Descending Brainstem Systems Contribute to Ankle Clonus in Humans with Spinal Cord Injury

Abstract

Ankle clonus is a sustained, involuntary, rhythmic muscle contraction frequently observed in humans with spinal cord injury (SCI). Although its pathophysiology remains incompletely understood, converging evidence suggests a role for brainstem systems in its generation. Following SCI, brainstem neuromodulatory inputs partially compensate for the loss of descending motor pathways by regulating motoneuron excitability during involuntary contractions, suggesting their involvement in the generation of clonus. To test this hypothesis, motoneuron excitability in response to Ia synaptic input was quantified using the soleus H-reflex and maximal motor response (H/M ratio). Brainstem involvement was assessed using the long-latency component of the cutaneous reflex (LLR), which is influenced by monoaminergic pathways, and the StartReact response, in which reductions in reaction time to a loud acoustic stimulus are thought to reflect descending drive from the reticular formation, measured in the tibialis anterior and soleus muscles. We studied males and females with chronic SCI, both with and without ankle clonus, using standardized clinical tests across 2 d. Participants with clonus showed elevated H/M ratios, indicating increased motoneuron excitability, whereas those without clonus exhibited lower values than controls. Additionally, individuals with clonus exhibited longer LLR duration and greater LLR magnitude in both muscles, along with shorter reaction times to startle stimuli, consistent with enhanced monoaminergic and reticulospinal contributions. Notably, LLR duration was positively correlated with both StartReact response and H/M ratio. Together, these findings suggest a potential contribution of descending brainstem systems to the maintenance of clonus after chronic SCI.

Introduction

Clonus is a common manifestation of upper motor neuron lesions, including those resulting from spinal cord injury (SCI; Therkildsen et al., 2025). It is characterized by involuntary, rhythmic muscle contractions, typically occurring at 3–8 Hz, most often at the ankle joint (Little et al., 1989; Bravo-Esteban et al., 2013). Clonus can be elicited by passive muscle stretch or during voluntary movement and is associated with reduced quality of life in individuals with SCI (Fleuren et al., 2009; Manella et al., 2013; Khan et al., 2016). Although its underlying mechanisms remain incompletely understood, converging evidence indicates contributions from both central and peripheral systems. This view is supported by findings that clonus frequency can remain stable despite variations in external load (Walsh, 1976; Dimitrijevic et al., 1980) and proprioceptive input (Beres-Jones et al., 2003). In contrast, other studies report that clonus frequency and amplitude scale with externally applied loads (Rack et al., 1984; Rossi et al., 1990), supporting models of reflex instability within a closed-loop sensorimotor feedback system (Hidler and Rymer, 1999, 2000). Despite these insights, the neural mechanisms underlying ankle clonus remain poorly understood. Differentiating clonus from generalized stretch hyperreflexia is important when evaluating spasticity after SCI using standardized assessments (Therkildsen et al., 2025).

Motoneuron properties are closely associated with the generation of clonus. Early studies proposed that clonus arises from self-reexcitation within hyperactive stretch reflex pathways (Clare et al., 1951; Cook, 1967; Lippold, 1970; Hagbarth et al., 1975; Nichols et al., 1978; Rack et al., 1984; Latash et al., 1989; Rossi et al., 1990). Computational models have shown that reducing latency in the reflex pathway—either by decreasing conduction time delays or by replacing the soleus slow-twitch muscle characteristics with those of fast-twitch muscles—can attenuate clonus (Hidler and Rymer, 1999). Motoneuron excitability is modulated by monoaminergic descending pathways, which facilitate intrinsic depolarizing currents known as persistent inward currents (PICs; Eken et al., 1989; Heckman et al., 2008; Hultborn et al., 2013). PICs amplify and prolong synaptic inputs, enabling sustained motoneuron firing and prolonged muscle contractions (Hounsgaard et al., 1988; Heckmann et al., 2005). Both animal (Bennett et al., 2001; Murray et al., 2010) and human (Gorassini et al., 2004; Murray et al., 2010) studies suggest that the long-lasting component of the cutaneous reflex (LLR) after SCI is influenced by multiple factors, including afferent drive, spinal network excitability, and motoneuron properties associated with PICs (D’Amico et al., 2014).

Descending brainstem pathways can modulate the excitability of motoneurons and interneuronal circuits and may influence their responsiveness to supraspinal motor tracts involved in voluntary movement through fast glutamatergic inputs (Rekling et al., 2000; Hultborn et al., 2003). Indeed, following upper motor neuron lesions, including stroke, increased excitability of motoneurons innervating the paretic limb may arise from both ionotropic reticulospinal and neuromodulatory mechanisms (McPherson et al., 2018a,b). In individuals with SCI, altered reticulospinal contributions have also been suggested, particularly in those exhibiting spasticity and hyperreflexia (Sangari and Perez, 2020; De Santis and Perez, 2026), although the underlying mechanisms remain incompletely understood. Therefore, we hypothesized that brainstem systems may contribute to the regulation of motoneuron excitability and may be involved in the maintenance of involuntary muscle contractions observed during ankle clonus after SCI.

Materials and Methods

Subjects.

