Axon guidance at the midline choice point
Abstract
The central nervous system (CNS) of higher organisms is bilaterally-symmetric. The transfer of information between the two sides of the nervous system occurs through commissures formed by neurons that project axons across the midline to the contralateral side of the CNS. Interestingly, these axons cross the midline only once. Other neurons extend axons that never cross the midline; they project exclusively on their own (ipsilateral) side of the CNS. Thus, the midline is an important choice point for several classes of pathfinding axons. Recent studies demonstrate that specialized midline cells play critical roles in regulating the guidance of both crossing and non-crossing axons at the ventral midline of the developing vertebrate spinal cord and the Drosophila ventral nerve cord. For example, these cells secrete attractive cues that guide commissural axons over long distances to the midline of the CNS. Furthermore, short-range interactions between guidance cues present on the surfaces of midline cells, and their receptors expressed on the surfaces of pathfinding axons, allow commissural axons to cross the midline only once and prevent ipsilaterally-projecting axons from entering the midline. Remarkably, the molecular composition of commissural axon surfaces is dynamically-altered as they cross the midline. Consequently, commissural axons become responsive to repulsive midline guidance cues that they had previously ignored on the ipsilateral side of the midline. Concomitantly, commissural axons lose responsiveness to attractive guidance cues that had initially attracted them to the midline. Thus, these exquisitely regulated guidance systems prevent commissural axons from lingering within the confines of the midline and allow them to pioneer an appropriate pathway on the contralateral side of the CNS. Many aspects of midline guidance are controlled by mechanistically and evolutionarily-conserved ligand-receptor systems. Strikingly, recent studies demonstrate that these receptors are modular; the ectodomains determine ligand recognition and the cytoplasmic domains specify the response of an axon to a given guidance cue. Despite rapid and dramatic progress in elucidating the molecular mechanisms that control midline guidance, many questions remain. © 2001 Wiley-Liss, Inc.
Introduction
The function of the mature central nervous system (CNS) relies upon the operation of highly complex neuronal circuits. Thus, a critical phase of early nervous system development is the establishment of specific connections between neurons and their target cells. In both vertebrates and invertebrates, the leading edge of an axon termed the “growth cone” navigates over significant distances with great precision. Growth cones guide axons by functioning as exquisite sensors that detect and subsequently respond to a variety of environmental guidance cues (Tessier-Lavigne and Goodman, 1996; Mueller, 1999). These cues exist in either diffusible or cell surface-associated forms that regulate long-range and short-range pathfinding, respectively (Tessier-Lavigne and Goodman, 1996). Cell surface receptors residing on growth cones and their associated axons interpret these signals as positive/attractive or negative/repulsive forces. Thus, long-range repulsive guidance cues referred to as chemorepellents are likely to provide growth cones with a “push” from behind, whereas long-range attractive cues known as chemoattractants would be expected to “pull” axons in from a distance. On the other hand, contact-dependent attractive and repulsive cues function to control the directionality of axon growth over short distances (Tessier-Lavigne and Goodman, 1996). Collectively, these four guidance forces are thought to shape the complete trajectory of a given pathfinding axon.
An important and relatively recent advance in the field has been the realization that complicated axonal trajectories can be separated into a series of short segments. Accordingly, the difficult problem of determining how a given axon reaches a distant target is reduced to understanding how the axon navigates each of these successive segments (Tessier-Lavigne and Goodman, 1996). In both vertebrates and invertebrates, so-called “intermediate targets” or “choice points” mark the end of one segment and the beginning of the next. Most intermediate targets consist of small groups of specialized guidepost cells that present growth cones with the guidance information required to steer the trailing axon along the next segment of its trajectory (see Palka et al., 1992). Consistent with this role, abrupt changes in growth cone morphology and the directionality of axonal growth are observed as axons travel in the immediate vicinity of (or, in some cases, after they make direct contact with) intermediate targets (see Caudy and Bentley, 1986; Bovolenta and Dodd, 1990; Colamarino and Tessier-Lavigne, 1995b; Mason and Erskine, 2000). Moreover, axon pathfinding at choice points is severely disrupted in the absence of these intermediate targets (see Bentley and Caudy, 1983; Bovolenta and Dodd, 1991; Mason and Sretavan, 1997 and references therein).
The midline of the developing CNS is an important choice point for pathfinding axons. In both vertebrates and invertebrates, many axons extend significant distances toward the midline and then must decide whether or not to cross over to the opposite side of the CNS. After reaching the midline, noncrossing and crossing axons execute right angle turns on the ipsilateral (same) and contralateral (opposite) side of the midline, respectively. Subsequently, these growth cones/axons must further decide whether to extend in the rostral or caudal direction parallel to the midline (Fig. 1). Both classes of axons are barred from the midline once they begin to extend in the longitudinal direction. Specialized midline cells function as intermediate targets that provide growth cones with long- and short-range guidance cues which positively or negatively influence axonal extension. Thus, proper axon guidance in this region relies upon the growth cone's ability to correctly integrate a variety of extracellular guidance signals.
Axon trajectories in the vicinity of the midline. A: Schematic three-dimensional representation of commissural axon trajectories in the developing vertebrate spinal cord. Commissural neurons located in dorsal regions of the spinal cord project axons along a circumferential/transverse path to the ventral midline/floor plate. After crossing the ventral midline, these axons turn at right angles to project in the longitudinal direction alongside the floor plate. Ipsilateral refers to the side of the midline from which these axons arise and contralateral refers to the side of the midline to which these axons project. B: Schematic “open-book” representation of commissural axon trajectories in the spinal cord. In this view, the floor plate occupies the midline and the roof plate occupies the lateral edges. Commissural axons extend in the transverse plane towards (1), into (2), and across the floor plate (3), and subsequently turn rostrally into the longitudinal plane (4). Axons emanating from only one side of the dorsolateral spinal cord are depicted in A and B. C: In the Drosophila ventral nerve cord, commissural neurons (c) extend axons across the midline through either the anterior commissure (AC) or the posterior commissure (PC). After crossing through the midline, these axons turn orthogonally and project parallel to the midline within longitudinally-oriented tracts or longitudinal connectives. Ipsilaterally-projecting neurons (i) extend axons in the longitudinal direction parallel to the midline and never cross the midline. Specialized midline cells separate the right and left sides of the nerve cord.
This review will primarily compare and contrast the molecular mechanisms that control axon guidance at the midline of the developing vertebrate spinal cord and the Drosophila ventral nerve cord. We will also highlight some of the guidance mechanisms which regulate the pathfinding of circumferentially- and longitudinally-directed axons in Caenorhabditis elegans. In particular, we will present the results of recent studies that have identified a variety of evolutionarily-conserved, as well as vertebrate- or invertebrate-specific guidance cues and their receptors (see Table 1; Figs. 2, 3). Since complicated axonal trajectories are more easily studied by analyzing their component parts, we will separately consider the molecular interactions that promote axonal extension toward the midline, control midline crossing, and regulate the transition from transverse to longitudinal growth which occurs in the immediate vicinity of the midline. In addition, we will discuss some of the major unresolved issues confronting researchers studying midline guidance and suggest important future directions for the field.
| Name | Organism | Localization | Role in midline guidance |
|---|---|---|---|
| Netrin-1 | Rat, mouse, chick | Floor plate (mRNA and protein) | Required for commissural axon guidance to the floor plate |
| Netrin-4 | Mouse | Lateral margins of the floor plate (mRNA) | Unknown |
| Netrin-A Netrin-B | Fly | Midline cells (mRNA), midline cells, commissural and longitudinal axon tracts (protein) | Required for proper formation of commissural and longitudinal axon tracts |
| UNC-6 | Nematode | Ventral midline-associated neuroglia and neurons, longitudinal axon tracts (protein) | Required for circumferential guidance of commissural axons along the dorsoventral body axis |
| DCC | Rat, mouse | Commissural neurons (mRNA), commissural axons (protein) | Required for commissural axon guidance to the floor plate |
| Frazzled | Fly | Commissural and longitudinal axon tracts (protein) | Required for proper formation of commissural and longitudinal axon tracts |
| UNC-40 | Nematode | Sensory and motor neurons/axons that project toward the ventral midline (protein) | Required for ventrally-directed guidance of axons that orient toward sources of UNC-6 |
| UNC5H1 UNC5H2 UNC5H3/RCM | Rat, mouse | Differentiating neurons in the ventral spinal cord (mRNA) | Unknown |
| UNC5H3/RCM | |||
| UNC-5 | Nematode | Sensory and motor axons that project away from the ventral midline (protein) | Required for dorsally-directed guidance of axons that orient away from sources of UNC-6 |
| TAG-1 | Rat, mouse | Commissural neurons (mRNA), uncrossed segments of commissural axons (protein) | Unknown |
| Axonin-1 | Chick | Commissural axons (protein) | Injection of anti- or soluble Axonin-1 into chick embryos prevents many commissural axons from entering the floor plate |
| NrCAM | Mouse, chick | Concentrated in floor plate and ventral commissure (protein) | Injection of anti-NrCAM into chick embryos prevents many commissural axons from entering the floor plate |
| L1 | Rat, mouse | Commissural neurons (mRNA), crossed segments of commissural axons (protein) | Unknown |
| NgCAM | Chick | Commissural axons (protein) | Injection of anti-NgCAM into chick embryos promotes defasciculation of commissural axons; injection of both anti-NgCAM and anti-NrCAM blocks extension of commissural axons along the longitudinal axis |
| dRobo-1, dRobo-2, dRobo-3 | Fly | High levels on longitudinal axon tracts and ipsilaterally-projecting axons, low levels on commissural axon tracts (protein) | Repulsive guidance receptors that function as gatekeepers to prevent ipsilaterally-projecting axons from crossing, and commissural axons from recrossing, the midline |
| rRobo-1, rRobo-2 | Rat | Commissural neurons (mRNA) | Unknown |
| Sax-3 | Nematode | Interneurons and motor neurons/axons that project to the ventral midline (protein) | Component of a repellent guidance system that regulates midline crossing of a variety of axonal populations |
| dSlit | Fly | Midline glia (mRNA), midline glia and commissural axons (protein) | Midline repellent that prevents ipsilaterally-projecting axons from crossing, and crossed commissural axons from recrossing, the midline |
| rSlit-1, rSlit-2, rSlit-3 | Rat, chick | Floor plate, roof plate (mRNA) | rSlit-2 selectively repels commissural axons which have passed through the floor plate in vitro |
| Commissureless | Fly | Midline cells (mRNA), midline cells and commissural axons (protein) | Required for commissural axons to cross the midline |
| Nidogen | Nematode | Body wall basement membranes (protein) | Required for axons to switch from circumferential to longitudinal migration; regulates positioning of longitudinal nerves |
| VEMA | Rat | Floor plate, roof plate (mRNA and protein) | Unknown |
| DPTP10D | Fly | High levels on longitudinal axon tracts, low levels on commissural axon tracts (protein) | In double mutants lacking both DPTP10D and DPTP69D many longitudinal axons abnormally cross the midline |
| DPTP69D | Fly | Commissural and longitudinal axon tracts (protein) | See DPTP10D |
| DLAR | Fly | High levels on longitudinal axon tracts, low levels on commissural axon tracts (protein) | Unknown |
| Derailed | Fly | Axon segments contained within the anterior commissure (protein) | Controls commissure choice displayed by midline-crossing axons |
| B-class ephrins | Mouse | Floor plate, roof plate (mRNA and protein) | Promote the collapse of EphB1-bearing commissural growth cones, in vitro |
| EphB1 | Mouse | Commissural neurons (mRNA) crossed segments of commissural axons (protein) | Unknown (see B-class ephrins) |
| Dek | Fly | High levels on longitudinal axon tracts, low levels on commissural axon tracts (protein) | Unknown |
| Neuropilin-2 | Mouse | Floor plate, roof plate, commissural neurons (mRNA and protein) | Required for normal commissural axon pathfinding during and after midline crossing |
| Semaphorin-3B | Mouse | Floor plate (mRNA) | Selectively repels commissural axons which have passed through the floor plate in vitro |
| F-Spondin | Rat, chick | Floor plate (mRNA), basal lamina underlying floor plate (protein) | Soluble F-Spondin injected into chick embryos prevents many commissural axons from crossing through the floor plate |
| GAD-65 | Rat | Commissural axons (protein) | Unknown |
| Annexin IV | Mouse | Floor plate, roof plate (protein) | Unknown |
| BMP-7 | Rat, mouse | Roof plate (mRNA) | Diffusible repellent for commissural axons that induces the collapse of their growth cones in vitro |
| BMP-6 | Rat, mouse | Roof plate and floor plate (mRNA) | Weak diffusible repellent for commissural axon in vitro |
| Gdf-7 | Rat, mouse | Roof plate (mRNA) | Unknown |
| UNC-129 | Nematode | Dorsal body wall muscles (protein) | Required to guide motor axons along dorsoventral axis |
| Gli2 | Mouse | Ventral and intermediate regions of the spinal cord (mRNA) | Required for induction of the floor plate and immediately adjacent interneurons |
| Math1 Ngn1 | Mouse | Commissural neurons (mRNA and protein) | Unknown; expression defines progenitors of dorsal commissural neurons |
| Dbx1 | Mouse | Commissural neurons (mRNA and protein) | Unknown; expression defines both ventral and dorsal commissural neuron populations |
| Pax-3 Pax-7 | Mouse | Commissural neurons (mRNA and protein) | In double mutants lacking both Pax-3 and Pax-7 the ventral commissure underlying the floor plate is greatly reduced |
| Lim 1 LH2A LH2B | Mouse | Commissural neurons (mRNA and protein) | Unknown; expression defines subpopulations of dorsal commissural neurons |
| LH2B |
- a The role (or proposed role) of these proteins in midline guidance is discussed in the text. Fly refers to Drosophila and nematode refers to C. elegans. Abbreviations: UNC, uncoordinated; DCC, deleted in colorectal cancer; TAG-1, transiently expressed axonal surface glycoprotein-1; NgCAM, neuron-glia cell adhesion molecule; NrCAM, NgCAM-related CAM; Robo, roundabout; Sax, sensory axon defect; VEMA, VEntral Midline Antigen; DPTP, Drosophila protein tyrosine phosphatase; Dek, Drosophila Eph kinase; GAD-65, glutamic acid decarboxylase (65 kD isoform); BMP, bone morphogenetic protein; Gdf-7, BMP family member; Math 1, Mammalian atonal homolog 1; Ngn 1, Neurogenin 1; Dbx 1, developing brain homeobox 1; Pax, paired-box-containing gene; Lim1, LH2A and LH2B, Lim homeodomain-containing genes. The reader is referred to the text for relevant references.
