Here we examine the role of Reelin, an extracellular protein involved in neuronal migration, in the formation of hippocampal connections. Both at prenatal and postnatal stages, the general laminar and topographic distribution of entorhinal projections is preserved in the hippocampus of reeler mutant mice, in the absence of Reelin. However, developing and adult entorhinal afferents show severe alterations, including increased numbers of misrouted fibers and the formation of abnormal patches of termination from the medial and lateral entorhinal cortices. At perinatal stages, single entorhinal axons in reeler mice are grouped into thick bundles, and they have decreased axonal branching and decreased extension of axon collaterals. We also show that the number of entorhino-hippocampal synapses is lower in reeler mice than in control animals during development. Studies performed in mixed entorhino-hippocampal co-cultures combining slices from reeler and wild-type mice indicate that these abnormalities are caused by the lack of Reelin in the target hippocampus. These findings imply that Reelin fulfills a modulatory role during the formation of layer-specific and topographic connections in the hippocampus. They also suggest that Reelin promotes maturation of single fibers and synaptogenesis by entorhinal afferents.
The formation of initial neural projections depends on diffusible and membrane-anchored factors, whereas neurotrophins and neuronal activity are believed to shape final axonal arbors (Katz and Shatz, 1996; Tessier-Lavigne and Goodman, 1996). In addition, cell adhesion molecules and extracellular matrix proteins contribute to the growth and targeting of developing axons (Stoeckli and Landmesser, 1995; Faissner, 1997; Gotz et al., 1997;Inoue and Sanes, 1997).
In many brain regions neural connections are organized in specific laminae. The main afferents to the hippocampus terminate in a laminar manner in which distinct synaptic inputs innervate nonoverlapping layers. Thus, entorhinal afferents innervate the stratum lacunosum-moleculare (SLM) and the outer molecular layer (OML) and, conversely, the commissural/associational fibers terminate in the stratum oriens (SO), stratum radiatum (SR), and inner molecular layer (Blackstad, 1956; Steward, 1976; Steward and Scoville, 1976; Swanson and Cowan, 1977; Swanson et al., 1978; Ruth et al., 1982; Amaral and Witter, 1995). Tracing studies in embryos have shown that hippocampal afferents invade their target layers as soon as they enter the hippocampus, which suggests that specific laminar cues are already present at these stages (Supèr and Soriano, 1994; Supèr et al., 1998a,b). This is consistent with slice culture experiments showing that the layer-specific targeting of hippocampal afferents does not depend on the temporal order of fiber arrival (Frotscher and Heimrich, 1993).
Cajal-Retzius (CR) cells are a special class of pioneer neuron in the marginal zone–layer I of the cerebral cortex (Marı́n-Padilla, 1978; Edmunds and Parnavelas, 1982; Marı́n-Padilla and Marı́n-Padilla, 1982; Derer and Derer, 1990). These neurons are also very abundant in the hippocampus, where they are arranged in a single layer, the outer marginal zone–SLM (Soriano et al., 1994; Del Rı́o et al., 1995, 1996). At the time of ingrowth, entorhinal fibers overlap extensively and form transient synaptic contacts with CR cells (Supèr and Soriano, 1994; Supèr et al., 1998a). The ablation of hippocampal CR cells in organotypic cultures prevents the ingrowth of entorhinal afferents to the hippocampus (Del Rı́o et al., 1997), indicating that pioneer CR cells play an important role in the guidance and targeting of layer-specific hippocampal connections.
The reelin gene, which is responsible for the reeler mutation and encodes a large extracellular matrix protein, is highly expressed in CR cells (D’Arcangelo et al., 1995, 1997; Hirotsune et al., 1995; Ogawa et al., 1995; Nakajima et al., 1997; Schiffmann et al., 1997; Alcántara et al., 1998). Incubation of entorhino-hippocampal slices with anti-Reelin antibodies resulted in decreased innervation and reduced branching and growth of entorhinal afferents (Del Rı́o et al., 1997). Preliminary studies in vivo suggested that similar alterations may occur in reeler mutant mice, suggesting a novel role for Reelin in axonal growth (Del Rı́o et al., 1997; Holt and Harris, 1998).
Here we analyze the development of the entorhino-hippocampal connection in reeler mice. We show that reelin is expressed in CR cells before and during the arrival of entorhinal afferents, and that its absence leads to alterations in the entorhino-hippocampal pathway, including reduced axonal branching, and an increase in the number of misrouted aberrant fibers. Some of these abnormalities are transient, because they are not detectable in adult reeler mice. Furthermore, we show that reeler mice have fewer entorhino-hippocampal synapses, which indicates a role for Reelin in synaptogenesis. Finally, becausereelin is expressed in the entorhinal cortex, mixed organotypic co-cultures of reeler and heterozygous slices were prepared to determine the contribution of Reelin produced by the target region to the pattern of afferent termination.