Thirty-three individuals with chronic SCI (≥1 year postinjury; mean age 47.5 ± 16.0 years; 28 males, 5 females) participated in the study. All participants provided written informed consent, and study procedures were approved by the Northwestern University Institutional Review Board in accordance with the Declaration of Helsinki. Neurological level of injury was determined using the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) examination, with all injuries at or above T10. Of the 33 participants, five were classified by the American Spinal Cord Injury Impairment Scale (AIS) as AIS A (motor and sensory complete), two as AIS B (motor complete, sensory incomplete), and 26 as motor and sensory incomplete (AIS C, n = 8; AIS D, n = 18; Table 1). To minimize the effects of antispasmodic or GABAergic medications on muscle tone and neural excitability, participants were asked to withhold these medications for at least 12 h before testing. Clonus was assessed using the drop test and manual stretch test by a trained examiner (Koelman et al., 1993; Manella et al., 2013). The drop test provides reliable measures of clonus that correlate with both the Spinal Cord Assessment Tool for Spastic Reflexes (SCATS) clonus score (Benz et al., 2005) and reflex responses in individuals with SCI (Manella et al., 2013, 2017). Tests were performed in a standardized position (see below) and repeated approximately five times on 2 separate days. Clonus severity was classified using SCATS criteria (Benz et al., 2005): 0, no reaction (<1 s); 1, mild clonus (<3 s); 2, moderate clonus (3–10 s); and 3, severe clonus (>10 s). Based on these criteria, participants were classified as non-clonus (<1 s; 0.35 ± 0.21 s, range 0–0.68 s; n = 14) or clonus (≥10 s, typically >30 s; n = 19). A clonus burst was defined as an EMG burst followed by a silent period (Fig. 1B). The ratio of the soleus maximum H-reflex (H-max) and maximal motor response (M-max; i.e., H/M ratio) was used as a proxy for motoneuron excitability in response to Ia synaptic input. Brainstem monoaminergic involvement was examined using cutaneous reflexes recorded from the tibialis anterior and soleus muscles, focusing on the long-polysynaptic reflex (LPR) and long-lasting reflex (LLR) components. In addition, the StartReact response—an involuntary release of a planned movement evoked by a startling stimulus and thought to reflect reticulospinal engagement (Tapia et al., 2022)—was assessed in participants with residual voluntary control of tibialis anterior and soleus muscles.

Table 1.

Spinal cord injury participants

ID	Age (years)	Sex	AIS	Level	Etiology	Time postinjury (years)	Group	Medication	Cutaneous reflex	Start React	H/M
1	36	M	C	C8	T	10	Clonus	None	X		X
2	34	M	C	C3	T	2	Clonus	None	X	X	X
3	56	M	D	C4	T	2	Clonus	BAC	X	X	X
4	54	M	C	C2	NT	4	Clonus	BAC, GBP	X		X
5	44	M	B	T8	T	4	Clonus	None	X		X
6	41	M	B	T8	T	8	Clonus	BAC, GBP	X		X
7	54	F	C	T10	NT	20	Clonus	BAC, GBP, TIZ	X	X	X
8	41	M	D	C7	T	14	Clonus	None	X	X	X
9	51	M	C	C5	T	12	Clonus	BAC, GBP	X	X	X
10	40	M	C	C2	T	4	Clonus	BAC, TIZ	X		X
11	34	M	D	T6	T	15	Clonus	BAC, GBP	X	X	X
12	38	M	D	C5	T	22	Clonus	None	X		X
13	21	M	D	C6	T	3	Clonus	BAC	X		X
14	23	M	A	C5	T	10	Clonus	BAC, GBP	X		X
15	68	F	D	C2	T	9	Clonus	None	X	X	X
16	51	M	D	C5	NT	22	Clonus	None	X	X	X
17	20	M	D	C4	T	4	Clonus	BAC	X		X
18	46	F	D	C5	T	2	Clonus	BAC	X	X	X
19	55	M	C	C3	T	13	Clonus	BAC, GAB, TIZ		X
20	36	M	A	T4	T	17	No clonus	None	X		X
21	50	M	A	T5	T	7	No clonus	None	X		X
22	77	M	D	C5	T	5	No clonus	GBP	X	X	X
23	78	M	D	C3	NT	3	No clonus	BAC, GBP	X	X	X
24	44	M	A	T10	T	4	No clonus	None	X		X
25	52	M	D	C3	NT	3	No clonus	None	X		X
26	41	M	D	C5	T	4	No clonus	BAC	X	X	X
27	77	M	C	C7	T	35	No clonus	None	X		X
28	21	F	A	T6	T	3	No clonus	None	X		X
29	61	M	D	C4	NT	15	No clonus	None	X	X	X
30	61	F	D	C1	T	8	No clonus	GBP	X		X
31	43	M	D	T8	T	8	No clonus	None	X	X	X
32	75	M	D	C3	T	15	No clonus	None	X	X	X
33	48	M	D	C8	T	9	No clonus	GBP	X	X	X

Demographic characteristics of individuals with spinal cord injury (SCI) enrolled in the study. Completion of each study procedure is indicated by an “X.”

M, male; F, female; T, traumatic; NT, nontraumatic; AIS, American Spinal Injury Association Impairment Scale; BAC, baclofen; GBP, gabapentin; TIZ, tizanidine.

Figure 1.

Experimental set-up. A, Two clinical tests were used to elicit clonus. The leg drop test (left side) involved lifting the leg 10 cm above the platform (blue square) and allowing the metatarsal heads of the foot to contact the platform. The manual stretch test (right side) was performed by the examiner applying a brisk dorsiflexion movement (turquoise arrow). B, Rectified EMG recordings from the soleus muscle in a representative participant with clonus (top trace), showing rhythmic clonic activity lasting over 30 s. For participants with clonus, clonic EMG activity was quantified by measuring burst duration and interburst duration. In contrast, a representative participant without clonus exhibited no rhythmic EMG bursts (bottom trace).

Electromyographic recordings and maximal voluntary contractions.

