disparate guidance signals at the receptor level. These in vitro data suggest that the pathway of corticofugal axons is defined by an attractive cue in the intermediate zone, where Sema3C is expressed alone. To directly test this hypothesis in vivo, we performed axon-tracing experiments in Sema3C-deficient mice. Compared with wild-type animals, corticofugal axons take a more superficial route in Sema3C−/− mice, and the corticofugal pathway is more compacted. This phenotype is expected when an attractive cue for cortical axons, Sema3C, is eliminated and a repulsive cue, Sema3A, becomes predominant.
During the development of the nervous system, axons grow long distances in a complex environment to innervate specific target regions. Before reaching their final destination, axons often first navigate to intermediate targets and make selective pathway choices by responding to the combined action of attractive and repulsive guidance cues (reviewed in Tessier-Lavigne and Goodman 1996; O'Donnell et al. 2009; Kolodkin and Tessier-Lavigne 2011). Several families of brain wiring molecules, including netrins, ephrins, slits, and semaphorins, have been implicated in growth cone guidance. Initially, these wiring molecules have been classified as either attractive (e.g., netrins) or repulsive (e.g., ephrins and semaphorins). However, there is now ample evidence that the same wiring molecules, depending on the biological context, can have both attractive and repulsive effects (reviewed in Bolz and Castellani 1997; Song and Poo 2001; Bolz et al. 2004). This bifunctionality can be attributed to different mechanisms, for instance the composition of the receptor complex, the spatial distribution of the guidance signal, or different levels of second messengers in the growth cone.
In most previous studies, different in vitro assays have been used to examine how growth cones respond to individual guidance cues. However, much less is known about how axons react to a combination of guidance signals, which more closely resembles the in vivo situation where navigating axons are simultaneously exposed to multiple cues. For example, spinal commissural axons first project to the floor plate, an intermediate target, before they cross the ventral midline and project alongside it. In initial experiments, it was discovered that commissural axons are attracted by netrin-1 released from floor plate cells (Kennedy et al. 1994; Serafini et al. 1996), but more recent work revealed that other factors, such as sonic hedgehog (Shh) and stem cell factor, also produced by floor plate cells, collaborate with netrin-1 in midline attraction (Charron et al. 2003; Gore et al. 2008). Upon reaching the midline, a number of repulsive signals, including slits, semaphorins, and ephrins, allow commissural axons to exit the midline and then never recross (Zou et al. 2000; Brittis et al. 2002; Long et al. 2004; Kadison et al. 2006; Nawabi et al. 2010). This cooperation of guidance signals of the same sign (attractive or repulsive) could reflect redundancy, but it might be also necessary for the fine tuning in the assembly of neuronal circuits. For example, an attractive cue could elicit outgrowth-promoting effects and/or instructive chemoattractive effects; diverse cellular processes that can be differentially controlled by distinct “attractive” signals.
But what happens when growth cones encounter opposing guidance cues? This seems to be a common theme during nervous system development, when axons are attracted to successive intermediate targets and then—after reaching one of these midway regions—do not stall but rather move away toward the next target. There are several possible solutions to this apparent paradox. It has been suggested that after reaching an intermediate target, growth cones loose their responsiveness to chemoattractants released from this region and simultaneously upregulate responsiveness to chemorepellents (reviewed in Dickson and Zou 2010). For commissural axons, it has been shown that they loose attraction and gain repulsion once they reach the midline by the interaction and dynamic expression of guidance receptors on the cell surface. There is convincing evidence that this switch in responsiveness is under the tight control of signals emanating from the floor plate (Stein and Tessier-Lavigne 2001; Nawabi et al. 2010). A variety of cellular mechanisms control the expression and activity of guidance receptors during midline crossing, including protein trafficking, local protein synthesis and degradation, formation of receptor complexes, and alternative splicing of guidance receptors (reviewed in O'Donnell et al. 2009). Since commissural axons express different combinations of guidance receptors before and after they cross the midline, they do not directly integrate attractive and repulsive effects; instead in this case, opposing guidance cues interact sequentially in a hierarchical fashion, with the response to one gating the response to the other.
In the present study, we examined the effects of 2 opposing guidance cues on a population of axons that possesses the receptor repertoire to respond to one cue by attraction and the other cue by repulsion when presented individually. We previously demonstrated that Sema3A acts as a repulsive and Sema3C as an attractive signal for this population of axons (Bagnard et al. 1998). In the same study, we also proposed that distinct spatial expression patterns of Sema3A and Sema3C define the initial pathway of corticofugal axons. However, the anatomical data presented here indicate that there is a considerable overlap between Sema3A and Sema3C expression in the developing cortex. This prompted us to study how cortical axons respond to different combinations of Sema3A and Sema3C presented simultaneously. Results from different in vitro assays indicated that the repulsive effect of Sema3A overrides the attractive response elicited by Sema3C. Experiments with function-blocking antibodies to different receptor components provided some insights in how cortical axons integrate their responses to Sema3A and Sema3C. Based on these data, we propose a refinement of our original model how Sema3A and Sema3C gradients delineate the early path of corticofugal projections. Finally, our data from axon-tracing experiments in Sema3C knockout animals revealed distinct alterations in efferent cortical projections in vivo, which are consistent with this model.
