The topography of corticopontine projections is controlled by postmitotic expression of the area-mapping gene Nr2f1

Neuronal populations responsible for fine motor coordination are arranged in topographically organized maps in the neocortex and cerebellum, with different body parts being represented in largely continuous maps in the somatosensory cortex (Chapin and Lin, 1984; Fabri and Burton, 1991; Welker, 1971; Woolsey and Van der Loos, 1970), but in discontinuous fractured maps in the cerebellum (Bower, 2011; Bower et al., 1981; Bower and Kassel, 1990; Leergaard et al., 2006; Nitschke et al., 1996; Shambes et al., 1978). The intercalated regions of this network, the pontine nuclei, deep cerebellar nuclei and thalamus, receive and integrate signals, ultimately resulting in coordinated and seamlessly executed behaviours (Buckner, 2013; Peterburs and Desmond, 2016; Stoodley and Schmahmann, 2010), including fine voluntary movements (Badura et al., 2013; Mottolese et al., 2013).

The pontine nuclei constitute the major synaptic relay for cerebro-cerebellar signals (Brodal and Bjaalie, 1992; Lemon, 2008; Mihailoff et al., 1985). Axonal projections originating from layer V pyramidal neurons across the neocortex are distributed in topographically organized clusters within the pontine nuclei, as shown in monkey (Brodal, 1978; Schmahmann and Pandya, 1997), cat (Bjaalie and Brodal, 1997), rat (Leergaard et al., 2000a,b) and, to some extent, mouse (Henschke and Pakan, 2020; Inoue et al., 1991; Proville et al., 2014). Within the pontine nuclei, the three-dimensional (3D) arrangement of clustered terminal fields, well described in rats, both preserves the overall topographical relationships of the cortical maps, and partially overlaps and introduces new spatial proximities among projections from different cortical areas (Bjaalie and Brodal, 1989; Leergaard, 2003; Leergaard and Bjaalie, 2007).

To date, the mechanisms responsible for establishing the topographic map between the neocortex and pontine nuclei are poorly understood. The leading proposition, referred to as chrono-architectonic hypothesis, postulates that the complex 3D topography is a product of straightforward spatio-temporal gradients, possibly combined with non-specific chemo-attractive mechanisms (Leergaard, 2003; Leergaard and Bjaalie, 2007; Leergaard et al., 1995). Recent new discoveries raise the possibility that other mechanisms are also in action during the establishment of the corticopontine maps. Several lines of evidence suggest a functional role of graded gene expression during topography of sensory maps in several systems (D'Elia and Dasen, 2018; Erzurumlu et al., 2010; Fritzsch et al., 2019; McLaughlin and O'Leary, 2005), but whether this process is also operative during establishment of corticopontine topography is not understood. A recent study has shown that postmitotic graded expression of the HOX gene Hoxa5 is directly involved in imparting an anterior-to-posterior identity to pontine neurons and in attracting corticopontine axons (Maheshwari et al., 2020). Whether gradient expression of molecular factors along the antero-posterior or medio-lateral axes of the cerebral cortex also intrinsically determines the topography of corticopontine projections is still not known.

Layer V neurons from the anterolateral cerebral cortex project to the central regions of the pontine nuclei, while more medially located cortical regions project to more external parts; motor area projections are distributed more medially and rostrally, while somatosensory projections reach the middle and caudal parts of the pontine nuclei. Finally, auditory and visual cortical projections innervate the dorsolateral regions of the pontine nuclei (Leergaard et al., 2004; Leergaard and Bjaalie, 2007). The fine-tuned and precise corticopontine topography raises the possibility that cortical neurons are intrinsically programmed to target specific groups of pontine neurons, possibly coupling both intrinsic and extrinsic mechanisms to direct proper topographical innervation to the pontine nuclei.

A new theme of cortical patterning has emerged, in which genetic factors direct the spatial and temporal establishment of topographically organized axonal connections between the cortex and subcortical brain regions (Cadwell et al., 2019). Area-mapping genes are expressed in gradients along the different axes of the cortical primordium, and are known to modulate the size and position of future cortical areas (Alfano and Studer, 2012; Cadwell et al., 2019; O'Leary and Sahara, 2008). They also determine areal fate and regulate expression of downstream molecules that, in turn, control the topographic organization of synaptic inputs and outputs of related structures (Assimacopoulos et al., 2012; Greig et al., 2013). These genes represent good candidates for modulating topographic mapping. In mice, the Nr2f1 gradient expression appears to be a particularly strong candidate for having a formative role during the establishment of topographic maps (Armentano et al., 2007; Liu et al., 2000; Zhou et al., 2001). For example, Nr2f1 is expressed in cortical progenitor cells from embryonic day E9.0 in a high caudo-lateral (future sensory) to low rostro-medial (future motor) gradient fashion, which is maintained in postmitotic descendants and postnatally, when the cortical area map is completed (Bertacchi et al., 2019; Flore et al., 2017; Tomassy et al., 2010). Previous studies show that Nr2f1 promotes somatosensory (S1) cortex identity by repressing motor identity in postmitotic neurons; in its absence, area size and thalamocortical topography are both affected (Alfano et al., 2014; Armentano et al., 2007; Bertacchi et al., 2019). We thus hypothesized that Nr2f1 could control topographic corticopontine mapping during corticogenesis.

