Abstract
Axon fasciculation is thought to be a critical step in neural circuit formation and function. Recent studies have revealed various molecular mechanisms that underlie axon fasciculation; however, the impacts of axon fasciculation, and its corollary, defasciculation, on neural circuit wiring remain unclear. Corticospinal (CS) neurons in the sensorimotor cortex project axons to the spinal cord to control skilled movements. In rodents, the axons remain tightly fasciculated in the brain and traverse the dorsal funiculus of the spinal cord. Here we show that plexinA1 (PlexA1) and plexinA3 (PlexA3) receptors are expressed by CS neurons, whereas their ligands, semaphorin-5A (Sema5A) and semaphorin-5B (Sema5B) are expressed in the medulla at the decussation site of CS axons to inhibit premature defasciculation of these axons. In the absence of Sema5A/5B-PlexA1/A3 signaling, some CS axons are prematurely defasciculated in the medulla of the brainstem, and those defasciculated CS axons aberrantly transverse in the spinal gray matter instead of the spinal dorsal funiculus. In the absence of Sema5A/Sema5B-PlexA1/A3 signaling, CS axons, which would normally innervate the lumbar spinal cord, are unbundled in the spinal gray matter, and prematurely innervate the cervical gray matter with reduced innervation of the lumbar gray matter. In both Sema5A/5B and PlexA1/A3 mutant mice (both sexes), stimulation of the hindlimb motor cortex aberrantly evokes robust forelimb muscle activation. Finally, Sema5A/5B and PlexA1/A3 mutant mice show deficits in skilled movements. These results suggest that proper fasciculation of CS axons is required for appropriate neural circuit wiring and ultimately affect the ability to perform skilled movements.
SIGNIFICANCE STATEMENT Axon fasciculation is believed to be essential for neural circuit formation and function. However, whether and how defects in axon fasciculation affect the formation and function of neural circuits remain unclear. Here we examine whether the transmembrane proteins semaphorin-5A (Sema5A) and semaphorin-5B (Sema5B), and their receptors, plexinA1 (PlexA1) and plexinA3 (PlexA3) play roles in the development of corticospinal circuits. We find that Sema5A/Sema5B and PlexA1/A3 are required for proper axon fasciculation of corticospinal neurons. Furthermore, Sema5A/5B and PlexA1/A3 mutant mice show marked deficits in skilled motor behaviors. Therefore, these results strongly suggest that proper corticospinal axon fasciculation is required for the appropriate formation and functioning of corticospinal circuits in mice.
Introduction
Fasciculation, the tight bundling of axons into fascicles, is thought to be a key step in neural circuit development enabling axons to traverse long distances within the mammalian body. Previous studies have elucidated various molecular mechanisms underlying axon fasciculation both in the peripheral and CNS. In mice lacking semaphorin-3A (Sema3A) or its receptor neuropilin-1 (Npn-1), sensory axons in the PNS are defasciculated during development (Fujisawa et al., 1997; Kitsukawa et al., 1997; C. Gu et al., 2003; Huber et al., 2005; Haupt et al., 2010), while in the CNS, various molecules are required for axon fasciculation in different neuronal populations (Behar et al., 1996; Kawauchi et al., 2003; Wolman et al., 2007). In contrast, little is known about how defects in axon fasciculation and defasciculation at appropriate points along their trajectory, before target innervation affect neural circuit wiring and function.
In mice, corticospinal (CS) axons form the thick axon bundles in the dorsal funiculus. CS axons must travel great distances within the dorsal funiculus to reach their appropriate spinal neuron targets and will not send their axon collaterals to the spinal gray matter until they have reached the appropriate spinal segmental levels and thereby form synapses with spinal neurons (Canty and Murphy, 2008). Timing disruptions in innervation of the spinal gray matter by CS axons could result in commands being sent to inappropriate spinal segments resulting in aberrant coordination of muscle activation. During normal development, CS neurons (CSNs) in the forelimb motor cortex project axons to cervical spinal levels via the corticospinal tract (CST), then, on exiting the CST, they are defasciculated within the cervical gray matter. In contrast, CS axons originating in the hindlimb motor cortex are primarily defasciculated when they reach the lumbar spinal cord, again, innervating spinal interneurons at this level. This initial fasciculation, followed by defasciculation at appropriate spinal levels must be driven by a delicate choreography of axonal guidance cues to ensure that axons bundle together, then unbundle and form connections at the appropriate spinal levels. Axon fasciculation may, in turn, facilitate targeting by promoting axon growth past areas that would be inappropriate to innervate (Canty and Murphy, 2008). The signaling mechanisms used by cells in the spinal cord to achieve fasciculation, then appropriate defasciculation and spinal segmentation of CS axons remain incompletely understood.
Semaphorins and their receptors, the plexins and neuropilins, play significant roles in nervous system development, in areas, such as neural proliferation, migration, polarization, synapse development, and axon guidance throughout the CNS and PNS (Tran et al., 2007; Yoshida, 2012). Semaphorin signaling performs a variety of roles in the development of the CST as well. For example, CS axons of plexinA3/A4 double-mutant mice and their semaphorin ligand, Sema6A, fail to turn dorsally and decussate at the pyramids (Faulkner et al., 2008; Runker et al., 2008). In addition, in cortical conditional KO of neuropilin-1, CS axons aberrantly extend to the ventral spinal cord (Z. Gu et al., 2019).
Here, we report that the transmembrane proteins semaphorin-5A (Sema5A) and semaphorin-5B (Sema5B), along with their receptors, plexin-A1 (PlexA1) and plexin-A3 (PlexA3) (Matsuoka et al., 2011), are required for proper axon fasciculation of CSNs in the medulla at the site of axon decussation. Our study demonstrates a previously unknown role for this subset of semaphorins and their plexin receptors in the CST targeting of appropriate spinal segments. Their repulsive signaling maintains CST fasciculation and prevents CS axons from the hindlimb motor cortex from innervating cervical spinal segments. Their roles in the timing of CS axon fasciculation and defasciculation on tract exit are is required for segment-specific CS circuit formation, which is necessary for skilled movements.
Materials and Methods
Mouse lines
The following mouse strains were used in this study: Emx1-Cre (Gorski et al., 2002), CAG-CAT-eGFP (ccGFP) (Nakamura et al., 2006), Sema5A−/− (Matsuoka et al., 2011), Sema5B−/− (Matsuoka et al., 2011), PlexA1fl/fl (Yoshida et al., 2006), and PlexA3−/− (Bagri et al., 2003). Mice were maintained on a C57BL/6J background. All animals were treated according to institutional and National Institutes of Health guidelines, with the approval of the Institutional Animal Care and Use Committee at Cincinnati Children's Hospital Medical Center and Burke Neurologic Institute/Weill Cornell Medicine.
Immunohistochemistry and imaging
Perfusion fixation (ice-cold PBS followed by 4% PFA) was used to harvest brains and spinal cords. Upon dissection, the brain and spinal cord were fixed overnight at 4°C. Brains and spinal cords were cryoprotected by immersion in 30% sucrose/PBS for 48 h and sectioned using a cryostat at 50 µm and 80 µm thickness, respectively. Free-floating immunohistochemistry was performed by incubating brain and spinal cord sections with primary antibodies overnight at room temperature, then with fluorophore-conjugated secondary antibodies for 4 h at room temperature. Sections were mounted with Vectashield Mounting Media (Vector Labs) and coverslipped for imaging. Confocal images were taken with a Nikon A1R GaAsP. The following primary antibodies were used in this study: chicken anti-eGFP (1:2000, Abcam); goat anti-mCherry (1:1000, Biorbyt); rat anti-Ctip2 (1:1000, Abcam); and mouse anti-NeuN (1: 2000, Millipore). The fluorophore-conjugated secondary antibodies were obtained from Jackson ImmunoResearch Laboratories and Invitrogen. Sections were scanned with Leica SP8 and Nikon A1R.
ISH
ISHs were performed on PFA-fixed cryosections (12-14 µm thickness) as described previously (Yoshida et al., 2006) using digoxigenin-labeled antisense riboprobes specific for the coding sequences of Sema5A (3178-3690 bp), Sema5B (808-1376 bp), PlexA1 (600-1149 bp), and PlexA3 (187-733 bp).