Electromyographic (EMG) activity was recorded from the tibialis anterior and soleus muscles using bipolar surface electrodes (2-cm-diameter Ag–AgCl; ConMed). Electrodes were placed over the tibialis anterior at one-third of the distance between the fibular head and medial malleolus and over the soleus on the muscle belly along the midline of the leg. A ground electrode was placed over the patella or medial malleolus. EMG signals were amplified (100 times), bandpass filtered (30–2,000 Hz), sampled at 4 kHz, and stored for offline analysis (CED 1401 with Signal software; Cambridge Electronic Design). Participants performed three brief maximal voluntary contractions (MVCs; 3–5 s) in dorsiflexion and plantarflexion, presented in random order and separated by ~60 s of rest. The maximal mean rectified EMG over a 1 s window was calculated for each MVC, and the average across the three trials was used for analysis.

Drop test.

The drop test is a standardized biomechanical method used to quantify ankle clonus by applying a controlled mechanical perturbation to the plantarflexor muscles (Manella et al., 2013; Figs. 1A, 2). Participants were seated with back support, with the hip and knee flexed to ~90°. The metatarsal heads of the test foot were placed on the edge of a 10-cm-high platform, and a padded horizontal bar positioned ~10 cm above the knee defined the release height. To elicit clonus, the examiner lifted the leg just below the knee until it contacted the bar and then released it, allowing the foot to drop onto the platform and produce a rapid dorsiflexion perturbation. Clonus duration was measured with a stopwatch from the onset of oscillatory movement until cessation or manual termination. Responses lasting longer than 30 s were manually stopped for participant comfort. Each session consisted of five trials on Days 1 and 2, separated by at least 30 s of rest to minimize fatigue.

Figure 2.

Clonus and non-clonus. Responses obtained during individual trials in participants with clonus (teal) and without clonus (red). Participants were categorized according to the Spinal Cord Assessment Tool for Spastic Reflexes (SCATS) clonus score (Benz et al., 2005). A rating of 0 indicated no reaction (lasting <1 s; red), a rating of 1 indicated mild clonus lasting <3 s (pink), a rating of 2 denoted moderate clonus persisting for 3–10 s (light teal), and a rating of 3 represented severe clonus lasting >10 s (teal). Based on these criteria, participants were divided into two groups: those with average oscillation durations of <1 s (0.35 ± 0.21 s, range 0–0.68 s; n = 14) and those with oscillations lasting ≥10 s (classified as “clonus,” with oscillations in most participants exceeding 30 s; n = 19).

Manual stretch test.

Clonus was elicited by applying a rapid, manually delivered stretch to the ankle joint (Koelman et al., 1993; Manella et al., 2013). Participants were seated with back support, with the hip and knee flexed to ~90°. The examiner held the participant’s foot at the heel while palpating the Achilles tendon and applied a brisk dorsiflexion to the ankle to trigger clonus (Figs. 1A, 2). Each session also consisted of five trials on Days 1 and 2, separated by at least 30 s of rest to minimize fatigue.

Cutaneous reflex.

Participants were seated with back support, with the hip flexed to 70° relative to the trunk and the knee flexed to 45° from full extension, and the foot secured with straps using a custom device. The reflex was evoked by electrical stimulation of the medial plantar nerve at the mid-arch using bipolar electrodes (2 cm diameter, 2 cm apart; ConMed). A train of 14 pulses (500 μs, 300 Hz) was delivered via a DS7R stimulator (Digitimer). We measured the mean amplitude of the LPR (0–500 ms) and LLR (500 ms to reflex offset), as well as LLR duration. Previous studies have reported that LPR and LLR amplitudes in lower-limb muscles plateau at ~40 mA for incomplete and 80 mA for complete SCI (DeForest et al., 2020). Consistent with these findings, stimulation intensity was set to 40 mA for the 26 participants with incomplete SCI (16 with clonus, 10 without) and 80 mA for the 7 participants with motor complete SCI (3 with clonus, 4 without). Reflex onset, offset, and duration were determined from the average EMG across trials. Onset was defined as the point where rectified EMG exceeded baseline by 5 SD over a 100 ms prestimulus window and offset as the point where EMG remained below this threshold for at least 500 ms (DeForest et al., 2020). Trials in which reflex offset could not be clearly identified due to sustained single motor unit firing were excluded, as this would affect the LLR duration. To enable comparisons of mean amplitudes between individuals with and without (or with shorter) LLRs, mean amplitudes were calculated using a fixed time window defined by the longest reflex duration observed across all participants, ensuring a consistent analysis window across all recordings. For the LPR, a fixed 500 ms analysis window was used. A total of 20 reflexes were acquired from each participant, and the average EMG was used for analysis.

StartReact.

In the same position described above, the StartReact response was assessed in participants with SCI using a previously established protocol (Baker and Perez, 2017). Participants performed rapid isometric ankle dorsiflexion or plantarflexion in response to a 20 ms light-emitting diode (LED) cue placed ~1 m in front of them. Visual reaction time (VRT) was defined as the latency from LED onset to rectified EMG onset in the tibialis anterior and soleus. The LED was presented alone or paired with either a nonstartling acoustic stimulus (80 dB, 500 Hz, 50 ms) or a startling acoustic stimulus (SAS; 115 dB, 500 Hz, 50 ms) via headphones. Stimuli were delivered in random order with inter-trial intervals of 5–15 s. Visual–auditory reaction time (VART) and visual–startle reaction time (VSRT) were defined as the intervals from stimulus onset to EMG onset. Reaction time was determined as the point where EMG exceeded 5 SD above baseline activity measured over 100 ms prestimulus. Each participant completed 20 trials per stimulus condition and contraction type. Three familiarization trials per condition were performed at the start of each session. The VART and VSRT are both mediated via the cochlear nuclei, but only the high-intensity sound of the VSRT activates the reticulospinal pathway (Davis et al., 1982; Brown et al., 1991; Valls-Solé et al., 1999; Tapia et al., 2022). The StartReact effect was estimated as the difference between VART and VSRT. Participants able to perform voluntary contractions participated in the testing (n = 10 clonus; n = 7 non-clonus). StartReact responses were also compared with previously collected data from control participants (n = 19, age 43.3 ± 18.4 years, 14 males, 5 females; manuscript in preparation). The same procedures used to assess StartReact in individuals with SCI were applied to the control participants.

