For this study, we employed a two-virus synaptic silencing strategy developed by Tadashi Isa and colleagues (Kinoshita et al., 2012). This system involves a highly efficient lentiviral vector (HiRet) with a tetracycline response element upstream of enhanced tetanus neurotoxin (eTeNt) and EGFP (eTeNT.EGFP) that infects neuron terminals. Next, an AAV2/2 virus that contains a tetracycline transactivator (rtTAV16) is delivered which infects neuron cell bodies. Doxycycline (Dox) then induces the production of eTeNT/EGFP which cleaves the vesicular docking protein VAMP2 thus preventing neurotransmitter release and silencing the neurons based only on their anatomy, any/only neurons with terminals exposed to the lentiviral vector and cell bodies exposed to the AAV2/2 will be silenced, as employed previously (Pocratsky et al., 2017; Pocratsky et al., 2020; Shepard et al., 2021). To silence the LDPNs, we performed bilateral injections at L2 and C6 spinal cord segments to doubly infect LDPNs at their terminals and cell bodies, respectively (Figure 1A). Only doubly-infected neurons that constitutively express rtTAV16 will then express enhanced tetanus neurotoxin (eTeNT) in the presence of Dox. At the level of the cell terminal, active eTeNT prevents synaptic vesicle release, leading to ‘silenced’ neurotransmission (Figure 1B). Removing Dox restores functional neurotransmission, allowing for reversible silencing of inter-enlargement LDPNs both before and after SCI (Figure 1C). We examined the coupling patterns of the limb pairs by comparing the temporal relationship, which can be expressed as a phase value. As described in our previous work (Pocratsky et al., 2017; Pocratsky et al., 2020), phase is determined by dividing the initial contact time of the trailing limb by the stride time of the leading limb. Phase values of 1 indicate synchrony and values of 0.5 indicate alternation of the limb pair. We determined the mean phase value of the limb pairs and any value >2  SD from this mean was considered ‘irregular,’ as defined by the blue boxes. All the locomotor assessments were done on two different walking surfaces as previously described: bare acrylic with a coefficient of friction (CoF) of 0.44 and Sylgard-coated acrylic with a CoF of 1.73, representing surfaces with relatively poor grip and extremely good grip, respectively (Pocratsky et al., 2020).

When the LDPNs were silenced in otherwise intact animals, we observed a disruption of right-left hindlimb and forelimb alternation that was, in many respects, similar to those observed for LAPN silencing. The two limb pairs demonstrated a substantial proportion of irregular steps and the disruption involved a similar bias towards the hindlimbs, with the forelimb pair showing less significant disruption (Figure 2A). However, in contrast to the highly context-dependent LAPN phenotype, LDPN silencing disrupted left-right hindlimb alternation for both the hindlimbs and forelimbs and occurred on both the low CoF (Figure 2a–c; small yellow circles; Videos 1 and 2) and the high CoF (Figure 2A, D and E; small red circles; Videos 3 and 4) surfaces. Also distinct from what we saw with the LAPNs, a second round of Dox administration led to greater disruptions in the alternation of both limb pairs, as well as the contralateral hindlimb-forelimb pair, such that there were a greater number of irregular steps during Dox2 than during Dox1, and again these disruptions occurred on both surfaces (Figure 2B–E; Figure 2—figure supplement 1; Pocratsky et al., 2020). During silencing, the disruption to interlimb coordination included modest but significant changes in heterolateral/contralateral and homolateral/ipsilateral forelimb-hindlimb coupling, as observed previously when silencing the LAPNs (Figure 2—figure supplement 1; Pocratsky et al., 2020). Based on the magnitude of the effect and the apparent hierarchy of impact, we consider the disruption of forelimb-hindlimb coupling to be secondary to the partial de-coupling of the forelimb and hindlimb pairs (see discussion). The silencing effects on limb pair relationships were restored when Dox was removed from the drinking water (Figure 2B–E, ‘PostD1’).

