Dbx1 triggers crucial molecular programs required for midline crossing by midbrain commissural axons

In the developing nervous system, navigating axons have to ‘decide’ whether or not to cross the midline in order to project to their specific cellular targets. The establishment of axonal projection laterality is a fundamental step for the construction of the proper neural circuits. In general, axon guidance depends upon the ability of individual axons to recognize specific guidance cues (Tessier-Lavigne and Goodman, 1996). Accumulating evidence now suggests that the intrinsic genetic programs for sensing these guidance cues are in most cases encoded by unique transcription factors that ultimately induce the expression of specific guidance receptors and receptor-associated co-factors (Shirasaki and Pfaff, 2002; Polleux et al., 2007). However, knowledge of the intrinsic genetic programs that specify the laterality of axonal projections is still limited. Importantly, for example, it remains obscure whether the expression of a single transcription factor is sufficient to trigger downstream transcriptional cascades required for midline crossing, or whether the parallel actions of several transcription factors are required to determine axonal projection laterality.

In the developing retina, the zinc finger transcription factor Zic2 is a determinant for the ipsilateral guidance programs of retinal ganglion cells (RGCs) at the ventrotemporal (VT) area (Herrera et al., 2003). Indeed, the expression of guidance receptor EphB1 is triggered by Zic2 in uncrossed RGCs, which therefore contributes to these axons’ avoidance of the midline by the repulsive action of ephrin B2 expressed at around the optic chiasm (García-Frigola et al., 2008; Lee et al., 2008). It has also been reported through the analysis of Isl2 knockout mice that the LIM homeodomain (LIM-HD) transcription factor Isl2 is involved in the assignment of the midline-crossing guidance program to crossed RGCs (Pak et al., 2004). It should be noted, however, that Isl2-mediated Zic2 repression seems to be restricted to the VT area (Pak et al., 2004). Thus, it is uncertain whether the expression of Isl2 alone is sufficient to confer axon guidance programs for midline crossing.

In the developing dorsal spinal cord, the basic helix-loop-helix (bHLH) transcription factor Atoh1 is crucially involved in the generation of dI1-type commissural neurons (Helms and Johnson, 1998; Bermingham et al., 2001; Gowan et al., 2001; Wilson et al., 2008). This suggested a possibility that Atoh1 acts as a dedicated factor that specifically triggers commissural neuron differentiation. However, it seems unlikely, as recent results have indicated that Atoh1 is also required for the generation of ipsilateral neurons (Wilson et al., 2008). Within the Atoh1 transcriptional cascade, both commissural and ipsilateral neurons postmitotically express two closely related LIM-HD transcription factors, Lhx2 and Lhx9 (Wilson et al., 2008). Interestingly, the analysis of the Lhx2 and Lhx9 double-knockout mice has indicated that these factors regulate the expression of Robo3 (Wilson et al., 2008), which is an essential regulator for midline crossing (Marillat et al., 2004; Sabatier et al., 2004; Tamada et al., 2008) by selectively silencing the repellent action of Slit-Robo1 signaling (Sabatier et al., 2004). Although this gene-targeting study suggests that these LIM-HD factors are required for midline crossing by commissural axons, it is currently obscure whether the expression of these factors is sufficient to specify axonal projection laterality. In addition, the mammalian Bar-class transcription factor Barhl2, which is expressed in both commissural and ipsilateral neurons of Atoh1-derived lineage, has recently been shown to be involved preferentially in the assignment of ipsilateral neuron identity (Ding et al., 2012). Therefore, it remains unclear whether there is a certain transcription factor assigned genetically, in a dedicated manner, to the acquisition of axon guidance programs required for midline crossing.