Structures of selected midline-/commissural axon-associated proteins. The reader is directed to Table 1 for the localization of these proteins in the developing CNS, as well as for their functional role(s) in midline guidance.
Distribution of selected midline- or commissural neuron/axon-associated molecules in the developing vertebrate spinal cord and Drosophila ventral nerve cord. A: Open-book view of the vertebrate spinal cord emphasizing the axon-segment specific expression of several cell surface receptors. Strikingly, TAG-1 is expressed on only those segments of commissural axons (red) that are extending to the floor plate, NrCAM is expressed at high levels on only those portions of commissural axons (black bar) that are contained within the floor plate, while the expression of L1 and EphB1 is restricted to crossed segments of commissural axons (black). In contrast, Axonin-1, DCC, NgCAM, Neuropilin-2, and GAD65 are expressed along the entire length of commissural axons. The expression of Robo proteins on these axons has not yet been reported. Roof plate cells express BMP6, BMP7, Gdf7, B-class ephrins, VEMA, Neuropilin-2, Slits, and AnnexinIV. Commissural neurons or their progenitors express the transcription factors MATH1, NGN1, Dbx1, PAX-3, PAX-7, LH2A, LH2B, and Lim1. Floor plate cells express netrins, NrCAM, Slits, B-class ephrins, F-spondin, Gli2, VEMA, AnnexinIV, Neuropilin-2, and BMP-6. B: In the Drosophila ventral nerve cord, Derailed is expressed on only those segments of commissural axons (blue) passing through the midline via the anterior commissure (AC), while significant levels of Robos, Dek, DPTP10D, and DLAR are detected on only those axons (commissural and ipsilateral) projecting within the longitudinal tracts (red). High levels of Frazzled are expressed on axons extending through the AC (blue) and PC (green), as well as those traveling within the longitudinal tracts. Midline cells express Netrins, Slit, and Commissureless. See text for relevant references and Table 1 for abbreviations.
Axon Pathfinding in the Vicinity of the Midline
Vertebrate Spinal Cord
In the developing CNS of a wide variety of bilaterally-symmetric organisms, many interneurons (neurons that connect with other neurons) project axons along trajectories that are either perpendicular or parallel to the midline. The vertebrate spinal cord represents a particularly good model system for the study of axon pathfinding in the vicinity of the midline. Commissural interneurons located at dorsolateral positions on either side of the spinal cord, extend axons that project along a circumferential path toward the floor plate, a small group of specialized columnar ependymal cells that span the width of the spinal cord at the ventral midline (Bovolenta and Dodd, 1990; Colamarino and Tessier-Lavigne, 1995b and references therein; Figs. 1A, B). The earliest commissural projections (“pioneer” axons), travel along the lateral edges of the spinal cord until they reach the floor plate. The later projections (“follower” axons) initially extend along the same pathway. However, these axons turn away from the lateral edges of the spinal cord upon reaching the position occupied by developing motor neurons, and then follow a more direct, ventromedial route to the floor plate (see Colamarino and Tessier-Lavigne, 1995b). After reaching the midline, commissural axons cross through the ventral-most third of the floor plate (Bovolenta and Dodd, 1990). At the contralateral margin of the floor plate, most of these axons turn orthogonally and extend, at least for a short distance (approximately 100 μm; see Bovolenta and Dodd, 1990), within the longitudinally-oriented ventral funiculus (axon bundle). Interestingly, decussated commissural axons extend exclusively in the rostral direction in rodents, and in both rostral and caudal directions in the chick (see Colamarino and Tessier-Lavigne, 1995b and references therein; Figs. 1A, B). Thus, the directionality of this portion of the commissural projection is likely to be species-specific. Commissural axons apparently ignore guidance cues located on the ipsilateral side of the spinal cord that they will subsequently respond to on the contralateral side, since they are never observed to turn into the longitudinal plane prior to crossing the midline and since they do not recross the midline.
A second early developing population of dorsolaterally-positioned spinal neurons, known as association interneurons, also initially extend axons along a circumferential/transverse path toward the floor plate. However, these axons execute right angle turns and extend parallel to the floor plate along the ipsilaterally-projecting lateral funiculus prior to reaching the midline (see Colamarino and Tessier-Lavigne, 1995b and references therein). More ventrally-positioned, ipsilaterally-projecting interneurons have been identified in both the developing chick and mouse spinal cord. For example, the Primitive Longitudinal (PL) neurons develop in a region located between the floor plate and motor neurons in early chick embryos. These unique neurons extend axons that directly pioneer an ipsilateral, longitudinally-directed pathway within the ventral funiculus prior to the arrival of commissural axons (Yaginuma et al., 1990). In the developing mouse spinal cord, the expression patterns of several transcription factors define the locations of specific populations of interneurons (Burrill et al., 1997; Matise and Joyner, 1997). Recently, the pathfinding of axons which extend from a small population of interneurons that express the homeodomain-containing protein, EN-1, was characterized through the use of the axonal reporter gene tau-lacZ (Saueressig et al., 1999). In transgenic mice that express tau-β-galactosidase under the control of EN1 regulatory sequences, EN-1-positive axons were observed to pioneer an ipsilateral projection in the ventral spinal cord. In contrast to the direct initiation of longitudinal growth displayed by PL cells in the chick, these axons initially extend in the ventral direction before turning to project rostrally for a short distance within the developing ventrolateral funiculus (see Saueressig et al., 1999).
Drosophila Ventral Nerve Cord
Axonal patterning in the Drosophila ventral nerve cord and in the vertebrate spinal cord share several common features. Just as in the spinal cord, most neurons contained within the ventral nerve cord are interneurons that initially extend axons toward the midline. Within each neuromere or segment of the ventral nerve cord (a single nerve cord is comprised of about 15 essentially identical segments) many of these neurons (e.g., RP1, SP1) project axons across the midline. However, in contrast to the known behavior of their vertebrate counterparts, commissural axons in the fly cross the midline through either the anterior or posterior commissure (AC or PC). As their names imply, these transversely-oriented fiber tracts connect both sides of the CNS and are located at distinct rostrocaudal positions within a given neuromere (see Klambt et al., 1991 and references therein; Fig. 1C). After crossing the midline, commissural axons turn orthogonally to project within the longitudinal connectives. These longitudinally-oriented fiber bundles are symmetrically positioned on either side of the midline and run the length of the CNS (Fig. 1C). As in the vertebrate spinal cord, commissural axons in the Drosophila ventral nerve cord do not recross the midline. In addition to decussated segments of commissural axons, the longitudinal connectives also contain ipsilaterally-projecting axons that arise from identified neurons (e.g., pCC, vMP2) and which turn into the longitudinal direction on their own side of the midline (Fig. 1C). The trajectories of these axons resemble those followed by ipsilaterally-projecting interneurons located in dorsal and ventral regions of the vertebrate spinal cord (see above). Collectively, large numbers of contralaterally- and ipsilaterally-projecting axons ultimately give rise to the conspicuous orthogonal array of axons which represents the Drosophila ventral nerve cord (Fig. 1C).
The Role of Midline Cells in Commissural Axon Guidance
Vertebrates
Analyses of mutant or experimentally-manipulated vertebrate embryos demonstrate that the floor plate is a critical source of guidance information for commissural axons. In the Danforth's Short-tail (Sd) mouse mutant, commissural axons properly extend to the midline in caudal segments of the spinal cord that lack a floor plate. However, these axons subsequently either fail to cross the midline, or successfully cross but fail to turn into the longitudinal plane (Bovolenta and Dodd, 1991). While these findings appear to support a direct role for the floor plate in midline guidance, it should be noted that Sd mice also lack a notochord and exhibit defects in motor neuron differentiation. Thus, at least some of the described guidance errors may not be directly attributable to the absence of the floor plate.
More recently, the fidelity of commissural axon pathfinding was assessed in the spinal cords of mice lacking Gli2, a zinc-finger transcription factor that functions downstream of the Sonic Hedgehog signaling pathway (see Matise et al., 1998; Matise et al., 1999 and references therein). Gli2 mutant mouse embryos display a selective loss of the floor plate and a small population of immediately adjacent interneurons [ventral intermediate region (VIR) cells; Matise et al., 1998; Matise et al., 1999]. Thus, these animals represent a model system for more directly determining the role of the floor plate in commissural axon guidance. After properly projecting to the ventral midline, most commissural axons inappropriately accumulate within the region formerly occupied by the floor plate in the spinal cords of Gli2 mutant mice (Fig. 4). In addition, while the very few axons which successfully cross through the floor plate turn appropriately into the longitudinal plane, they aberrantly project in both rostral and caudal directions (Matise et al., 1999). Importantly, most populations of commissural axons exhibit similar guidance defects in the Gli2 mutant mice. This was demonstrated by using either the lipophilic dye, DiI, or specific antibodies to label large numbers of axons extending from multiple dorsal populations of commissural neurons. Furthermore, regulatory sequences derived from the homeobox-containing gene, Dbx1, were used to target tau-lacZ to a specific population of ventrally-located commissural neurons/axons in transgenic mice (referred to as TgDbx1/TLZ; Matise et al., 1999), in order to determine whether the proper pathfinding of this class of axons is also dependent on the presence of the floor plate. In the progeny of crosses between TgDbx1/TLZ and the Gli2 mutant mice, β-gal expression revealed pathfinding defects that closely resembled those observed with DiI labeling (Matise et al., 1999). Taken together, these findings directly demonstrate that floor plate (and possibly VIR) cells are important regulators of local guidance events in the vicinity of the midline. Analyses of commissural axon pathfinding performed in floor plate-deficient chick, Xenopus, and zebrafish embryos support these conclusions (see Colamarino and Tessier-Lavigne, 1995b and references therein).
Most commissural axons aberrantly accumulate at the ventral midline, adjacent to a low level of ventricular zone-associated Netrin-1 expression, in Gli2 homozygous mutant mouse embryos. Cross-sectional schematic views depicting the developing spinal cord of wild-type and Gli2−/− mutant mice. In both views, commissural axons are red and Netrin1 expression is represented by purple shading. Left: In wild-type embryos, dorsal and ventral classes of commissural axons are guided to the floor plate (FP). The combined expression of Netrin1 in the FP and the ventricular zone (vz) presumably gives rise to a ventral (high)-to-dorsal (low) Netrin 1 protein gradient (plus marks). Commissural axons appropriately turn (yellow arrows) into the longitudinal plane after crossing through the floor plate. Right: In Gli2 homozygous mutant embryos, a gradient of Netrin1 expression established exclusively by cells of the vz is sufficient to guide commissural axons to the ventral midline. Upon reaching the ventral midline, most commissural axons cluster within, and remain confined to, the position formerly occupied by the floor plate, where they exhibit disorganized projections (red and yellow lines). “Reprinted with permission from Matise MP, Lustig M, Sakurai T, Grumet M, Joyner AL. 1999. Ventral midline cells are required for the local control of commissural axon guidance in the mouse spinal cord. Development 126:3649–3659. Copyright 1999 The Company of Biologists Limited.”
A recent in vitro study has presented fascinating evidence suggesting that floor plate cells provide commissural neurons with trophic support (Wang and Tessier-Lavigne, 1999). Importantly, dependence on this support begins at the time when commissural axons contact the floor plate and is maintained during the period when these axons are thought to extend alongside this structure (Wang and Tessier-Lavigne, 1999). These findings predict that only those axons in the vicinity of floor plate and that consequently receive trophic support en passant [in passing; a term coined by the authors to classify this type of neurotrophic activity; Wang and Tessier-Lavigne, 1999] will survive, thus ensuring the fidelity of commissural projections in vivo. Accordingly, commissural neurons might be expected to die soon after their axons reach the ventral midline in floor plate-lacking Sd and Gli2-deficient mice. An analysis of cell death performed within dorsal spinal cord regions in these mice should represent an important test of the en passant hypothesis.
The findings described above demonstrate that short-range, presumably contact-dependent interactions between commissural growth cones and floor plate (and/or VIR) cells regulate commissural axon guidance at the midline. Multiple populations of ipsilaterally-projecting interneurons travel in the longitudinal plane parallel to the floor plate. However, the extent to which their growth cones make significant contacts with floor plate cells is unknown. Thus, it is certainly possible that the floor plate does not play a major role in controlling the guidance of ipsilaterally-projecting axons. Consistent with this hypothesis, ipsilaterally-projecting dorsal association interneurons have been observed to extend along their normal trajectories in Gli2 mutant mice (Matise et al., 1999). However, the use of DiI in these particular studies may have obscured potentially subtle guidance defects displayed by specific populations of ipsilaterally-projecting axons which represent a small proportion of the total number of DiI-labeled axons. In the future, it should be possible to overcome this potential limitation by assessing the pathfinding of molecularly-distinct (e.g., EN-1-positive; Saueressig et al., 1999), ipsilaterally-projecting axonal populations in isolation within a Gli2 mutant background.
