All experiments were performed in accordance with the European Community Council directive of 24 November 1986 (86-609/ECC) and the decree of 20 October 1987 (87-848/EEC). Mice were treated according to the principles of laboratory animal care, and experiments and mouse housing were approved by the Animal Welfare Committee of the Université catholique de Louvain (Permit Number: 2013/UCL/MD/11). The day of vaginal plug was considered as embryonic day (e)0.5. Minimum numbers of three embryos of the same genotype were analyzed in each experiment. The Ascl1^+/−, Hnf6^+/−; onecut 2(Oc2)^+/−, Presenilin-1 (PS1)^+/−, Pax6^+/Sey and Vsx1^+/τLacZ mutant mice were previously described (Hill et al., 1992; Guillemot et al., 1993; Wong et al., 1997; Jacquemin et al., 2000; Gong et al., 2003; Chow et al., 2004; Li et al., 2004; Clotman et al., 2005). Although β-galactosidase production was evident in a ventral population in Vsx1^{τLacZ/τLacZ} spinal cords, it was barely detectable in Vsx1^+/τLacZ heterozygous embryos, probably due to the negative auto-regulatory loop reported to control Vsx1 expression levels in the retina (Chow et al., 2004). Furthermore, β-galactosidase distribution was diffuse and punctuated, hindering the identification of the cells wherein it was present (data not shown). Therefore, a novel Vsx1^+/nlsLacZ line was generated using the PG00233_Z_5_A10 allele developed by the Knock-Out Mouse Project (KOMP). Vsx1 inactivation was confirmed by genotyping PCR and by complete loss of the Vsx1 protein. However, β-galactosidase was never detected in this line (data not shown). Nevertheless, Vsx1^{nlsLacZ/nlsLacZ} embryos were analyzed for the development of V2 interneuron populations.

Mouse embryos were fixed in PBS/4% PFA at 4°C for 15–30 min according to the developmental stage. Fixed mouse embryos were washed in PBS before incubation in PBS/30% sucrose overnight at 4°C. They were embedded in PBS/7.5% gelatin/15% sucrose and frozen at −55°C. Embryos were cut at 14 μm in a Leica CM3050 cryostat.

Cryosections were saturated with PBS/0.1% Triton/10% horse serum for 30 min and incubated with the primary antibodies diluted in the same solution at 4°C overnight. For Vsx1 labeling, cryosections were permeabilized with PBS/1% Triton for 30 min at room temperature and saturated for 30 min with PBS/0.1% Triton/1% horse serum. Anti-Vsx1 antibody diluted in the same solution was incubated for 2 h at room temperature. After three washes in PBS/0.1% Triton, the secondary antibodies, diluted in PBS/0.1% Triton/10% horse serum, were added for 30 min at room temperature. Slides were washed three times in PBS/0.1% Triton before a final wash in PBS/DAPI, and were mounted with Fluorescent mounting medium (DAKO).

The following primary antibodies and dilution were used: mouse anti-Ascl1 at 1:200 (BD #556604), guinea-pig anti-Ascl1 at 1:10,000 (Kim et al., 2008), mouse anti-beta III tubulin at 1:5000 (Chemicon #MAB1637), goat anti-BhlhB5 at 1:1000 (Santa Cruz #sc-6045), rabbit or rat anti-BhlhB5 at 1:2000 (Ross et al., 2012), sheep anti-Chx10 (Vsx2) at 1:500 (Exalpha Biologicals #X1179P), goat anti Dll4 at 1:200 (R&D System #AF1389), rabbit or guinea pig anti-Foxd3 at 1:5000 (Müller et al., 2005), rat anti-Gata3 at 1: 20 (Panayi et al., 2010), guinea-pig anti-Insm1 at 1:10,000 (Welcker et al., 2013), mouse anti-Islet 1/2 at 1:6000 (DSHB #39.4D5), mouse anti-Lhx3 at 1:1000 (DSHB #67.4E12), mouse anti-Lhx1/5 at 1:2000 (DSHB # 4F2), rat anti-Nkx6.1 at 1:2 (Ono et al., 2007), guinea pig anti-OC-1 at 1:6000 (Espana and Clotman, 2012), sheep anti-OC-1 at 1:250 (R&D System #AF6277), rat anti-OC-2 at 1:400 (Clotman et al., 2005), mouse anti-p27^kip1 at 1:2000 (BD), mouse anti-Pax6 at 1:1000 (DSHB #PAX6), mouse anti-phospho-Histone H3 at 1:1000 (Abcam ab14955), rabbit or guinea pig anti-Prdm8 at 1:1000 (Ross et al., 2012), goat anti-Prox1 at 1:100 (R&D Systems # AF2727), mouse anti-Shox2 at 1:500 (Abcam #ab55740), goat anti-Sox 1 at 1:500 (Santa Cruz #sc-17318), rabbit anti-Vsx1 at 1:500 (Clark et al., 2008).

