Neurogenesis order of V1 clades assayed by EdU birthdating.

A, Experimental design. Timed pregnant En1cre-Ai9:tdT females were EdU injected at one of seven time points between E9.5 and E12.5, and the spinal cords harvested at P5. Tissue sections were processed for EdU (Click-iT™) and immunostained for representative transcription factors (TFs) of major V1 clades. MafB was used to identify the MafA-Renshaw cell clade by location (“ventral MafB-V1 cells”). B, EdU labeling at different embryonic times. Spinal neurons are born in a ventrolateral to dorsomedial sequence. C, Example of E11 EdU labeling in an En1-tdT spinal cord. EdU integrated in the DNA at the time of injection is diluted with subsequent cell divisions. To ensure we sampled V1 cells that incorporated EdU during S-phase after their final division, we only counted V1 cells with nuclei filled by EdU Click-iT™ reaction (arrows). D, Percentages of V1 cells labeled with “strong” EdU in each mouse. The x-axis indicates individual animals (“<litter number>.<animal number>”). The percentage of EdU-labeled V1 interneurons at each time point was consistent, although we also noted variability between litters and among animals within a litter. One animal (459.2) showed the wrong EdU pattern for its injection age and was discarded (indicated by an X). E, EdU birthdating reveals a peak in V1 neurogenesis around E11 (error bars = SD; each dot represents one animal). At the time points flanking the peak there is a larger amount of variability, suggesting a fast-changing pace in V1 neurogenesis. F, Representative images of TF antibody staining combined with EdU labeling to determine birthdates of defined V1 clades. The time points represented were selected according to the maximal or near-maximal generation of V1 interneurons in each clade. G-I, V1-clade neurogenesis quantification. Graphs represent the average ±SEM calculated from n = 3.9 ±0.3 mice per TF/date (not all TFs were tested in all mice). For each mouse average, we analyzed four ventral horns in Lumbar 4 or 5 segments. G, Percentage of V1s expressing each clade-specific TF labeled with EdU at each embryonic time point. Ventral MafB-V1s (Renshaw cells) and Pou6f2-V1s are mostly born before E11 (dorsal MafB-V1s are a subgroup of Pou6f2-V1s). Foxp2-V1 and Sp8-V1 interneurons have wider windows of neurogenesis, but most are generated after E11. H, Data normalized to the maximum percentage of V1s born in each group showing peak generation in each clade. Ventral MafB-V1s, Pou6f2-V1s, Foxp2-V1s and Sp8-V1s have progressively later times of peak generation. I, Cum-sum graphs of V1-clades neurogenesis. Between 50% to 68% of all neurons in each V1 clade are labeled across all ages. By E11 nearly all neurons in early-clades are generated, while fewer than half of neurons in late-clades are.

V1 interneurons born at different embryonic times settle in different locations.

Top row in each panel: blue open dots indicate all V1 interneurons positive for each TF at P5, and filled red dots indicate those with strong EdU. Each plot is from one representative animal injected at each of the indicated embryonic times (four ventral horns from each animal are superimposed in the diagram). Bottom rows: cellular density profiles of V1 interneurons positive for each TF and with strong EdU. Cellular density profiles represent all cells sampled in all animals EdU-injected at the indicated embryonic times; n indicates the number of animals in each plot (4 ventral horns analyzed per animal). Contour plots are derived from 2D kernel density estimates of interneuron positions; lines encompass 10% increments. Only timepoints with representative numbers of TF+/EdU+ cells are shown. Blue arrows highlight major directional change in settling locations for V1s born at each time point with respect to the previous time point. A, MafB-V1s are divided into two groups: ventral Renshaw cell MafB-V1s and dorsal MafB-Pou6f2-V1s with different birthdates. B, Pou6f2-V1s (taken as a whole) are born between E10 and E11.5 and always settle dorsally. C, Foxp2-V1s show variations in location according to birthdate. Cells of increasingly older birthdates settle dorsally, laterally, ventro-laterally and then ventro-medially. D, A few Sp8-V1s are born early and are located ventrally, but the majority are born later and settle dorsally. Directional changes in settling positions according to birthdate differ among V1 clades.