To this purpose, we made use of cortico-specific Nr2f1 conditional knockout mice as an in vivo model system and a paradigm to investigate the contribution of cortical genetic programs in the establishment of topographic corticopontine projections. Two distinct conditional mouse lines were crossed to the Thy1-eYFP-H reporter line (Feng et al., 2000), in which YFP is highly expressed in cortical layer V pyramidal neurons and their axonal projections (Porrero et al., 2010). The distribution of fluorescent YFP signals as well as anterogradely labelled corticopontine projections were evaluated by comparison of spatially corresponding microscopic images and 3D visualization of extracted point-coordinate data representing labelling. Our results indicate that cortical Nr2f1 expression plays a dual role in controlling the spatio-temporal development of corticopontine projections. Although early expression in progenitor cells influences the ratio between corticofugal fibres passing the pontine nuclei, loss of postmitotic late expression specifically affects topographic pontine mapping. Overall, our data demonstrate that intrinsic genetic programs and postmitotic graded expression of cortical area-mapping genes are implicated in the establishment of area-specific targeting of corticopontine neurons.

To first establish a 3D reference of the topographical organization of corticopontine projections in wild-type adult mice, we used tract tracing data from the Allen Institute Mouse Brain Connectivity Atlas (Wang et al., 2020), see flowchart in Fig. 1A. These data allowed us to visualize the spatial distribution of the pontine projections of motor and somatosensory neocortical areas. Corticopontine projections digitized from sagittally oriented microscopic images (matching the orientation used for our experimental data, see below) were co-visualized as 3D data points in a 3D viewer tool (see, for example, Fig. 2A-D). Experimental data were selected and coloured according to tracer injection localizations in the cerebral cortex: first, to visualize and compare the pontine distribution of corticopontine projections from motor and somatosensory areas; and, second, to determine the organization of projections arising from progressively more medial and caudal locations in the cerebral cortex, following the cortical neurogenetic gradient that ripples out from the anterolateral cortex (Leergaard and Bjaalie, 2007; Smart, 1984). Our findings confirm that the somatosensory and motor neurons of the mouse cortex project to largely separate parts of the pontine nuclei (Henschke and Pakan, 2020; Inoue et al., 1991; Proville et al., 2014), with clustered terminal fields topographically distributed in the same concentric fashion (Fig. 2L-P) as previously reported in rats (Leergaard et al., 2000a; Leergaard and Bjaalie, 2007). These 3D point data were also used below as additional control data, and as benchmarks for interpreting YFP expression and tract-tracing results in Nr2f1 mutant mice.

To assess the influence of cortical area mapping on the establishment of topographical organization in mouse corticopontine projections, we used Nr2f1 conditional knockout mice as an experimental model system and their littermates as controls (Alfano et al., 2014; Armentano et al., 2007). Two well-established conditional Nr2f1 mouse mutants were used: the Nr2f1^fl/fl::Emx1-Cre mouse, in which Nr2f1 expression is abolished from early cortical progenitor cells at mouse embryonic stage (E) 9.5 (Armentano et al., 2007); and the Nr2f1^fl/fl::Nex-Cre mouse, in which Nr2f1 expression is inactivated at later stages (E11.5-E12), solely in cortical postmitotic neurons (Alfano et al., 2014; Goebbels et al., 2006). Both mouse lines were crossed to the Thy1-eYFP-H reporter line to specifically restrict signal expression to layer V pyramidal neurons (Harb et al., 2016; Porrero et al., 2010). In these mice, YFP expression follows the physiological distribution of subcortical layer V projections, including corticospinal and corticopontine fibres, with a high density in motor cortex and a gradually decreased distribution in S1 and more caudal areas. For simplicity, Nr2f1^fl/fl::Emx1-Cre and Nr2f1^fl/fl::Nex-Cre are referred to as Emx1-cKO and Nex-cKO, respectively.

In agreement with a previous study (Porrero et al., 2010), we observed substantial YFP signal expression in the hippocampus, tectum and pontine nuclei, as well as in the globus pallidus, claustrum, endopiriform nucleus, nucleus of the lateral olfactory tract, mammillary nuclei, piriform area and the substantia innominata in adult mice (Fig. 3A). Signal expression was also seen in the vestibular nuclei, deep cerebellar nuclei and cerebellum. Although YFP expression was present in almost the same cortical regions in 2-month-old mutant mice as in controls, detailed analysis of layer V expression revealed some distinct differences in the spatial distribution of Emx1-cKO and Nex-cKO brains relative to their respective controls, and between the two conditional lines. We used an ImageJ macro to automatically count YFP⁺ nuclei by area of interest in the brain using a threshold based on intensity and shape of the elements. This allowed us to estimate the number of YFP⁺ neurons in seven cortical areas (corresponding to prefrontal, motor, somatosensory, auditory, visual and retrosplenial cortices in control mice) defined by delineations from spatially registered overlay images from the Allen Mouse Brain Atlas (Wang et al., 2020) (Fig. 1; Fig. 3B-D′). In control animals, the number of YFP⁺ neurons follows the physiological distribution of subcortical layer V projections (Fig. 3B,B′), as previously reported (Polleux et al., 1997; Porrero et al., 2010; Shepherd, 2009). Strong staining in motor and somatosensory cortices resulted in bright signal expression in the cerebral peduncle (CP) and corticospinal tract (CST) (Fig. 3A).