Secondary antibodies from Life Science were used at 1:2000 and were donkey anti-mouse/AlexaFluor 594 or 488 or 647, anti-rabbit/AlexaFluor 647, anti-rat/AlexaFluor 594, anti-goat/AlexaFluor 488, anti-sheep/AlexaFluor 594, anti-mouse IgG₁/AlexaFluor 594 or anti-mouse IgG_2a/AlexaFluor 488. Secondary antibodies from Jackson ImmunoResearch were used at 1:1000 and were donkey-anti chicken/Dylight 488 or 594, anti-mouse/AlexaFluor 647, anti-rat/AlexaFluor 647, anti-sheep/AlexaFluor 594 or 647, anti-rabbit/AlexaFluor 488 or 594, anti-guinea pig/AlexaFluor 488 or 647. All secondary antibodies gave signals in the spinal cord only in the presence of corresponding primary antibodies.

Mouse embryos were fixed overnight in PBS/4% PFA at 4°C and processed as for immunolabeling experiments. Fourteen micrometer cryostat sections were cut and double in situ hybridization (ISH) protocol was performed essentially as previously described with slight modifications (Beguin et al., 2013; Pelosi et al., 2014). Sections were simultaneously hybridized overnight at 65°C with a DIG-conjugated Foxn4 (NM_148935.2, nucleotides 78–1643 (Francius et al., 2015)) and a fluorescein-labeled Vsx1 (NM_054068.2, nucleotides 121–2728, provided by C. Cepko) riboprobe. After hybridization, sections were washed four times in 50% Formamide, 1× SSC, 0.1% Tween-20 for 1 h at 65°C, twice in MABT buffer (100 mM maleic acid, 150 mM NaCl, 0.1% Tween20, pH 7.5) for 30 min before blocking in blocking buffer (MABT, 2% blocking reagent from Roche, 20% inactivated horse serum) for 2 h at room temperature. Sections were then incubated overnight with anti-DIG-alkaline phosphatase (AP)-conjugate antibody (Roche) at 4°C. After washing for 30 min in MABT, Foxn4 probe was visualized by AP-catalyzed chromogenic reaction using NBT-BCIP substrates (Roche) according to manufacturer’s instructions. The color reaction was stopped in 1× PBS and AP was inactivated by incubating the slides for 30 min with 0.1 M glycine/HCl pH 2.2. The sections were then washed twice in MABT buffer for 30 min, blocked again in blocking buffer for 2 h, and incubated to a 1:2000 dilution of anti-Fluo-AP-conjugate antibody (Roche) in blocking buffer overnight at 4°C. Slides were washed and incubated with HNPP/Fast Red kit (Roche) according to the manufacturer’s instructions to visualize Vsx1 expression. The reaction was stopped by washing in 1× PBS. Sections were then counterstained with DAPI and slides were mounted with Fluorescent mounting medium (DAKO).

Immunofluorescence and double ISH images of cryosections at thoracic level of the developing spinal cord were acquired on a Zeiss Axio Cell Observer Z1 confocal microscope with the Zeiss AxioVision Rel. 4.8 software and processed with Adobe Photoshop CS5 software. Double ISH images were also acquired using an EVOS^® FL Auto Imaging System (ThermoFisher Scientific) and related software. Quantifications were performed on red or green or blue layer of acquired confocal images and double or triple labeling cells were processed by subtractive method (Francius and Clotman, 2010). For each embryo (n ≥ 3), both sides of three sections at thoracic level were quantified using the count analysis tool of Adobe Photoshop CS5 software.