Surgery and virus injections
Mice were anesthetized with isoflurane during surgery and injections. The left forelimb and hindlimb motor cortices of P4 WT, Sema5A−/−; Sema5B−/−, and PlexA1fl/fl; PlexA3−/− mice were injected with 300 nl of AAV1-CAG-GFP and AAV1-CAG-tdTomato using the following coordinates: forelimb: AP, 1.5 mm from bregma; ML, −1.0 mm; and depth, 0.4 mm below the surface of the brain; hindlimb: AP, −1.0 mm from bregma; ML, −1.0 mm; and depth, 0.4 mm below the surface of the brain. AAV solutions (Penn Vector Core) were diluted to a final titer of 1 × 1012 genome copies/ml (GC/ml) for injections. Brains and spinal cords were collected at P16 (see Fig. 1), P23 (see Fig. 1), and P56 (see Figs. 1, 7, 8).
Decussation analysis
AAV1-CAG-tdTomato (Addgene, #59462, 1.9 × 1013 viral genomes (vg)/ml, 200 nl/site) was injected into the cortex at P7, and mice were killed at P14 (see Fig. 6). The following coordinates were used to target corticospinal neurons with bregma as 0 for anterior origin for AP, ML, and DV axes: (1.5 mm, 1.2 mm, 0.5 mm), (0.5 mm, 1.5 mm, 0.5 mm), and (1.0 mm, 1.3 mm, 0.5 mm).
Intracranial microstimulation (ICMS) and electromyography (EMG) recordings
ICMS experiments were performed on 3-month-old WT, Sema5A−/−; Sema5B−/−, and PlexA1fl/fl; PlexA3−/− mice. Electrical responses were recorded from biceps (Bi) and rectus femoris (Rf) muscles using percutaneous Ni-chrome wire electrodes (de-insulated 1 mm from the tip) in response to motor cortex threshold stimulation as previously reported (Z. Gu et al., 2017a,b). EMG recording wires were inserted using a 27 gauge hypodermic needle guide. We recorded differentially, with two-wire electrodes within each muscle and a separate ground, and verified the adequacy of the EMG recordings and muscle placements by noting increased EMG activity with passive joint rotation. EMG recordings were made with a differential AC amplifier with low- and high-pass filtration (model 1700; A-M Systems). EMG signals were acquired using an analog-to-digital converter (CED) and processed using Signal software (version 6.05; CED). Averages of rectified EMGs were generated from individual sites from each animal. To quantify the functional connectivity between the cortex and muscles, we calculated the integrated EMG values for each muscle and derived the following connectivity indices: the forelimb cortex and forelimb muscle index (the integrated EMG value for the Bi muscle divided by the average integrated EMG value for the Bi muscle in control mice); the hindlimb cortex and hindlimb muscle index (the integrated EMG value for the Rf muscle divided by the average integrated EMG value for the Rf muscle in control mice); and the hindlimb cortex and forelimb muscle index (the integrated EMG value for the Bi muscle divided by the average integrated EMG value for the Bi muscle in control mice). To illustrate the functional connectivity between the forelimb cortex and Bi muscle, a 1.66 mm × 2.0 mm portion of the forelimb motor cortex was divided equally into 42 squares. To illustrate the functional connectivity between the hindlimb cortex and Rf muscle, a 1.5 mm × 2.0 mm portion of the hindlimb motor cortex was divided equally into 35 squares. To illustrate the functional connectivity between hindlimb cortex and Bi muscle, a 1.5 mm × 2.0 mm portion of the hindlimb motor cortex was divided equally into 35 squares. Individual sites were assigned to different squares based on their anterior to posterior and medial to lateral coordinates (relative to bregma). An average connectivity index was calculated from all sites in each square. Connectivity index maps were generated from plots of the average laterality indices from all squares in each map (see Figs. 9, 10).
Beam-walking test. The beam-walking test was conducted on 3-month-old WT, Sema5A−/−; Sema5B−/−, and PlexA1fl/fl; PlexA3−/− mice as previously described (Z. Gu et al., 2017b). Beams with a flat surface of 16, 8, and 4 mm width were used in this test. Performance on the beam was quantified by measuring the time required for each mouse to traverse the beam.
Single pellet reaching test. We assessed the performance of 3-month-old WT, Sema5A−/−; Sema5B−/−, and PlexA1fl/fl; PlexA3−/− mice in a single pellet reaching test as previously described (Xu et al., 2009), with minor modifications. Briefly, mice were food-restricted to maintain 90% of their free-feeding weight before the training. We determined the preferred limb for each mouse during the shaping phase (2-4 d), which was followed by a 7 d training protocol. We recorded 30 reaches from each mouse per day during training. Only when the mouse successfully retrieved the seed and put it into its mouth was the attempt considered a success.
Accelerating rotarod. The accelerating rotarod (Med Associates) was used to assess motor coordination of 3-month-old WT, Sema5A−/−; Sema5B−/−, and PlexA1fl/fl; PlexA3−/− mice (Med Associates). Mice were placed on a 3-cm-diameter rod with an initial rotation speed of 4 rpm that accelerated to 40 rpm over 5 min. Mice were tested for tumble latency (time before falling off the rod) in 8 trials over 2 consecutive days as described previously (Daily et al., 2011).
Statistics
Results are expressed as the mean ± SEM. Two-way repeated-measures ANOVAs were used for Figure 8B, C (followed by post hoc comparisons). Student's t tests were used for other statistical analyses. Error bars indicate the SEM.
Results
Segmental specificity of different CS circuits revealed by dual-color anterograde tracing
CS circuits exhibit somatotopic organization. In the hindlimb motor cortex, CSNs project axons to, and preferentially the lumbar spinal cord, whereas in the forelimb motor cortex CS axons extend to the cervical spinal cord (Tyner, 1974; Wise et al., 1979; Li et al., 1990). To simultaneously visualize the descending axons from these two distinct CSN populations, we injected postnatal day 4 (P4, n = 15) mice with AAVs expressing either GFP (AAV1-CAG-eGFP) or tdTomato (AAV1-CAG-tdTomato) targeting the forelimb and hindlimb motor cortices, respectively (Fig. 1A). This dual-color labeling resulted in restricted expression of GFP and tdTomato in the forelimb and hindlimb motor cortices of adult mice at P56 (n = 5) (Fig. 1B–E). We observed distinct, nonoverlapping expression of GFP and tdTomato in Ctip2+ CSNs in the forelimb and hindlimb motor cortices (Fig. 1F–L). We then investigated descending axons from these virally labeled CSNs in distinct cortical locations for their segmental targeting specificity in juvenile (P16, n = 6; and P23, n = 4) and adult (P56, n = 5) mice. At P16, CSNs from the forelimb motor cortex projected down to the thoracic spinal cord, whereas CSNs from the hindlimb motor cortex projected to the lumbar spinal cord (Fig. 1M–P). At this stage, CSNs from the forelimb motor cortex sent out extensive axon collateral branches to innervate the medial and ventral regions of the cervical and thoracic spinal segments (Fig. 1N,O), whereas those from the hindlimb motor cortex profusely innervated the dorsal and medial lumbar spinal cords (Fig. 1P). Interestingly, CSN axons arising from the hindlimb motor cortex also exhibited slight arborization on the dorsal side of the cervical and thoracic spinal segments (Fig. 1N,O). The developing and adult CS circuits exhibited some forelimb-hindlimb segmental specificity at P16 (Fig. 1N–P), P23 (Fig. 1Q–S), and P56 (Fig. 1T–V). In WT mice, the tdTomato+ CS axons from the hindlimb motor cortex were in the dorsal cervical spinal cord at both P23 and P56 with a sharp developmental decrease in axon collaterals from P16 to P56 (p < 0.0001, unpaired t test) and from P23 (p = 0.0025, unpaired t test) to P56 (p < 0.0001, unpaired t test) (Fig. 1W), consistent with a previous investigation (Kamiyama et al., 2015). In contrast, GFP+ axonal collaterals from the forelimb motor cortex innervated the cervical spinal cord and increased from P16 to P23 (p = 0.0004, unpaired t test) and from P16 to P56 (p < 0.0001, unpaired t test) (Fig. 1X). Therefore, our dual-color axon tracing experiments revealed early segmental specificity at P16 and P23 with further refinement in adults.