MATERIALS AND METHODS
Animals. OF1 mice (IFFA-Credo, Lyon, France) were used for the mRNA in situ hybridization studies. For tracing studies, reeler and heterozygous mice were obtained by mating reeler (Relnrl, BALB/c) homozygous (rl/rl) males with heterozygous (rl/+) females; BALB/c and OF1 mice were used as controls (+/+). The day on which a vaginal plug was detected was considered embryonic day 0 (E0), and the day of birth was considered postnatal day 0 (P0).
In situ hybridization and immunocytochemistry. Embryos from E12, E14, E16, and E18 stages, mice from postnatal stages P0, P5, P10, P15, and P21, and adults (two to three animals each) were perfused with phosphate-buffered (PB) 4% paraformaldehyde. The brains were post-fixed with the same fixative, cryoprotected with 30% sucrose, and sectioned at 25–60 μm. A reelin antisense riboprobe was labeled with digoxigenin-d-UTP (Boehringer Mannheim, Mannheim, Germany) by in vitro transcription of a 2.2 kbp fragment encoding mouse reelin (D’Arcangelo et al., 1995) using T3 polymerase (Ambion).
In situ hybridization was performed on free-floating sections essentially as described elsewhere (de Lecea et al., 1994;Alcántara et al., 1996). Briefly, sections were permeabilized, deproteinized, and acetylated. Thereafter, sections were prehybridized in a solution containing 50% formamide, 10% dextran sulfate, 5× Denhardt’s, 0.62 m NaCl, 10 mm EDTA, 20 mm PIPES, pH 6.8, 50 mm1,4-dithio-dl-threitol (DTT), 250 mg/ml yeast RNA, and 250 mg/ml denatured salmon sperm DNA for 3 hr at 60°C. Sections were then incubated overnight at 60°C in the same solution containing the labeled antisense RNA riboprobe (200–500 ng/ml). Thereafter, sections were treated with RNase A, washed in 0.5× SSC/50% formamide (55°C) and in 0.1× SSC/0.1% sarkosyl (60°C). Sections were then rinsed and incubated with an anti-digoxigenin antibody conjugated to alkaline phosphatase. After washing, sections were developed with nitroblue tetrazolium salt and 5-bromo-4-chloro-3-indolyl phosphate, toluidinium salt, mounted onto gelatinized slides, and coverslipped with Mowiol. Some sections were immunostained with rabbit antibodies against calbindin (1:6000) or calretinin (1:3000) (Swant antibodies, Bellizona, Switzerland). Primary antibodies were visualized using biotinylated anti-rabbit antibodies and the avidin–biotin–peroxidase complex (ABC; Vector Labs, Burlingame, CA). The peroxidase reaction was developed using diaminobenzidine (DAB) and H2O2 as substrates. Sections were mounted and coverslipped as above.
Controls including hybridization with a sense riboprobe or omission of the primary antibodies prevented hybridization or immunohistochemical signals.
Tracing of the entorhino-hippocampal pathway in vivo. Pregnant mice were killed by an overdose of ether at embryonic stages E16–E18. Embryos were perfused with 4% paraformaldehyde in 0.1m PB. A single injection of DiI (Molecular Probes, Eugene, OR) was delivered into the entorhinal cortex via a glass micropipette (Supèr and Soriano, 1994), under inspection through a microscope (8 rl/+, 15 rl/rl, 7 +/+). Then, brains were stored in fixative for 3–4 weeks in the dark. Coronal vibratome sections (80 μm thick) were counterstained with bisbenzimide, mounted, and coverslipped with Mowiol. Postnatal and adult reeler, heterozygous, and wild-type mice (P1: 15 rl/rl, 13rl/+, 7 +/+; P4: 10 rl/rl, 10 rl/+, 6 +/+; P11: 8 rl/rl, 15 rl/+, 4 +/+) and adult mice (13 rl/rl, 6 rl/+, 3 +/+) were iontophoretically injected with biocytin in the entorhinal cortex (+7.6 μA; 2 sec on, 1 sec off). After 24 hr of survival, animals were perfused with 4% paraformaldehyde, and the brains were post-fixed overnight at 4°C. After cryoprotection, brains were cut either coronally or horizontally (50 μm thick). After blocking endogenous peroxidase activity, sections were incubated with ABC and developed with a DAB nickel-enhanced reaction (black reaction product); sections were counterstained with Nissl stain or immunostained for calretinin, as described elsewhere (Del Rı́o et al., 1996) using DAB (brown product). In some cases, sections were incubated with anti-calretinin antibody and streptavidin–Texas Red overnight, rinsed several times, and incubated with a secondary antibody bound to fluorescein. These sections were analyzed with a confocal microscope.