Genetic mouse models to label the Foxp2-V1 lineage.

A, P5 mouse intersection of En1::Cre and Foxp2::Flpo with two reporter alleles (Ai9-tdTomato and RCE:dualEGFP). Foxp2-V1s express EGFP (green) and non-Foxp2-V1s tdTomato (red). A few “yellow cells” correspond with Foxp2-V1 neurons that failed to remove the tdTomato Ai9 reporter. B, Density contours demonstrate high spatial overlap between V1 neurons expressing Foxp2 (EGFP+) or not (tdT+ and EGFP-) (n=6 ventral horns). C, Expression of lineage labels are stable throughout postnatal development, suggesting no additional foxp2 gene upregulation in V1s after P0 (n=6 ventral horns per age in one animal, except for P5 in which 6 animals and 36 ventral horns are included; error bars show SEM). D, Lineage labeling in P15 mice is uniform across all spinal cord segments from thoracic 13 to sacral 1 (n = 6 ventral horns in each segment from 2 mice; error bars show SEM). E, Foxp2-V1 (EGFP) and non-Foxp2 V1 (tdT) lineage labeling with antibody staining for the transcription factors defining the four major V1 clades: Foxp2, Pou6f2, MafB, and Sp8. The Foxp2-V1 lineage contains all P5 Foxp2-expressing V1s and excludes almost all those expressing the markers of other clades. F, Percentages of Foxp2-V1s (EGFP) expressing the clade markers at P5. Around half of the Foxp2-V1s maintain expression of the Foxp2 protein at P5, and a minimal number of these cells express transcription factors definitive of other V1 clades. G, Percentages of non-Foxp2 lineage V1s (tdT) expressing the different V1 clade markers. Cells outside of the Foxp2-V1 lineage do not express Foxp2 at P5, and this subset contain V1s from the three other clades (for both plots n = 6.5 ± 2.6 mice, 4 ventral horns each; error bars show SD). H, FLTG reporter mice reveal lineage-labeled non-V1 Foxp2 cells in the spinal cord. EGFP is expressed in Foxp2-V1s and tdTomato in non-V1 Foxp2 cells. The zoomed images of the highlighted region with and without NeuN-IR demonstrate that non-V1 Foxp2 cells include non-neuronal cell types with astrocyte morphologies. Only neurons (NeuN-IR) are included in the cell density contour plots. non-V1 Foxp2 neurons are located in the medial ventral horn and a few in the deep dorsal horn (n = 2 animals at P5, 6 ventral horns in each). I, Most Foxp2-neurons (red, non-V1s or green, V1s) are in the ventral horn and their number in 30 μm thick L4-5 sections decreases with age as the neuropil matures and expands in size. Dorsal horn Foxp2-neurons maintain their numbers despite the growth of the spinal cord, suggesting de novo postnatal Foxp2 expression in this population. Each point is one animal analyzed through 6 ventral horns (errors bars are SD). J, Percentage of Foxp2-lineage-labeled cells that are Foxp2-V1s remains constant throughout postnatal development (n=2 mice, 6 ventral horns each; error bars show SD).

Foxp2-V1 interneurons are closely associated with shifting motor columns throughout thoracic, lumbar and sacral levels of the spinal cord.

