Synaptic imbalance and increased inhibition impair motor function in SMA

Abstract

Movement is executed through balanced excitation-inhibition in spinal motor circuits. Short-term perturbations in one type of neurotransmission are homeostatically counteracted by the opposing type, but prolonged excitation-inhibition imbalance causes dysfunction at both single neuron and circuit levels. However, whether dysfunction in one or both types of neurotransmission leads to pathogenicity in neurodegenerative diseases characterized by select synaptic deficits is not known. Here, we used functional, morphological, and viral-mediated approaches to uncover the pathogenic contribution of unbalanced excitation-inhibition in a mouse model of spinal muscular atrophy (SMA). We show that vulnerable SMA motor circuits fail to respond homeostatically to reduced excitation and instead increase inhibition. This imposes an excessive burden on motor neurons and further restricts their recruitment. Reducing inhibition genetically or pharmacologically improves neuronal function and motor behavior in SMA mice. Thus, the disruption of excitation-inhibition homeostasis is a major maladaptive mechanism that diminishes the capacity of premotor commands to recruit motor neurons and elicit muscle contractions in SMA.

Fig. 1.. Reduced activation of Renshaw cells by proprioceptive synapses at the onset of SMA.

(A) Excitatory postsynaptic potentials (EPSPs) in Renshaw cells following ventral root stimulation in WT (blue) and SMA (red) mice. (B) Superimposed EPSPs from (A) at an expanded timescale. The maximum monosynaptically induced EPSP amplitude was measured at 3 ms from its onset (vertical dashed line and arrows). (C) Amplitude of ventral root (VR)-induced EPSPs between WT (n = 8 Renshaw cells, N = 8 mice) and SMA (n = 9 Renshaw cells, N = 9 mice) Renshaw cells. P = 0.62, unpaired two-tailed t test. (D) EPSPs in Renshaw cells following dorsal root (DR) stimulation in WT (blue) and SMA (red) mice. (E) Superimposed EPSPs from (B) at an expanded timescale. Similar to (B), the EPSP amplitude was measured at 3 ms from its onset. (F) Values of DR-induced EPSP amplitude in WT (blue; n = 5, N = 5) and SMA (red; n = 6, N = 6) Renshaw cells. *P = 0.0115, unpaired two-tailed t test. Point of stimulation is denoted by a black arrow. (G) Superimposed traces of voltage responses (top traces) to current injections (bottom traces) in a WT and a SMA Renshaw cell. (H) Resting membrane potential (RMP), input resistance (R_IN), rheobase (I_rh), voltage threshold (V_Th), time constant (τ), and capacitance for WT and SMA Renshaw cells. *P < 0.05, all unpaired two-tailed t tests (WT: n = 5, N = 5; SMA: n = 6, N = 6; P = 0.0251 in R_IN, P = 0.0395 in I_rh, and P = 0.0376 in τ, respectively). MΩ, megohms. (I and J) Recorded cells intracellularly filled with Neurobiotin (Nb; visualized in blue) in WT (I) and a SMA (J) mice together with immunoreactivity against VAChT⁺ (red) and VGluT1⁺ (green) synapses (respective arrows) in apposition onto their somata (insets) and dendrites (dashed box and insets).

Fig. 2.. Decrease in synaptic density of proprioceptive synapses on Renshaw cells at SMA onset.

(A) Spinal cord image showing calbindin (green), VAChT (red) immunoreactivity, and retrogradely filled motor neurons (MNs; blue) in one hemicord at P3. (B_1-4) Higher-magnification images of retrogradely filled MN axon collaterals (B₁, blue), a calbindin⁺ Renshaw (B₂, green), VAChT⁺ (B₃, red), and merged (B₄). (C) Neurolucida reconstruction of a WT Renshaw (green) with VAChT boutons (red). (D) Density of VAChT⁺ synapses on the somata of WT and SMA Renshaw cells at P3. Each dot represents one Renshaw (n = 10 cells per mouse per genotype). (E) As in (A) but for a P3 SMA mouse. (F_1-4) as in (B_1-4) for an SMA spinal cord. (G) Neurolucida reconstruction of a SMA Renshaw cell, similar to (C). (H) Density of VAChT⁺ synapses on the dendrites of WT and SMA Renshaw cells at P3 (n as in D). No significance in (D) or (H) between the groups (unpaired two-tailed t test). (I) Image of a WT spinal cord showing unilateral retrograde fill of MNs (blue), calbindin (green), VGluT1 (red), and Pv (white) immunoreactivity. (J_1-5) Images of MN axon collaterals (J₁, blue), a calbindin⁺ Renshaw (J₂, green), VGluT1 (J₃, red), Pv (J₄, white), and merged (J₅). Red arrows in J₅ denote proprioceptive synapses (VGluT1⁺ and Pv⁺) on Renshaw cell. (K) Neurolucida reconstruction of Renshaw shown in J_1-5 with VGluT1⁺/Pv⁺ synapses (red). (L) VGluT1⁺/Pv⁺ synaptic density on the somata of WT and SMA Renshaw cells at P3; *P = 0.0165, unpaired two-tailed t test (WT: n = 19 cells, N = 2 mice; SMA: n = 29, N = 3). (M) SMA spinal cord at P3, as in (I). (N_1-5) Images of a SMA Renshaw cell, as in (J_1-5). (O) Neurolucida reconstruction of a SMA Renshaw with VGluT1+ synapses, as in (K). (P) VGluT1⁺/Pv⁺ synaptic density on the dendrites of the same Renshaw cells; *P = 0.0245, unpaired two-tailed t test. n.s., not significant.