Electron microscopy. Postnatal mice (P2: 2 rl/+, 1 rl/rl; P5: 6 rl/+, 4 rl/rl) were injected in the entorhinal cortex with biocytin for the tracing of the entorhino-hippocampal pathway. After 24 hr of survival animals were anesthetized with ether and perfused with 4% paraformaldehyde and 0.1–0.2% glutaraldehyde in PB. Vibratome sections were processed for the visualization of biocytin as above. Uninjected animals were also perfused and vibratome-sectioned (P2: 4 rl/+, 2rl/rl; P5: 4 rl/+, 5 rl/rl). Tissue slices were post-fixed with osmium tetroxide, stained with uranyl acetate, and flat-embedded in araldite. Thin sections were collected onto Formvar-coated slot grids and stained with lead citrate. Quantitative analyses were performed on a series of electron micrographs taken at random in the SLM and OML (final magnification 16,000×).
Reeler co-culture experiments. Entorhino-hippocampal slice co-cultures were prepared from P0–P1 mice essentially as described (Frotscher and Heimrich, 1993; Del Rı́o et al., 1996, 1997). Animals were anesthetized by hypothermia, and the hippocampus and prospective parietal cortex were dissected out. Tissue pieces were cut into transverse slices (300–350 μm thick) using a McIlwain tissue chopper, and the slices were maintained in Minimum Essential Medium (MEM) supplemented with l-glutamine (2 mm) for 45 min at 4°C. Selected slices were co-cultured using the membrane interface technique (Stoppini et al., 1991). Incubation medium was 50% MEM, 25% horse serum, 25% HBSS, supplemented withl-glutamine (2 mm). The reeler or normal (heterozygous) phenotype of the pups was determined by the cytoarchitectonics of the hippocampus, clearly discernible in slices under dark-field optics. Slices from reeler mice displayed a disrupted dentate gyrus and a typical double-layered pyramidal layer. Five types of co-cultures were prepared as follows: (1) wild-type co-cultures (E+/+/H+/+) (n = 27) and (2) heterozygous co-cultures (Erl/+/Hrl/+) (n = 25), where both slices were from wild-type and heterozygous animals, respectively; (3) reeler co-cultures (Erl/rl/Hrl/rl) (n = 40), in which the tissue pieces were from reeler pups; (4) mixed co-cultures using hippocampi from reeler and entorhinal cortices from heterozygous pups (Erl/+/Hrl/rl) (n = 44); and (5) mixed co-cultures combining heterozygous hippocampi with reeler entorhinal cortex slices (Erl/rl/Hrl/+) (n = 56). To assess the formation of entorhino-hippocampal connections, a crystal of biocytin was injected into the entorhinal slice 24 hr before fixation (Del Rı́o et al., 1997). Co-cultures were fixed after 7 d in vitro(DIV) and 15 DIV with 4% paraformaldehyde in 0.1 m PB. After several rinses, 40-μm-thick sections were obtained, blocked with 10% normal goat serum, and incubated with the ABC overnight at 4°C. After development with a nickel-enhanced DAB reaction, sections were counterstained with Nissl stain. In some cases, sections were processed for the detection of calretinin as described (brown product).
Quantitative analysis. Selected biocytin-labeled fibers were drawn using a 100× objective lens and a camera lucida (23 to 47 fibers per group and age). Lengths were measured using a planimeter and the branching index (number of branching points per 100 μm), lengths of side collaterals (in micrometers), and density of boutons (number of axonal varicosities per 100 μm) were calculated.
The density of synaptic contacts was counted in 16,000× electron micrographs (each of 80 μm2; 26 to 30 photographs per group, obtained from two animals). The length of the synaptic contacts was drawn and measured using an IMAT image analyzer (Scientific and Technical Services, University of Barcelona). ANOVA least significant difference (LSD) tests were run to analyze the significance of differences.
