Intermuscular coherence in spinocerebellar ataxias 3 and 6: a preliminary study

Objective: Spinocerebellar ataxias (SCAs) are familial neurodegenerative diseases involving the cerebellum and spinocerebellar tracts. While there is variable involvement of corticospinal tracts (CST), dorsal root ganglia, and motor neurons in SCA3, SCA6 is characterized by a pure, late-onset ataxia. Abnormal intermuscular coherence in the beta-gamma frequency range (IMCbg) implies lack of integrity of CST or the afferent input from the acting muscles. We test the hypothesis that IMCbg has the potential to be a biomarker of disease activity in SCA3 but not SCA6. Methods: Intermuscular coherence between biceps and brachioradialis muscles was measured from surface EMG waveforms in SCA3 (N=16) and SCA6 (N=20) patients, and in neurotypical subjects (N=23). Results: IMC peak frequencies were present in the b range in SCA patients and in the g range in neurotypical subjects. The difference between IMC amplitudes in the g and b ranges was significant when comparing neurotypical control subjects to SCA3 (p < 0.01) and SCA6 (p = 0.01) patients. IMCbg amplitude was smaller in SCA3 patients compared to neurotypical subjects (p<0.05), but not different between SCA3 and SCA6 patients or between SCA6 and neurotypical subjects. Conclusion/significance: IMC metrics can differentiate SCA patients from normal controls.


Introduction
Spinocerebellar ataxias (SCAs) are a group of familial neurodegenerative diseases that involve the cerebellum and the brainstem, with their most common clinical presentation being a progressive ataxia.
The most common type of SCA in the USA and worldwide is SCA3, followed by SCA2 and SCA6 (1), with SCA3 and SCA6 together accounting for more than a third of familial ataxias (2).SCA3 has an onset in childhood to mid-adulthood with heterogeneous clinical manifestations, including pyramidal and extrapyramidal features, amyotrophy and peripheral neuropathy, in addition to ataxia (3).SCA6, on the other hand, is primarily an ataxic disorder with a mean age of onset of 52 years (1,4).Classically, diagnosis of SCA subtype has relied on clinical presentation, with de nitive diagnosis now done by genetic testing.
There has been substantial expansion in therapeutic approaches for inherited neurodegenerative diseases with targeted treatments including antisense oligonucleotides and stem cell-based approaches in the preclinical stage for different types of SCA (5).As these therapies come to clinical trials there will be a need for biomarkers of degeneration in central and peripheral pathways involving the cerebellum, pyramidal system, and sensory pathways in symptomatic and presymptomatic stages of SCA.Although clinical outcome measures are commonly used to capture the severity of symptoms in SCAs, an increasing number of imaging and uid biomarkers are being identi ed for future clinical trials (6).The Scale for the Assessment and Rating of Ataxia (SARA) is a validated clinical scale used for monitoring the progression of ataxia (4,7).SARA has rather low effect sizes in longitudinal studies, therefore necessitating a large sample size to detect changes in progression, which is di cult in SCAs given their rarity (8).For example, in a large European cohort, the rate of increase of SARA was steepest in SCA1 and only ~ 4% /year for SCA3 (9).SARA is therefore not an ideal tool for monitoring disease progression in clinical trials, which typically last 6-12 months.Certain MRI sequences such as volumetric analysis of the cerebellar cortex, superior cerebellar peduncle and brainstem, diffusion tensor imaging and tractography have shown promising results in longitudinal studies on SCA (6, 8), but these are not validated and available only in a limited number of centers.A combination of neurophysiological methods including nerve conduction studies (NCS), somatosensory evoked potentials (SEPs), and motor cortical evoked potentials using transcranial magnetic stimulation (TMS), are reliable methods for assessment of peripheral nerves, dorsal root ganglia, and corticospinal tracts (10).However, these tests usually cannot be done on a repetitive basis in the setting of a clinical trial because of the associated discomfort and -in the case of TMS -lack of widespread availability.Ideal biomarkers to objectively monitor SCA disease progression are therefore not yet available.
Coherence analysis has been extensively used to study the integrity of pathways involved in performance of a motor act.Coherence in neurophysiological signals measured from sensorimotor cortex, subcortical regions, and lower motor neurons represents synchronization of neuronal activity in these spatially distant structures involved in a coordinated movement (11).Corticomuscular coherence (CMC) in the β frequency range (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) is present between oscillatory activity of motor cortex and active motor units in the contralateral limbs (12)(13)(14)(15)(16).The cortically originated signals can also be assessed by measuring the coherence between co-activated muscles (intermuscular coherence; IMC) in a subset of the βγ frequency ranges (20-40 Hz) (12,14,17,18).IMCβγ becomes abnormal in diseases involving upper motor neurons including primary lateral sclerosis and amyotrophic lateral sclerosis (ALS) (19,20), and also by sensory deafferentation of the acting muscles.Such changes could be the result of dysfunction at the level of the peripheral nerves, dorsal root ganglia or dorsal columns (21,22).Determination of βγ coherence using IMC or CMC is therefore a potential biomarker not only of motor neuron diseases but also for other diseases of the nervous system that affect the motor system such as SCAs.
In this study we apply a method for assessing IMC using a conventional electromyography (EMG) apparatus (20) to determine if IMCβγ has potential as a new biomarker for SCA3 or SCA6.In particular, we hypothesized that IMCβγ would be abnormal in SCA3 patients, who often have signi cant involvement of corticospinal tract and/or sensory pathways.

