Draxin, a Repulsive Guidance Protein for Spinal Cord and Forebrain Commissures

See allHide authors and affiliations

Science  16 Jan 2009:
Vol. 323, Issue 5912, pp. 388-393
DOI: 10.1126/science.1165187


Axon guidance proteins are critical for the correct wiring of the nervous system during development. Several axon guidance cues and their family members have been well characterized. More unidentified axon guidance cues are assumed to participate in the formation of the extremely complex nervous system. We identified a secreted protein, draxin, that shares no homology with known guidance cues. Draxin inhibited or repelled neurite outgrowth from dorsal spinal cord and cortical explants in vitro. Ectopically expressed draxin inhibited growth or caused misrouting of chick spinal cord commissural axons in vivo. draxin knockout mice showed defasciculation of spinal cord commissural axons and absence of all forebrain commissures. Thus, draxin is a previously unknown chemorepulsive axon guidance molecule required for the development of spinal cord and forebrain commissures.

Although axon guidance proteins, including netrins, semaphorins, ephrins, and Slits (also slits), and morphogens, such as sonic hedgehog (Shh), Wnts, and bone morphogenic proteins (BMPs), are known to play roles in the correct wiring of the nervous system during development (13), the immense complexity of the nervous system makes it likely that there are more unidentified axon guidance cues to be discovered. In our search for novel axon guidance proteins, we performed signal sequence trap screening, which enabled us to identify secreted and transmembrane proteins (table S1). With this method, we have identified a molecule named draxin (dorsal repulsive axon guidance protein, fig. S1A) from a cDNA library of enriched motoneurons, floor plate, and roof plate of chick embryos. Chick draxin mRNA was expressed transiently during development of the brain and spinal cord (Fig. 1A), especially in the roof plate and the dorsal lip of the dermomyotome (Fig. 1B). Mouse draxin mRNA was expressed in a manner similar to that of the chick (fig. S2, A and B). We examined the expression of draxin in the brain by β-galactosidase (β-gal) staining of heterozygous mice (fig. S3). Mouse draxin expression was observed in many brain regions, including the olfactory bulb, cortex, midbrain, cerebellum, and pontine nuclei in postnatal day 0 (P0) mice (Fig. 1D and fig. S2, D and E).

Fig. 1.

Expression of draxin transcripts and protein during nervous system development. (A to C) draxin mRNA expression [(A) and (B)] and protein distribution (C) in chick embryos. draxin is expressed in the brain and spinal cord. (D) draxin expression by β-gal staining in a sagittal section of draxin heterozygous P0 mouse brain. draxin is expressed in the cortex (Ctx), midbrain (Mb), cerebellum (Cb), olfactory bulb (Ob), and pontine nuclei (Pn). Scale bars in (A) and (D) indicate 1 mm; in (B) and (C), 100 μm.

The deduced draxin amino acid sequence (fig. S1A) indicates that chick draxin consists of 349 amino acids with a putative signal peptide sequence at the N-terminal end but no membrane anchoring sequence, which suggests that draxin is a secreted protein. We confirmed this hypothesis via detection of the recombinant protein in conditioned medium of COS7 cells transfected with chick draxin expression vector (fig. S1B). Immunohistochemistry using antibodies against draxin (anti-draxin) revealed an interesting attribute of the draxin protein. In addition to its detection in mRNA-positive regions, draxin protein was detected at the dorsolateral basement membrane of the spinal cord (Fig. 1C and fig. S2C), indicating that the protein diffuses from its site of production and has high affinity for basement membranes.