In this study, we have addressed the issue of whether a single transcription factor has an ability to trigger downstream transcriptional cascades required for midline crossing in vivo. For this, we employed gain- and loss-of-function strategies using an in vivo electroporation technique developed for mouse embryos (Tabata and Nakajima, 2001; Saito, 2006; Shimogori and Ogawa, 2008). We examined development of commissural neurons in the midbrain for the current purpose because of the following advantages. First, the size of the midbrain neural tube is the largest compared with that of the other axial levels, such as the spinal cord. This enabled us to perform precise and reproducible in vivo electroporation at early stages of the mouse embryo. Second, because the axonal trajectory of commissural and ipsilateral neurons in the midbrain is anatomically and evolutionarily well documented as tectospinal and tectobulbar tracts, respectively (Murray and Coulter, 1982; Kröger and Schwarz, 1990; Shepherd and Taylor, 1995; Mastick and Easter, 1996; Shirasaki et al., 1996), the contrasting phenotype (i.e. to cross or not to cross) after gene manipulation is easily distinguishable. Here, we show that a single progenitor homeodomain factor Dbx1, expressed by a subset of progenitor cells in the dorsal midbrain, has an essential role in assigning midline-crossing identity by activating the downstream molecular programs that ultimately control the expression of Robo3 in commissural neurons.

In the developing midbrain, both commissural (i.e. tectospinal) and ipsilateral (i.e. tectobulbar) neurons are generated from the same dorsal area (Murray and Coulter, 1982; Kröger and Schwarz, 1990; Shepherd and Taylor, 1995; Mastick and Easter, 1996). It has previously been shown in mice that the long projection neurons in the dorsal midbrain are generated at around embryonic day 10 (E10) (Mastick and Easter, 1996). We therefore first examined the development of midbrain commissural axons in mice by introducing green fluorescent protein (GFP)-expression vector using in vivo electroporation at E10.75. Three days after the electroporation, embryos were taken out and flat-mounted to analyze the entire trajectories of GFP-labeled axons (Fig. 1A). As expected, we found that commissural as well as ipsilateral axons were labeled by GFP (Fig. 1B,B′). After commissural axons had crossed the floor plate, these axons made a sharp turn to project caudally in the region between the floor plate and the oculomotor neurons, as revealed by GFP-labeled axonal trajectory together with immunostaining for expression of ALCAM, the molecular marker for motoneurons and floor plate cells (Weiner et al., 2004) (supplementary material Fig. S1A-C′). This observation is consistent with the previously reported DiI-labeling of tectofugal tracts in the mouse and chick midbrain (Shepherd and Taylor, 1995; Mastick and Easter, 1996). By contrast, when we performed electroporation at E11.5, the majority of GFP-labeled axons were ipsilateral, with few commissural axons labeled (Fig. 1C,C′,D). Here we took advantage of this experimental condition (i.e. the developmental stage used for in vivo electroporation), in order to examine the genetic program that directs the differentiation of midbrain commissural neurons in the mouse.

We have previously shown that commissural neurons in the dorsal neural tube, from the spinal cord rostrally to the midbrain where floor plate cells are found, share axon guidance mechanisms at least in the context of Netrin-mediated chemoattraction signaling (Shirasaki et al., 1995; Shirasaki et al., 1996). Here, we further characterized the expression of cell surface molecules that may be involved in the guidance of midbrain commissural axons. We first examined Robo3 expression using whole-mount immunohistochemistry on flat-mounted midbrain preparations, as Robo3 has been shown to be a crucial regulator for midline crossing by commissural axons in the spinal cord and the hindbrain (Marillat et al., 2004; Sabatier et al., 2004; Tamada et al., 2008). We found that Robo3 was selectively expressed on axons growing toward the floor plate, as judged by expression analysis of the pan-neuronal marker Tuj1 and Robo3 (Fig. 2A-C′). In addition, Robo1 was found to be expressed on the pre-crossing segment of these Robo3-expressing commissural axons (Fig. 2D-F′), as well as on the post-crossing portion of these axons at high levels (supplementary material Fig. S1D-F′). Furthermore, Robo1 was also expressed on Robo3-negative ipsilateral axons (Fig. 2D-F′; supplementary material Fig. S2).

