HoxD transcription factors define monosynaptic sensory-motor specificity in the developing spinal cord

Proprioceptive sensory pathways relay important information about the state of muscle contraction to motor neurons in the spinal cord (Arber, 2012; Catela et al., 2015; Imai and Yoshida, 2018; Zampieri and Nooij, 2021). This enables motor coordination through the precise harmonization and orchestration of limb movements (Akay et al., 2014; Mayer et al., 2018; Santuz et al., 2019; Akay, 2020). Proprioceptive sensory neurons, which are located in the dorsal root ganglion (DRG), innervate muscle spindles and Golgi tendon organs in the periphery. In the spinal cord, proprioceptive sensory neurons and motor neurons project axons to the same or synergistic muscles to form monosynaptic connections, whereas they do not form monosynaptic connections when they project to antagonistic muscles (Ladle et al., 2007; Mendelsohn et al., 2015; Imai and Yoshida, 2018; Balaskas et al., 2020). The specificity of these monosynaptic sensory-motor connections has been suggested to be encoded genetically through coordinated regulation of motor pool positioning, motor neuron dendrite patterning, and the angular approaches of proprioceptor axons to their motor neuron dendrite targets. (Frank, 1990; Mendelson and Frank, 1991; Vrieseling and Arber, 2006; Sürmeli et al., 2011; Mendelsohn et al., 2015; Baek et al., 2017; Balaskas et al., 2019). Loss of transcription factor FoxP1, for example, affects motor pool locations and results in inappropriate sensory-motor connections (Sürmeli et al., 2011). In addition, another study demonstrates a strong correlation between monosynaptic sensory-motor specificity and the angles of proprioceptive sensory axons and directionality of motor neuron dendrites (Balaskas et al., 2019).

Several molecules involved in the finer facets of sensory-motor specificity have also been identified. The repellent signaling molecule, semaphorin 3E (Sema3E) has been shown to repel proprioceptive sensory axons expressing its receptor, plexin D1 (PlxnD1), thereby regulating sensory-motor specificity (Pecho-Vrieseling et al., 2009; Fukuhara et al., 2013). Inactivation of either Sema3E or PlxnD1 causes defects in sensory-motor specificity without affecting either the positioning or dendritic development of motor neurons (Pecho-Vrieseling et al., 2009; Fukuhara et al., 2013). More recently, expression of Hoxc8 in proprioceptive sensory neurons has been shown to play a role in the monosynaptic sensory-motor specificity of particular forelimb sensory motor circuits, possibly by regulating gene expression in proprioceptive sensory neurons (Shin et al., 2020).

Transcription factors have well-documented roles in spinal cord development. The Hox family, for example, participates in regulating a wide range of developmental processes including rostro-caudal axis development, motor pool specification and guidance of skeletal and limb development (Dasen and Jessell, 2009; Jung et al., 2010; Lacombe et al., 2013; Philippidou and Dasen, 2013; Catela et al., 2015; Sweeney et al., 2018; Baek et al., 2019). Despite the various roles of Hox transcription factors in motor neuron patterning and specification, it remains unknown whether the expression of Hox transcription factors in motor neurons directly regulate monosynaptic sensory-motor specificity.

In this report, we examined whether HoxD transcription factors impact the development of monosynaptic sensory-motor specificity by focusing on the well-characterized quadriceps and obturator sensory-motor reflex circuit in the lumbar spinal cord (Mears and Frank, 1997). Hoxd9 and Hoxd10 are expressed in the lumbar spinal cord and locomotor deficits have been reported in Hoxd10 single- and Hoxd9, 10 double-mutant mice (Carpenter et al., 1997; de la Cruz et al., 1999; Tschopp et al., 2012). As there are many classes of Hox genes, the effects of deleting a single gene may be compensated for by other closely related members of the Hox family. To overcome this issue, Hoxd9, 10, and 11 triple mutant mice have been generated (hereafter, Hoxd9, 10, 11^−/−; Raines et al., 2013). These triple mutant mice are documented to have defects in reproductive tract development (Raines et al., 2013), but the effects of Hoxd9, Hoxd10, and Hoxd11 deletion on motor circuits and behavior are unknown. Thus, we tested the idea of compensatory functioning of multiple Hox proteins through an analysis of the quadriceps and obturator sensory-motor reflex circuit in Hoxd9, 10, 11^−/− mice.