Co-cultures were classified as shown in Table 1. Co-cultures receiving similar biocytin injections were viewed with a 40× oil-immersion objective lens and a millimetric eyepiece. Axons crossing a 75 μm bar were counted on a single focus plane (three to six counts per culture).
reelin is expressed before and during the formation of hippocampal connections
To investigate the developmental expression pattern ofreelin in the hippocampal region, embryos and postnatal mice were processed for in situ hybridization. reelinmRNA was detected at E12–E14 (preplate stage) (Soriano et al., 1994) in a thick layer of cells at the surface of the hippocampal primordium (Fig. 1 A). At E16–P0,reelin mRNA-positive cells were densely packed around the hippocampal fissure, particularly in the outer marginal zone (prospective SLM) (Soriano et al., 1994), which at these stages is densely populated by CR cells (Fig. 1 B). Double-labeled sections positive for calretinin immunoreactivity, a marker of murine CR cells (Soriano et al., 1994; Del Rı́o et al., 1995), confirmed that these intense reelinmRNA-positive cells were indeed CR cells displaying horizontal cell bodies and dendrites (Fig. 1 C,D). These data show thatreelin mRNA is expressed in CR cells before and during the growth of entorhinal axons into the hippocampus (Supèr and Soriano, 1994).
From E16 onward, a few cells expressing a low level ofreelin mRNA were also seen in the inner marginal zone (prospective SR) and below the pyramidal layer (prospective SO) (Fig.1 B,C). Double-labeling studies revealed that these neurons were calretinin-negative, although some expressed calbindin (data not shown), a marker for a population of GABAergic pioneer neurons present in these layers (Soriano et al., 1994; Supèr et al., 1998a,b). Thus, in addition to CR cells, reelinis expressed at low levels in a subpopulation of GABA-positive neurons.
reelin mRNA was also expressed in CR cells in layer I of the entorhinal cortex from E12 onward. In addition, weak hybridization was observed from E18 onward in neurons located in layer II of the entorhinal cortex (Fig. 1 B). Because these neurons are known to project to the dentate gyrus (Steward, 1976; Steward and Scoville, 1976; Swanson and Cowan, 1977; Ruth et al., 1982; Amaral and Witter, 1995), we conclude that reelin mRNA is also expressed in neurons that originate in the entorhino-hippocampal pathway.
During the early postnatal period (P2–P10), the pattern of expression and intensity of labeling in the hippocampal area remained essentially similar. At later stages (P15–P21), the expression ofreelin mRNA decreased progressively in both CR cells and GABAergic neurons of the hippocampus (data not shown). In adult mice, few weakly labeled cells were observed in the hippocampus and in the entorhinal cortex (Alcántara et al., 1998).
Reeler mice display abnormalities in the entorhino-hippocampal pathway
To study the role of Reelin in the formation of the entorhino-hippocampal projection in vivo, reeler embryos and postnatal animals were injected in the entorhinal cortex with the lipophilic dye DiI or with biocytin. In mouse embryos, entorhinal afferents reach the hippocampal white matter by E14–E15 (Supèr and Soriano, 1994). In wild-type and heterozygous embryos (E16–E18), DiI-labeled entorhinal afferents completely filled the outer marginal zone, where fibers had numerous collaterals and produced dense, uniform innervation restricted to the target layer (Fig.2 A,C). Injections of DiI in reeler embryos revealed labeled entorhinal axons in the outer marginal zone (Fig. 2 B) that were concentrated in a narrow zone near the hippocampal fissure (especially at dorsal levels; data not shown). These axons followed rather straight courses and had few axon collaterals (Fig. 2 B,D). Consistent with this, the density of innervation was always lower in reeler embryos than in wild-type and heterozygous embryos. In addition, in reeler embryos a substantial number of fibers invaded layers other than the marginal zone (the SR and the SP) (Fig. 2 B). These observations indicate that Reelin contributes to the targeting of entorhinal axons.
The pattern of entorhinal termination in wild-type and heterozygous animals remained similar between P2 and P5, although the innervation in the SLM became denser at P5, with axons forming elaborate arbors. Similarly, entorhinal afferents, which began to invade the dentate molecular layer (ML) at P2, had already formed a dense innervation at P5. At both stages, entorhinal axons were restricted to the entorhinal termination layers, with virtually no invasion of the underlying layers such as the SR or SO (Figs. 2 E,F,I, 3). In contrast, in P2–P5 reeler mice, most entorhinal fibers at dorsal hippocampal levels were densely packed in the outer SLM near the hippocampal fissure, where they formed tight axonal bundles (Figs. 2 G,O,3). At more ventral hippocampal levels, entorhinal fibers were also packed near the hippocampal fissure, from which numerous fibers descended to innervate deep layers (Figs. 2 H,J, 3). These fibers innervated the SLM but were also distributed within the SR, the pyramidal layer, and the SO (Figs. 2 H,J, 3). Perhaps the most dramatic abnormality in rl/rl mice was the finding that entorhinal axons running through the alveus (the alvear pathway) (Lorente de Nó, 1934) formed a patch of termination in the SR and SO of the CA2 region (Figs. 2 P, 3), a feature never observed in control mice. In the rl/rl dentate gyrus, entorhinal afferents innervated the OML (Figs.2 J, 3), but misrouted fibers terminating in the inner molecular layer and the hilus were very frequent (data not shown). As at embryonic stages, single fibers in reeler mice displayed fewer axonal branches than those in wild-type and heterozygous animals. Another distinguishing feature of rl/rl entorhinal axons is that they were tipped with very large, complex growth cones with long filopodia and large lamellipodia, which contrasted with the small size of growth cones in control mice (Fig. 2 K–N). Furthermore, the characteristic topographic projections from the medial and lateral entorhinal cortices (see below) began to be recognizable at P5 in rl/+ and wild-type animals but not in rl/rlmutant mice (Fig. 2 I,J).