To examine whether draxin has guidance activity for commissural axons in the spinal cord, we cultured dorsal spinal cord explants from stages 19 and 20 chick embryonic spinal cords, obtained from the thoracic level, in collagen gels. Netrin-1 (4) was added to the cultures to stimulate neurite outgrowth of commissural neurons from the explants. Neurites emerged from dorsal spinal cord explants, and dissociated cells were stained with chick TAG-1 antibody (chick anti–TAG-1), a marker for commissural axons (fig. S4), suggesting that they were commissural axons. Neurite outgrowth from dorsal spinal cord explants was greatly inhibited in draxin-conditioned medium (Fig. 2B), whereas there was robust neurite outgrowth in the control mock-transfected conditioned medium (Fig. 2A). After replacing draxin-conditioned medium with fresh culture medium, we observed robust neurite growth within 24 hours (Fig. 2C). These data excluded the possibility of a secondary effect of cell death in the presence of draxin-conditioned medium and indicated that draxin did indeed inhibit neurite outgrowth from dorsal spinal cord explants. Purified recombinant chick draxin also inhibited neurite outgrowth from dorsal spinal cord explants in a dose-dependent manner (Fig. 2, D, E, and P). We co-cultured the dorsal spinal cord explants with COS7 cell aggregates expressing chick draxin in collagen gels. Explants were dissected without adjacent roof plate tissue for radial outgrowth of neurites (5). When explants were co-cultured with mock-transfected cell aggregates, neurites grew radially from all sides of the explants (Fig. 2, F and Q). In contrast, when co-cultured with cell aggregates expressing draxin, neurites did not grow out of the proximal side to the COS7 cell aggregates; rather, they grew out of the distal side (Fig. 2, G and Q). To test whether draxin could induce growth cone collapse, we cultured chick embryonic dorsal spinal cord explants on a laminin-coated dish without netrin-1 addition. Purified draxin protein was added to the culture medium after neurites had grown out from the explants (Fig. 2H), and growth cones were followed by time-lapse video microscopy. Growth cone collapse was observed within 30 min after the addition of purified draxin (Fig. 2I), and the neurites gradually retracted (movie S1). About 70% of the growth cones collapsed, and the remaining seemed to be insensitive to draxin. These results indicated that draxin might directly bind to the neurites and growth cones. We confirmed draxin binding to the neurites and growth cones by a binding assay using alkaline phosphatase (AP)–tagged draxin protein (Fig. 2, J and K). Importantly, draxin did not bind to and did not repel the neurites of dorsal root ganglion (Fig. 2, L and M). Next, we checked whether the above three events—neurite outgrowth inhibition, growth cone collapse, and draxin binding—were correlated in terms of dose dependency. We used draxin-AP protein for this analysis (6) and observed that these three events were correlated with each other (Fig. 2S). We also examined whether mouse draxin had repulsive activity against cortical neurites. Cortical explants from E17 mouse brains were co-cultured with COS7 cell aggregates expressing mouse draxin. Draxin repelled neurites from mouse cortical explants (Fig. 2, N, O and R). These results indicated that draxin might function as a repulsive axon guidance molecule for subpopulations of neurons in vivo.

Fig. 2.

Draxin inhibits neurite outgrowth. (A to E) Inhibition of neurite outgrowth by draxin in conditioned medium (B) or purified [(E), 20 nM] form from dorsal spinal cord (dSC) explants of chick embryos. (F and G) Repulsion of neurite growth from chick dorsal spinal cord explants by draxin. (H and I) Growth cone collapse induced by draxin (n = 15). The black spots are landmarks in (H) and (I). (J to L) Draxin-AP binding to neurites from dorsal spinal cord and dorsal root ganglion (DRG) explants. (M) Coculture of dorsal root ganglion explant, dorsal spinal cord explant, and draxin COS7 cell aggregates were performed at least three times. (N and O) Repulsion of neurite outgrowth by draxin from mouse cortical explants. (P) Quantification of neurite outgrowth inhibition (mean ± SEM, *P = 0.009, **P = 0.005, ***P = 0.001, ****P< 0.001, t test, n = 8, P values were calculated by comparing with control). (Q) Quantification of repulsive activity observed in (F) and (G) measured as previously described (16)(mean ± SEM, *P< 0.001, t test, n = 10). (R) Quantification of repulsive activity observed in (N) and (O), measured as described in (Q) (mean ± SEM, *P< 0.001, t test, n = 15). (S) Dose-dependent activity of draxin (6). CM, conditioned medium. Scale bars in (A) to (G), (N), and (O), 200 μm; in (H) and (I), 10 μm; in (J) to (L), 100 μm; and in (M), 500 μm.