Materials and Methods
All animals were treated and maintained in accordance with Society for Neuroscience resolutions on the use of animals in research, the National Institutes of Health guidelines, as well as institutional protocols. In brief, animals used were timed pregnant NOR, C57/BL6, and Sema3C knockout mice. The day of detection of the vaginal plug was destined as embryonic day zero (E0). Pregnant mice were deeply anaesthetized using peritoneal injection of 10% chloral hydrate. The embryos were dissected, and the mother was then culled by an overdose of chloral hydrate.
In Situ Hybridization
The pregnant females were culled, and the embryos rapidly removed and decapitated. The E14 and E16 brains were frozen in 2-methylbutane at −40 °C. Digoxigenin-labeled RNA probes for Sema3A, Sema3C, PlexinA1 (PlexA1), PlexinA2 (PlexA2), PlexinA3 (PlexA3), PlexinA4 (PlexA4), Neuropilin-1 (Nrp-1), and Neuropilin-2 (Nrp-2) were used for in situ hybridization. Eighteen micrometers coronal cryostat sections were mounted on superfrost plus slides (Menzel-Gläser). Sections were dried 2–3 h at 56 °C, annealed, and fixed for 10 min in fresh prepared 4% formaldehyde in 1 M Diethylpyrocarbonate (DEPC)–phosphate-buffered saline (PBS, pH 7.4). After washing in DEPC–phosphate-buffered saline (PBS) and permeabilization in 0.2 M HCl for 10 min, the slides were acetylated in 0.1 M Triethanolamine (TEA) with 5 mM acetanhydrid for 15 min. The following washing was performed in DEPC-PBS for 5 min. Then sections were hybridized overnight at 66 °C with 50% formamide, 5× Denhardt's solution, 5× 3 M sodium chloride, 0.3 M sodium citrate (SSC), 100 μg/mL bakers yeast transfer RNA, and 400 μg/mL Torula RNA containing 3 ng/μL Digoxigenin (DIG)-labeled riboprobe. On the next day, sections were washed 5 min in 5× SSC at room temperature, 30 min at 66 °C in 2× SSC/50% formamide, 60 min at 66 °C in 0.2× SSC, and 15 min in 0.2× SSC at room temperature. After short washing in 1× maleic acid–buffered saline (MaBS), the blocking reaction was done in a 2% blocking solution/1× MaBS for 2–3 h. To detect the DIG-labeled riboprobes in the tissue, slides were incubated overnight at 4 °C with an anti-DIG Fab fragment conjugated with alkaline phosphatase (1:750). On the following day, sections were washed 3 times for 10 min in 1× MaBS. After blocking, the endogenous phosphatase activity with levamisole (Sigma) slides was covered with the color substrate (3.75 μL 5-Bromo-4-chloro-3-indolyl phosphate (Roche) and 5 μL Nitro blue tetrazolium chloride (Roche) in 1 mL reaction buffer). The color reaction was performed overnight at room temperature. Finally, slides were washed in 1 M PBS.
Immunohistochemistry was carried out on 20 μm coronal cryosections of heads of E14 wild-type embryonic brains that were immersion fixed with 4% paraformaldehyde (PFA) in PBS for 4 h at room temperature. Fixed brains were then cryoprotected overnight with 15% and 30% sucrose at 4 °C before freezing in liquid nitrogen for cryosectioning. Slices were treated with the blocking reagent (3% BSA, 10% normal goat or donkey serum, in PBS with 0.1% Tween and 0.1% Triton-X-100) for 2 h. Then, the incubation with the primary antibodies rat anti-Nrp-1 (R&D Systems, 1:100) and rabbit anti-Nrp-2 (Cell Signaling, 1:100) was performed overnight at room temperature. Secondary antibodies (donkey anti-goat Cy3, 1:1000 and donkey anti-rabbit Dylight488, 1:1000, both from Jackson Immuno Research Lab. Inc.) were applied for 2 h at room temperature. Nuclei were stained with DAPI (100 ng/mL in PBS; Sigma) for 15 min.
Pictures of alternating slices were taken with a digital camera (Spot, Diagnostic Instruments) in combination with the Spot software and an inverted microscope (Zeiss Axiovert S100). The images were overlaid using Adobe Photoshop 6.0, converted to RGB images, and colorized with corresponding colors (green and red), with overlapping gene expression appearing as yellow.
Qualitative Image Analysis
Sections stacks of 10 overlaid sections, that is, corresponding to 20 original sections alternately stained for Sema3A and Sema3C, were used for each age. Then, we chose a region in the lateral cortex where coexpression of both semaphorins was obvious for measuring signal intensity across cortical layers using ImageJ analysis software. In the selected region, pixel intensity analysis was performed along line profiles constructed from ventricular zone (VZ) to superficial marginal zone (MZ). The data for Sema3A and Sema3C at E14 and E16 are reported as means across sections.