At P12 the pattern of entorhinal termination was similar to that in adults, and so both ages will be described together. At these ages, biocytin was selectively injected into the medial and lateral entorhinal areas so that the specific topographic projections from these entorhinal subfields could be analyzed (Amaral and Witter, 1995). Entorhinal injections in wild-type and heterozygous animals resulted in a dense innervation restricted to the SLM in the hippocampus proper. Consistent with previous studies (Steward, 1976; Swanson and Cowan, 1977; Ruth et al., 1982; Amaral and Witter, 1995), injections in the lateral entorhinal cortex defined patches of termination in two subzones of the SLM corresponding to the CA1–subicular interface (data not shown) and the CA3–CA2 region (Fig.4 A). Conversely, medial entorhinal projections innervated the SLM in the subiculum and the CA1 region (Fig. 4 C). In reeler mice, lateral entorhinal fibers occupied the SLM uniformly throughout the CA1 and subicular fields, without forming region-specific patches, and with a virtual absence of innervation in the CA3 subfield (Fig.4 B). The innervation of the SLM was narrower inrl/rl (70 ± 8 μm) than in heterozygous mice (125 ± 6 μm).
Injections in the medial entorhinal area in rl/rl mice gave rise to the typical two-patch pattern of termination in the SLM of the CA1 and subiculum (Fig. 4 D). In addition, numerous aberrant fibers remained present in the pyramidal layer and SO of the CA2–CA1 subfields (Fig. 4 D). Moreover, after injections in the lateral and medial entorhinal cortices there were many axons in inappropriate layers, including the SR, SO, and pyramidal layer. These aberrant projections were more frequent at P12 than in the adult.
In the dentate gyrus of +/+ and rl/+ mice, lateral entorhinal projections terminated in the OML, whereas medial entorhinal fibers were restricted to the middle molecular layer (MML) (Fig.4 E,G,I,K). In reeler mice, axons arising from the lateral entorhinal area terminated in an outer band near the hippocampal fissure, which was narrower than in control mice. Medial entorhinal fibers terminated in the MML. However, afferents arriving from both entorhinal subfields gave rise to aberrant fibers that terminated in inappropriate layers of the dentate gyrus, such as the hilus (Fig. 4 F,H,J,L).
This in vivo analysis indicates that reeler entorhinal fibers show aberrant trajectories and termination patterns as well as decreased fiber growth, which appear to be partially corrected as development progresses. Similarly, although the region-specific topographic projections from the lateral and medial entorhinal areas are preserved to a certain extent in reeler mice, they also show abnormal patterns of termination.
Entorhinal axons in reeler mice transiently exhibit reduced axonal arbors
To confirm these findings, a quantitative analysis was undertaken on single entorhinal axons present in the CA1 field and in the dentate gyrus. In the CA1 field, reeler entorhinal fibers had a reduced branching index at P2 and more so at P5 (69.7% less) (Fig.5 B). At this age, entorhino-dentate axons also had fewer axon collaterals in reeler mice than in heterozygous animals (51.2% less). At P12 the differences between reeler and heterozygous mice disappeared in the dentate gyrus, whereas they persisted in the CA1 field. The branching index in adult mutant mice was similar to that in rl/+ mice, in both hippocampal regions.
Entorhinal axon side branches in reeler mice were slightly, but significantly, shorter than in heterozygous mice at all stages examined, except at P5 in the SLM (Fig. 5 C). The lack of a significant difference at P5 may be because this is the period of greatest side branch formation in heterozygous mice (Fig.5 B), and thus most newly formed collaterals may be short. These findings indicate that Reelin promotes the branching and extension of developing entorhinal axons in vivo.