A, Foxp2-V1 lineage labeling and ChAT antibody staining for motoneuron identification in adult mouse spinal cords from thoracic to sacral levels. Foxp2-V1 neurons accumulate at the lateral edge of the ventral spinal cord but their locations shift when the LMC expands from L3 to L6. In addition, a distinct group of Foxp2-V1 interneurons is dispersed at ventro-medial locations in lumbar segments. B, Plots of Foxp2-V1 and ChAT-IR motoneuron cell body positions in x,y coordinates with 0,0 at the top of the central canal (n=4 ventral horns, 1 representative animal). C, Contour plots of kernel Foxp2-V1 cell density estimations. The highest density of Foxp2-V1 neurons cluster close to LMC motoneurons from L2 to L5 (contours enclose 10% increments, closer lines indicate steeper changes in density). Motoneuron numbers progressively increase from Th13 to L5 and drop in number in L6 and S1. D, Number of Foxp2-V1s per 50 μm thick section (ventral horn) significantly increases in lower lumbar segments from L3 to L6 compared to S1 (dots represent individual mice; n = 3-5 mice in different segments, each mouse estimate is from 6 ventral horns; bars show SEM). One-way ANOVA, post-hoc Bonferroni corrected t-tests are summarized in the right-hand table. See Table S1 for details). E, Ratios of Foxp2-V1 neurons to MNs remain constant at roughly 2.5:1 with no significant changes throughout the lumbar cord. Significance was only found for L3-L5 compared to S1 (*p<0.05; **p<0.01; post-hoc Bonferroni tests). High variability in S1 is likely due to the sharp rostro-caudal decrease in motoneuron numbers in S1.

Limb and axial motoneurons are densely innervated by Foxp2-V1s and Renshaw cells.

