During spinal cord development, numerous neuronal populations that progressively subdivide into smaller and discrete subsets are generated in a space- and time-regulated manner from a collection of distinct progenitor domains orderly distributed along the ventro-dorsal axis of the neural tube (Lewis, 2006; Lu et al., 2015). Each population (Grossmann et al., 2010), and each subset it generates (Bikoff et al., 2016), migrates along stereotyped pathways and settles at well-defined locations in the developing spinal cord. Several observations suggest that the position of spinal neurons is a critical determinant of microcircuit organization. Indeed, the clustering and dorsoventral settling position of motor neuron pools serve as a determinant of the pattern of sensory input specificity (Sürmeli et al., 2011). Similarly, position of dorsal interneuron (dINs) along the mediolateral axis in lamina V determines their connectivity with sensory afferents (Ladle et al., 2007; Tripodi et al., 2011; Hilde et al., 2016). Furthermore, positional distinctions among V1 interneuron subsets (Bikoff et al., 2016) or Lbx1-derived premotor interneurons (Goetz et al., 2015) constrain patterns of input from sensory and motor neurons. Finally, multiple interneuron subsets are differentially distributed along the antero-posterior axis of the spinal cord, consistent with their integration into specific but distinct local microcircuits (Francius et al., 2013). The genetic programs that support the generation and the diversification of spinal interneurons have been extensively investigated (Lu et al., 2015). However, in contrast to cortical interneurons whose migration mechanisms have been largely deciphered (Guo and Anton, 2014; Barber and Pierani, 2016), the molecular determinants that regulate the distribution of interneuron populations in the developing spinal cord remain elusive.

Spinal interneurons generated from dorsal progenitor domains of the neural tube integrate and relay proprioceptive, mechanosensory or nociceptive information to motor circuits or to higher brain centers (Windhorst, 2007; Todd, 2010; Abraira and Ginty, 2013; Bui et al., 2015; West et al., 2015; Lai et al., 2016). Dorso-ventral (DV) gradients of BMP and Wnt signaling contribute to specify three dorsal progenitor domains (pdI1–3) according to their position along the DV axis (Wine-Lee et al., 2004; Timmer et al., 2005; Zechner et al., 2007; Tozer et al., 2013), whereas generation of the three more ventral domains (pdI4–6) occurs independently of these signals (Zhuang and Sockanathan, 2006; Le Dréau and Martí, 2012; Lai et al., 2016). Specific combinations of pro-neural factors and of bHLH or homeodomain proteins direct the identity of the post-mitotic neuronal populations generated from these domains (Zhuang and Sockanathan, 2006; Lai et al., 2016). Subsequently, dorsal cells acquire an inhibitory or excitatory phenotype through a complex integrated network of factors (Cheng et al., 2004, 2005; Mizuguchi et al., 2006; Henke et al., 2009; Chang et al., 2013; Borromeo et al., 2014; Hanotel et al., 2014). In contrast to spinal ventral interneurons, for which numerous subpopulations and subsets have been identified (Francius et al., 2013; Lu et al., 2015) and partly functionally characterized (Gosgnach et al., 2006; Zagoraiou et al., 2009; Borowska et al., 2013; Dougherty et al., 2013; Talpalar et al., 2013), the subdivision of dIN populations into more discrete subsets has only been sparsely investigated (Ding et al., 2004; Andersson et al., 2012; Dougherty et al., 2013; Bourane et al., 2015; Hilde et al., 2016). In addition, although these cells migrate according to highly specific and distinct directions and patterns, and display various final settling positions ranging from ventral to the dorsal most regions of the spinal cord (Lai et al., 2016), the molecular mechanisms that regulate their distribution remain unknown.

Recent observations in our laboratory indicate that transcription factors of the Onecut (OC) family are present in numerous neuronal populations, including motor neurons and interneurons (Francius and Clotman, 2010; Roy et al., 2012; Francius et al., 2013) and regulate neuronal diversification and migration in the developing spinal cord (Roy et al., 2012; Francius and Clotman, 2014); unpublished data). Onecut factors, namely Hepatocyte Nuclear Factor-6 (HNF-6, or OC-1), OC-2 and OC-3, are transcriptional activators detected in liver, pancreas and CNS during embryonic development (Lemaigre et al., 1996; Landry et al., 1997; Jacquemin et al., 1999, 2003; Vanhorenbeeck et al., 2002). In the CNS, they regulate the production (Espana and Clotman, 2012a), the diversification (Roy et al., 2012; Francius and Clotman, 2014), the distribution (Espana and Clotman, 2012a,b; Audouard et al., 2013) and the maintenance (Espana and Clotman, 2012a,b; Stam et al., 2012) of specific neuronal populations, as well as the formation of neuromuscular junctions (Audouard et al., 2012). In motor neurons, they are required to maintain levels of Isl1 expression necessary to support the diversification of this population into distinct subsets (Roy et al., 2012). Their presence was also detected in the dorsal embryonic spinal cord (Francius and Clotman, 2010; Francius et al., 2013). However, their distribution in dINs and their possible functions in the development of these cell populations remain unknown. Here, we demonstrate that OC factors are produced in a temporally- and spatially-defined manner in most early populations of spinal dINs. Furthermore, our analyses of mutant mice devoid of OC factors suggest that these proteins control genetic programs that contribute to the regulation of dIN diversification and distribution during embryonic development.