Reeler mutant mice show altered hippocampal synaptogenesis
To determine whether Reelin influences the sequence of synaptogenesis, we first counted the number of axonal varicosities in biocytin-labeled entorhinal fibers (Fig. 5 D). Both in the dentate gyrus and in the CA1 region, axons of reeler mice at P2–P5 had a lower density (34–52% less) of axonal varicosities than heterozygous mice. These differences were reduced at P12, but they were still detectable in adult mutant mice (10–28% reduction). These data suggest that Reelin participates in the hippocampal synaptogenesis.
To examine the role of Reelin in early synaptogenesis, biocytin was injected into the entorhinal cortex, and labeled axons were examined by electron microscopy. Identified entorhino-hippocampal afferents established morphologically mature synaptic contacts at early postnatal stages (P2–P5) in both reeler and heterozygous animals (Fig.6 C,D). In both groups the labeled entorhinal boutons displayed synaptic vesicles clustered near the active synaptic zones, with typical presynaptic and postsynaptic densities. Synaptic contacts were asymmetric, and postsynaptic targets included dendritic shafts and spines. Moreover, tight axonal bundles were frequently observed in the SLM of rl/rl mice (Fig.6 E), whereas this feature was rarely found in heterozygous mice.
To estimate the density of entorhinal synapses we quantified the number of synaptic contacts observed in the SLM and OML of the dentate gyrus (Figs. 6 A,B, 7). At P2–P5 we found a 26–53% reduction in the numbers of synaptic contacts in the SLM and OML of reeler mice, which was more marked at P2. In contrast, no significant differences were observed in the length of the synaptic contacts between the two groups (e.g., at P2, in the SLM, 0.287 ± 0.013 and 0.333 ± 0.018 μm inrl/rl and rl/+ mice, respectively). These findings indicate that Reelin is involved in the regulation of synaptogenesis in the developing hippocampus.
Reelin produced in the target region is responsible for fiber abnormalities in reeler mutant mice
Because reelin transcripts are expressed in both the hippocampus and the entorhinal area (including entorhino-hippocampal neurons) (Fig. 1 B), the fiber abnormalities reported in rl/rl mice could be attributable to the lack of Reelin in either the target region or at the site of axonal origin. Also, the mispositioning of neurons in the entorhinal cortex of reeler mice may influence the pattern of entorhino-hippocampal innervation. To discern among these possibilities we prepared mixed organotypic slice co-cultures obtained from reeler and heterozygous newborn mice. After 7–15 DIV, homogenetic co-cultures (E+/+/H+/+, Erl/+/Hrl/+, Erl/rl/Hrl/rl) displayed patterns of entorhinal innervation similar to those describedin vivo (Fig.8 A,B). Thus, in +/+ andrl/+ co-cultures entorhino-hippocampal fibers formed a dense innervation that completely filled the SLM and OML (Fig.8 A). Fibers displayed numerous axonal varicosities and axon collaterals (Table 1, Fig.9) and were observed only rarely in inappropriate layers.
The pattern of innervation in rl/rl co-cultures was quite different. These co-cultures displayed a compacted entorhinal projection to the dentate gyrus, near the hippocampal fissure, with axons well collateralized and with numerous varicosities. The innervation of the SLM, in contrast, was very scarce, with fibers occupying a narrow zone of termination and giving rise to a reduced density of innervation (Table 1, Figs. 8 B, 9). Entorhinal fibers in the SLM showed straight courses and few axonal side branches. In addition, there were many aberrant entorhinal fibers in the hilus, the SR, and the pyramidal layer (data not shown). Quantitative analyses showed a lower branching index of entorhinal fibers in the SLM, but not in the ML, than in rl/+ co-cultures (Fig. 9). Thus, the abnormalities of the entorhino-hippocampal projection in reeler slice co-cultures are reminiscent of those in reeler mice in vivo. Because entorhino-hippocampal fibers are severed during the preparation of the slice cultures, these results also indicate that Reelin plays a similar role both in developing and in regenerating axotomized entorhino-hippocampal axons.
When heterozygous entorhinal slices were co-cultured with reeler hippocampi, both the pattern of innervation and the characteristics of single axons were identical to those of reeler co-cultures (Table 1). Thus, the entorhinal projection matched the reeler-like pattern described above, and single axons showed less collateral branching in the SLM (Figs. 8 D, 9). In contrast, co-cultures of reeler entorhinal cortex with heterozygous hippocampus resulted in normal patterns of entorhinal innervation (Table 1, Fig. 8 C) and axonal branching (Fig. 9). In neither of the co-culture combinations did we observe major differences when the slices were cultured for 7 or 15 DIV, except that the density of innervation increased after longer incubation. These in vitroexperiments indicate that the abnormalities of the entorhino-hippocampal projection in reeler mice are caused by the lack of Reelin in the target hippocampal region.