Segmental specificity and developmental reorganization of the CS system revealed by dual-color anterograde tracing. A, Schematic diagram showing the dual-color anterograde tracing experiment to label forelimb and hindlimb motor cortices with GFP and tdTomato, respectively. B-E, Epifluorescence and color images represent the expression of GFP and tdTomato in forelimb and hindlimb motor cortices, respectively. C, Representative sagittal section from the injected brain (indicated by a black line in B) showing the expression of GFP and tdTomato in forelimb and hindlimb motor cortices. G-I, High-magnification images of the boxed areas in F showing that many GFP (green) labeled cortical neurons are positive for Ctip2 (blue). J-L, High-magnification images of the boxed areas in F showing that many tdTomato (red) labeled cortical neurons are positive for Ctip2 (blue). M, Schematic diagram represents the labeling of CSNs in the forelimb and hindlimb motor areas with GFP and tdTomato, and their axonal projections to different segmental levels of the spinal cord. N-P, Spinal cord sections from juvenile (P16) mice were immunostained for tdTomato (red) and GFP (green) showing innervations of GPF+ and tdTomato+ CS axons at cervical level 2 (N), thoracic (O), and lumbar spinal cords (P). Q-S, Spinal cord sections from [juvenile/adolescent?] P23 mice were immunostained for tdTomato (red) and GFP (green) showing innervations of GPF+ and tdTomato+ CS axons at cervical level 2 (Q), thoracic (R), and lumbar spinal cords (S). T-V, Spinal cord sections from adult (P56) mice were immunostained for tdTomato (red) and GFP (green) showing innervations of GPF+ and tdTomato+ CS axons at cervical level 2 (T), thoracic (U), and lumbar spinal cords (V). W, Quantification of tdTomato+ axon density, showing a significant increase (p = 0.0025) in tdTomato+ axon collaterals from P16 to P23 and a marked decrease (p < 0.0001) in tdTomato+ axon collaterals from P16 to P56 in dorsal spinal cords at cervical levels. X, Quantification of GFP+ axon density, showing a significant increase in GFP+ axon collaterals from P16 to P23 (p = 0.0004) and [P16 to] P56 (p < 0.0001). Scale bars: E, 2 mm; F, 200 µm; L, 100 µm; V, 200 µm. **p < 0.01. ***p < 0.001.
Expression of Sema5A and Sema5B in the cortex, brainstem, and spinal cord during development
Recent studies have revealed some important roles, such as axon guidance for Class 3 secreted and Class 6 transmembrane semaphorins during CS development (Finger et al., 2002; Sibbe et al., 2007; Faulkner et al., 2008; Runker et al., 2008; Z. Gu et al., 2017a, 2019). In this study, we explored the possible contributions of Class 5 transmembrane semaphorins (Sema5A and Sema5B) in CS circuit formation and function. Two Class 5 semaphorins, Sema5A and Sema5B, are phylogenetically conserved membrane-bound proteins that display high overall amino acid similarity and an identical arrangement of semaphorin domains (Adams et al., 1996; Tran et al., 2007). Both Sema5A and Sema5B can act as guidance cues to either attract or repel neuronal processes, depending on the neuronal populations involved (Oster et al., 2003; Goldberg et al., 2004; Kantor et al., 2004; Hilario et al., 2009; Lett et al., 2009), and have been shown to play functionally redundant roles in inhibiting retinal neurite outgrowth through their PlexA1 and PlexA3 receptors (Matsuoka et al., 2011). To investigate the possible contributions of Sema5A and Sema5B to CS circuit formation, we began by examining the expression of Sema5A and Sema5B at distinct points along the trajectory of CS axons: in the sensorimotor cortex, the brainstem where CS axons decussate, and the spinal cord during mouse postnatal development.
In the sensory motor cortex of postnatal day 0 (P0) mice, PlexA1 and PlexA3 were both expressed across multiple layers of sensorimotor cortex, including in Ctip2+ neurons in layer Vb, where CSNs are located (Fig. 2A,B,E,F). Sema5A and Sema5B, however, were not detectable in these areas (Fig. 2C,D).
Expression of PlexA1, PlexA3, Sema5A, and Sema5B in the cerebral cortex. A-D, Expression of indicated genes in the cerebral cortex of WT mice at P0 was assessed by RNA ISHs. The signals for Sema5A and Sema5B were much more scattered and weaker compared with signals of PlexA1 and PlexA3 in the motor cortex area. HP, Hippocampus; MO, putative motor cortex area. E, F, Immunofluorescent image of Ctip2 (E) or nuclear staining by DAPI. Serial sections from the same sample were used in all images (A-F). Scale bars, 500 µm.
Interestingly, we found that these ligands were expressed in the caudal most region of the medulla at P0 (Fig. 3A–J). Both Sema5A and Sema5B were expressed bilaterally specifically in the dorsal region, confined near the area where CS axons traverse the decussation around P0 to P2 (Fig. 3A–N). However, by P14, the expression of Sema5A and Sema5B in the medulla was dramatically decreased (Fig. 3O–Q), suggesting that Sema5A/5B-PlexA1/A3 signaling may influence CS projections and/or fasciculation at the site of the decussation in the medulla.
Specific expression of Sema5A and Sema5B in the medulla. A-E, L, P, Expression of Sema5A in the medulla of WT mice at P0 (A-E,L) and P14 (P) was assessed by RNA ISHs. F-J, M, Q, Expression of Sema5B in the medulla of WT mice at P0 (F-J,M) and P14 (Q) was assessed by RNA ISHs. C, H, Black and brown dotted lines indicate approximate positions of the sagittal section shown in L and M, respectively. Black and brown arrows indicate specific bilateral and centered expression, respectively. Specific expression continues up to the cervicomedullary junction. K, O, Immunofluorescent images of PKCγ (K,O) and DAPI staining (K) at P14 are shown to visualize the tract of CS axons. Serial sections from the same sample are used in images (A-J), images (L,M), and images (O-Q). N, A schematic representation of the putative route of CS axons (in red) and Sema5A/B expression. Blue lines indicate approximate positions of the coronal section shown in A-J. Med, Medulla; SC, spinal cord. Scale bars, 250 µm.
In the spinal cord, Sema5A and Sema5B were uniformly expressed in the cervical, thoracic, and lumbar spinal cords at all stages (P0-P14) (Fig. 4A–R), although it is not clear which cell types expressed Sema5A and Sema5B. Together, the expression patterns of PlexA receptors by CSNs and Sema5A/5B ligands in the medulla and spinal cords suggest that Sema5A/5B-PlexA1/A3 signaling is involved in regulating CS development during early postnatal mice.
A-R, Expression of Sema5A and Sema5B in the spinal cord. Expression of Sema5A (top panel set) and Sema5B (bottom panel set) in the spinal cords of WT mice at P0, P6, and P14 assessed by RNA ISHs. Sema5A was ubiquitously expressed in the cervical, thoracic, and lumbar spinal cords at all stages with reduced expression at P14. Sema5B was also ubiquitously expressed in the cervical, thoracic, and lumbar spinal cords at all stages with reduced expression from P6 to P14. Scale bars, 100 µm.
Defects in fasciculation of CS axons in the spinal cord in the absence of Sema5A/5B-PlexA1/A3 signaling
To examine how CS axons descend and innervate the spinal cords in the absence of Sema5A and/or Sema5B, we examined CS axons that were genetically labeled with GFP in the cervical spinal cords from Sema5A−/−; Emx1-Cre::ccGFP (n = 7), Sema5B−/−; Emx1-Cre::ccGFP (n = 6), Sema5A−/−; Sema5B−/−; Emx1-Cre::ccGFP (n = 8), or Emx1-Cre; ccGFP (control, n = 9). Most of the GFP+ CS axons traveled within the ventral most part of the dorsal funiculus (vDF) of Emx1-Cre::ccGFP control mice (Fig. 5A,B). We found that a subset of CS axons in both Sema5A−/− (Fig. 5C,D) and Sema5B−/− (Fig. 5E,F) mice were slightly defasciculated from the main CS axon bundle in the vDF, with axon bundles traversing the gray matter of the cervical spinal cord (Fig. 5A,C,E). This defasciculation phenotype had smaller CSTs in the vDF of Sema5A−/− (p = 0.0136, unpaired t test) and Sema5B−/− mice (p = 0.0300, unpaired t test) compared with controls (Fig. 5I). Sema5A−/−; Sema5B−/− mice exhibited stronger defects in CS axon fasciculation compared with control mice (p = 0.0014, unpaired t test) (Fig. 5A,G,I). Together, these data indicate that Sema5A and Sema5B exhibit dynamic expression patterns in the spinal cord during postnatal development of the CS circuits and that they synergistically function to mediate fasciculation of CS axons.