We further examined whether silencing non-SCI animals would adversely affect the coordination of the limbs in time (temporal measures) and space (spatial measures). Swing time, stance time, stride time, stride frequency, and stride distance comprise well-described gait indices that have been shown to have specific relationships with speed (Gillis and Biewener, 2001). As observed previously for the LAPNs, silencing LDPNs did not affect the fundamental gait relationships of swing, stance, or stride times regardless of walking surface (Figure 3A–C and F–H). Stride distance and frequency were also unaffected (Figure 3D, E, I and J). It is interesting to note that overall, even at control time points, the LDPN animals exhibited greater variability in many of the outcome measures, as compared to the LAPNs, but both groups showed a small, but statistically significant increase in mean speed during silencing regardless of stepping surface (Figure 3K). Thus, these findings suggest that both the LAPNs and LDPNs have roles in securing interlimb coordination but that removing them from the circuitry does not affect velocity-dependent step cycle characteristics or the locomotor rhythm in an otherwise intact rat.

In addition to stance, swing, and stride times, interlimb coordination involving the homolateral and heterolateral HL-FL pairs is a fundamental speed-dependent gait characteristic. These components of interlimb coordination were investigated by plotting the phases of each limb pair against each other at the control and Dox time points, and for each stepping surface. This comparison was not made previously for the LAPNs, so we include those results here (see Pocratsky et al., 2020; Shepard et al., 2021 for LAPN methods). During control time points, steps were clustered in the center of the plot for the FL/FL and HL/HL comparisons, indicating that alternating forelimb steps were concomitant with alternating hindlimb steps (Figure 2—figure supplement 1C and I). Similarly, alternating hindlimb steps were clustered with synchronous/alternate forelimb steps, contralateral or ipsilateral, respectively (Figure 2—figure supplement 1A, B, G and H). This held true for control testing on both the low and high CoF surfaces. These phase plots demonstrate that during silencing the HL and FL pairs could adopt any possible phase value, but with some over-riding control system that maintained stable stepping even as the two limb pairs were partially decoupled. This can most easily be discerned by noticing that certain combinations of HL/HL – FL/FL phase values simply didn’t occur (Figure 2—figure supplement 1C, F, I and L) even though each pair expressed the entire possible range of phase values. The spaces in these graphs, most notably centrally in D & J, and middle top and bottom in E & K, also support the concept that forelimb-hindlimb coordination was only disrupted when the forelimb and hindlimb pairs were partially de-coupled and thus was a secondary rather than primary effect of silencing. In addition, there are several specific regions in these phase-phase plots that highlight the differences between the silenced phenotypes for the LAPNs and LDPNs. First, for the LAPNs, there were no perturbations in left-right alternation on the low CoF surface as illustrated by the closely clustered clouds of yellow markers in Figure 2—figure supplement 1A, E and F. In contrast, Figure 2—figure supplement 1J, K and I demonstrate that phase was disrupted for both walking surfaces when the LDPNs were silenced. Second, the appearance of steps with HL/HL and FL/FL phases ~0.25 & 0.75 is clear in Figure 2—figure supplement 1F as compared to 1 l, given the smaller number of steps analyzed for the LAPN animals. This is also reflected by comparing Figure 2—figure supplement 1E and K. In k, it is apparent that the ipsilateral HL-FL pair maintained a phase value around 0.5 even while the HL-HL phase varied throughout the entire range of values on both walking surfaces. Finally, no differences in the base-of-support were detected between the low and high CoF surfaces during silencing for either pathway, suggesting that balance/postural changes likely do not account for these fascinating results (LAPN Dox: 18.36±2.97° vs 21.44±4.48°, p>0.05, n.s., paired t-test; LDPN Dox: 18.95±7.41 vs 20.11±11.91, p>0.05, n.s., paired t-test).

Finally, we examined the effects of LDPN silencing on left-right hindlimb coordination in water (Figure 4A). Swimming is a bipedal task where the hindlimbs provide the major propulsive force while the forelimbs steer (Gruner and Altman, 1980). During swimming, the limbs are unloaded and the proprioceptive and cutaneous feedback associated with stepping is altered (Miller and van der Burg, 1973; Duysens and Stein, 1978; Akay et al., 2014). In contrast to our overground findings, and in keeping with our previous LAPN study (Pocratsky et al., 2020), silencing the LDPNs had no effect on left-right hindlimb alternation during swimming (Figure 4B and C). This observation suggests that the LDPN locomotor phenotype is not rigidly context-independent. Thus, the circuitry responsible for securing alternation during swimming remains unperturbed when either set of long propriospinal neuron is silenced.