To understand whether the spatial organization of YFP⁺ layer V cortical neurons was affected upon loss of cortical Nr2f1 gradient expression, we quantified and compared the number of YFP⁺ cells across mutant and control brains (Fig. 3E,F). As previous reports showed that the distribution of layer V neurons changes in the absence of Nr2f1 (Alfano et al., 2014; Armentano et al., 2007; Tomassy et al., 2010), we expected that the Thy1-YFP signal would also be altered, as a read-out of layer V changes. Indeed, the layer V gradient was disrupted in both mutant strains, and YFP⁺ cells were more homogenously distributed along the antero-posterior cortical axis (Fig. 3B-F). Although Emx1-cKO brains showed a significant increase in YFP⁺ cells in parietal and occipital regions, compared with controls, this was not the case for Nex-cKO brains, which instead revealed a decrease in only frontal areas (Fig. 3F). To better quantify the differences between corticopontine projections from frontal and parietal areas, where normally the motor and somatosensory areas develop (Leergaard et al., 2000a,b), we compared the ratio of YFP⁺ cells over the total cell populations counted in the frontal (motor) and parietal (somatosensory) areas (Fig. 3G). We found that the percentage of YFP⁺ cells was significantly decreased in frontal (motor) cortex and concomitantly increased in parietal (somatosensory) cortex in both Emx1-cKO and Nex-cKO compared with controls, but no differences were found between the two lines (Fig. 3E,G). This lack of a difference in the distribution of YFP⁺ layer V neurons between cortical regions normally containing the motor and somatosensory areas is in line with the acquisition of a motor-like identity in the somatosensory cortex of Nr2f1-deficient brains, as previously reported (Alfano et al., 2014). We next asked whether and how alterations in layer V organization observed in the two mutant lines would be translated into layer V corticospinal projections and/or corticopontine topographic mapping.

We hypothesized that the disordered cortical distribution of YFP-expressing layer V neurons in mutant mice might influence the integrity of subcortical axonal projections. In all cases, strong YFP signal expression was seen bilaterally in the main corticofugal pathways (Fig. 3A), visible as longitudinally oriented fibre bundles coursing towards the CP (Fig. 4A-D), passing dorsal to the pontine nuclei as the longitudinal fasciculus (Fig. 4B′-D′), and continuing through the brain stem towards the spinal cord as the CST. Because a large fraction of the corticobulbar fibres terminate in the pontine nuclei (Tomasch, 1968, 1969), we reasoned that abnormal distribution of YFP⁺ layer V neurons observed in mutant mice (Fig. 3) might affect corticopontine innervation, and could be reflected by an abnormal size of the pontine longitudinal fascicle as it enters the CP and exits the pons in rostral and caudal positions to the pontine nuclei, respectively. To evaluate this, we measured the dorso-ventral width of the longitudinal fascicle of the pons in sequential sections along the medio-lateral axis. The measurements were taken at rostral and caudal levels to the pontine nuclei (Fig. 4A′). Surprisingly, we found the lateral part of the fascicle to be wider at both rostral and caudal levels in Nex-cKO mice compared with Emx1-cKO and controls, while being narrower medially (see red area chart in Fig. 4E,F), suggesting that the longitudinal fascicle of the pons is flattened and expanded laterally upon Nr2f1 inactivation in postmitotic neurons. Only minor differences were observed in the Emx1-cKO fascicles (blue area chart in Fig. 4E,F). This is also supported by quantification of the total surface of the longitudinal fascicle of the pons at rostral and caudal levels, which shows a significant surface area reduction at caudal levels, but only a tendency towards a reduced surface area at rostral levels, in Emx1-cKO mice (Fig. 4G,H). These data indicate that genetic inactivation of Nr2f1 in cortical progenitors, but not in neurons, results in fewer YFP⁺ fibres passing the pontine nuclei towards the brain stem to form the CST.

Moreover, we observed, caudal to the pontine nuclei, abnormally widespread fibre fascicles in the CST of mutant animals (arrowheads in Fig. 4C′,D′). To determine whether there was a significant difference between animal groups, we estimated the degree of fibre bundle fasciculation in the CST of Emx1-cKO and Nex-cKO mice. At locations of 250 µm and 500 µm caudal to the pontine nuclei (Fig. 4A′), we measured the total dorso-ventral width of the CST at several medio-lateral levels and subtracted the gaps between the YFP-expressing fibre bundles at the same levels. The ratio of the total width of the CST and fibres was used as a measure of the fasciculation index (Fig. 4I). Notably, in both groups of mutant mice, we found a lower degree of fasciculation in the CST that was more pronounced at the most caudal level (Fig. 4I). Together, these data show that loss of Nr2f1 expression affects the diameter, shape and degree of fasciculation of the CST originating from layer V neurons.