Drosophila
As in the vertebrate spinal cord, specialized cells located at the midline of the Drosophila ventral nerve cord physically separate the left and right sides of the CNS. In contrast to the apparent homogeneity of floor plate cells, a variety of cell types, including three pairs of midline glia, as well as the MP1 and VUM neurons, are present at the midline of the Drosophila ventral nerve cord (Klambt et al., 1991). Nevertheless, Drosophila midline cells appear to be functional counterparts of floor plate cells. Analyses of mutants defective in midline cell development reveal essential roles for these cells in the formation of the Drosophila ventral nerve cord. For example, in the single-minded (sim) mutant, midline cells fail to differentiate and ultimately die. Consequently, commissures do not form and the longitudinal connectives collapse into a single fused tract at the midline (Thomas et al., 1988; Klambt et al., 1991). This dramatic phenotype is reminiscent of the aberrant accumulation of commissural axons at the ventral midline in the spinal cords of Gli2 mutant mice (Matise et al., 1999; Fig. 4). The similarities between these two phenotypes underscore the evolutionarily-conserved role that midline cells play in regulating midline guidance.
Long-Range Guidance to the Midline
Vertebrate Spinal Cord: Netrins
Both contralaterally- and ipsilaterally-projecting interneurons in the vertebrate spinal cord initially extend axons toward the midline. Accordingly, midline-derived chemoattractants are good candidates for regulating these long-range pathfinding events. Several years ago, the membrane-associated laminin-like proteins Netrin-1 and Netrin-2 were isolated from chick brain based on their ability (like the floor plate) to promote the outgrowth of, and reorient, commissural axons in a collagen gel-based in vitro assay system (Kennedy et al., 1994; Serafini et al., 1994). Consistent with their putative roles as diffusible signals, it was also shown that COS (immortalized monkey cell line) cells transfected with Netrin-1 or Netrin-2 cDNAs secrete the corresponding proteins into the medium and chemoattract commissural axons in vitro (Kennedy et al., 1994). Interestingly, while recombinant forms of both proteins have been shown to regulate commissural outgrowth in vitro, only netrin-1 mRNA expression is detected in the chick floor plate (netrin-2 is expressed in the ventral two-thirds of the spinal cord excluding the floor plate; Kennedy et al., 1994).
More recent studies have now provided compelling evidence that Netrin-1 functions as a floor plate-derived chemoattractant which directs the pathfinding of commissural axons in vivo. For example, Netrin-1 binds a cell surface receptor, referred to as DCC (for Deleted in Colorectal Cancer), that is expressed on commissural axons and their associated growth cones (Keino-Masu et al., 1996). DCC belongs to a family of transmembrane proteins that possess four immunoglobulin (Ig) domains and six fibronectin type III (FNIII) repeats (Keino-Masu et al., 1996). Consistent with the dose of Netrin-1 required to promote commissural axon outgrowth, Netrin-1 binds DCC with a dissociation constant (Kd) of 10−8 M. Furthermore, an antibody against DCC blocks Netrin-1-dependent axon outgrowth in vitro (Keino-Masu et al., 1996). Most importantly, inactivation of netrin-1 (Serafini et al., 1996) or DCC (Fazeli et al., 1997) in mice leads to pathfinding defects that ultimately prevent most commissural axons from reaching the floor plate (see Fig. 5). Consistent with these findings, floor plate explants derived from Netrin-1-deficient mice fail to promote commissural axon outgrowth in vitro (while retaining their ability to reorient these axons; Serafini et al., 1996). A role for Netrin-1 in regulating the guidance of ipsilaterally-projecting axons has also been recently established. Specifically, the Netrin-1 receptor DCC (and possibly UNC5H1, see below) is expressed on En-1-positive axons in vivo and the initial ventral projections extended by EN-1 interneurons are disorganized and fasciculate improperly in Netrin-1-deficient mice (Saueressig et al., 1999). Taken together, these studies strongly support a role for Netrin-1 as a floor plate-derived chemoattractant that guides both contralaterally- and ipsilaterally-projecting axons to the ventral midline of the developing mouse spinal cord in a DCC-dependent manner.
Most commissural axons fail to reach the floor plate in Netrin-1 homozygous mutant mouse embryos. Transverse sections of wild-type and Netrin-1 homozygous mutant spinal cords labeled with an antibody against TAG-1. Left: In wild-type embryos, TAG-1-positive commissural axons (C axon) grow in a directed manner to the floor plate. Right: In mutant embryos, C axon growth within the spinal cord is disorganized and many commissural axons fail to reach the floor plate. “Reprinted from Cell, 87, Serafini T, Colamarino, SA, Leonardo ED, Wang H, Beddington R, Skarnes WC, Tessier-Lavigne M, Netrin-1 is required for commissural axon guidance in the developing vertebrate nervous system, 1001–1014, Copyright (1996), with permission from Elsevier Science.”
Despite these compelling data, several issues regarding the in vivo localization and the function(s) of Netrins remain to be clarified. For example, since commissural growth cones/axons respond to a spatially-restricted source of diffusible Netrin-1 (and Netrin-2) in vitro, it has been assumed that a dorsoventral gradient of Netrin-1 protein regulates their directed growth in vivo. However, a substantial fraction of recombinant Netrin-1 (and Netrin-2) remains associated with the surfaces of COS cells transfected with the corresponding cDNAs (Kennedy et al., 1994). While this result is consistent with the biochemical characteristics of the Netrins, it calls into question the precise distribution of Netrin-1 protein in the spinal cord. Interestingly, a recent immunohistochemical study failed to detect a ventral (high) to dorsal (low) gradient of Netrin-1 protein in the embryonic chick spinal cord. Rather, Netrin-1 was localized to the lateral edges of the spinal cord along the paths that early-developing commissural axons take to reach the ventral midline (MacLennan et al., 1997). This finding suggests that Netrin-1 may be more likely to function as a haptotactic (substrate-anchored), as opposed to a chemotactic, guidance cue in vivo (MacLennan et al., 1997). However, it should be pointed out that technical limitations inherent in immunohistochemical analyses might make it difficult to accurately document the distribution of diffusible, as opposed to cell-associated, Netrin-1. In this regard, it is also important to reconsider the interpretation of pathfinding defects observed in the Gli2-knockout mice. In these animals, commissural axons project normally to the ventral midline in the absence of the structure (floor plate) which provides the highest source of Netrin-1. As a possible explanation of this apparent paradox, it was carefully noted that a ventral (high) to dorsal (low) gradient of netrin-1 mRNA originating in the ventricular zone (previously described in wild-type mice; Serafini et al., 1996) persists in the spinal cord of Gli2-deficient mice (Matise et al., 1999; see Fig. 4). This observation supports the notion that a dorsoventral gradient of Netrin-1 may, in fact, guide commissural axons to the ventral midline even in the absence of the floor plate (Matise et al., 1999).
While Netrin-1 is likely to play a major role in regulating the circumferential pathfinding of vertebrate commissural axons, several observations support the existence of additional guidance cues. For example, some commissural axons extend to the floor plate in Netrin-1-deficient mice, and floor plate explants from these mice retain their ability to reorient commissural growth cones/axons in vitro (Serafini et al., 1996). Furthermore, a function-blocking anti-DCC antibody does not abrogate the Netrin-dependent commissural growth cone turning response displayed in vitro (Keino-Masu et al., 1996). Netrin homologs are good candidates for guidance cues that may partially compensate for the loss of Netrin-1 in the knockout mice, and which might regulate commissural axon pathfinding through a DCC-independent signaling mechanism. Despite the lack of a mouse ortholog of chick Netrin-2, cDNA clones encoding three additional mouse Netrins have recently been isolated (Puschel, 1999; Wang et al., 1999; Nakashiba et al., 2000; Yin et al., 2000). While Netrin-3 (Puschel, 1999; Wang et al., 1999) and Netrin-G1, a novel glycosyl phosphatidylinositol (GPI)-linked Netrin (Nakashiba et al., 2000) are not expressed at the ventral midline of the developing spinal cord, mouse Netrin-4 is selectively expressed at the lateral margins of the floor plate (Yin et al., 2000). Thus, Netrin-4 is well-positioned to potentially function as a long-range chemoattractant for a subpopulation of commissural axons in the netrin-1 knockout mice. Alternatively, the so-called Netrin-synergizing activity (NSA), which is capable of potentiating Netrin-elicited commissural axon outgrowth in vitro (Serafini et al., 1994), may modulate the function of a Netrin-1-independent guidance system in the mutant mice. While it has recently been reported that NSA is likely to be composed of a denaturation-resistant basic protein(s) of about 25–35kD, its molecular identity has not been established (Galko and Tessier-Lavigne, 2000a). Thus, it is not yet possible to assess commissural axon pathfinding in mice lacking this activity.
Vertebrate Spinal Cord: Bone Morphogenetic Proteins
The essentially normal ventrally-directed growth exhibited by commissural axons in the Netrin-1 knockout mice suggests that non-floor plate-derived guidance cues play a significant role in regulating the early pathfinding of these axons. Since the collective effects of multiple positive and negative guidance cues are likely to shape the complete trajectory of a given class of axons (Tessier-Lavigne and Goodman, 1996), it is reasonable to expect that repulsive forces also make an important contribution to at least the initial phase of commissural axon pathfinding. For example, diffusible repulsive guidance cues located in dorsal regions of the embryonic spinal cord may provide commissural growth cones/axons with a “push from behind” to ensure that they properly extend in the ventral direction toward the midline.
The roof plate is a small, triangular-shaped structure that is comprised of primitive glial cells located at the dorsal midline of the embryonic spinal cord. This midline structure has long been thought to represent a barrier to axonal growth in the developing CNS (see Snow et al., 1990; Steindler, 1993). The roof plate-enriched expression of several chondroitin (Snow et al., 1991; Katoh-Semba et al., 1995; Engel et al., 1996; Katoh-Semba et al., 1998) and keratan (Cole and McCabe, 1991; McCabe and Cole, 1992; McCabe et al., 1992) sulfate proteoglycans which are capable of inhibiting neurite outgrowth in vitro, provides support for this hypothesis. More recently, it was demonstrated that the roof plate is capable of deflecting and reorienting Netrin-1-responsive commissural axons from E11 rat dorsal spinal cord explants (Augsburger et al., 1999). The ability of roof plate explants to deflect commissural axons from as far away as 100 μm suggested that this repulsive activity was associated with a diffusible molecule secreted by roof plate cells. Importantly, the repellent activity was shown to be restricted to the roof plate and specific for commissural axons (Augsburger et al., 1999). By employing a candidate molecule approach, it was further demonstrated that BMP7, a member of the TGFβ family of signaling molecules referred to as bone morphogenetic proteins (BMPs), mimics the repellent roof plate activity. While BMP6, BMP7, and Gdf7 (a member of a distinct group of BMPs; see Lee and Jessell, 1999) mRNA are each expressed within the roof plate at a time when commissural axons are initiating ventrally-directed growth, only recombinant BMP7 was shown to repel and reorient commissural axons as effectively as isolated roof plate explants (Augsburger et al., 1999). Importantly, both an antibody specific for BMP7 and a BMP antagonist were used to convincingly demonstrate that this secreted factor represents a major component of the roof plate-associated repellent activity. Additionally, roof plate explants derived from mice lacking BMP7 show a markedly decreased ability to deflect axons in vitro. A rapid and direct effect on commissural axon growth was demonstrated by the ability of BMP7 to promote the collapse of cultured commissural growth cones (Augsburger et al., 1999). Consistent with an in vivo role for BMP7, commissural axon pathfinding defects have apparently been observed in the spinal cords of BMP7 knockout mice (see Augsburger et al., 1999).
These findings represent an important new advance in our understanding of the molecular mechanisms that control commissural axon pathfinding. Interestingly, it now appears that two midline structures (floor plate and roof plate) play complementary roles in regulating early commissural axon pathfinding events. Additionally, by establishing them as repellent guidance molecules (at least in vitro), these data expand the role of BMPs in mammalian systems well beyond their known effects on the patterning of early neural cell types (see Lee and Jessell, 1999). However, it is important to note in this regard that BMP-related proteins have previously been implicated as regulators of axon guidance in invertebrates. Specifically, the recently identified unc-129 gene encodes a TGF-β superfamily member that is required for the guidance of pioneer motor axons along the dorsoventral axis of C. elegans (Colavita et al., 1998). Analogous to the dorsally-restricted distribution of BMP7 in the rodent spinal cord, unc-129 is expressed in dorsal, but not ventral, rows of body wall muscles (Colavita et al., 1998). Thus, TGF-β family members may play evolutionarily-conserved roles in controlling the guidance of circumferentially-projecting axons in the developing CNS.
