Ephrins regulate the formation of terminal axonal arbors during the development of thalamocortical projections

One of the basic features of the mammalian neocortex is its lamina-specific connectivity both with subcortical structures and other cortical areas. For example, in the primary somatosensory area (S1) of the cortex, thalamocortical projections arising from the ventrobasal nucleus (VB) terminate in layer 4 and extend collaterals in layer 6, skipping layer 5. The question of how these connections are established has been examined in the developing cortex of various species, and results indicate that the laminar pattern of thalamocortical projections develops precisely from the outset (Bicknese et al., 1994; Ghosh and Shatz, 1992; Lund and Mustari, 1977; Miller et al., 1993; Agmon et al., 1993; Catalano et al., 1996; Kageyama and Robertson, 1993). In the murine brain, the first thalamic axons enter the cortex at embryonic day (E) 14, and grow along a pathway centered on the subplate layer underlying the undifferentiated cortical plate (Bicknese et al., 1994). Once they reach their target area of the cortex, thalamic axons adopt a radial growth and invade progressively the differentiating cortical layers. At birth, thalamocortical axons extend though developing layers 6 and 5 and start branching within layer 6 (Agmon et al., 1993). They then stop their radial growth at postnatal day (P) 2 after reaching layer 4, form terminal branches and elaborate synaptic contact with their appropriate target cells (Agmon et al., 1993). Over the following week, axons emit more branches within layer 4, but the extent of terminal arbors remains within the width of layer 4 and does not extend across into the adjacent non-target layers (Agmon et al., 1993).

In vitro studies have implicated intracortical cues in directing the lamina-specific growth of thalamic axons. It has been proposed that guidance of thalamic afferents in the subplate results from the expression of growth-promoting factors along their pathway (Götz et al., 1992; Henke-Fahle et al., 1996; Hübener et al., 1995; Kinnunen et al., 1999) and inhibitory signals in the undifferentiated cortical plate (Emerling and Lander, 1996; Tuttle et al., 1995). Later on, upregulation of growth-promoting molecules in the cortical plate allows thalamic axons to enter the cortical layers (Götz et al., 1992; Hübener et al., 1995). The observation of precise layer-specific targeting of thalamocortical projections in organotypic thalamus-cortex co-cultures (Bolz et al., 1992; Molnar and Blakemore, 1991; Yamamoto et al., 1992; Yamamoto et al., 1997) have led to the hypothesis that molecular signals confined to individual cortical layers allow thalamic axons to distinguish between target and non-target cells. In support of this idea, in vitro experiments revealed the existence of layer-specific cues that regulate the assembly of intrinsic cortical circuits (Castellani and Bolz, 1997; Dantzker and Callaway, 1998). However, the molecular identity of the factors that serve this function remains to be elucidated.

Over the past few years, a large number of cell surface and secreted molecules have been identified that orchestrate the development of specific connections in the central nervous system. Among them, the ephrins is a family of putative axon guidance ligands that fall into two subclasses: glycosylphosphatidylinositol (GPI)-linked ephrin-As and transmembrane ephrin-Bs. These interact respectively with A-type and B-type Eph tyrosine kinase receptors. Eph receptors and ephrins have been implicated in a range of developmental processes, including topographic mapping, brain commissure formation and axon guidance at the midline (reviewed by Klein, 2001; Wilkinson, 2001). Recent studies have provided evidence for a key role of ephrins in guiding thalamocortical axons to their appropriate target cells in the developing cortex. A role in regulating topographic mapping of thalamocortical projections within the primary somatosensory area of the neocortex has been proposed based on the analysis of ephrin-A5 knockout mice (Vanderhaeghen et al., 2000). It has also been suggested that ephrin-A5 could function in the patterning of thalamocortical connections between areas of the limbic cortex and neocortex (Gao et al., 1998; Mackarehtschian et al., 1999). In addition, members of the ephrin family may play a role in regulating the laminar pattern of local circuits within the neocortex (Castellani et al., 1998) and the laminar termination of entorhinal afferents in the hippocampus (Stein et al., 1999). Together, these data suggest that ephrins are candidate molecules for providing information for thalamic afferents during targeting to the proper layer in the cortex.

In the present study, we used different in vitro assays to explore the influence of layer-specific cell surface molecules on selective growth and targeting of thalamic axons. We show that multiple membrane-bound signals, acting either as attractive or repulsive cues, cooperate to specify the patterning of thalamocortical projections. Furthermore, our results furnish evidence suggesting that ephrin ligands stimulate branch formation of thalamic axons and thereby may contribute to the formation of terminal arborizations in their target layer.