Next, we evaluated whether topographical organization of corticopontine projections depended on proper cortical area mapping and layer V distribution (Fig. 5). To achieve this, we assessed the spatial distribution of YFP signal expression within the pontine nuclei by comparing intensity-normalized microscopic images of spatially corresponding sagittal sections from the brains of Emx1-cKO, Nex-cKO and control littermate animals (Fig. 1B; Fig. 5A). A complete documentation of spatially comparable and reproducible images is provided in Figs S1 and S2. In control brains, we observed a strong YFP signal in central parts of the pontine nuclei, with the densest expression surrounding a centrally located zone exhibiting less-dense signals (Fig. 5B-E). This region of the pontine nuclei typically receives strong projections from somatosensory areas (green dots in Fig. 2G-K). Some signal expression was also visible in medial parts of the pontine nuclei (Fig. 5D,E), which receive projections primarily from cortical motor areas (purple dots in Fig. 2G-K). By contrast, signal expression was lower in rostral and lateral parts of the pontine nuclei (Fig. 5B,C,E), known to receive projections from visual and auditory areas of the cerebral cortex (Inoue et al., 1991; Leergaard and Bjaalie, 2007).

Interestingly, Emx1-cKO mice showed a relatively homogeneous signal distribution across all parts of the pontine nuclei, and displayed more expression in the dorso-lateral regions (Fig. 5F-I). Signal expression was also present in the medial part of the nuclei (Fig. 5G-I). This observation fits well with the finding of a higher number of YFP⁺ neurons in the occipital cortex (Fig. 3C,C′,F), which projects to the dorso-lateral pontine nuclei. By comparison, Nex-cKO animals showed more-constrained signal expression that predominated in rostrally and caudally located clusters extending from the cerebral peduncle towards the ventral surface of the pons, and medially surrounding a central core with low expression (Fig. 5J-M). These clusters were more peripherally located than those observed in control animals (Fig. 5J-M). Notably, little signal expression was seen in the central region of the pontine nuclei of all Nex-cKO cases (unfilled arrowheads in Fig. 5J-L), despite the presence of YFP⁺ layer V neurons in somatosensory cortex (Fig. 3D,D′,F,G). This central region is normally innervated by projections from the face representations located in S1 (Fig. 2A-F).

Taken together, these findings show that corticopontine projections are abnormally distributed in Nr2f1-deficient mice, with more homogenously (non-specific) distributed expression in Emx1-cKO mice (schematically summarized in Fig. 5N,O), and more peripherally distributed signal expression in Nex-cKO mice, which display reduced expression in the central region of the pontine nuclei that normally receives somatosensory projections (schematically summarized in Fig. 5N,P; Figs S1 and S2). In both mutant groups, the signal expression was expanded to dorso-lateral regions of the pontine nuclei that normally are innervated by projections from occipital cortical areas. This suggests that cortical Nr2f1 graded expression in postmitotic neurons might be directly involved in the establishment of topographically organized corticopontine projections.

To further support that Nr2f1 gradient cortical expression might be directly involved in topographical pontine mapping, we focused on the Nex-cKO genetic model to unveil the cortical origin of the innervation defect. Compared with the Emx1-cKO mice, in which projections are more disorganized and abundant, the Nex-cKO mouse model exhibits externally shifted projections, more suitable for investigating changes in topographical mapping using experimental tract tracing techniques. Therefore, we injected the AAV9-CAGtdTomato anterograde viral tracer (Pourchet et al., 2021) in the cortex of 5-day-old (P5) Nex-cKO mice and littermate controls, in frontal and parietal locations corresponding to the motor or S1 areas in control mice, respectively (Figs 6 and 7). The mice were then sacrificed at P21 and brain sections analysed microscopically (Fig. 1B). All histological sections were spatially registered to the Allen Mouse Brain Atlas (common coordinate framework, CCF3; Wang et al., 2020), and the location of tracer injections sites were mapped in the same atlas space (Figs 6A and 7A). For each injection site location in a Nex-cKO brain, we selected the most corresponding control experiment or wild-type tract-tracing data from the Allen Mouse Brain Connectivity Atlas as additional controls. Representative examples of injections and comparisons between controls and littermate Nex-cKO brains are shown in Figs 6B-C′ and 7B-C′. Complete documentation of all tracing experiments cases is provided in Figs S3-S5.

In controls, the spatial distributions of corticopontine projections (dark grey point clouds in Figs 6 and 7) were highly comparable with the labelling patterns seen in corresponding wild-type tracing data from the Allen Mouse Brain Connectivity Atlas (Wang et al., 2020). As expected, tracer injections into areas corresponding to the motor cortex in control mice gave rise to labelled axonal clusters located rostrally, caudally and medially in the pontine nuclei (purple dots in Figs 2G-K and 6D-F; Figs S3 and S4). In all Nex-cKO mice receiving tracer injections into areas normally corresponding to the motor cortex, the overall distribution of corticopontine labelling was found to be essentially similar to that observed in control cases (compare grey point with red point clouds in Fig. 6D-F; Figs S3-S5).

Tracer injections into S1 areas of control animals gave rise to labelled axonal clusters located centrally or caudally in the pontine nuclei (grey point clouds in Fig. 7D-I). The spatial distribution of labelling varied systematically as a function of the location of the cortical injection sites, such that (1) the face representation in the lateral part of S1 projects centrally in the pontine nuclei (Fig. 7G,H; Fig. S3I,J), (2) the whisker representations lateral and more posterior in S1 project to regions surrounding the central core (Fig. 7F; Fig. S3K), while (3) the more medially located forelimb and hindlimb representations of S1 projected to medio-laterally oriented regions in the caudal part of the pontine nuclei (Fig. 7D,E; Fig. S3G,H). By contrast, tracer injections in Nex-cKO brains, at locations corresponding to control S1 representations of the whiskers or upper limb, or into the S1/M1 (sensorimotor) lower limb representation, gave rise to abnormal distribution of corticopontine fibres (Fig. 7D-G). Specifically, corticopontine projections from medial parts of parietal cortex (corresponding to the S1 forelimb and hindlimb representations in control mice) were shifted towards more rostral locations in Nex-cKO experiments (Fig. 7D-F, red points), resembling the distributions observed after tracer injections into motor areas in the control experiments (Fig. 6D-F).