H/M ratio.

The soleus H-max and M-max were evoked by stimulating the posterior tibial nerve at the popliteal fossa (0.2 Hz) using a constant-current stimulator (DS7R, Digitimer). Stimulus intensity was increased in 0.05 mA steps from below H-reflex threshold until H-max was reached and then further increased to elicit M-max. The M-max amplitude was confirmed by increasing stimulation intensity until the M-wave no longer increased in size. The H/M ratio was calculated by dividing H-max by M-max. Values were obtained from almost all participants (n = 18 clonus; n = 14 non-clonus) and compared with previously collected control data (n = 26, age 43.8 ± 14.9 years, 16 males, 10 females; Chen and Perez, 2022). The same procedures used to assess the H/M ratio in individuals with SCI were applied to the control participants.

Clonus analysis.

Clonus was analyzed in the soleus and tibialis anterior using a custom MATLAB script (v.2020a, MathWorks) by measuring burst duration, interburst interval, mean EMG amplitude, and clonus frequency. In the clonus group, 10 s of EMG activity were analyzed, and trials contaminated by spasm (defined as continuous EMG activity for over 1 s) were excluded. Raw EMG was filtered (20–200 Hz, fourth-order IIR), rectified, and smoothed with a 30 Hz low-pass filter. Bursts were identified as peaks in the EMG envelope, with peaks closer than 120 ms (8.5 Hz) excluded based on typical clonus frequencies (4.5–8.4 Hz; Mummidisetty et al., 2012; Wallace et al., 2012). Burst onset and offset were defined as the timepoint at which the envelope exceeded and fell below the mean threshold for ≥25 ms, respectively. Interburst duration was measured from the offset of one burst to the onset of the next, and burst amplitude was calculated as the mean rectified EMG over 20 ms centered on the peak. Clonus frequency was estimated using autocorrelation of the smoothed envelope, defined as the reciprocal of the time lag of the first autocorrelation peak exceeding 0.4. All automated measurements were verified by visual inspection.

Statistical analysis.

Normality of distributions was tested with the Shapiro–Wilk test, and homogeneity of variances with Levene’s test. Sphericity was assessed using Mauchly’s test, and the Greenhouse–Geisser correction was applied when sphericity was violated. All data meet normality assumptions. Intraclass correlation coefficients (ICC) with 95% confidence intervals were calculated using a two-way mixed-effects model to assess absolute agreement between the drop test and manual stretch test across 2 d. A split-plot repeated-measures ANOVA was used to examine the effects of MUSCLE (within-subject: tibialis anterior, soleus) and GROUP (between-subject: clonus, nonclonus) on LPR and LLR amplitude, LLR duration, MVCs, and StartReact responses. Within the clonus group, repeated-measures ANOVA assessed the effect of MUSCLE on burst duration, interburst interval, burst amplitude, and burst frequency. One-way ANOVA tested GROUP effects on H-max, M-max, and H/M ratio. Pearson’s correlations were used to examine relationships between LLR, StartReact responses, and H/M ratio. Bonferroni’s adjustments were applied for post hoc comparisons. Additional analyses compared StartReact responses, H-max, M-max, and H/M ratio in participants with SCI to previously collected age-matched control data. Statistical analyses were performed in SPSS (v26.0; IBM), with significance set at p < 0.05. Effect sizes for ANOVA are reported as partial η², representing the proportion of variance in the dependent variable explained by a specific factor, accounting for other variables and error variance. Group data are presented as mean ± SD.

Results

Clonus

Figure 1B illustrates rectified EMG traces from soleus muscle. The top trace shows clonus contractions in a representative participant from the clonus group, lasting over 30 s (only one portion is shown), while the bottom trace shows EMG activity from a participant in the non-clonus group, demonstrating the absence of EMG bursts. Figure 2 provides detailed information on EMG burst duration in individual trials for both groups, assessed using the drop test and the manual stretch test, with timing obtained from EMG recordings or visual detection. Most participants in the clonus group (17 of 19) exhibited bursts lasting >10 s in all trials, while 2 participants (2 of 19) showed bursts lasting >10 s in 90% of trials. In contrast, all participants in the non-clonus group (14 of 14) showed no EMG bursts or clonus activity recorded by a timer in any trial. Reliability testing demonstrated excellent consistency for both assessments: the manual stretch test showed an ICC of 1.0 (95% CI [1.0, 1.0]) between Day 1 and Day 2, and the drop test showed an ICC of 0.97 (95% CI [0.957, 0.98]), suggesting that both assessments provide reliable tests for the presence of clonus.

Repeated-measures ANOVA in the clonus group revealed a significant effect of MUSCLE on burst amplitude (F(_1,14) = 16.27, p = 0.001, partial η² = 0.54), indicating that EMG burst amplitude differed substantially between the soleus and tibialis anterior muscles (Fig. 3). Burst amplitude was 16.12 ± 9.43% M-max (range 4.20–38.95% M-max) in the soleus and 5.34 ± 4.01% M-max (range 1.17–15.97% M-max) in the tibialis anterior (p = 0.001). In contrast, MUSCLE had no significant effect on burst frequency (F(_1,14) = 0.97, p =0.34, partial η² = 0.065), burst duration (F(_1,14) = 1.12, p = 0.30, partial η² = 0.07), or interburst interval (F(_1,14) = 1.23, p = 0.29, partial η² = 0.081), suggesting that these parameters were similar across muscles. Interburst duration was 141 ± 18 ms (range 104–168 ms) in the soleus and 143 ± 17 ms (range 118–172 ms) in the tibialis anterior (p = 0.34). Clonus frequency was also comparable, at 5.2 ±0.6 Hz in both the soleus (range 4.4–6.5 Hz) and tibialis anterior (range 4.5–6.4 Hz; p = 0.29; Fig. 3).