Fasciculation defects of CS axons in Sema5A−/−; Sema5B−/− and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. A, Image of the spinal cord at cervical level 2 from P10 Emx1-Cre; ccGFP mice (n = 9) showing CS axons (green) and NeuN (red). B, High magnification view of the dotted line boxed area in (A) showing GFP+ CS axon collaterals in the white matter. C, Image of the spinal cord at cervical level 2 from P10 Sema5A−/−; Emx1-Cre; ccGFP mice (n = 7) showing CS axons (green) and NeuN (red). D, High magnification view of the dotted line boxed area in (C) showing GFP+ CS axon collaterals and defasciculated CS axon bundles in the white matter. E, Image of the spinal cord at cervical level 2 from P10 Sema5B−/−; Emx1-Cre; ccGFP mice (n = 6) showing CS axons (green) and NeuN (red). F, High magnification view of the dotted line boxed area in (E) showing GFP+ CS axon collaterals and defasciculated CS axon bundles in the white matter. G, Image of the spinal cord at cervical level 2 from P10 Sema5A−/−; Sema5B−/−; Emx1-Cre; ccGFP mice (n = 8) showing CS axons (green) and NeuN (red). H, High magnification view of the dotted line boxed area in (G) showing GFP+ CS axon collaterals and defasciculated CS axon bundles in the white matter. Note that GFP+ defasciculated CS axon bundles were present in mutant mice (C, E, and G) but absent from wild-type mice (A). I, Quantification of the size of the CST (boxed areas in A, C, E, and G) showing a significant reduction in CST size in Sema5A−/− (p = 0.0136), Sema5B−/− (p = 0.0300), and Sema5A−/−; Sema5B−/− (p = 0.0014) mutant mice. J, The sizes of the vlCST in Sema5A−/− (p = 0.0860), Sema5B−/− (p = 0.3961), and Sema5A−/−; Sema5B−/− (p = 0.1124) mice were similar to that of control mice. K, Image of the spinal cord at cervical level 2 from P10 Emx1-Cre; ccGFP mice (n = 9) showing CS axons (green) and NeuN (red). L, High magnification view of the dotted line boxed area in (K) showing GFP+ CS axon collaterals in the white matter. M, Image of a spinal cord section at cervical level 2 from P10 PlexA1fl/fl; Emx1-Cre; ccGFP mice (n = 8) showing CS axons (green) and NeuN (red). N, High magnification view of the dotted line boxed area in (M) showing GFP+ CS axon collaterals and defasciculated CS axon bundles in the white matter. O, Image of the spinal cord at cervical level 2 from P10 PlexA3−/−; Emx1-Cre; ccGFP mice (n = 9) showing CS axons (green) and NeuN (red). P, High magnification view of the dotted line boxed area in (O) showing GFP+ CS axonal collaterals and defasciculated CS axon bundles in the white matter. Q, Image of the spinal cord at cervical level 2 from P10 PlexA1fl/fl; PlexA3−/−; Emx1-Cre; ccGFP mice (n = 9) showing CS axons (green) and NeuN (red). R, High magnification view of the dotted line boxed area in (Q) showing GFP+ CS axonal collaterals and defasciculated CS axon bundles in the white matter. Defasciculation of GFP+ CS axon bundles was present in PlexA1fl/fl; PlexA3−/−; Emx1-Cre; ccGFP mice (Q) but absent from wild-type (K) and PlexA3−/−; Emx1-Cre; ccGFP mice (O). Mild defasciculation defects were also observed in PlexA1fl/fl; Emx1-Cre; ccGFP mice (M). S, Quantification of the size of the CST (boxed areas in K, M, O, and Q) showing a significant reduction in CST size in PlexA1fl/fl; Emx1-Cre (p = 0.0157) and PlexA1fl/fl; PlexA3−/−; Emx1-Cre (p = 0.0055) but not in PlexA3−/− (p = 0.0725) mice. T, Quantification showing a significant increase in vlCST size in PlexA1fl/fl; Emx1-Cre (p = 0.0001) and PlexA1fl/fl; PlexA3−/−; Emx1-Cre (p < 0.0001) but not in PlexA3−/− (p = 0.3139) mice. Scale bars, 100 µm (Q), 20 µm (R). N.S. = not significant.
Defects in axon fasciculation in the brainstem of Sema5A−/−; 5B−/− mice. A-H, Coronal sections of WT (A-D) and Sema5A−/−; 5B−/− mice at P14. Corticospinal axons were labeled with tdTomato (red). Sections were labeled with DAPI (blue). A'-H', Magnification views of boxed area in A-H.
Since PlexA1 and PlexA3 are shown to be functional receptors for Sema5A and Sema5B (Matsuoka et al., 2011), we examined CS axons in the cervical spinal cord from Emx1-Cre::ccGFP (n = 9), PlexA1fl/fl; Emx1-Cre::ccGFP (n = 8), PlexA3−/−; Emx1-Cre::ccGFP (n = 9), or PlexA1fl/fl; PlexA3−/−; Emx1-Cre::ccGFP (n = 9) mice at P10 (Fig. 5K–T). We found that a small subset of CS axons in PlexA1fl/fl; Emx1-Cre:: ccGFP mice were defasciculated from the main CS axon bundle in the vDF and traveled in the gray matter of the cervical spinal cord, leading to a decrease in CST size (p = 0.00157, unpaired t test) (Fig. 5K,Q,S). Similar to control mice, the CS axons in PlexA3−/−; Emx1-Cre; ccGFP mice were mainly confined to the vDF, resulting in CSTs that were similar in size to those of control mice (p = 0.0725, unpaired t test) (Fig. 5K,O,S). A substantial portion of CS axons were defasciculated from the main CS axon bundle and were present in the gray matter of the cervical spinal cord in PlexA1fl/fl; PlexA3−/−; Emx1-Cre:: ccGFP mice, causing a significant reduction in CST size in the vDF (p = 0.0055, unpaired t test) (Fig. 5K,Q,S). Consistent with our previous observations, we also observed a presence of the ventral lateral CST (vlCST) specifically in PlexA1fl/fl; Emx1-Cre:: ccGFP (p = 0.0001, unpaired t test) (Fig. 5M,T) and PlexA1fl/fl; PlexA3−/−; Emx1-Cre:: ccGFP (p < 0.0001, unpaired t test) (Fig. 5Q,T) mice compared with controls and other mutants (Fig. 5C,E,G,J,K,O). This is likely because of the presence of unpruned transient vlCST axons that are normally eliminated by Sema6D-PlexA1 signaling but persist in PlexA1 mutant mice (Z. Gu et al., 2017a). Together, these data suggest that PlexA1 and PlexA3 play functional roles in mediating Sema5A and Sema5B signaling to regulate the fasciculation of CS axons.
Defects in CS axon fasciculation in the medulla in the absence of Sema5A/Sema5B-PlexA1/PlexA3 signaling
Since we observed CS axon fasciculation defects in the spinal cord, we wondered whether the absence of Sema5A/Sema5B-PlexA1/PlexA3 signaling causes CS axon fasciculation defects elsewhere along their trajectory at levels rostral to the spinal cord. We examined CS axons in the medulla at the decussation site above the spinal cord levels by injecting AAV1-CAG-tdTomato into the cortex of WT and Sema5A−/−; Sema5B−/− mice at P7, and analyzing the mice at P14. In WT mice, CS axon bundles were tightly fasciculated in the medulla at the decussation site and traveled through the dorsal funiculus in the spinal cord (Fig. 6A–D). In contrast, in Sema5A−/−; Sema5B−/− mice, some CS axons seemed to be prematurely defasciculated in the medulla at the decussation site and those prematurely defasciculated axons entered the gray matter of the medulla, and are likely the axons that we observed within the cervical spinal gray matter (Fig. 6E–H). Therefore, initial defects in CS axon fasciculation likely occur in the medulla at the decussation site of CS axons, and these defasciculated axons then remain separated from the main CST and traverse into the spinal gray matter.