To understand the function of draxin in vivo, we overexpressed myc-tagged draxin in the chick spinal cord at stages 14 and 15 at the thoracic level by in ovo electroporation. Embryos were fixed at stages 23 to 25, when many commissural axons cross the floor plate. Anti-myc signals from ectopic draxin were detected in the electroporated side (fig. S5, A to C). Anti-myc signals were also detected in the dorsolateral basement membrane of the control side (fig. S5B arrowhead). This result further supported the diffusible nature of draxin and its tendency to deposit in the dorsolateral basement membrane of the spinal cord. Immunohistochemical analyses using chick anti–TAG-1 showed partial inhibition of commissural axon growth in the experimental side (Fig. 3D and fig. S5G) compared with the control side or after expression of enhanced green fluorescent protein (EGFP) alone (Fig. 3B and fig. S5G). Next, we analyzed the effects of draxin on high-level ectopic expression by constructing an expression vector for a membrane-bound form of draxin. Anti-myc signals were localized only in the EGFP-expressing area, and there was no diffusion of the ectopic protein to the control side (fig. S5, D to F). Anti–TAG-1 staining showed a stronger effect of ectopic membrane-bound draxin compared with native draxin on commissural axon growth, which was almost completely inhibited on the experimental side in the presence of the membrane-bound form (Fig. 3F and fig. S5G). This result suggests that the signal induced by membrane-bound draxin is stronger than that of the ectopically expressed native draxin. Moreover, this result suggests that membrane-bound draxin inhibited axonogenesis in the same way as bath application of draxin in the context of explant culture. However, overexpression of membrane-bound draxin at stages 19 and 20, after substantial commissural axonal growth initiation, resulted in the growth of many TAG-1–positive axons into the lumen of the central canal of the cord (Fig. 3H arrowheads). Whole-mount immunohistochemistry and EGFP fluorescence in open-book configuration after expression of EGFP alone showed the following: clear commissural axonal growth with parallel axons and a ventral funiculus along the floor plate after crossing over (fig. S5H arrowheads) and thick dorsal funiculi of the dorsal root ganglia (fig. S5I). In the case of membrane-bound draxin expression, EGFP-positive parallel axons and the ventral funiculus were not observed (fig. S5, J and L). Commissural axonal growth was completely inhibited on the experimental side; however, these axons grew normally toward the floor plate in the control side. The dorsal funiculi formed normally on both sides (fig. S5, K and M). Anti–TAG-1 staining of the electroporated side showed a clear stage difference in the effects of ectopic draxin expression, which is consistent with the data obtained from staining of the sections (Fig. 3, F and H). Earlier electroporation of membrane-bound draxin substantially inhibited commissural axonal growth (fig. S5K); however, later electroporation did not inhibit axonal growth itself, and axons were distributed in such a disorganized manner that the staining density was much higher than that in the control side (fig. S5M).

Fig. 3.

Ectopic draxin inhibits growth and disrupts the routing of commissural axons in vivo. (A to H) Transverse sections of chick spinal cord, fixed at stages 23 and 24 after electroporation. Commissural axon (TAG-1–positive) growth was normal when the control vector was electroporated [compare arrowheads in (B)]. In contrast, their growth was partially inhibited by secreted draxin [compare arrowheads in (D)]. They were completely inhibited [compare arrowheads in (F)] or severely misrouted [arrowheads in (H)] (n = 12) by membrane-bound draxin. draxin-Tm indicates membrane-bound draxin. EP, electroporation. Scale bar, 100 μm.

Next, we examined whether postcrossing commissural axons are affected by draxin. We co-cultured spinal cord explants, including the floor plate, from stages 25 and 26 chick embryos and COS7 cell aggregates in collagen gel. We did not observe any significant difference in the intensity of growth of postcrossing axons toward the COS7 cell aggregates transfected with control vector or chick draxin (fig. S6, A to C). We also labeled commissural axons with 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) after electroporation of membrane-bound draxin at stages 19 and 20. From the control side, labeled axons crossed the floor plate and turned to the anterior direction in a normal manner within the ectopic draxin environment, whereas projection patterns of precrossing axons were severely disrupted in the electroporated side (fig. S6, G to I). In the case of control vector electroporation, projection patterns of both precrossing and postcrossing axons were normal (fig. S6, D to F). These data suggest that postcrossing commissural axons are not sensitive to draxin.