LDPNs with cell bodies in the cervical enlargement and axon terminals in the more rostral segments of the lumbar cord have axons that travel in the ventrolateral funiculus (VLF) (Menétrey et al., 1985; Reed et al., 2006; Flynn et al., 2017). Lesions to the thoracic VLF in the cat disrupt forelimb–hindlimb coupling (Brustein and Rossignol, 1998), further confirming the location of these long descending propriospinal projections. Given this location, it was essential to determine whether the VLF, the area of white matter in which LDPN axons project, was preserved after SCI. To examine this, animals were euthanized during post-injury Dox administration. Their spinal cords were harvested and were blocked in three sections: caudal cervical spinal cord (cell bodies), caudal thoracic spinal cord (injury epicenter and penumbra), and rostral lumbar spinal cord (cell terminals). Thoracic tissue sections were stained at the injury epicenter using eriochrome cyanine (EC) to quantify to amount of spared white matter and imaged to determine where the spared white matter was located. Amongst all animals, the average spared white matter percentage (SWM%) was 26%, slightly higher than that previously described in the LAPN SCI study (Figure 5A). The proposed location of LDPN axons within the VLF was spared, as evidenced by the remaining spared white matter in each animal (Figure 5B–I). Thus, a proportion of LDPN axons are likely spared after SCI, a critical finding for the interpretation of any behavioral assessments moving forward.

Knowing that LDPN axons were likely preserved at the level of injury, we next confirmed that eTeNT.EGFP-expressing putative LDPN axons were present in the rostral lumbar spinal cord, as well as eTeNT.EGFP-expressing putative LDPN cell bodies in the intermediate gray matter of the caudal cervical segments. Histological analysis for EGFP immunoreactivity showed that putatively positive eTeNT.EGFP fibers were found to surround neuronal processes in the rostral lumbar enlargement (Figure 6A). eTeNT.EGFP co-localized with synaptophysin (Figure 6B; synaptic marker), vesicular GABA transporter (Figure 6D, VGAT, inhibitory neurotransmitter), and vesicular glutamate transporter 1/2 (Figure 6E; VGlut2, excitatory neurotransmitters). For D and E, arrows indicate two neuron terminals shown in X, Z and Y, Z focal planes in the boxes below and to the right of these photomicrographs. Isotype controls revealed minimal-to-no immunoreactivity (Figure 6C).

Using the same immunohistological protocol as described for lumbar spinal tissue, caudal cervical spinal cord segments were assessed for the presence of eTeNT.EGFP within LDPN somata. Putatively eTeNT-EGFP-positive LDPN cell bodies co-localized with fluorescent Nissl stained (NeuroTrace) neurons in the intermediate gray matter (Figure 6f–H, Reed et al., 2006). Isotype controls showed no immunoreactivity (Figure 6I–K). Taken together with the lumbar spinal cord histology, these data suggest that double-infected LDPNs maintained expression of eTeNT at the level of both the cell soma and axon terminals following SCI. Furthermore, any post-SCI spared LDPNs axons were functionally silenced, indicating that any behavioral changes seen during Dox administration were concomitant with active eTeNT.EGFP expression.

Having validated that (1) the viral-based silencing system is active post-SCI and (2) LDPN axons are likely intact post-SCI, we began to explore the effects of LDPN silencing on locomotion after injury. In our previous study of the LAPNs, we found that moderate contusive injuries at T10 resulted in disruptions of hindlimb alternation that were not dissimilar to the effects of LAPN silencing. However, silencing LAPNs post-SCI unexpectedly resulted in improved stepping, and in particular restored hindlimb alternation so that it was not significantly different from the uninjured state (Shepard et al., 2021). Thus, given that LDPN silencing also resulted in disrupted right-left alternation, we hypothesized that silencing LDPNs after a T10 SCI would again result in improved recovered locomotion.