With regard to the signaling mechanisms underlying the actions of BMP7 and UNC-129, it is important to note that BMPs typically interact with heterodimeric serine/threonine kinase receptors whose activation leads to nuclear translocation of SMAD proteins that then regulate the transcription of target genes (see Whitman, 1997; Kretzschmar and Massague, 1998 and references therein). However, the rapid time course of growth cone collapse is inconsistent with a transcriptional cascade mediating the repellent effects of BMP7 in rodents; additionally, in C. elegans, SMAD mutants do not exhibit pathfinding defects similar to those observed in unc-129 mutants (Colavita et al., 1998; Branda and Stern, 1999; Merz and Culotti, 2000). Thus, both BMP7 and UNC-129 may activate unconventional signaling pathways when they function as guidance cues for circumferentially-projecting axons in mammals and worms, respectively. It is anticipated that future work will not only elucidate these potentially novel signaling mechanisms, but also will more clearly define the in vivo role(s) of TGF-β proteins in axon guidance.
Drosophila Ventral Nerve Cord
Netrins are also required for commissural axon guidance in the developing Drosophila CNS. Netrin-A and netrin-B are expressed by midline cells during the initial stages of commissure formation in the ventral nerve cord (Harris et al., 1996; Mitchell et al., 1996). Deletion of both genes results in thinner than normal commissures (an indication that fewer than normal axons have crossed the midline), as well as occasional breaks in the longitudinal connectives (Harris et al., 1996; Mitchell et al., 1996). Genetic analyses demonstrate that netrin-A and netrin-B presumably play redundant roles at the midline. Deficiencies that delete either gene do not lead to defects in commissure formation, whereas each protein can rescue phenotypes associated with deletions that eliminate both genes. Interestingly, pan-neural ectopic expression of either Netrin-A or Netrin-B leads to defects in commissural and longitudinal axonal tracts that resemble those seen in the double mutants (Harris et al., 1996; Mitchell et al., 1996). This important result elegantly demonstrates that the precise spatial distribution of Netrin-A and Netrin-B, and not simply their presence, is required for the proper formation of commissural (and some longitudinal) tracts. Taken together, these data provide additional compelling support for the evolutionarily-conserved role of Netrins in commissural axon guidance. However, with regard to this latter point, it is important to note that many commissures appear normal in flies lacking both Netrins (Harris et al., 1996; Mitchell et al., 1996). Thus, just as is likely to be the case in mammals, Netrin-independent guidance cues are also capable of guiding commissural axons toward the midline of the developing Drosophila CNS. Good candidates include the recently identified schizo and weniger genes, each of which is required for proper commissure formation in flies (Hummel et al., 1999a; Hummel et al., 1999b).
The observation that the Drosophila DCC ortholog, Frazzled, is expressed at high levels on commissural and longitudinal axon tracts in the ventral nerve cord (Kolodziej et al., 1996) provides additional support for an evolutionarily-conserved role for Netrins in midline guidance. Reminiscent of the CNS phenotype detected in the absence of both netrin genes, thin or missing commissures characterize the ventral nerve cord of frazzled null mutants (Kolodziej et al., 1996). These findings suggest that Frazzled functions as a putative Netrin receptor in flies. However, an intriguing recent study provides support for an alternative Frazzled-dependent guidance mechanism that may also operate in the Drosophila CNS. This work was motivated by the surprising finding that a substantial fraction of Netrin protein was colocalized with Frazzled on axons in dorsal regions of the CNS, far from the site of Netrin synthesis at the midline (Hiramoto et al., 2000). Importantly, the dorsal accumulation of Netrin was not observed in the absence of Frazzled, and ectopic Netrin protein was detected on the surfaces of axons forced to misexpress Frazzled (Hiramoto et al., 2000). These data suggest that Frazzled may be capable of capturing and localizing Netrins at specific sites within the CNS. Strikingly, pioneer axons that extend from dMP2 neurons first grow laterally away from the midline and then turn into the longitudinal plane along a path that precisely abuts the dorsal domain of relocated Netrin protein (Hiramoto et al., 2000). This observation is further consistent with the possibility that Frazzled indirectly regulates the guidance of specific axonal populations by capturing and presenting Netrin at various points along their trajectory. Support for this hypothesis comes from the finding that dMP2 axons extend too far laterally and fail to turn posteriorly in flies lacking netrins and frazzled. Moreover, dMP2 axons respond in an attractive manner to an ectopic region of Netrin protein generated by mis-expression of the Frazzled extracellular domain (Hiramoto et al., 2000). Since dMP2 neurons do not themselves express Frazzled, these data are consistent with their axons responding to captured Netrin through a distinct receptor (Hiramoto et al., 2000). The novel “capture/relocation” mechanism elucidated in these studies could facilitate the efficient and widespread use (and reuse) of guidance cues, some of which may be selectively synthesized by midline cells, in the developing CNS.
C. elegans Body Wall
Netrins and their receptors were originally identified in screens for nematode locomotor mutants which exhibit uncoordinated (unc) movements due to defects in commissural and motor neuron guidance along the dorsoventral body axis. Genetic analyses have now established that interactions between the products of three genes, unc-5, unc-6, and unc-40, regulate the circumferential guidance of growth cones/axons (and the migration of cells) along the C. elegans body wall (for reviews see Culotti and Kolodkin, 1996; Hedgecock and Norris, 1997; Culotti and Merz, 1998; Branda and Stern, 1999; Merz and Culotti, 2000). More specifically, the vertebrate Netrin ortholog, UNC-6, is required for both dorsal and ventral guidance events (Hedgecock et al., 1990; Ishii et al., 1992; Wadsworth et al., 1996), the vertebrate DCC ortholog, UNC-40, is required for ventral (and to a lesser extent dorsal) guidance events (Hedgecock et al., 1990; Chan et al., 1996), and a novel Ig-superfamily member, UNC-5, is required exclusively for dorsally-directed axon guidance (Leung-Hagesteijn et al., 1992; Hamelin et al., 1993). Furthermore, UNC-40 and UNC-5 are each capable of acting cell autonomously to orient growth cones either towards or away, respectively, from sources of UNC-6 (see Culotti and Kolodkin, 1996; Hedgecock and Norris, 1997; Merz and Culotti, 2000 and references therein). Importantly, ectopic expression of UNC-5 in neurons that normally only express UNC-40 and are attracted to UNC-6, resulted in a reorientation of axon growth that reflects a repulsive response to UNC-6 (Hamelin et al., 1993). This observation formed the basis of a new genetic screen that has recently led to the isolation of eight genes whose products are required for UNC-5-directed axon outgrowth (Colavita and Culotti, 1998; Colavita et al., 1998). Given the phylogenetic conservation of other guidance cues and their receptors, further characterization of these genes should reveal significant insights into the mechanisms underlying pioneer axon guidance in both invertebrates and vertebrates.
A significant finding that has emerged out of the study of circumferentially-directed axon guidance in C. elegans is that UNC-6 can function as both an attractive and a repulsive guidance cue. Interestingly, a similar bifunctionality has been attributed to Netrin-1; commissural axons are attracted to a source of Netrin-1 both in vitro and in vivo (see above), while trochlear motor axons which grow dorsally away from the floor plate are repelled by Netrin-1 in vitro (Colamarino and Tessier-Lavigne, 1995a). However, the absence of trochlear motor axon pathfinding defects in Netrin-1 mutant mice (Serafini et al., 1996) suggests that Netrin-1 may not function as a major chemorepellent for these axons in vivo. Nevertheless, vertebrate homologs of UNC-5 are good candidates for functioning as receptors that mediate repellent activities of Netrin-1. Consistent with this possibility, two rat homologs of UNC-5, UNC5H1 and UNC5H2, have been shown to bind Netrin-1 and are expressed in multiple classes of neurons within ventral regions of the developing CNS (Leonardo et al., 1997).
Short-Range Guidance at the Midline
The ventral midline of the developing CNS represents a binary choice point for pathfinding axons. Upon reaching the midline, each axon must decide whether or not to cross over to the contralateral side of the CNS. There is now growing evidence from both vertebrate and invertebrate systems that a variety of short-range guidance mechanisms control midline crossing. Additional contact-dependent interactions ensure that commissural axons make the transition from circumferential to longitudinal growth only after they cross the midline and prevent recrossing events.
Vertebrate Spinal Cord
Axonin-1, NrCAM, and NgCAM.
The defects in commissural axon pathfinding detected in Sd (Bovolenta and Dodd, 1991) and Gli2-deficient (Matise et al., 1999; Fig. 4) mice strongly suggest that proper midline crossing requires contact-dependent interactions between commissural growth cones/axons and floor plate cells. Consistent with this notion, antibody perturbation experiments performed in ovo have shown that a direct interaction between two cell adhesion molecules of the Ig superfamily (or IgCAMs), axonin-1 on commissural axons, and NrCAM on floor plate cells, normally renders the floor plate permissive for commissural growth cone entry and passage (Stoeckli and Landmesser, 1995; Lustig et al., 1999). In these studies, soluble axonin-1 or function-blocking antibodies specific for either axonin-1 or NrCAM were repeatedly injected into chick embryos during the period of commissural axon pathfinding. Subsequently, DiI labeling revealed that up to 50% of commissural axons (as compared to control embryos) failed to cross the floor plate, and instead turned to project in the longitudinal direction on the ipsilateral side of the midline (Stoeckli and Landmesser, 1995). On the other hand, injection of antibodies against NgCAM, a chick L1 homolog reported to be expressed on both uncrossed and crossed segments of commissural axons, promoted the defasciculation of commissural axons, but resulted in no guidance defects (Stoeckli and Landmesser, 1995).
The results of these perturbation experiments suggested the possibility that upon reaching the midline, commissural axons must choose to extend along a NrCAM-lined pathway across the floor plate or, alternatively, along a longitudinally-oriented tract of NgCAM on the ipsilateral side of the floor plate. Interestingly, commissural axons choose the NgCAM pathway only when the NrCAM route is unavailable. This observation suggested that commissural axons prefer to grow on NrCAM (see Fitzli et al., 2000). Perhaps unexpectedly, a stripe/choice assay was recently used to show that in vitro, commissural axons grow equally well on NrCAM and NgCAM substrates, but that they prefer to grow on a mixed NrCAM/NgCAM substrate versus one containing NgCAM alone (Fitzli et al., 2000). Collectively, these observations suggest two distinct possibilities: 1) commissural axons grow across the midline because NrCAM is a more permissive substrate than NgCAM or 2) through its interaction with axonin-1, NrCAM plays a more active role in instructing axons to extend across the floor plate (Fitzli et al., 2000). The latter possibility is supported by the results of recent in ovo perturbation experiments (Fitzli et al., 2000).
In order to address the mechanism by which axonin-1 and NrCAM regulate midline crossing, an in vitro assay system was used to demonstrate that commissural growth cones/axons emanating from dorsal spinal cord explants fail to enter cocultured floor plate explants in the presence of reagents (anti-axonin-1, anti-NrCAM, or soluble axonin-1) that interfere with heterophilic axonin-1/NrCAM interactions; in these experiments, commissural growth cones situated at the periphery of floor plate explants displayed an intact and well-spread morphology (Stoeckli et al., 1997). Furthermore, in similar experiments, it was shown that commissural growth cones not only fail to enter floor plate explants, but also collapse in the presence of anti-axonin-1 antibodies (Stoeckli et al., 1997). Taken together with the in ovo results, these findings suggest that positive interactions between axonin-1 and NrCAM not only promote the entry of commissural growth cones/axons into the floor plate, but also render these axons insensitive to a collapse-inducing activity (CIA) that is unmasked by antibody perturbation (Stoeckli et al., 1997). It is important to note in this regard that a midline-associated inhibitory activity for commissural growth cones has previously been identified in grasshopper embryos (Myers and Bastiani, 1993), and that floor plate-derived B-class ephrins have been shown to promote the collapse of commissural growth cones in vitro (Imondi et al., 2000). Furthermore, Slit functions as a midline repellent for commissural axons in the Drosophila ventral nerve cord (see Battye et al., 1999; Kidd et al., 1999 and below) and, possibly, in the vertebrate spinal cord (see Zou et al., 2000).
F-spondin.
The results of recent studies aimed at perturbing the function of other floor plate-associated proteins further support a role for contact-dependent attractants/repellents in the regulation of midline crossing. F-spondin, an extracellular matrix protein that was originally identified in a subtractive hybridization screen for cDNAs encoding proteins enriched in the embryonic rat floor plate, is produced and secreted by floor plate cells (Klar et al., 1992). More recently, a chick homolog of F-spondin has been identified and soluble fusion proteins representing the thrombospondin type I repeat (TSR)-containing domain of this protein have been shown to bind to, and promote the outgrowth of, commissural growth axons in vitro (Burstyn-Cohen et al., 1999). In addition, it was also shown that repeated injection of the TSR fusion proteins into the central canal of chick embryos prevented many commissural axons from crossing the midline. Consequently, these axons were observed to turn orthogonally at the ipsilateral margin of (or within) the floor plate (Burstyn-Cohen et al., 1999). Thus, it appears that positive interactions between F-spondin and an as yet unidentified receptor(s) are also required for midline crossing in the chick spinal cord. Ultimately, it will be interesting to compare the phenotypes of mice deficient in mouse orthologs of F-spondin, axonin-1 or NrCAM, to determine which, if any, of these proteins are necessary and sufficient for midline crossing in mammals.
Altered-responsiveness.