To determine whether HoxD transcription factors are involved in the assembly of sensory-motor reflex circuits, we first examined the expression patterns of Hoxd9, Hoxd10 and Hoxd11 in the mouse spinal cord, DRG and muscles. Hoxd9 and Hoxd10 were expressed in the rostral lumbar spinal cord (Fig. 1A,B; Fig. S1) as previously reported (Choe et al., 2006; Wu et al., 2008). Hoxd11 was expressed in the caudal but not the rostral lumbar spinal cord (Fig. 1C; Fig. S1) (Davis and Capecchi, 1994; Tschopp et al., 2012). Expression of these three Hox genes was not detected in the DRG or the quadriceps muscles (Fig. 1A-F). Hoxd10 was expressed in both spinal interneurons and motor neurons, with strong expression in Islet1^off lateral motor column (LMC) neurons (Fig. 1G,H) (Choe et al., 2006; Wu et al., 2008).

To assess the roles of Hoxd9, Hoxd10 and Hoxd11 in motor neuron development at lumbar levels, we analyzed the preganglionic column (PGC), hypaxial motor column (HMC), median motor column (MMC) and LMC (Fig. 1I-N,S,T) in Hoxd9, 10, 11^−/− mice (Raines et al., 2013). Motor neurons from the PGC, HMC, MMC and LMC project their axons into sympathetic ganglia, hypaxial muscles, axial muscles and limb muscles, respectively. PGCs, identified by neuronal nitric oxide synthase (nNOS), were located in the thoracic spinal cord (Fig. 1I,Q), whereas HMC and MMC neurons were detected with Islet1 (Fig. 1I,K). Finally, LMC neurons at lumbar levels were identified by FoxP1 (Fig. 1M). Based on nNOS, Islet1 and FoxP1 immunohistochemistry, we observed no overt differences in the positioning or numbers of PGC, HMC/MMC or LMC motor neurons between control and Hoxd9, 10, 11^−/− mice (Fig. 1I-N,Q,R). To further characterize LMC, Pea3^on motor neurons were analyzed. Pea3^on motor neurons were located in lateral LMC and projected into rectus femoris muscles, which is one of quadriceps muscles (Fig. 1O) (Arber et al., 2000). Pea3^on motor neurons were similarly segregated both in control and Hoxd9, 10, 11^−/− mice (Fig. 1O,P). In addition, at lumbar levels, the number of Pea3^on motor neurons was similar between control and mutant embryos. (Fig. 1Q,R). We also assessed the expression of a motor neuron pool marker, Sema3e, which, at hindlimb levels, is expressed in gluteus motor neurons (Pecho-Vrieseling et al., 2009; Fukuhara et al., 2013), but we observed no differences in Sema3e expression between control and Hoxd9, 10, 11^−/− mice (Fig. 1S,T).

Taken together, these data indicate that Hoxd9-11 transcription factors are not necessary for general motor column or motor pool patterning within the thoracic and lumbar spinal cord during mouse development.