We next analyzed the expression of transcription factors in the dorsal midbrain that have previously been shown to be expressed by commissural neurons in the dorsal spinal cord (Liem et al., 1997; Helms and Johnson, 1998; Gowan et al., 2001; Wilson et al., 2008). We found in flat-mounted preparations that Lhx2 and Lhx9 were expressed in the dorsal midbrain, where cell bodies of commissural neurons reside (supplementary material Fig. S3A-B′). Moreover, the POU transcription factor Brn3a (Pou4f1 - Mouse Genome Informatics), a definitive marker of midbrain ipsilateral neurons (Fedtsova and Turner, 1995; Fedtsova et al., 2008), was also expressed in the dorsal midbrain (supplementary material Fig. S3C,C′). To characterize the expression of these molecules in more detail, we employed immunohistochemistry on transverse cryostat sections of the dorsal midbrain. We found that, among Lhx2-positive postmitotic cells (supplementary material Fig. S3D-I), there were two subclasses categorized by the presence or absence of Robo3 (Fig. 3A-C′). This is reminiscent of the observation in the spinal cord in which Lhx2 is expressed not only in dI1 commissural neurons but also in dI1 ipsilateral neurons (Wilson et al., 2008). We therefore asked whether the Lhx2-positive population in the dorsal midbrain encompasses ipsilateral neurons by focusing on the expression of Brn3a. We found that cells that were positive for Lhx2 also expressed Brn3a (Fig. 3D-F′). Importantly, Brn3a-positive cells never expressed Robo3 (Fig. 3G-I; supplementary material Fig. S4A-C). These results therefore indicate that, in the dorsal midbrain, Lhx2 is expressed not just in Robo3-positive commissural neurons but also in ipsilateral neurons. We also found that Lhx9 was expressed in the dorsal midbrain, and this Lhx9-positive population similarly included Brn3a-positive and Brn3a-negative cells (supplementary material Fig. S3J-L).

Thus, in subsequent analyses, we have used these expression profiles to distinguish between commissural and ipsilateral neurons in the dorsal midbrain.

To identify a crucial genetic determinant that triggers downstream molecular programs required for midline crossing by commissural neurons, we searched for transcription factors whose expression has been reported in the dorsal midbrain. Here, we focused on the homeodomain transcription factor Dbx1, because it has been shown that Dbx1 mRNA is expressed by progenitor cells in the dorsal midbrain during the time when midbrain commissural neurons are generated (Mastick et al., 1997; Prakash et al., 2009). We first examined the overall expression pattern of Dbx1 in the midbrain using whole-mount immunohistochemistry on flat-mounted midbrain preparations at E11.5. As reported previously, we found that Dbx1 expression was restricted to the dorsal region of the midbrain (Fig. 4A). To characterize the Dbx1 expression in more detail, we next performed immunohistochemistry using transverse sections of the midbrain. We found that Dbx1 was expressed by progenitor cells, as judged by the expression of proliferation marker Ki67 (Mki67 - Mouse Genome Informatics) and postmitotic neuronal marker Tuj1 (Fig. 4B,D-I). Consistent with this, Dbx1-positive cells did not express postmitotically expressed molecules such as Lhx2 (Fig. 4J-L; supplementary material Fig. S4D-F). Moreover, among the progenitor cells, Dbx1 was expressed in subsets of cells positive for the proneural bHLH factor Ngn1 (Neurog1 - Mouse Genome Informatics) (Fig. 4B,C,M-O), but not in Ascl1-positive cells (Fig. 4P-R; supplementary material Fig. S4G-I). A similar Dbx1 expression pattern was also observed at E10.75 (supplementary material Fig. S5). Thus, these results raise the possibility that Dbx1 is involved in the generation of unique subsets of cells in the developing dorsal midbrain.

Next, to examine which neuronal class employs Dbx1-mediated transcriptional program, we performed a lineage-tracing analysis using the Dbx1-enhancer element. It has been shown that a distal 3.5 kb of the 5.7 kb Dbx1 regulatory sequence acts as an enhancer element that controls unique expression of Dbx1 in the forebrain, midbrain, hindbrain and spinal cord (Lu et al., 1996). We therefore used this unique enhancer to generate an expression vector in which ZsGreen expression is under the control of the Dbx1 enhancer. We then electroporated Dbx1 enhancer-driven ZsGreen1 (Dbx1 enhancer::ZsGreen) vector into the dorsal midbrain at E10.75, because at this stage both commissural and ipsilateral neurons can be transfected by our in vivo electroporation technique in mice (Fig. 1B,B′). We found in the electroporated embryos that axons growing toward the floor plate were selectively labeled with ZsGreen (Fig. 5A,A′). In addition, these ZsGreen-expressing commissural axons also expressed Robo3 (Fig. 5A-C′), indicating that Dbx1-positive progenitors in the dorsal midbrain give rise to Robo3-positive commissural neurons. Furthermore, expression analyses using transverse sections of the electroporated midbrain revealed that ZsGreen-positive postmitotic cells expressed Robo3, but did not express ipsilateral neuron marker Brn3a (Fig. 5D-L; supplementary material Fig. S4J-L). Together, these results therefore suggest that commissural neurons in the dorsal midbrain are generated selectively from Dbx1-positive progenitors.