Two molecular mechanisms have been postulated to account for the sharp transition in the direction of growth (i.e., transverse to longitudinal) exhibited by commissural axons after they cross the midline of the vertebrate spinal cord. In one model, decussated commissural axons turn orthogonally and grow in the longitudinal direction in order to maintain maximal contact with the floor plate which represents a more favorable substrate than adjacent regions of the spinal cord (Bovolenta and Dodd, 1990; Stoeckli and Landmesser, 1995 see Figs. 1A, B). In fact, the preferential expression of a number of IgCAMs in the floor plate supports such a selective adhesion mechanism (Colamarino and Tessier-Lavigne, 1995b). However, this model does not readily account for several aspects of commissural growth cone/axon pathfinding. For example, commissural axons execute orthogonal turns only after they cross the midline (see Figs. 1A, B). A straightforward interpretation of this observation is that commissural growth cones respond to turning cues that are selectively localized to the contralateral borders of the floor plate. However, given the bilateral symmetry of the spinal cord, commissural growth cones would be expected to encounter turn signals at both the contralateral and ipsilateral margins of the floor plate. Thus, a simple adhesion-based guidance system is not likely to account for the apparent ability of commissural growth cones to distinguish between the left and right sides of the floor plate. In addition, the observation that commissural axons do not recross the floor plate is inconsistent with a substrate preference model of midline guidance.
On the other hand, a model that does appear well-suited to account for these complex guidance behaviors postulates that floor plate contact alters the responsiveness of growth cones to bilaterally-distributed environmental guidance cues (see Dodd et al., 1988; Colamarino and Tessier-Lavigne, 1995b; Stoeckli and Landmesser, 1995). Mechanistically, this could be achieved by modifying the molecular composition of commissural growth cone surfaces. Thus, a prediction of this model is that commissural growth cones/axons up-regulate the expression of specific guidance receptors only after they cross the midline and consequently become responsive to guidance cues which direct them to turn into the longitudinal plane and prevent them from recrossing the floor plate.
Tag-1, L1, and NrCAM.
The transient expression of several IgCAMs on distinct segments of commissural axons supports an altered-responsiveness model of midline guidance. Strikingly, commissural axons extending toward the ventral midline in the rodent spinal cord express TAG-1, but not L1, while axonal segments on the contralateral side of the floor plate (representing a mixture of crossed-segments of commissural axons and ipsilaterally-projecting axons) express L1, but not TAG-1 (Dodd et al., 1988; Imondi et al., 2000; Fig. 3A, but also see Tran and Phelps, 2000). Consistent with an altered-responsiveness mechanism, it has been postulated that the switch in expression from TAG-1 to L1, presumably triggered by contact with the floor plate, delays the rostral turn exhibited by commissural axons until after they cross the floor plate (Dodd et al., 1988). NrCAM, another commissural axon-associated IgCAM, is expressed at low levels on commissural axons both as they extend toward the floor plate on the ipsilateral side of the midline, and as they project in the longitudinal direction on the contralateral side of the floor plate in the chick (Stoeckli and Landmesser, 1995) and mouse (Matise et al., 1999) spinal cord. In stark contrast, those segments of commissural axons residing within the ventral commissure (VC; short axonal tract coursing through the basal third of the floor plate; see Fig. 3A), express high levels of NrCAM (Stoeckli and Landmesser, 1995; Matise et al., 1999). These observations suggest that floor plate contact induces high levels of NrCAM on commissural axon segments contained within the VC (Matise et al., 1999). Despite these compelling expression patterns, functional roles for any of these IgCAMs in regulating midline guidance within the mammalian spinal cord have not been elucidated. On the other hand, recent analyses of mice deficient in L1 do reveal dramatic alterations in the formation of commissural tracts in the brain (for a recent review see Kamiguchi et al., 1998).
Careful analyses of commissural axon pathfinding in the spinal cords of Sd and Gli2-deficient mouse embryos do, however, support a role for the floor plate in regulating at least some of these axon segment-specific expression patterns. For example, in Sd mice, commissural axons that do not encounter the floor plate in more caudal regions of the spinal cord continue to express TAG-1 as they aberrantly extend either out of, or into, contralateral regions of the spinal cord (Bovolenta and Dodd, 1991). On the other hand, TAG-1 expression is appropriately down-regulated once commissural axons reach the ventral midline in more rostral, floor plate-containing regions of the these mice (Bovolenta and Dodd, 1991). Consistent with these observations, the few commissural axons that turn into the longitudinal direction after passing through the region formerly occupied by the floor plate in Gli2-knockout mice continue to express TAG-1 (Matise et al., 1999). Interestingly, these in vivo findings are consistent with the recent observation that only those commissural axons which encounter the floor plate as they extend out of E13 rat spinal cord explants, in vitro, down-regulate TAG-1 expression (see Zou et al., 2000 and below). Furthermore, in contrast to the enriched expression of NrCAM on axons coursing within the VC in wild-type mice, low levels of NrCAM are detected at the ventral midline in floor plate-lacking Gli2-deficient mice (Matise et al., 1999). Taken together, these findings demonstrate a clear in vivo role for the floor plate (and/or VIR cells) in modulating the segment-specific expression of TAG-1 and NrCAM on commissural axons.
In contrast, conflicting data make it difficult to establish whether floor plate contact regulates the expression of L1 on commissural axons. While L1 has been reported to be expressed on only those segments of commissural axons that pass through the floor plate in vitro (see Zou et al., 2000 and below), it appears that L1 is inappropriately expressed on commissural axons that have crossed through the midline in Gli2-deficient mice (Matise et al., 1999). Given these findings, it is important to consider floor plate-independent mechanisms which might modulate L1 levels on commissural axons. Reasonable alternatives include contact between commissural axons and other cells that they encounter en route to the floor plate, as well as axon-axon interactions. It is also important to point-out that in the spinal cords of Gli2 mouse mutants, some commissural axons have been observed to turn into the longitudinal direction in the absence of (and within the position formerly occupied by) the floor plate (Matise et al., 1999). These findings indicate that prior floor plate contact is not strictly required for commissural axon turning, in vivo. However, since the vast majority of axons fail to make the transition to longitudinally-directed growth in the spinal cords of these embryos, ventral midline cells are likely to play a major role in regulating the orientation of decussated commissural axon segments (see Matise et al., 1999).
In vitro assays.
Direct functional support for an altered-responsiveness model of midline guidance is provided by the results of two recent in vitro studies. The novel observation which facilitated the first study was that commissural axons were capable of extending across the midline in cultured explants of rat hindbrain (Shirasaki et al., 1998). This assay system was then used to show that once commissural growth cones/axons cross through the floor plate, they lose their responsiveness to an ectopic floor plate or source of Netrin-1 (Shirasaki et al., 1998: Fig. 6). In control experiments, commissural growth cones/axons that encountered an ectopic floor plate or source of Netrin-1 prior to crossing the midline, were capable of responding to chemoattractants secreted by the floor plate (Shirasaki et al., 1998; Fig. 6). One interpretation of these findings that is consistent with an altered-responsiveness model of midline guidance is that floor plate contact down-regulates the expression of the Netrin-1 receptor, DCC, on commissural growth cones/axons. However, since DCC is known to be expressed on both uncrossed and crossed segments of commissural axons in vivo (Keino-Masu et al., 1996), and since the distribution of DCC on the cultured commissural axons was not directly examined (Shirasaki et al., 1998), there exists no empirical support for this model. It is interesting to note in this regard that the metalloprotease inhibitor, IC-3, has recently been shown to increase DCC levels on spinal commissural axons, and consequently, potentiates Netrin-1-mediated outgrowth in vitro (Galko and Tessier-Lavigne, 2000b). This raises the intriguing possibility that commissural axon-containing spinal cord/hindbrain explants possess an endogenous metalloprotease activity that mediates the proteolytic degradation and functional inactivation of DCC (Galko and Tessier-Lavigne, 2000b). Thus, it is conceivable that this type of mechanism might account for the altered-responsiveness of commissural axons to Netrin-1 observed in this assay system. In any case, the finding that commissural axons lose their responsiveness to Netrin-1 after contacting the floor plate provides a plausible explanation for why these axons leave, and do not linger within, the confines of this chemoattractant-rich ventral midline structure. In fact, the finding that commissural growth cones/axons accumulate at the ventral midline in Gli2-knockout mice, adjacent to the high levels of ventricular zone-associated Netrin-1, is consistent with these axons remaining responsive to Netrin-1 in the absence of a floor plate (Matise et al., 1999).
Commissural axons lose responsiveness to chemoattractants after crossing through the floor plate. A and C: Schematic diagrams depicting two different manipulations performed on open-book explants of E13 rat hindbrain. B and D: Fluorescent micrographs representing the behavior of DiI labeled commissural axons in the experimentally-manipulated explants depicted in A and C, respectively, that were cultured for 2 days. A: A portion of the ventral spinal cord including the floor plate (hatched region in top diagram) was removed from an open-book explant of the developing hindbrain and the remaining pieces of the explant were abutted (bottom diagram). B: DiI-labeled commissural axons that enter the contralateral side of the midline without encountering a floor plate are attracted to an ectopic floor plate explant (eFP). Asterisks indicate DiI injection sites. C: A portion of the ventral spinal cord excluding the floor plate was removed from an open-book explant of the developing hindbrain (hatched region in top diagram) and the remaining pieces of the explant were juxtaposed (bottom diagram). D: DiI-labeled commissural axons that cross through the floor plate (FP) grow strictly in the transverse direction on the contralateral side of the midline and are not attracted to an eFP. Scale bar = 250 μm. “Reprinted with permission from Shirasaki R, Katsumata R, Murakami F. 1998. Change in chemoattractant responsiveness of developing axons at an intermediate target. Science 279: 105–107. Copyright 1998 American Association for the Advancement of Science.”
According to an altered-responsiveness model of midline guidance, commissural axons might also be expected to gain responsiveness to midline repellents after crossing through the floor plate. An interesting variation of the hindbrain assay system was recently developed in order to test this hypothesis. Specifically, explants consisting of hemisected spinal cord with or without an attached floor plate were cultured within a collagen gel (Zou et al., 2000). Strikingly, the so-called “post-crossing” (commissural) axons which extended out of the floor plate-attached explants, but not the “precrossing” axons which extended out of the floor plate-lacking explants, were shown to be repelled by both isolated floor plate and ventral spinal cord tissue (Zou et al., 2000). It was further demonstrated that post-crossing, but not precrossing, axons were repelled by recombinant forms of the known inhibitory guidance cues Slit-2 (see below), SEMA 3B (Semaphorin) and SEMA 3F (see Table 1 and Fig. 2). The spatiotemporal expression patterns of Slit-2, SEMA 3B and SEMA 3F in the embryonic mouse spinal cord suggest that Slit-2 and SEMA 3B are good candidates for mediating the repulsive activity associated with the floor plate, while SEMA 3F may account for the inhibitory activity associated with the ventral spinal cord (Zou et al., 2000). Interestingly, mice deficient in NPN-2, a high affinity receptor for SEMA 3B and SEMA 3F (Chen et al., 1997), exhibit pathfinding defects that are consistent with commissural axons encountering repulsive guidance cues in the floor plate and within the ventral spinal cord after they cross the midline (Zou et al., 2000). While the in vitro functional results provide strong support for commissural axons selectively gaining responsive to repellent guidance cues after crossing through the floor plate, a role for NPN-2 in mediating these responses is difficult to reconcile with the observation that NPN-2 protein is expressed on both uncrossed and crossed segments of commissural axons in vivo and in vitro (see Zou et al., 2000). Furthermore, it is possible that vertebrate Robo homologs (see Table 1, Figs. 2, 3 and below) or other commissural axon-associated receptors mediate Slit-2-dependent repulsion in vitro or in vivo. Nevertheless, taken together with the results of the hindbrain studies, these data suggest that commissural axons not only lose responsiveness to chemoattractants, but also gain responsiveness to chemorepellents after they cross the midline in the vertebrate spinal cord.
Ephs and ephrins.
Receptor-ligand pairs that mediate repulsive axon guidance through short-range interactions are likely to represent additional candidates for the molecular components of an altered-responsiveness-based guidance system. The Eph (erythropoietin-producing hepatocellular) receptor protein tyrosine kinases (RPTKs) constitute the largest class of RPTKs and are characterized by an extracellular domain containing a unique cysteine-rich motif and two fibronectin type III repeats (see Tuzi and Gullick, 1994). There exists at least 14 Eph receptors and eight membrane-associated ligands referred to as ephrins. Eph receptors are divided into two subclasses: EphA receptors that preferentially interact with A-class ephrin ligands and EphB receptors that preferentially interact with B-class ephrin ligands. While both types of ephrins are associated with the cell surface, A-class ephrin ligands are tethered to the membrane through a glycosyl phosphatidylinositol (GPI) linkage, whereas B-class ephrin ligands possess transmembrane domains (see Gale et al., 1996; Flanagan and Vanderhaeghen, 1998 and references therein). The ability of B-class ephrins to mediate contact-dependent repulsion in a number of neural systems (see Flanagan and Vanderhaeghen, 1998; Holder and Klein, 1999; Wilkinson, 2000) makes them particularly well-suited to function as highly localized repellent guidance cues at the ventral midline.
We have recently demonstrated by in situ hybridization that three B-class ephrins are expressed in the floor plate and that EphB1 is expressed by multiple populations of commissural and ipsilaterally-projecting interneurons in the developing mouse spinal cord (Imondi et al., 2000). Moreover, we have confirmed that B-class ephrin protein is expressed at the lateral margins of the floor plate, and further demonstrated that EphB1 protein is specifically expressed on decussated commissural axon segments (Imondi et al., 2000). The segregation of EphB1 to crossed axonal segments is strikingly similar to that displayed by L1 in the developing rodent spinal cord (Dodd et al., 1988; Imondi et al., 2000; Fig. 3A) and is consistent with the recent finding that the only identified Drosophila Eph receptor, referred to as Dek, is selectively expressed on longitudinally-projecting axons in the developing ventral nerve cord (Scully et al., 1999; Fig. 3B). Together with our finding that B-class ephrins promote the collapse of commissural growth cones in vitro (Imondi et al., 2000), these expression patterns support the existence of an Eph/ephrin-based repellent guidance system that operates through an altered-responsiveness mechanism at the midline of the developing vertebrate spinal cord.