To determine whether Hoxd9-11 transcription factors influence monosynaptic sensory-motor specificity, we examined the motor neurons that innervate the quadriceps and obturator muscles as the monosynaptic sensory-motor specificity within this muscle pair has been well-characterized at lumbar levels (Mears and Frank, 1997). To assess cell body positions and dendritic morphologies of the quadriceps and obturator motor neurons, we employed a viral retrograde labeling approach. We injected AAV6-CAG-tdTomato into the quadriceps and obturator (gracilis) muscles at postnatal day (P) 3 mice and analyzed them at P10 (Fig. 2A-D). First, we examined cell body positions along the medio-lateral axis of the lumbar spinal cord. In P10 wild-type mice, quadriceps and obturator motor neurons were located laterally (60.5±6.8% of motor neurons were located 600-800 µm from the midline; mean±s.d.) and medially (72.2±14.8% of motor neurons were 0-600 µm from the midline), respectively (Fig. 2E,G). Similar distributions of motor neurons were observed in Hoxd9, 10, 11^−/− mice (Fig. 2F,H; quadriceps: 64.4±17.1% of motor neurons were 600-800 µm from the midline, P=0.32; obturator: 74.1±16.0% of motor neurons were 0-600 µm from the midline, P=0.85). Motor neuron cell body positions along the rostro-caudal axis were also similar between wild-type and mutant mice (Fig. 2I,J; distance from L1 for quadriceps motor neurons in wild type: 1555±207 µm, mutants: 1556±233 µm; distance from L1 for obturator motor neurons in wild-type: 1219±267 µm, mutants: 1185±321 µm). Thus, proper quadriceps and obturator motor neuron positions were maintained in Hoxd9, 10, 11^−/− mice.

As motor neuron dendrite morphology has been shown to affect synaptic specificity between proprioceptors and their appropriate spinal motor neuron targets (Vrieseling and Arber, 2006; Balaskas et al., 2019), we subsequently analyzed the dendritic arborization patterns of quadriceps and obturator motor neurons. Radial plot analyses of the densities of motor neuron dendrites or of dendrites and cell bodies combined showed similar arched dendrite distributions for quadriceps and obturator motor neurons in wild-type and Hoxd9, 10, 11^−/− mice (Fig. 2K-T). This strongly suggests that loss of Hoxd9-11 does not affect dendrite development or patterning in quadriceps and obturator motor neurons.

As previous studies have suggested that retrograde signals from muscles to motor neurons and proprioceptive sensory neurons influence sensory-motor circuit specificity (Haase et al., 2002; Vrieseling and Arber, 2006; Wang et al., 2007), we then examined the axonal projections of quadriceps and obturator motor neurons to their cognate muscle targets using Hb9-GFP reporter mice, in which GFP is expressed in motor neurons (Wichterle et al., 2002; De Marco Garcia and Jessell, 2008). At embryonic day (E) 15.5, motor nerves projecting to quadriceps (vasti and rectus femoris) muscles were thinner and less branched in Hoxd9, 10, 11^−/− mice compared with wild-type mice (Fig. 2U,V) (axon volume: wild-type, 100±26%; mutants, 40±16%; P<0.01, n=4, see Materials and Methods for details). In contrast, we detected no obvious differences in the obturator nerves between control and Hoxd9, 10, 11^−/− mice (Fig. 2W,X) (wild-type, 100±11%; mutants, 99±14%; P=0.87, n=4). These data indicate that Hoxd9-11 transcription factors are crucial for quadriceps, but not obturator motor axon projections and branches.