Embryos at E14 (day of insemination=E1) were removed from anesthetized pregnant OF1 (IFFA-Credo, Lyon, France) and NMRI albino mice (from Harlan Winkelmann, Borchen, Germany), and the brains were transferred in Gey’s Balanced Salt Solution (GBSS) supplemented with 6.5 mg/ml glucose. Neurons in the thalamus are generated following a lateral to medial gradient over a period of time extending from E11 to E16 in the mouse (Angevine, 1970). At E14 most of the neurons in the thalamus are destined to lateral nuclei that provide inputs to the neocortex. In some experiments, medial (limbic) nuclei of the thalamus were isolated from E16 mouse fetuses as previously described (Mann et al., 1998).

To prepare thalamic explants, blocks of diencephalic tissue were isolated and cut in 200 μm³ explants with a McIlwain tissue chopper. To prepare dissociated thalamic cells, thalami were excised and collected in ice cold Hank’s Balanced Salt Solution (HBSS) containing 6.5 mg/ml glucose and 0.025% trypsin, then incubated at 37°C for 25 minutes. Thalamic tissue was dissociated into single cells and filtered through a nylon gauze to remove cell aggregates. Cells were centrifuged at 8 g for 8 minutes and cell density was adjusted to 2×10⁶ cells/ml.

Mice at P6-8 (day of birth=P0) were decapitated and the brains were transferred in GBSS with 6.5 mg/ml glucose. The dorsal neocortex was dissected and cut in 200 μm slices with a McIlwain tissue chopper. Cortex slices from the primary somatosensory area in which a barrel pattern in layer 4 was clearly visible were selected. Cortical layer 4 and layer 5 were isolated as described previously (Castellani and Bolz, 1997; Götz et al., 1992).

NIH3T3 cell line transfected with a retroviral vector, pLIG*, containing the human ephrin-A5 gene (EPHA5) or, as a control, transfected with the vector alone (Gao et al., 1996) were grown in the following medium: DMEM/F12 (1:1), 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin.

Membranes from isolated cortical layers and confluent NIH3T3 cells were prepared as described previously (Mann et al., 1998). To prepare homogenous substrates of membrane, pairs of glass coverslips were coated with 1.9 μg of laminin and 0.1 μg of poly-L-lysine in 100 μl GBSS as a ‘sandwich’ for 1 hour at 37°C, then incubated for 2 hours at 37°C with 100 μl of membrane suspension used at a concentration of 60 μg/ml. In some experiments, a 1:1 mixture of layer 4 and layer 5 membranes was used as a substrate for neurite growth. In the latter case, the total membrane concentration was adjusted to 120 μg/ml. Thus, the amount of target membranes was unchanged when layer 4 membranes were used individually or mixed with layer 5 membranes. For stripe assay, polycarbonate filters containing alternating lanes of two membrane species were prepared according to Walter et al. (Walter et al., 1987). In some cases, membrane stripes were transferred on glass coverslip as described by Wizenmann et al. (Wizenmann et al., 1993). For the enzyme treatment, membranes were incubated with phosphatidylinositol-specific phospholipase C (PI-PLC; 3 U/5 mg proteins) for 1 hour at 37°C and washed in phosphate-buffered saline (PBS) before use in substrate preparation. Control membranes were processed in the same way but without PI-PLC. Coated coverslips were placed in Petriperm dishes in 750 μl of medium (60% Eagle’s basal medium, 30% HBSS and 10% fetal bovine serum, supplemented with 0.1 mM glutamine, 1 mg/ml glucose, 4 mg/ml methylcellulose, 100 U/ML penicillin and 100 μg/ml streptomycin). Thalamic explants or dissociated thalamic cells were pipetted onto the substrate and incubated for 10-30 minutes at 37°C before adding 1.25 ml of medium. In some experiments, purified ephrin-A5-Fc fusion protein was added to the culture medium at a final concentration of 2 μg/ml. Cultures were grown at 37°C in an incubator with a humid atmosphere containing 5% CO₂ and fixed after 2 days in 2 ml of 4% paraformaldehyde with 30% sucrose.