To examine the function of draxin by in vivo loss-of-function analysis, we established draxin knockout mice (fig. S3). Homozygous draxin (–/–) mice are viable and fertile. We analyzed projection patterns of spinal commissural axons by TAG-1 staining at embryonic day 11.5 (E11.5). In homozygous draxin (–/–) mice, commissural axons projected in a defasciculated manner toward the floor plate (Fig. 4, Ba and Bb, and fig. S7, A and B), resulting in expansion of the TAG-1–positive area medially, whereas they projected in a tightly fasciculated form in wild-type (Fig. 4, Aa and Ab, and fig. S7, A and B) and heterozygous mice. In addition, thick bundles of TAG-1–positive axons along the basement membrane were observed more frequently in homozygous draxin (–/–) mice than in wild-type littermates (compare arrows in Fig. 4, Ab and Bb). Whole-mount anti-TAG-1 immunohistochemistry of a dissected spinal cord in open-book configuration also showed the defasciculation of commissural axons in homozygous draxin (–/–) mice (fig. S7, E and F) compared with those in wild-type mice (fig. S7, C and D). Postcrossing commissural axon projections seemed normal in homozygous draxin (–/–) mice (fig. S8, A and B).

Fig. 4.

Abnormal development of spinal cord and forebrain commissures in draxin deficient mice. (Aa to Bb) Transverse section at the upper thoracic level of E11.5 spinal cord stained with anti-TAG-1. Boxed areas in (Aa) and (Ba) are shown at high magnification in (Ab) and (Bb), respectively. The arrow in (Bb) indicates an axon bundle along the basement membrane that is thicker than that in the wild type [arrow in (Ab)]. Serial coronal [(Ca) to (Dc)] and horizontal [(Ja) to (Kc)] sections of P0 brains stained with hematoxylin and eosin. Coronal [(E) and (F)] and horizontal [(L) to (Nb)] sections of brains after injection of DiI into the neocortex [(E) and (F)] or olfactory bulb [(L) to (Nb)]. The corpus callosum (CC) and hippocampal commissure (HC) failed to cross the midline in the knockout mice [(Da) to (Dc)]. Arrow and arrowhead in (Da) and (F) indicate tangled corpus callosum axons and misprojected corpus callosum axons, respectively. (G to I) Coimmunostaining for GFAP and L1 in coronal sections of wild-type mice (G) and strongly (H) and weakly (I) affected mice at P0 to visualize the glial wedge (GW), indusium griseum glia (IGG), and midline zipper glia (MZG). Note the absence of indusium griseum glia in the strongly affected mice (H). The boxed areas in (Ma) and (Na) are shown at higher magnification in (Mb) and (Nb), respectively. Arrowheads in (Kc) indicate rudiments of anterior [AC(a)] and posterior pars [AC(p)]ofthe anterior commissure that never cross the midline. The dotted white line in (Nb) indicates the midline of the forebrain. Scale bars in (Aa) and (Ba), 200 μm; in (Ab) and (Bb), 100 μm; in (Ca) to (Dc) and (Ja) to (Kc), 500 μm; and in (E) to (F), (G) to (I), (L), (Ma), and (Na), 1 mm.

We next examined whether the projections of brain commissures were impaired in the draxin knockout mice with use of hematoxylin-eosin staining (Fig. 4, Ca to Dc and Ja to Kc) and immunostaining for the axonal maker L1 (fig. S9, A′ to D‴). All homozygous draxin (–/–) mice showed abnormal development of the corpus callosum, hippocampal commissure, and anterior commissure. We classified these phenotypes into two groups, depending on their severity. Severely affected mice had complete agenesis of these commissures (Fig. 4, Da to Dc and Ka to Kc and fig. S9, C′ to D‴), and weakly affected ones had partial defects in the formation of these commissures (table S2). About half of the heterozygous draxin mice showed abnormalities in corpus callosum and hippocampal commissure formation, and some showed complete agenesis (table S2). Anterograde tracing by DiI injection into the cortex revealed that, in the draxin knockout mice, corpus callosum axons failed to cross the midline and instead abnormally directed ventrally before they reached the midline (Fig. 4F), whereas corpus callosum axons in control mice crossed the midline (Fig. 4E). Previous studies have described that midline glial structures are intermediate targets of corpus callosum axons (7). To examine these structures, we coimmunostained with an antibody against glial fibrillary acidic protein (anti-GFAP) in wild-type and knockout mice at P0. All three midline glial populations—the glial wedge, indusium griseum glia, and midline zipper glia—were present in wild-type (Fig. 4G, n = 8) and weakly affected draxin knockout mice (Fig. 4I, n = 4). In contrast, indusium griseum glia was absent in strongly affected knockout mice (Fig. 4H, n = 6). This result suggests that the lack of indusium griseum glia might correlate with the severe defect in corpus callosum development. In the anterior pars of the anterior commissure, the axons in control mice turned to the midline and crossed toward the contralateral side, and this was confirmed by DiI injection into the olfactory bulb (Fig. 4L). The anterior commissure axons in strongly affected knockout mice turned to the lateral side but not the midline (Fig. 4, Ma and Mb), whereas the majority of anterior commissure axons in weakly affected knockout mice took a normal course toward the midline but misprojected rostrally at the midline (Fig. 4, Na and Nb). In addition, anterior commissure neurons were retrogradely labeled in the olfactory bulb contralateral to the injection site in wild-type mice (Fig. 4L); however, such labeled neurons were not detected in the corresponding area of strongly affected knockout mice (Fig. 4Ma). In contrast to the severe defects in the forebrain commissure, the posterior commissure and habenular commissure appeared to develop normally in the draxin knockout mice (fig. S8, D and F).