To determine whether there were any functional defects in the synaptic specificity of monosynaptic sensory-motor connections in Hoxd9, 10, 11^−/− mice, we performed intracellular recordings from both quadriceps and obturator motor neurons after peripheral nerve stimulation in P5-P7 mice (Fig. 3A) (Mears and Frank, 1997). We identified the quadriceps and obturator motor neurons by virtue of their antidromic responses to electrical stimuli applied to their respective muscle nerves. After detection of antidromic responses, motor responses to other stimulated muscle nerves were recorded. We found no aberrant antidromic responses in either wild-type or Hoxd9, 10, 11^−/− mice (quadriceps motor neurons, wild type=0/13, mutants=0/15; obturator motor neurons, wild type=0/8, mutants=0/13), indicating that quadriceps and obturator motor neurons project appropriately to their target muscles. To identify the latencies of monosynaptic sensory-motor connections, we analyzed the homonymous connections of obturator and quadriceps sensory-motor circuits in wild-type mice. Latencies of monosynaptic connections following obturator and quadriceps nerve stimulation ranged between 4.41 to 5.89 ms and 3.96 to 5.98 ms, respectively (red dotted lines in Fig. 3D,E; see Materials and Methods for details). To determine the specificity of monosynaptic sensory-motor connections, excitatory postsynaptic potentials (EPSPs) and shortest latencies were recorded from quadriceps motor neurons with obturator nerve stimulation (Fig. 3B-D). Intriguingly, we found aberrant monosynaptic connections between obturator sensory nerves and quadriceps motor neurons in Hoxd9, 10, 11^−/− mice (wild-type: 0/13, mutants: 6/15, Fig. 3D, right panel). We also analyzed obturator motor neurons with quadriceps nerve stimulation; however, we found no monosynaptic connections in either control or Hoxd9, 10, 11^−/− mice (wild-type: 0/8, mutants: 0/13, Fig. 3E, right panel). There were also no significant differences in homonymous EPSPs between wild-type and Hoxd9, 10, 11^−/− mice (Ob-Ob wild-type: 4.78±2.70 mV, mutants: 3.50±2.44 mV, P=0.289; Q-Q wild-type: 5.79±2.44 mV, mutants: 3.96±2.94 mV, P=0.083, Fig. 3F left, middle). Monosynaptic EPSPs from quadriceps motor neurons after obturator sensory nerve stimulation in Hoxd9, 10, 11^−/− mice were similar to homonymous quadriceps circuits in wild-type mice (4.63±2.96 mV, Fig. 3F right). Taken together, these results reveal that quadriceps motor neurons receive aberrant inputs from obturator sensory nerves in Hoxd9, 10, 11^−/− mice (Fig. 3G).

What is the mechanism underlying synaptic specificity regulated by HoxD9-11 transcription factors? Hoxd9, 10, 11^−/− mice showed no defects in motor neuron topography (Figs 2 and 3). Dendrite morphologies of quadriceps and obturator motor neurons in Hoxd9, 10, 11^−/− mice were also identical to control mice. However, we found defects in peripheral projections of quadriceps but not obturator motor neurons in Hoxd9, 10, 11^−/− mice, raising the possibility that retrograde signaling may be impaired between the quadriceps muscle and its innervating motor neurons. It is possible that the reduction in quadriceps motor axons in the quadriceps muscle may affect expression of signaling molecules or axon guidance molecules in the quadriceps motor neurons, and degrade monosynaptic sensory-motor specificity.