Neocortical tissue of the presumptive somatosensory cortex was prepared from E15 mouse embryo in Gey's balanced salt solution supplemented with 6.5 mg/mL glucose. To obtain explants, the cortical tissue was cut into 200 μm3 cubes using a McIlwain tissue chopper. Cortical explants were kept at 37 °C and 5% CO2 for minimum 2 h before use in coculture medium consisting of 5% fetal bovine serum, 1% penicillin–streptomycin (penicillin: 10 000 U/mL and streptomycin: 10 000 μg/mL), 1% 200 mM L-glutamine, 6.5 mg/mL glucose, and 0.4 mg/mL methylcellulose in Dulbecco's Modified Eagle Medium (all from GIBCO).
Cell Culture and Coculture Assay
Human embryonic kidney (HEK) 293 cells expressing recombinant Semaphorin 3A (Sema3A) or Semaphorin 3C (Sema3C) were used (Bagnard et al. 1998). Culture medium was composed of 1% penicillin–streptomycin (penicillin: 10 000 U/mL and streptomycin: 10 000 μg/mL), 1% 200 mM L-glutamine and 10% fetal bovine serum in Minimum Essential Medium (all from GIBCO). Confluent dishes of each cell type were used to produce cell aggregates (about 750 000 HEK cells per aggregate). First, cells were centrifuged 5 min 800× to obtain a pellet and resuspended in 400 μL culture medium. Twenty microliters drops of this cell suspension were applied to the cover of a dish, which was then repositioned to the bottom allowing the cells in the “hanging drops” to form cell aggregates. The hanging drops were incubated for at least 12 h (37 °C, 5% CO2). For all mixed coculture experiments, cells were counted using a Neubauer chamber. According to the cell numbers, cell lines expressing Sema3A or Sema3C were mixed at a ratio of 50:50, 25:75, 10:90, and 5:95 (3A:3C).
Cell aggregates were transferred to a 20 μL drop of chicken plasma (Sigma-Aldrich) on a glass coverslip in a 33-mm dish. Cortical explants were arranged around the aggregates at a distance of about 1 mm. For coagulation of the plasma clot, 25 μL thrombin was added to the chicken plasma. After 45 min at room temperature, 2.5 mL of coculture medium was added to the cocultures. Blocking experiments were performed by incubating the cultures with 1 μg/mL goat anti-Nrp-1 polyclonal antibody (R&D Systems) and/or 0.5 μg/mL rabbit anti-Nrp-2 polyclonal antibody (Santa Cruz Biotechnology). Cocultures grew 48 h at 37 °C, and 5% CO2; fixation was achieved in 4% PFA (Merck).
Analysis of Coculture Assay
Cocultures were analyzed with an inverted microscope (Zeiss Axiovert S100) equipped with 1.6 optovar, ×20 phase contrast objective, and a SPOT camera (Visitron Systems). For detailed quantitative analysis of axonal outgrowth, all axons growing toward the cell aggregate were counted, and the 10 longest axons were measured via ImageJ software. Student's t-test was used to perform statistical analysis for axonal length. To quantify effects of the blocking experiments and the mixture of Semaphorin 3A– and Semaphorin 3C–secreting cells, numbers according to the growth behavior of the cortical explants were awarded: 0 for explants with radial equally outgrowth; −1 for repulsion which is characterized by a decreased outgrowth toward the cell aggregates, and +1 for attraction of cortical axons growing preferential in the direction of the cell aggregates. Thus, a positive index implies attraction, whereas negative values repulsion of cortical axons. Statistical analysis of these experiments was performed with a chi-square test.
Cortical explants (E15) were prepared as described above and incubated for 42 h at 37 °C and 5% CO2. Then, 2.5 μg/mL anti Nrp-2 (Santa Cruz Biotechnology) antibody was added for 2 h at 37 °C in culture medium. Explants were briefly rinsed with culture medium (37 °C), and the 5 μg/mL of recombinant Sema3A-Fc, Sema3C-Fc (R&D Systems), or Fc-protein (Rockland Inc.) was added. Explants were incubated for 2 h with the recombinant proteins, briefly rinsed with culture medium (37 °C) and fixed in 4% PFA in PBS (37 °C). The 10 longest axons of each explant were analyzed. Therefore, pictures were taken with a Zeiss Axiovert S100 inverted microscope (Zeiss, Germany) using the ×40 phase contrast objective (Zeiss Plan Neofluar, NA 0.75). The number of collapsed growth cones was quantified blindly for every condition.