To test whether the downstream transcriptional program initiated by Dbx1 ultimately regulates midline crossing by midbrain commissural axons, we first performed Dbx1 gain-of-function experiments by ectopically expressing Dbx1 in progenitors of ipsilateral neurons. For this, we first took advantage of the fact that cells mainly transfected by in vivo electroporation at E11.5 are ipsilateral neurons (Fig. 1C,C′). We found that Dbx1 misexpression in the dorsal midbrain at this stage dramatically increased midline-crossing axons compared with controls (Fig. 6A-B′,H; supplementary material Figs S6 and S7). We also found that Dbx1 electroporation at E10.75, the stage when both commissural and ipsilateral neurons are abundantly labeled by the electroporation (Fig. 1B,B′), resulted in more effective induction of midline-crossing axons at the expense of ipsilateral axons (Fig. 6C-D′,I).

Next, we carried out Dbx1 loss-of-function experiments using in vivo electroporation at E10.75. For this, we first misexpressed a dominant-negative form of Dbx1 (dnDbx1) in the dorsal midbrain. We found that expression of dnDbx1 caused loss of midline crossing without affecting caudally directed axon growth (Fig. 6E,E′,I). We next performed Dbx1 knockdown by introduction of siRNA specifically targeting Dbx1, and found that this similarly caused an absence of midline crossing (Fig. 6F,F′,J; supplementary material Fig. S8). In addition, the loss of midline crossing observed in the Dbx1-siRNA electroporated embryos was rescued by co-introduction of the siRNA-resistant variant of Dbx1 (Fig. 6G,G′,J).

Together, these results therefore suggest that the expression of Dbx1 is necessary and sufficient to trigger downstream molecular programs required for midline crossing by midbrain commissural axons.

To examine the molecular programs triggered by Dbx1, we next focused our attention on the expression of transcription factors that may be under the control of Dbx1. It has been shown in the chick ventral spinal cord that Dbx1 misexpression promotes a switch in transcription factor profile from V1 ipsilateral to V0 commissural neurons by suppressing the expression of transcription factors unique to ipsilateral neurons (Pierani et al., 2001). This raises the possibility that an ectopic generation of commissural neurons in the Dbx1-misexpressed dorsal midbrain (Fig. 6) is induced at the expense of ipsilateral neuron generation. To test this, we examined the expression of Brn3a after Dbx1 electroporation in the midbrain. We found that the expression of Brn3a was significantly repressed in the Dbx1-transfected cells compared with GFP-transfected controls (Fig. 7A-F,H). Consistent with this, when analyzed using whole-mount immunohistochemistry on flat-mounted preparations, Brn3a expression was dramatically suppressed in the Dbx1-electroporated area (Fig. 7G,G′). These results therefore suggest that the Dbx1-triggered transcriptional program may include suppression of ipsilateral neuron genetic programs to consolidate commissural neuron identity.