To assess whether Hoxd9, 10, 11^−/− mice have locomotor deficits, hindlimb joint angles were analyzed on a treadmill at 25 cm/s (Fig. 4A). Hoxd9, 10, 11^−/− mice showed foot-dragging behaviors (red arrows in Fig. 4C) similar to what has been previously reported in Hoxd9,d10 double-mutant mice (de la Cruz et al., 1999). In addition to altered toe trajectories, maximum and minimum joint angles of hindlimbs were also affected during locomotion in Hoxd9, 10, 11^−/− mice (Fig. 4B). Minimum joint angles of hips and ankles (hips: wild-type, 79.4±8.7°, mutants, 61.8±7.2°; ankles: wild-type, 58.9±5.9°, mutants, 96.9±4.7°) and maximum joint angles of the knees and the metatarsal phalangeal joints (MTP) (knees: wild-type, 113.7±9.2, mutants, 95.4±9.3°; MTP: wild-type, 163.6±6.2°, mutants, 152.0±4.5°) were also affected during locomotion in Hoxd9, 10, 11^−/− mice (Fig. 4B). No significant differences were observed in joint angles during the swing to stance transitions (Fig. 4D,E). Hip and knee angles were more flexed and ankle angles were extended during the transitions from stance to swing phases in Hoxd9, 10, 11^−/ mice (Fig. 4F,G). These results demonstrate that deletion of Hoxd9-11 caused changes in locomotor behaviors in terms of foot-dragging and altered joint angles during the transitions between the swing and stance phases of locomotion. Surprisingly, these locomotor defects in Hoxd9, 10, 11^−/− mice appear to be milder than those observed in Hoxd10 single- or Hoxd9, 10 double-mutant mice (Carpenter et al., 1997; de la Cruz et al., 1999). This may be due to the different gene knockout strategies employed. Indeed, Hoxd9, 10, 11^−/− mice do not show defects in hindlimb development, unlike Hoxd10 single- or Hoxd9, 10 double-mutant mice which show hindlimb skeletal abnormalities (Carpenter et al., 1997; de la Cruz et al., 1999; Raines et al., 2015). Moreover, as the behavioral deficits in Hoxd9, 10, 11^−/− mice do not seem to simply be explained by defects in quadriceps and obturator sensory-motor circuits or through aberrant development of quadriceps motor axons, it is plausible that other sensory-motor circuits could be affected in Hoxd9, 10, 11^−/− mice.

Although the precise mechanisms underlying Hoxd9-11-mediated sensory-motor specificity are not yet clear, this study has identified a group of transcription factors that can influence synaptic specificity in the developing monosynaptic sensory-motor circuit in the spinal cord. This study, along with previous studies (Fukuhara et al., 2013; Baek et al., 2017), suggests that transcriptional and repellent molecular programs contribute to the finer-grain aspects of sensory-motor specificity, and a delicate interaction between positional and molecular cues is likely necessary to achieve the highly complex neural circuit wiring seen in the mammalian central nervous system.

The following mouse lines were used in this study: Hoxd9, 10, 11^−/− (Raines et al., 2013) and Hb9-GFP (Wichterle et al., 2002). Mouse handling and procedures were approved by the Institutional Animal Care and Use Committee at the Cincinnati Children's Hospital Research Foundation and Burke Neurological Institute. We used wild-type litter mates as controls and both sexes were used for this study.

After shaving the adult mice (2-3 months old), joints (hip, ankle, MTP) and the tips of digits (toes) were marked and recorded with a high-speed video camera (Blackfly S, 226fps, FLIR). Treadmill speed (Harvard Apparatus) was set to 25 cm/s after a 5-min warm up at 15 cm/s. Joint markers were tracked with DeepLabCut (Mathis et al., 2018; Nath et al., 2019) and marker positions and joint angles were analyzed using R statistical software (https://www.R-project.org/). Knee angles were calculated by triangulation with hip and ankle coordinates, as well as the lengths of femurs and tibias.

To label motor neurons (Fig. 2), AAV6-CAG-tdTomato (4.5×10¹² vg/ml, 1 µl, UNC Vector Core) was injected into quadriceps or gracilis muscles at P3 and mice were sacrificed at P10.

Spinal cords, DRGs and the surrounding tissues were fixed with 4% paraformaldehyde (PFA)/phosphate buffer (PB) for 20 min (embryos) or overnight (postnatal mice) for immunostaining or in situ hybridization. Samples were then cryoprotected in 30% sucrose, embedded in OCT compound and sectioned with a cryostat (Leica VT1200) at a thickness of 16 µm. For analyses of motor neuron cell body positions and dendrite projections, spinal cords were embedded in 4% low-melting agarose then sectioned into 200 µm slices using a vibratome (Leica CM1860).

Digoxigenin (DIG)-labeled RNA probes were synthesized using a DIG labeling kit (Roche). In situ hybridization was performed according to standard protocols (Schaeren-Wiemers and Gerfin-Moser, 1993).