DiI Labeling in Fixed Brains
For carbocyanine tract tracing, E16 brains were fixed in 4% PFA for 24 h at 4 °C, rinsed in PBS, and kept in PBS/0.02% sodium azide at 4 °C. 1,1′-dioctadecyl-3,3,3′3′-tetramethylindocarbocyanine perchlorate (DiI) crystals were diluted in dimethyl sulfoxide. The DiI solution was injected using pulled glass pipettes. Brains were stored for 3 weeks in the dark at room temperature. Then, they were embedded in 4% agarose and were cut coronally in 100 μm slices using a Vibratome. Brain slices were stained with DAPI for 10 min before embedding in Mowiol. Pictures were taken using the LSM 510 multiphoton laser scanning microscope (Zeiss, Germany) using a ×20 objective (NA 0.75). A helium–neon laser (wavelength 543 nm) in combination with an emission long-pass filter set (560 nm) was used for scanning the DiI signals. The DAPI-stained nuclei were visualized using a 2-Photon Titanium-Saphire laser (Mai-Thai; SpectraPhysics) with a wavelength of 780 and a 435–485 nm band-pass filter set. Images were taken using the LSM510 software and exported in Tagged Image File Format formats for analysis. Image processing was performed with Adobe Photoshop 6.0, compressing all sections to the same height from the VZ to the MZ. The dimension of the DiI signal in the cortex was measured with the open domain ImageJ analysis software (W. S. Rasband, National Institutes of Health, Bethesda, MD; http://rsb.info.nih.gov/ij/, 1997–2006). The highest value of the absolute fluorescence intensity in each slice was set to 100%, while the lowest value was defined as 0%. The relative fluorescence intensities were determined in relation to these set points.
Expression Patterns of Sema3A and Sema3C in the Developing Dorsolateral Cortex
Previous studies have already reported expression of different members of the class 3 semaphorin family, including Sema3A and Sema3C, in the cerebral cortex of the developing mouse brain (Bagnard et al. 1998; Skaliora et al. 1998; Wright et al. 2007; Kolk et al. 2009). Since we were interested in the interactions of Sema3A and Sema3C on growing cortical axons, we first performed in situ hybridizations on consecutive coronal sections to precisely compare the exact expression patterns of these class 3 semaphorins. We examined brains at E14, the stage when the first corticofugal axons reached the internal capsule (Molnar et al. 1998; Auladell et al. 2000; Deng and Elberger 2003). We also analyzed expression patterns of the Sema3-ligand–binding receptors, Neuropilin-1 (Nrp-1) and Neuropilin-2 (Nrp-2), and members of the plexin family, PlexinA1 (PlexA1)–PlexinA4 (PlexA4), which are known to act as signal transducers for Sema3A and/or Sema3C (Kruger et al. 2005; Zhou et al. 2008). In this analysis, we focused on the dorsolateral region of the posterior cortex, the presumptive somatosensory cortex. This cortical region was also used for the functional in vitro studies described in this paper.
At E14, Sema3A and Sema3C transcripts were found in adjacent cortical zones. The Sema3A signal was mainly restricted to the VZ and subventricular zone (SVZ), whereas a Sema3C labeling was mainly confined to the intermediate zone (IMZ; Fig. 1A,B). However, a closer analysis of Sema3A and Sema3C signals at higher magnification reveals some overlap in the upper SVZ/IMZ of these transcripts (Fig. 1A′,B′). To quantify these results, we generated overlays of consecutive sections and used ImageJ to measure the signal intensity and the amount of overlap across the cortical layers at E14 and at E16. The Sema3A and Sema3C expression in pseudocolors in the cortex is illustrated in Figure 2A,D at E14 and E16, respectively. The location of the magnified sections shown in Figure 2B,E, which were used for quantification of fluorescence intensity, is indicated by the white rectangle in Figure 2A,D. The plots of the corresponding signal intensities are presented in Figure 2C,F. This plot reveals that at E14 Sema3A and Sema3C are coexpressed in the upper SVZ and IMZ. About 25% of the total Sema3C signal, the integral of the signal intensity over the whole cortical thickness overlaps with Sema3A expression zones. At E16, expression of Sema3A in the VZ/SVZ is similar to what we found at E14, while there is an additional signal in the MZ (Fig. 2D). The Sema3C expression at E16 was mainly restricted to upper SVZ and the lower IMZ (Fig. 2D,F). At both stages (E14 and E16), there were two distinct gradients of the Sema3C signal, with a relatively weak expression in the dorsal cortex and a high expression in the lateral and medial cortex (Fig. 2A,B). A comparison of the Sema3A and Sema3C signals at E16 showed a pronounced overlapping expression in the upper SVZ (yellow region in Fig. 2D,E). The quantitative signal intensity plot shown in Figure 2F revealed that the overlap of the Sema3C signal with Sema3A at E16 increased to about 60%.
We also detected Nrp-1 and Nrp-2 transcripts at both developmental stages examined. At E14, both Nrp-1 and Nrp-2 are expressed in the IMZ (Fig. 1C,C′,D,D′). The expression persisted largely to E16 (data not shown). A similar expression pattern was found at the protein level, as illustrated in Figure 1E–E′′′ with immunohistochemical stainings with Nrp-1 and Nrp-2 antibodies. Finally, transcripts of plexins known to associate with neuropilins to form functional Sema3 holoreceptors (PlexA1–PlexA4) have also been detected in the cortex at E14 and E16 (Supplementary Fig. 1).