Methods
This is a cross sectional study on a cohort of patients with SCA3 (N = 16), SCA6 (N = 20) and neurotypical control subjects (N = 23).Subjects were excluded from the study if they were < 18 years, unable to provide informed consent, or were unable to perform the study task.The data collected included demographic information, symptom duration, results of the genetic testing (CAG trinucleotide expansion size), the Scale for the Assessment and Rating of Ataxia (SARA) which ranges from 0 (no cerebellar symptoms) to 40 (most severe cerebellar symptoms) to measure the severity of ataxia (7), and assessment of IMC as previously described (20).The study was approved by the Institutional Review Board of University of Chicago Biological Science Division, and subjects provided written informed consent before researchrelated data were collected.

Intermuscular coherence measurements
Surface EMG was recorded simultaneously from the biceps and brachioradialis muscles from the dominant arm.Bipolar surface EMG electrodes with a 3 cm spacing (Nicolet bipolar disk bar electrodes; Natus Medical, Inc., Pleasanton, California) were placed over the muscle bellies, and a ground electrode was placed on the volar surface of the wrist.The nearest contacts on the two different muscles were separated by at least 8 cm.
Muscle activity was recorded while the subject held their forearm parallel to the ground and perpendicular to the upper arm.The arm was at their side, with the shoulder adducted and neutrally rotated.Recording sessions consisted of three epochs each lasting 30 seconds and separated from each other by 30 seconds of rest with the hand in a semi-pronated position (20).Surface EMG signals were acquired using a Cadwell EMG system (model Sierra-Summit, Cadwell Industries, Inc., Kennewick, WA) with a high-pass lter of 5 Hz, a low-pass lter of 200 Hz, a notch lter at 60 Hz, and at a sampling rate of 64,000 samples/second.Signals were down-sampled to 4267 Hz using the Matlab downsample function prior to calculation of IMC.Surface EMG activity was segmented into blocks of ~ 0.25 seconds.The IMC calculation was implemented in Matlab using the mscohere (magnitude-squared coherence) algorithm.The con dence limit is the coherence level below which coherence is not statistically different from zero.The con dence limit is de ned as: in which α is the desired signi cance level (0.05) and S is the number of segments analyzed (23).
Coherence is de ned for stationary signals, meaning changes in load or arm movement during the task would degrade intermuscular coherence calculations.This was especially an issue in patients with tremor, for whom the more frequent arm movements could arti cially reduce coherence measures.We used the same process as in Issa et al. 2017 (20) to exclude segments of data that violate stationarity, with modi cations to reduce the amount of data excluded, set a maximum allowable variance, and ensure a low con dence limit.The exclusion process was iterative, with the goal of having the coe cient of variation (COV) for root mean squared (RMS) amplitudes (standard deviation of block RMS amplitudes / mean of block RMS amplitudes) be smaller than 0.15.Individual 0.25 second blocks of EMG data that had an RMS amplitude greater than four standard deviations from the mean RMS amplitude of a subject's EMG data set were initially removed.If this post-hoc processing step did not achieve a COV below 0.15, then blocks with progressively smaller deviations from the mean were removed until the COV was less than 0.15.Only subjects in whom a su cient number of segments were retained to provide an IMC con dence limit below 0.03 were included in the nal analysis.
The average IMC was calculated in speci c frequency bands.IMCβγ was calculated as the average IMC between 20 and 40 Hz; this frequency range crosses the boundary between the β-band and γ-band but does not encompass the full β-or γ-bands.IMCγ was calculated as the average IMC in the portion of the γ band between 30 and 40 Hz and IMCβ was calculated as the average IMC in the portion of the β band between 15-25 Hz.IMCγ-IMCβ was calculated as the difference between IMCγ and IMCβ.