Raw data were exported from Adobe Photoshop CS5 software to SigmaPlot v12.3 software and processed in order to generate histogram figures. All data were analyzed and histograms were made with SigmaPlot software. Adequate statistical tests were applied based on the number of comparisons and on the variance in each group. For analysis of cell quantification based on comparison of two groups (control or mutant), standard Student’s t-tests were performed. Quantitative analyses were considered significant at p < 0.05. Three asterisks (***) indicate p < 0.001.

In mouse embryonic spinal cord, neurogenesis starts around e9.5. The earliest differentiating neurons are motor neurons, which originate from the pMN progenitor domain. We recently described that the production of V1 interneurons is also initiated at e9.5 (Stam et al., 2012). At this early stage, the presence of the OC transcription factors, namely OC-1 (or HNF-6), OC-2 and OC-3, was restricted to differentiating neurons and was observed in motor neurons and V1 interneurons (Francius and Clotman, 2010; Roy et al., 2012; Stam et al., 2012; Francius et al., 2013). Surprisingly, OC-1 was additionally detected in cells located between motor neurons and prospective Renshaw cells (Figures 1A–A″′; Stam et al., 2012). As the p2 progenitors are located between the pMN and the p1 domains, we postulated that these unidentified cells were most likely early-born V2 cells containing OC factors. To test this hypothesis we assessed the distribution of Ascl1, which is specifically detected in the p2 domain and early V2 interneurons (Gong et al., 2003; Wildner et al., 2006; Parras et al., 2007). Co-detection of OC-1 and Ascl1 confirmed that early-born OC-1⁺ cells derived from the p2 domain (Figures 1B–B″′). Given that OC factors are detected only in post-mitotic neurons (Francius and Clotman, 2010), we concluded that these cells were early V2 interneurons. However, V2 interneurons are reported to be generated from e10.5 onward (Zhou et al., 2000; Li et al., 2005; Peng et al., 2007). Accordingly, the number of V2a or V2b identified by the presence of Chx10 or Gata3, respectively, was restricted at e9.5 (1–5 cells per hemisection; Figures 1C–C″,F–F″, 2C) but significantly increased from e10.5 on (Figures 2A,C; Nardelli et al., 1999; Zhou et al., 2000; Smith et al., 2002). At e9.5, OC-1 was present in V2a and V2b interneurons, as well as in other cells located lateral to the p2 progenitor domain (Figures 1C–C″). Taken together, these observations suggested either that OC-1 was detected in early V2a/V2b interneurons before the onset of Chx10/Gata3 expression or that OC-1 was present in an additional unknown V2 subset. The latter hypothesis was consistent with results from a genetic lineage-tracing of Foxn4⁺ progenitors, which indicated that the p2 domain produces additional unidentified V2 interneurons (Li et al., 2010).

To distinguish between these possibilities, we searched for specific markers of these cells. Several transcription factors present in V2 interneurons are also detected in differentiating neurons of the retina (Glaser et al., 1992; Liu et al., 1994; Tomita et al., 1996; Gouge et al., 2001). Thus, specific markers of this V2 subset may be found among retinal proteins that are not yet associated with a specific spinal population. Vsx1 is the single paralog of the V2a marker Chx10 in the mouse genome and is expressed as Chx10 in the developing retina (Chow et al., 2001; Ohtoshi et al., 2001). Interestingly, Vsx1 expression in retinal cells is inhibited by Chx10 (Clark et al., 2008) and Chx10 expression is repressed by Vsx1 in Type 7 cone bipolar cells (Shi et al., 2011). In the zebrafish embryonic spinal cord, the Vsx1 ortholog is detected in common V2a/V2b progenitors and is transiently maintained in the V2a lineage (Batista et al., 2008; Kimura et al., 2008). In Xenopus and chick, Vsx1 expression was attributed to V2a interneurons based on its distribution similar to Chx10 (Chen and Cepko, 2000; D’Autilia et al., 2006). However, co-labeling of Vsx1 and other V2 markers in the mouse embryonic spinal cord was not reported yet.