Response of Cortical Axons to Sema3A and Sema3C Cocktails
More than 10 years ago, we first described an attractive effect of a class 3 semaphorin, and we proposed a model according to which growing efferent cortical axons are attracted by a Sema3C gradient that drives them deep into the IMZ. Additionally, a repulsive Sema3A gradient prevents cortical axons from entering the proliferative zones, the VZ and SVZ (Bagnard et al. 1998). This model was based on functional in vitro assays with Sema3A and Sema3C and the limited expression data of the Sema3s available at that time. The present results demonstrate, however, that there is a considerable overlap of Sema3A and Sema3C in the SVZ at E14 and E16. This raises the obvious question how cortical axons might respond in a situation in which they are simultaneously exposed to Sema3A and Sema3C. We therefore analyzed the behavior of cortical axons in coculture experiments in the presence of defined mixtures of Sema3A and Sema3C molecules.
We first performed coculture experiments with Sema3A- and Sema3C-secreting cell lines and cortical explants (E15) to confirm previous results that Sema3A acts as a repulsive and Sema3C as an attractive signal for growing cortical axons (Bagnard et al. 1998; Castellani et al. 2000; Niquille et al. 2009; Piper et al. 2009). As illustrated in Figure 3B,D, in the presence of Sema3A-secreting cell aggregates, there are fewer cortical axons growing toward increasing Sema3A gradients. In addition, cortical axons growing toward Sema3A-secreting cells are shorter than axons growing away from these cells (Fig. 3E). The opposite effect was observed within Sema3C gradients: There are more and longer cortical axons when growing toward than away from the Sema3C gradient (Fig. 3C–E).
We then produced cell aggregates with defined mixtures of Sema3A- and Sema3C-secreting cells. Previous work demonstrated that Sema3A or Sema3C released at the same time from transfected cells form disulfide-linked homodimers, whereas Sema3A/Sema3C heterodimers are not formed (Takahashi et al. 1998). Thus, combining Sema3A and Sema3C does not produce artificial semaphorin ligands. As illustrated in Figure 4, in the presence of 50% and 75% of Sema3C-expressing cells in the aggregates, the repulsive effect of Sema3A was still present, but it was reduced in a concentration-dependent manner. When the Sema3A and Sema3C aggregates contained 90% or 95% of Sema3C-expressing cells, there was neither a repulsive nor an attractive effect on cortical axons. In other words, the attractive effect of Sema3C was negated in the presence of even very low concentrations of Sema3A.
Integration of Opposing Sema3A and Sema3C Guidance Cues
With few exceptions (Chauvet et al. 2007), Sema3s require first neuropilins as direct binding partners for cell signaling. Then plexins of the A and D family, associated with neuropilins, are obligatory coreceptors for transmitting cell signaling for Sema3s (Kruger et al. 2005). Ligand/receptor-binding studies suggested that Sema3A binds exclusively to Nrp-1, but not to Nrp-2, whereas Sema3C binds to Nrp-1 and Nrp-2 with equal affinities (He and Tessier-Lavigne 1997; Chen et al. 1998; Takahashi et al. 1999). Therefore, in a first step to analyze how cortical axons integrate opposing effects of Sema3A and Sema3C, we studied the effects of these semaphorins after blocking Nrp-1 and Nrp-2 function with antibodies either alone or in combination in coculture assays. As illustrated in Figure 5B, the repulsive effect of Sema3A was almost completely abolished by antibodies directed against Nrp-1. As expected from previous studies, antibodies directed against Nrp-2 had no effect on the repulsive effects of Sema3A, and the combination of Nrp-1 and Nrp-2 antibodies had the same effects as Nrp-1 antibodies alone (Fig. 5A). In contrast, the attractive response of Sema3C on cortical axons was blocked by antibodies directed either against Nrp-1 alone or by a combination of Nrp-1 and Nrp-2 antibodies. Surprisingly, however, in the presence of Nrp-2–blocking antibodies, the attractive response of cortical axons to Sema3C was converted in a repulsive response to this cortical wiring molecule (Fig. 5A).
To confirm the finding that cortical axons switch their response to Sema3C from attraction to repulsion after blocking Nrp-2 function, we performed an independent set of experiments. For this, we turned to the growth cone collapse assay, which has been widely applied to study repulsive semaphorin effects. In accordance with our previous results (Bagnard et al. 1998), addition of recombinant Sema3A induced growth cone collapse, whereas Sema3C did not lead to an increased collapse above background level (Fig. 5B). In contrast, in the presence of Nrp-2–blocking antibodies, Sema3C now also induced growth cone collapse (Fig. 5B). The collapse activity of Sema3A was not affected by function-blocking Nrp-2 antibodies. However, as already noticed with the coculture assays, the repulsive effects of Sema3C after blocking Nrp-2 function in the collapse assay, although highly significant, was weaker than the repulsive effects elicited by Sema3A.