In very recent studies, we have re-examined the trajectory of commissural axons on the contralateral side of the midline in relation to B-class ephrin expression. In contrast to most models of midline guidance (but see Bovolenta and Dodd, 1990), which presuppose that commissural axons extend alongside the floor plate throughout most of their trajectory, we now find that decussated segments of commissural axons travel in close proximity to the floor plate for only a short distance. Subsequently, these axons grow away from the midline into more dorsal regions of the spinal cord, where they execute a second rostral turn into the longitudinal plane alongside a dorsal domain of B-class ephrin expression (R Imondi and Z Kaprielian, unpublished observations). Thus, these new findings suggest that B-class ephrins may demarcate barriers to commissural axon growth alongside the floor plate (for short distances), as well as in more dorsal regions of spinal cord, where they may act to specify the dorsoventral position of the longitudinal commissural axon tract. Interestingly, recent genetic studies in C. elegans support a positive role for the basement membrane component Nidogen in establishing the placement of longitudinally-projecting nerves at various positions along the dorsoventral body axis (Kim and Wadsworth, 2000). With regard to the specific mechanism of Eph-ephrin mediated guidance events in the spinal cord, it is interesting to consider the recent observation that regulated, metalloprotease-dependent cleavage of A-class ephrins rapidly converts the initial binding of A-class receptors and ligands into a repulsive interaction (Hattori et al., 2000). Given the presence of conserved metalloprotease recognition motifs in B-class ephrins, it is tempting to speculate that the initial attractive interaction between EphB1 and B-class ephrins accounts for growth along the contralateral margin of the floor plate, and that the subsequent cleavage and repulsive guidance events account for growth away from this structure into the dorsal spinal cord.
While validation of an Eph/ephrin-based model of repulsive midline guidance in the spinal cord awaits the analyses of loss-of-function mutants, recent studies of commissure formation in the brain support a role for B-class Eph receptors and ephrins in the control of midline crossing. For example, mice deficient in EphB2 (Henkemeyer et al., 1996), or both EphB2 and EphB3 (Orioli et al., 1996), exhibit defects in major midline-crossing tracts in the forebrain, and inner ear efferents (IEE) select inappropriate pathways along which to extend at the midline of the hindbrain (Cowan et al., 2000). Furthermore, misexpression studies in Xenopus tadpoles demonstrate that precocious expression of B-class ephrins in the optic chiasm leads to misrouting of Eph-expressing retinal axons (Nakagawa et al., 2000).
Drosophila Ventral Nerve Cord
Comm, Robo, and Slit.
Molecular genetic studies performed in Drosophila provide the most compelling support for an altered-responsiveness guidance system operating at the midline of the developing CNS. Large-scale screens for Drosophila mutants in which too many or too few axons cross the midline have resulted in the isolation of three genes that collectively control midline crossing. In commissureless (comm) mutants, the CNS is devoid of essentially all commissural tracts and contains only the two longitudinal connectives located on either side of the midline (see Seeger et al., 1993; Fig. 7). Consistent with this striking phenotype, commissural growth cones/axons properly orient to, but never cross the midline in this mutant (Seeger et al., 1993; Tear et al., 1996). The comm gene product, Comm, is likely to be directly required for midline crossing since the differentiation of midline-associated glia and neurons are normal in these embryos. Comm is a novel type I transmembrane protein that is initially localized to midline cells and then is apparently transferred (by an unknown mechanism) to commissural axons as they cross the midline (Tear et al., 1996). Interestingly, the primary structure of Comm does not contain motifs that are present within other proteins which mediate axon guidance. While biochemical analyses reveal that Comm is strictly associated with membranes, anti-Comm antibodies strongly label intracellular organelles (e.g., Golgi complex and endosomes), and only weakly label the cell surface (Tear et al., 1996). One interpretation of these findings that is consistent with Comm functioning as an extracellular guidance cue is that Comm transiently exists at the cell surface before being cleared via endocytosis (see below). Support for this hypothesis comes from the recent identification of a four residue motif, YXX∅︁ (a tyrosine followed by two random amino acids and a hydrophobic amino acid) within the cytoplasmic domain of Comm (Wolf et al., 1998a). This site mediates the association of various proteins with Adaptin/Clathrin complexes, which consequently facilitates their rapid endocytosis (see Kirchhausen et al., 1997). Consistent with this notion, wild-type Comm, but not a truncated mutant that lacks the YXX∅︁ sequence, is rapidly endocytosed in embryonic Drosophila muscle cells (Wolf et al., 1998a).
Schematic representations of midline guidance phenotypes observed in the developing ventral nerve cord of several Drosophila mutants. The wild-type ventral nerve cord consists of longitudinal connectives on either side of the midline separated by a pair of commissures (the AC and PC) in each segment. In comm LOF and robo GOF++ mutants, no axons cross the midline as indicated by the absence of the AC and PC in each segment. In robo LOF and comm GOF+ mutants, many axons inappropriately cross and recross the midline. This is reflected by the appearance of thick commissures and thin longitudinal connectives. In slit LOF and comm GOF+++ mutants, there exists a single longitudinal tract at the midline. LOF refers to loss-of-function and GOF refers to gain-of-function. The number of plus signs following GOF indicates the relative dosage of the overexpressed transgene. See text for relevant references.
In robo mutants, the ventral nerve cord contains thickened commissures that reflect excessive midline crossing events (Seeger et al., 1993). Antibody labeling demonstrates that axons which normally pioneer ipsilateral projections (e.g., pCC, vMP2) now cross the midline, while contralaterally-projecting axons (e.g., SP1) recross the midline multiple times (Seeger et al., 1993; Kidd et al., 1998a; Kidd et al., 1998b; Fig. 7). Interestingly, only those axons (both crossed segments of commissural axons and ipsilaterally-projecting axons) which project within the innermost longitudinal connectives aberrantly cross and re-cross the midline in robo mutants (Kidd et al., 1998b). The protein encoded by robo (Robo) is a novel member of a unique family of IgCAMs (Kidd et al., 1998a). Strikingly, robo mRNA is widely expressed by many or most developing neurons, but Robo protein is regionally expressed in an axon-segment-specific manner in wild-type embryos. Specifically, axons extending toward or across the midline express very low levels of Robo, whereas crossed segments of commissural axons as well as ipsilaterally-projecting axons express high levels of Robo (see Kidd et al., 1998a; Kidd et al., 1998b; Fig. 3B). These studies are consistent with high levels of Robo being present on those axon segments which do not normally cross the midline or cross the midline only once.
Recent evidence demonstrates that Comm and Robo act in concert to control midline crossing. The first suggestion that this might be the case was the observation that comm; robo double mutants exhibit a robo phenotype (excessive midline crossing; Kidd et al., 1998b). This implies that in the absence of Robo, Comm is no longer required for axons to cross the midline. Misexpression experiments also provide evidence for a regulatory interaction between Robo and Comm. Specifically, ectopic expression of Comm throughout the CNS results in a significant reduction in Robo protein levels and consequently aberrant midline crossing and recrossing events. Furthermore, overexpression experiments demonstrate that inappropriately high levels of Comm lead to more severe robo-like phenotypes (Kidd et al., 1998b). Taken together, these findings suggest that Comm may act locally to keep Robo levels low on commissural axons (Kidd et al., 1998b). Thus, Robo apparently functions as a “gatekeeper” that controls midline crossing events; growth cones/axons expressing high levels of Robo are prevented from crossing the midline, while midline crossing is permitted for growth cones/axons that do not express, or express low levels of, Robo (Kidd et al., 1998a; Kidd et al., 1998b). In this way, high Robo expression on ipsilaterally-projecting axons, as well as on crossed-segments of commissural axons, would normally prevent midline crossing.
The mechanism by which Comm regulates Robo levels remains speculative. It has been proposed that at the midline, Comm is transferred from midline cells to the surfaces of growth cones. Consistent with the possibility that Comm is rapidly endocytosed, it seems reasonable to expect that after being transferred to commissural axons, Comm and Robo may be cointernalized; the clearance of Robo from the cell surface would render only those segments of commissural axons that contact Comm-expressing midline cells unresponsive to a presumably repellent Robo ligand.
The analyses of comm and robo mutants outlined above provide strong support for an altered-responsiveness model of midline guidance operating in the Drosophila CNS. However, a recent study has demonstrated that axon pathfinding can proceed normally despite aberrant growth cone decisions at the midline in both comm and robo mutants. For example, RP3 and V motor axons that are inappropriately prevented from crossing the midline in comm mutants follow mirror image trajectories and ultimately innervate corresponding muscle targets on the ipsilateral side of the CNS. Conversely, RP2 and aCC motor axons that abnormally cross the midline in robo mutants subsequently extend along mirror image pathways and connect with appropriate synaptic targets on the contralateral side of the CNS (Wolf and Chiba, 2000). Thus, at least for the identified motor neurons analyzed in this study, it appears that the previous molecular experience of a given growth cone does not effect its ability to respond normally to guidance cues encountered along subsequent segments of its trajectory (Wolf and Chiba, 2000).
Until recently, the identity of the Robo ligand had remained obscure. A re-evaluation of previous mutants has now identified Slit as a candidate midline-associated repulsive ligand for Robo (for a recent review see Brose and Tessier-Lavigne, 2000). The slit loss-of-function phenotype is characterized by a fusion of the longitudinal connectives and their ultimate collapse onto the midline of the CNS (Rothberg et al., 1990; Fig. 7). This phenotype was originally thought to represent a defect in the specification of midline cell fate (Rothberg et al., 1988; Rothberg et al., 1990). Although a ventral displacement of midline cells results in the fusion of the longitudinal connectives shortly after commissure formation, the primary axonal defect in slit loss-of-function mutants is the inability of commissural axons to leave the midline (Sonnenfeld and Jacobs, 1994). In retrospect, this is precisely the type of defect expected for a gene that purportedly functions as a midline repellent.
More recent studies have now provided genetic evidence supporting a role for Slit as the repulsive Robo ligand. The key result which suggested this possibility was the finding that the strongest gain-of-function Comm phenotype resembles the “collapsed-midline” phenotype exhibited by slit mutants (Kidd et al., 1999). Subsequently, it was shown that flies which carry a single mutant copy of robo and slit display a robo-like phenotype (Kidd et al., 1999). This observation supports a receptor-ligand relationship between Robo and Slit (Kidd et al., 1999). Consistent with this notion, Robo and Slit proteins serve as binding partners for each other (Brose et al., 1999). Taken together, these data strongly suggest that Slit is the Robo ligand. Given this conclusion, it is important to reconsider the phenotypes of robo and slit mutants. In robo mutants, axons freely cross and recross the midline, whereas in slit mutants, axons enter, but never leave, the midline. Thus, in the absence of Robo, another receptor must respond to Slit to ensure that growth cones/axons do not linger at the midline, without interfering with their ability to cross the midline (see Kidd et al., 1999). A good candidate for this additional Slit receptor is Robo2, a close relative of Robo that is expressed on developing CNS neurons. A straightforward prediction of this model is that the phenotype of robo/robo2 double mutants should resemble the phenotype of slit single mutants (see Kidd et al., 1999).
Slit is a large extracellular matrix protein that is secreted by midline glia and, interestingly, is also found associated with the surfaces of axons (Rothberg et al., 1990). Consistent with a role as an inhibitory guidance cue at the midline, it has recently been shown that commissural axons avoid ectopically-expressed Slit in the Drosophila CNS (Battye et al., 1999). These findings further support the notion that Slit functions as a midline-associated repellent ligand for Robo. However, the recent identification of Drosophila mutants such as karussell, in which axons are observed to inappropriately “circle” in the vicinity of the midline, suggests that multiple repulsive guidance systems operate at the midline (Hummel et al., 1999b).
Several studies support the existence of evolutionarily-conserved Robo/Slit midline guidance systems. For example, the C. elegans Robo ortholog, sax-3, was recently identified in a screen for axon guidance mutants. Similar to the phenotype displayed by robo loss-of-function flies, axons are observed to inappropriately cross between the left and right axon bundles of the ventral nerve cord in sax-3 mutants (Zallen et al., 1998). While a single slit gene has been identified in the nematode, its role in axon guidance remains to be established (see Brose et al., 1999 and references therein). Nevertheless, these findings suggest that Robo, and possibly Slit, may also mediate midline guidance in C. elegans. The zebrafish astray mutant was originally identified in a large-scale screen for defects in the retinotectal projection (see Karlstrom et al., 1996). Retinal axons in astray mutants exhibit a variety of pathfinding defects, including a propensity to recross the midline. The midline recrossing phenotype suggested that astray could represent a defect in a zebrafish robo homolog. Very recent studies have, in fact, identified the gene responsible for astray as a robo-2 ortholog, and cDNAs encoding three zebrafish Robos have now been isolated by degenerate RT-PCR (C Fricke, J-S Lee, and C-B Chien, unpublished observations). In addition, a recent time-lapse analysis reveals that astray growth cones exhibit an abnormal morphology, an increased tendency to leave the optic pathway, and fail to correct initial pathfinding errors (L. Hutson and C-B Chien, unpublished observations).