During embryonic development, the dorsal neural tube generates interneuron populations that contribute to motor circuits or that integrate and relay proprioceptive, sensory or nociceptive information to motor circuits or to higher brain centers. The molecular mechanisms regulating the production of those interneurons have been extensively deciphered (Helms and Johnson, 2003; Lewis, 2006; Zhuang and Sockanathan, 2006), but the developmental determinants controlling their diversification, migration and final distribution (Lewis, 2006; Lai et al., 2016) remain elusive. Here, we provide evidence that the OC transcription factors are present in subsets of differentiating dINs and regulate some aspects of the diversification and of the distribution of these populations, likely impacting on neural circuit formation.

In spinal motor neurons and ventral interneurons, OC transcription factors are present in fractions of homogeneous subpopulations irrelevant to their molecular characteristics, neurotransmitter phenotype, location or contribution to neural circuits (Roy et al., 2012; Francius et al., 2013). Similarly, we detected OC factors in multiple dIN subsets including excitatory (dI2, dI3 and dI5) or inhibitory (dI4 and dI6) populations, generated at various levels along DV axis of the neural tube (from dI2 to dI6), involved in modulating motor control (dI3, dI4, dI5 and dI6) or processing sensory information (dI2), and migrating to ventral (dI3, dI5 and dI6) or deep dorsal (dI2 to dI5) locations in the mature spinal cord (Helms and Johnson, 2003; Lewis, 2006; Zhuang and Sockanathan, 2006; Lu et al., 2015; Lai et al., 2016). As for the other neuronal populations wherein OC factors are detected, this raises the question of the upstream regulators generating such a complex and ill-defined expression pattern in the developing CNS. To date, factors controlling Oc expression in the spinal cord remain unknown. Furthermore, the cis-acting sequences that stimulate HNF-6 production in the developing liver or pancreas are ineffective in the CNS (Poll et al., 2006), suggesting that Oc expression in the neural tissue is regulated by distant enhancers.

Although OC factors are present in differentiating neurons (Francius and Clotman, 2010; Roy et al., 2012; Francius et al., 2013), their distribution in dINs correlates with the DV morphogen gradients known to determine the production of distinct progenitor domains in the dorsal neural tube. Indeed, OC are present in a larger proportion of cells and for a longer period of time in the more ventral populations of dINs than in their dorsal complement. The fraction of dorsal populations containing OC proteins increases from dI2 to dI6 and the number of cells that maintain Oc expression at e12.5 also rises from dI2 to dI6 (Figures 1–3). This is reminiscent of the gradient of Wnt and BMP morphogens initially produced by the dorsal ectoderm and subsequently maintained by the roof plate, which determines the identity of the three dorsal-most progenitor compartments without affecting the more ventral populations (Wine-Lee et al., 2004; Timmer et al., 2005; Zechner et al., 2007; Tozer et al., 2013). While expression of receptor Ia to BMP (BMPRIa) is restricted to spinal progenitors, BMPRIb is additionally expressed in the intermediate zone and in differentiating neurons (Yamauchi et al., 2008). Furthermore, Wnt receptors of the Frizzled family are detected at early stages in the whole spinal cord (Hu et al., 2009) and Wnt signaling is active in postmitotic neurons (Galli et al., 2014; Avilés and Stoeckli, 2016). This indicates that, beyond patterning of the neural tube, BMP and Wnt signaling is maintained in differentiating spinal interneurons. Therefore, Oc expression may be inhibited in dIN populations by BMP/Wnt signaling that would remain active in differentiating cells. Consistent with this, exposure of human ES cells to Noggin, a BMP antagonist, combined to retinoic acid and FGF-10 increases Hnf6 expression (Mfopou et al., 2010). In contrast, Wnt1 signaling inhibition by Dkk-1 in ventral midbrain neural stem cells reduces Hnf6 expression, suggesting that Wnt1 may rather stimulate OC production in neural cells (Yuan et al., 2015). Similar to OC factors, Pax6 is absent from the most dIN populations because its expression is repressed by BMP signaling (Bel-Vialar et al., 2007). In the retina, OC factors act downstream of Pax6 to maintain horizontal cell identity (Klimova et al., 2015). However, regulation of Oc expression by proneural factors or by transcriptional regulators differentially expressed in distinct interneuron populations along the DV axis of the spinal cord has not been reported yet. Hence, identification of the factors upstream of the Oc genes in the neural tube constitutes a prerequisite to understand how their complex distribution pattern is achieved in the developing CNS.