Statistical analysis
Stata version SE17 (StataCorp LLC, College Station, TX) was used for statistical analysis.Continuous data were presented with mean and standard deviation (SD), and categorical data as number and percentile.Comparisons of age, SARA score, disease duration, and IMC amplitudes were done using Mann-Whitney U test, and gender and handedness were compared with a test of proportions.The associations between CAG trinucleotide expansion size and IMC metrics were assessed with linear regression.Alpha was set to 0.05.Receiver-operating characteristic (ROC) curves were calculated to assess the diagnostic ability of IMC metrics to differentiate subtypes of SCA from each other and from neurotypical subjects (24).An area under the ROC curve (AUC) of 0.9-1.0 was considered excellent, 0.8-0.9good, 0.7-0.8fair, 0.6-0.7 poor, and between 0.5-0.6 non-diagnostic.

Results
Intermuscular coherence was measured in patients with SCA3 (N = 16), SCA6 (N = 20), and in neurotypical control subjects (N = 23).Group characteristics are shown in Table 1.All pairwise comparison results are shown in supplemental tables 1-2.Gender and handedness proportions across the three study groups and disease duration among the SCA subgroups were not signi cantly different (p > 0.05 for all comparisons).SCA6 subjects were signi cantly older than the neurotypical subjects (p = 0.01), but not the SCA3 group (p = 0.28); and the average age of the SCA3 group was not different from the neurotypical control group (p = 0.12).The difference in SARA scores between the SCA6 and SCA3 groups was not statistically signi cant (p = 0.065).
Intermuscular coherence varied with frequency in all measured groups (Fig. 1).The pro le for neurotypical controls shows a peak in coherence between 20 and 40 Hz, similar to that previously reported for an independent set of neurotypical subjects (20).SCA3 and SCA6 subjects had a peak between 15 and 25 Hz, about 10-15 Hz lower than the peak in the average neurotypical IMC pro le.
IMCβγ, de ned as the average coherence measured between 20 and 40 Hz, has been identi ed as a potential marker for upper motor neuron dysfunction (19,20).To determine if IMCβγ could help differentiate among the SCA subtypes or between patients with SCAs and neurotypical subjects, IMCβγ was measured for each subject.As suggested by the plots in Fig. 1, the average IMCβγ was smallest in SCA3 patients and largest in neurotypical control subjects (Table 1).When the distributions of individual IMCβγ were compared among groups (Fig. 2A), there was no signi cant difference between the IMCβγ amplitudes of SCA6 patients and neurotypical control subjects (P = 0.35).SCA3 patients had a signi cantly smaller IMCβγ than neurotypical subjects (P = 0.024), but there was no statistical difference between SCA3 and SCA6 patients (P = 0.18).
The IMC pro les for SCA3 and SCA6 appeared qualitatively different from that of neurotypical control subjects (Fig. 1), and different from previous ndings from amyotrophic lateral sclerosis (ALS) subjects who showed a loss of coherence in the βγ range (20).The difference between IMCγ (30-40 Hz) and IMCβ (15-25 Hz) was calculated for each subject to determine whether differences in the coherence peaks in the IMC pro le could differentiate SCA3 or SCA6 from neurotypical subjects.The distributions of IMCγ-IMCβ (Fig. 2B) were signi cantly different between neurotypical subjects and SCA3 patients (P = 0.0003) and between neurotypical subjects and SCA6 patients (P = 0.013), but not between SCA3 and SCA6 groups (P = 0.50).
Linear regression was used to test whether IMC metrics were associated with CAG triplet repeat number.For the SCA6 group there was little variability in triplet repeat number (Table 1), with all but two patients having 22 repeats, one had 21 repeats, and one had 25 repeats.As would be expected based on the limited variability in the independent variable, neither IMCβγ (p = 0.85) nor IMCγ-IMCβ (p = 0.29) were correlated with triplet repeat number in the SCA6 group.For the SCA3 group the triplet repeat number ranged between 62 and 77 (Table 1), but was not correlated with either IMCβγ (p = 0.80) or IMCγ-IMCβ (p = 0.92).There was therefore no signi cant association between triplet repeat number and IMC metrics in either SCA3 or SCA6 groups.
To determine whether IMCβγ or IMCγ-IMCβ have the potential to be markers for disease severity, they were compared to SARA scores and disease duration.SARA scores were strongly correlated with disease duration for SCA6 patients (R 2 = 0.60, p = 0.00), but not for SCA3 patients (R 2 = 0.02, p = 0.56).Neither SARA scores nor disease duration was correlated with IMCβγ or IMCγ-IMCβ in the SCA6 group.In SCA3 patients, neither IMC metric correlated with disease duration.While the correlation between SARA scores and IMCβγ was not statistically signi cant (R 2 = 0.17, p = 0.11), SARA scores were signi cantly correlated with IMCγ-IMCβ for the SCA3 group (R 2 = 0.26, p = 0.04).Therefore, IMC measures showed a signal of correlating with disease severity in SCA3 but not SCA6 patients.