Notably, tracer injections placed in the antero-lateral parietal cortex in Nex-cKO mice, in regions normally representing sensory surfaces of the head, gave rise to labelled axons distributed in the central part of the pontine nuclei, with more subtle difference to the matching control experiments (Fig. 7G). In two Nex-cKO cases, projections from regions corresponding to the S1 head representation were distinctly medially shifted, relative to control experiments (red dots in Fig. 7G,I), attaining a distribution resembling corticopontine projections from head representations in M1 cortex, located significantly more anteriorly in the cortex (grey dots in Fig. 7I). Interestingly, two tracer injections in similar locations, corresponding to the border between nose and whisker representations, gave rise to labelling located in central and rostral parts of the pontine nuclei (Fig. 7F,G, red dots, see also Fig. 7C′) with predominantly rostrally shifted fibres in one case (Fig. 7F,C′; Fig. S5J), and centrally located and medially shifted fibres in the other case (Fig. 7G; Fig. S5K). Despite the differences in density distributions, both cases represent a distinct shift of fibre distributions towards regions normally receiving motor projections. Taken together, our observations indicate that corticopontine projections from antero-lateral cortex in Nex-cKO mice display abnormal topographical distributions resembling the normal mice projections from homologous representations in the more anterior and medially located primary motor cortex of control mice. Finally, one tracer injection placed in the most antero-lateral part of S1 in a Nex-cKO mouse, at a location corresponding to the perioral surface representation in controls, yielded corticopontine labelling highly similar to that of a control experiment (Fig. 7H).

Our findings show that corticopontine projections from frontal (motor) areas and the most antero-laterally located parts of the parietal (S1) cortex are topographically similar in Nex-cKO and control brains, whereas corticopontine projections from most parts of the parietal cortex, where head, whisker, upper limb and lower limb representations of S1 are located in control mice, are abnormally shifted towards rostral and medial regions of the pontine nuclei, normally receiving projections from cortical motor areas (schematically summarized in Fig. 7J,K). This is in overall agreement with the spatial and temporal control of Nr2f1 in area mapping. Indeed, the absence of changes in corticopontine projections from the frontal (motor) cortex might be due to low Nr2f1 expression in this area (spatial control), whereas lack of changes in the projections originating from the most antero-lateral part of the parietal (S1) cortex, from which the earliest cortical projections to innervate the forming pontine nuclei originate, might be explained by the late Nr2f1 genetic inactivation occurring after the earliest layer V neurons have been produced (temporal control).

Our present study questions whether and how spatio-temporal cortical expression gradients are involved in the establishment of normal topographical organization of corticopontine projections. By combining genetically modified mice and public mouse brain connectivity data with tract-tracing techniques and digital brain atlas tools, we have provided novel evidence of an intrinsic molecular control of layer V cortical neurons during the establishment of topographical organization of corticopontine projections. Abnormal areal organization in the neocortex induced by Nr2f1 inactivation is reflected in altered corticopontine projections, as well as impaired structural integrity of the CST. Although loss of Nr2f1 from the early progenitor cell pool leads to increased and abnormal corticopontine innervation at the expense of corticospinal projections, only late postmitotic Nr2f1 inactivation reveals altered topographic pontine mapping from medially located parts of somatosensory cortex controlling whisker and limb representations. No shifts in projections from the earliest generated antero-lateral cortical areas were observed in these mice, in line with a spatial and temporal control of Nr2f1 expression, respectively. Overall, our data show that proper area mapping of the neocortical primordium is a pre-requisite for preserving the cortical spatial and temporal segregation within the pontine nuclei, and thus correct corticopontine topographic organization.

To ensure accurate 3D data in wild-type and genetically modified mice, we relied on spatial alignment of serial microscopic section images to a common reference atlas achieved through non-linear image registration method (Puchades et al., 2019). The use of non-linear registration compensated for minor shape differences among brains and allowed comparison of distribution patterns among spatially relevant data. The focus on the location rather than the amount of signal expression/axonal labelling also compensated for the variation in signal expression intensity and size of tracer injections among cases. By representing signal expression and axonal labelling as 3D point clouds, it became possible to directly explore and compare location and distribution patterns in 3D in different combinations of datasets. For the additional benchmark data extracted from the Allen Mouse Brain Connectivity Atlas (Wang et al., 2020), we used the same sagittal image orientation as in our microscopic data to facilitate comparison of microscopic images in addition to the 3D comparisons. The relevance and accuracy of the approach was confirmed by demonstrating that similarly located cortical tracer injections in control animals gave rise to similarly distributed labelling patterns in the pontine nuclei. Given the distinct patterns of topographical organization of corticopontine projections, the interpretation and comparison of spatial distribution patterns and variability in our tract tracing experiments critically depended on the analysis of tracer injection locations and on a 3D understanding of the topographical mapping of the cortical surface onto the pontine nuclei, which were derived from our analyses of control experiments and earlier studies of the rat corticopontine system (Leergaard and Bjaalie, 2007).