Therefore, we characterized the distribution of Vsx1 in the mouse developing spinal cord. First, we assessed whether Vsx1 is present in the early OC-1⁺ V2 cells identified at e9.5. At this stage, Vsx1 was detected in cells containing OC-1 that were neither motor neurons nor early V1 interneurons (Figures 1D–D″′). As observed for OC-1 (Figures 1B–B″′), these cells contained Ascl1, confirming that they derived from the p2 domain (Figures 1E–E″). To determine whether these may correspond to V2a or V2b interneurons, the distribution of Vsx1 was compared to that of Chx10 and Gata3. However, Vsx1 was detected in cells lacking Chx10 or Gata3 (Figures 1F–F″), leaving open the question whether they corresponded to early V2 precursors or to a distinct population of V2 interneurons.

To further investigate this question, we compared the distribution of Vsx1 and V2 markers at later developmental stages (Figures 2A–E). We first noticed that the number of Vsx1⁺ cells increased from e9.5 to e10.5, then decreased to e12.5 and became undetectable at later stages (Figures 2A–C and data not shown), indicating a transient expression of Vsx1 in the mouse spinal cord as observed in zebrafish embryos (Passini et al., 1998). At all developmental stages, Vsx1 was found in cells located in close vicinity to V2a and V2b interneurons. However, co-detection of Vsx1 with Chx10 or Gata3 was never observed (Figures 2A–B″). Vsx1 was also present at e11.5 in cells located nearby V2d interneurons, characterized by the presence of Shox2 in the absence of Chx10 (Dougherty et al., 2013), but Shox2 was not detected in Vsx1⁺ cells (Figures 2D–D″). Finally, Vsx1 distribution was compared to that of Sox1, which labels V2c interneurons in addition to spinal progenitors (Panayi et al., 2010). However, V2c are located more ventrally and were distinct from cells containing Vsx1 (Figure 2E, arrow). Hence, Vsx1 is a specific, albeit transient, marker of a novel cell subset distinct from described V2 populations.

Spinal neurons that share lineage relationship or developmental origin often exhibit common markers. To assess whether Vsx1⁺ cells might relate to any V2 population, we established a repertoire of markers detected in these cells. As OC-1 was detected in Vsx1⁺ cells at early developmental stages (Figures 1D–D″), we assessed whether OC factors were co-detected with Vsx1 in more differentiated cells. At e10.5, OC-1 levels decreased sharply but the protein was still detected in 66% of the Vsx1⁺ cells (Figures 2F–F″), while OC-2 (Figures 2F–F″) and OC-3 (data not shown) were absent. As observed at e9.5, Ascl1 was present at e10.5 in the Vsx1⁺ cells, confirming their p2 origin (Figures 2G–G″′). Lhx3 is present in pMN and in p2 progenitors and is maintained in the medial motor column and in V2a interneurons but not in other V2 subsets. Lhx3 was detected in 95% of the Vsx1⁺ cells (Figures 2H–H″′), suggesting that they might relate to V2a interneurons. Accordingly, Prdm8 and Bhlhb5, which are present in V2a interneurons (Skaggs et al., 2011; Francius et al., 2013), were detected in 97% and 87% of the Vsx1⁺ cells, respectively (Figures 2G–H″′). In contrast Lhx1/5, which is present in V2b but is excluded from V2a interneurons, was barely detectable in 7% of the Vsx1⁺ complement (Figure 2I, arrowheads). Thus, as previously reported in other species (Chen and Cepko, 2000; D’Autilia et al., 2006; Batista et al., 2008; Kimura et al., 2008), Vsx1⁺ cells share some characteristics with V2a interneurons but Vsx1 is not detected in known V2 populations.