We next investigated draxin expression during development of the spinal cord and forebrain commissures by β-gal staining of heterozygous mice. β-gal expression was detected in the dorsal spinal cord and commissural axons (fig. S10, A to F). Antidraxin staining revealed draxin protein expression in the same area and heavy deposition in the lateral basement membrane (fig. S10, G to L). In the case of forebrain commissures, β-gal expression was observed in the regions that surround the corpus callosum, hippocampal commissure, and anterior commissure, such as the midline glial cells, indusium griseum glia, and glial wedge, whereas β-gal expression was not detected in these commissural axons (fig. S11). Antidraxin staining and draxin-AP binding on sections revealed the presence of draxin proteins and its receptors in the forebrain commissural axons (fig. S12).

Spinal cord commissural axonal growth has been well studied and found to be guided by the attractive cues netrin-1 (8, 9) and Shh (10), which emanate from the floor plate, and also by the repulsive cues BMP7 and growth differentiation factor 7 (GDF7), which emanate from the roof plate (11, 12). Our in vitro and in vivo gain-of-function data suggest that draxin is a chemorepulsive guidance protein for commissural axons. The homozygous draxin (–/–) mouse showed defasciculated projections of commissural axons toward the floor plate (Fig. 4 and fig. S7). This defasciculation might be due to the deficient repulsive activity of environmental draxin, mainly in the lateral basement membrane. Immunochemical analyses indicate expression of draxin mRNA and protein in commissural axons (fig. S10). The importance of this expression is unknown, although draxin may function in an autocrine manner to regulate the sensitivity of commissural axon to draxin in the surrounding milieu.

The data presented here demonstrate that draxin is required for the midline crossing of forebrain commissures (Fig. 4 and fig. S9). draxin is expressed in midline glial cells, which have been thought to act as intermediate guideposts for corpus callosum axons via the expression of axon guidance molecules (13), such as Slit2 (14) and Wnt5a (15). In addition, draxin repels neurite outgrowth from cortical explants at E17 (Fig. 2O), when corpus callosum axons cross the midline. These results suggest that draxin is a chemorepulsive molecule that is responsible for corpus callosum development in midline glial cells. We speculate that misprojection of corpus callosum axons at the midline, observed in almost all of the knockout mice (Fig. 4F), is caused by the deficient draxin repulsive activity from the glial wedge. A similar presumption probably applies to the draxin roles in anterior commissure and hippocampal commissure development, as judged from analyses of the draxin expression and the mutant phenotypes. Thus, we propose that draxin repulsion from the regions surround the trajectories of forebrain commissures is essential for proper guidance of their commissural axons, preventing them from misprojecting before reaching the midline. Because all forebrain commissures were frequently misprojected at the midline in the knockout mice, midline cells expressing draxin may be particularly critical for the midline crossing. Furthermore, corpus callosum axons in the knockout mice are defasciculated in the ipsilateral side (Fig. 4F). draxin is expressed in deep cortical layers and cingulate cortex, in addition to the midline glial cells. Because most of corpus callosum axons arise from neurons in layers 2/3 and 5, some of the projecting neurons may express draxin. Thus, draxin is required for the fasciculation of corpus callosum axons in a paracrine and/or autocrine manner, which may be consistent with draxin functions on spinal commissural axons. It is also important to note that indusium griseum glia is missing only in the severely affected knockout mice (Fig. 4H). This result suggests that not only the deficient repulsive activity from the glial cells but also the lack of indusium griseum glia might be involved in the disruption of the corpus callosum formation. Further investigation is needed to clarify draxin functions on the formation of indusium griseum glia and its involvement on the commissure formation.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Tables S1 and S2

Movie S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article