Figure 3.

Clonus. A, Representative raw EMG recordings demonstrating clonus activity in the soleus and tibialis anterior muscles. B, Group data showing in the ordinate burst frequency (top left), burst amplitude (top right), burst duration (bottom left), and interburst duration (bottom right) for the soleus and tibialis anterior muscles in participants with clonus. The abscissa indicates the muscle tested (soleus or tibialis anterior). Open circles represent mean values for individual participants for each measurement and muscle. *p < 0.05.

Cutaneous reflexes

Figure 4, A and B, illustrates average EMG traces of cutaneous reflexes in the tibialis anterior and soleus muscles from representative participants in the clonus and non-clonus groups. The LLR duration was clearly prolonged in both muscles in the participant with clonus compared with the participant without clonus.

Figure 4.

Cutaneous reflex. Raw averaged EMG traces illustrate the cutaneous reflex in an individual with spinal cord injury (SCI) with and without clonus in the tibialis anterior (A) and soleus (B) muscles. Traces show the long-polysynaptic reflex (LPR) and long-lasting reflex (LLR) components of the cutaneous reflex in both muscles. The duration of the LLR component was calculated starting from 500 ms to reflex offset. Arrows indicate the time of reflex offset for each muscle. Note that LLR duration is shorter in the individual without clonus compared with the individual with clonus for both muscles. Group data are shown for LPR amplitude, LLR amplitude, and LLR duration in the tibialis anterior (A) and soleus (B) muscles. The abscissa indicates the groups tested (clonus, teal; non-clonus, red). Open circles represent mean values for individual participants for each measurement and muscle. *p < 0.05.

We first compared LLR duration and amplitude across groups and muscles. Repeated-measures ANOVA revealed significant effects of MUSCLE (F(_1,27) = 19.43, p < 0.001, partial η² = 0.42), GROUP (F_(1,27) = 40.70, p < 0.001, partial η² = 0.60), and their interaction (F_(1,27) = 12.76, p = 0.001, partial η² = 0.32) on LLR duration. Post hoc analyses showed that LLR duration was significantly longer in the clonus group compared with the nonclonus group for both the tibialis anterior (1.72 ± 1.02 s vs 0.17 ± 0.11 s; p < 0.001; Fig. 4A) and the soleus (0.56 ± 0.24 s vs 0.06 ± 0.05 s; p < 0.001; Fig. 4B). Similarly, we found that LLR mean amplitude normalized to the M-max was significantly influenced by MUSCLE (F_(1,27) = 32.08, p < 0.001, partial η² = 0.56), GROUP (F_(1,27) = 7.84, p = 0.01, partial η² = 0.24), and their interaction (F(_1,27) = 7.52, p = 0.011, partial η² = 0.23). Post hoc analyses showed that normalized LLR amplitude was significantly larger in the clonus group compared with the non-clonus group for both the tibialis anterior (0.66 ± 0.50% M-max vs 0.25 ± 0.10% M-max, p = 0.007; Fig. 4A) and the soleus (0.09 ± 0.05% M-max vs 0.04 ± 0.04% M-max, p = 0.01; Fig. 4B). Within each group, both LLR duration (clonus: p < 0.001; non-clonus: p = 0.001) and amplitude (clonus: p < 0.001; non-clonus: p < 0.001) were larger in the tibialis anterior compared with the soleus.

We also compared the LPR (the initial part of the cutaneous reflex) between muscles and groups. Repeated-measures ANOVA revealed significant effects of MUSCLE (F(_1,27) = 96.05, p < 0.001, partial η² = 0.79), GROUP (F_(1,27) = 11.23, p = 0.003, partial η² = 0.31), and their interaction (F(_1,27) = 9.86, p = 0.004, partial η² = 0.28) on LPR mean amplitude normalized to the M-max. Post hoc analyses showed that normalized LPR amplitude was significantly larger in the clonus group compared with the non-clonus group for both the tibialis anterior (4.31 ± 1.79% M-max vs 2.23 ± 1.45% M-max, p = 0.003; Fig. 4A) and the soleus (0.29 ± 0.18% M-max vs 0.15 ± 0.10% M-max, p = 0.016; Fig. 4B).

StartReact responses

Figure 5A illustrates the average rectified EMG traces in the soleus and tibialis anterior muscles for both groups. In the participant with clonus, the StartReact response, calculated as the difference between VART and VSRT, was larger in both muscles compared with participants without clonus, suggesting a greater contribution of the reticulospinal pathway.

Figure 5.

StartReact. A, Raw mean EMG traces illustrating visual reaction time (VRT; black), visual–auditory reaction time (VART; gray), and visual–startle reaction time (VSRT; red) in the soleus and tibialis anterior muscles from participants with SCI with and without clonus. For each stimulus condition (VRT, VART, and VSRT) and contraction type (plantarflexion and dorsiflexion), 20 responses were averaged. The StartReact response was calculated as the difference between VART and VSRT (VART – VSRT) in both muscles (double-headed arrow). Notably, in the participant with clonus, reaction time was further reduced and the difference between VART and VSRT was larger in both muscles compared with the participant without clonus. B, Group data for the StartReact response in the soleus and tibialis anterior muscles. The abscissa indicates the groups tested (clonus, teal; non-clonus, red). Open circles represent mean values for individual participants for each measurement and muscle. *p < 0.05.