Mammalian homologs of Robo and Slit have also been identified and the results of preliminary expression studies are consistent with these proteins regulating midline guidance in the developing spinal cord (see Brose and Tessier-Lavigne, 2000). Consistent with the distribution of their Drosophila counterparts, mRNA expression studies reveal that two rodent Robo homologs (Robo-1 and Robo-2) appear to be expressed by spinal commissural neurons (Kidd et al., 1998a; Brose et al., 1999), and three rodent Slit homologs (Slit-1, Slit-2, and Slit-3) are expressed in the floor plate (Brose et al., 1999). However, selective expression of Robo-1 and Robo-2 proteins on decussated segments of commissural axons has not been demonstrated (see Fig. 3A). Consistent with interactions between mammalian Robos and Slits mediating repulsive axon guidance, Slit-2 binds the surfaces of Robo-expressing cells (Brose et al., 1999) and repels spinal motor axons in vitro (Brose et al., 1999). More importantly, Slit-2 selectively repels commissural axons that have passed through the floor plate in vitro (see Zou et al., 2000 and above). Thus, a Robo-Slit repulsive guidance system that is mechanistically similar to the one thought to operate in the Drosophila, zebrafish (and possibly the worm) CNS may also regulate midline crossing in the developing mammalian spinal cord. However, it is important to note that the C. elegans genome does not encode a Comm ortholog (see Chisholm and Tessier-Lavigne, 1999), and that a clear vertebrate Comm homolog has not yet been identified (Tear et al., 1996; Kidd et al., 1998b). Thus, Comm may not be required for midline crossing in either worms or vertebrates.
RPTPs.
Receptor-linked tyrosine phosphatases (RPTPs) also regulate midline crossing in the Drosophila ventral nerve cord. RPTPs regulate tyrosine dephosphorylation in growth cones and thus reverse reactions catalyzed by tyrosine kinases. Consistent with potential roles as repellent receptors, it has recently been demonstrated that the Drosophila RPTPs, DLAR and DPTP10D are, like Robo, selectively localized to longitudinal axonal tracts in the embryonic ventral nerve cord (see Sun et al., 2000; Fig. 3B). Furthermore, many longitudinally-growing axons are rerouted across the midline in flies lacking DPTP10D and another neural RPTP, DPTP69D, (see Sun et al., 2000; Fig. 8). Interestingly, antibody labeling and lineage tracing studies reveal that the innermost (or most medial) longitudinal tracts, which contain the axons that aberrantly recross the midline in robo mutants, are not affected in RPTP double mutants. Rather, more laterally-positioned axons are observed to cross and recross the midline in flies lacking particular combinations of RPTPs (Sun et al., 2000). The finding that DPTP10D and DPTP69D genetically interact with robo, slit and comm, provides direct support for the possibility that these two RPTPs regulate Robo/Slit repulsive signaling at the midline, possibly by modulating tyrosine phosphorylation events mediated by repulsive Robo/Slit interactions (Sun et al., 2000). This is consistent with the observation that pharmacological inhibition of tyrosine kinase activity in grasshopper embryos results in aberrant midline recrossing events (Menon and Zinn, 1998; Sun et al., 2000).
Midline crossing defects in DPTP10D DPTP69D double mutant Drosophila embryos. Lineage tracing was performed by using DiI to label all the progeny of identified neuroblasts (NBs) in the Drosophila ventral nerve cord. Individual neuroectodermal cells were labeled at stage 8 and the embryos were allowed to develop until stage 17. DiI-labeled NBs arising from the injected cells were then identified based on their positions, and the cell bodies and axons of the NB progeny were visualized by confocal microscopy. Each large panel consists of a confocal z-series on the left and a diagram of the clone in relation to morphological landmarks on the right (the ventral nerve cord is depicted on the left side and the body wall musculature is represented on the right side). The NB 2–5 lineage generates 15–22 cells by stage 17, of which 8–16 are intersegmental interneurons. Top: In wild-type embryos, 4–8 intersegmental interneurons from the NB 2–5 lineage extend axons anteriorly toward the brain on both the contralateral (for up to 10 segments) and ipsilateral (for up to about 5 segments) sides of the midline, and a single motorneuron from this lineage projects an axon via the ISNd that innervates muscles 15–17. Bottom: In DPTP10D DPTP69D double mutant embryos, contralateral interneurons cross the midline and project anteriorly as in wild-type embryos. Strikingly, however, after traveling for only about 2 segments, these axons recross the midline and project posteriorly within the ipsilateral longitudinal connective. Ipsilateral interneurons grow anteriorly for only a short distance and then stop without crossing segmental boundaries. The motorneuron fails to exit the CNS. “Reprinted with permission from Sun Q, Bahir S, Schmid A, Chia W, Zinn K. 2000. Receptor tyrosine phosphatases regulate axon guidance across the midline of the Drosophila embryo. Development 127:801–812. Copyright 2000 The Company of Biologists Limited.”
Derailed.
An important recent study has identified a midline repulsive guidance system that controls the commissure choice displayed by midline-crossing axons in a Robo/Comm/Slit-independent manner (Bonkowsky et al., 1999). The novel RPTK known as Derailed (Drl) is expressed by a unique subset of interneurons and motor neurons that selectively extend axons through the AC (Callahan et al., 1995; Bonkowsky et al., 1999). Strikingly, Drl protein is detected on only those axonal segments that are actively projecting into and along the AC, and not on these same axons as they project out of the AC (see Bonkowsky et al., 1999; Figs. 3B, 9). In the absence of Drl, these axons often project aberrantly into the PC, and misexpression of Drl forces axons that normally project within the PC to cross the midline through the AC (Bonkowsky et al., 1999). Taken together with the presence of a putative ligand for Drl in the immediate vicinity of the PC, these data suggest that Drl functions as a repellent guidance receptor that dictates the commissure choice displayed by a subset of midline-crossing axons (Bonkowsky et al., 1999). Since a likely mammalian homolog of Drl, referred to as RYK (Hovens et al., 1992), is expressed in the developing rat CNS (Kamitori et al., 1999), it is tempting to speculate that this unique RPTK also plays a significant role in regulating axon guidance in the developing vertebrate CNS. However, no pathfinding defects were reported in the recent characterization of RYK null mice (Halford et al., 2000). Nevertheless, the craniofacial abnormalities observed in these mice, as well as the finding that RYK exists in a complex with EphB2 and EphB3, suggest a role for RYK in Eph receptor-mediated signaling.
Anterior commissure-specific expression of Drl in the embryonic Drosophila ventral nerve cord. a: In a wild-type embryonic ventral nerve cord, anti-HRP labels all axons contained within the AC and the PC, as well as those traveling within the longitudinal connectives. b: Anti-Drl selectively labels those axons contained within the AC. c: Merge of the images presented in a and b. “Reprinted with permission from Nature, Bonkowsky JL, Yoshikawa S, O'Keefe DD, Scully AL, Thomas JB. 1999. Axon routing across the midline controlled by the Drosophila Derailed receptor. 402:540–544. Copyright 1999 Macmillan Magazines Limited.”
Attraction Versus Repulsion.
The studies outlined above strongly support the view that commissural growth cones alter their responsiveness to attractive and repulsive environmental guidance cues as they cross the midline in both vertebrates and invertebrates. In the vertebrate spinal cord, for example, commissural growth cones are first attracted to floor plate-derived chemoattractants; subsequently, these growth cones appear to lose responsiveness to positively-acting guidance cues and simultaneously gain responsiveness to repulsive signals as they cross the midline. Presumably, these processes prevent commissural axons from stalling at the midline and ultimately facilitate their growth toward appropriate targets. Several recent studies indicate that the response of a growth cone to a given guidance cue is determined by the internal state of the neuron and that this response can be regulated by second messenger systems. Additional experiments reveal that attraction and repulsion are encoded by the cytoplasmic domains of specific guidance receptors (for recent reviews see Garrity, 1999; Mueller, 1999; Seeger and Beattie, 1999; Song and Poo, 1999). Below, we summarize some of the most significant results.
One key set of studies was motivated by the observation that growth cones which arise from cultured Xenopus spinal neurons exhibit a chemoattractive response to a pulsatile application of Netrin-1 in vitro (Ming et al., 1997). Strikingly, this positive response was converted into a repulsive response in the presence of a competitive analog of cAMP (Rp-cAMPS) or an inhibitor of protein kinase A (KT5720; Ming et al., 1997). Thus, the same growth cone may exhibit opposite turning responses to Netrin-1 depending on the levels of cytosolic cAMP. The adenosine receptor, A2b, has recently been shown to induce cAMP production after binding to Netrin-1 in vitro (Corset et al., 2000). Accordingly, it has been suggested that this interaction may represent one of the critical signaling events that regulates the attraction of growth cones to Netrin-1 (see Corset et al., 2000).
Regardless of the nature of the response induced by Netrin-1 (i.e., attractive vs. repulsive) DCC appears to play a crucial role in mediating downstream signaling events. Thus, both attractive and repulsive responses to Netrin-1 in the growth cone turning assay were abolished in the presence of a blocking antibody against DCC (Ming et al., 1997). In addition, the presence of extracellular calcium was shown to be absolutely required for both attraction and repulsion (Ming et al., 1997). These important findings suggest that the response of a pathfinding growth cone to a specific guidance cue may not be predetermined. Rather, the “sign” of the response is likely to be critically dependent on the status of cytosolic cAMP activity which, in turn, may be dependent on the full complement of environmental signals that the growth cone receives at a given point along its trajectory (Ming et al., 1997). For example, since laminin is capable of converting Netrin-1-induced growth cone attraction into repulsion through a decrease in cytosolic levels of cAMP in vitro, it has recently been suggested that growth cones navigating in regions of the CNS that coexpress laminin and Netrin-1 may be driven into areas where only Netrin-1 is expressed (Hopker et al., 1999). This type of mechanism might ultimately contribute to the proper placement of axonal tracts in molecularly complex regions of the developing CNS.
Complementary approaches were recently used in two related studies to address the question of how a growth cone responds to a bifunctional guidance cue in vitro and in vivo (for a recent review see Seeger and Beattie, 1999). In one set of experiments, the turning assay described above was used to show that exogenous expression of UNC-5 in Xenopus spinal neurons converts the attractive response of their axons/growth cones to Netrin-1 into a repulsive response (Hong et al., 1999). Importantly, the repulsive response of UNC-5-expressing axons was eliminated in the presence of a function-blocking antibody against DCC (Hong et al., 1999). It was further demonstrated that a ligand-induced interaction between the cytoplasmic domains of UNC-5 and DCC underlies the conversion of Netrin-1-induced attraction into repulsion (Hong et al., 1999). These findings suggest that UNC-5 proteins play an evolutionarily-conserved role in mediating repulsive events that require DCC. Interestingly, when a chimeric receptor consisting of the ectodomain of DCC and the cytoplasmic domain of UNC-5 was expressed in the Xenopus neurons, the corresponding growth cones also exhibited a repulsive response to Netrin-1 (Hong et al., 1999). Thus, repulsion appears to be encoded in the cytoplasmic domain of UNC-5.
An independent study used the Drosophila midline as an in vivo system within which to test whether the cytoplasmic domains of Frazzled (Fra) and Robo mediate attractive and repulsive growth cone responses, respectively (Bashaw and Goodman, 1999). Specifically, chimeric receptors representing the extracellular domain of Fra and the intracellular domain of Robo (Fra-Robo) or, conversely, the extracellular domain of Robo and the intracellular domain of Fra (Robo-Fra), were generated and expressed in all neurons of the embryonic Drosophila CNS (Bashaw and Goodman, 1999). Strikingly, axons expressing high levels of Fra-Robo were repelled by Netrin-expressing midline cells and failed to cross the midline (Bashaw and Goodman, 1999). On the other hand, axons expressing high levels of Robo-Fra were attracted to Slit-expressing midline cells and many of them inappropriately crossed the midline (Bashaw and Goodman, 1999). These findings indicate that the cytoplasmic domains of Robo and Fra encode repulsion and attraction, respectively. Taken together, the results of these two provocative studies demonstrate that guidance receptors (at least UNC-5, Robo and Fra) are modular; the cytoplasmic domains determine the nature of the response to a given guidance cue, while the ectodomains determine ligand binding specificity (Bashaw and Goodman, 1999; Hong et al., 1999; Seeger and Beattie, 1999). However, it is important to point-out in this context that the kinase domain of EphB2 has been shown to be dispensable for the formation of commissural tracts in the mouse brain (Henkemeyer et al., 1996). More generally, these new studies suggest that it may not be possible to classify a given guidance cue as strictly attractive or repulsive (Bashaw and Goodman, 1999).
Perspectives
Over the last several years, dramatic progress has been made in elucidating molecular mechanisms that control axon guidance at the midline of the developing vertebrate and invertebrate CNS. However, these new findings raise new questions and many mysteries remain. Below, we discuss several of the unresolved issues which future studies in this field must address, particularly with respect to axon guidance at the midline of the vertebrate CNS.