Segmental specificity is disrupted in the absence of Sema5A/5B-PlexA1/A3 signaling
We next investigated whether the ectopic axon bundles observed in the cervical cord of Sema5A−/−; Sema5B−/− mice originated from the forelimb or hindlimb motor cortex or both. We performed dual-color anterograde axon tracing in P4 WT (control, n = 6) and Sema5A−/−; Sema5B−/− (n = 8) mice to express GFP and tdTomato in CSNs in the forelimb and hindlimb motor cortex, respectively. We then followed the CSN projections and traced their innervation of the spinal cord and found that CSN axons originating in the forelimb motor cortex mainly terminate at the cervical and thoracic spinal cord levels at P56, while CSNs in the hindlimb motor cortex predominantly project to the lumbar spinal cord (Fig. 7A–C,G–I). Ectopic CS axons in the gray matter of the cervical spinal cord in Sema5A−/−; Sema5B−/− mice mainly contained tdTomato+ axons, suggesting that CS axons from the hindlimb motor cortex prematurely enter the cervical spinal gray matter from the main axon bundle in the vDF (Fig. 7D–F). Accordingly, we observed an increase in tdTomato+ axon collaterals of CSNs from the hindlimb motor cortex innervating the dorsal (p < 0.0001) and ventral (p = 0.0030) cervical gray matter of Sema5A−/−; Sema5B−/− mice compared with WT mice (Fig. 7A–F,N). In the lumbar spinal cord, we observed a marked decrease of tdTomato+ CS axons from the hindlimb motor cortex innervating the dorsal (p = 0.0019) and ventral (p < 0.0001) lumbar gray matter in Sema5A −/−; Sema5B −/− mice. These results are consistent with the finding that many CS axons from the hindlimb motor cortex were defasciculated from the CST and inappropriately innervated the cervical spinal gray matter in Sema5A−/−; Sema5B−/− mice (Fig. 7D–F). Quantitative analyses of CS axons revealed that GFP+ axon collaterals of CSNs in the forelimb motor cortex innervating the dorsal (p = 0.2175) and ventral (p = 0.3794) cervical spinal cord were similar between control and Sema5A −/−; Sema5B −/− mice. Interestingly, there was a significant increase of GFP+ CS axons from the forelimb motor cortex in the dorsal (p = 0.0004) and ventral (p = 0.0002) lumbar spinal cords of Sema5A−/−; Sema5B−/− mice compared with those observed in control mice (Fig. 7G–L,O,P), suggesting that Sema5A/5B also function to prevent the defasciculation of CSNs from the forelimb motor cortex in the CST at lumbar levels. Together, these data suggest that the loss of function of Sema5A and Sema5B leads to premature defasciculation of CS axons from the hindlimb motor cortex at the cervical spinal cord level, resulting in the innervation of the cervical spinal cords by CSNs from the hindlimb motor cortex and a marked decrease of anatomic connections between the hindlimb motor cortex and lumbar spinal cords. Thus, segmental specificity was severely disrupted in Sema5A−/−; Sema5B−/− mice.
We next investigated this segmental specificity in PlexA1 and PlexA3 mutant mice. In P56 WT mice, GFP+ CSNs from the forelimb motor cortex mainly arborized in the cervical gray matter, whereas tdTomato+ CSNs from the hindlimb motor cortex primarily arborized in the lumbar gray matter (Fig. 8A–L). Unlike control mice, many tdTomato+ CS axons from the hindlimb motor cortex in P56 PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice were defasciculated from the main CS axon bundle in the vDF and innervated the cervical spinal cord (Fig. 8D–F). As a consequence, we observed a significant increase in tdTomato+ CS axon density in the dorsal (p = 0.0004) and ventral (p = 0.0001) cervical gray matter of PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with control mice (Fig. 8A–F,M,N). In contrast, we found marked decreases in tdTomato+ CS axons in the dorsal (p = 0.0008) and ventral (p = 0.0009) lumbar gray matter of PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice (Fig. 8G–L,P). We also observed a significant increase in GFP+ CS axons in the dorsal (p = 0.0091) but not ventral (p = 0.2318) lumbar spinal cord of PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice (Fig. 8G–L,O). Thus, our anatomic analyses demonstrate that deletion of PlexA1 and PlexA3 leads to a disruption in segmental specificity of CS circuits similar to that observed in Sema5A and Sema5B double-mutant mice.
Functional defects in CS circuits in the absence of Sema5A/5B-PlexA1/A3 signaling
To determine whether the anatomic disruptions of CS circuits by axon defasciculation in Sema5A−/−; Sema5B−/− mice affect CS circuit wiring, we performed ICMS in the motor cortex and recorded EMGs from forelimb (Bi) and hindlimb (Rf) muscles in 3-month-old adult WT (n = 5) and Sema5A−/−; Sema5B−/− (n = 3) mice (Fig. 9A,G). The ICMS threshold for the forelimb motor cortex was similar between control and Sema5A−/−; Sema5B−/− mice (p = 0.5815, Student's t test) (Fig. 9F), while the hindlimb ICMS threshold was significantly higher in Sema5A−/−; Sema5B−/− mice compared with control mice (p = 0.0091, Student's t test) (Fig. 9L). In WT mice, threshold stimulation of the forelimb motor cortex (44 sites) evoked robust Bi, but not Rf, EMG responses, whereas threshold stimulation of the hindlimb motor cortex (35 sites) threshold stimulation only induced strong Rf, but not Bi, EMG responses (Fig. 9B,C,G,H). Similar to those observed in control mice, we found equivalent EMG responses from Bi and Rf muscles in Sema5A−/−; Sema5B−/− mice on threshold stimulation of the forelimb (70 sites) and hindlimb (101 sites) motor cortices, respectively (Fig. 9D,J). However, we found that threshold stimulation of the hindlimb motor cortex in Sema5A−/−; Sema5B−/− (101 sites) but not control (35 sites) mice evoked strong Bi muscle responses, suggesting that the CS fibers from the hindlimb motor cortex that aberrantly innervated the cervical spinal cord in those mutant mice were functional (Fig. 9G,I).
Segmental specificity of corticospinal axonal projections is disrupted in Sema5A−/−; Sema5B−/− mice. A-F, Spinal cord sections at cervical level 2 from 8-week-old WT (A-C, n = 8) and Sema5A−/−; Sema5B−/− (D-F, n = 6) mice, showing GFP+ and tdTomato+ CS axons and their collaterals in the cervical spinal cord. Note the presence of defasciculated tdTomato+ CS axonal bundles and increased tdTomato+ CS axon collaterals in Sema5A−/−; Sema5B−/− mice. M, Quantification of GFP+ axon density, showing that similar densities of GFP+ axon collaterals were present at the dorsal (white dashed box in A, p = 0.2172) and ventral (white dashed box in A, p = 0.3794) cervical spinal cord in WT and Sema5A−/−; Sema5B−/− mice. N, Quantification of tdTomato+ axon density, showing a significant increase in tdTomato+ axon collateral density in Sema5A−/−; Sema5B−/− mice compared with WT mice in the dorsal (p < 0.0001) and ventral (p = 0.0030) regions of the cervical spinal cord. G-L, Spinal cord sections at lumbar level 2 from WT (G–I) and Sema5A−/−; Sema5B−/− mice (J–L) mice, showing GFP+ and tdTomato+ CS axons and their collaterals in the lumbar spinal cord. O, Quantification of GFP+ axon density, showing a significant increase in GFP+ axon collaterals in Sema5A−/−; Sema5B−/− mice compared with WT controls in the dorsal (p = 0.0004) and ventral (p = 0.0002) regions of the lumbar spinal cord. P, Quantification of tdTomato+ axon density, showing a significant decrease in tdTomato+ axon collaterals in Sema5A−/−; Sema5B−/− mice compared with WT mice in the dorsal (p = 0.0019) and ventral (p < 0.0001) regions of the lumbar spinal cord. Scale bar: L, 200 µm. **p < 0.01. ***p < 0.001.
Segmental specificity of corticospinal axonal projections is disrupted in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. A-F, Spinal cord sections at cervical level 2 from 8-week-old WT (A-C, n = 6) and PlexA1fl/fl; PlexA3−/−-Emx1-Cre (D-F, n = 10) mice, showing GFP+ and tdTomato+ CS axons and their collaterals in the cervical spinal cord. Note the presence of defasciculated tdTomato+ CS axonal bundles and increased tdTomato+ CS axon collaterals in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. M, Quantification of GFP+ axon density, showing that similar GFP+ axon collaterals were present at the dorsal (p = 0.8751) and ventral (p = 0.2318) spinal cord at cervical levels in WT and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. N, Quantification of tdTomato+ axon density, showing a significant increase in tdTomato+ axon collaterals in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with WT controls in the dorsal (p = 0.0004) and ventral (p = 0.0001) regions of the cervical spinal cord. G-L, Lumbar level 2 spinal cord sections from WT (G-I) and PlexA1fl/fl; PlexA3−/−-Emx1-Cre (J-L) mice, showing GFP+ and tdTomato+ CS axons and their collaterals in the lumbar spinal cord. O, Quantification of GFP+ axon density, showing a significant increase in GFP+ axon collaterals in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with WT mice in the dorsal (p = 0.0091) spinal cord at lumbar levels. GFP+ axon collaterals in the ventral spinal cord at lumbar levels were similar (p = 0.2318) between the two groups. P, Quantification of tdTomato+ axon density, showing a significant decrease in tdTomato+ axon collaterals in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with WT controls in the dorsal (p = 0.0008) and ventral (p = 0.0009) regions of the lumbar spinal cord. Scale bar: L, 200 µm. **p < 0.01. ***p < 0.001.