A repeated-measures ANOVA revealed a significant main effect of GROUP on the StartReact response (F_(1,15) = 12.21, p = 0.003, partial η² = 0.45), but no main effect of MUSCLE (F(_1,15) = 3.00, p = 0.104, partial η² = 0.17) and no interaction (F(_1,15) = 0.81, p = 0.38, partial η² = 0.05). Post hoc analyses showed that the StartReact response was larger in the clonus group compared with the non-clonus group for both the soleus (clonus: 0.035 ±0.010 s; non-clonus: 0.021 ± 0.010 s; p = 0.02; Fig. 5B) and tibialis anterior (clonus: 0.032 ± 0.018 s; non-clonus: 0.012 ± 0.006 s; p= 0.01; Fig. 5B) muscles. Within-group comparisons revealed no difference between soleus and tibialis anterior muscles in either the clonus group (p = 1.0) or the non-clonus group (p = 0.10). Comparisons with control participants indicated that the StartReact response was significantly larger in the clonus group relative to controls in both the soleus (clonus: 0.035 ± 0.010 s; controls: 0.022 ± 0.008 s; p = 0.001) and tibialis anterior (clonus: 0.032 ± 0.018 s; controls: 0.017 ± 0.012 s; p = 0.03) muscles. In contrast, no differences were observed between the non-clonus group and controls for either the soleus (p = 1.0) or tibialis anterior (p = 0.94) muscles.

We also compared the VRT between groups and muscles, which reflects reaction time in the absence of an auditory stimulus. A repeated-measures ANOVA revealed no main effect of MUSCLE (F(_1,15) = 2.54, p = 0.13, partial η² = 0.15), no main effect of GROUP (F_(1,15) = 0.0004, p = 0.99, partial η² = 0.00003), and no interaction (F_(1,15) = 0.03, p = 0.86, partial η² = 0.002) on VRT. Post hoc analyses showed that, in the clonus group, VRT did not differ between the tibialis anterior (0.26 ± 0.06 s) and soleus muscles (0.28 ± 0.08 s; p = 0.82). Similarly, in the nonclonus group, VRT was not different between the tibialis anterior (0.26 ± 0.06 s) and soleus muscles (0.28 ± 0.06 s; p = 1.0). Additionally, no differences in VRT were observed between the clonus and non-clonus groups for either the tibialis anterior or soleus muscles (both p = 1.0), suggesting that participants in both groups had a similar ability to response to the LED stimulus in the absence of sound.

H/M ratios

Figure 6A shows representative raw H-max and M-max traces from the soleus muscle in participants from the clonus and non-clonus groups. A one-way ANOVA revealed a significant effect of GROUP on the H/M ratio (F_(1,30) = 49.19, p < 0.001, partial η² = 0.62; Fig. 6B). Post hoc analyses showed that the H/M ratio was higher in the clonus group (0.64 ± 0.12) compared with the non-clonus group (0.25 ± 0.18; p < 0.001). Compared with controls (0.49 ± 0.17), the H/M ratio was higher in the clonus group (p = 0.015) and lower in the non-clonus group (p < 0.001).

Figure 6.

H-max, M-max, and H/M ratio. A, Representative raw mean EMG traces illustrating the maximal H-reflex (H-max) and maximal motor response (M-max) recorded from the soleus muscle in individuals with SCI with and without clonus. In this example, M-max amplitudes were similar between participants, whereas H-max was smaller in the participant without clonus compared with the participant with clonus. B, Group data for H-max, M-max, and the H/M ratio (left to right) in the soleus muscle. The abscissa indicates the groups tested (clonus and non-clonus). The ordinate shows H-max amplitude (millivolts), M-max amplitude (millivolts), and the H/M ratio, respectively. Open circles represent mean values for individual participants for each measurement and muscle. *p < 0.05.

A one-way ANOVA also revealed a significant effect of GROUP on H-max amplitude (F(_1,30) = 10.91, p = 0.002, partial η² = 0.27; Fig. 6B). H-max was greater in the clonus group (5.99 ± 2.87 mV) compared with the non-clonus group (2.70 ± 2.67 mV; p = 0.002). Relative to controls (6.41 ± 2.52 mV), H-max was reduced in the non-clonus group (p < 0.001), whereas no difference was observed between the clonus group and controls (p = 1.0). In contrast, no significant effect of GROUP was found for M-max amplitude (F(_1,30) = 0.32, p = 0.58, partial η² = 0.01; Fig. 6B). M-max did not differ between the clonus (9.47 ± 4.24 mV) and non-clonus groups (8.52 ± 5.25 mV; p = 0.58). However, compared with controls (13.23 ± 3.20 mV), M-max was smaller in both the clonus (p = 0.012) and non-clonus groups (p = 0.002).

Relationship between electrophysiological measures

Figure 7 illustrates the correlations between the StartReact response, LLR duration, and H/M ratio in both muscles. In the soleus muscle, LLR duration was positively correlated with the StartReact response (r = 0.52, p = 0.047; Fig. 7A) and with the H/M ratio (r = 0.58, p = 0.001; Fig. 7B). In addition, a positive correlation was observed between the StartReact response and the H/M ratio in the soleus muscle (r = 0.69, p = 0.003; Fig. 7C). However, no significant correlations were observed between LLR normalized magnitude and the StartReact response (r = −0.32, p = 0.245) or between LLR normalized magnitude and the H/M ratio (r = −0.13, p = 0.504). In the tibialis anterior muscle, no significant correlation was found between LLR duration and the StartReact response (r = 0.30, p = 0.28; Fig. 7D). Similarly, LLR normalized magnitude was not significantly correlated with the StartReact response (r = 0.46, p = 0.097).