Our previous work showed overall areal organization impairments in cortical Nr2f1 mutant brains, whether Nr2f1 inactivation was carried out in progenitors or postmitotic neurons. Here, we report for the first time that Nr2f1 drives corticopontine connectivity differently in progenitors versus postmitotic neurons. While Nr2f1 expressed by progenitor cells modulates the ratio between corticopontine and corticospinal axonal projections, similarly to what happens in C. elegans with the ortholog UNC-55 (Petersen et al., 2011; Zhou and Walthall, 1998), postmitotic Nr2f1 expression specifically acts on somatosensory topographic organization of corticopontine neurons (Figs 4 and 5). This suggests that early Nr2f1 expression in progenitor cells is mainly required in the initial axonal pathfinding of layer V subtypes, while later postmitotic expression is more implicated in the later refinement of corticopontine topographical organization. Since YFP+ cells follow the physiological distribution of subcortical layer V projections (Porrero et al., 2010), Thy1-driven YFP fluorescence will be induced in cortical subpopulations related to the specificity of their axonal target region; a tight relationship which will not be altered in the conditional KO models. Accordingly, a higher number of layer V neurons in parietal Emx1-cKO cortices leads to disorganized corticopontine innervation, in accordance with increased Lmo4 expression, known to drive layer V neurons versus the pontine nuclei (Cederquist et al., 2013; Harb et al., 2016). By contrast, Nex-cKO parietal axonal projections reach pontine targets normally innervated by motor-derived cortical areas. Finally, increased number of YFP⁺ cells in visual and auditory areas in the occipital cortex, corresponds to an augmented innervation in dorso-lateral regions of pontine nuclei known to receive projections from the occipital cortex. Together, these data indicate a dual role for Nr2f1 in layer V corticofugal connectivity: an early role in subtype specification (corticopontine versus corticospinal) and a later role in topographical mapping.

Previous data in developing rats have shown that pontine neurons settle in the forming pontine nuclei in a shell-like fashion according to their birthdate, with early-born neurons forming the central core of the pontine nuclei, and later-born neurons consecutively settling around and forming concentric rings (Altman and Bayer, 1987). At early postnatal stages, corticopontine axons are chemotropically attracted as collateral branches from corticospinal axons (Heffner et al., 1990; O'Leary and Terashima, 1988), innervating the pontine nuclei in a topographic inside-out pattern (Leergaard et al., 1995). Neurons in the frontal (motor) cortex project rostrally and medially in the pontine nuclei, neurons in the parietal (somatosensory) cortex project to central and caudal parts, neurons in the temporal (auditory) cortex project to central and lateral regions, and neurons in the occipital (visual) cortex to lateral and rostral parts of the pontine nuclei (Leergaard and Bjaalie, 2007). This concentric organization of corticopontine projections suggests that the birthdate of pontine neurons and their inside-out genesis is linked to the spatial organization of cortical inputs. However, intrinsic differences in pontine neurons born at different times might also have an instructive role for corticopontine innervation. A recent study in mice showed that postmitotic expression of the HOX gene Hoxa5 guides pontine neurons to settle caudally within the pontine nuclei, where they are targeted by projections from limb representations in the somatosensory cortex (Maheshwari et al., 2020). Moreover, ectopic Hoxa5 expression in pontine neurons is sufficient to attract cortical somatosensory inputs, regardless of their spatial position, showing that pontine neurons can play an instructive and attractive role in topographic input connectivity of corticopontine neurons (Maheshwari et al., 2020).

Nevertheless, maturational gradients in the pontine nuclei cannot fully explain the complexity of the fine-grained somatotopic topographic connectivity pattern between cortical input and pontine neuron targets. As the establishment of topographic maps requires multiple processes and structures, it is conceivable that the position and specific intrinsic molecular programs of both presynaptic afferents and postsynaptic target neurons contribute to this complex corticopontine connectivity map. Indeed, our data show that, without affecting the development and maturation of pontine neurons, corticopontine Nr2f1-deficient layer V axons originating from the parietal S1 cortex will abnormally target the pontine region, a function normally assigned to corticopontine motor axons. By contrast, Nr2f1-deficient axons originating from the frontal and medial cortex will innervate the expected pontine region allocated to motor axons. This strongly suggests that, during the establishment of corticopontine topography, both structures – the neocortex and the pons – need to be properly pre-patterned by factors involved in spatial and temporal control of neurogenesis, such as Nr2f1 for the cortex and Hoxa5 for the pontine nuclei (model in Fig. 8).

By showing that gradient cortical expression of the transcriptional regulator Nr2f1 is directly involved in the establishment of corticopontine topographic mapping, we provide new insights into the development of neural systems. However, other factors regulating the size and location of cortical areas are also likely implicated. We conclude that distinct molecular mechanisms in the source (cerebral cortex) and target (pontine nuclei) regions must be coordinated during the establishment of corticopontine topography. Identifying the molecular pathways within the cortex and pontine nuclei, as well as the mechanisms and molecules governing their interaction remains an open question for further studies.