Reelin and the formation of layer-specific and topographic connections in the hippocampus
A distinguishing feature of developing hippocampal connections is that ingrowing afferents invade the appropriate target layer when they reach the hippocampus. Thus, entorhinal fibers invade the SLM from the beginning, with no growth into other hippocampal layers, and commissural/associational afferents terminate specifically in the SR and SO from E18 onward (Supèr and Soriano, 1994; Supèr et al., 1998a,b). These findings, together with a recent study showing that entorhinal cells selectively adhere to their termination layers in the hippocampus (Förster et al., 1998), suggest the presence of layer-specific positional cues that target particular hippocampal afferents. In a previous study we showed that CR cells, a special class of pioneer neuron (Soriano et al., 1994; Del Rı́o et al., 1995;Supèr et al., 1998a), have an essential role in the ingrowth and targeting of entorhinal afferents (Del Rı́o et al., 1997). A crucial role for these neurons is supported by experiments in which ectopically placed CR cells perturb the targeting of entorhino-hippocampal fibers (J. A. Del Rı́o, unpublished observations). These studies prompted the present analysis of the function of Reelin, an extracellular protein highly expressed in CR cells (D’Arcangelo et al., 1995, 1997; Alcántara et al., 1998).
The present analyses of reeler mutant mice show that the entorhino-hippocampal pathway is formed in the absence of Reelin, indicating that this protein is not essential for the ingrowth of entorhinal axons to the hippocampus. Thus, in accordance with current views on axonal guidance (Goodman, 1996; Tessier-Lavigne and Goodman, 1996), the early trajectory of entorhinal fibers toward the hippocampus may be mediated by long-range molecular cues, including members of the Netrin and Semaphorin families (Chédotal et al., 1998). The present study also shows that, to a certain extent, entorhinal fibers in reeler mice terminate in the appropriate target layer (but see below). This indicates that, in addition to Reelin, other positional cues may determine the guidance of entorhinal afferents toward their specific termination layers [see also Förster et al. (1998)]. Some of the factors that provide such guidance in other brain regions are members of the ephrin and semaphorin families (Cheng et al., 1995;Drescher et al., 1995; Messersmith et al., 1995; Culotti and Kolodkin, 1996; Brennan et al., 1997; Wang and Anderson, 1997). Indeed, we have recently shown that several secreted Semaphorins and Neuropilin receptors are expressed in the developing hippocampal formation, and that Semaphorin III and IV have strong repulsive effects on both entorhinal and hippocampal axons (Chédotal et al., 1998).
In agreement with earlier studies (Stanfield et al., 1979), we found that the general topography of hippocampal connections is preserved in reeler mice. However, there are a number of marked abnormalities, which include a dramatic increase in ectopic fibers innervating inappropriate layers and the formation of aberrant projections from the medial entorhinal cortex to the SR and SO of the CA2 region. In addition, the lateral entorhinal projection to the SLM of the CA3 region is clearly absent in these mutant mice. The CR-50 blocking studies on wild-type co-cultures, demonstrating a reeler-like axonal phenotype without disruption of hippocampal layers, indicate that the fiber abnormalities in reeler mice are unlikely to be caused by the mispositioning of hippocampal neurons (Del Rı́o et al., 1997). Thus, there are at least two possible explanations for these abnormalities: (1) Reelin contributes to the layer-specific and topographic targeting of entorhinal axons via direct interaction with developing fibers, or (2) extracellular Reelin facilitates the mechanism of action of other, crucial factors that provide positional information. Although further studies are clearly needed to discern between these possibilities, we favor a direct interaction of Reelin with growing axons, as suggested by our finding of abnormally large, complex growth cones in reeler mice.
Reelin and the elaboration of axonal arbors
One remarkable finding of this study is that, in the absence of Reelin, entorhinal axons form tight bundles and have decreased axonal branching and elongation of side collaterals. Although Reelin has been proposed to provide a stop signal for migrating neurons (Ogawa et al., 1995; D’Arcangelo et al., 1997), the present study on the formation of hippocampal connections in reeler mice does not support a role for Reelin as a repulsive signal for developing entorhinal axons. On the contrary, extracellular Reelin appears to favor the growth, extension, and branching of entorhinal axons in their termination layers. It remains to be established whether Reelin only provides a permissive substrate for axon elongation or whether it signals the formation and growth of collateral branches.