Corticofugal Projections in Sema3C-Deficient Mice
The expression data and the in vitro experiments described so far suggest that the initial path of corticofugal axons is under the control of 2 opposing and partially overlapping semaphorin gradients. First, the Sema3C gradient attracts cortical axons into the narrow IMZ, the path of efferent cortical axons. Beneath the IMZ, in the SVZ, the expression of Sema3A prevents corticofugal fibers from entering this inappropriate pathway.
One prediction of this model is that in the absence of Sema3C attraction, corticofugal axons might be misdirected and/or choose a more superficial pathway. To test this hypothesis, we performed axon-tracing experiments in Sema3C knockout animals. For this, we made DiI injections in fixed E16 brains of Sema3C−/− and Sema3C+/+ animals and performed a quantitative analysis of the DiI-labeled cortical axons, as described in the Materials and Methods. Examples of DiI-labeled cortical axons wild type and in mutant animals are illustrated in Figure 6A,B, respectively. The plot of the distribution of efferent cortical fibers, averaged from 3 mutant and 3 wild-type brains, revealed that the bulk of the axons in Sema3C−/− animals, compared with Sema3C+/+ mice, is shifted toward the cortical plate (CP) (Fig. 6C). Thus, efferent cortical axons in Sema3C-deficient mice do not penetrate deep into the IMZ. In addition, in 2 of 3 brains from mutant animals, we also observed several efferent cortical fibers that initially traveled for a considerable distance in the CP, which was never observed in the 3 brains examined from wild-type mice.
To test if the misdirected axonal pathways shifted toward the CP in Sema3C−/− animals might be due to alterations in the architecture of the developing cortex, DAPI stainings were performed to identify the proliferative regions (VZ and SVZ), the IMZ, the cortical plate, and the MZ in knockout and wild-type mice at E14 The relative thickness of these embryonic cortical layers was calculated using ImageJ. There were no significant differences between Sema3C knockout and wild-type brains (Supplementary Fig. 2), supporting the hypothesis that the defects in the corticofugal trajectories observed in the Sema3C−/− brains are a direct result of alterations in brain wiring molecules. Taken together, this phenotype is consistent with the idea that an attractive Sema3C gradient pulls cortical fibers into the IMZ, the pathway of corticofugal axons.
During brain development, growing axons simultaneously encounter multiple guidance cues, thus growth cones have to integrate cell signaling mediated by various brain wiring molecules at the same time. With some notable exceptions (reviewed in Dontchev and Letourneau 2003), most previous studies concentrated on the effects of single guidance cues or examined different cues individually. In this study, we examined how cortical axons respond to different concentrations of Sema3A and Sema3C presented concurrently. Previous work demonstrated that these 2 semaphorins exert opposing effects on cortical axons, with Sema3A acting as a chemorepellent` and Sema3C as a chemoattractant. Our expression analysis revealed that during embryonic development, cortical axons are exposed to Sema3A and Sema3C concurrently. Functional assays with defined mixtures of Sema3A and Sema3C indicated that even low concentrations of Sema3A can override the attractive effects of Sema3C. Blocking experiments with antibodies directed against the ligand-binding receptor components of Sema3s, Nrp-1, and Nrp-2, provided some insights in how cortical axons integrate the opposing Sema3A and Sema3C cues at the receptor level. Together, the results of the present study lead to a refinement of our previous model on the coordinated function of Sema3A and Sema3C during the development of corticofugal pathways (Bagnard et al. 1998), and our axonal tracing studies in Sema3C-deficient animals were consistent with this hypothesis.
Opposing Effects of Sema3s on Growing Axons
The first vertebrate semaphorin discovered, Sema3A, was identified by its ability to induce a transient growth cone collapse in vitro and has therefore been originally named collapsin (Luo et al. 1993). As more members of this gene family were identified, it became clear that in some cases, semaphorins do not repel but instead attract growing axons. Since the first report that Sema3C has an attractive effect on cortical neurons (Bagnard et al. 1998), many other studies provided evidence that Sema3s-induced attraction for different population of neurons (de Castro et al. 1999; Wolman et al. 2004; Chauvet et al. 2007). The effects of Sema3s can be cell type specific, for example, Sema3E can be repulsive for corticofugal neurons and attractive for subiculo-mammillary neurons (Chauvet et al. 2007). Moreover, the effects of a given Sema3 acting on the same population of axons can even cycle between attraction and repulsion, dependent on the intracellular concentration of cyclic nucleotides (reviewed in Song and Poo 1999). Semaphorins, in addition to acting as guidance cues for growth cones and migrating neurons, can also influence axonal fasciculation, cell migration, and many other biological processes; for a recent review, see Zhou et al. (2008).