This study shows a limited number of viral injections per cortical area that was partially compensated by adapting mouse tract-tracing data from the Allen Brain Institute to our analysis. Even though both groups of samples showed a highly consistent pattern of topographical organization, the two types of data resulted from different experimental conditions. Because of some variability of the fluorescent signal among samples, we chose to focus on 3D spatial distribution patterns that turned out to be consistent across all control experiments. We found the Emx1-cKO mouse model less suitable for further topographical analysis due to the disorganized and chaotic innervation pattern observed in the Thy1-eYFP-H reporter background. Although more sophisticated methods are needed to pinpoint the cellular and molecular mechanisms involved in the establishment of corticofugal topography, our study represents a useful starting point and resource for further studies of the development of corticopontine and corticospinal projections in mice.

To establish a 3D benchmark map of corticopontine projections from somatosensory and motor areas in adult wild-type mice, we used a selection of public experimental tract-tracing data available from the Allen Institute mouse brain connectivity atlas (http://connectivity.brain-map.org/). We selected 11 experiments in which the anterograde tracer EGFP was injected in the right primary/secondary motor cortex (n=6) or primary somatosensory cortex (n=5) of wild-type C57BL/6J mice (Table S1). Serial two-photon fluorescence images were interactively inspected using the Projection High Resolution Image viewer of the Allen Institute and, from each case, five sagittally oriented images of the right pontine nuclei spaced at ∼100 µm were captured by screen shot from the largest 3D multiplane thumbnail viewer. The resolution of the captured images was up-sampled to three times the original size before their spatial alignment to the CCFv3 was optimized using the tools QuickNII (Puchades et al., 2019) and VisuAlign (RRID: SCR_017978), as described below. These images were used to create 3D representations of the axonal labelling in the pontine nuclei (Fig. 2; see below).

All mice used were bred in a C57BL/6J background. Male and female animals at any stage of development were used. All experiments were conducted in accordance with the French Animal Welfare Act and European guidelines for the use of experimental animals, using protocols approved by the French Ministry of Education, Research and Innovation and the local ethics committee (CIEPAL NCE/2019–548, Nice) under authorization #15 349 and #15 350. Nr2f1 mice were crossed with Emx1-Cre-recombinase mice to inactivate Nr2f1 exclusively in cortical progenitors and their progeny (Armentano et al., 2007) or with Nex-Cre-recombinase mice to abolish Nr2f1 expression from postmitotic neurons (Alfano et al., 2014). Littermate Nr2f1 mice without the presence of the Cre-recombinase gene (Cre-negatives) were considered controls (Table S2). For postnatal day (P) 21 and adult topographic map analysis, Emx1-cKO and Nex-cKO animals were further crossed with Thy1-eYFP-H mice to specifically label layer V projection neurons, as previously reported (Harb et al., 2016; Porrero et al., 2010). Mice were genotyped as previously described (Alfano et al., 2014; Armentano et al., 2007; Harb et al., 2016). Control and mutant littermates were genotyped as Nr2f1^fl/fl:Thy1-eYFP-H^T/+ (control) and Nr2f1^fl/fl:Emx1-Cre:Thy1-eYFP-H^T/+ (mutant) or Nr2f1^fl/fl:Nex-Cre:Thy1-eYFP-H^T/+ (mutant). For simplicity, mutant mice are named Emx1-cKO and Nex-cKO throughout the text. Midday of the day of the observed vaginal plug was considered as embryonic day (E) 0.5.

P4-P5 animals were anesthetized on ice for 5 min and kept on ice during the whole procedure. Viral particles were produced from the AAV9-CAGtdTomato plasmid by Alexis Bemelmans (CEA, France) and diluted 1:50 in TE-Buffer (Qiagen, 1018499) to a final concentration of 1.75e12 vg/ml (kindly donated by I. Dusart, Pierre and Marie Curie University, Paris, France). Approximately 0.5/1 μl was injected unilaterally in different rostral-caudal and medio-lateral brain locations of control and Nex-cKO pups, as previously described (Gu et al., 2017).

Mosaic microscopic images were acquired using an Axio Imager M2 epifluorescence microscope (Zeiss) equipped with a halogen lamp, a MCU 2008 motorized stage and an EC Plan-Neofluar 10×/0.30 (Figs 3A-D,B′-D′, 4B-D,B′-D′, 5B-L, 6B′,C′, 7B′-C′; Figs S1, S2; S3 columns 3 and 4, S4 columns 3 and 4) or an EC Plan-Neofluar 2.5×/0.075 (Figs 6B-C, 7B-C; Figs S3 column 2, S4 column 2) and an AxioCam MRm camera. ZEN blue software was used for imaging and automatic stitching. Images were exported in TIFF format and serially ordered from lateral to medial, rotated and, if needed, mirrored to consistent anatomical orientation using Adobe Photoshop CS6 (RRID: SCR_014199), before being converted to PNG format and resized to 60% of original size using ImageJ (RRID: SCR_003070) with bilinear interpolation. The resized serial images were loaded into Adobe Photoshop as a stack, spatially aligned using the ventral surfaces of the pons and cerebral peduncle as landmarks, and cropped and exported as individual PNG files. For comparative analyses of topographical organization (see below), variations in YFP signal expression intensity within and between groups were normalized by adjusting the brightness and contrast of images to equal levels using a custom-made histogram matching script available for ImageJ (National Institutes of Health; https://imagej.nih.gov/). One selected, a representative case (Experiment 5, Cre-negative, nr: 14250, Table S2) was used as reference.