To further unravel the Dbx1-triggered transcriptional cascade, we next searched for a transcription factor that is selectively expressed in midbrain commissural neurons. In the dorsal midbrain, Lhx2 is expressed not only by commissural neurons but also by ipsilateral neurons (Fig. 3), suggesting that the expression of Lhx2 is not regulated within the Dbx1 cascade. Interestingly, it should be noted that, although V0 commissural neurons in the ventral spinal cord never express Lhx2 (Moran-Rivard et al., 2001), Dbx1 specifically activates the expression of homeodomain transcription factors Evx1/2 in V0 neurons (Pierani et al., 2001). Indeed, among the Evx gene family, it has been shown in the dorsal midbrain that Evx2 is expressed by cells with unknown characters (Dollé et al., 1994; Kmita et al., 2002). This prompted us to examine whether the expression of Evx2 is restricted to commissural neurons in the dorsal midbrain. We found that Evx2 was expressed by postmitotic cells in the dorsal midbrain (Fig. 8A-C; supplementary material Fig. S9), the expression of which did not overlap with that of Dbx1 expressed by progenitor cells (Fig. 8D-F; supplementary material Fig. S10A-C). In addition, the Evx2-positive population was a subset of Lhx2-positive cells (Fig. 8G-I). Strikingly, these Evx2-positive cells corresponded to Robo3-expressing cells (Fig. 8J-L; supplementary material Fig. S10D-F), and the expression of Evx2 never overlapped with that of the ipsilateral neuron marker Brn3a (Fig. 8M-O; supplementary material Fig. S10G-I). Together, these results indicate that Evx2 selectively marks Robo3-positive commissural neurons in the dorsal midbrain, thus raising the possibility that Evx2 is one of the transcription factors acting downstream of Dbx1.

Next, we directly addressed whether the expression of Robo3 is a crucial outcome regulated within the Dbx1-triggered transcriptional cascade. For this, we employed a misexpression (gain-of-function) approach using in vivo electroporation. As expected, we found that the misexpression of Dbx1 in the dorsal midbrain induced an ectopic expression of Evx2 compared with the non-electroporated control side of the midbrain (Fig. 9A,A′; supplementary material Fig. S11). Furthermore, Evx2 misexpression elicited an ectopic expression of Robo3 in the electroporated region (Fig. 9B,B′; supplementary material Fig. S12). We then asked whether the expression of Robo3 is ultimately regulated by the Dbx1-triggered molecular program. We found that Robo3 expression was dramatically induced on the Dbx1-electroporated side compared with the control side (Fig. 9C,C′), on which endogenous expression of Robo3 was downregulated in the dorsal midbrain at this stage (Fig. 9B′,C′). Moreover, commissural axons generated by the Dbx1 electroporation strongly expressed Robo3 when they grew toward the floor plate (Fig. 9C,C′, arrowheads). Together, these results therefore suggest that the transcriptional program triggered by Dbx1 at the progenitor stage eventually regulates the expression of Robo3 on midbrain commissural axons.

Finally, because the role of Robo3 expressed by midbrain commissural neurons has been unexplored, we examined whether Robo3 is required for midline crossing by midbrain commissural axons. To this end, we carried out Robo3 knockdown using in vivo electroporation of Robo3-specific siRNA (Chen et al., 2008). As expected, we found that midline crossing was strongly inhibited by the Robo3 loss of function (Fig. 9D-E′,G). The specificity of the siRNA effect was ascertained by a rescue experiment in which the midline-crossing defect was reversed by co-introduction of the siRNA-resistant variant of Robo3 (Fig. 9D,D′,F,F′,G). Thus, these results indicate that Robo3 is essential for midline crossing by midbrain commissural axons (supplementary material Fig. S13).

Taken together, these gain- and loss-of-function analyses therefore demonstrate that a single progenitor homeodomain factor Dbx1 has a crucial role in assigning midline-crossing identity by activating the downstream molecular programs that ultimately control the expression of Robo3 in midbrain commissural neurons.

In the developing nervous system, unique expression of transcription factors ultimately determines the axonal projection patterns of individual neurons by directing the expression of specific guidance programs (Shirasaki and Pfaff, 2002; Polleux et al., 2007). Although commissural neurons in the vertebrate and invertebrate central nervous system have provided us with a wealth of information on the molecular mechanism of axon guidance (Dickson and Gilestro, 2006; Evans and Bashaw, 2010; Nawabi and Castellani, 2011), knowledge of the intrinsic genetic programs employed is still limited. For instance, it has remained obscure whether the action of a single transcription factor can activate downstream molecular programs required for midline crossing, or whether the parallel actions of several transcription factors are needed for this task. There has also been a crucial issue as to the timing of assignment of midline-crossing identity (e.g. at the progenitor stage or at the postmitotic stage). In this study, we show that a single transcription factor, Dbx1, expressed by progenitor cells in the dorsal midbrain acts as a crucial genetic determinant for midline-crossing identity. Interestingly, it should be noted that, in other classes of commissural neurons (e.g. spinal commissural neurons, contralateral RGCs, and cortical callosal neurons), the identity of ‘cross or not to cross’ seems to be assigned at the postmitotic stage (Herrera et al., 2003; Pak et al., 2004; Alcamo et al., 2008; Britanova et al., 2008; Wilson et al., 2008; Ding et al., 2012). Thus, our findings reveal the existence of a novel regulatory layer within the transcriptional cascade that contributes to the establishment of the wiring laterality of the developing nervous system.