Nevertheless, OC regulation must follow stereotyped mechanisms, as similar spatial and temporal progressions are observed among the different neuronal populations wherein OC are detected. Indeed, the distribution area of the three OC proteins follow a HNF-6 > OC-2 > OC-3 hierarchy and Hnf6 is expressed earlier than Oc2 detected itself before Oc3 in spinal motor neurons (Francius and Clotman, 2010; Roy et al., 2012) and ventral interneurons (Francius et al., 2013). A similar distribution dynamic was generally observed in dINs (Figures 1–3), with the single exception for OC-3 detected in a broader fraction of Phox2a-positive dI5 interneurons than OC-2 at e12.5 (Figure 2P). This suggests that similar mechanisms contributing to control Oc expression are acting in different cell types, raising the possibility that these factors may regulate generic aspects of neuronal development rather than the acquisition of specific characteristics by distinct neuronal subsets.

However, analysis of the few early dIN subpopulations identified to date suggested that OC factors contribute to dIN diversification. Indeed, the Phox2a subset of dI5 interneurons was reduced in the absence of OC proteins whereas the total number of dI5 cells was maintained. This alteration is not caused by apoptosis, as cell death is not increased in OC mutant embryos (Roy et al., 2012). Lack of markers for other dI5 subsets prevented to investigate whether this resulted from defective production of Phox2a-positive cells or from conversion into other dI5 subtypes. OC factors are known to regulate neuronal diversification in the developing spinal cord. They critically regulate the differentiation ratio between visceral and somatic motor neurons at thoracic levels of the spinal cord and the production of ventral limb innervating motor neurons at brachial and lumbar levels (Roy et al., 2012; Francius and Clotman, 2014). Interestingly, OC control these two developmental events by maintaining significant expression levels of the Lim-homeodomain transcription factors Isl1 in the differentiating motor neurons (Roy et al., 2012). Isl1 maintenance results from direct stimulation of Isl1 expression by the OC factors, as evidenced by direct binding of OC proteins (Roy et al., 2012; Kim et al., 2015) to the CREST-2 enhancer of Isl1 (Uemura et al., 2005) and by the dependence of CREST-2 activity on the OC factors (Roy et al., 2012; Kim et al., 2015). Isl1 is also produced in dI3 interneurons during their differentiation (Liem et al., 1997; Bui et al., 2013), and determines their axonal projections at later developmental stages (Avraham et al., 2010). The CREST-2 enhancer of Isl1 is active in the dI3 interneurons (Kim et al., 2015), suggesting that OC factors may also regulate Isl1 expression in dI3 neurons and possibly later steps of dI3 differentiation. However, Isl1 was detected in dI3 neurons in the absence of OC factors and the number of dI3 cells in Hnf6/Oc2-/- mutant embryos was comparable to that observed in control littermates, indicating that OC are dispensable for dI3 production and for Isl1 expression in dI3. This demonstrates that Isl1 regulation by the OC factors is highly cell-type dependent and suggests that other factors that remain to be identified activate the CREST-2 enhancer in dI3 interneurons independently from OC proteins.

Similarly, OC factors regulate the diversification of spinal ventral interneurons, the production of some ventral interneuron subsets being impaired in the absence of OC proteins (AH, VR and FC, unpublished observations). The characterization of distinct ventral neuronal subsets has been frenetically pursued in the recent years and led to the identification of an extensive, although yet incomplete, collection of dozens of discrete subpopulations with specific molecular, electrophysiological, distribution or functional characteristics (Gosgnach et al., 2006; Zagoraiou et al., 2009; Panayi et al., 2010; Borowska et al., 2013; Dougherty et al., 2013; Francius et al., 2013; Panayiotou et al., 2013; Talpalar et al., 2013; Bikoff et al., 2016). This knowledge provided molecular and genetic tools to specifically identify, record or functionally challenge these different subsets and to characterize their phenotype in mutant embryos. In contrast, only few early dIN subsets have been identified to date, based on molecular markers (Ding et al., 2004; Rebelo et al., 2007; Andersson et al., 2012; Dougherty et al., 2013; Hilde et al., 2016). Thus, a possible involvement of OC proteins in dorsal diversification awaits a more complete characterization of the different subpopulations produced by the known cardinal populations of dINs.