The Basso, Beattie, Bresnahan (BBB) Locomotor Rating Scale was used to grossly evaluate locomotor recovery (Basso et al., 2002). Control BBB scores (post-injury, pre-silencing) were concentrated around a score of 12, consistent with previous literature for this injury severity and with our previous LAPN silencing study (Shepard et al., 2021; Basso et al., 2002; Smith et al., 2006). Analogous to our LAPN study, average BBB scores were modestly increased during silencing (Figure 7A), with a greater proportion of raw scores for each limb between 13–18 (Figure 7B). The increased BBB scores were due to improved weight support of the hindlimbs as well as better ‘coordination,’ resulting in the appearance of improved stepping during silencing. Some BBB scores remained around 13, indicating that not all animals’ stepping improved at all Dox time points. However, a significant proportion of BBB scores were increased, demonstrating that better stepping occurred in multiple animals at multiple time points as a result of silencing.

To further assess the overall locomotor function post-SCI and the impact of LDPN silencing, we evaluated the step sequence patterns using the regularity index (RI) and coordinated pattern index (CPI). As previously described, RI scores plantar steps according to paw placement order and represents the gait as the ratio of plantar steps that are in order (one of four different orders that normal rats exhibit) over the total number of steps, while CPI represents the number of correctly patterned step cycles (dorsal and plantar) to the total number of cycles (dorsal and plantar) (Hamers et al., 2001; Caudle et al., 2015). CPI accounts for whether the animal can achieve coordination regardless of their ability to attain plantar placement, while RI only accounts for correctly patterned plantar steps. RI and CPI were both modestly improved during Dox administration (Figure 7C and D). We also examined the plantar stepping index (PSI), an index that simply compares the number of hindlimb plantar steps as compared to forelimb plantar steps, and it was also modestly improved during silencing (Figure 7E).

When using RI and CPI, an important differentiator is the presence of dorsal steps, which occur when the walking surface is contacted by the dorsum of the foot and/or toes because they remain flexed or curled at stance. Dorsal steps are common after SCI, especially acutely, and are one important measure of recovery from low-thoracic contusive SCI (Basso et al., 1996). Dorsal stepping index (DSI) is simply the ratio of dorsal steps to the number of total hindlimb steps, thus the DSI in normal uninjured animals should be zero. We compared DSI during post-injury control and silenced time points. The DSI was significantly reduced from 24.31±21.02–9.64±8.70 during silencing indicating a significant reduction in the overall proportion of dorsal steps (Figure 7F). This decrease closely resembles the decrease seen in dorsal stepping indices during LAPN post-SCI silencing.

We also compared the sidedness of dorsal steps. Dorsal step-sidedness refers to the percentage of dorsal steps that occur on the right and the percentage that occurs on the left. The sidedness did not change between control and silencing, with ~55% of dorsal steps occurring in the right hindlimb and ~45% of dorsal steps occurring in the left hindlimb during both control and silenced time points (Figure 7G). This is an important observation, as it shows that one limb did not improve to a greater extent than the other. For the purposes of simplicity and ease of comparison with pre-injury data, we utilize the left limb as lead to demonstrate the principle behavioral changes post-SCI, as we did with LAPN SCI data. To further elucidate these improvements in coordination indices, we examined more specific intralimb kinematics and interlimb gait analyses.

We next examined intralimb coordination using the peak-trough excursions and the temporal relationship between the proximal (iliac crest – hip – ankle) and distal (hip – ankle – toe) angles of the three-segment, two-angle kinematic hindlimb model we employ (Figure 8A–D; Videos 5 and 6). Interestingly, the excursions of both the distal and proximal angles were not significantly improved during silencing (Figure 8E and F). Next, we explored the temporal relationship between the proximal and distal angles. In uninjured animals, the peak extension of the distal angle typically occurs almost in-phase with the peak extension of the proximal angle during the stance phase of the step cycle, resulting in a coordination value ranging from 0.9 to 0.1 on a linear scale. Following injury, the average coordination value of these two angles was unchanged between Control and Dox, even while the variability was somewhat reduced (Figure 8G), indicating that silencing the LDPNs had a negligible effect on intralimb coordination post-SCI.