Our results show that molecular programs triggered by the progenitor transcription factor Dbx1 are necessary and sufficient to confer midline-crossing identity in the developing midbrain. Strikingly, Dbx1 misexpression resulted in an ectopic generation of Robo3-positive commissural neurons at the expense of ipsilateral neuron differentiation, as represented by suppression of the expression of Brn3a, an ipsilateral neuron marker in the midbrain. This suggests that, by activating several molecular programs in parallel, Dbx1 has a crucial role not only in promoting the specification of commissural neuron identity but also in suppressing the expression of ipsilateral neuron character, which may contribute to consolidating the identity of commissural neurons in the dorsal midbrain. In this context, it is noteworthy that many of the transcription factors possessing a role for fate specification have repressive function toward cells with contrasting characteristics (Arber et al., 1999; Thaler et al., 1999; Moran-Rivard et al., 2001; Pierani et al., 2001). For example, among dI1-class dorsal interneurons in the spinal cord, Barhl2 has been shown to facilitate differentiation of ipsilateral neurons by selectively suppressing the features of commissural neurons (Ding et al., 2012). Likewise, during neocortical development, Satb2 acts as a genetic determinant of callosal projection neurons by repressing the expression of Ctip2 (Bcl11b - Mouse Genome Informatics), a crucial factor for specification of subcortical projection neuron identity (Alcamo et al., 2008; Britanova et al., 2008). Thus, at least in some cellular contexts, the repressive function of the dedicated transcription factors may represent one broadly applicable strategy for the binary fate decision between commissural and ipsilateral neurons in order to restrain the potential to express conflicting genetic programs during development.

In the developing ventral spinal cord, it has been shown that Dbx1 is specifically expressed by a subset of progenitor cells that give rise to commissural neurons (V0 interneurons) (Pierani et al., 1999; Pierani et al., 2001). In addition, in Dbx1 mutant mice, the progenitor cells fail to generate V0 commissural neurons (Pierani et al., 2001). These findings suggest that Dbx1 is required for the generation of a subset of ventral commissural neurons in the spinal cord. However, it remains to be determined whether the expression of Dbx1 alone is sufficient to activate downstream molecular programs that directly couple with the expression of guidance programs for midline crossing. In the present study, we show that Dbx1-activated molecular programs ultimately induce the expression of Robo3 on midbrain commissural axons, which is hierarchically regulated via the expression of Evx2 at the postmitotic stage. Furthermore, we also show that the midline crossing by midbrain commissural axons depends crucially on the expression of Robo3. Interestingly, similar molecular program is also the case with dI1 commissural neurons generated from Atoh1-expressing progenitors in the dorsal spinal cord, although the expression of Robo3 is controlled by Lhx2 and Lhx9 at the postmitotic stage (Wilson et al., 2008). Because all classes of commissural axons in the spinal cord are misrouted and fail to cross the floor plate in Robo3 mutant mice (Sabatier et al., 2004; Wilson et al., 2008), it is predicted that axon guidance programs employed by V0 commissural neurons should also include the expression of Robo3. Thus, although upstream transcriptional programs for commissural neuron differentiation seem different among commissural neuron classes, we speculate that one of the essential outcomes from these molecular programs is to activate the expression of Robo3 to silence Slit-Robo1 repulsive signaling when commissural axons approach and cross the floor plate.