Functional deficits in the CS circuits in Sema5A−/−; Sema5B−/− mice. A, Experimental schematic for the forelimb motor cortex ICMS and evoked muscle EMG recordings from Bi (green) and Rf (red) muscles in 3-month-old adult mice. B, C, Representative EMG traces from Bi (green) and Rf (red) muscles in WT mice (n = 5), showing strong activation of Bi but no activation of Rf in response to forelimb motor cortex ICMS (green arrow indicates the onset of the Bi EMG response, B). D, E, Representative EMG traces from Bi (green) and Rf (red) muscles in Sema5A−/−; Sema5B−/− mice (n = 3), showing weak activation of Bi but no activation of Rf in response to forelimb motor cortex ICMS (green arrow indicates the onset of the Bi EMG response, D). F, Stimulation currents used for the 14-pulse forelimb ICMS were indistinguishable (p = 0.5815) between WT (44 sites from 5 mice) and Sema5A−/−; Sema5B−/− (70 sites from 3 mice) mice. G, Experimental schematic for the hindlimb motor cortex ICMS and evoked muscle EMG recordings from Bi (green) and Rf (red) muscles in 3-month-old adult mice. H, J, Representative EMG traces from Bi (green) and Rf (red) muscles in WT mice showing strong activation of Rf but no activation of Bi in response to hindlimb motor cortex ICMS (red arrow indicates the onset of Rf EMG response, I). J, K, Representative EMG traces from Bi (green) and Rf (red) muscles in Sema5A−/−; Sema5B−/− mice, showing strong activation of Bi and weak activation of Rf in response to hindlimb motor cortex ICMS (green and red arrow indicates the onset of Bi [J] and Rf [K] EMG responses, respectively). L, Hindlimb ICMS threshold currents used for Sema5A−/−; Sema5B−/− (101 sites from 3 mice) mice were significantly (p = 0.0091) higher than those of WT (35 sites from 5 mice) mice. M, N, Color-coded plots of the forelimb motor cortex and forelimb muscle connectivity index in WT and Sema5A−/−; Sema5B−/− mice, showing strong forelimb cortex and forelimb muscle connectivity in WT mice (M, n = 44 sites from 5 animals) and weak connectivity in Sema5A−/−; Sema5B−/− mice (N, n = 70 sites from 3 animals). O, Quantification showing that the forelimb motor cortex and forelimb muscle connectivity index is significantly (p < 0.0001) reduced in Sema5A−/−; Sema5B−/− mice. P, Q, Color-coded plots of the hindlimb motor cortex and hindlimb muscle connectivity index in WT and Sema5A−/−; Sema5B−/− mice, showing strong hindlimb cortex and hindlimb muscle connectivity in WT mice (P, n = 35 sites from 5 animals) and weak connectivity in Sema5A−/−; Sema5B−/− mice (Q, n = 101 sites from 3 animals). R, Quantification showing that the hindlimb motor cortex and hindlimb muscle connectivity index is significantly (p < 0.0001) reduced in Sema5A−/−; Sema5B−/− mice. S, T, Color-coded plots of the hindlimb motor cortex and forelimb muscle connectivity index in WT and Sema5A−/−; Sema5B−/− mice, showing almost no hindlimb cortex and forelimb muscle connectivity in WT mice (S, n = 35 sites from 5 animals) and strong connectivity in Sema5A−/−; Sema5B−/− mice (T, n = 101 sites from 3 animals). U, Quantification showing that the hindlimb motor cortex and forelimb muscle connectivity index is significantly (p < 0.0001) enhanced in Sema5A−/−; Sema5B−/− mice. **p < 0.01. ***p < 0.001.
To quantify the functional connectivity between the motor cortex and muscles, we calculated the integrated EMG values for each muscle and derived the following connectivity indices: the forelimb motor cortex and forelimb muscle index (the integrated EMG value for the Bi muscle divided by the average integrated EMG value for the Bi muscle in control mice); the hindlimb motor cortex and hindlimb muscle index (the integrated EMG value for the Rf muscle divided by the average integrated EMG value for the Rf muscle in control mice); and the hindlimb motor cortex and forelimb muscle index (the integrated EMG value for the Bi muscle divided by the average integrated EMG value for the Bi muscle in control mice). A plot of the forelimb motor cortex and forelimb muscle connectivity index revealed a marked reduction in connections between the forelimb cortex and forelimb Bi muscle in Sema5A−/−; Sema5B−/− mice compared with control mice (p < 0.0001, Student's t test) (Fig. 9B,D,M–O). Similarly, a plot of the hindlimb motor cortex and hindlimb muscle connectivity index demonstrates a significant decrease in connections between the hindlimb cortex and hindlimb Rf muscle in Sema5A−/−; Sema5B−/− mice compared with control mice (p < 0.0001, Student's t test) (Fig. 9I,K,P–R). In contrast, we observed a significant increase in functional connections between the hindlimb motor cortex and forelimb Bi muscle in Sema5A−/−; Sema5B−/− mice compared with controls (p < 0.0001, Student's t test) (Fig. 9C,I,S–U). Together, these data strongly suggest that the segment level-specific functional connections of CS circuits were disrupted in Sema5A−/−; Sema5B−/− mice.
To assess the effects of deletion of PlexA1 and PlexA3 on the functional connections between the motor cortex and spinal cords in mice, we measured EMGs from the Bi and Rf muscles in response to ICMS of the motor cortex in 3-month-old adult WT (control, n = 5) and PlexA1fl/fl; PlexA3−/−-Emx1-Cre (n = 5) mice (Fig. 10A,G). The ICMS thresholds for both the forelimb (p = 0.004, Student's t test) and hindlimb (p = 0.0021, Student's t test) motor cortex were significantly higher in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with control mice (Fig. 10F,L). The Bi, but not the Rf, muscle in control mice responded vigorously to the forelimb motor cortex (39 sites) threshold stimulation, whereas Rf, but not Bi, of control mice responded to threshold stimulation of the hindlimb motor cortex (38 sites) motor cortex threshold stimulation (Fig. 10B,C,H,I). Similar to the findings in control mice, the Bi and Rf muscles in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice evoked EMG responses to threshold stimulation of the forelimb (48 sites) and hindlimb (40 sites) motor cortices, respectively (Fig. 10D,J). However, we found that threshold stimulation of the hindlimb motor cortex (40 sites) in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice but not control (40 sites) mice evoked strong Bi muscle responses, indicating that the CS fibers from the hindlimb motor cortex innervating the cervical spinal cord in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice were functional (Fig. 10H,I). Next, we quantified the functional connectivity between the motor cortex and muscles. A plot of the forelimb motor cortex and forelimb muscle connectivity index revealed a marked decrease in connections between the forelimb motor cortex and the forelimb Bi muscle in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with controls (p < 0.0001, Student's t test) (Fig. 10B,D,M–O). Similarly, a plot of the hindlimb motor cortex and hindlimb muscle connectivity index revealed a significant reduction in connections between the hindlimb motor cortex and the hindlimb Rf muscle in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with controls (p < 0.0001, Student's t test) (Fig. 10I,K,P–R). In contrast, we observed a marked increase of functional connections between the hindlimb motor cortex and the forelimb Bi muscle in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice compared with control mice (p < 0.0001, Student's t test) (Fig. 10G,H,J,S–U). Together, these data strongly suggest that the segment level-specific functional connections of CS circuits were disrupted in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice, similar to the disruptions observed in Sema5A−/−; Sema5B−/− mice.