A, Motor column identification from lower thoracic to upper sacral spinal cord in P20 mice following labeling with ChAT antibodies: PGC = preganglionic cell column; MMC = medial motor column; HMC = hypaxial motor column; LMCd/v = lateral motor column (dorsal/ventral). B, Schematic representation of the rostro-caudal span of each motor column in the spinal segments studied. C, Synapse quantification. Axons of Foxp2 and non-Foxp2 V1 interneurons were respectively labeled with EGFP and tdT in en1Cre::foxp2flpo::R26 RCE:dualGFP/Ai9tdT mice. In en1Cre::R26 Ai9tdT mice we identified V1-Renshaw cell axons using calbindin antibodies. Synaptic locations were labeled with VGAT antibodies and the postsynaptic motoneurons with ChAT antibodies. Synapse densities were analyzed in a ribbon of membrane at mid-cell body level (7 optical planes, 1 μm z-step). C1, Single optical plane of a L4/5 LMCv motoneuron surrounded by genetically labeled Foxp2-V1 and non-Foxp2-V1 axons. Inhibitory synapses on ChAT-IR motoneurons are VGAT+. C2, Single optical image of a L4/5 LMCv ChAT-IR motoneuron receiving synapses from V1 Renshaw cells (genetically labeled V1 axons with calbindin-IR and VGAT). C3-4, Method for estimating synapse densities on motoneuron cell bodies using C2 as example. C3, V1-VGAT (red arrowheads) and V1-CB-VGAT synapses (yellow arrowhead) are marked (VGAT-IR is not shown for clarity), and the cell body contour annotated with regions corresponding to dendrite exits. This process was repeated in 7 consecutive mid-cell body optical planes (cross-sections with well-defined nucleus and nucleolus). C4, A membrane surface slab is reconstructed in 3D (two different rotations shown). The surface area corresponding to dendrite exits is subtracted from the total surface area of the slab to calculate the available surface area on the motoneuron cell body. V1-VGAT synapses (red), V1-CB-VGAT synapses (yellow), and CB-VGAT synapses (green) are marked. A similar process was followed for calculating Foxp2-V1 synapse density. D, Quantification of total V1-VGAT synapse densities on motoneuron cell bodies in different motor columns (n = 21-30 motoneurons per motor column, n = 5 animals with 4-9 motoneurons per animal per motor column). Each data point is one motoneuron color coded by mouse origin. Average synaptic densities ±SD indicated to the right of scatter plots. A nested ANOVA found significant differences among motor column/segments (p < 0.0001) with no inter-animal variability (p = 0.4768). The table summarizes all post-hoc pairwise comparisons for average V1 synaptic densities of each motor column and segment (Bonferroni corrected t-tests) (Statistical details are in Table S2). Colors indicate increased (>1, red) or decreased ratios (<1, blue) of column motoneurons vs row motoneurons. PGC neurons receive significantly fewer V1 synapses than MMC or LMC motoneurons. The LMC (ventral and dorsal) in lower lumbar (L4/L5) had significantly more V1 contacts than MMC motoneurons or L6 dorsal LMC. E, Comparison of synaptic densities from Foxp2-V1 and non-Foxp2-V1 neurons (top) or Renshaw cells (bottom). All motoneurons sampled in 2 to 3 animals for each comparison were pooled together. Densities of V1-VGAT synapses from Foxp2-V1s, non-Foxp2 V1s, or calbindin (CB)+ V1s (Renshaw cells) (n = 6-17 motoneurons sampled per motor column/segments, average = 12.1 ±2.9 SD) were compared using a two-way ANOVA for axon type vs motor column and segment. Foxp2-V1 vs non-Foxp2-V1 synapses: significant differences in density were found for type of synapse (p=0.001), motor column location (p<0.0001), and their interaction (p<0.0001). Significant differences after post-hoc Bonferroni tests are indicated (*p<0.05; ****p<0.0001). In general, synapses from Foxp2-V1 axons have higher density than non-Foxp2-V1 axons on HMC and LMC columns at all spinal segments except for L1/L2 LMC. MMC motoneurons receive similar synaptic densities from both types of V1 axons, except at the sacral level in which non-Foxp2 V1 synapses predominate. PGC neurons receive very low densities of V1 axons and there are no significant differences between either type in any region. Foxp2-V1 vs CB+V1 synapses: significant density differences were found for type of synapse (p<0.0001), motor column location (p<0.0001), and their interaction (p<0.0001). Significant differences between Foxp2-V1 and CB+V1 synapses after post-hoc Bonferroni tests are indicated (*p<0.05; ****p<0.0001). Synapses from Foxp2-V1 axons have higher density than CB+V1 axons in HMC and LMC columns at all spinal segments except for L1/L2 LMC. MMC motoneurons receive similar synaptic densities from both types of V1 axons in upper lumbar regions, but Foxp2-V1 synapse predominate in lower lumbar. In S1 the density of CB+/V1 synapses is significantly higher. The low synaptic densities estimated in PGC neurons for Foxp2-V1s and CB+ V1s are not significantly different. Details of all statistical comparisons are Supplementary tables S3 and S4. F, Comparing the numbers of Foxp2 and CB+ (Renshaw) V1 synapses to the total number of V1 synapses, we estimated their respective percentages. From these estimates we calculated that the remainder belongs to non-Foxp2 and non-CB+ Renshaw cells. The large majority of V1 synapses on the cell bodies of LMC, HMC and MMC motoneurons are either from Renshaw cells or Foxp2-V1s. Asterisks denote significant differences as found in E. G, Summary diagram of major V1 clade connectivity to motoneuron cell bodies. Foxp2-V1s and Renshaw cells form the majority of inhibitory V1 contacts on LMC and HMC motoneurons, with slightly higher density from Foxp2-V1s. The MMC receives roughly equal portions of V1 contacts from Foxp2-V1s and Renshaw cells. V1s provide only sparse inhibition on preganglionic sympathetic neurons and most originate in V1 clades other than Renshaw cells and Foxp2-V1s.

Subgroups of Foxp2-V1 interneurons defined by transcription factor expression at P5 and birthdate.