After SCI, the temporal relationship between hindlimb and hindlimb-forelimb limb pairs (HL-HL and HL-FL) is highly variable, even as the forelimbs maintain alternation. This variability is accounted for, in part, by the presence of dorsal steps. However, many plantar steps also fall into irregular gait patterns after SCI. We included both plantar and dorsal steps for the purposes of quantifying interlimb coordination, with left dorsal steps identified as teal throughout the remaining figures. As was seen during LAPN silencing, the hindlimb-hindlimb pair and the hindlimb-forelimb pairs are partially decoupled at all post-SCI Control time points (Figure 9A, gray circles). We found that the coupling of the hindlimbs was significantly restored during the initial round of post-injury silencing (D2D5 & D2D8). Heterolateral coupling (HL-FL) was also improved, but reached significance only on D8. Interestingly, the improvement in HL-HL coupling was not significant at any Dox3 time point, suggesting that there may be some plastic changes occurring within the circuitry over this 3 week period from D2D5 to D3D14. Importantly, hindlimb coupling is seen as significantly improved when Control and Dox time points are collapsed, suggesting that the total number of steps that fall outside normal variability are reduced overall by LDPN silencing from ~23 – ~12% (Figure 9A). In addition, the coupling of the heterolateral hindlimb-forelimb pair was also significantly improved overall, when the control and Dox time points were collapsed (Figure 9B), even though the absolute reduction in abnormal steps was modest (from ~43 – ~37%; Figure 9B). Finally, forelimb-forelimb alternation was maintained during both control and Dox time points (Figure 9C).

Two animals were removed from the dataset prior to injury as they showed no perturbations to left-right alternation at any pre-injury Dox time point (Figure 9—figure supplement 1A-D, see Methods for exclusion criteria). Interestingly, these animals showed an injury-induced disruption and did not show improvements in coordinated stepping during LDPN silencing post-SCI, suggesting that the pre- and post-SCI changes are due to LDPN silencing (Figure 9—figure supplement 1).

To summarize the findings on interlimb coordination, post-injury silencing significantly improved hindlimb-hindlimb alternation, had only a modest effect on the hindlimb-forelimb relationship and no impact on right-left forelimb coordination, which remained alternating throughout after injury. Thus, the injury eliminated the impact of LDPN silencing on forelimb circuitry even though the cell bodies of the neurons being silenced are rostral to the injury. In contrast, in our previous work, we showed that silencing the LAPNs after injury improved the interlimb coordination (alternation) of the hindlimbs with essentially no impact on homo- or heterolateral interlimb (HL-FL) coupling (Shepard et al., 2021). Finally, the reduced prevalence of dorsal steps during post-injury silencing of both LAPNs and LDPNs accounted for some, but not all, of the improvement in hindlimb interlimb coordination.

Populations of descending cervico-lumbar projections have been implicated as essential for postural stability during overground locomotion in the uninjured mouse (Ruder et al., 2016). To determine if this holds true for the LDPNs post-SCI, we examined measures that are associated with balance and postural control. Uninjured animals typically have a relatively narrow base of support. In conditions with increased postural instability, such as SCI, the paws become externally rotated and the base of support widens (Basso et al., 1996). Thus, we looked at gait angle, the angle made by three subsequent hindlimb placement points (R-L-R), diagonal length (distance from L hindpaw to R forepaw placement), and rear track width (distance between R and L hindpaw placements) and found that all three were modestly improved during silencing (Figure 10A–C).

We next challenged the animals’ ability to maintain balance and posture by testing them on a horizontal ladder task (Figure 10D). To successfully traverse the ladder apparatus, animals must maintain postural control to accurately place their feet on the fixed-space ladder rungs. The number of foot slips, or times that the foot doesn’t maintain contact with a rung, was reduced for both the left and right hindlimbs (Figure 10D) during post-injury LDPN silencing.