Functional deficits in the CS circuitry of PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. A, Experimental schematic for the forelimb cortex ICMS and evoked muscle EMG recordings from Bi (green) and Rf (red) muscles in 3-month-old adult mice. B, C, Representative EMG traces from Bi (green) and Rf (red) muscles in WT mice (n = 5), showing strong activation of Bi but no activation of Rf in response to forelimb motor cortex ICMS (green arrow indicates the onset of the Bi EMG response, B). D, E, Representative EMG traces from Bi (green) and Rf (red) muscles in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice (n = 5), showing weak activation of Bi but no activation of Rf in response to forelimb motor cortex ICMS (green arrow indicates the onset of the Bi EMG response, D). F, Quantification showing that forelimb ICMS threshold currents used for PlexA1fl/fl; PlexA3−/− -Emx1-Cre (48 sites from 5 mice) mice were significantly (p = 0.004) higher than those for WT (39 sites from 5 mice) mice. G, Experimental schematic for hindlimb motor cortex ICMS and evoked muscle EMG recordings from Bi (green) and Rf (red) muscles in 3-month-old adult mice. H, I, Representative EMG traces from Bi (green) and Rf (red) muscles in WT mice, showing strong activation of Rf but no activation of Bi in response to hindlimb motor cortex ICMS (red arrow indicates the onset of the Rf EMG response, I). J, K, Representative EMG traces from Bi (green) and Rf (red) muscles in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice, showing strong activation of Bi and weak activation of Rf in response to hindlimb motor cortex ICMS (green and red arrow indicate the onset of the Bi [J] and Rf [K] EMG responses, respectively). L, Quantification showing that hindlimb ICMS threshold currents used for PlexA1fl/fl; PlexA3−/−-Emx1-Cre (40 sites from 5 mice) mice were significantly (p = 0.0021) higher than those for WT (40 sites from 5 mice) mice. M, N, Color-coded plots of the forelimb motor cortex and forelimb muscle connectivity index in WT and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice, showing strong forelimb motor cortex and forelimb muscle connectivity in WT (M, n = 39 sites from 5 mice) and weak connectivity in PlexA1fl/fl; PlexA3−/−-Emx1-Cre (N, n = 48 sites from 5 mice) mice. O, Quantification showing that the forelimb motor cortex and forelimb muscle connectivity index is significantly (p < 0.0001) reduced in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. P, Q, Color-coded plots of the hindlimb motor cortex and hindlimb muscle connectivity index in WT and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice, showing strong hindlimb motor cortex and hindlimb muscle connectivity in WT (P, n = 38 sites from 5 mice) and weak connectivity in PlexA1fl/fl; PlexA3−/−-Emx1-Cre (Q, n = 40 sites from 5 mice) mice. R, Quantification showing that the hindlimb motor cortex and hindlimb muscle connectivity index is significantly (p < 0.0001) reduced in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. S, T, Color-coded plots of the hindlimb motor cortex and forelimb muscle connectivity index in WT and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice, showing weak hindlimb motor cortex and forelimb muscle connectivity in WT (S, n = 38 sites from 5 mice) and strong connectivity in PlexA1fl/fl; PlexA3−/−-Emx1-Cre (T, n = 40 sites from 5 mice) mice. U, Quantification showing that the hindlimb motor cortex and forelimb muscle connectivity index is significantly (p < 0.0001) enhanced in PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. **p < 0.01. ***p < 0.001.
Sema5A/5B and PlexA1/A3 double-mutant mice exhibit skilled movement deficits
To determine whether the anatomic and physiological defects in CS circuits in Sema5A/5B and PlexA1/A3 double-mutant mice affect motor behaviors, we subjected control and mutant mice to a beam walking test, a seed-reaching test, and an accelerating rotarod test.
For the beam walking test, 3-month-old WT (control, n = 8), PlexA1fl/fl; PlexA3−/−-Emx1-Cre (n = 8), and Sema5A−/−; Sema5B−/− (n = 4) mice traversed increasingly narrow, elevated beams of varying widths (16, 8, and 4 mm widths) to reach a safe platform. Performance was quantified by measuring the time taken to cross each beam as described previously (Z. Gu et al., 2017b). We found that the time to cross the 16 mm beam was indistinguishable between control and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice (p = 0.4775), while Sema5A−/−; Sema5B−/− mice took significantly longer to cross the beam (p = 0.0042) (Fig. 11A). Importantly, compared with control mice, both Sema5A−/−; Sema5B−/− (8 mm: p = 0.0014; 4 mm: p < 0.0001) and PlexA1fl/fl; PlexA3−/−-Emx1-Cre (8 mm: p < 0.0001; 4 mm: p < 0.0001) mice took much longer to cross the 8 mm and 4 mm beams (Fig. 11A).
Adult Sema5A/5B and PlexA1/A3 double-mutant mice exhibit skilled movement deficits. A-C, Performance on behavioral tests by 3-month-old adult WT (n = 8), Sema5A−/−; Sema5B−/− mice (n = 4), and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice (n = 8). A, Comparison of crossing times on the beam walking task. The time to cross the 16 mm beam was indistinguishable (p = 0.4775) between WT and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice. However, PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice required more time to cross the 8 mm (p = 0.0014) and 4 mm (p < 0.0001) beams than WT mice. Similarly, Sema5A−/−; Sema5B−/− mice required greater time to cross the beams than control mice (16 mm, p = 0.0042; 8 mm, p < 0.0001; 4 mm, p < 0.0001). B, Success rates of Sema5A−/−; Sema5B−/− mice (p < 0.0001), and PlexA1fl/fl; PlexA3−/−-Emx1-Cre (p < 0.0001) mice were significantly lower than that of WT mice during the reaching task. C, Line graph indicates changes in latency before falling off of an accelerating rotarod. Sema5A−/−; Sema5B−/− mice (p = 0.0001) and PlexA1fl/fl; PlexA3−/−-Emx1-Cre (p < 0.0001) mice seemed to experience greater difficulty staying on the rotarod and had significantly shorter latencies before falling than WT mice. **p < 0.01. ***p < 0.001.
Next, we conducted the single-seed reaching task (Xu et al., 2009). Both control and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice learned the task and improved their performance over the 7 d period. However, success rates for the PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice were significantly lower than those for control mice (p < 0.0001, two-way ANOVA) (Fig. 11B). Sema5A−/−; Sema5B−/− mice failed to learn the task, and their success rates were significantly lower than control mice (p < 0.0001, two-way ANOVA) (Fig. 11B).
Finally, we used an accelerating rotarod to examine motor learning and coordination (Daily et al., 2011). Control mice showed progressive improvement on the rotarod test over the course of eight trials (Fig. 11C). In contrast, both Sema5A−/−; Sema5B−/− and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice failed to improve during the trials (Fig. 11C). Importantly, the falling latencies for both double-mutants were significantly shorter than that of control mice (Sema5A−/−; Sema5B−/−, p < 0.0001; PlexA1fl/fl; PlexA3−/−-Emx1-Cre, p = 0.0001, two-way ANOVA) (Fig. 11C). Together, these data reveal motor behavioral deficits in Sema5A−/−; Sema5B−/− and PlexA1fl/fl; PlexA3−/−-Emx1-Cre mice, likely because of the anatomic and physiological defects in CSN wiring in these double-mutant mice.
Discussion
How axon fasciculation or bundling plays a role in neural circuit formation and function is an enduring question in the mammalian nervous system development. In particular, axons fasciculate in long tracts and then leave the tracts by defasciculation to enter target regions and make synapses. Although we now have a better understanding of how axon fasciculation occurs at the molecular level (Honig et al., 1998; Bonanomi and Pfaff, 2010; Miller et al., 2010; Raper and Mason, 2010; Huettl et al., 2012; Tan et al., 2017), the effects of aberrant axon defasciculation on neural circuit wiring and function remain largely unknown.