A, Otp (blue) and Foxp4 (white) expression in lineage-labeled Foxp2-V1s (EGFP, green) and non-Foxp2-V1s (tdT, red). The boxed area is shown at higher magnification with different color combinations for clarity. It shows that Foxp4-IR cells in the Foxp2-V1 population always expressed Otp (arrows). Quantification: 49.8-51.2% of Foxp2-V1s express Otp and 20.7-21.7% Foxp4 (n=2 mice each examined in 3 ventral horns in L4/5). Very few non-Foxp2-V1 cells express either TF (Otp: 1.3-4.2% and Foxp2: 0.5-2.9%). B, Images of P5 spinal cords containing Foxp2-V1 lineage labeling (EGFP, omitted for clarity) and double immunolabeled for Otp/Foxp2, Foxp4/Foxp2 and Otp/Foxp4. C, Quantification of Foxp2-V1 interneurons with different combinations of TF expression at P5. For each combination, the left panels show cell distributions, and the right graphs the percentage of Foxp2-V1s with each combination (n=3 mice each analyzed in 6 ventral horns). The data was analyzed with one-way ANOVAS followed by Bonferroni corrected pair-wise comparisons (Statistical details in Supplemental Tables S5, S6 and S7. ****p<0.0001; ***p<0.001; **p<0.01; *p<0.05). There are five main groups defined by TF expression patterns: Otp-Foxp2 (44%), Otp-Foxp2-Foxp4 (23%), Otp only (15%), Foxp2 only (9%) and no TF labeling (32% by subtracting all other groups). Some groups associate with specific locations: Otp-Foxp2 cells and the smaller proportion of Otp only cells are located laterally; Otp-Foxp4-Foxp2 cells are located latero-ventrally; medial cells either contain only Foxp2 or nothing; some dorsal cells are either Foxp2 only or do not express any of these TFs. D, Example image of EdU birthdating in en1Cre::foxp2flpo::R26 RCE:dualGFP/Ai9tdT dual-color mice pulse labeled at E11. Foxp2-V1s (green and yellow cells) are born between E10 and E12, with peak birthdate around E11 and after non-Foxp2 V1s (tdTomato only). The lower graph’s data is normalized to highlight the time of peak neurogenesis for each population (n=15.0 ± 4.1 ventral horns from one mouse per time point from E10 to E11 and 2 mice per time from E11.5 to E12.5; total 9 mice, bars show SEM). There is no difference between green (EGFP only) and the smaller population of yellow cells (EGFP and tdTomato). E, EdU labeling in the Foxp2-V1 dual-color genetic model combined with Otp and Foxp4 antibody staining. F, V1s expressing Foxp2, Otp, and Foxp4 at P5 are mostly born between E10.5 and E11.5 with neurogenesis time courses largely overlapping. The normalized plot indicates that peak neurogenesis for all three populations occurs at E11.5, although a marginally higher number of Otp V1 cells are born earlier (n=4 ventral horns from 3.27 ±1.34 mice per timepoint; error bars show SEM). Contour plots to the right show settling locations of Otp and Foxp4-IR populations born at each time point. G, Schematic of the L4-5 ventral horn summarizing Foxp2-V1 subgroups according to location and combinatorial expression of Otp, Foxp4, and Foxp2 at P5. LMCD: dorsal lateral motor column; LMCV: ventral lateral motor column; MMC: medial motor column.

Proprioceptive (VGLUT1+) inputs preferentially target lateral Foxp2-V1 interneurons.