Novel Midline Markers
The floor plate is a key source of guidance information for pathfinding commissural axons in the developing vertebrate spinal cord. A large number of proteins that are preferentially expressed in the floor plate (see Colamarino and Tessier-Lavigne, 1995b; Table 1) are presumably good candidates for guidance cues that commissural growth cones encounter as they approach, cross through, and exit this structure. However, the function(s) of many of these proteins, in the context of midline guidance, has not been elucidated. VEMA (for VEntral Midline Antigen) is a particularly interesting example of a floor plate-associated protein whose function remains to be determined. We originally identified VEMA through the use of a monoclonal antibody (mAb) that preferentially labels the floor plate in the developing rat CNS (Zhu et al., 1998). VEMA is a 28 kD membrane-associated protein that possesses one putative transmembrane domain near its N-terminus. Consistent with our original mAb labeling patterns (Zhu et al., 1998), in situ hybridization demonstrates that VEMA mRNA is specifically localized to the floor plate in the developing rat spinal cord (see Runko et al., 1999). In addition to being expressed in both the floor plate and roof plate in the developing mouse spinal cord, we have recently found that VEMA is selectively expressed in guidepost cells which control the pathfinding of retinal ganglion cell axons in the vertebrate optic chiasm (E Runko and Z Kaprielian, unpublished results). Consistent with nonexistent or low levels of VEMA on the cell surface, the primary structure of VEMA does not contain any motifs that might suggest an obvious role for this protein in mediating contact-dependent interactions which regulate axon guidance. Nevertheless, VEMA does contain several distinct sorting motifs that are thought to facilitate the rapid clearance of a variety of proteins from the cell surface (see Runko et al., 1999). These structural characteristics are consistent with the predominantly intracellular localization displayed by the VEMA protein (Runko et al., 1999). Taken together, these findings support the possibility that VEMA might be targeted to, and function at, the cell surface where it may exist transiently before being cleared via endocytosis. It is interesting to note in this regard that the midline cell-associated Comm protein is also predominantly associated with intracellular membranes and contains internalization motifs of the type present in the VEMA sequence (Tear et al., 1996; Wolf et al., 1998a). Ongoing studies in our lab are aimed at analyzing commissural axon pathfinding in loss-of-function VEMA mutants.
Interestingly, the C. elegans genome contains a putative VEMA ortholog. Through the use of GFP reporter constructs, we have recently found that VEMA is likely to be expressed in a subpopulation of early-developing neurons that extend axons which pioneer the ventral nerve cord (E Runko and Z Kaprielian, unpublished observations). The expression of VEMA in multiple intermediate targets within the developing vertebrate CNS, and in neurons which pioneer early axonal tracts in the worm, suggests that VEMA may play an evolutionarily-conserved role in regulating axonogenesis and/or axon pathfinding.
A small subset of genes/proteins are known to be expressed in both the floor plate and roof plate. These include Slit-1, Slit-2, Slit-3 (see Brose and Tessier-Lavigne, 2000 and references therein), NPN-2 (Chen et al., 1997), B-class ephrins (Imondi et al., 2000), Annexin IV (Hamre et al., 1996), BMP-6 (see Lee et al., 2000), and VEMA (Runko et al., 1999). Given that the roof plate, like the floor plate, is thought to be an important source of guidance cues for extending axons (Snow et al., 1990; Augsburger et al., 1999), each of these proteins may play important roles in regulating the pathfinding of axons that grow near, or across, the dorsal midline of the spinal cord. In fact, the Slits (Brose et al., 1999; Zou et al., 2000), Neuropilin-2 (Chen et al., 1997), B-class ephrins (Flanagan and Vanderhaeghen, 1998) and BMPs (Augsburger et al., 1999) have previously been shown to mediate repulsive axon guidance events in several neural systems. Thus, despite the fact that functions for Annexin IV and VEMA have yet to be established, it seems possible that their expression in both ventral and dorsal midline structures may be indicative of an important role in axon guidance.
Interneuron Heterogeneity
The results of anatomical, histological, and gene expression studies emphasize the functional and structural complexity of the vertebrate spinal cord, which contains many distinct populations of both ipsilaterally- and contralaterally-projecting interneurons (Silos-Santiago and Snider, 1992; Silos-Santiago and Snider, 1994). With specific regard to commissural interneuron heterogeneity, recent studies have established that the basic-helix-loop proteins, Math1 (Helms and Johnson, 1998; Helms et al., 2000) and Ngn1 (JE Johnson, unpublished observations), and the paired-box domain-containing genes Pax-3 and Pax-7 (Mansouri and Gruss, 1998), are expressed in overlapping populations of dorsal commissural neuron progenitors, while the homeobox-containing gene, Dbx1 (Matise et al., 1999), is expressed in both ventral and dorsal progenitor populations. Importantly, regulatory elements derived from each of these genes were shown to be capable of driving reporter gene expression in neuronal populations that ultimately extend axons across the floor plate in transgenic mice. Furthermore, several classes of postmitotic commissural neurons that are likely to arise from MATH1 and Ngn1-positive progenitor populations may be distinguished by the expression of the LIM homeodomain genes LH2A and LH2B (see Lee and Jessell, 1999) and references therein) and possibly Lim1 (see Kania et al., 2000). While it has recently been shown that the roof plate controls the specification of dorsal commissural neuron progenitor populations (Lee et al., 2000; Millonig et al., 2000), we currently know little about the mechanisms which regulate their migratory routes or their differentiation into postmitotic neurons. The recent localization of several commissural axon-associated guidance receptors in the developing rodent spinal cord also reveals a previously unreported molecular heterogeneity among commissural neurons. Specifically, the overlapping expression pattern of DCC and TAG-1 protein define dorsomedial populations of commissural neurons, while the distribution of Robo-2, Neuropilin-2, EphB1 and L1 mRNA appear to mark multiple, partially overlapping subsets of commissural neurons in more lateral regions of the spinal cord (see Imondi et al., 2000 and references therein). In addition, the 65-kD isoform of glutamic acid decarboxylase (GAD65) has recently been shown to be a specific marker of an early-developing ventrally-located population of commissural axons in the developing rat spinal cord (Phelps et al., 1999; Tran and Phelps, 2000).
An important goal for future studies is to correlate the trajectories of specific populations of ipsilaterally- and contralaterally-projecting interneurons with the expression patterns (mRNA and protein) of specific regulatory genes, guidance receptors, or neurotransmitters in wild-type mice. These data will firmly establish the presence of multiple, distinct populations of ipsilaterally-projecting and commissural interneurons within the spinal cord, and may identify good molecular candidates for the control of specific patterns of connectivity in the developing CNS. Furthermore, these comparative studies may motivate the generation of a library of transgenic mice in which the directed expression of tau/lacZ (Callahan et al., 1998) or green fluorescent protein GFP (Feng et al., 2000) to specific interneuron populations can be used to characterize the pathfinding of distinct subsets of ipsilaterally-projecting and commissural axons in both wild-type and mutant backgrounds. In addition, the availability of these reagents should make it possible to not only clarify the migratory routes of early interneuron precursors, but also to identify synaptic targets for distinct classes of interneurons.
Axon Segment-Specific Protein Expression
In both the developing vertebrate spinal cord and the Drosophila ventral nerve cord, the expression of several distinct cell surface receptors is spatially-segregated along axons that navigate in the vicinity of the midline (see Fig. 3). The tight spatial control of receptor expression provides a powerful mechanism by which pathfinding growth cones may respond, at restricted points along their trajectory, to symmetrically distributed guidance cues. While the results of recent studies suggest that axon contact with specialized midline cells is likely to regulate axon segment-specific expression of various guidance receptors (Matise et al., 1999), it is not at all clear how the underlying molecular mechanisms will ultimately be elucidated. Unfortunately, the unsuccessful initial attempts aimed at expressing Robo along the entire length of commissural axons in the Drosophila ventral nerve cord strongly suggest that this may prove to be a difficult task (Kidd et al., 1998a). Further complicating matters, it appears that (at least in vertebrates) the spatial regulation of axonal receptor expression has not been extensively characterized, and may even be species-specific. For example, the “TAG-1-L1 switch” documented in the rodent spinal cord (see Dodd et al., 1988) implies that TAG-1 and L1 are present on distinct, nonoverlapping segments of the same axon. However, this has not, in fact, been directly demonstrated. In addition, the TAG-1 to L1 switch does not appear to occur in chick embryos and TAG-1 continues to be expressed on crossed segments of commissural axons in the developing human spinal cord (Karagogeos et al., 1997). Thus, important future goals include the careful description of receptor distribution along the lengths of axons that navigate in the vicinity of the midline, and the elucidation of the molecular mechanisms that establish axon segment-specific expression patterns.
Protein Transfer to Commissural Axons
A number of studies have reported the surprising finding that proteins which are produced by midline cells subsequently appear on commissural axon/growth cones surfaces. For example, Comm (Tear et al., 1996), Slit (Rothberg et al., 1990), and Netrins (Harris et al., 1996) have been detected on decussated commissural axon segments in the Drosophila ventral nerve cord. In addition, expression of the Tg4 transgene, which is normally restricted to floor plate cells in the mouse spinal cord, begins to appear on commissural axons at about the time they cross the ventral midline (Campbell and Peterson, 1993). These observations provide support for an intimate link between midline cells and commissural axons. The functional importance of such a link is suggested by the exquisite manner in which Comm is thought to regulate Robo expression. In the proposed model (see above), transfer of Comm to commissural axons facilitates a down-regulation of Robo, which in turn promotes midline crossing. Given that the reduction of Robo must occur coincident with the very first contacts that commissural growth cones/axons make with midline cells, the proposed transfer of Comm and the subsequent cointernalization of a Comm and Robo complex may, in fact, represent the most efficient way to achieve the observed rapid modulation of receptor levels; a Comm-independent mechanism would presumably be inconsistent with the time frame of midline crossing, since it would most likely involve the relay of a signal from the commissural neuron cell body to the axon.
While the presence of floor plate-derived proteins on the surfaces of commissural axons is thought to reflect an active process of macromolecular transfer, the underlying mechanism(s) remains obscure (Tear et al., 1996). Thus, an important future goal will be to develop an accessible system(s) within which it may be possible to elucidate the relevant molecular interactions. One possibility would be to reconstitute the transfer of Comm (for example) in an in vitro system, by using epitope-tagged forms of the protein to monitor its movement (see Tear et al., 1996) between apposed cell surfaces. Whatever approach is taken, uncovering the mechanisms which regulate protein transfer to commissural axons will provide fascinating new insights into the control of midline guidance.
Commissural Axon Pathfinding on the Contralateral Side of the Floor Plate
In the developing mammalian spinal cord, commissural axons turn exclusively in the anterior direction after crossing the floor plate. Interestingly, neither the cellular nor the molecular mechanisms underlying this directed pathfinding decision have been elucidated. However, recent studies suggest that local, contact-dependent interactions may account for the observed polarity of these decussated projections. For example, of the small number of commissural axons which extend longitudinally in the region formerly occupied by the floor plate in Gli2 homozygous mice, a significant fraction inappropriately project in the caudal direction. This recent finding supports a role for the floor pate and/or VIR cells in regulating the directionality of commissural axon growth on the contralateral side of the ventral midline (see Matise et al., 1999). Consistent with this finding, the results of recent in vitro studies suggest that local directional cues guide the rostrally-directed projections displayed by dopaminergic axons in the rat midbrain (Nakamura et al., 2000). Rostrocaudal polarity errors have also recently been described in the spinal cords of NPN-2-deficient mice. However, these errors were apparently corrected over a one day period and they may simply be a secondary consequence of axons stalling within the floor plate (see Zou et al., 2000). Clearly, an important area of future study will be the elucidation of the molecular mechanisms which control the directionality of commissural axon growth on the contralateral side of the ventral midline. In this regard, it is interesting to note that the vab-8 gene controls posteriorly-directed axon outgrowth in C. elegans and encodes for multiple intracellular kinesin-like proteins (Wightman et al., 1996; Wolf et al., 1998b). Thus, it is tempting to speculate that mammalian homologs of vab-8 or a closely related gene may regulate the polarity of decussated commissural projections in the spinal cord.
In most models of vertebrate midline guidance, decussated segments of commissural axons travel alongside the floor plate for significant distances (see Colamarino and Tessier-Lavigne, 1995b and references therein). However, it was previously observed that commissural axons grow in close proximity to the floor plate for distances of only about 100 μm after crossing the ventral midline in the developing rat spinal cord (Bovolenta and Dodd, 1990). Importantly, subsequent stages of the contralateral commissural projection were not examined in this study. In the zebrafish spinal cord, commissural axons cross the ventral midline and then follow a contralateral diagonal trajectory that is directed anteriorly and dorsally. Upon reaching the dorsal spinal cord, these axons turn to extend along a longitudinal fiber tract (see Bernhardt et al., 1990; Bernhardt, 1994). We have recently used DiI labeling to re-examine the pathfinding of commissural axons on the contralateral side of the midline in the developing mouse and chick spinal cord. Surprisingly, we find that most decussated commissural axons in the mouse and chick extend along a complex trajectory that closely resembles that observed in the zebrafish spinal cord (R Imondi and Z Kaprielian, unpublished observations). The results of these studies should not only provide a new framework within which to consider the molecular mechanisms that regulate commissural axon guidance on the contralateral side of the floor plate, but also may call into question previously described midline guidance defects characterized as lateral drifting (Burstyn-Cohen et al., 1999) or wandering (Zou et al., 2000) of descussated commissural axons.
Acknowledgements
We thank the following investigators for kindly supplying us with their previously published data figures: MP Matise (UMDNJ; Fig. 4), M Tessier-Lavigne (UCSF; Fig. 5), Fujio Murakami (Osaka University; Fig. 6), K Zinn and A Schmid (Caltech; Fig. 8) and JB Thomas (Salk Institute; Fig. 9), as well as JE Johnson and C-B Chien for allowing us to cite their unpublished results. Work from our lab was supported by an NIH FIRST award (R29 NS34847), an R01 (NS 38505) and a pre-doctoral training grant (NS 07098) from the NINDS, as well as by grants from the Christopher Reeve Paralysis Association, the Spinal Cord Research Foundation, and the Whitehall Foundation.