In this study, we addressed how CS axon bundle integrity in the medulla is molecularly determined and how defects in this axon fasciculation in the medulla affects CS circuit formation and function. First, we showed that PlexA1/A3 expressed by CS axons, likely interacts with Sema5A/5B which are expressed during early postnatal development in the medulla (Fig. 12). Deletion of either PlexA1/3 or Sema5A/5B leads to premature defasciculation in the medulla and early termination of CS axons in the cervical rather than lumbar spinal cord (Fig. 12). This results in a rewiring of standard cortical output, in which CSNs in the hindlimb motor cortex aberrantly innervate cervical spinal gray matter, and no longer drive proper electrical activity in the lumbar spinal cord, thus removing functional CS inputs to the motor neurons controlling hindlimb muscles. However, we still cannot exclude the possibility that axon pathfinding deficits (instead of axon fasciculation defects) in the medulla cause CS circuit defects in the absence of Sema5A/5B-PlexA1/A3 signaling. For example, some CS axons may aberrantly project to incorrect regions in the medulla. Therefore, it remains unclear what exactly causes the defects in CS axons in the absence of Sema5A/5B-PlexA1/A3 signaling. Considering the expression of Sema5A/5B in the medulla, they could function as either repulsive or attractive molecules. Moreover, it has been shown that Sema5A is expressed by proprioceptive sensory neurons in the DRG and controls sensory-motor specificity in the spinal cord (Poliak et al., 2016). Therefore, it remains unclear whether and how much defects in sensory-motor circuits influence CS circuits until we analyze the conditional mutant mice.
Schematic illustration of defects in corticospinal axon fasciculation and segmental specificity of corticospinal axonal projections in adult Sema5A/5B and PlexA1/A3 double-mutant mice. A, B, Schematic diagrams of sagittal sections of the medulla at P0-P2 when the fasciculation of CS axons occurs during axon decussation. The bilateral expression of Sema5A/B may have repulsive qualities (A blue allow) to promote tight bundling at the dorsal turning site of CS axons. If Sema5A/5B-PlexA1/A3 signaling is lost at this site (B), CS axons may innervate more rostral regions in the medulla, leading to weak bundling and greater axon straying. These wayward axons may ultimately form synaptic connections in aberrant locations of the spinal cord.
A key feature of CS connectivity is somatotopic organization, with CS axons that originate in the hindlimb motor cortex projecting preferentially to the lumbar spinal cord and those in the forelimb motor cortex projecting to the cervical spinal cord. These axons carry signals that are crucial for individual limb motor performance and coordination. This likely requires the proper guidance of axon growth cones during embryonic development to ensure that cortical outputs target specific areas and do not exit too early and thus fail to innervate the further, more caudal spinal cord. Both chemical and mechanical factors likely play a role in proper axonal targeting.
Recent studies have identified key molecules (Crim1 and Kelch-like 14) that control differential projections of CSNs to the cervical versus lumbar spinal cord (Sahni et al., 2021a, 2021b). In addition, we show that CS axon fasciculation, or bundling, helps maintain directionality and reduce early termination of axons by preventing them from finding synaptic targets. Thus, deficits in defasciculation in the medulla could result in nonsomatotopic projections from the motor cortex. Axon fasciculation and defasciculation may aid proper targeting of axon terminals over long distances during neural development.
Interestingly, in the absence of Sema5A/5B-PlexA1/A3 signaling, ectopic CS axons from the hindlimb motor cortex terminated at cervical levels, and they seem to be segregated from CS axons from the forelimb motor cortex (Fig. 7F). This suggests that these distinct CS axons express some cues or cell-surface molecules to prevent potential intermixing of their axon fibers. Another interesting observation is that CS axons that normally exclusively innervate cervical and thoracic levels extend ectopically into the lumbar spinal cord in both Sema5A/Sema5B and PlexA1/A3 double-mutant mice. This may indicate that Sema5A/5B-PlexA1/A3 signaling has an additional role in the dorsal funiculus. However, electrophysiological analysis did not show functional connectivity between the forelimb motor cortex and the hindlimb muscle in the absence of Sema5A/5B-PlexA1/A3 signaling (Figs. 9, 10), suggesting that the connectivity is not functional or not strong enough for detection by this electrophysiological assay. These data show that Sema5A/5B-PlexA1/A3 signaling is involved in somatotopic CS innervation of the spinal cord in WT mice. Moreover, we detected behavioral deficits in both individual forelimb use and forelimb-hindlimb coordination, suggesting that the addition of CS synapses to the cervical spinal cord generates improper integration of signals for dexterous forelimb movement, and that the loss of input to the lumbar spinal cord contributes to a loss of functional coordination in the hindlimbs.
In normal development, a bundle of CS axons travels within the dorsal funiculus, then enters the spinal gray matter and arborizes despite the continued expression of PlexA1/A3 on CS axons and Sema5A/5B in the spinal cord during the early postnatal period. There may be several different explanations for this process. For example, CS axons might passively exit from the CS axon bundle within the dorsal funiculus without any specific molecular programming. Or, an attractive chemical gradient begins promoting axonal arborization during the early postnatal phase. This could result in attractive gradients forming of other guidance cues over time and space in the spinal cord, or differential ratios of repulsive and attractive cues forming over time that could affect axon growth and arborization in the spinal cord. Ryk-Wnt signaling has been shown to function as guidance cues through a rostro-caudal spatial gradient in the spinal cord (Liu et al., 2005). This signal could additionally help promote posterior-bound fasciculation and its gradual dissipation may cause greater CS axon arborization in the posterior spinal gray matter. Alternatively, expression of PlexA1/A3 coreceptors may change over development. Npn-1 continues to be expressed until at least P7 (Z. Gu et al., 2019), but previous work in the subiculum indicated that VEGFR-2 could act as a coreceptor that switches Sema3E/PlexD1 signaling from repulsive to attractive (Bellon et al., 2010). A similar mixture of attractive and repulsive functions may allow CS axons to be stimulated to exit the tract and then be guided into the spinal gray matter at appropriate levels. VEGFR-2 mRNA expression was recently shown to gradually decrease during postnatal development in rat spinal motor neurons (Glaesel et al., 2020). Further research is needed to determine whether these or other factors, including the localization of signals, changes during postnatal development. Another possibility is that the effects of plexin intracellular signaling may change during development, leading to less repulsion of axons during later developmental stages.
Previously, we showed that PlexA1 mutant mice have better manual dexterity in forelimb tasks (Z. Gu et al., 2017a). This behavioral enhancement is likely mediated by the preservation of monosynaptic connections between CS axons and spinal motor neurons that are normally eliminated in rodents by PlexA1-Sema6D signaling (Z. Gu et al., 2017a). That is, a behavioral enhancement occurs when a structural feature is rescued from elimination. As we have knocked out an additional signaling receptor, PlexA3, in this study, we have generated deficits in CS axon tract structure and spinal targeting, which likely contributed to the behavioral deficits observed here. Thus, morphologic deficits seem to lead to behavioral deficits. Behaviors that require forelimb and hindlimb coordination were affected, such as beam-walking and rotarod learning, as was reaching for food pellets, a predominantly forelimb task. This suggests that excess innervation of the cervical spinal cord by larger portions of the motor cortex could lead to an inability to refine cortical input to the spinal cord, which may lead to inappropriate muscle activity during behavior.
Finally, why do CS axon bundles stay in the dorsal funiculus of the spinal cord? Proper axon fasciculation and defasciculation at the right time and place in the CNS may be necessary for appropriate neural circuit formation. Bundled CS axons may move more easily within the dorsal funiculus compared with the gray matter where other axons or neurons could inhibit normal CS axon extension via a variety of axon guidance cues. Spinal neurons or other cell types in the spinal gray matter may produce axon attractants and repellents that could interfere with appropriate CS axon targeting. Indeed, in Sema5A/5B or PlexinA1/A3 mutant mice, CS axons from the forelimb motor cortex aberrantly terminate at the cervical spinal cord, suggesting that an axon guidance cue for CS axons could be causing them to terminate prematurely. The typical gradient of attractive and repellant signaling near the CST may allow for only a few axons to exit the CST and terminate at each spinal level along the anterior-posterior axis. Sema5A/5B or PlexinA1/A3 mutant mice, whose CS axons defasciculate too early in the medulla, develop a larger number of axons that may bypass this regulation in the spinal cord, resulting in premature termination patterns in the cervical spinal cord.
Footnotes
This work was supported by the Structural and Functional Imaging Core at Burke Neurological Institute and National Institutes of Health S10 Shared Instrumentation Grant OD028547-01. Y.Y. was supported by National Institute of Neurological Disorders and Stroke Grants NS100772, NS115963, NS119508, and NS093002. We thank Carol Ann Mason, Alexandra Rebsam, and Vibhu Sahni for critical comments on the manuscript; and Ryota L. Matsuoka, Sun Lu, Roman J. Giger, and Alex L. Kolodkin for providing Sema5 reagents and mice.
The authors declare no competing financial interests.
- Correspondence should be addressed to Yutaka Yoshida at yoy4001{at}med.cornell.edu