To visualize processes extending from the dissociated thalamic cells and to confirm their neuronal origin, thalamic cultures were immunostained with the neuron-specific marker SMI 312 (Sternberger Monoclonals, Lutherville) using the following protocol. After washing three times for 20 minutes in 0.5% Triton X-100/PBS, thalamic cell cultures were incubated with the first antibody SMI 312 at 1:1000 in blocking solution (0.5% Triton X-100, 1% bovine serum albumin in PBS) for 48 hours at 4°C. After washing, cells were incubated with the secondary antibody (Cy3-conjugated goat anti mouse IgG (H+L), Jackson Immuno Research) diluted at 1:1000 together with 50 ng/ml DAPI in blocking solution for 24 hours at 4°C. Cultures were washed and mounted in moviol containing n-propylgallate.

The number of axons extending from thalamic explants was counted under an inverted microscope with 20× phase-contrast objective [Zeiss Plan-Neofluar, numerical aperture (NA) 0.50]. Because fiber outgrowth is very dense close to the explant, only main axons extending on the substrate were taken into account. To estimate axonal elongation, the five longest axons extending from each explant were measured from the external border of the explant to the tip of their growth cone. Statistical comparison between average number of axons and average axonal length under different culture conditions were determined using a two-tailed Student’s t-test.

Axonal arborization was examined with a 40× phase-contrast optic (Zeiss Plan-Neofluar, NA 0.75) in combination with additional magnification lenses (1.6× Optovar). For each explant, five isolated and unfasciculated axons were randomly chosen. Axonal length was measured and the number of side branches was counted. In the stripe assays with dissociated neurons, SMI 312 labeled axons were traced using a high-power continuous-focus camera lucida, and we reported on the same drawings, the position of the stripe boundaries. Side branches were counted for single axons which extended about orthogonal to the membrane stripes and crossed at least two membrane lanes. The branching density was calculated as the ratio between total number of branches and total axonal length. Statistical differences were determined with a Fisher’s permutation tests.

Finally, axon guidance in stripe assays was assessed by counting for a pair of stripes the number of axons extending on each type of membrane species. Axonal numbers were compared with a two tailed Student’s t-test and the data are presented in percentage. Values reported in the present study represent results collected in two to four independent experiments.

For in situ hybridization, fresh-frozen brains of P6 animals were coronally sectioned into 20 μm thick slices. Alternating cryostat sections containing the barrel field of the somatosensory cortex were stained with Cresyl Violet, stained for cytochrome oxidase and used for in situ hybridization respectively. For in situ analysis, sections were thaw-mounted onto Superfrost Plus slides (Menzel Gläser, Germany), dried at 55°C for 3 hours and fixed with 4% paraformaldehyde in PBS for 10 minutes at room temperature. Permeabilisation was done in 0.2 M HCl for 10 minutes. Sections were digested with 1 μg/ml proteinase K (Roche) in 0.1 M Tris HCl (pH 8.0) for 5 minutes at 37°C and treated with 5 mM acetanhydride in 0.1 M triethanolamine (pH 8.0) for 10 minutes at room temperature. Between each of these steps, slices were rinsed with PBS. Digoxigenin-labeled riboprobes were made by in vitro transcription from pBlueskript SK(–) hephnA5 kindly provided by J. G. Flanagan and used for hybridization at a final concentration of 3 ng/μl (Vanderhaegen et al., 2000). To allow optimal tissue penetration, sense and antisense riboprobes were hydrolyzed at 60°C for 32.5 minutes. The approximate length of the hydrolyzed probes was 0.2 kb. The in situ hybridization was performed as described by Weth et al. (Weth et al., 1996). Only background staining was obtained when the sense probe was used as a negative control.

To assay the expression of EphA receptors and ephrin-A ligands on thalamic axons and in cortical layers, supernatant from COS cells expressing the zebrafish ephrin-A5 ectodomain fused to alkaline phosphatase (ephrin-A5-AP) (kindly provided by Caroline Brennan) and the EphA3-Fc fusion protein (R&D Systems) were used as described (Flanagan and Leder, 1990; Marcus et al., 2000).

We used an in vitro strategy to examine the molecular mechanisms underlying lamina-specific termination of thalamocortical projections in the primary somatosensory cortex. As a first step, we asked whether membrane-bound molecules expressed in the cortical target layer of thalamocortical axons provide signals that cause thalamic afferents to stop their growth and elaborate terminal arbors. To this end, thalamic explants harvested from E14 mouse embryos were cultured on membrane substrates from isolated cortical layers. For technical reasons, cortical layers were dissected at P6-8, shortly after the initial growth of thalamic axons in layer 4, when cortical layers are well differentiated. Previous co-culture experiments have shown that the factors allowing target recognition are present in the cortex during the first postnatal week (Götz et al., 1992). Fig. 1 shows photomicrographs of typical growth patterns observed on membranes from layer 4, the target layer for thalamic inputs, and on membranes from layer 5, a non-target layer that thalamic axons normally cross without branching. Thalamic explants exhibited differential growth patterns on both membrane substrates, with a reduced outgrowth on membranes from layer 4 compared with layer 5. Moreover, examination of individual fibers at high magnification indicated a profuse branching of those axons that grew on layer 4 membranes. To quantify these observations, mean number of fibers per explant and average axonal length were measured. On target membranes, we found 44% less thalamic fibers (Fig. 2A) and a 19% reduction in axonal length (Fig. 2B) compared with non-target membranes. Consistent with our qualitative observations, measurements of branching frequency indicated that thalamic axons emitted 57% more side branches on membranes isolated from layer 4 than on membranes from layer 5 (Fig. 2C).