In the uninjured animal, silencing did not affect the fundamental relationships between speed and spatiotemporal features of limb movements (e.g. swing time, stance time, etc.). However, these relationships are disrupted by the SCI model chosen, such that a substantial proportion of dorsal and some plantar steps fall outside the typically described relationship (Figure 11A-E). Silencing LDPNs restored these relationships and resulted in a dramatic reduction in the variability of abnormal steps, regardless of dorsal or plantar stepping (Figure 11F–J). Importantly, average swing (Figure 11K) and stance times (Figure 11L) were unchanged during post-SCI silencing as was the duty cycle (Figure 11M). Finally, the average locomotor speed was modestly increased during post-injury silencing (Figure 11N) suggesting that speed may be a factor in the improved spatiotemporal relationships. Together with the previous findings, these data suggest that silencing LDPNs after SCI positively influences the temporal coordination of multiple aspects of locomotion, primarily interlimb (HL-HL) coordination. It is imperative to note that while the pattern of improvements observed with LDPN silencing post-SCI is very similar to those seen when the LAPNs were silenced, the overall magnitude of improvement is modest by comparison, where the LAPN silencing induced a more robustly improved phenotype.

Animals were housed two per cage under a 12 hr light/dark cycle with ad libitum food and water throughout the course of the study. Power analysis for gait measures revealed that n=6–10 could detect a true significant difference with a power of 85–95% and an alpha of 0.05. Additional animals were used to mitigate animal mortality following repeat surgical exposure. A total of n=19 adult female Sprague-Dawley rats (215–230 g) were used in this study (Figure 1c). After viral vector injection and initial Dox administration (‘Dox1’; n=19), animals were split into two groups. Animals were assigned into the uninjured (n=8) and injured groups (n=11) with their cage mate after Dox1 administration and behavioral assessments. The uninjured group received one additional round of Dox (Dox2) lasting 13 days and were subsequently euthanized. In the injured group, n=1 animals died as a result of SCI injury before post-injury behavioral assessments. Eight of 10 remaining animals displayed a locomotor phenotype pre-SCI. The two animals that did not display behavioral changes were removed from the main dataset to be analyzed separately as a post-injury control. These animals had fewer than 10% of their steps outside of normal variability during Dox testing. Animals that lacked behavioral outcomes pre-injury (n=2) did not show improvements in locomotion post-SCI. However, both animals had mild control injury phenotypes, clouding deeper interpretations of post-injury silencing data for these animals.

Viruses were constructed and titered following previously described methods (Pocratsky et al., 2017; Pocratsky et al., 2020).

Intraspinal injections and power analyses for kinematic measures are based on previous literature. The procedural details are described in the Nature Protocol Exchange (Pocratsky et al., 2018).

Animals received a moderate spinal cord contusion at T9/T10 spinal cord approximately 2 weeks after the conclusion of uninjured Dox^ON assessments. For SCI surgery, animals were re-anesthetized (ketamine:xylazine:acepromazine, 40 mg/kg:2.5 mg/kg:1 mg/kg; I.P., Henry Schein Animal Health, Dublin, OH; Akorn Animal Health, Lake Forest, IL). Spinal cord contusion injuries were performed as described in our previous paper (Shepard et al., 2021).

Doxycycline hydrochloride (Dox, 20 mg/ml; Fisher Scientific BP2653-5, Pittsburgh, NH) was dissolved in 3% sucrose and provided ad libitum for 8 days pre-injury and 8–14 days post-injury. Dox water was made fresh and replenished daily and monitored for consumption. All behavioral assessments were performed during the light cycle portion of the day and concluded several hours before the dark cycle began.

Prior to SCI, behavioral assessments were performed prior to viral injections (BL), prior to Dox (PD1), during Dox (D1D5-D1D8), and 14 days post-Dox (PostD1). For the uninjured LDPN animals, behavioral assessments were performed again prior to the second Dox administration (PD2) and during Dox (D2D5, D2D8, D2D13). Following SCI in the injured group, pre-Dox and Dox^ON time point assessments were reproduced twice (Dox2 and Dox3) following SCI to assess the reproducibility of any behavioral changes that were seen. Post-injury control time points included assessments prior to Dox (PD2, PD3), during Dox (D2D5, D2D8, D3D5, D3D8, and D3D14), and after the second Dox administration (PostD2). Some data shown are compiled from pre-injury Control and Dox time points and post-injury Control and Dox time points. Control vs Dox uninjured and Control vs Dox injured time point comparisons were made both on an individual and group basis. Behavioral analyses began on day 5 of Dox (DoxD5) administration and were repeated on Dox day 8 (DoxD8). For terminal assessments after injury, behavioral assessments occurred on Dox Day 14 (D3D14).