Our lineage-tracing studies revealed that axons of midbrain commissural neurons generated from Dbx1-expressing progenitor cells grow caudally after crossing the floor plate, as opposed to axonal behavior of V0 commissural neurons generated from Dbx1-postive progenitor cells in the ventral spinal cord (Moran-Rivard et al., 2001; Pierani et al., 2001). Interestingly, in the ventral spinal cord of Dbx1 mutant mice, it has been shown that the molecular program for rostrally directed axon growth by V0 neurons is under the control of Dbx1-triggered transcriptional programs (Pierani et al., 2001). However, this is in a sharp contrast with transcriptional program employed by midbrain commissural neurons, since these midbrain axons still grow caudally under conditions where Dbx1 function is lost (Fig. 6). In addition, ipsilateral neurons in the dorsal midbrain that are independent of Dbx1-triggered genetic programs also extend axons caudally (Fig. 1). It is therefore likely that a unique transcriptional program acting independently of Dbx1 directs the expression of molecular programs required for caudally oriented axon growth. Whether this transcriptional program is triggered by the action of a single dedicated transcription factor or by the parallel actions of several transcription factors remains to be determined.

Timed pregnant Institute for Cancer Research (ICR) mice were obtained from Japan SLC (Hamamatsu, Japan). Noon of the day on which a vaginal plug was found was designated as embryonic day 0.5 (E0.5). All experiments were performed in accordance with the guidelines of the animal welfare committees of Osaka University and the Japan Neuroscience Society.

In vivo electroporation in mice was performed as described (Saba et al., 2003; Saito, 2006). In brief, after plasmid DNA was injected into the canal of the midbrain, electric pulses were applied to the dorsal midbrain using square pulse electroporator (CUY-21, Bex) with a pair of 1 mm diameter platinum electrode (CUY650P1, Nepa Gene). For cell-lineage-tracing analyses in transverse cryostat sections, we used a pair of 3 mm diameter electrode (CUY650P3, Nepa Gene) to electroporate the Dbx1-enhancer vector into a large area of the dorsal midbrain. Two to 3 days after electroporation, the embryos were taken out from the mother, fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) for 2 hours at 4°C and subjected to immunohistochemistry for further analyses.

Whole-mount immunohistochemistry on flat-mounted midbrain/hindbrain preparations was performed as described (Shirasaki et al., 2006). The flat-mounted preparations enable ready recognition of the entire trajectories of labeled axons. The procedure for flat-mounted midbrain/hindbrain preparation was described previously (Shirasaki et al., 1995). Immunostaining was also carried out on 12 μm-thick transverse cryostat sections of the midbrain as described (Shirasaki et al., 2006). Commercially available primary antibodies used are listed in supplementary material Table S1. Primary antibody against mouse Dbx1 was generated in rabbits using a keyhole limpet hemocyanin (KLH)-conjugated peptide containing the 15 amino acids of the mouse Dbx1 (CDEDEEGEEDEEITVS) (Vue et al., 2007). An Evx2 guinea pig polyclonal antibody was generated using a KLH-conjugated peptide that includes 20 amino acids of the mouse Evx2 (EIASATESRKKPSHYSEAAAC), the amino acid sequence of which is unique to Evx2 among the Evx-class homeodomain proteins. The Vector MOM Immunodetection Kit (Vector Laboratories) was used when primary antibodies from mouse were used. Images were taken by a fluorescence microscope (Olympus, BX61N) with a high-resolution digital cooled charge-coupled device (CCD) camera (Hamamatsu Photonics, ORCA-AG) and by a confocal laser scanning microscope (Olympus, Fluoview FV300).