In a second set of experiments, dissociated thalamic cells were placed on a substrate consisting of alternating stripes of membranes from cortical layers 4 and 5. This assay allowed us to analyze branch formation of single thalamic axons that grew perpendicular to the membrane stripes and thus encountered successively non-target and target substrates, as they normally do in vivo. As shown in Fig. 3, thalamic axons extended on both type of membranes and freely crossed the stripe borders, but branch formation was higher on stripes from target layer than from non-target layer (Fig. 3A,B). Quantitative analysis indicated a 35% increased in branching density on layer 4 membranes compared with layer 5 membranes (Fig. 2D). Taken together, our findings indicate the presence of cell surface cues restricted to individual cortical layers that control the elongation and branch formation of thalamocortical axons.

Recent studies have provided evidences for a role of GPI-anchored ephrin ligands as topographic guidance cues in the developing thalamocortical system (Gao et al., 1998; Mackarehtschian et al., 1999; Vanderhaeghen et al., 2000). We were interested in further exploring the possibility that Eph ligands might also function in setting up the layer specificity of thalamocortical connections. Vanderhaeghen et al. (Vanderhaeghen et al., 2000) analyzed ephrin-A5 mRNA expression from E18 to P3 in mouse brains, and they report that at postnatal stages labeling in the somatosensory cortex was strongest in layer 6 and layer 4, with only moderate expression in layer 5 especially in the motor areas. We examined whether ephrin-A5 is also present at later postnatal stages, when thalamocortical arbors are still being elaborated. For this in situ hybridization on coronal sections from P6 brains was performed to detect ephrin-A5 mRNA. Adjacent sections were stained with Cresyl Violet to identify the cortical layers and stained for cytochrome oxidase to reveal the barrel pattern in S1, the primary somatosensory cortex. As illustrated in Fig. 4A, at P6 ephrin-A5 mRNA is clearly detectable in layer 4 and there is also a somewhat weaker staining in layer 6 of S1. In the adjacent motor cortex, ephrin-A5 mRNA is most prominently expressed in some large cells in layer 5. Ephrin-A ligand protein distribution in cortical layers was determined by affinity probe in situ with an EphA3-Fc fusion protein. In P6 somatosensory cortex, EphA3-Fc binding activity was restricted to layer 4 and layer 6, similar to ephrin-A5 transcripts (Fig. 4B).

Finally, to detect EphA receptors on thalamic axons, we used a probe consisting of the ephrin-A5 ectodomain fused to alkaline phosphatase to label thalamic explants from E14 animals in vitro. Labeling was seen on virtually all thalamic neurites (Fig. 4C), confirming the presence of EphA receptors on growing thalamic axons and growth cones.

The detection of ephrin-A ligands in the cortical layers where thalamocortical axons elaborate branches suggest a role of ephrins as branching factors for thalamic afferents. To address directly a possible function of ephrin-A5 in stimulating branch formation for thalamic axons, we compared the branching pattern of thalamic fibers on membrane substrates from NIH3T3 cells expressing the transfected ephrin-A5 ligand or from control cells. As depicted in Fig. 5, there was a strong effect of the recombinant ephrin-A5 on the branching behavior of thalamic fibers. On a total number of 181 axons examined, there was an overall increase by 101% of the branching density on ephrin-A5 containing membranes compared to control membranes (Fig. 6C). By contrast, we found no significant effect of ephrin-A5 on the average growth of thalamic axons (Fig. 6A,B). However, previous experiments showed that a subset of fibers from the ventrobasal nucleus of the thalamus was repelled by ephrin-A5 in vitro (Vanderhaeghen et al., 2000). We found that short thalamic axons (<100 μm), which may represent a subset of fibers that is sensitive to the inhibitory activity of ephrin-A5, exhibited a similar branching density (39.1±4.9 branches/mm, n=29) as did axons that grew longer on NIH3T3-ephrin-A5 membranes (>300 μm, 34.0±4.3 branches/mm, n=33). This suggests that all or most of the thalamic axons showed a strong branching response to ephrin-A5 that was independent of their growth rate on the ephrin-A5 substrate. This branching response was abolished after removal of ephrin-A5 from NIH3T3 cell membranes using the enzyme PI-PLC (Fig. 6D). These data demonstrate that ephrin-A5 acts as a branch promoter for thalamic axons in vitro.