Fig. 3.. Unexpected increase of VGluT1⁺ synaptic coverage on Renshaw cells at the end stage of SMA.

Low-magnification images of the ventral horn showing iliopsoas (IL) motor neurons, labeled by retrograde muscle injection with CTb488 (green), calbindin (blue), and VGluT1 (red) immunoreactivity in a WT (A_1-3) and a SMA (B_1-3) mouse, at P11. Single plane confocal images of a calbindin (blue) Renshaw cell and VGluT1⁺ synapses (red) in the WT (A₂) and a SMA (B₂) mouse, from the dashed boxes in (A₁) and (B₁). Neurolucida reconstruction (A₃, white) of WT Renshaw cell shown in (A₂), and a SMA Renshaw cell (B₃, white). VGluT1⁺ synaptic appositions are indicated as red dots. (C and D) VGluT1⁺ synaptic density increased on the somata (C) and dendrites (D) of SMA Renshaw cells (n = 29 cells, N = 3 mice) compared to WT (n = 29 cells, N = 3 mice); somata: **P = 0.0044, unpaired two-way t test; dendrites: **P = 0.0029, unpaired two-tailed t test. Statistical comparisons were between the average for each mouse in each genotype. (E) Experimental protocol for the spinal cord transection performed in the thoracic segment T4/5 at P8, followed by morphological examination of Renshaw cells located in the L1/2 spinal segments at P10. (F) Low-magnification image of a transverse section of a spinal cord labeled with calbindin (blue), VGluT1 (red), and VAChT (green) antibodies. The Renshaw areas are shown bilaterally in the dashed oval circles. High-magnification confocal images of a WT (G) and a SMA (H) Renshaw cell (blue), together with VGluT1 (red) and VAChT (green) immunoreactivity after spinal transection. (I) VGluT1⁺ synaptic density is decreased on the somata of SMA Renshaw cells at P11 (n = 17, N = 3) compared to WT (n = 14, N = 3); **P = 0.0065, unpaired two-tailed t test. (J) VGluT1⁺ synaptic density is also decreased on the dendrites; *P = 0.0127, unpaired two-tailed t test.

Fig. 4.. Putative Ia inhibitory interneurons are hyperexcitable and receive fewer proprioceptive synapses in SMA at the disease onset.

(A) Superimposed voltage responses (top) to current injections (bottom) for a putative WT (left) and a SMA (right) Ia inhibitory interneuron. Input resistance (B), rheobase (C), voltage threshold (D), and EPSP amplitude (E) in WT (blue; n = 3 neurons, N = 3 mice) and SMA (red; n = 3, N = 3) putative Ia inhibitory interneurons were significantly different; *P = 0.048 in (B), *P = 0.019 (C), and *P = 0.049 (E), unpaired two-tailed t test. The small “n” was compensated by large effect sizes and Hedges’ g calculated used estimating statistics and bootstrapping (see Materials and Methods). MΩ, megohms. (B) Mean difference of 59.8% increase in SMA [95% confidence interval (CI), 24.6 to 90.8%]; Hedges’ g = 1.83. (C) Mean difference 63.4% decrease in SMA (95% CI, 88.9 to 37.9%); Hedges’ g = −2.47. (D) No significant (n.s.) difference was observed. (E) Mean difference 41.5% decrease in SMA (95% CI, 65.5 to 17.5%); Hedges’ g = −1.82. To the right of each graph, we show the bootstrap distribution of sample differences with the mean difference (black dot) and 95% confidence limits (bars). P values (orange) show results of permutation t test. (F and G) Low-magnification images from one side of the spinal cord of a WT (F_1-5) and a SMA (G_1-5) mouse, showing Foxp2 (white, F₁ and G₁), NeuN (blue, F₂ and G₂), calbindin (red, F₃ and G₃), VGluT1 (green, F₄ and G₄) immunoreactivity, as well as the merged image (F₅ and G₅). (H and I) Higher magnification of Ia inhibitory interneurons, showing calbindin⁺ and VGluT1⁺ synapses in a WT (H_1-5) and a SMA (I_1-5) mouse. (J) The number of VGluT1⁺ synapses on the soma of Ia inhibitory interneurons in WT (n = 12 cells, N = 3 mice) and SMA (n = 13 cells, N = 3 mice) mice differ at P4; **P = 0.004, unpaired two-tailed t test. Statistical comparison performed between WT and SMA mice.