To visualize trajectory of GFP-labeled axons from the dorsal midbrain, we first constructed a pCAGGS-AcGFP1 vector. Briefly, an open reading frame (ORF) of AcGFP1 was amplified from pIRES2-AcGFP1 vector (Clontech) by PCR and cloned into the EcoRI sites of pCAGGS vector (Niwa et al., 1991). For gain- and loss-of-function experiments, pCAGGS-IRES2-AcGFP1 vector was generated. For this, a fragment containing IRES2-AcGFP1 was isolated from pIRES2-AcGFP1 (Clontech) by digesting with XhoI and NotI sites and inserted into the pCAGGS vector. For gain-of-function experiments for Dbx1 and Evx2, ORF clones of mouse Dbx1 (GenBank accession no. BC082541) and human Evx2 (GenBank accession no. NM_001080458) were obtained from Open Biosystems and OriGene, respectively. After an HA-tag was fused to the N-terminus of Dbx1 and Evx2, these fragments were subcloned into a pCAGGS-IRES2-AcGFP1 vector. To generate dnDbx1 for loss-of-function experiments, we fused the activation domain of the herpesvirus protein VP16 resides (amino acids 446-490) to the C-terminus of the Dbx1, and the dnDbx1 fragment was subcloned into pCAGGS-IRES2-AcGFP1 vector. An ORF of human Robo3 (GenBank accession no. BC008623) was purchased from Open Biosystems in order to generate an siRNA-resistant variant (see below). We fused Myc-tag to the C-terminus of Robo3 and inserted into the pCAGGS vector. For cell lineage-tracing analysis of Dbx1-positive progenitors, we utilized a distal 3.5 kb of the 5.7 kb Dbx1 regulatory sequence as a Dbx1 enhancer element (Lu et al., 1996). The 3.5 kb Dbx1 enhancer was isolated from a Mus musculus molossinus (MSM) mouse bacterial artificial clone (BAC) clone that contains Dbx1 gene (clone name: MSMg01-341M02, RIKEN BioResource Center DNA Bank) (Abe et al., 2004). Then we generated Dbx1 enhancer::ZsGreen vector to drive the expression of fluorescent protein ZsGreen1 (Clontech) under the control of minimum β-globin promoter combined with the Dbx1 enhancer.

Stealth siRNAs for mouse Dbx1 were designed using BLOCK-iT RNAi Designer (Invitrogen). The sequences of the sense strand were: 5′-CCGGCCACUCUAGUUUCCUAGUAGA-3′, 5′-GGUAAACCGUCAGACUUCUCUGAUU-3′. Because both of these siRNAs yielded a similar axon guidance phenotype (Fig. 6F,F′; supplementary material Fig. S8), we routinely used the former for the loss-of-function and the rescue experiments. A Stealth siRNA for mouse Robo3 was also designed using BLOCK-iT RNAi Designer. The target sequence and the specificity of Robo3 siRNA were as described (Chen et al., 2008). The siRNA and plasmid DNA were co-electroporated as described above. In rescue experiments, to generate the siRNA resistant variants of Dbx1 and Robo3, several silent mutations were introduced into the siRNA target regions using PrimeSTAR Mutagenesis Basal Kit (TaKaRa). The resistant sequences of the sense strand were: Dbx1, 5′-GUGGACAUAGCUCCUUUUUGGUCGA-3′; Robo3, 5′-CUACUUGGUCAGGAUGGUUGAAAAU-3′.

Flat-mounted preparations were made from the electroporated embryos and immunostained with anti-GFP. Non-saturated fluorescent images were taken using a fluorescence microscope with a ×2 objective lens to capture overall trajectory of the labeled axons. Mean fluorescence intensity of cohort of commissural and ipsilateral axons at the isthmus (midbrain/hindbrain boundary) was measured using ImageJ software (Schneider et al., 2012). After mean fluorescence intensity of unstained (background) areas was subtracted, the ratio of commissural to ipsilateral axons was calculated, and then this ratio was defined as commissural index (Wilson et al., 2008).

To analyze the identity of Dbx1-transfected cells, the number of GFP-labeled cells and Brn3a-positive cells was counted on 12 μm-thick transverse cryosections of the dorsal midbrain using ImageJ software (Schneider et al., 2012). For this, a 150 μm square was placed on the Brn3a expression domain in the postmitotic layers of the dorsal midbrain, as the minimum vertical length of Brn3a-positive layers is approximately 150 μm at the embryonic stage examined. The ratio of Brn3a-positive cells to GFP-labeled cells within the square was then calculated. Three or four embryos of each experimental condition were used for the analysis.

We thank Fujio Murakami for anti-Robo3 and anti-Robo1 antibodies; Jun-ichi Miyazaki for pCAGGS vector; and Yusuke Kamachi for pCMV/SV2-HA vector. We are grateful to Nobuhiko Yamamoto for comments on the manuscript and encouragement of this work. We also thank members of the Shirasaki group for helpful comments. The mAb 99.1-3A2 developed by Thomas M. Jessell and Susan Brenner-Morton was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242, USA.