Discussion
Abnormally low IMCβγ is thought to represent a lack of integrity in the corticospinal tract or proprioceptive loss (14,18,19,22,25).As these pathways are differently involved in the various SCA subtypes, we hypothesized that IMC could provide differentiating information between the subtypes or may serve as a biomarker for disease progression for certain SCAs.
The preliminary ndings reported here suggest that IMCβγ is not signi cantly altered in patients with SCA6.This was the expected nding for SCA6 patients, in whom pathological changes are primarily cerebellar with minimal changes to the corticospinal system or peripheral nerves.Conversely, IMCβγ was signi cantly lower in SCA3 patients than in neurotypical control subjects, consistent with the mixed cerebellar, corticospinal, and peripheral nerve dysfunction found with SCA3.
While the main question addressed in this study was whether IMCβγ would be useful as a marker for SCA3 and/or SCA6, another characteristic of the IMC pro le was identi ed as a good marker for SCAs.In both SCA3 and SCA6 groups there was an IMC peak in the β frequency range, with a comparatively lower peak in the γ frequency range.This pattern was qualitatively different from the IMC pro le in neurotypical subjects and the pattern seen in ALS patients (20).The difference between IMCs the γ and β frequency bands (IMCγ-IMCβ) differentiated the SCA subtypes from the neurotypical group better than IMCβγ did, with the best performance in differentiating SCA3 from neurotypical subjects having an area under the ROC curve of 0.85.
Our ndings suggest that IMC metrics might be useful in monitoring disease progression in SCA3 but not SCA6.SARA scores were tightly coupled to SCA6 disease duration, consistent with the primacy of cerebellar symptoms in SCA6.Neither IMCβγ nor IMCγ-IMCβ were correlated with either SARA scores or SCA6 disease duration, suggesting that they are not good markers for purely cerebellar pathology.For SCA3 subjects IMCγ-IMCβ was weakly correlated with disease severity as approximated by SARA score.Interestingly, neither SARA scores nor IMCγ-IMCβ were correlated with disease duration for SCA3 patients.This is consistent with the ndings of a previous study which did not nd a signi cant correlation of disease duration as an independent variable with SARA score in SCA3 patients (4).This might be because cerebellar dysfunction is only a part of the neurological impairment in SCA3.The correlation between SARA scores and IMCγ-IMCβ in SCA3 might be the result of a secondary correlation between SARA scores and corticospinal dysfunction.Future studies to address this issue should include a larger study group, longitudinal follow up with repeated IMC measurements, and assessments of non-cerebellar symptoms in SCA, like the Inventory of Non-Ataxia Symptoms (INAS).