Moreover, our data indicate that the OC factors regulate the distribution of different dIN populations during spinal cord development. Although limited to short-distance cell movements, spinal neuronal migration is a very complex process. On the one hand, each population or subset of motor neurons and of interneurons migrates according to a specific trajectory, in a specific direction, and eventually settles at a specific location. In addition, some populations assume their general migratory trajectory along the ventricular zone as soon as they exit their progenitor domain whereas others migrate first in a radial orientation before turning in ventral or dorsal direction. These combined processes result in an intense intermingling of cells with distinct identities. On the other hand, cells from a single subset do not migrate as a cohesive compact cluster but rather form a relatively loose population that progresses in a common general direction. Eventually, these cells settle at a more or less focused final location depending on each subset. To analyze this highly dynamic and complex process, we adapted a distribution analysis routine initially developed for motor neurons (Palmesino et al., 2010). This routine enabled to obtain a highly reproducible integrated view of a cell population distribution on the transversal plane of the spinal cord and to precisely compare the migration pathway and the final location adopted by molecularly-defined cell populations in wild type or mutant embryos.

Consistent with the presence of OC proteins in small proportions of each dI population, only fractions of dIN subsets were mislocated in Oc mutant embryos. Lack of OC does not affect the capacity of cells to migrate nor their intrinsic motility. Indeed, we did not observe any delay in dIN migration nor any premature arrest of migration at inappropriate settling location in Oc mutant embryos, suggesting that their migration abilities are conserved. In contrast, three types of migration defects were observed in Hnf6/Oc2-/- embryos. First, the migration trajectory of the dI3 interneurons was shifted to a more medial pathway than in control embryos (Figures 4P–Q). Second, the organization of migrating dI3 (Figures 4P–Q) and WT1-positive dI6 interneurons (Figure 6S) was altered, since cells migrated as a more compact cluster with the highest cell density closer to the migration front. Third, Phox2a-positive dI5 interneurons settling position was modified (Figures 5M–P), with a reduced number of cells in the lateral cluster and significant migration in ventro-medial locations where Phox2a-positive dI5 are never found in control embryos. Slight changes in settling location were also observed for dI3 (Figures 4P–Q) and dI6 (Figure 6S) subsets. These observations suggest that the production of environmental cues that determine the migration pathway and the settling location of dIN populations or their interpretation by migrating cells is altered in the absence of OC factors. They also raise the possibility that factors involved in clustering or in cohesion of these migrating populations may be affected, as previously proposed in the dopaminergic A13 nucleus of Hnf6/Oc2-/- embryos (Espana and Clotman, 2012b), although changes in cell clustering may be secondary to alterations in the migration of individual cells. The molecular mechanisms regulated by OC actors in these processes remain to be identified.

Requirement of proper neuronal migration for adequate formation of neural circuits in the developing spinal cord has been repeatedly questioned (Kania, 2014). Indeed, some mutant lines show mispositioned neuron cell bodies but proper connectivity of the affected cells and organization of the corresponding neural circuits (Coonan et al., 2003; Demireva et al., 2011). However, increasing amount of evidence demonstrate that the location of spinal neurons is critical for proper neuronal connectivity and adequate integration into sensory-motor circuitries (Sürmeli et al., 2011; Bikoff et al., 2016). In dorsal spinal cord, the spatial organization of Lbx1-derived premotor interneurons determines their role, i.e., flexor interneurons born at e10.5 migrate laterally in lamina VI while extensor interneurons born at e12.5 remain closer to the central canal (Tripodi et al., 2011). EphA4 in these cells determines their localization as well as their axonal projections and their contribution to left-right motor alternation (Satoh et al., 2016). Likewise, the sensory cues that trigger motor responses are linked to the ML position of dorsal premotor interneurons in the deep dorsal horn (Hilde et al., 2016). Therefore, identification of the molecular regulators of neuronal migration and distribution in the spinal cord is important to understand circuit formation and neural activity. Interestingly, the interneuron populations altered in absence of OC factors are involved in motor control, either modulating motor neurons (Andersson et al., 2012; Bui et al., 2013; Goetz et al., 2015; Satoh et al., 2016) or ventral premotor interneuron activity (Levine et al., 2014; Hilde et al., 2016) or mediating presynaptic inhibitory control of proprioceptive sensory neurons (Betley et al., 2009; Fink et al., 2014). Hence, OC factors could have been recruited in different neuronal subsets to coordinate the development of distinct populations eventually involved in motor circuits.

KUK, GM, AH and FC designed the experiments. BJ generated Matlab routines for the distribution analyses of dorsal interneurons. KUK, GM, AH and VR performed the experiments. GM performed additional experiments for the revision of the manuscript, and all the authors discussed the data. KUK and FC drafted the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.