To test for a possible functional role of ephrins in the patterning of thalamocortical projections, we used two different in vitro approaches. First, we examined the growth and branching of thalamic axons on membranes from layer 4 and layer 5 after removal of GPI-anchored molecules with the enzyme PI-PLC. Enzyme treatment increased the growth of thalamic axons both on layer 5 and layer 4 membranes, compared to control substrates (Table 1). However, the differential growth of thalamic explants on target and non-target membranes was unaffected, as axons grew shorter on treated layer 4 than on treated layer 5 membranes (Table 1). On PI-PLC treated layer 4 membranes, branching of thalamic fibers, decreased by 43% compared to control layer 4 membranes (Fig. 7A). By contrast, elimination of GPI-linked molecules did not affected axonal sprouting on layer 5 membranes (Fig. 7A). This indicates that GPI-anchored branching factors are expressed in the target layer for thalamocortical fibers.

Second, to test whether these GPI-anchored factors belong to the subfamily of GPI-linked ephrin-A ligands, thalamic explants were grown on cortical membranes in the presence of soluble ephrin-A5-Fc fusion protein, added to the culture medium at 2 μg/ml. Exogenous ephrin-A5-Fc did not affect the general growth rate of thalamic explants (Table 1) and, unlike membrane-bound ephrin-A5 (which elicited a strong branch promoting activity on thalamic axons), soluble ephrin-A5-Fc did not increase thalamic axon branching. This suggests that membrane attachment (or higher cluster forms) might be required for ephrin-A5 to act as a branching factor. Instead, we found that in the presence of ephrin-A5-Fc the branching rate of thalamic fibers on layer 4 membranes decreased by 48%, and was similar to the branching level observed on layer 5 membranes with or without ephrin-A5-Fc (Fig. 7B). This finding suggests that ephrin-A5-Fc interferes with normal Eph/ephrin interactions in cultures of thalamic axons on their target membranes. The inability of thalamic axons to preferentially branch on layer 4 membranes could be explained by the internalization of EphA receptors after activation by ephrin-A5-Fc and the subsequent desensitization of thalamic axons to ephrin-A cues expressed in their target membranes (Ciossek et al., 1998).

The experiments described above indicate that membrane-associated factors promote the branching of thalamocortical axons in their target layer. It is also possible that additional signals in layer 5 may prevent thalamic afferents to branch in this inappropriate layer. Such an inhibition of branch formation in inappropriate territories has already been demonstrated in the formation of topographic retinotectal projection (Roskies and O’Leary, 1994). To address this issue, thalamic explants were confronted with a substrate consisting of a 1:1 mixture of layer 4 and layer 5 cortical membranes (see Materials and Methods). Under this condition, the branching density of thalamic axons was not significantly different from the branching observed on layer 5 membranes alone (Fig. 7C). This result provides evidence for the presence of inhibitory signals in non-target layer 5 membranes that mask or reduce the branching properties of layer 4 membranes.

Terminal branches of thalamic axons are confined to layer 4 and do not extend across into the adjacent non-target layers, suggesting that mechanisms operate that guide axon collateral extension within the boundaries of their appropriate territory. To test this hypothesis, the in vitro stripe assay was used to offer growing thalamic axons a choice between alternating lanes of membranes from their target layer 4 and non-target layer 5. We observed that thalamic axons that grew parallel to the membrane stripes tend to restrict their growth on their target substrate. On a total number of 126 pair of stripes and 586 axons examined, we found that 339 fibers (58%) grew on membranes from cortical layer 4 and only 247 (42%) on layer 5 membranes (Fig. 8A,B). PI-PLC treatment of membranes from layer 5, but not from layer 4, leads to an equal distribution of growing axons on target and non-target membranes (Fig. 8B). As enzyme treatment of non-target membranes abolished the preferential growth of thalamic axons on their target membranes, we postulated that GPI-anchored molecule(s) in layer 5 act as repulsive guidance cue(s) for thalamocortical axons.