Previous studies have demonstrated the usefulness of coherence analysis in SCA.Velazquez-Perez, et al demonstrated abnormal β-band CMC in the lower limbs in 19 SCA2 patients, which correlated with prolongation of central motor conduction time, consistent with degeneration of the corticospinal tracts (26).Abnormal CMC was present even in the absence of clinical evidence of corticospinal tract involvement in the aforementioned study.Other studies from the same investigators demonstrated lower IMCβ in symptomatic SCA2 patients as well as those in the prodromal state of the disease (27,28).
Although the cerebellum plays a critical role in the timing of motor tasks (29), it is not clear if and how cerebellar output modify CMC and IMC.Marsden et al found coherence in the 8-27 Hz range between the local eld potentials of nucleus ventralis intermedius (VIM) of the thalamus (i.e., cerebellar thalamus), ipsilateral EEG and contralateral surface EMG of the hand muscles in patients with isolated tremors (30).The coherence between VIM and EMG activity was less apparent in patients with cerebellar disease.The EEG-EMG coherence was absent in the majority of the patients in that study, so it could not be determined if cerebellar disease could affect the CMC.
The use of recti cation in calculating intermuscular coherence has been heavily debated (27,(31)(32)(33)(34), with most studies using recti ed EMG.We chose not to rectify EMG signals because a previous study showed a clear difference between neurotypical subjects and ALS patients with upper motor neuron dysfunction using IMC from unrecti ed EMG signals (20), and because recti cation has been shown to signi cantly attenuate IMC in the γ frequency band (27).We also have con rmed that recti cation abolishes the γ band coherence (data not shown), and thus recti cation could not be used to differentiate SCA patients from neurotypical subjects.Recti cation has been suggested to result in distortion of the coherence in βγ band due to high levels of amplitude cancellations in those frequencies (34).
Given these preliminary ndings, future studies will be able to address several limitations.The crosssectional nature prevents determination of whether IMC metrics change with time.In addition, future studies should include measures of non-cerebellar symptoms that might correlate with IMC better than does the SARA score.While signi cant correlations were found with a small number of the subjects, a larger sample size will be needed to both validate and better characterize the relationship between IMC and clinical ndings in SCA patients.One potential confounder is that IMC was measured in only a single limb, while disease progression might be unequal across limbs.Subsequent studies could address this question by measure IMC in all four limbs.

Conclusion
In this proof of concept, single center, cross sectional study, we demonstrated that IMC patterns are different in SCA3 and SCA6 patients compared to neurotypical control subjects.Both IMCβγ and IMCγ-IMCβ were able to distinguish SCA3 from control subjects, but IMCγ-IMCβ performed better than IMCβγ at identifying patients with SCA6.IMC, which can be done in a simple and inexpensive manner, is a potential biomarker for SCA3 and SCA6, but longitudinal studies will be needed to assess the viability of this biomarker.

Figures
Figures

Figure 1 Average
Figure 1