The lack of differentiated V2 interneuron markers in Vsx1⁺ cells and their medial location in the intermediate zone (Figures 1, 2) prompted us to assess whether they were differentiating neurons. Co-labeling experiments from e9.5 to e11.5 showed that only a minority of the Vsx1⁺ cells, decreasing from 14% to 0%, respectively, contained the neuronal differentiation marker β-_III-tubulin (Figures 3A–D). This suggested that Vsx1 was present before the onset of neuronal differentiation in p2-derived cells. Accordingly, Sox1, which labels the spinal progenitors in the ventricular zone, was detected in medial Vsx1⁺ cells located within the subventricular zone (Figures 2E, 3C–C″). These observations supported the hypothesis that Vsx1 is expressed in early V2 precursors before the onset of expression of their population-specific differentiation markers. Consistent with this idea, β-_III-tubulin was detected in all the V2a and V2b cells at e10.5 (Figures 3E–E″) and the cyclin-dependent kinase inhibitor protein p27^Kip1 was present in V2a and V2b interneurons but not in Vsx1⁺ cells (Figures 3F–F″, Chx10 and Gata3 both in blue). This suggested that Vsx1 expression occurs earlier than that of Chx10 and Gata3 in V2 interneurons. Hence, Vsx1 may be restricted to an early V2 compartment prior to V2 interneuron diversification.

To test this idea, we compared the distribution of Vsx1 with that of Prox1 and Insm1. These two factors define an early intermediate compartment containing late progenitors (i.e., progenitors in their last division) and newborn neurons (Duggan et al., 2008; Misra et al., 2008). Indeed, we found that Vsx1 was present in cells containing both Prox1 and Insm1 (Figures 4A–A″), demonstrating that Vsx1 is restricted to an early intermediate V2 subset. The presence of Ascl1 in this subset but not in differentiating V2a or V2b interneurons (Figures 1B–B″, 2G–G″, 4B–B″, Chx10 and Gata3 both in blue) suggested that the Vsx1⁺ cells may correspond to precursors of the described V2 populations. We thus investigated whether Vsx1⁺ cells contained other determinants of V2 development. The segregation of V2a and V2b interneurons depends on a regulatory network involving Ascl1 and Foxn4 that eventually controls production of the Notch ligand Dll4 and thereby differentially modulates activation of Notch signaling in V2 precursors (Li et al., 2005; Del Barrio et al., 2007; Peng et al., 2007; Misra et al., 2014; Zou et al., 2015). Dll4 was present in Vsx1⁺ cells at the time of V2a/V2b production (Figures 4C–C″, arrowheads), although it was not exclusive to the Vsx1⁺ population. Similarly, Foxn4 expression was detected by double ISH in the cells expressing Vsx1 (Figures 4D–F′). As V2 precursors expressing Foxn4 have been shown to generate all the described V2 interneuron populations as well as additional V2 subsets (Li et al., 2010), our data demonstrate that Vsx1⁺ cells are precursors of V2a/V2b interneurons and of other V2 subsets that also derive from Foxn4- and Dll4-containing progenitors (Panayi et al., 2010; Panayiotou et al., 2013; Zou et al., 2015).

We then assessed whether Vsx1 is detected in dividing cells of the ventricular zone or in intermediate precursor cells. Pax6 and Nkx6.1, which are present in the p2 progenitors, were detected in the Vsx1⁺ cells. However, these cells were clustered nearby the lateral border of the ventricular zone rather than the lumen (Figures 4G–G″′). Consistently, although Sox1 was present at low levels (Figures 2E, 4H–H″′), the proliferation marker Ki-67 was barely detectable in Vsx1⁺ cells (Figures 4H–H″′) and only remaining traces of phosphorylated-Histone H3, specific to the metaphase of mitotic division, were found (Figures 4I,I′, arrowhead). Taken together, these observations suggest that Vsx1 is present in cells that just completed their last mitotic division but did not yet initiate neuronal differentiation, defining an early intermediate V2 interneuron compartment. As observed respectively in motor neurons and in V1 interneurons, Nkx6.1 and Pax6 are maintained in these cells even though they are no longer dividing.