All experiments described so far were performed with explants from somatomotor thalamic nuclei that project to the neocortex (see Materials and Methods). In a previous study, it was demonstrated that the Eph ligand ephrin-A5 selectively inhibited outgrowth of a sub-population of thalamic neurons that project to the limbic cortex (Gao et al., 1998). We were interested in further examining the response of limbic thalamic axons to recombinant ephrin-A5 and to compare it with the behavior of non-limbic fibers described above. In accordance with the results of Gao et al. (Gao et al., 1998), we found a 13% reduction in the length of limbic thalamic axons on NIH3T3-ephrin-A5 cell membranes compared with control membranes (Table 2). In addition, in the stripe assay, limbic thalamic fibers extended preferentially on control lanes and avoided the membrane stripes containing recombinant ephrin-A5 (Fig. 9B), indicating that ephrin-A5 functions as a repellent guidance cue for limbic thalamic axons. We found no guidance effect of ephrin-A5 on fibers extending from explants prepared from the whole non-limbic thalamus (Fig. 9A). Finally, whereas in the stripe assay ephrin-A5 exhibits a repellent guidance activity on limbic fibers, on homogenous membrane preparations we found that ephrin-A5 activates branch formation of limbic thalamic axons, as measured by a 63% increased branching on NIH3T3-ephrin-A5 membranes versus control membranes (Table 2).

The layer-specific patterning of thalamocortical projections develops with a high degree of specificity. The present in vitro experiments provide evidence for molecular cues selectively expressed in either target or non-target layers that regulate elongation, branching and targeting of thalamic axons. Blocking endogenous ephrin-A ligands abolishes the preferential branching of thalamic axons on substrates from their target layer, while ephrin-A5, which is expressed in layer 4 of the somatosensory cortex, induces collateral formation of thalamic axons. This suggests that ephrin-As are instrumental for the formation of terminal arborizations of thalamic axons in their cortical target layer.

Previous experiments revealed that thalamic axons terminate in layer 4 of cortical slice cultures (Bolz et al., 1992; Götz et al., 1992; Molnar and Blakemore, 1991; Yamamoto et al., 1992). Together with the fact that thalamic growth cones often pause and transiently collapse at the time they reach layer 4 (Yamamoto et al., 1997), these observations support the argument that thalamic fibers recognize a ‘stop signal’ expressed at the surface of their target cells in layer 4 (Molnar and Blakemore, 1991). Götz et al. (Götz et al., 1992) have shown that axonal outgrowth from thalamic explants on membrane substrates from layer 4 is reduced compared with the outgrowth on membranes from the deep cortical layers. The present study confirms this result.

Shortly after thalamic axons have ceased their extension in their target layer, lateral branches emerge at a short distance behind the leading growth cone (Yamamoto et al., 1997). This process, called backbranching, was initially described in the frog visual system (Harris et al., 1987), and appears to be a common mechanism used to elaborate terminal arbors. We show that cultured thalamic axons form more lateral branches on membranes from their target layer than on membranes from layer 5, a non-target layer. Our experiments implicated the activity of both branch-promoting cues in layer 4 and branch-inhibiting cues in layer 5. It is thought that the formation of the filopodia-like extensions we observed along the shaft of thalamic axons represents the initial stage of new branch formation (Davenport et al., 1999). In our culture system, however, these lateral extensions remained short (2-10 μm), with only a few of them developing a new growth cone, suggesting that additional cues might be necessary in vivo for the stabilization and the expansion of these branches into terminal arborization.

Removal of GPI-linked molecules and addition of exogenous ephrin-A5-Fc strongly inhibited thalamic axon branching on membranes from their target layer, but did not interfere with the differential outgrowth of thalamic explants on membranes from layer 4 and layer 5. Thus, growth arrest and arborization in layer 4 are separate processes controlled by independent molecular mechanisms. This might be a common principle in the development of layer-specific neuronal projections. In the tectum, for example, cues for the laminar-specific termination of retinal axons comprise N-cadherin and Vicia villosa agglutinin-B4 binding glycoconjugates, whereas neurotrophins regulate the size of retinal arbors without affecting their laminar distribution (Inoue and Sanes, 1997). Similarly, in the hippocampus, the extracellular molecule Reelin acts as a branching factor for entorhinal afferents. However, in the absence of Reelin, entorhinal axons still terminate in the appropriate layer (Borrell et al., 1999; Del Rio et al., 1997).

