Functional Local Proprioceptive Feedback Circuits Initiate and Maintain Locomotor Recovery after Spinal Cord Injury  Aya Takeoka, Silvia Arber  Cell Reports  Volume 27, Issue 1, Pages 71-85.e3 (April 2019) DOI: 10.1016/j.celrep.2019.03.010 Copyright © 2019 The Author(s) Terms and Conditions

Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 1 Adult-Induced PA Ablation Permanently Alters Locomotor Ability (A) Genetic strategy to ablate proprioceptive afferents using PVcre::AdvilliniDTR mice with diphtheria toxin. Pan-somatosensory neuron expression of advillin allows for the conditional expression of human diphtheria toxin receptor (DTR) in PAs by intersectional crossing with PVCre mice. (B) DTX eliminates tdTomatoON/PVON neurons and their tdTomatoON/vGlut1ON terminals in the spinal cord, respectively, from PVcre::AdvilliniDTR::RosaLSL-tdTomato mice. This manipulation does not affect AdvillinOFF/PVON spinal interneurons or PVOFF/vGlut1ON cortical and other non-proprioceptive somatosensory terminals. (C) Percentage of PVON DRG neurons at different spinal levels. Cervical, thoracic, and lumbar segments are pooled into groups for quantification: C1–3, C4–6, C7–8, T1–4, T5–9, T10–13, L1–2, L3–4, and L5–6 segments. Normalized vGlut1 density in cervical, thoracic, and lumbar gray matter in control and PVcre::AdviDTR mice after DTX injection (n = 3 mice per group). (D) Marked joints for kinematic reconstruction and timeline of locomotor tests. (E) PCA for volitional overground locomotion (15–25 gait cycles per mouse, n = 4 pre- and post-DTX). (F) Distance between pre- and each post-ablation time point in the PC space is not different. (G) The maximum locomotor speed achieved on a motorized treadmill overtime. Data are represented as means ± SEMs. See also Figure S1. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 2 Local PAs Are Needed for High-Speed Interlimb Coordination (A) Ablation of PAs using PVcre::AdviDTR::RosaLSL-tdTomato mice with spatially confined intraspinal delivery of diphtheria toxin leading to segmentally restricted ablation of tdTomatoON/PVON DRG neurons. (B) Percentage of PVON DRG neurons at different spinal segments (n = 3 mice per group). (C) PCA for volitional overground locomotion (15–25 gait cycles per time point per mouse, n = 4 each for controls, cervical- and lumbar-centric ablation groups). Histogram plot shows mean values of PC1 scores for segment-specific ablation. (D) Both cervical- and lumbar-centric ablation groups show limited speed adaptability. (E) 3D plots of right hindlimb (RHL), right forelimb (RFL), and left forelimb (LFL) phase values referencing left hindlimb (LHL) gait timing. A single dot represents one analyzed gait cycle. The more centered a dot is placed in the 3D space, the more tightly coordinated each limb is to LHL. 3D plots show that interlimb coordination deteriorates in segment-specific PA ablation groups, but not proprioceptionnull group at maximum speed (n = 4, 4, and 3, mice with cervical- or lumbar-specific PA ablation and proprioceptionnull mice, respectively, 20–30 steps per mouse per speed). Pre- versus post-ablation: cervical-centric 10 cm RHL p = 0.07, RFL p = 0.20, and LFL p = 0.11; cervical-centric maximum speed RHL p < 0.001, RFL p < 0.001, and LFL p < 0.01; lumbar-centric 10 cm RHL p = 0.2, RFL p = 0.06, and LFL p = 0.08; lumbar-centric maximum speed RHL p < 0.001, RFL p < 0.001, and LFL p < 0.001; i.p. 10 cm RHL p = 0.30, RFL p = 0.15, and LFL p = 0.12; i.p. maximum speed RHL p = 0.58, RFL p = 0.78, and LFL p = 0.21. Asterisks denote statistical difference between pre- and post-ablation groups. Data are represented as means ± SEMs. See also Figure S2. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 3 PAs below the Lesion Are Critical for Locomotor Recovery (A) Experimental timeline following left lateral hemisection at T10. (B) Representative stick decomposition of pre-injury, 3 days and 7 weeks post-injury. (C) PCA was applied to 100 gait parameters measured during volitional overground locomotion from pre-, 3 days, and 7 weeks post-lesion time points (15–25 gait cycles per left hindlimb per mouse, n = 4 each for all groups). Gait patterns of ipsilesional hindlimb of each mouse are represented as individual dots in the new space created by PC1–2. (D) Histogram plot shows mean values of PC1 scores for each experimental group. (E) Control versus cervical-centric ablation and their factor loading of individual parameter (individual dots) to PC1 show high correlation, capturing the similarity of the locomotor recovery process between the two groups. (F) Diagram illustrating ΔG-rabies virus injection strategy to retrogradely label cervical and thoracic descending neurons projecting to ipsilesional lumbar spinal cord (yellow). Quantification of % ipsilesional neuron to total number (ipsilesional + contralesional neurons) of spinal neurons (n = 3 each for intact and acute control and PVcre::AdviDTR mice for each ablation condition, n = 4 each for chronic control and PVcre::AdviDTR mice for each ablation condition) displaying the percentage of ipsilesional neurons of entire rabies-marked population. Asterisks indicate the statistical difference from the percentage of ipsilesional rabiesON neurons found in control mice at the chronic condition. Data are represented as means ± SEMs. See also Figure S3. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 4 Proprioceptive Feedback Is Required for the Initiation and Maintenance of Recovery (A) Timeline of DTX administration (i.p.) and locomotor analysis during recovery. Histogram plots show mean values of PC1 scores at different time points (n = 4 for 7-week delayed ablation, n = 3 for control groups; 100 parameters per gait cycle). (B) Percentage of ipsilesional hindlimb dragging during stance phase at different time points. (C) A retrograde tracing approach with ΔG-rabies labels descending spinal, magnocellular, and gigantocellular brainstem projection neurons that innervate ipsilesional lumbar spinal circuits. (D) A representative stick decomposition for hindlimb movement during volitional and treadmill locomotion is shown for acute and chronic time points for mice with PA ablation before injury and 7 weeks post-injury (bottom: joint oscillation; gray bars, stance; brown bars, dragging). (E) Timeline of hemisection with respect to PA removal time point. Three representative parameters that capture the facilitation of locomotion on the treadmill at a chronic time point (12 weeks post-lesion) in mice with delayed PA ablation, compared to volitional locomotion. TM, treadmill locomotion; RW, volitional locomotion on a runway. Data are represented as means ± SEMs. See also Figure S4. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 5 PA Input Reorganization to Motor Neurons Is Specific after Lateral Hemisection Injury (A) Scheme of lateral hemisection at T10 level and how it affects ipsilesional and contralesional LVe access to lumbar spinal segments combined with ΔG-rabies, with fluorescent tag in Sol muscle. (B) vGlut1ON input increases exclusively to ipsilesional but not to contralesional soleus motor neurons (n = 3 mice per condition, 7–11 motor neurons [MNs] analyzed per mouse). (C) No change in PA input to ipsilesional soleus MNs in Egr3 mutants following injury (n = 3 per group, 5–8 MNs analyzed per mouse). (D) Scheme of specific muscle targeting with ΔG-rabies (soleus) and CTB (GS) to map heteronymous input from GS afferents to soleus MNs. Quantification of vGlut1ON/CTB ON (heteronymous) to vGlut1ON/CTBOFF (homonymous and other unidentified heteronymous sources) ratio using the number of inputs at the intact condition as a reference data point (n = 3 mice per group, 6–11 MNs analyzed per mouse). Scheme of PA input increase to soleus MNs, both from heteronymous and likely homonymous sources. (E) Scheme of specific muscle targeting with ΔG-rabies (soleus) and CTB (TA) to map antagonistic PA input from TA afferents to soleus MNs (n = 3 mice per group, 4–8 MNs analyzed per mouse). (F and G) Within the soleus (F) and gluteus (G) MN pools, PA input increases to MMP9OFF slow MNs but not to MMP9ON fast MNs (n = 3 per group, 5–9 MNs analyzed per mouse). (H) Quantification of L5–6 lumbar LMC ChATON α MNs (n = 3 per group, 15–25 MNs analyzed per mouse). (I) The increase in PA input to soleus MNs is not due to the change in MMP9ON/OFF ratio after spinal cord injury (n = 3 per group). (J) Summary scheme representing the change in PA terminals onto MN pools. Data are represented as means ± SEMs. See also Figure S5. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 6 PA Input Reorganization to Premotor Neurons Is Specific after Lateral Hemisection (A) Injection scheme to visualize spinal premotor network innervating a specific muscle group (TA or GS) using ChATcre mice. ΔG-rabies viruses with different fluorescent tags were used. (B) Ipsilesional premotor organization in intact or hemisected spinal cord. (C and D) PA input to ipsilesional TA premotor neurons at the level of both somata (C) and dendrites (D; n = 3 mice per premotor group, 6–9 neurons analyzed per mouse). (E) PA input quantification to ipsilesional Renshaw cells (n = 3 mice per group, 21–28 neurons analyzed per mouse). Data are represented as means ± SEMs. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions

Figure 7 Plasticity of Locomotor Circuit Functions after PA Ablation Is Age Dependent (A) Timeline of DTX administration to eliminate PA neurons. DTX administration time points are shown as 1, 2, and 3. (B) Least-squares elliptical fitting (95% confidence) was computed in a new space created by PC1–2. The histogram plot shows the mean values of PC1 scores for age-dependent ablation. Experimental groups consist of n = 4 each for P14 DTX and their control littermates, n = 4 pre- and post-adult DTX (the same dataset as Figure 1), n = 5 P1 ablation, and n = 3 control littermates. (C) The highest speed achieved for each experimental group. Data are represented as means ± SEMs. See also Figure S6. Cell Reports 2019 27, 71-85.e3DOI: (10.1016/j.celrep.2019.03.010) Copyright © 2019 The Author(s) Terms and Conditions