Hindlimb joint identification and hindlimb kinematics were acquired as previously described (Pocratsky et al., 2020; Shepard et al., 2021). The analyzed points from the two sagittal cameras were exported to a Microsoft Excel workbook and 2D average angles were calculated for each digitized frame. Maximum and minimum angles (maximum extension and flexion, respectively) were identified using a custom Microsoft Excel macro. Excursion for the proximal and distal hindlimb joint angles (Maximum Angle – Minimum Angle) were also calculated using the same macro. The temporal relationship between the proximal and distal hindlimb joint angles was calculated using the peak-to-peak duration of the lead angle during a single-step cycle. Within this duration, the maximum excursion of the first angle was determined and depended on which angle peaked first (proximal or distal angles). The time of onset of the second angle was divided by the peak-to-peak distal angle duration to determine the temporal relationship between intralimb angles. A coordination value of 1 indicates in-phase coordination of the proximal and distal intralimb joints, while a phase value of 0.5 indicates anti-phase, uncoordinated joint movements. Intralimb joint phase was calculated for each step cycle of the left and right hindlimbs independently.

Overground gait analyses were performed as described previously (Shepard et al., 2021). During recordings of injured animals (control and silenced), a step was classified as dorsal when the dorsum of the foot came into contact with the ground during the stance portion of the step cycle and is considered a step if it maintains contact with the surface and completes the swing portion of the step cycle. For both uninjured and injured conditions, gait analysis was always performed relative to the left hindlimb (i.e. RLRR and RLFR). Consequently, for the overground gait (phase) analysis dorsal steps were identified only for the left hindlimb and any dorsal steps on the right side would not be identified in the analyzed dataset. In separate analysis (CPI, DSI, right-left dorsal step ratio) dorsal steps for both right and left hindpaws were considered (see Figure 7).

Interlimb phase was calculated by dividing the initial contact time of the trailing limb by the stride time (initial contact to initial contact) of the leading limb. Alternating gaits were defined by phase values concentrated around 0.5 (Lemieux et al., 2016).

Phases for individual steps were transformed from a circular scale and were plotted linearly. Blue boxes on graphs represent >2 standard deviations (SD) as calculated from the uninjured control average and SD for the specified limb pair. Values found within the blue box are considered outside of normal variability and were quantified using pie charts to indicate the percent of total steps that existed outside of that range. Plantar steps are indicated by gray circles during control and red circles during silenced time points. Post-injury dorsal steps were indicated on both circular and linear phase plots as teal circles.

BBB assessments were performed by individuals blinded to experimental time points. All raters were aware that assessments were being performed post-injury, but were blind to Control and Dox. Overground stepping was assessed using the BBB Open Field Locomotor Scale as previously described (Basso et al., 2002; Caudle et al., 2015).

RI, CPI, PSI, and DSI were calculated as described previously (Shepard et al., 2021).

Several measures were used to determine the postural stability of animals during locomotor bouts. Gait angle is the angle between two consecutive initial contacts of a rear hindlimb in reference to the other hindlimb. Typically, a smaller gait angle would indicate higher stability during locomotion, as the hindlimbs are in the line with the body. Larger gait angles would suggest diminished stability as the hindlimbs would be set wider. Gait angle is calculated using the Pythagorean theorem and the X, Y coordinates of the foot while it is in contact with the ground. Statistics were performed on the group means for the left and right hindlimbs (bars: average ± SD; circles: individual means).

The diagonal length is the distance between the initial contacts of diagonal paws. For this calculation, the diagonal pair used was left hindlimb–right forelimb. This pair was chosen to remain consistent with hindlimb phase data in which the left hindlimb is the lead limb. The diagonal length is calculated using the Pythagorean Theorem and the X, Y coordinates of each foot.