When confronted with a choice between target and non-target substrates, thalamic fibers are able to distinguish between layers 4 and 5 membranes and progressively restrict their growth on their appropriate substrate. Thus, although on uniform substrates the outgrowth of thalamic axons is reduced on layer 4 membranes compared with layer 5 membranes, when given a choice the axons prefer to grow on their target layer substrate. A similar behavior has been observed in co-cultures of thalamus and cortex: thalamic axons entering the lateral side of a cortical slice follow a pathway centered on layer 4 and avoid the adjacent non-target layers (Yamamoto et al., 1992). Our results with stripe assays indicate that this preferential growth on target membranes is at least in part due to repulsive guidance signals expressed in non-target layers.

Taken together, the present results provide evidence for the existence of multiple membrane-associated molecules expressed in individual layers that influence the growth and branching of thalamic axons. The cooperation of positive and negative cues allows the layer-specific termination of thalamocortical afferents and restricts the extension of terminal arbors within the appropriate target layer.

The molecular nature of factors involved in setting up layer specificity of thalamocortical connections is not known. In experiments with soluble ephrin-A5-Fc, we showed that branching of thalamic axon is strongly reduced on membranes from cortical layer 4, consistent with the idea that A-type ephrin ligands mediate branch formation of thalamic afferents in their appropriate target layer. This result is reminiscent of previous studies showing that several ephrin-A ligands act as branch factors for cortical, hippocampal and retinal neurons in vitro (Castellani et al., 1998; Gao et al., 1999; Davenport et al., 1999). Interestingly, ephrin-A5 was found to specifically promote branching of cortical axons that target the cortical layer 4, whereas cortical axons that normally avoid layer 4 show no branching responses to ephrin-A5 (Castellani et al., 1998). Growth cone collapse induced by ephrin-As was shown to be the leading event that initiates the formation of backbranches along the axonal shaft (Davenport et al., 1999). However, a separate study has demonstrated that ephrin-A ligands can also inhibit branch formation in the retinotectal system (Yates et al., 2001).

Our experiments do not indicate which individual ephrin-A ligand or which specific combination of ligands is likely to operate during thalamic axon branching in layer 4 of the cortex. As recombinant ephrin-A5 stimulates thalamic axons sprouting in our culture system, it is tempting to speculate that ephrin-A5 influences the formation of terminal arborization in vivo. There is some controversy, however, about the laminar expression of ephrin-A5 in the developing cortex. Studies by Castellani et al. (Castellani et al., 1998) and Vanderhaeghen et al. (Vanderhaeghen et al., 2000) in the mouse primary somatosensory area reported ephrin-A5 expression in cortical layers 6 and 4, whereas Yabuta et al. (Yabuta et al., 2000) described that ephrin-A5 is most intensively distributed in the deep cortical layers but not in layer 4. In the present study, we re-examined ephrin-A5 expression in the mouse barrel cortex with a more sensitive in situ hybridization technique using hydrolyzed riboprobes. Our results confirmed expression in cortical layers innervated by thalamic axons (layers 6 and 4), suggesting a possible involvement of ephrin-A5 in thalamic afferent branching. ephrin-A5^–/– (Efna5^–/–) knockout mice show no gross defects in layer-specific thalamocortical targeting in S1 (Vanderhaeghen et al., 2000), and anterograde tracing of VB afferents revealed terminal arbors forming correctly in layer 4 (Muehlfriedel et al., 2000). However, VB terminal arbors are less complex and contain fewer branches in the ephrin-A5^–/– mice than in control mice (Muehlfriedel et al., 2000) (D. Uziel and J. Bolz, unpublished), indicating that ephrin-A5 regulates branch formation of thalamocortical axons in cortical layer 4.

Supporting the idea that ephrin ligands might play a role on thalamocortical branch formation in other areas of the neocortex, studies investigating the developmental expression of Eph ligands in rodents have reported significant expression of ephrin-As across sensory and motor areas (Mackarehtschian et al., 1999; Vanderhaeghen et al., 2000). In some neocortical regions, however, expression was also found outside layer 4 (Mackarehtschian et al., 1999; Vanderhaeghen et al., 2000) (present study). How can these observations be integrated into the idea that ephrin-A ligands are involved in layer-specific branching of thalamic afferents? One possibility is that branch-inhibiting factors co-expressed in non-target layers might mask the branch-promoting activity of ephrin-A ligands. We show that signals present in membranes from layer 5 are sufficient to interfere in vitro with the increased branching of thalamic axons induced on membranes from layer 4, a result that is consistent with this interpretation. Potential candidates for such inhibitory function are the semaphorins, a family of cell-surface and secreted molecules that can function as inhibitors or chemorepellents. In the developing cortex, several members of the semaphorin family exhibit a layer-specific distribution restricted to deep layers 6 and 5 and superficial layers 2/3 (Skaliora et al., 1998). As some semaphorins also inhibit axonal arborization (Bagnard et al., 1998; Kolodkin et al., 1992; Matthes et al., 1995), it has been suggested that they might prevent ingrowing thalamic fibers to form and extend collaterals in inappropriate cortical layers (Skaliora et al., 1998).

