The current model of axon guidance postulates a collaboration of attractive and repulsive guidance cues that can act over some distance as long-range guidance cues, or locally as short-range guidance cues [1, 2]. One of the preferred systems for axon guidance studies has been commissural axons that cross the floor plate, the ventral midline of the spinal cord [3, 4]. Midline crossing is a conserved feature of axonal navigation between invertebrates and vertebrates [4–6]. In vertebrates, axons are guided toward the ventral midline by long-range guidance cues. These include roof plate-derived BMPs and Draxin which repel commissural axons from the dorsal midline [7–9] and the chemoattractants Netrin-1 and Shh which are released from the floor plate . In both, the ventral nerve cord of invertebrates and the vertebrate spinal cord, midline crossing is controlled by a balance between positive and negative signals derived from the interaction between growth cone receptors and ligands expressed by midline cells [3, 11]. Negative regulators of midline crossing were first identified based on genetic screens in Drosophila[12, 13]. Characterization of the genes responsible for this repulsive activity identified robo receptors  (roundabout receptors) and their ligand Slit  but also the transmembrane protein comm (commissureless), which was found to regulate Robo expression [16–19]. In vertebrates, positive regulators of midline crossing were first identified . Both in vivo and in vitro interactions of Axonin-1/TAG-1/Contactin-2 and NrCAM were shown to mask a repellent activity of the floor plate [21, 22]. The repellent activity was later attributed to Semaphorin 3B and 3F, mediated by Neuropilin-2 , and to orthologs of Drosophila Slit, mediated by Robo receptors [15, 23–29]. Vertebrates express three Slits[25, 30–33] and four Robos: Robo1, Robo2, and Robo3/Rig1 are expressed in the developing nervous system [34, 35]. Robo4 (Magic Roundabout) differs markedly in its domain structure from the other Robos and is expressed exclusively in endothelial cells [36, 37]. A role for Robo4 in angiogenesis has been described in mice  and zebrafish .
In the developing nervous system, Robos were mainly described as receptors for Slits which mediate a repellent signal. For midline crossing, commissural neurons face the problem of regulating Robo expression temporally in such a way that Robo is not expressed on the axonal surface before they have reached and entered the floor plate. However, upon floor-plate contact Robo has to be expressed on commissural growth cones in order to expel them from the floor plate that was previously perceived as an attractive environment.
The model of Robo regulation put forth in invertebrates postulates that midline crossing is controlled by Comm, which prevents surface expression of Robo before midline contact [16, 17, 40–44]. According to the sorting model, comm is specifically and transiently expressed in contralaterally but not ipsilaterally projecting neurons. In the presence of Comm, Robo is not inserted into the plasma membrane but rather transported to the endosomal-lysosomal compartment directly, thus allowing axons to cross the midline [18, 19].
Interestingly, an ortholog of comm is not found in vertebrate genomes [41, 43], and therefore, it has been unclear how Robo levels are controlled in vertebrate commissural axons. A role for Robo3/Rig-1 in regulating the function of Robo1 as receptor for midline Slits has been suggested, but the proposed mechanism does not include the regulation of Robo1 levels on precommissural axons . Instead, alternative splicing was recently reported to produce different Robo3 isoforms with antagonistic function with respect to midline crossing . Robo3.1 was shown to be expressed on axons before, whereas Robo3.2 is expressed after midline crossing. Based on loss- and gain-of-function experiments, the authors suggested that Robo3.1 silences the effect of Robo1 and Robo2, while Robo3.2 enhances their effect and perhaps additionally counteracts Robo3.1 function. Still, it remains unclear how Robo1 protein levels are kept low on pre-crossing compared to post-crossing axons, a finding that was confirmed in several studies.
Here, we show that levels of Robo1 on commissural axons are regulated by RabGDI (Rab Guanine Nucleotide Dissociation Inhibitor, GDI1). RabGDI is a component of the vesicle fusion machinery [46, 47]. It is required for the recycling of hydrolyzed RabGDP to RabGTP. RabGDI retrieves RabGDP from the plasma membrane and shuttles it to new donor vesicles, where RabGDP is activated by a guanine nucleotide exchange factor (GEF). The GEF exchanges the GDP for a GTP, thus recycling the active RabGTP required for a subsequent round of vesicle fusion.
In humans, loss of RabGDI function results in mental retardation . In mice, loss of RabGDI function has been associated with defects in associative memory . These abnormalities are linked to changes in Rab-mediated vesicle trafficking. Here, we provide in vivo and in vitro evidence that loss of RabGDI function during midline crossing prevents the fusion of a subset of vesicles required for the insertion of Robo1 into the growth cone membrane. Thus, in both invertebrates and vertebrates, Robo levels on precommissural axons are regulated post-translationally to allow midline crossing. However, the mechanisms and the molecules involved in the regulation of Robo1 surface levels differ: In flies commissural axons can cross the midline, because the transient expression of Comm prevents Robo1 surface expression by directing it to the lysosomal pathway. In chicken, RabGDI is required for membrane-insertion of Robo1. In the absence of RabGDI, Robo1 is not inserted into the growth cone surface.