Figure 7.

Correlations. Individual data from participants with SCI with clonus (teal) and without clonus (red) are shown. In the top graphs, the abscissa indicates LLR duration in the soleus muscle, and the ordinate shows the StartReact response (A) and H/M ratio (B). The solid line represents the Pearson’s or Spearman correlation calculated using all data points from both clonus and non-clonus groups. Notably, LLR duration was positively correlated with both the StartReact response and the H/M ratio. In the bottom graphs, the abscissa indicates the StartReact response in the soleus muscle (C) and LLR duration in the tibialis anterior muscle (D), and the ordinate shows the H/M ratio and the StartReact response, respectively. *p < 0.05.

Discussion

Our findings suggest that descending brainstem systems play a role in the maintenance of ankle clonus following chronic SCI. Individuals with clonus exhibited elevated H/M ratios, indicating increased motoneuron excitability in response to Ia synaptic input, whereas those without clonus showed smaller ratios than uninjured controls. Participants with clonus demonstrated larger and more prolonged LLRs in the tibialis anterior and soleus muscles, as well as shorter StartReact response latencies, compared with participants without clonus. Notably, LLR duration was positively correlated with both StartReact response and H/M ratio. Together, these findings support the hypothesis that differences in descending brainstem system contributions to motoneuron excitability may, at least in part, explain the presence or absence of ankle clonus following chronic SCI.

Descending brainstem systems contribute to motoneuron excitability and clonus

Earlier studies suggest clonus arises from self-reexcitation of hyperactive stretch reflex pathways (Clare et al., 1951; Cook, 1967; Lippold, 1970; Hagbarth et al., 1975; Nichols et al., 1978; Rack et al., 1984; Rossi et al., 1990). The coactivation of soleus and tibialis anterior EMG bursts observed here and previously (Beres-Jones et al., 2003) is consistent with reciprocal facilitation after SCI (Crone et al., 2003), also shown in rodents (Rotterman et al., 2024). Consistent with this, H/M ratios are elevated in individuals with clonus compared with those without clonus and controls (Sangari et al., 2019; Chen and Perez, 2022; Nito et al., 2026). However, H/M ratios were reduced in SCI participants without clonus (Koelman et al., 1993; Manella et al., 2013). A key question is why H/M ratios distinguish individuals with and without clonus, yet do not consistently separate those with and without enhanced stretch reflexes (Schindler-Ivens and Shields, 2004; Sangari et al., 2019; Chen and Perez, 2022; Nito et al., 2026). The H/M ratio reflects motoneuron excitability to Ia afferent input (Pierrot-Deseilligny and Burke, 2005) whereas the stretch reflex involves additional monosynaptic, polysynaptic, transcortical, and slower sensory pathways (Birnbaum and Ashby, 1982; Matthews, 1984). Unlike the stretch reflex, the H-reflex is evoked by electrical stimulation of Ia afferents, generating a temporally synchronized afferent volley that enhances spatial summation of excitatory postsynaptic potentials (EPSPs) across motoneurons (Burke et al., 1983). Thus, stretch reflex amplitude reflects both temporal and spatial summation, whereas the H-reflex primarily reflects spatial summation, which may help explain why H/M ratios do not consistently differ in individuals with heightened stretch reflexes. Other mechanisms, such as afferent sprouting after corticospinal tract damage, have been linked to reduced frequency-dependent depression of the H-reflex and hyperreflexia in animals (Tan et al., 2012). However, human studies show no consistent differences in frequency-dependent depression between individuals with SCI with and without hyperreflexia (Nito et al., 2026). Likewise, H-reflex amplitude (Schindler-Ivens and Shields, 2004; Sangari et al., 2019; Chen et al., 2020) and presynaptic inhibition during voluntary contraction (Morita et al., 2001; Chen and Perez, 2022) appear similar across SCI groups. Thus, while these mechanisms cannot be fully excluded, they are unlikely to explain the observed group differences.

We favor the hypothesis that differences in H-max and H/M ratios between the clonus and non-clonus groups reflect, at least in part, differential contributions from descending brainstem systems. To probe brainstem-derived monoaminergic contributions to motoneuron excitability, we assessed the LLR in tibialis anterior and soleus. LLR amplitude and duration are largely enhanced by PICs, which amplify synaptic inputs and sustain motoneuron firing, producing prolonged muscle contractions (Hounsgaard et al., 1988; Hultborn et al., 2003; Li et al., 2004; Heckmann et al., 2005). Animal studies further support a role for PICs in generating involuntary muscle activity and LLRs following SCI (Murray et al., 2010; Tysseling et al., 2017). In humans, similar stimulation intensities (40/80 mA; ~3–5 × T) used to elicit cutaneous reflexes show that cyproheptadine reduces the LLR without affecting the LPR, consistent with a PIC-related contribution (Murray et al., 2010). However, our stimulation may also recruit high-threshold afferents and spinal interneuronal pathways (de Willer, 1977; Sandrini et al., 2005), along with PIC-mediated increases in motoneuron excitability, which may partly explain reduced LPR in individuals without clonus.

Regardless of additional pathways that may contribute to the LLR, it was prolonged in participants with clonus across both muscles. This aligns with prior reports that individuals with SCI and spasms show prolonged LLRs (DeForest et al., 2020), suggesting shared mechanisms of motoneuron hyperexcitability and sustained involuntary activity. Illustrative examples (Fig. 8) show transitions from clonus-like activity to sustained spasms, consistent with PIC “warm-up,” where repeated activation increases PIC amplitude (Bennett et al., 1998; Gorassini et al.,2002). Another possibility is that movements triggered by prolonged reflex activity repeatedly reactivate low-threshold afferents, generating additional polysynaptic EPSPs. Through temporal summation and wind-up potentiation, these EPSPs may progressively increase in amplitude, further sustaining rhythmic motoneuron activation. This repeated reexcitation could contribute to clonus, characterized by ~5 Hz oscillations consistent with ~200 ms duration of each clonus cycle.