As suggested previously, we find that the Eph ligand ephrin-A5 selectively inhibits outgrowth of neurons from limbic thalamus (Gao et al., 1998). Moreover, results from stripe assay experiments in this study indicate that ephrin-A5 also serves as a repulsive guidance cue for limbic thalamic axons, but not for non-limbic thalamic axons. How might the selective effect of ephrin-A5 on axon guidance be related to the development of thalamocortical projections in vivo? During embryonic development, ephrin-A5 expression occurs predominately as a gradient across the subplate zone and the cortical plate of sensory and motor cortex, but there is very little expression in the limbic cortex (Gao et al., 1998; Mackarehtschian et al., 1999), suggesting that ephrin-A5 could function in the patterning of thalamocortical connections to specific cortical areas. In an earlier study we reported the existence of membrane-bound signals differentially expressed in limbic cortex and neocortex that allow thalamic afferents to distinguish between target and non-target cortical regions (Mann et al., 1998). The limbic-associated membrane protein (LAMP), a glycoprotein that is selectively expressed in the limbic cortex and other limbic regions, has been suggested to function as a recognition signal for limbic thalamic afferents and as a repulsive signal for thalamic afferents which project to neocortical areas (Barbe and Levitt, 1992; Mann et al., 1998). The present study also implicates ephrin-A5 as a positional cue in neocortical areas, which through repulsive activity prevents limbic thalamic afferents from invading inappropriate neocortical region.

The observation of multiple cell type specific activities of ephrin-A5 raises the question of how a single ligand can mediate these different responses. Because of the redundancy in Eph receptor and ephrin interactions (Gale et al., 1996), the mechanisms by which ephrin-A5 stimulates branching or repels growth cones may involve distinct Eph receptors. Three putative receptors for ephrin-A5 are expressed in the developing thalamus: EphA3 and EphA4 are found in almost all thalamic nuclei, whereas little expression of EphA5 is found outside the limbic thalamus (Gao et al., 1998; Mackarehtschian et al., 1999; Vanderhaeghen et al., 2000; Zhou et al., 1994). Thus EphA5 is a potential candidate to mediate the repulsive activity of ephrin-A5 for limbic fibers. By contrast, the branching response to ephrin-A5 may occur independently of EphA5 expression, as both populations of limbic and non-limbic neurons recognize ephrin-A5 as a branch promoter. Another possibility is that multiple cellular responses require the recruitment of distinct co-factors functionally associated to Eph receptors. Such a link between Eph receptor activation and function of member of the L1 family of cell adhesion molecules have been suggested (Zisch et al., 1997), and co-expression of ephrins and Eph receptors on axons has been found to regulate responsiveness to ephrin ligands in vitro (Hornberger et al., 1999). Finally, differences between intrinsic properties of axonal populations could govern different responses to ephrin-A5 ligand through a single receptor mechanism. For example, it has been shown that by increasing cytosolic concentration of either cAMP or cGMP, repulsive actions of some axonal guidance signals are converted to attractive effects (Ming et al., 1997; Song et al., 1998).

The present results indicate that the effects of ephrin-A5 on limbic thalamic fibers depend critically on the spatial context in which it is presented. When presented uniformly, ephrin-A5 increased axon branching. By contrast, when distributed as a sharp boundary, it repelled limbic axons. Similarly, in a recent study it was found that cortical axons respond differently to semaphorins when they grow towards increasing or decreasing semaphorin gradients (Bagnard et al., 2000). It is now an intriguing question to examine whether the spatial distribution of wiring molecules in the environment of growing axons influences expression levels or turnover rates of receptors, co-factors and/or second messengers, and thereby causes differential responses to the same extrinsic signal.

We thank Caroline Brennan for providing the ephrin-A5-AP probe, John Flanagan for providing the ephrin-A5 plasmid, and Naura Chounlamountri for excellent technical assistance. Special thanks to Fanco Weth for his enduring support in the in situ experiments, to Bill Harris and Christine Holt for their help and advice during the publication process. This work was supported by HFSP and IZKF Jena (J. B.), and a Fellowship from the French Ministry of Research (F. M.).