Given the possibility that Vsx1⁺ cells may constitute precursors of V2 populations, we assessed whether their production depends on factors necessary for proper V2 interneuron development. The segregation of the V2a and V2b lineages relies on the asymmetrical activation of the Notch pathway by Dll4, which stimulates Notch signaling in surrounding cells and thereby promotes V2b differentiation, restricting V2a differentiation to cells with lower Notch activity (Li et al., 2005; Del Barrio et al., 2007; Peng et al., 2007; Misra et al., 2014; Zou et al., 2015). Proper activation of this pathway requires mosaic expression of several transcriptional regulators including Foxn4 and Ascl1 in the p2 domain (Li et al., 2005; Del Barrio et al., 2007; Xiang and Li, 2013; Misra et al., 2014). To assess whether the production of the Vsx1⁺ cells also depends on this network, we evaluated the phenotype of these populations in embryos mutant for Foxn4, Ascl1 or PS1. Lack of Foxn4 resulted as expected in a strong imbalance in the generation of V2 interneuron subsets, with V2a being produced in excess at the expense of V2b cells (Figures 5A–B″). However, the generation of Vsx1⁺ cells was not affected (Figures 5A–C). Similarly, inactivation of Ascl1 resulted in excessive production of V2a interneurons at the expense of V2b, but the number of Vsx1⁺ cells was not altered (Figures 5D–F). Consistently, loss of PS1, which results in severe diminution of Notch activation (Struhl and Greenwald, 2001), did not alter the specification of Vsx1⁺ cells (Figures 5G–I). These observations indicate that the generation of Vsx1⁺ cells does not depend on Notch signaling, but remain consistent with the hypothesis that these cells may constitute the precursor population wherein differential activation of the Notch pathway determines further V2 interneuron diversification.

We then turned our analyses to the other transcription factors expressed during Vsx1⁺ cell development. OC factors are present in the early Vsx1⁺ cells (Figures 1, 2). The role of OC factors has been evaluated in Oc1/Oc2^−/− double-mutant embryos, which also lack Oc3 in the CNS (Roy et al., 2012). However, the number of Vsx1⁺ cells was not changed in the spinal cord of these mutants, suggesting that OC proteins are not necessary for their production (Figures 6A–C).

Next, we investigated the role of Pax6 in Vsx1⁺ cell generation. Pax6 is required for patterning the ventral spinal cord during neurogenesis (Ericson et al., 1997) and for proper generation of V1 interneurons (Burrill et al., 1997; Sapir et al., 2004; Gosgnach et al., 2006). In Small eye (Sey/Sey) mutant embryos, which lack Pax6 (Glaser et al., 1992), the production of V0 and V2b populations is not affected (Burrill et al., 1997; Sapir et al., 2004; Gosgnach et al., 2006; Zhang et al., 2014) while the number of V2a interneurons is only slightly decreased (Ericson et al., 1997; Gosgnach et al., 2006). The persistence of Pax6 in the early Vsx1⁺ cells suggested that this factor may contribute to their development. Accordingly, Vsx1 was barely detectable in Sey/Sey embryos at e11.5 (Figures 6D–F) and e12.5 (data not shown), although the production of V2a and V2b interneurons was conserved. The presence of V2a and V2b interneurons indicated that the p2 progenitor domain was not re-specified and that the V2a/V2b precursors were preserved. In contrast, the loss of Vsx1 unveiled a selective requirement for Pax6 in the expression of Vsx1 in this early intermediate V2 compartment. Furthermore, these observations suggest that Vsx1 may be dispensable for proper generation of V2 interneuron subsets.

To test this hypothesis, we assessed the role of Vsx1 in V2 interneuron development. In the absence of any other specific marker of the Vsx1⁺ cells, this compartment could not be analyzed in the Vsx1 mutant embryos. In contrast, as Vsx1-containing precursors generate all the V2 subsets, we addressed a possible role for Vsx1 in V2 interneuron development. An inverse relationship between the levels of Chx10 and Vsx1 was previously reported in bipolar cells of the retina (Clark et al., 2008). To investigate whether a similar relationship also exists in spinal cord, the distribution of Chx10 was first assessed in Vsx1 mutant embryos. However, the number of Chx10-positive cells was similar at e12.5 in control and in mutant embryos, indicating that the absence of Vsx1 does neither impact on V2a interneuron production nor on Chx10 expression. Accordingly, the number of V2b interneurons was not affected and the V2a/V2b ratio was preserved (Figures 6G–I). Taken together, these observations confirm that Vsx1 is not required for the production of described V2 interneuron populations, suggesting that the integrity of the early V2 precursor compartment is also preserved in Vsx1 mutant embryos. Consistently, Vsx1 mutant mice did not display any alteration in their motor behavior (data not shown).