The intricacy of Sema3s actions is matched in complexity by the composition of their receptors. With very few exceptions (Gu et al. 2005), Sema3s require either Nrp-1 and/or Nrp-2 as ligand-binding partners, but there are different binding affinities for Sema3s to Nrp-1 and Nrp-2 (Feiner et al. 1997; He and Tessier-Lavigne 1997; Kolodkin et al. 1997). However, neuropilins have only a short cytoplasmatic domain, thus they are not able to activate downstream signal-transduction pathways by themselves but require additional molecular partners to form a functional Sema3 receptor. One important class of coreceptors are the plexins, which in vertebrates consist of 9 members divided in 4 classes A–D. Neuropilins can complex with PlexA1–PlexA4 and PlexD1 (reviewed in Tran et al. 2007). Moreover, Ig superfamily cell adhesion molecules, such as L1 and NrCAM, have also been found to contribute to Sema3 responses (Castellani et al. 2000, 2004; Falk et al. 2005; Wright et al. 2007; Hernandez-Miranda et al. 2011).
Our expression data demonstrated that all potential components of the Sema3A and Sema3C receptors are present in the cortex at the time when the neurons send their axons to distant targets. Binding studies revealed that Sema3A binds only to Nrp-1, but not to Nrp-2, whereas Sema3C binds to both neuropilins with high affinity (He and Tessier-Lavigne 1997; Chen et al. 1998; Takahashi et al. 1999). This raises the immediate question how cortical axons can discriminate between these class 3 semaphorins and exhibit differential responses. Since Sema3C can bind to Nrp-1 with high affinity, it should also activate the Sema3A receptor, whatever signaling coreceptors are associated with Nrp-1 on cortical axons. As will be discussed below, our Nrp-1 and Nrp-2 antibody blocking experiments provide a possible answer to this question.
How Cortical Axons Integrate Opposing Effects of Sema3A and Sema3C
Although Sema3A and Sema3C bind with equal affinity to Nrp-1 alone, neuropilin/plexin interactions can modify their binding specificity with neuropilins (Takahashi et al. 1999; Rohm et al. 2000). For example, coexpression experiments of plexins and neuropilins in transfected 293T cells indicated that PlexA1 and PlexA2 substantially increased Sema3A binding to Nrp-1 compared with Nrp-1 alone. In contrast, whereas for Sema3C, a considerable augmentation of the binding affinity to Nrp-1 was also observed in the presence of PlexA1, coexpression of Nrp-1 with PlexA2 even decreased Sema3C binding. Thus, given an appropriate association of specific plexins with Nrp-1, neurons could assemble a Sema3A holoreceptor that exhibits only little cross reactivity to Sema3C.
Like the Sema3 ligands, which in vivo form disulfide-linked homodimers, neuropilins are also dimerized. When both neuropilins are present in neurons, all studies performed so far could not distinguish whether cells express homo- or hetero-neuropilins dimers or a combination of both (Kolodkin et al. 1997; Chen et al. 1998; Giger et al. 1998). Thus, Sema3C could activate holoreceptors containing either Nrp-1 homodimers or Nrp-1/Nrp-2 heterodimers. As argued below, the antibody blocking experiments of the present study suggest that it is most plausible that cortical neurons possess Nrp-1/Nrp-2 heterodimers.
The repulsive response to Sema3A was blocked in the presence of antibodies directed against Nrp-1, whereas Nrp-2 antibodies did not influence the Sema3A effects on cortical neurons. As illustrated in Fig. 7A, a straightforward explanation of these results is that Sema3A activates a receptor complex consisting of Nrp-1 homodimers associated plexins that increase the binding of Sema3A to Nrp-1 and at the same time decrease Sema3C/Nrp-1 binding. The situation for Sema3C is more complicated, and we suggest the following scenario. First, Sema3C also binds with low affinity to the Nrp-1 homodimers and elicits a weak repulsive response (Fig. 7B). This “cross talk” is superseded by a strong attractive response mediated by Nrp-1/Nrp-2 heterodimers associated with plexins that lead to high-affinity Sema3C binding (Fig. 7C). According to this model, Nrp-1 antibodies then would block both the strong attractive and the weak repulsive response. In contrast, Nrp-2 antibodies would only block the attractive effects of Sema3C, whereas the weak repulsive effect is not affected. As a result, antibodies directed against Nrp-2 should convert the attractive effects of Sema3C on cortical neurons to repulsion, as observed in our in vitro experiments. This model would also explain why the repulsion mediated by Sema3C in the presence of Nrp-2 antibodies is always weaker than the repulsion mediated by Sema3A. Finally, Figure 7D illustrates why it is less likely that the Sema3C response of cortical neurons is mediated by Nrp-2 homodimers. In this case, Nrp-2 antibodies would also lead to a switch from attraction to repulsion. However, Nrp-1 antibodies should have no effect on the attractive response component of Sema3C. On the contrary, the attractive Sema3C response should even increase because the cross talk with Nrp-1 homodimers mediating the weak repulsive effects is blocked. Since Nrp-1 antibodies abolished the response to Sema3C, we postulate that Nrp-1/Nrp-2 heterodimers are the coreceptors for this ligand on cortical axons.