The following antibodies were used in this study: rabbit anti-Hoxd10 (ab76897, Abcam, 1/200), goat anti-Islet1 (AF1837, R&D Systems, 1/50), rabbit anti-nNOS (#24287, Immunostar, 1/1000), goat anti-FoxP1 (AF4534, R&D Systems, 1/100) and rabbit anti-Pea3 (kindly provided by Dr Silvia Arber, University of Basel, Switzerland, 1/20,000). Sections were stained with Alexa488 or Alexa568-conjugated secondary antibodies (A21206, A110042, A11055, A11057, Thermo Fisher Scientific, 1/500), then mounted with Vectashield (Vector Laboratories). To analyze motor axon projections in muscles, embryos were fixed and cleared using SeeDB or CUBIC (Ke et al., 2013; Kubota et al., 2017). Vibratome sections (200 µm) were treated with CUBIC-L (T3740, TCI America) at 37°C for 30 min, then sections were treated with 50% then 100% CUBIC-R+ (T3741, TCI America) for final clearing. Samples were scanned using Nikon A1R and Leica SP8 confocal microscopes.

Dissection of spinal cords and intracellular recordings were performed as previously described (Imai et al., 2016a,b). Briefly, after perfusion, spinal cords from P5-P7 pups were removed and hemi-sectioned in an oxygenated cerebrospinal fluid (aCSF) bath containing NaCl (127 mM), KCl (1.9 mM), KH₂PO₄ (1.2 mM), CaCl₂ (2 mM), MgSO₄ (1 mM), NaHCO₃ (26 mM) and D-glucose (20.5 mM). Using tightly-fitting glass pipettes, obturator and quadriceps nerves were stimulated independently (10 mA, S88X, SIU-C Grass Technologies). Intracellular potentials were recorded using glass micropipettes (90-180 MΩ) filled with 2 M potassium acetate with 0.5% Fast Green and 300 mM of lidocaine N-ethyl bromide for blocking antidromic potentials (Sigma-Aldrich). Synaptic potentials evoked by 1 Hz stimulation of either the quadriceps or obturator nerves were recorded 20-60 times and then averaged (MultiClamp 700B, Digidata 1440A, Clampex 10, Molecular Devices). Obturator and quadriceps motor neurons were identified by antidromic activation. Recordings were accepted only from neurons in which the resting membrane potential was below −40 mV (Mears and Frank, 1997). The time window for monosynaptic EPSP onset latencies was defined as the mean latency of wild-type homonymous EPSPs±2 s.d., as previously described (Vrieseling and Arber, 2006; Pecho-Vrieseling et al., 2009; Fukuhara et al., 2013).

For counting motor neurons (Fig. 1), every fourth section (thickness: 16 µm) was labeled, scanned and counted using ImageJ software. Motor neuron morphologies (Fig. 2) were analyzed with surface tools in the IMARIS software package (Bitplane). For motor neuron cell body density analyses (Fig. 2), bin sizes were set to 100 µm. The densities of motor neuron dendrite projections (Fig. 2) were evaluated in 30° bins using ImageJ software. For quantifications of quadriceps and obturator motor axons (Fig. 2), volumes were analyzed with surface tools in IMARIS software (quadriceps: wild-type, 3,106,551±800,904 µm³, mutants, 1,231,847±496,490 µm³; obturator: wild-type, 3,061,520±326,758 µm³, mutants, 3,024,534±421,486 µm³). For joint angle analyses (Fig. 4), swing/stance and stance/swing transition phases were manually defined, then analyzed, averaged (10-15 cycles) and compared using an unpaired two-tailed Student's t-test.

We are grateful to Silvia Arber (University of Basel) for providing us with the Pea3 antibody. We also thank David Ladle (Wright State University), Turgay Akay (Dalhousie University), Joriene De Nooij (Columbia University) and Niccolo Zampieri (Max-Delbrück-Center for Molecular Medicine Berlin-Buch) for critically reading this manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.191122