A, Left, low magnification confocal image of Foxp2-V1 lineage-labeling (EGFP) combined with VGLUT1 antibody staining to identify primary proprioceptive afferent synapses. Center, the boxed area is expanded and shown with the magenta OTP-IR channel overlaying the GFP channel instead of merging with the green to maximize visual discrimination of OTP-IR Foxp2-V1s (denoted by asterisks). The overlay method used here is fully described in Methods, “Figure composition”. Right, the center image’s inlay is shown at high magnification to demonstrate variability of VGLUT1 synapse density on Foxp2-V1 interneurons (I = absent, II = medium or low, III = high). For simplicity and rigor, we classified Foxp2-V1 interneurons as receiving or not receiving VGLUT1 synapses. Asterisks here indicate which of these Foxp2-V1s are OTP-IR based on the lower magnification images. OTP-IR Foxp2-V1s are laterally biased in their positioning and tend to have higher VGLUT1 synapse densities. B, Distribution of Foxp2-V1 interneurons with and without VGLUT1 synapses (green and magenta, respectively). The positioning of Foxp2-V1 interneurons receiving VGLUT1 synapses is laterally biased. C, Percentages of Foxp2-V1s (GFP+ interneurons), OTP-IR Foxp2-V1s, and Foxp2-IR Foxp2-V1s receiving VGLUT1 synapses (both proximal and/or distal). Each dot is an animal estimate from 6 or 7 ventral horns with respectively 591 and 525 Foxp2-V1 interneurons sampled. Lines indicate the averages of both animals. D, Percentages of lineage labeled Foxp2-V1 interneurons receiving VGLUT1 synapses and with Otp expression (left graph) or Foxp2 (right graph). Each dot represents one mouse, and the lines indicate averages. The numbers of sections and genetically labeled Foxp2-V1 interneurons sampled in each mouse are as in C. In each mouse, this includes 237 and 236 Otp-IR cells and 256 and 225 Foxp2-IR cells.

Some Foxp2-V1 interneurons are reciprocal Ia inhibitory interneurons (IaINs).

A, Experimental design to label spinal neurons that receive inputs from tibialis anterior (TA) muscle primary afferents and connect to lateral gastrocnemius (LG) motoneurons, forming Ia reciprocal inhibitory connections from TA to LG. TA sensory afferents are labeled anterogradely with CTB followed by antibody detection of CTB and the presynaptic marker VGLUT1. Interneurons premotor to LG motoneurons are labeled by monosynaptic retrograde labeling with RVΔG-mCherry. B, Foxp2-V1 IaIN with the most TA/VGLUT1 contacts (31) in our sample (n = 5). Left, low magnification image of Foxp2-V1 interneurons (EGFP, green), RV-mCherry labeling (red) of LG muscle afferents in the dorsal horn and interneurons presynaptic to the LG and of TA afferents anterogradely labeled with CTB (white). The Foxp2-V1 interneuron contains mCherry (yellow cell, inside box). This cell is magnified in two panels to the right, one showing Foxp2-V1 and RV-mCherry and the other RV-mCherry and CTB labeling. Arrows in the zoomed image show examples of CTB synapses (confirmed with VGLUT1) on the dendrites of this neuron. Far right image is the 3D reconstructed cell (Neurolucida) with CTB/VGLUT1 synapses indicated on its dendrites by yellow stars. The axon initial trajectory is indicated (the axon is lost at the section cut surface). The blue area highlights lamina IX. C, Low magnification of section serial to B, showing TA-CTB afferents (white) and LG-RV-mCherry labeled interneurons (red). Transsynaptically labeled interneurons are categorized according to position and Foxp2-V1 lineage labeling: images below show superimposed Foxp2-V1 EGFP (green) and additional VGLUT1 immunolabeling (blue). The location of the LMC is indicated. Contralateral interneurons were found in LX (as the one in this section) and in LVIII (the other two in this animal, not shown). D, High magnification images of two LVII LG-coupled interneurons (RV-mCherry, red) receiving synapses from TA afferents (dendrites in boxed regions are shown at high magnification demonstrating CTB-TA labeling and VGLUT1 content). The most medial interneuron belongs to the Foxp2-V1 lineage (see inset with +EGFP). Insets shown VGLUT1 synapses with (arrows) and without CTB from the TA muscle. E, Neurolucida neuronal reconstructions showed that the Foxp2-V1 interneuron contained a medium number of TA/VGLUT1 synapses (13 contacts) in our sample of putative IaINs derived from the Foxp2-V1 lineage (n=5), while the non-Foxp2-V1 interneuron contained the largest (115) of any LVII reconstructed interneuron with mCherry, including many proximal synapses. For further examples of labeling in serial section from this animal, see Supplementary Figure 6.

Mouse models

Antibodies.