Fig. 5.. SMA motor neurons receive increased inhibitory synaptic drive.

(A) Repetitive firing following current injection at 40 pA above threshold in a WT and a SMA Renshaw cell at P4. (B) Average firing frequency after current injection in WT (n = 5 cells, N = 5 mice) and SMA (n = 5 cells, N = 5 mice) Renshaw cells at P4. **P = 0.009 (10-pA step), *P = 0.013 (20-pA step), and *P = 0.045 (30-pA step), unpaired two-tailed t test. (C) Spontaneous voltage activity in a WT and a SMA Renshaw cell at P4. (D) Spontaneous firing frequency in WT and SMA Renshaw cells. *P = 0.0377, unpaired two-tailed t test (WT: n = 5 cells, N = 5 mice; SMA: n = 8 cells, N = 8 mice). (E) Pharmacologically isolated mIPSCs in a WT and SMA motor neuron located from the L2 spinal segment at P4. Amplitude (F) and frequency (G) of mIPSCs in WT (n = 3 cells, N = 3 mice) and SMA (n = 3 cells, N = 3 mice) L2 motor neurons at P4. *P = 0.016 and **P = 0.004, unpaired two-tailed t test. Small n was compensated by large effect sizes and Hedges’ g calculated used estimating statistics and bootstrapping (see Materials and Methods). (F) Mean difference 228.5% increase in SMA (95% CI, 170.3 to 293.5%); Hedges’ g = 3.72. (G) Mean difference 101.1% increase in SMA (95% CI, 68.8 to 150.8%); Hedges’ g = 2.58.

Fig. 6.. SMA motor neurons are covered by a higher number of GABAergic and glycinergic synapses.

(A to C) Single plane confocal images from a WT (A), SMA (B), and SMA::Pv^CRE (C) motor neuron (MN) showing GAD65/67 (red), gephyrin (green), and ChAT (blue) immunoreactivity at P4. Insets at the bottom are higher-magnification images from the dashed boxed area, showing the individual fluorochromes. (D to F) Single plane confocal images from a WT (D), SMA (E), and SMA::Pv^CRE (F) MN showing GlyT2 (red), gephyrin (green), and ChAT (blue) immunoreactivity at P4. Insets are higher-magnification images from the dashed boxed area, as in (A) to (C). (G) The number of GAD65/67 synapses per 10 μm of MN membrane in WT (n = 33 MNs, N = 4 mice), SMA (n = 24 MNs, N = 4 mice), and SMA::Pv^CRE (n = 20 MNs, N = 3 mice) MNs. ***P < 0.0001, WT versus SMA; ***P = 0.0004, SMA versus SMA::Pv^CRE; one-way analysis of variance (ANOVA), Tukey’s post hoc test. (H) The number of gephyrin clusters associated with GAD65/67⁺ synapses per 10 μm of MN membrane in WT and SMA MNs (“n” and “N” identical as in G). ***P < 0.0001, WT versus SMA; ***P = 0.0001, SMA versus SMA::Pv^CRE; one-way ANOVA, Tukey’s post hoc test. (I) The number of GlyT2 synapses per 10 μm of MN membrane in WT (n = 30 MNs, N = 4 mice), SMA (n = 26 MNs, N = 4 mice), and SMA::Pv^CRE (n = 16 MNs, N = 3 mice) MNs. *P = 0.032, SMA versus SMA::Pv^CRE; ***P = 0.0008, WT versus SMA, one-way ANOVA, Tukey’s post hoc test. (J) The number of gephyrin clusters associated with GlyT2⁺ synapses per 10 μm of MN membrane in WT and SMA MNs (n and N identical as in I). *P = 0.023, SMA versus SMA::Pv^CRE; **P = 0.0018, WT versus SMA; one-way ANOVA, Tukey’s post hoc test. Statistical comparison was performed between the average values from mice.