Other molecules, including Ng-CAM, fasciclin II, polysialic acid, ephrins, semaphorins, receptor protein tyrosine phosphatases, and neurotrophins, contribute to axonal bundling, defasciculation, and axonal branching (Tang et al., 1992; Lin et al., 1994; Cabelli et al., 1995; Cohen-Cory and Fraser, 1995; Matthes et al., 1995; Püschel et al., 1995; Stoeckli and Landmesser, 1995; Winslow et al., 1995;Desai et al., 1996; Fujisawa et al., 1997). This suggests that there may be redundancy of signals regulating these processes, which may explain the finding that some fiber abnormalities in reeler mice are more dramatic at perinatal stages than in the adult. For instance, the partial recovery of fiber abnormalities in reeler mice from P12 onward may be caused, at least in part, by the action of neurotrophins BDNF and NT-3, which also influence the branching of hippocampal afferents at comparable stages (Martı́nez et al., 1998).
Reelin and synaptogenesis of hippocampal connections
Although reeler entorhinal axons appear to form normal axonal varicosities with synaptic vesicles and morphologically mature synaptic contacts, their number is dramatically reduced. The reduced number of synaptic contacts might result simply from the decreased branching and maturation of single entorhinal axons in these mutant mice. However, the observation that reeler entorhinal fibers have fewer axonal varicosities along their length points to a direct effect of Reelin in the formation of synapses. Little is known about the molecular factors that regulate synapse formation and the number of synaptic inputs. Very recently, it has been proposed that neurotrophins and N-cadherins may regulate the density of synaptic innervation (Snider and Litchman, 1996; Causing et al., 1997; Inoue and Sanes, 1997; Martı́nez et al., 1998). Thus, like other processes contributing to the development of neuronal connections, the formation of synapses appears to be multifactorial and dependent on a combination of factors.
Reelin, an extracellular protein involved in neuronal migration and axonal growth
Recent data have shown that the same factors may play a role in both neuronal migration and axonal growth. For example, netrin-1 exerts chemoattractive or chemorepulsive effects on both developing axons and migrating neurons (Serafini et al., 1996; Wadsworth et al., 1996). The strong abnormalities apparent in reeler mice demonstrates that Reelin plays a role in neuronal migration (D’Arcangelo et al., 1995; Ogawa et al., 1995; Goffinet, 1997). However, Reelin is a large extracellular protein with distinct domains, which suggests that it may be involved in different processes (D’Arcangelo et al., 1995). Here we provide evidence that Reelin participates in the targeting of topographic connections, in the branching and growth of single fibers, and in the synaptogenesis of hippocampal afferents. The manner in which Reelin contributes to these developmental processes is unknown, but the growth of entorhinal axons onto a Reelin-rich termination zone might suggest molecular interactions between target-derived Reelin and putative Reelin-binding proteins present in developing axons. Recent studies have shown that mutations in genes encoding signal transduction-associated proteins, including the cdk5,p35, and mdab1 genes, lead to deficits in neuronal migration resembling those of reeler mice (Ohshima et al., 1996; Chae et al., 1997; Goldowitz et al., 1997; González et al., 1997; Howell et al., 1997a; Sheldon et al., 1997). The recent observation of mDab1 protein in fiber tracts during ontogenesis (Howell et al., 1997b; Rice et al., 1998) suggests that these proteins may also act in signal transduction pathways triggered by Reelin in developing axons.
This work was supported by The Marató de TV3 Foundation, the Direccion General de Investigación Científica y Técnica (Grant P.M.95-0102) and Comisión Interministerial de Ciencia y Tecnologı́a, Spain (Grant SAF98-0106), the Ramón Areces Foundation and the International Institute for Research in Paraplegia (E.S.), by the President’s Special Research Grant (The Institute of Physical and Chemical Research) and the Ministry of Education, Science, and Culture of Japan (K.N.), National Institutes of Health Cancer Center Support CORE grant (T.C.), a grant from the National Institute of Neurological Diseases and Stroke (NINDS) (T.C.), the American Lebanese Syrian Associated Charities (ALSAC) (G.D., T.C.), and a National Research Service Award from NINDS (G.D.). V.B. was supported by a Comisió Interdepartamental de Recerca i Tecnologia-FPI fellowship. We thank R. Rycroft for editorial assistance.
Correspondence should be addressed to Dr. Eduardo Soriano, Department of Animal and Plant Cell Biology, Faculty of Biology, University of Barcelona, Av. Diagonal 645, Barcelona 08028, Spain.