This interpretation of our data can also explain some of the results described in 2 recent studies on the effects of Sema3A and Sema3C on pioneer axons from the cingulate cortex that initiate corpus callosum development (Niquille et al. 2009; Piper et al. 2009). In analogy with the effects reported here for neocortical axons, Sema3A has also a repulsive and Sema3C an attractive effect on these limbic cortical axons. Also in accordance with the present results, antibodies directed against Nrp-1 blocked both the Sema3A and Sema3C response. To explain the contrasting effects of Sema3A and Sema3C on callosum axons, the authors argue that different Nrp-1 receptor complexes coexist in these axons, with different combinations of subunits that confer specific binding properties either to Sema3A or Sema3C (Niquille et al. 2009). Nrp-2 antibodies were not tested in these studies because the authors could not detect Nrp-2 at the early (E15) stages of corpus callosum development. However, our in situ data indicate that Nrp-2 transcripts are present in the cingulate cortex already at E14 (S Barchmann, J Bolz, unpublished results). Based on the findings presented here, we would predict that Nrp-2 antibodies should also influence the effect of Sema3C on callosal axons, either by blocking Sema3C attraction or even by converting the response from attraction to repulsion.
The Initiation of Corticofugal Projections
During embryonic development, cortical neurons extend their efferent axons deep in the IMZ, the future white matter, whereas thalamic afferents invading the cortex travel in upper IMZ and subplate zone. This segregation of afferent and efferent projections remains preserved in the white matter of adult mice (Miller et al. 1993). At E14, when the first pioneering cortical axons start to project out of the cortex, the IMZ consists of a very thin band, which cannot be detected with conventional histological techniques. Our in situ hybridizations with Sema3A and Sema3C probes indicate that at E14, Sema3C is expressed in the presumptive IMZ and in the SVZ and that Sema3A is present throughout the VZ and lower SVZ. This then suggests that the pathway of corticofugal projections is defined by the zone in which Sema3C is expressed exclusively, whereas the adjacent zone, in which Sema3C and Sema3A are coexpressed, are avoided by efferent cortical axons. Likewise, at E16, the only zone where Sema3C does not overlap with Sema3A is a zone deep in the IMZ, the presumptive pathway of efferent cortical axons, whereas in the adjacent SVZ, Sema3C is coexpressed with Sema3A.
When we first described the contrasting effects of Sema3A and Sema3C on cortical axons, we proposed that Sema3C attracts these axons toward the IMZ and Sema3A prevents them from entering the SVZ and VZ (Bagnard et al. 1998). We noted already at that time that since Sema3s are diffusible molecules, cortical axons might be exposed to overlapping antagonistic gradients of Sema3A and Sema3C. What was not known at that time, however, was how cortical axons respond when they are exposed to a combination of Sema3A and Sema3C. The present data provide a clear answer to this question: even at very low concentration, the repulsive Sema3A effect dominates over the attractive Sema3C effect. Thus, our new data lead to a refinement of our original model: The pathway of corticofugal fibers is defined by a zone where Sema3C is expressed alone and where it can act as an attractant for this population of axons. The adjacent zone, where Sema3A and Sema3C gradients overlap, is not entered by corticofugal axons because the repulsive Sema3A effect overrides the attractive Sema3C effect.
This hypothesis is based on the functional characterization of Sema3A and Sema3C with in vitro assays, together with their expression patterns in the developing cortex. To directly test this prediction in vivo, we analyzed the initial trajectories of corticofugal axons in Sema3C knockout mice. In these animals, we could detect relatively subtle but consistent alterations in pathway selection of efferent cortical fibers. First, compared with wild-type animals, cortical axons take a more superficial route in Sema3C−/− mice, and second, the whole corticofugal corridor was compressed. This phenotype is expected when an attractive force for these axons, Sema3C, is eliminated and a repulsive force, Sema3A, becomes more dominant. However, corticofugal fibers in Sema3C−/− mice, with a few exceptions, still navigate within the more superficial IMZ, suggesting that other molecular cues also contribute to define their pathway. In future studies, it will be interesting to study the Sema3C knockout animals in more detail and examine, for example, whether there are changes in expression patterns of Sem3A or other members of the semaphorin gene family. Adaptive and compensatory changes have been described for the Eph/ephrin gene family in a mouse line where ephrin-A5 was selectively mutated (Peuckert et al. 2008). It will also be worth examining corticofugal axons in Sema3A knockout animals, which should exhibit an opposite phenotype to Sema3C mutants, with more fibers navigating more deeply in the IMZ or perhaps even entering the SVZ. There are indications that cortical axons are abnormally oriented in Sema3A−/− mice (Polleux et al. 1998), but so far there is no detailed analysis of corticofugal pathway selection in these animals.
IZKF Jena to J.B.; Carl-Zeiss Foundation to G.Z.
We are grateful to André Steinecke for many helpful comments on the manuscript. We would like to thank Christine Raue for excellent technical assistance. Conflict of Interest: None declared.