Fig. 7.. Knockdown of gephyrin in vivo abolishes the enhancement of inhibitory synapses in SMA MNs and provides phenotypic benefit in SMA mice.

(A_1-4) Single plane images of WT (A_1,2) and SMA (A_3,4) motor neurons (MNs) transduced with AAV9-GFP (A_1,3) or AAV9-Gephy_RNAi-GFP (A_2,4). Gephyrin clusters/MN soma (B) for four groups shown in (A). In each group N = 3 mice; WT + AAV9-GFP (n = 10 MNs), WT + AAV9-Geph_RNAi-GFP (n = 14 MNs), SMA + AAV9-GFP (n = 9 MNs), and SMA + AAV9-Geph_RNAi-GFP (n = 10 MNs). **P = 0.0049, WT versus WT + AAV9-Geph_RNAi; ***P = 0.0006, SMA versus SMA + AAV9-Geph_RNAi; one-way ANOVA, Tukey’s post hoc test. Amplitude (C) and frequency (D) of mIPSCs from L1/2 MNs (n = 3 MNs per group) at P3/4 (N = 3 mice per group). In (C): *P = 0.018, WT versus SMA; *P = 0.035, SMA versus SMA + AAV9-Geph_RNAi; one-way ANOVA, Tukey’s post hoc test. In (D): *P = 0.049, WT versus SMA; *P = 0.021, SMA versus SMA + AAV9-Geph_RNAi; one-way ANOVA, Tukey’s post hoc test. Significance confirmed by evaluating effect sizes. (C) The estimated mean difference: WT versus SMA showed 251.2% increase in mIPSC amplitude (95% CI, 115.8 to 357.0%) and Hedges’ g = 2.12; the difference, WT versus AAV9-Geph_RNAi, not significant (5.26; 95% CI, −2.79, 10.3; Hedge’ g = 0.92). (D) Estimated mean difference WT versus SMA showed 164.9% increase in mIPSC frequency (95% CI, 131.3 to 224.5%) and Hedges’ g = 3.61; the difference, WT versus AAV9-Geph_RNAi, not significant (−0.61; 95% CI, −2.11, 1.14; Hedges’ g = −0.39). Estimation statistics are plotted as Fig. 4. (E_1-3) in vivo EMGs from IL muscle during righting in WT + AAV9-GFP, SMA + AAV9-GFP, and SMA + AAV9-Geph_RNAi-GFP mice (at P10). Expanded traces are shown (red bar). Blue arrows indicate righting. (F) EMG burst duration during righting. ***P < 0.0001, WT + AAV9-GFP versus SMA + AAV9-GFP; SMA + AAV9-GFP versus SMA + AAV9-Geph_RNAi-GFP; one-way ANOVA, Tukey’s post hoc test. (G) EMG amplitude. *P = 0.0167, WT + AAV9-GFP versus SMA + AAV9-GFP; *P = 0.0500, SMA + AAV9-GFP versus SMA + AAV9-Geph_RNAi-GFP mice; one-way ANOVA, Tukey’s post hoc test. (H) Righting time: WT + AAV9-GFP (N = 10 mice), SMA + AAV9-GFP (N = 9), and SMA + AAV9-Geph_RNAi-GFP (N = 15). *P < 0.05 and **P < 0.01, unpaired two-tailed t tests. (I) Righting in vivo Org-25543 treatment. *P < 0.05, **P < 0.01, and ***P < 0.001, unpaired two-tailed t tests, SMA versus SMA + Org-25543 for individual ages (WT, N = 17 mice; SMA, N = 7; WT + Org, N = 7; SMA + Org, N = 9).

Fig. 8.. Neuronal circuit changes conferring an increase in tonic inhibitory drive on motor neurons in a severe SMA mouse model.

Under healthy conditions, lumbar motor neurons (gray) receive excitatory-glutamatergic synaptic drive from proprioceptive fibers (green) and inhibitory synapses from Renshaw cells (red) and Ia inhibitory interneurons (yellow). In SMA, vulnerable motor neurons receive decreased proprioceptive synaptic drive while their inhibitory synapses from Renshaw cells and Ia inhibitory interneurons increase. Furthermore, Renshaw cells receive higher than normal excitation from corticospinal glutamatergic synapses (blue/green).