Serial sectional images were spatially registered to the Allen Mouse Common Coordinate Framework, version 3, 2017 edition of the delineations (CCFv3; Wang et al., 2020) using the QuickNII software tool (RRID:SCR_016854; Puchades et al., 2019). Multiple anatomical landmarks (hippocampus, caudate-putamen, inferior and superior colliculus, and the external surface of the neocortex) were used to determine the medio-lateral position and orientation of the sagittal section images. For each section image, custom atlas diagrams were aligned to anatomical landmarks in the experimental images using affine transformations, with emphasis on matching the ventral surface of the pons and white matter tracts close to the pontine nuclei and exported as PNG images. As a secondary step, to further optimize registration, the custom atlas images were non-linearly transformed using the software tool VisuAlign v0.8 (RRID:SCR_017978), with particular focus on fitting the template to the outer brain surface, subcortical white matter and the outer boundaries of the pontine nuclei. To co-display images and the spatially registered custom atlas images, we used the software tool LocaliZoom, which is embedded in the Navigator3 image management system (bit.ly/navigator3), developed and hosted by the Neural Systems Laboratory at the University of Oslo, Norway.

Serial section images from Nex-cKO and Emx1-cKO mutants co-registered to the Allen Mouse Brain Connectivity Atlas were analysed using an ImageJ macro allowing automatic counting of spots by area of interest in the brain. The spots are considered as specific staining using a threshold based on intensity and shape of the elements, and the composite RGB segmentation atlas is used as a mask for the region of interest. The macro pre-processes an atlas plate to delineate the regions of interest (ROI), based upon their unique colour-code and corrects the corresponding raw image by subtracting the background, generating binary images of the signal and combining these to the ROI maps. Finally, all objects within circularity range of 0.5-1 were counted per ROI using the ‘find maxima’ tool. The process, reiterated for each atlas plate-raw image combination, produced a summary table containing the quantification of particles per area per section, shown as graphs in Fig. 3.

Serial section images from Nex-cKO and Emx1-cKO mutants were analysed using the Fiji-ImageJ Software tool (Schindelin et al., 2015) to determine the total dorsoventral width of the bundle expressing fluorescent signal in the descending fibre tract in different positions: rostrally and caudally to the pontine nuclei; and 250 µm and 500 µm caudal to the nuclei. The width of separate fibre fascicles was also measured 250 µm and 500 µm from the terminal edge of the pontine nuclei (Fig. 4A′).

Serial section images of cortical tracer injections in Nex-cKO brains (Table S3) and experiments taken from the Allen Mouse Brain Connectivity Atlas, were spatially aligned using QuickNII and VisuAlign, as described above. The centre positions of the injection sites were annotated as a point-coordinate using LocaliZoom and co-displayed with the CCFv3 atlas in the 3D viewer tool MeshView (RRID:SCR_017222). These visualizations were used to select spatially corresponding injection site locations for analyses of spatial distribution of corticopontine projections.

At P21 and adulthood, animals were anesthetized by intraperitoneal injection of a mixture of Tiletamine-Zolazepam-Xylazine-Buprenorphine and intracardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA) in PBS. Specifically, 15 ml of PBS was used for both P21 and adult animals, followed by 20 or 30 ml of PFA for P21 and adults, respectively, and according to the average body weight of the animals. Brains were removed from the skull and postfixed for 4 h at 4°C in 4% PFA, before being vibratome sagittally sectioned at 100 μm (adult samples) or 150 μm (P21 samples). All sections were incubated overnight at 4°C in a solution of 0.5% Triton X-100, 3% BSA, 10% goat serum in PBS, for permeabilization and reduction of non-specific binding of antibodies. For immunofluorescence (IF), sections were incubated for 2 days at 4°C with primary antibodies in a solution of 0.5% Triton X-100, 10% goat serum in PBS and then overnight at 4°C with relative secondary antibodies and Hoechst diluted in PBS. For the complete list of primary and secondary antibodies, see Table S5. Sections were washed several times in PBS, then transferred on Superfrost plus slides (ThermoScientific), covered and dried for 30 min to 1 h, and finally mounted with the Mowiol (Sigma-Aldrich) mounting medium.

To investigate and compare the 3D distributions of YFP signal expression or anterograde axonal labelling within the pontine nuclei, we used the annotation functionality in the LocaliZoom tool to semi-quantitatively record YFP signal expression or labelled axons in all sections through the pontine nuclei as point coordinates (specified in the coordinate system of the reference atlas, CCFv3), reflecting the overall density of signal/labelling observed in the images (Fig. 2C,D). To compensate for the spacing between sections and allow inspection of point distributions perpendicularly to the section angle, the z-coordinate of each point was randomly displaced within the thickness of the gap between sections using a custom Python script. The point coordinates were co-displayed in the MeshView 3D viewer tool (Figs 2, 5, 6 and 7).

We thank Isabelle Dusart in Paris for providing us with AAV9-CAG-tdTomato viral particles produced by Alexis Bemelmans (CEA, France); Nicolaas Groeneboom and Gergely Csucs for expert technical assistance; Baptiste Monterroso and Sameh Ben Aicha of the iBV PRISM microscopy platform for the set-up and optimization of automatic image analysis tools used in this work; and Sharon Yates and Maja A. Puchades for useful discussions.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200026.