Rear track width is the distance between the hindlimbs during consecutive initial contacts. Track width is calculated by taking the absolute value of the Y-coordinate of the left hindlimb initial contact subtracted from the Y-coordinate of the subsequent right hindlimb initial contact. Larger track width indicates a wider stance during stepping, while a smaller track width indicates stance phases more in line with the body.

We quantified the animals’ ability to effectively traverse a ladder with fixed-spacing rungs (Columbus Instruments, Columbus OH, Chen et al., 2012; Metz and Whishaw, 2002). Behavioral testing was performed on the same time points as BBB testing (PD2, D2D5, D2D8, PostD2, PD3, D3D5, D3D8, and D3D13). Each animal received five stepping trials in each direction. The total number of footfalls was calculated for the left and right hindlimbs, respectively, for each animal across the time points. We then calculated each animal’s average number of foot slips during the Control and Dox time points listed above. Left and right limbs were not combined to demonstrate that the number of foot slips decrease by a similar amount for each hindlimb (i.e. one foot is not different from the other in terms of recovery). Statistics were performed on the group means (bars: average ± SD; circles: individual means).

Spatiotemporal gait analyses were performed as described previously (Shepard et al., 2021).

Animals were killed on D3D14 (n=10) following terminal BBB and kinematic behavioral assessments. Animals were overdosed with sodium pentobarbital, followed by pneumothorax and transcardial perfusion with 0.1  M phosphate-buffered saline (PBS) (pH 7.4) followed by 4% paraformaldehyde diluted in PBS solution. Spinal cords were dissected, post-fixed for 1.5 hr, and transferred to 30% sucrose for a minimum of 4 days at 4 °C. Spinal segments C5-C8, T8-T12, and T13-L3/L4 were dissected, embedded in tissue freezing medium, and stored at –20 °C until they were cryosectioned at 30 µm.

Histological analysis was performed using the immunohistochemistry protocol described previously (Shepard et al., 2021).

Statistical analyses were performed using the SPSS v22 software package from IBM. Additional references for parametric and non-parametric testing were used in complementation to SPSS (Hays, 1981; Siegel and Castellan, 1988; Ott and Longnecker, 1977). Differences between groups were deemed statistically significant at p≤0.05. Two-tail p-values are reported.

The Binomial Proportion Test was used to detect significant differences in the proportion of coordination values beyond the control threshold for the raw and transformed interlimb coordination data of various limb pairs prior to and post-SCI. It was also used to determine statistical significance for per-step changes in left-right coordination and change in interlimb phase, raw BBB score differences, intralimb phase and per-step changes in intralimb phase, dorsal steps as a percentage of total steps, and percentage of categorically organized steps (anti-phase, out of phase, in phase).

Regression analyses were used to compare speed versus spatiotemporal gait indices datasets, including speed vs. swing time, stance time, stride time, and stride distance. Analyses were performed for hindlimb-hindlimb relationships prior to and after SCI. For regression analyses post-SCI, plantar, and dorsal steps were included in the analysis and dorsal steps are shown in blue on graphs for identification. Trend lines are shown on graphs with 95% prediction intervals indicated by dashed lines.

Mixed model analysis of variance (ANOVA) followed by Bonferroni post hoc t-tests (where appropriate) were used to detect a significant difference in the BBB scores based on sidedness and condition (i.e. Control vs Dox; graph not shown). Repeated measures ANOVA analyses were used to compare the number of average steps per animal during uninjured and injured time points.

Paired t-tests were used to detect significant differences in proximal and distal angle excursion for uninjured and injured intralimb coordination, average gross stepping measures including RI, CPI, PSI, DSI, and combined BBB scores, average intralimb phase, percentage of dorsal step sidedness, average swing time, average stance time, average duty cycle, average number of ladder errors, average gait angle, average diagonal length, average rear track width, and overall average speed at Control and Dox combined time points.

Circular statistics were performed on the raw phase data to analyze phase distribution of forelimb, hindlimb, heterolateral and homolateral limb pairs as described previously (Pocratsky et al., 2017). Time point comparisons were performed using the nonparametric two-sample U2 tests. The p-values are reported as ranges based on critical values of Watson’s U2 as reported in Table D.44 of Zar, 1974.