Figure 8.

Clonus to spasms. Individual data from participants with SCI and clonus showing representative raw EMG recordings illustrating transitions in which clonus activity in the soleus muscle evolves into a sustained spasm. The representative participants shown correspond to Participants 3, 15, and 17 in Table 1. These illustrative examples demonstrate how rhythmic clonus-like activity can progressively transform into prolonged spasmodic muscle activation.

To probe brainstem contributions via the reticulospinal pathway, we assessed the StartReact response, an involuntary release of a preplanned movement triggered by a startling stimulus that engages reticulospinal circuits (Davis et al., 1982; Brown et al., 1991; Valls-Solé et al., 1999). Previous studies using motoneuron firing-generated simulated electromyograms have shown that the reduction in reaction time in response to a loud sound increases with the proportion of descending drive originating from the reticular formation (Tapia et al., 2022). Group differences in StartReact may also reflect differences in reticulospinal contribution or reliance during voluntary ankle control after SCI in the clonus versus non-clonus groups. Overall, this is consistent with evidence of enhanced reticulospinal involvement in SCI with spasticity (Sangari and Perez, 2020; De Santis and Perez, 2026) and with stroke studies linking increased motoneuron excitability to both reticulospinal and neuromodulatory mechanisms (McPherson et al., 2018a,b).

A critical question is why LLRs were absent or reduced in participants without clonus. One possibility is reduced preservation of descending serotonergic drive, leading to lower motoneuron excitability and weaker reflex responses. LLRs are influenced by constitutive serotonergic receptor activity after incomplete SCI (Tysseling et al., 2017), which can be further amplified by descending serotonergic input. Consistent with this, motoneurons and interneurons below the lesion become supersensitive to serotonin after SCI, with enhanced PIC responsiveness to 5-HT (Harvey et al., 2006; Husch et al., 2012). Clinically, selective serotonin reuptake inhibitors can exacerbate hyperreflexia and spasms (Thompson and Hornby, 2013), elevated serotonin has been linked to clonus (Dunkley et al., 2003), and reduced brainstem pathway integrity has been associated with fewer spasticity manifestations (Schading-Sassenhausen et al., 2025). Together, these findings are consistent with the possibility that reduced preservation of brainstem-derived monoaminergic influences contribute to the absence or reduction of LLRs in individuals without clonus. However, additional mechanisms cannot be excluded. For example, increased spinal serotonin production has been shown in animal models of SCI (Wienecke et al., 2014), and serotonin receptor expression on motoneurons increases after SCI (Ren et al., 2013); these changes may differ between individuals with and without clonus. Inflammatory processes after SCI may also alter serotonin receptor activity on motoneurons (Di Narzo et al., 2015), which may also vary between individuals with and without clonus. Although participants withheld antispasmodic and GABAergic medications for at least 12 h before testing, most individuals in the non-clonus group were not taking baclofen. Thus, residual baclofen effects or withdrawal-related rebound cannot be fully excluded as contributors to the observed group differences.

Functional considerations

Our findings suggest three clinical implications for individuals with SCI with and without clonus. First, they highlight the importance of distinguishing clonus from generalized stretch hyperreflexia in spasticity assessment after SCI. Although definitions vary, increased resistance to rapid muscle stretch is widely accepted as a key feature (Gracies, 2005; Dietz and Sinkjaer, 2007, 2012; Nielsen et al., 2007). However, clinical assessments reflect both neural factors, such as exaggerated stretch reflexes (Lorentzen et al., 2010; Yamaguchi et al., 2018), and non-neural factors, including increased muscle stiffness (Mirbagheri et al., 2001) and shortened muscle–tendon units (Diong et al., 2012), complicating interpretation (Therkildsen et al., 2025).

Second, standardized tests such as the drop test and manual stretch test showed good-to-excellent testretest reliability, supporting their use for clonus quantification (Koelman et al., 1993; Manella et al., 2013). We observed strong agreement across days, though some trial-to-trial variability suggests that multiple measurements may be needed to accurately determine clonus. As our cohort included participants with severe or no clonus, future work should examine mild-to-moderate cases.

Third, differences in H-max amplitude and H/M ratios suggest that clonus is a sensitive marker of altered motoneuron excitability to Ia input. Prior studies show these measures often do not differ between individuals with and without other spasticity signs, particularly when defined by increased stretch reflexes (Schindler-Ivens and Shields, 2004; Sangari et al., 2019; Chen and Perez, 2022; Nito et al., 2026). Similarly, measures of Ia afferent transmission—such as postactivation depression at rest (Nito et al., 2026) and presynaptic inhibition during voluntary contraction (Morita et al., 2001; Chen and Perez, 2022)—are often similar across SCI groups. Clonus therefore may provide a more objective indicator of motoneuron excitability than commonly used clinical scales (Therkildsen et al., 2025).

Significance Statement.

Ankle clonus is a repetitive, involuntary muscle contraction often seen after spinal cord injury (SCI). Its cause is not fully understood, but brainstem pathways may contribute by increasing the excitability of spinal motoneurons. In this study, we measured reflex responses in leg muscles and reactions to sensory and startling stimuli in people with chronic SCI, with and without clonus. Individuals with clonus showed stronger and longer reflex responses and faster reactions to startling sounds. These findings suggest that brainstem pathways involved in movement control may help maintain ankle clonus after SCI.